Words and Other Wonders
≥
Cognitive Linguistics Research 33
Editors Dirk Geeraerts Rene´ Dirven John R. Taylor Hono...
106 downloads
1667 Views
27MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Words and Other Wonders
≥
Cognitive Linguistics Research 33
Editors Dirk Geeraerts Rene´ Dirven John R. Taylor Honorary editor Ronald W. Langacker
Mouton de Gruyter Berlin · New York
Words and Other Wonders Papers on Lexical and Semantic Topics by Dirk Geeraerts
Mouton de Gruyter Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin
앝 Printed on acid-free paper 앪
which falls within the guidelines of the ANSI to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data Geeraerts, Dirk, 1955⫺ Words and other wonders : papers on lexical and semantic topics / by Dirk Geeraerts. p. cm. ⫺ (Cognitive linguistics research ; 33) Includes bibliographical references. ISBN-13: 978-3-11-019042-7 (hardcover : alk. paper) ISBN-10: 3-11-019042-7 (hardcover : alk. paper) 1. Semantics. 2. Lexicology. I. Title. P325.G43 2006 4011.43⫺dc22 2006018773
Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at ⬍http://dnb.ddb.de⬎.
ISBN-13: 978-3-11-019042-7 ISBN-10: 3-11-019042-7 ISSN 1861-4132 쑔 Copyright 2006 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 or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany
Preface
Granunar and lexicon being traditionally separated in mainstream twentiethcentury linguistics, one of the great things about Cognitive Linguistics is the opening it has created towards a renewed study of the lexicon and an integration of lexical semantics into theoretical granunar at large. In fact, the study of the lexicon plays a crucial role in Cognitive Linguistics: questions about conceptual structure and semantic mechanisms that lie at the heart of Cognitive Linguistics, belong to the time-honored expertise of lexical semantics. In the present collection, I bring together a representative sample of my own contributions to lexicology and lexical semantics within the framework of Cognitive Linguistics (including the extension of insights from lexical semantics to other areas of linguistics). In the fifteen years covered by this collection, from 1988 to 2003, the central topics of my research led to a number of monographs: Diachronic Prototype Semantics (1997) synthesizes the research into historical semasiology that dominated much of my work in the 1980s, and two co-authored monographs, The Structure of Lexical Variation (1994) and Convergentie en Divergentie in de Nederlandse Woordenschat (1999) embody the corpus-based studies of onomasiological variation and change that I developed with Dirk Speelman, Stefan Grondelaers and the other members of my Leuven research group in the 1990s, and that continue to dominate our present research efforts. While these two central fields of research are only indirectly represented in the present volume, this collection predominantly brings together a number of papers devoted to lexicological and semantic topics that I dealt with more sporadically or in parallel to these major foci. Thematically speaking, I have consistently tried to choose papers that represent a variety of issues and topics within the broadly defined field of lexicology and lexical semantics. Accordingly, the chapters are organized in thematic groups. The first section deals with prototypicality as a theoretical and practical model of semantic description. The second section discusses polysemy and criteria for distinguishing between meanings. The third section tackles questions of meaning description beyond the level of words, on the level of idioms and constructions. The following section casts the net even wider, dealing with the cultural aspects of meaning. Moving away from the theoretical and descriptive perspective towards applied concerns, the fifth section looks at lexicography from the point of view of Cognitive Linguis-
VI
Preface
tics. The final section has a metatheoretical orientation: it discusses the history and methodology of lexical semantics. Theoretically, the papers are interconnected by a number of accents that mark their specific position within the varied landscape of Cognitive Linguistics: a continued attempt to provide solid empirical foundations for semantics, an emphasis on the multivariate nature of semantic structure, and an insistence on the social and cultural background of language. In the research goals that we are currently pursuing in Leuven, these accents combine in multifactorial corpus-based studies in the field of cognitive sociolinguistics - studies, that is, in which corpus materials provide the firm empirical basis for research, in which linguistic phenomena are statistically analysed with the help of multivariate techniques, and in which social and cultural variation is explicitly included into the multifactorial model. Except for necessary corrections, small additions, and some typographical reorganization (and unless otherwise stated), the chapters are reprinted in their original form. All chapters are reprinted with permission of the original publishers. Each paper is preceded by a brief introduction that situates the text against the period in which it was first written, but that also points to further developments, in my own research or in Cognitive Linguistics at large. In this way, the collection should give the reader an idea of the dynamism of Cognitive Linguistics in the crucial period when it developed from a more or less marginal approach to one that is firmly situated in the mainstream of contemporary linguistics. Knowledge is not produced in isolation, and it befits any researcher to thank those who have provided guidance and motivation in the course of his career. In gratitude, I dedicate this collection to the memory of Brygida Rudzka-Ostyn (1939-1998). Brygida's impact on the international expansion of Cognitive Linguistics can hardly be underestimated. With her infinite enthusiasm and her sense for initiative, she played a crucial rol in the second half of the 1980s in bringing Cognitive Linguistics from its Californian homeground to Europe and beyond. Without the warm stimulus she then provided for a young scholar to place his work in a wider context, none of the papers collected here might ever have been written.
Table of contents
Preface Publication sources
V
IX
Section 1. Prototypicality and salience
1 2 3 4
Prospects and problems of prototype theory Where does prototypicality come from? The semantic structure of Dutch over Salience phenomena in the lexicon. A typology
3 27
48
74
Section 2. Polysemy
5 6
Vagueness's puzzles, polysemy's vagaries Classical definability and the monosemic bias
99 149
Section 3. Constructions and idioms 7 8
The semantic structure of the indirect object in Dutch The interaction of metaphor and metonymy in composite expressions
175 198
Section 4. Meaning and culture
9 10 11 12
Looking back at anger. Cultural traditions and metaphorical patterns Beer and semantics Cultural models of linguistic standardization Caught in a web of irony: Job and his embarassed God
227 252 272 307
Section 5. Lexicography
13 14
The lexicographical treatment of prototypical polysemy The definitional practice of dictionaries and the cognitive semantic conception of polysemy
327 345
Section 6. Theory and method in lexical semantics
15 16 17
Cognitive grammar and the history oflexical semantics The theoretical and descriptive development of lexical semantics Idealist and empiricist tendencies in cognitive semantics
367 398 416
Vlll
Table ofcontents
References Index of names Index of subjects
445 479 485
Publication sources
Chapter 1 - Geeraerts, Dirk. 1989. Prospects and problems of prototype theory. linguistics 27: 587-612. Chapter 2 - Geeraerts, Dirk. 1988. Where does prototypicality come from? In Topics in Cognitive Linguistics, Brygida Rudzka-Ostyn (ed.), 207-229. AmsterdamJPhiladelphia: John Benjamins. Chapter 3 - Geeraerts, Dirk. 1992. The semantic structure of Dutch over. Leuvense Bijdragen 81: 205-230. Chapter 4 - Geeraerts, Dirk. 2000. Salience phenomena in the lexicon. A typology. In Meaning and Cognition, Liliana Albertazzi (ed.), 125-136. AmsterdamJPhiladelphia: John Benjamins. Chapter 5 - Geeraerts, Dirk. 1993. Vagueness's puzzles, polysemy's vagaries. Cognitive Linguistics 4: 223-272. Chapter 6 - Geeraerts, Dirk. 1994. Classical definability and the monosernic bias. Rivista di Linguistica 6: 189-207. Chapter 7 - Geeraerts, Dirk. 1998. The semantic structure of the indirect object in Dutch. In The Dative 11. Theoretical Approaches, Willy Van Langendonck and Wiliam Van Belle (eds.), 185-210. AmsterdamJPhiladelphia: John Benjamins. Chapter 8 - Geeraerts, Dirk. 2002. The interaction of metaphor and metonymy in composite expressions. In Metaphor and Metonymy in Comparison and Contrast, Rene Dirven and Ralf Parings (eds.), 435-465. BerlinINew York: Mouton de Gruyter. Chapter 9 - Geeraerts, Dirk, and Stefan Grondelaers. 1995. Looking back at anger. Cultural traditions and metaphorical patterns. In Language and the Construal of the World, John Taylor and Robert E. MacLaury (eds.), 153-180. BerlinINew York: Mouton de Gruyter. Chapter 10 - Geeraerts, Dirk. 1999. Beer and semantics. In Issues in Cognitive Linguistics, Leon De Stadler and Christoph Eyrich (eds.), 35-55. BerlinINew York: Mouton de Gruyter. Chapter 11 - Geeraerts, Dirk. 2003. Cultural models of linguistic standardization. In Cognitive lvfodels in Language and Thought. Ideology, Metaphors and Meanings, Rene Dirven, Roslyn Frank, and Martin Piitz (eds.), 25-68. BerlinINew York: Mouton de Gruyter.
x
Publication sources
Chapter 12 - Geeraerts, Dirk. 2003. Caught in a web of irony: Job and his embarassed God. In Job 28. Cognition in Context, Ellen Van Wolde (ed.), 37-55. Leiden: Brill. Chapter 13 - Geeraerts, Dirk. 1990. The lexicographical treatment of prototypical polysemy. In Meanings and Prototypes. Studies in Linguistic Categorization, Savas L. Tsohatzidis (ed.), 195-210. London: Routledge. Chapter 14 - Geeraerts, Dirk. 2001. The definitional practice of dictionaries and the cognitive semantic conception of polysemy. Lexicographica 17: 6-21. Chapter 15 - Geeraerts, Dirk. 1988. Cognitive grammar and the history oflexical semantics. In Topics in Cognitive Linguistics, Brygida Rudzka-Ostyn (ed.), 647677. Amsterdam/Philadelphia: John Benjamins. Chapter 16 - Geeraerts, Dirk. 2002. The theoretical and descriptive development of lexical semantics. In The Lexicon in Focus. Competition and Convergence in Current Lexicology, Leila Behrens and Dietmar Zaefferer (eds.), 23-42. Frankfurt: Peter Lang Verlag. Chapter 17 - Geeraerts, Dirk. 1999. Idealist and empiricist tendencies in cognitive linguistics. In Cognitive Linguistics: Foundations, Scope, and Methodology, Theo Janssen and Gisela Redeker (eds.), 163-194. BerlinINew York: Mouton de Gruyter.
Section 1 Prototypicality and salience
Chapter 1 Prospects and problems of prototype theory
Originally published in Linguistics 1989,27: 587-612. Parallel to the research that I did on the application of prototype theory to questions of semantic change (which eventually resulted in my Diachronic Prototype Semantics monograph of 1997), I paid attention on a number of occasions to the theoretical clarification of the notion of prototypicality as such (and of the related notion of polysemy: see the second section of this collection). In the paper reprinted here, I try to bring some systematicity into the many uses of the concept 'prototypicality' by distinguishing between two cross-classifying dimensions. First, I make a distinction between two crucial structural phenomena underlying prototypicality effects: flexibility (i.e. the absence of clear boundaries and demarcations) and salience (i.e. differences of structural weight). Second, I suggest that both prototypicality phenomena may be found on an intensional level (the level of definitions) and on an extensional level (the level of referents). The crossclassification of the two dimensions defines four basic types of prototypicality effects. The chapter originally appeared as the introductory paper of a thematic issue of the journal Linguistics. A section of the original paper describing the various contributions to the thematic issue has been omitted from the present reprint.
1.
Prototype theory within linguistics
The starting-point of the prototypical conception of categorial structure is sununarized in the statement that when describing categories analytically, most traditions of thought have treated category membership as a digital, all-or-none phenomenon. That is, much work in philosophy, psychology, linguistics, and anthropology assumes that categories are logical bounded entities, membership in which is defined by an item's possession of a simple set of criterial features, in which all instances possessing the criterial attributes have a full and equal degree of membership. In contrast, it has recently been argued ... that some natural categories are analog and must be represented logically in a manner which reflects their analog structure (Roscll and Mervis 1975: 573574).
4
Prototypicality and salience
As we shall see in section 2, the exact definition of the concept of prototypicality as used in linguistics is not without problems. The major part of this introduction to the prototypicality-based studies collected here will, in fact, consist of an attempt at clarification of some of the problematic aspects of the way in which the notion of prototype has been used in linguistics. To begin with, however, we shall be concerned with a brief overview of the state of the art in linguistic prototype theory. I The theory originated in the mid 1970s with Eleanor Rosch' s research into the internal structure of categories. (Overviews may be found in Rosch 1978, 1988, and Mervis and Rosch 1981; the basic research is reported on mainly in Heider 1972, Rosch 1973, 1975, 1977, Rosch and Mervis 1975, Rosch, Simpson and Miller 1976, Rosch et al. 1976.) From its psycholinguistic origins, prototype theory has moved mainll in two directions. On the one hand, Rosch's findings and proposals were taken up by formal psycholexicology (and more generally, information-processing psychology), which tries to devise formal models for human conceptual memory and its operation, and which so, obviously, borders on Artificial Intelligence. Excellent overviews of the representational and experimental issues at stake here are Smith and Medin (1981), and Medin and Smith (1984); an interesting sample of current research may be found in Neisser (1987). On the other hand, prototype theory has had a steadily growing success in linguistics since the early 1980s, as witnessed by a number of recent monographs and collective volumes in which prototype theory and its cognitive extensions play a major role (Wierzbicka 1985, Lakoff 1987, Langacker 1987, Craig 1986, Holland and Quinn 1987, Rudzka-Ostyn 1988, Lehmann 1988a, Hullen and Schulze 1988, Tsohatzidis 1989, Taylor 1989). It is with the latter development that we shall be concerned with here. Against the background of the development of linguistic semantics, prototype theory may be defined primarily in contrast with the componential model of semantic analysis that was current in transformational grammar and that is stereotypically associated with Katz and Fodor's analysis of bachelor (Katz and Fodor 1963); in an early defense of a prototypical approach, Fillmore (1975) called this the 'checklist theory' of meaning. The prototypists' reaction against this featural approach had, however, the negative side-effect of creating the impression that prototypical theories rejected any kind of componential analysis. This is a misconception for the simple reason that there can be no semantic description without some sort of decompositional analysis. As a heuristic tool for the description and comparison of lexical meanings, a componential analysis retains its value (a value
Prospects and problems ofproto(vpe theory
5
that, incidentally, it did not acquire with the advent of componential analysis as an explicit semantic theory, but which had been obvious to lexicographers from time immemorial). Rather, the difficulties with the neostructuralist kind of feature analysis that grew out of structuralist field theory lie elsewhere; it is not the use of decomposition as a descriptive instrument that causes concern, but the status attributed to the featural analysis. Two important points have to be mentioned. In the first place, as suggested by the quotation at the beginning of this introduction, featural definitions are classically thought of as criterial, i.e. as listing attributes that are each indispensable for the definition of the concept in question, and that taken together suffice to delimit that concept from all others. In contrast, prototype theory claims that there need not be a single set of defining attributes that conform to the necessity-cum-sufficiency requirement. 3 In the second place, prototype theory is reluctant to accept the idea that there is an autonomous semantic structure in natural languages which can be studied in its own right, in isolation from the other cognitive capacities of man. In particular, meaning phenomena in natural languages cannot be studied in isolation from the encyclopedic knowledge individuals possess; it is precisely the presupposition that there exists a purely linguistic structure of semantic oppositions that enables structuralist and neostructuralist semantics to posit the existence of a distinction between semantic and encyclopedic knowledge. Prototype theory tends to minimize the distinction primarily for methodological reasons: because linguistic categorization is a cognitive phenomenon just like the other cognitive capacities of man, it is important to study it in its relationship to these other capacities. More specific arguments have also been formulated to show that the distinction between an encyclopedic and a semantic level of categorial structure is untenable. 4 For instance, given that the flexible extendibility of prototypical concepts is a synchronic characteristic of linguistic structure, and given the fact that these extensions may be based indiscriminately on allegedly encyclopedic or on allegedly semantic features, the distinction between both kinds of information loses its synchronic relevance. Take the case of metaphor: before lion acquires the meaning 'brave man', the feature 'brave' is not structurally distinctive within the semasiological structure of lion, and hence, it has to be considered encyclopedic according to structuralist theories. But if it can be accepted (and this is of course the crucial point) that the metaphorical extension of lion towards the concept 'brave man' is not just a question of diachronic change, but is merely an effect of the synchronic flexibility of lexical items,
6
Prototypicality and salience
the feature clearly acquires semantic status. If, furthermore, the argument can be repeated in the sense that such synchronic metaphorical extensions may be based on any allegedly encyclopedic attribute, the distinction be5 tween semantic and encyclopedic concepts as a whole falls. The matter need not, to be sure, be settled here. What is important for our introductory purposes is rather to see what exactly prototype theory objects to in componential theories of the Katzian type. First, the suggestion that lexical concepts are criterial in the classical sense, and second, the suggestion that there exists a purely linguistic level of conceptual structuring that is neatly separated from other, 'encyclopedic' forms of conceptual information, and that may thus be studied autonomously, in methodological isolation from other kinds of cognitive research. As against these points of view, prototype theory defends a non-criterial conception of categorial structure, and an interdisciplinary methodological perspective that takes into account relevant research from the other cognitive sciences. (The very transposition of the prototypical approach from experimental psychology to linguistics derives from this attitude.) But this historical positioning of prototype theory with regard to its immediate predecessors within the field of lexical semantics clearly does not explain why it has turned out to be such a successful alternative. Why did (and does) the prototypical approach appeal to a sizeable part of the linguistic community? On the one hand, the historical development of generative grammar had raised a considerable amount of interest in semantic matters. It should not be forgotten, in fact, that it was only after the incorporation of a semantic component into the transformational framework that Chomskyanism became internationally populac the universal appeal of the generative Standard Theory was at least partly due to the promises held by its Katzian semantic component. On the other hand, the promises were not fulfilled. Within the generative paradigm, Generative Semantics (which most strongly embodied the semantic approach) withered in favor of Autonomous Syntax, in which semantics hardly played a role worthy of note. Outside the generative approach, formal semantics of the Montagovian kind was too narrowly restricted to sentential meaning to be able to hold the attention of those who were interested primarily in the internal structure of natural language categories (and not primarily in the way these categories combine into larger unities).6 In short, as far as semantics was concerned, there was a gap in the linguistic market of the early 1980s that was not filled by the major approaches of the day. 7
Prospects and problems ofprototype theory
7
But again, recognizing that there was an interest in the semantics of natural language categories to which prototype theory could appeal does not tell the whole story. Why didn't people simply stick to the componential theory popularized by Katz, or to the rival axiomatic method of representation even if these gradually moved out of the centre of the linguistic attention as Autonomous Syntax and Formal Semantics took over? In general, there are a number of methodological requirements people nowadays expect of linguistic theories: descriptive adequacy (mainly in the form of a broad empirical scope), explanatory depth, productivity, and formalization. Although prototype theory rates much lower on the formalization scale than either the axiomatic or the featural approach, its assets with regard to the other three points are considerable. In the first place, it tackles a number of semantic phenomena that had been swept under the rug by the more structurally minded approaches. The fuzzy boundaries of lexical categories, the existence of typicality scales for the members of a category, the flexible and dynamic nature of word meanings, the importance of metaphor and metonymy as the basis of that flexibility - these are all intuitively obvious elements of the subject matter of semantics that were largely neglected by structural semantics. It is true that they were occasionally pointed at as an indispensable aspect of any fullfledged semantic theory: think, for instance, of Weinreich's remark (1966: 471) that a semantic theory should be able to deal with 'interpretable deviance', or Uhlenbeck's plea (1967) for a dynamic conception of word meaning. 8 These remarks did not, however, have much effect as far as theory formation was concerned. In particular, it is only with the advent of prototype theory that contemporary linguistics developed a valid model for the polysemy of lexical items. This is perhaps the single most appealing characteristic of prototype theory: here at last is a descriptive approach to lexical meaning in which our pretheoretical intuitions about gradedness, fuzziness, flexibility, clustering of senses etc. receive due attention. In the second place, prototype theory appears to be a productive theory not just in the sense that its insights into the structure of lexical categories can be easily applied in various fields of the lexicon, but also in the sense that it may be extended towards other aspects of linguistics. Whereas prototype theory started with being descriptively fruitful in lexical semantics, it soon became theoretically fruitful in the sense that other areas of linguistics were taken into consideration. A few recent examples of such extensions may suffice: phonology (Nathan 1986), morphology (Bybee and Moder 1983, Post 1986), syntax (Van Oosten 1986, Ross 1987), historicallinguis-
8
Prototypicality and salience
tics (Winters 1987, Aijmer 1985), markedness theory (Van Langendonck 1986), theoretical lexicography (Geeraerts 1985b). Through these and similar extensions,9 prototype theory has become one of the cornerstones of Cognitive Linguistics, which tries to account for the interaction between language and cognition on all levels of linguistic structure: one need only have a look at the prominent place attributed to a prototypical conception of categorial structure in Langacker (1987) (one of the basic works of the Cognitive Linguistic approach) to appreciate its importance. IQ In this sense, the development of prototype theory into Cognitive Linguistics contains exciting promises of a unified cognitive theory of linguistic categorization. In the third place, the explanatory depth of prototype theory resides partly in its generalizable character, but also in its interdisciplinary nature. The importance of its genetic link with psycholinguistics can only be fully appreciated against the background of the Chomskyan requirements with regard to theories of grammar. Chomsky' s methodology is, in fact, in the awkward position of declaring linguistics a cognitive science, but refusing to deal directly with the findings of the other sciences of the mind. Roughly stated, Chomskyan linguistics claims to reveal something about the mind, but imperviously prefers a strictly autonomist methodology over the open dialogue with psychology that would seem to be implied by such a claim. Prototype theory's linguistic application of psycholinguistic findings, on the other hand, takes the Chomskyan ideal of cognitive explanatory depth to its natural consequences, viz. of giving up the methodological autonomy of linguistics in favor of an interdisciplinary dialogue with the other cognitive sciences. l1 Prototype theory takes the cognitive claims of Chomskyanism methodologically seriously by its interdisciplinary openness. This is all the more important at a moment when Cognitive Science is emerging as an interdisciplinary cluster of psychology, neuroscience, Artificial Intelligence, and philosophy. It is probably one of the reasons for the appeal of prototype theory that its interdisciplinary connections hold the promise of linking linguistics to the most important development that the human sciences are currently witnessing.
2. Definitional problems, first series: 'Prototype' as a prototypical notion The appeal of prototype theory should not, however, obscure the fact that the exact definition of prototypicality is not without problems. The purpose
Prospects and problems ofprototype theory
9
of this section (and the following) is to analyze the sources of the confusion by making clear that prototypicality is itself, in the words of Posner (1986), a prototypical concept. As a first step, we shall have a look at four characteristics that are frequently mentioned (in various combinations) as typical of prototypicality. In each case, a quotation from early prototype studies is added to illustrate the point. (i) Prototypical categories cannot be defined by means of a single set of criterial (necessary and sufficient) attributes: We have argued that many words ... have as their meanings not a list of necessary and sufficient conditions that a thing or event must satisfy to count as a member of the category denoted by the word, but rather a psychological object or process which we have called a prototype (Coleman and Kay 1981: 43). (ii) Prototypical categories exhibit a family resemblance structure, or more generally, their semantic structure takes the form of a radial set of clustered and overlapping meanings: 12 The purpose of the present research was to explore one of the major structural principles which, we believe, may govern the fonnation of the prototype structure of semantic categories. This principle was first suggested in philosophy; Wittgenstein (1953) argued that the referents of a word need not have common elements to be understood and used in the nonnal functioning of language. He suggested that, rather, a family resemblance might be what linked the various referents of a word. A family resemblance relationship takes the form AB, BC, CD, DE. That is, each item has at least one, and probably several, elements in common with one or more items, but no, or few, elements are common to all items (Rosch and Mervis 1975: 574-575).
(iii) Prototypical categories exhibit degrees of category membership; not every member is equally representative for a category: By prototypes of categories we have generally meant the clearest cases of category membership defined operationally by people's judgments of goodness of membership in the category ... we can judge how clear a case something is and deal with categories on the basis of clear cases in the total absence of information about boundaries (Rosch 1978: 36). (iv) Prototypical categories are blurred at the edges: New trends in categorization research have brought into investigation and debate some of the major issues in conception and learning whose solution
10
Proto(vpicality and salience
had been unquestioned in earlier approaches. Empirical findings have established that ... category boundaries are not necessarily definite (Mervis and Rosch 1981: 109). As a first remark with regard to these characteristics, it should be noted that they are not the only ones that may be used in attempts to define the prototypical conception of categorization. Two classes of such additional features should be mentioned. On the one hand, there are characteristics that do not pertain (as the four mentioned above) to the structure of categories, but that rather pertain to the epistemological features of so-called non-Aristotelian categories. 13 For instance, the view that prototypical categories are not 'objectivist' but 'experiential' in nature (Lakoff 1987) envisages the epistemological relationship between concepts and the world rather than the structural characteristics of those concepts. In particular, it contrasts the allegedly classical view that 'categories of mind ... are simply reflections of categories that supposedly exist objectively in the world, independent of all beings', with the view that 'both categories of mind and human reason depend upon experiential aspects of human psychology' (Lakoff 1982: 99). Such an epistemological rather than structural characterization of natural concepts also has a methodological aspect to it; it entails that prototypical categories should not be studied in isolation from their experiential context. While such an epistemological or methodological conception of prototypical categorization is extremely valuable, we shall take a structural point of view in the following pages; we shall try to determine whether it is possible to give a coherent, structurally intrinsic characterization of prototypical categories. On the other hand, there are structural characteristics of prototypical concepts that can be reduced to the four basic structural features mentioned above. For instance, in my own work on prototypical categorization, I have repeatedly stressed the flexibility of prototypical concepts (1983a, 1985a), together with the fact that a distinction between semantic and encyclopedic components of lexical concepts cannot be maintained in the case of prototypical concepts (l985b). But the flexibility of prototypical categories is linked in a straightforward manner with the fourth characteristic: uncertainties with regard to the denotational boundaries of a category imply that it need not be used in a rigidly fixed manner. Similarly, the absence of a clear dividing line between encyclopedic and purely semantic information follows from this very flexibility together with the first and second characteristic. As illustrated in the previous section, the possibility of incorporating members into the category that do not correspond in every definitional respect with the
Prospects and problems ofprototype theory
11
eXlstmg members entails that features that are encyclopedic (nondefinitional) with regard to a given set of category members may turn into defmitional features with regard to a flexibly incorporated peripheral category member. The resemblance between central and peripheral cases may be based on allegedly encyclopedic just as well as on allegedly 'semantic' features. In short, features of prototypicality that are not included among the ones mentioned in (i)-(iv) may often be reduced to those four, and this in turn justifies a preliminary restriction of the discussion to the latter. A second remark with regard to the four characteristics is concerned with the fact that they are systematically related along two dimensions. On the one hand, the third and the fourth characteristic take into account the referential, extensional structure of a category. In particular, they have a look at the members of a category; they observe, respectively, that not all referents of a category are equal in representativeness for that category, and that the denotational boundaries of a category are not always determinate. On the other hand, these two aspects (centrality and non-rigidity) recur on the intensional level, where the definitional rather than the referential structure of a category is envisaged. For one thing, non-rigidity shows up in the fact that there is no single necessary and sufficient definition for a prototypical concept. For another, family resemblances imply overlapping of the subsets of a category. To take up the formulation used in the quotation under (ii) above, if there is no definition adequately describing A, B, C, D, and E, each of the subsets AB, BC, CD, and DE can be defined separately, but obviously, the 'meanings' that are so distinguished overlap. Consequently, meanings exhibiting a greater degree of overlapping (in the example: the senses corresponding with BC and CD) will have more structural weight than meanings that cover peripheral members of the category only. In short, the clustering of meanings that is typical of family resemblances implies that not every meaning is structurally equally important (and a similar observation can be made with regard to the components into which those meanings may be analyzed). The systematic links between the characteristics mentioned at the beginning are schematically summarized in Table 1. As a third remark, it should be noted that the four characteristics are often thought to be co-extensive, in spite of incidental but clear warnings such as Rosch and Mervis's remark that a family resemblance structure need not be the only source of prototypicality (1975: 599). Admittedly, it is easy to consider them to be equivalent; already in the quotations given above, partial reasons for their mutual interdependence can be found. More systematically,
12
Prototypicality and salience
the following links between the four characteristics might be responsible for the idea that prototypicality necessarily entails the joint presence of all four. Table 1. Characteristics of prototypicality NON-EQUALITY
NON-RIGIDITY
differences in structural weight
flexibility and vagueness
EXTENSIONAL
degrees of representativity
absence of clear boundaries
INTENSIONAL
clusters of overlapping senses
absence of classical definition
First, linking the first to the second characteristic is the argument mentioned above: if there is no single definition adequately describing the extension of an item as a whole, different subsets may be defined, but since the members of a category can usually be grouped together along different dimensions, these subsets are likely to overlap, i.e., to form clusters of related mearungs. Second, linking the second to the third characteristic is the idea that members of a category that are found in an area of overlapping between two senses carry more structural weight than instances that are covered by only one meaning. Representative members of a category (i.e., instances with a high degree of representativity) are to be found in maximally overlapping areas of the extension of a category. (In the example, A and E are less typical members that B, C, and D, which each belong to two different subsets.) Third, linking the third to the fourth characteristic is the idea that differences in degree of membership may diminish to a point where it becomes unclear whether something still belongs to the category or not. Categories have referentially blurred edges because of the dubious categorial status of items with extremely low membership degrees. And fourth, linking the fourth to the first characteristic is the idea that the flexibility that is inherent in the absence of clear boundaries prevents the formulation of an essence that is common to all the members of the category. Because peripheral members may not be identical with central cases but may only share some characteristics with them, it is difficult to define a set of attributes that is common to all members of a category and that is sufficient to distinguish that category from all others. These circular links between the four characteristics are, however, misleading. A closer look at some (familiar and less familiar) examples of prototypicality reveals that they need not co-occur.
Prospects and problems ofprototype theory
13
Bllill
The concept bird (one of Rosch's original examples of prototypicality) shows that natural categories may have clear-cut boundaries. At least with regard to our own, real world, the denotation of bird is determinate; educated speakers of English know very well where birds end and non-birds begin. They know, for instance, that a bat is not a bird but that a penguin is. Of course, the principled indeterminacy described by Waismann (1952) as 'open texture' remains: when confronted with an SF creature (a post-World War III mutant) that looks like a bird but talks like a man, we would not be sure whether it should be called a bird or not. A boundary problem that is typical for a prototypical organization of the lexicon would then arise. As it functions now, however, in present-day English, bird is denotationally clearly bounded, the archaeopteryx notwithstanding. 14 As has been remarked elsewhere (Lakoff 1987), the existence of prototypicality effects in clearly bounded concepts such as bird implies that a strict distinction has to be made between degree of membership and degree of representativity. Membership in the category bird is discrete; something is or is not a bird. But some birds may be birdier than others: the swallow does remain a more typical bird than the ostrich. RED
Color terms such as red constituted the starting-point for prototypical research; drawing on the views developed in Berlin and Kay (1969), Rosch's earliest work is an experimental demonstration of the fact that the borderline between different colors is fuzzy (there is no single line in the spectrum where red stops and orange begins), and of the fact that each color term is psychologically represented by focal colors (some hues are experienced as better reds than others) (Heider 1972, Heider and Olivier 1972). These prototypical characteristics on the extensionallevel are not matched on the definitionallevel. If red can be analytically defined at all (i.e., if it does not simply receive an ostensive definition consisting of an enumeration of hues with their degree of focality), its definition might be 'having a color that is more like that of blood than like that of an unclouded sky, that of grass, that of the sun, that of ... (etc., listing a typical exemplar for each of the other main colors)'. Such a definition (cp. Wierzbicka 1985: 342) does not correspond with either the first or the second characteristic mentioned above. ODD N1J};ffiER
Armstrong, Gleitman and Gleitman (1983) have shown experimentally that even a mathematical concept such as odd number exhibits psychological
14
Prototypicality and salience
representativity effects. This might seem remarkable, since odd number is a classical concept in all other respects: it receives a clear definition, does not exhibit a family resemblance structure or a radial set of clustered meanings, does not have blurred edges. However, Lakoff (1982) has made clear that degrees of representativity among odd numbers are not surprising if the experiential nature of concepts is taken into account. For instance, because the even or uneven character of a large number can be detennined easily by looking at the final number, it is no wonder that uneven numbers below 10 carry more psychological weight: they are procedurally of primary importance. VERS
As I have tried to show elsewhere (1988a), the first characteristic mentioned above is not sufficient to distinguish prototypical from classical categories, since, within the classical approach, the absence of a single definition characterized by necessity-cum-sufficiency might simply be an indication of polysemy. This means that it has to be shown on independent grounds that the allegedly prototypical concepts are not polysemous, or rather, it means that prototypical lexical concepts will be polysemous according to a definitional analysis in tenns of necessary and sufficient conditions (the classical definition of polysemy), but univocal according to certain other criteria. These criteria may be found, for instance, in native speakers' intuitions about the lexical items involved, intuitions that may be revealed by tests such as Quine's (1960) or Zwicky and Sadock's (1975). In this sense, the first characteristic has to be restated: prototypical categories will exhibit intuitive univocality coupled with analytical (definitional) polysemy, and not just the absence of a necessary-and-sufficient definition. Once this revision of the first characteristic is accepted, it can be demonstrated that the first and the second criterion need not co-occur. Lexical items that show clustered overlapping of senses may either confonn or not confonn to the revised first characteristic. An example of the first situation is the literal meaning of bird, an example of the second situation the Dutch adjective vers, which corresponds roughly with Englishfresh (except for the fact that the Dutch word does not carry the meaning 'cool'). Details of the comparison between both categories may be found in the paper mentioned above; by way of summary, Figures 1 and 2 represent the definitional analysis of both items. The distinction in intuitive status between vers and bird can be demonstrated by means of the Quinean test (roughly, a lexical item is ambiguous if it can be simultaneously predicated and negated of something
Prospects and problems ofprototype theory ............................................................................................................························6····· 7
15 ,
kiwi o_s_tr_i_ch_-t_:hiCkeJ
y
3---+-------------+-----+-----'
penguin 5---t--------------' L--------------4------'
1 being able to fly 4 having wings 7 having a beak or bill
2 having feathers 5 not domesticated
3 being S-shaped 6 being born from eggs
Figure J. A definitional analysis of bird
in a particular context). Thus, taking an example based on the corresponding ambiguity in the English counterpart of vers, it would be quite normal to state that the news meant in the sentence there was no fresh news from the jightini 5 is fresh in one sense ('recent, new') but not in another ('in optimal condition'): it makes sense to say that the news is at the same time fresh and not fresh. By contrast, it would be intuitively paradoxical to state that a penguin is at the same time a bird and not a bird (disregarding figurative extensions of the semantic range of bird). Nevertheless, the definitional analyses in Figures 1 and 2 make clear that both concepts exhibit prototypical clustering. In both cases, too, the structural position of the instances just discussed (news, penguin) is not in the central area with maximal overlapping. In short, then, the revised version of the first characteristic need not coincide with the second characteristic. The insight derived from a closer look at the four examples just described may be summarized as in Table 2. It is now easy to see to what extent 'prototypicality' is itself a prototypical notion. There is no single set of attributes that is common to all of the examples discussed here. Rather, they exhibit a family resemblance structure based on partial similarities. In this sense, the set of prototypical concepts characterized by clustering of senses overlaps
16
Prototypicality and salience
e.g. infonnation
1
2
e.g. fruit and other foodstuffs
'---- 1
e.g. air 1 new, novel, recent 2 in an optimal condition, pure, untainted
Figure 2. A definitional analysis of vers with the subset characterized by fuzzy boundaries (because of vers), and so on. At the same time, some concepts are more typically prototypical than others. (Bird and vers are more prototypical than red.) Notice, in particular, that the category fruit makes a good candidate for prototypical prototypicality, in the sense that it seems to combine all four characteristics. It shares the prototypical characteristics of bird, but in addition, things such as coconuts and, perhaps, tomatoes, seem to point out that the denotational boundary of fruit is less clear-cut than that of bird. However, although the examples considered above do not have a set of attributes in common, they do share a single feature, viz. degrees of membership representativity. It is highly dubious, though, whether this feature alone suffices to distinguish prototypical concepts from classical concepts. If the possibility of a single necessary-and-sufficient definition is one of the features par excellence with which the classical conception has been identified, it might justifiably be claimed that degrees of representativity are entirely compatible with the classical conception of categorization. It is, in fact, in that sense that Annstrong, Gleitman and Gleitman (1983) deal with a category such as odd number. The experiments used by Rosch to measure degrees of representativity are not, they claim, indicative of prototypicality since they occur with classical, rigidly definable concepts such as odd number. To say the least, representativity effects are only a peripheral prototypi-
Prospects and problems ofprototype theory
17
cal attribute according to Table 2 (cp. Lakoff 1986). But at the same time, the debate over the status of odd number shows that the concept 'prototypical concept' has no clear boundaries: it is not immediately clear whether a concept such as odd number should be included in the set of prototypical concepts or not. Table 2. The prototypicality of 'prototypicality' P::: ~
Cl P:::
iQ
absence of classical definition clusters of overlapping senses degrees of representativity absence of clear boundaries
+ + +
o:l
P:::
Cl
>
P:::
~
+ + +
~
+ +
Cl Cl
0
::2 ~
z
+
Of course, contrary to the situation in everyday speech, such a boundary conflict should not be maintained in scientific speech. A discipline such as linguistics should try to define its concepts as clearly as possible, and the purpose of this section is precisely to show that what has intuitively been classified together as instances of prototypical categories consists of distinct phenomena that have to be kept theoretically apart. In line with prototype theory itself, however, such an attempt at clear definition should not imply an attempt to define the 'true nature' or the 'very essence' of prototypicality. Determining an 'only true kind' of prototypicality is infinitely less important than seeing what the phenomena are and how they are related to each other by contrast or similarity. Still, there might seem to be one way in which decent sense could be made of the question what the true meaning of prototypicality would be. To begin with, let us note that the prototypical character attributed to the concept of prototypicality also shows up in the fact that the notion 'prototype' is an extremely flexible one. This can be illustrated in two ways. First, the lexical item prototypical is spontaneously used to name a number of phenomena that are linked by metonymy, next to the phenomena linked by similarity that are brought together in Table 2. The lexical item does not only characterize structural features of concepts, and the concepts exhibiting those features themselves, but sometimes even particular (viz., highly representative) instances of the categories in question (the robin as a prototypical
18
Prototypicality and salience
bird). Second, context may stress one feature of prototypical organization rather than another (cp. the priming effects in Rosch 1975). The general purpose of one's investigations may lead one to devote more attention to one aspect of the prototypical cluster than to another. To name a few examples: degrees of representativity are important for language development studies (if it is taken into account that most concepts in early language development are acquired via their exemplars), while clustered overlapping of senses will come to the fore in linguistic or lexicographical studies into the structure of polysemy. And a cognitive interest into the epistemological principles underlying natural language will attach more weight to the decoupling of intuitive univocality and analytical, definitional polysemy.16 In this respect, the question with regard to the true nature of prototypicality might be transformed into the question what might be the most interesting (or perhaps even the most important) perspective for studying and defining prototypicality. But here again, the 'ultimate essence fallacy' exposed by prototype theory itself lurks round the corner: there will be different preferences for one perspective rather than another, but there will be no single ultimately and eternally most important conception of prototypicality. In short, the foregoing analysis corroborates Wierzbicka's remark that there are 'many senses' to the notion prototype, and that 'the notion prototype has been used in recent literature as a catch-all notion' (1985: 343). However, a more systematic analysis than Wierzbicka's reveals that this very multiplicity of usage also supports Cognitive Semantics, in the sense that it shows that the same categorization principles may guide common sense and scientific thinking. This is, then, a further indication of the metatheoretical relevance of a cognitive conception of linguistic categorization, which I have explored at length elsewhere (1985b). At the same time, it has become clear that one of the major tasks for the further development of prototype theory is the closer investigation of the prototypically clustered characteristics of prototypicality. A major reference in this respect is Lakoff's attempt (1987: Chapter 4-8) to determine which different kinds of conceptual models may lie at the basis of prototypicality effects.
3. Definitional problems, second Series: 'Prototype theory' as a prototypical notion Whereas the previous section made clear that prototypicality as used in linguistic semantics is a prototypically structured concept, it should now be
Prospects and problems ofprototype theory
19
noted that the prototype-theoretical movement as well is a prototypically structured approach to semantics. There are, in other words, central as well as more peripheral examples of prototypical theories. In particular, there exist a number of theories that combine aspects of the classical approach to semantic structure with aspects of the prototypical conception. In this section, two approaches will be considered that are to some extent semiclassical as well as semi-prototypical; each of both embodies a strategy for reinstating particular aspects of the classical view against the background of an overall cognitive point of view. To begin with, some of the clarity and neatness of the classical approach may be recovered by concentrating on the prototypical centre of a category. If the non-classical indeterminacy of lexical concepts stems primarily from the flexible extendibility of concepts, discreteness may be reinstalled by avoiding the problems of clustered polysemy, i.e., by restricting the definitional analysis to the prototypical centre of the category. This approach is vigorously carried through by Wierzbicka (1985), who explicitly defends the discreteness of semantics by introspectively considering only the clear, salient centre of lexical categories. In a discussion of Labov's experimental investigation into the non-classical characteristics of everyday concepts (1973), she notes: 17 To state the meaning of a word, it is not sufficient to study its applicability to things; what one must do above all is to study the structure of the concept which underlies and explains that applicability. In the case of words describing natural kinds or kinds of human artefacts, to understand the structure of the concept means to describe fully and accurately the idea (not just the visual image) of a typical representative of the kind: the prototype. And to describe it fully and accurately we have to discover the internal logic of the concept. This is best done not through interviews, not through laboratory experiments, and not through reports of casual, superficial impressions or intuitions ... but through methodical introspection and thinking (1985: 19). It should be noted immediately that Wierzbicka's reinstatement of discreteness does not imply that her definitions do in fact always consist of necessary-and-sufficient conditions, and she acknowledges as much (1985: 60). In this respect, Wierzbicka's approach is only partly a departure from the hard core of prototype-theoretical studies: the absence of necessary-andsufficient conditions for the definition of certain core concepts is accepted, but the avoidance of the clustered polysemy problem 'tidies up' the semantic description and reinstates some of the classical neatness. Neither does
20
Prototypicality and salience
Wierzbicka's approach imply that lexical items are always univocal; in her dictionary of English speech act verbs (1987a), several items receive multiple definitions. Each of the definitions does, however, constitute a highly salient meaning, and again, by disregarding peripheral kinds of usage, the clustered or radial structure of the polysemy of lexical items does not enter the picture. The question to be asked, then, is whether Wierzbicka's restriction of the description to the salient meanings of a category is useful and adequate from a cognitive point of view. From a methodological point of view, the periphery of natural, nonuniquely definable categories is as interesting as their salient centre(s), because it is precisely the relationship between both that typically characterizes natural categories. Cognitive Linguistics is not only interested in what constitutes the centre of a category, but also in how this centre can be extended towards peripheral cases, and how far this extension can go. The mechanisms for incorporating marginal cases into a category at the same time restrict the flexibility of that concept; it is only by studying peripheral cases, for instance, that an answer may be found with regard to the question how dissimilar things can be before they are no longer recognized as basically the same. If, in other words, flexible polysemization is indeed one of the major characteristics of natural language categories, a deliberate restriction of the description to the salient meanings of a category is methodologically less propitious, as it may lead to a neglect of this basic feature. 18 A second strategy for salvaging aspects of the classical approach is to invoke sociolinguistic mechanisms such as Putnam's 'division of linguistic labor' (1975). According to Putnam, ordinary language users possess no more than 'stereotypical' knowledge about natural kinds, that is to say, they are aware of a number of salient characteristics, such as the fact that water is a transparent, thirst-quenching, tasteless liquid. The technical definition of water as H 20, on the other hand, is to be located primarily with scientific experts. It is the experts' knowledge that ultimately determines how natural kind terms are to be used. On the one hand, a 'division of linguistic labor' ensures that there are societal experts who know that water is H 2 0, that there is a distinction between elms and beech, how to recognize gold from pyrites, and so on. On the other hand, laymen attune their own linguistic usage to that of the expert scientists, technicians, etc.. The members of the non-specialized group are not required to have expert knowledge, but if they wish to be considered full-fledged members of the linguistic community, they are supposed to know the 'stereotype' connected with a category. A stereotype is, thus, a socially determined minimum set of data with regard to the
Prospects and problems ofproto(vpe theory
21
extension of a category. Given the similarity between Putnam's stereotypes and the prototypes of Cognitive Linguistics (both consist roughly of the most salient information connected with a category), the division of linguistic labor might be used to rescue the classical view of concepts. 19 Expert definitions being classical (they specify an essentialist 'hidden structure' for natural kinds), the stereotypical concepts of everyday language users might now be seen as hardly more than a sloppy derivative of those classically defined expert categories. 'True' (expert) definitions would be classical, and stereotypical/prototypical concepts might be dismissed as sociolinguistically secondary phenomena. It should be remarked immediately that such a reinstatement of the classical view is not as obvious for other words than the natural kind terms for which Putnam' s theory is in fact intended (what is the expert definition of the preposition for?). Moreover, as a sociolinguistic theory about the social factors that determine how lexical items may be used, the 'division of linguistic labor' theory is incomplete to say the least. The primacy of expert definitions would seem to imply that natural language follows the developments and discoveries of science in a strict fashion. In actual fact, however, natural language categorization is not only determined by the state of affairs in the sciences, but also by the communicative and cognitive requirements of the linguistic community in its own right. One of Putnam's own examples may serve as an illustration. Although science has discovered that jade refers to two kinds of materials (one with the 'hidden structure' of a silicate of calcium and magnesium, the other being a silicate of sodium and aluminium), ordinary usage continues to refer to both substances indiscriminately as jade. That is to say, categorization in everyday language is not entirely dependent upon scientific research, but seems to be determined at least in part by independent criteria: if the classificatory exigencies of everyday communicative interaction do not call for a distinction between the two kinds of jade, the scientific splitting of the category is largely ignored. This implies that an investigation into everyday language categorization as an independent cognitive system is justified. More generally, if Putnam's view is seen as a theory about the sociolinguistic structure of semantic norms, his hierarchical model (with experts at one end and laymen at the other) is only one among a number of alternatives, some of which (such as the one described by Bartsch 1985) link up closely with a prototypical conception of categorial structure. At the same time, however, it should be admitted that the relationship between classical scientific categorization and prototypical commonsense categorization may be explored in more depth than is yet the case. 20
22
Prototypicality and salience
To summarize: the confusion associated with the notion of prototypicality is further increased by the fact that more straightforwardly prototypical approaches are surrounded by hybrid theories that contain particular strategies for combining classical discreteness with typically prototypical phenomena. We have discussed two such approaches (one in which the strategy in question is methodological, and another one in which it is sociolinguistic), but 21 this does not mean that these are the only ones that might be mentioned. The two approaches mentioned here are, however, particularly revealing, as they link up with two important currents in the history of Western thought. The first one simplifyingly boils down to the view that the mind is neat (if you look hard enough into it), but that the world is fuzzy: the nondiscreteness that Cognitive Linguistics is concerned with arises from the fact that we have to apply clear-cut mental categories to an external reality that is so to say less well organized. The conception that the world of mental entities is somehow better organized than the outside world is obviously an idealistic one (though it does not constitute the only possible kind of idealism); Wierzbicka herself stresses the Platonist character of her approach. On the other hand, Putnam' s view that science is neat whereas everyday language is fuzzy, links up with the empiricist objectivism of the Ideal Language branch of analytical philosophy: the objective structure of reality is best described by the language of science, and everyday language is at best a weak derivative of scientific categorization, at worst a conceptual muddle teeming with philosophical pseudo-problems. As the previous discussion suggests that hard-core Cognitive Linguistics steers clear of both the idealist and the objectivist option, we have here one more indication 22 for the necessity of a further investigation into the epistemological, philosophical background of the prototypical conception of categorial structure.
Notes 1.
2.
The discussion in section 2 will make clear that the term proto~ype theory should be used with care, since the theoretical uniformity that it suggests tends to obliterate the actual distinctions between the diverse forms of prototypicality discussed in the literature. The term is used here as a convenient reference mark only, to indicate a number of related theoretical conceptions of categorial structure that share an insistence on any or more of the various kinds of prototypicality effects discussed in section 2. Though not exclusively: see Rosch (1988: 386).
Prospects and problems o/prototype theory
3.
4. 5.
6.
7.
8.
23
Notice that this claim applies just as well to the axiomatic, postulate-based form of description that developed as the major representational alternative for Katzian componential analysis. The notion of criteriality is just as much part and parcel of the classical versions of the axiomatic alternative as it is of Katzian feature analysis. See, among others, Haiman (1980a) and Geeraerts (1985b). The distinction between semantic and encyclopedic concepts against which Cognitive Semantics reacts is often misconstrued. In particular, in the statement that there is no principled distinction between semantic and encyclopedic information, the words semantic and encyclopedic are not used (as implied by Lehmann 1988b) in the senses 'as may be found in dictionaries' and 'as may be found in encyclopedias', respectively. Rather, the rejected distinction refers to an alleged distinction within an individual language user's conceptual memory; it involves the presupposition that there is an independent level of semantic information that belongs to the language and that is distinct from the individual's world knowledge. The kind of information that is typically found in encyclopedias involves scientific information ofthe kind 'ovulation triggered by copulation' for the item cat (the example is Lehmann's); but while the distinction between scientific and laymen's knowledge is primarily a social one, this kind of 'encyclopedic' information is only relevant for the psychological perspective of Cognitive Semantics if the individual lexicon to be described is that of someone with a certain amount of scientific knowledge of cats (or if, through sociolinguistic idealization, the average language user's lexicon may be supposed to contain that piece of scientific information). There are, of course, exceptions such as Dowty (1979) to confirm the rule. The historical sketch of the advent of prototype theory given here is treated more thoroughly in Geeraerts (1988a). As the semantic interests of the former audience of Generative Semantics were so to say no longer envisaged by the leading theories of the day, it does not come as a total surprise, from this point of view, to find George Lakoff, one of the leading Generative Semanticists, again as one of the leading cognitivists. These antecedents are not the only ones that might be mentioned. I have elsewhere (1988c) drawn the attention to the similarities between the prestructuralist, historical tradition of semantic research and present-day Cognitive Semantics, but there are other (admittedly non-mainstream) traditions of semantic research with which Cognitive Semantics is methodologically related: think, e.g., of the anthropological research of Malinowski, Firth, and the London School in general. Even a structuralist such as Reichling has held views about the structure of polysemy that come close to the point of view of prototype theory: his influential work on the word as the fundamental unit of
24
9.
10.
11.
12.
13.
Prototypicality and salience
linguistics (1935) contains an analysis of the Dutch word spel that is awkwardly similar to Wittgenstein's remarks about the German equivalent Spiel. The point to be stressed is this: as a theory about the (radial, clustered, dynamically flexible) structure of polysemy, prototype theory is to a considerable extent a rediscovery of views that were paramount during the prestructuralist era of the development of lexical semantics, and that lingered on below the surface in the structuralist and transfonnationalist periods. Because of their large scope, the functionalist approach of Seiler (1986) and the naturalist approach of Dressier (1985) are particularly interesting for the use of prototypicality with regard to various aspects of the fonnal organization of language. A bibliography of work in Cognitive Linguistics is to be found in Dirven (1988). It is worth mentioning that Cognitive Linguistics is currently in a stage of organization: a first international conference of Cognitive linguistics was held in Duisburg in March 1989, and a new journal entitled Cognitive Linguistics, published by Mouton, is scheduled to start appearing in the beginning of 1990. Next to the link with psycholinguistics, there is a connection with Artificial Intelligence research, through the correspondences between the notion of prototypicality and that of frame; see Fillmore (1977a). It needs to be stressed, though, that the link is relatively weak; specifically, the correspondence just mentioned is to a certain extent counterbalanced by Lakoff's criticism (1987) of the objectivist assumptions of mainstream Artificial Intelligence research (but then again, one of Lakoff's current research projects involves a connectionist approach to the fonnal modeling of Cognitive Semantic notions such as metaphorical image schemata). In general, sorting out the relationship between Cognitive Semantics and Artificial Intelligenceoriented Cognitive Science will be one of the major tasks for the further development of Cognitive Semantics. See Lakoff (1987: Chapter 6) for the notion of a radial set and compare Givon (1986) for a comparison between the views of Wittgenstein and those of prototype theory. The stress Givon places on the distinctions between both is slightly exaggerated, as it tends to obscure their mutual rejection of the socalled classical theory. See also the next footnote. The 'so-called' is added to stress, first, that the views of Aristotle also contain features that correspond rather with a cognitive than with a 'classical' approach, and second (more generally), that the philosophical position of prototype theory is in need of further elucidation. The present situation is rather muddled: while the classical Roschian position is to characterize prototype theory as non-Aristotelian and Wittgensteinian, Givon (1986) has argued that prototype theory is non-Wittgensteinian (see the previous note), but whereas Givon also describes prototype theory as non-Platonic, Wierzbicka
Prospects and problems ofprototype theory
14. 15. 16.
17. 18.
19.
25
(to whom we shall come back in section 3) precisely presents an explicitly Platonic version of prototype theory. More generally, the philosophical position of prototype theory has so far been discussed mainly against the background of classical philosophy (Aristotle and Plato), and against the background of contemporary analytical philosophy (see Lakoff 1987). This means that a large part of the history of Western philosophy passes unmentioned; this is to be regretted, as the post-Cartesian period in the history of philosophy is concerned with epistemological questions that are of immediate interest to Cognitive Semantics. In particular, if it can be accepted that one of the major epistemological aspects of a prototypical conception of categorial structure resides in the fact that categories are interpretive schemata that are used flexibly and dynamically in our encounters with reality, a major philosophical reference point for prototype theory will lie with those philosophical theories that recognize the constitutive role of existing knowledge with regard to new experiences. As I have argued elsewhere (1985b), the Husserlian phenomenological movement (as represented, specifically, by Maurice Merleau-Ponty) provides a good starting-point for a further confrontation with philosophy. The archaeopteryx is probably regarded as a species separate from either bird or reptile. The example is taken from the Longman Dictionary of Contemporary English. Considered from this point of view, Lakofrs radial sets as such are not particularly unclassical: structured polysemy as such is entirely compatible with the classical view. Kleiber (1988) offers an insightful discussion of the theoretical consequences of the growing importance of the structure of polysemy in prototype-theoretical research. For a more extended discussion of Wierzbicka's views, see Geeraerts (1988c). Notice that the restriction to the prototypical centre of categories correlates with Wierzbicka's Platonic, introspective methodology: it seems probable that the applications of a category that can be accessed introspectively are only the more salient ones; peripheral cases probably do not always pass the threshold of conscious attention. What is interesting from a cognitive point of view, however, is the way people spontaneously categorize and classify things, not the way in which they introspectively reflect upon their own conceptualizations. Any attempt to describe the peripheral instances of a category together with its prototypical centre can therefore not be restricted to an introspective methodology. This is not say that Putnam actually intended his stereotypical theory as such an attempted rescue: his problems lay with the notion of reference rather than with those of polysemy and categorial structure. My remarks about Putnam
26
Prototypicality and salience
are an investigation into some of the possible consequences of the notion of division of linguistic labor, not an attempt to give an account of Putnam's view in its original setting. Further, it has to be mentioned that some of Putnam's later philosophical views open up entirely different perspectives for a confrontation with Cognitive Semantics; in particular, see Lakoff (1987) on Putnam and anti-objectivism. 20. An interesting contribution to such an exploration is found in Lakoff (1987: Chapter 12), where it is claimed that scientific categories are far from being as classical as is usually assumed. 21. Again, see Lakoff (1987: Chapter 9) for some more examples; they are situated within formal psycholexicology rather than within linguistics. 22. Next, that is, to the remarks made in footnote 13.
Chapter 2 Where does prototypicality come from?
Originally published in Brygida Rudzka-Ostyn (ed.), 1988, Topics in Cognitive Linguistics 207-229. Amsterdam/Philadelphia: John Benjamins. The prototype-theoretical case study reprinted in the present chapter is not just included because it provides a concrete illustration of the theoretical model discussed in Chapter 1. The paper makes two points of a more general nature. First, on the theoretical level, the paper argues for a functional explanation of prototypicality effects, i.e. an explanation in tenns of general requirements on the organization of human cognition: It is cognitively advantageous to maximize the conceptual richness of each category through the incorporation of closely related nuances into a single concept because this makes the conceptual system more economic. Second, on the methodological level, the paper introduces an onomasiological perspective in lexical semantics: it is only by comparing the near-synonyms vernielen and vernietigen that we get a correct insight into the factors that might explain the presence of prototypicality effects. Over the years, I have become more and more convinced that such an onomasiological perspective is essential in Cognitive Linguistics: categorization primarily involve onomasiological choices, i.e. the question 'Why does someone use the category x rather than the category y for talking about phenomenon zT is cognitively more realistic than the question 'Does z belong in the semantic range of application of xT Starting with the monograph The Structure of Lexical Variation (Geeraerts, Grondelaers, and Bakema 1994), the lines that I have tried to set out with our Leuven research group have increasingly been geared towards the analysis of onomasiological variation (see Geeraerts 2002b, Grondelaers and Geeraerts 2003 for a general statement and an overview, and compare Chapter 11).
1.
Hypotheses about the sources of prototypicality
Prototype theory is as it were part of the prototypical core of the cognitive paradigm in semantics, particularly in lexical semantics. I think it is safe to say that it is by now quite obvious that gradience and salience are among the linguistically relevant aspects of semantic structure. One need only recall the early experimental work by Rosch (1973) and Labov (1973) to appreciate
28
Prototypicality and salience
the importance of gradualness and vagueness for the adequate description of word meaning. But what about explanatory adequacy? Can we move beyond the descriptive level and explain why prototypicality exists at all? There are at least four different hypotheses that have been proposed to explain prototypical phenomena. Each of these hypotheses has been formulated (or at least hinted at) by Eleanor Rosch herself; this is an indication that the hypotheses might well be complementary rather than mutually contradictory. I will call these four hypotheses the physiological, the referential, the statistical, and the psychological one. Let us have a look at them. The physiological hypothesis says that prototypicality is the result of the physiological structure of the perceptual apparatus (Rosch 1973). This hypothesis has been formulated with regard to the prototypicality effects in the domain of color terms (the first major field in which prototypicality phenomena have been observed). Particular colors are thought to be focal because the human eye is more sensitive to certain light frequencies than to others. The scope of the physiological explanation is probably fairly limited; it may only be applicable to concepts immediately referring to perceptual phenomena, or at least to bodily experiences that have a distinct physiological basis. Since this is most likely not the majority of cases, additional hypotheses will have to be invoked to explain the prototypical structure of concepts that have no immediate physiological basis. The referential hypothesis states that prototypicality results from the fact that some instances of a category share more attributes with other instances of the category than certain peripheral members of the category (or share attributes with more other instances than these peripheral cases). The peripheral applications of a category share attributes with relatively few other cases, or share only a relatively small number of attributes with other, more central members of the category. This is the family resemblance model of prototypicality (Rosch and Mervis 1975); in psychological terms, it states that the prototypical instances of a category maximize cue validity. I have dubbed this view 'referential' because it considers prototypicality to be an automatic consequence of the structure of the range of application of a concept. Once you know what objects, events etc. a concept can refer to, you can compute differences in salience by comparing the number of shared attributes among those things. One might even say that prototypicality is a secondary phenomenon: it is a side-effect of the mutual attribute relations among the instances in the referential range of application of the concept. Statistical explanations of prototypicality state that the most frequently experienced member of a category is the prototype. At least, this is the sim-
Where does prototypicality come from?
29
pIe fonn of the frequency model. It can also be combined with the family resemblance model; the weight of an attribute within a concept is then not only detennined by its role within the family of applications constituting the category, but also by the relative frequency with which it is experienced (Rosch 1975). The psychological hypothesis is a functional one. It states that it is cognitively advantageous to maximize the conceptual richness of each category through the incorporation of closely related nuances into a single concept because this makes the conceptual system more economic. Because of the maximal conceptual density of each category, the most infonnation can be provided with the least cognitive effort (Rosch 1977). In what follows, I would like to show that the functional explanation of prototypicality is more general than the other ones because it can explain cases of prototypicality that are counter-examples to the other modes. I will elaborate the psychological hypothesis by indicating some more functional sources of prototypicality; I will try to make clear that prototypicality is the outcome of some deep-seated principles of cognitive functioning.
2. A case study in synonymics Dutch has a pair of synonyms vernielen and vernietigen, which both roughly mean 'to destroy'. Though they exhibit some degree of phonetic similarity, their origin is quite diverse. Vernielen is the older fonn. It is already to be found in Middle Dutch, and it is fonned by means of the common verbfonning prefix ver- and the adjective niel, only a few examples of which survive, but which probably meant 'down to the ground'. Etymologically, then, vernielen means 'to throw down to the ground, to tear down'. Vernietigen, on the other hand, makes its first appearance in the sixteenth century; it is fonned by means of the same prefix ver- and the adjective nietig, which is itself a derivation from the negation particle niet (English not) and the suffix -ig (which corresponds with English -y). Vernietigen gradually replaces a third fonn verniefen, which is a straightforward derivation from niet with ver-, and which is extinct by the end of the seventeenth century. Vernietigen literally means 'to annihilate, to bring to naught'. The best way to study both words is to turn to the Woordenboek der Nederlandsche Taal (hence WNT), the major dictionary of Dutch that covers the period from 1500 up to 1920 and that, by the way, is still uncompleted after about a century of editorial work. This dictionary is being com-
30
Prototypicality and salience
piled on the basis of a huge corpus of quotations; there are as yet no equally representative corpora for contemporary Dutch, so that it is rather more difficult to get an adequate picture of twentieth-century usage than it is to study the semantic history of the vocabulary of Dutch. For the purpose of this paper, this is not very important: it suffices to pick out one synchronic period and to see how both words relate to each other in that period. For a number of practical reasons (among others the amount of available material), I will concentrate on the nineteenth century, stretching the temporal borders of that period with approximately one decade at each end. In this way, a period from 120 to 130 years can be considered, ranging from roughly 1790 to 1910. In the light of the history of culture, this seems quite justified; we more or less envisage the cultural period from the French revolution up to the First World War: the nineteenth century in the broadest sense. To get a good picture of the development ofvernietigen and vernielen, it would be necessary to present and discuss the entire articles that I have compiled for the WNT, and the complete set of quotations on which they arc based. For obvious lack of space, I will only give illustrative quotations; bibliographical references are as in the WNT. The set of quotations shows that vernielen and vernietigen can be used indiscriminately with the same range of application. Each numbered pair of quotations gives examples of one particular kind of usage. These examples should be studied from two points of view. On the one hand, the question has to be asked whether vernielen and vernietigen exhibit any syntagmatic differences, i.e. differences in their collocational properties. On the other hand, the question arises whether they are paradigmatically different, i.e. whether they exhibit purely conceptual differences. (More details on the analysis of both verbs can be found in Geeraerts 1985e.) (1)
Dat huis was ... evennmin als de naburige tegen de verwoestende veeten dier tijd bestand. Reeds onder den zoon en opvolger des stichters werd het ... tot den grand toe vernield (Veegens, Hist. Stud. 2, 282, 1869). [Like the neighboring one, this house was not able to stand up against the destructive quarrels of the age. Already under the son of the founder, it was demolished down to the ground.] Alleen zijn de vroegere kruisvensters door vensterramen van nieuweren trant vervangen en hebben de vrijheidsmannen van 1795 ... het wapen des stichters in den voorgevel met ruwe hand vernietigd (Veegens, Hist. Stud. 1, 125, 1864). [Only, the earlier cross-windows have been replaced by windows in a
Where does prototypicality come from? newer style, and in 1795, the freedom fighters demolished the founder's arms in the facade with their rough hands.] (2)
Er gaat dan stroom op den daarvoor gevormden zijweg over, waarbij genoeg warmte ontwikkeld wordt om de draadwindingen in zeer korten tijd te vernielen (Van Cappelle, Electr. 214, 1908). [Electric current is then transferred to the diversion that has been construed to that end, in which case enough warmth is produced to destroy the coils of the wire in a very short time.] Zonder deze voorzorg zou het draadje door de enorme hitte van den gloeidraad vernietigd worden (Van Cappelle, Electr. 295, 1908). [Without this precaution, the wire would be destroyed by the enormous heat of the filament.]
(3)
Hoeveel het wild vernielt wordt door een Engelschman zeer goed uiteengezet bij gelegenheid van een aanval op de beschenning die het wild aldaar geniet (Volksvlijt 1872, 175). [How much is destroyed by game is aptly expressed by a certain Englishman on the occasion of an attack on the protection these animals enjoy in his country.] Bij het vernietigen van de onkruiden door het bewerken dient op hunne voortplanting en ontwikkeling te worden gelet (Reinders, Land. I, 309, 1892). [During the destruction of weeds by cultivating the land, one should bear in mind their reproduction and development.]
(4)
Wel wat hamer! Wordt door zulke sentimentele zotternyen niet al de inwendige kracht vernield? (Wolff en Deken, Blank. 3, 220, 1789). [By golly! Does not such sentiment al foolishness destroy all our inner strength?] Stel mij niet zoo hoog, zei ze onthutst, ik zou daaraan niet beantwoorden; ik zou uw ideaal vernietigen (Vosmaer, Amaz. 175, 1880). [Do not put me on a pedestal, she said disconcertedly, I would not live up to that; I would destroy your ideal.]
(5)
De bergstroom in zijn grammen loop Verscheurt zijn zoom, verdrinkt de dalen: Alzoo vernielt Gij 's menschen hoop! (Ten Kate, Job 53, 1865). [The mountain stream in its angry course rends its banks, drowns the valleys: thus, Thou destroys the hope of Man!] Zy is dan, van kindsbeen af, opgevoed om mynheer Daniel's echtgenote te worden, en nu is die hoop van een geheelleven vernietigd! (Conscience, Kwael d. T. 2, 65, 1859) [She has been raised from childhood to become Master Daniel's wife,
31
32
Prototypicality and salience and now this hope of a lifetime has been annihilated!] Dit toeval vernietigde ons geheele plan (Haafner, Ceilon 103, 1810). [This coincidence annihilated our entire plan.]
(6)
Mij gedenkt ook nog dat Nicolaas Gaal ... mij placht te verhalen ... dat de oude man om deze ontstolen eer zich zoo ontstelde en vergramde, dat het ook scheen ofhij dezen diefwel had willen vernielen (Fruin, Geschr. 1,1974, 1888). [I still remember that Nicolas Gaal was fond oftelling me that the old man used to get so angry and upset about this stolen honor, that it seemed that he would have liked to kill that thief.] Intusschen heeft de Godin de Natuur besloten nu voor altijd de Drijvende Eilanden en al hun inwoners te vernietigen (Quack, Soc. 1,246, 1875). [Meanwhile, the goddess Nature has decided to destroy the inhabitants of the Floating Islands once and for all.]
(7)
De beroerte, die haar zwakke levenskrachten in een half uur tijds vernielde, had reeds in het eerste oogenblik hare spraak verlamd (Beets, C.O. 206, 1840). [The stroke that destroyed her weak life force in half an hour. had from the first moment paralyzed her speech.] Hy moet rusten. Zulke driften vernietigen het sterkste gestel (Wolff en Deken, Leev. 1,290, 1784). [He has to take a rest. Such passions undennine the strongest constitution.]
(8)
De vrouwen, Lus, zijn zonen, al de anderen bleven stom, vernield van ontsteltenis, op hun stoelen genageld (Buysse, NeefPerseyn 4545, 1893). [The women, Lus, his sons, all the others remained silent, destroyed by disconcertedness, nailed to their chairs.] Toen antwoordde zij langzaam met een doffe stem, als vernietigd door haar eigene woorden: 'la, indien het nog mogelijk is' (Buysse, Mea Culpa 68, 1896). [Then she answered slowly, in a dull voice, as if struck down by her own words: 'Yes, if it is still possible'.]
(9)
De uitslag van den stryd was ditmael hem niet gunstig: geheel zyn leger werd vernield of uiteen geslagen (Conscience, Gesch. v. Belgie 110, 1845). [This time, the result of the battle was not favorable to him: his entire army was destroyed or dispersed.]
Where does prototypicality comefrom?
33
Het gansch leger der Turken was vernietigd! (Conscience, Gesch. v. Belgie 352, 1845). [The entire army of the Turks was destroyed! 1 Syntagmatically, we not only see that both words can be used by the same author in the same context without noticeable differences, as in (1), (2), (8) and (9), but also that the range of application of each word can be divided into three identical major groups, which can moreover be subdivided along parallel lines. There is a set of applications in which the words are used with regard to concrete, material objects (1, 2, 3); a set in which they are used with regard to abstract objects (4, 5), and a set in which they are used with regard to persons (6, 7, 8, 9). Within the first set, frequently occurring applications relate to buildings (1), other human artifacts (2), and natural objects, in particular plants and crops (3). With regard to the abstract applications, we can distinguish between the annihilation of the existence of certain abstract objects as such (4), and applications in which the realization or fulfillment of certain abstract notions that contain an aspect of expectation or intention with regard to the future is prevented (5). With regard to persons, (6) expresses their death as such; (7) and (8) indicate how someone's bodily or mental health, respectively, are undermined. (9) expresses how armies are beaten; this application is half-way between the abstract group (the armies cease to exist as functional entities), and the personal group (individual soldiers are killed). The existence of analogous subdivisions within each of the major groups shows that the syntagmatic equivalence of vernielen and vernietigen is not a coincidence, but that it is an essential part of their relationship. Furthermore, the examples also show that there is a paradigmatic, strictly conceptual equivalence between both: they do not only have the same collocational properties, but they also seem to express the same concepts in the same contexts. (The distinction between syntagmatic and paradigmatic meaning is used here for purposes of analysis only; it does not imply any particular view with regard to the theoretical relation between both aspects of lexical meaning and particularly with regard to the question whether selectional restrictions are always an automatic consequence of a concept's paradigmatic characteristics.) As a preliminary step, notice that the concept 'to destroy' does not only appear as the notion 'to annihilate the existence of someone or something, to cause someone or something to disappear out of existence', but that it also exhibits the weaker nuance 'to undermine someone or something with regard
34
Prototypicality and salience
to some aspect of his existence' (without a complete destruction or a complete removal out of existence being implied). The distinction can easily be discovered within the personal group of applications. In (6), a person is killed, taken out of existence, while in (8) (and most likely also in the second quotation from (7), someone's existence is undermined from one point of view or another, but not entirely annihilated. Likewise, we can see that within the abstract group, (4) signifies the suppression of the existence of some abstract things as such, whereas in (5), plans, hopes, and expectations are undermined with regard to their realization and fulfillment: the plan as such is not removed (at least not to begin with), but it is reduced to ineffectiveness and futility. In short, both vernielen and vernietigen express the notions of complete destruction and partial damage, that is to say, the complete removal out of existence of something or someone, and the less drastic undermining in some respect, of the existence of people or objects. According to the syntagmatic context, these notions receive further specifications. For instance, with regard to persons, complete destruction means killing, but with regard to concrete things, destruction signifies material demolition, and so on. (For the sake of completeness it should be added that the equivalence of vernielen and vernietigen is less straightforward in present-day Dutch than it is in nineteenth-century Dutch. Some of the quotations discussed here are now felt to be rather awkward; in particular, it would be difficult to use vernielen with regard to persons. ) On the basis of the foregoing observations, one might be tempted to conclude that the semantic structure of vernielen and vernietigen in nineteenthcentury Dutch is completely identical: both syntagmatically and paradigmatically, they have the same range of application. However, a number of facts testify that both words have different prototypical structures, i.e., that they have different conceptual centers. There are two sets of facts to be considered: corpus-based facts relating to the way in which both words are used in our corpus of quotations, and introspective facts relating to the way in which the words are perceived by the speakers of the language. In general, consideration of these facts will lead to the conclusion that the abstract applications are central within the structure of vernietigen, and that the material applications are central in the cases ofvernielen. As such, each verb has a different semantic structure in spite of the fact that the elements of these structures appear to be the same.
Where does prototypicality comefrom?
35
3. Usage as evidence for prototypicality Five observations support the prototypical hypothesis. In the first place, the abstract group of applications is quantitatively more prominent within the structure of vernietigen than the material set of applications, while the reverse is true of vernielen, in which the material group is the most frequently occurring one. In both cases, the major group is represented by approximately three times the quotations of the less central group. In the second place, the differences in centrality show up in the fact that the prominent applications exhibit specifications and particular nuances that they do not have when they are peripheral within the structure of the lexical item. Thus, the material group of vernielen contains a metonymical extension of the application with regard to plants and crops, towards an application in which the fields and gardens where these plants and crops grow appear as the direct object of the verb. Likewise, the application with regard to buildings receives a figurative extension towards an application with regard to an allegorical 'wall' that separates two people. These extensions are probably not impossible within the concrete set of applications of vernietigen, but the fact that they do not appear there is statistically interesting: it indicates that the concrete application is more productive in the case of vernielen than in the case ofvernietigen. Conversely, the abstract group has nuances and additional specifications in the case ofvernietigen that are lacking in the same group with vernielen, although it is quite easy to imagine that they would in fact occur there. For example, vernietigen has a fairly large set of applications in which social movements, institutions, activities and so on are abolished, one quotation in which it is said that railway transport destroys distances (obviously, distances do not disappear as such, they are only functionally overcome), and one quotation in which a philosopher is said to destroy the soul (again, the soul is not destroyed as such, but the idea that the soul exists is metonymically abolished by the philosopher in question). None of these extensions of the abstract use of the concept 'to destroy' can be found in the case of vernielen, which is indicative of the fact that the abstract use is less prominent in the latter verb than in the semantic structure of vernietigen. In the third place, the salience of the material kind of usage can be derived indirectly from the nominalizations of both verbs. Both vernietiging and vernieling have the verbal sense 'the fact, the act or the process of destroying or being destroyed', but only vernieling exhibits the metonymical extension towards the concept expressing the result of that process or that
36
Prototypicali~yand salience
act, i.e., the concrete damage that issues from it. (In the latter case, the word is typically used in the plural: vernielingen more or less equals the notion 'damage'.) In the fourth place, the internal structure of the set of personal applications reflects the differences in prototypical structure between both verbs. To begin with, notice that the personal group contains concrete as well as abstract applications; to kill someone is clearly more concrete than to undermine someone's psychological well-being or his social position. If we then have a look at the mutual relationship between the abstract and the concrete subgroups of the application with regard to persons, we find that the abstract subgroup is proportionally dominant in the case of vernietigen, whereas the reverse is true in the case of vernielen. Also, we find that extensions of the concrete subgroup of the personal application with regard to other living beings than people or with regard to personifications, are not as strongly present in the case of vernietigen than in the case of vernielen. (It should be added that these observations have to be considered with more care than the previously mentioned points, since there is a general tendency throughout the centuries covered by the WNT-material, to re move the personal application from the structure of vernielen. There are relatively less personal applications in the structure of nineteenth-century vernielen than in the structure of either nineteenth-century vernietigen or sixteenth-century vernielen; as has al ready been mentioned, it is even more difficult to use vernielen with regard to persons in present-day Dutch. In any case, the nineteenth-century material does seem to show that the material subgroup of the personal application of vernielen is more resistant to the tendency in question than the abstract subgroup, as can be predicted from our centrality hypothesis.) Finally, the importance of prototypicality can be derived from the fact that different nuances play a central role within the core of each concept, whereas those nuances are not particularly important within the corresponding group in the other concept. Thus, the destruction of buildings and other human constructions is prominent within the material use ofvernielen, but is only rarely present within the material group of vernietigen. Within the structure vernielen as a whole, demolishing buildings is the single most frequently represented kind of usage, but within the structure of vernietigen, it is merely one among many equally important nuances of the material set of applications. In the same way, the central, abstract group within the structure of vernietigen is itself centered round applications relating to the dissolution, the cancellation, the annulment of agreements, commitments, engagements, obligations, permissions, rights, and so on, and of the laws,
Where does prototypicality comefrom?
37
orders, contracts etc. in which they are contained and through which they come into existence. Whereas vernielen only rarely exhibits this kind of usage, it is the most frequently occurring sense within the abstract group of vernietigen as well as within that word as a whole. In general, these facts of linguistic usage clearly favor the hypothesis that the abstract applications of the concept 'to destroy' are prototypical within the structure of vernietigen, whereas the concrete applications are prominent in the case of vernielen. Taking into account that each central group is itself concentrated round a dominant kind of usage, it seems plausible to say that the latter is the prototypical sense for each of the verbs in question. It should furthermore be noted that these prototypical phenomena seem to be connected with the etymology of the words. On the one hand, the abstract prototype ofvernietigen may well be connected with the abstract character of the words niet 'not', and nietig 'null and void, insignificant', on which it is based. Moreover, the common phrase nietig verklaren 'to declare something to be null and void, dissolve, annul something' corresponds pretty closely with the central notion within the abstract group ofvernietigen. On the other hand, the centrality of the application with regard to buildings in the structure ofvernielen seems to correspond with the etymological meaning 'to tear down, to throw to the ground' that we reconstructed above as the original meaning of the verb.
4. Introspective evidence of prototypicality Before we can deal with the introspective evidence in favor of the prototypicality hypothesis, two preliminary questions have to be answered. In the first place, how trustworthy is the introspective methodology? The paradoxical fact of the matter is that it is exactly the unreliability of introspection that makes it interesting for our purposes. If introspection were able to yield a completely adequate picture of the facts of linguistic usage (which is doubtful), it would simply reduplicate the results reached in the previous paragraphs on the basis of a direct examination of linguistic usage. But given the presupposition that introspection yields only a partial insight into the semantic structure of the words that are investigated, we can also presuppose that it will be exactly the prototypical kinds of usage of those words, that reach the introspective consciousness of the language user. We can use the results of the introspective method as support for the prototypical hypothesis if we presuppose that prototypical kinds of usage (precisely because they are more
38
Prototypicality and salience
salient than other applications) will more easily pass the threshold of conscious attention. Given this presupposition, the introspective judgments of native speakers may shed light on the question which kinds of usage are predominant within a certain concept. In the second place, how can the introspective method be used with regard to historical material? There are no nineteenth-century speakers of Dutch around to be asked what they think is the meaning of particular words, so how are we going to get introspective judgments at all? The fact is that we do have information on how the nineteenth-century speakers of Dutch perceived the near-synonyms that we are investigating, viz. in the form of synonym dictionaries. Synonym dictionaries (at least the older ones) are notoriously unreliable as descriptions of actual patterns of usage; most of the time, the compilers of synonym dictionaries rationalize away the actual identity of words by imposing distinctions that cannot be discovered in the actual facts of usage. However, these rationalizations need not always have proceeded out of the blue: it seems quite plausible that they were guided by the introspective judgments of the compilers. So, if we like to know something of the introspective insights of the nineteenth-century speakers of Dutch, we can have a look at the synonym dictionaries of that time to see whether the distinctions they make between vernielen and vernietigen (however inadequate as a picture of the complete set of possible kinds of usage) do indeed reflect the differences in prototypical structure of both words. And indeed, the nineteenth-century synonym dictionaries of Dutch do distinguish between vernielen and vernietigen along lines that fit into our hypothesis. On the one hand, there are those that draw the line syntagmatically, such as Weiland and Landre (1825), who state that vernielen can only be used with regard to 'lighamelijke dingen' (material things), whereas vernietigen is more widely used, in particular also with regard to 'menschelijke instellingen' (human institutions). De Beer (1897) expresses an analogous point of view. On the other hand, there are those that describe the distinction along paradigmatic lines, so that there would be an actual notional difference between the verbs in question, rather than merely a distinction in selectional restrictions. Whereas vernietigen is defined as 'to bring to naught, to annihilate' , vernielen is defined as 'to damage, to smash to pieces, to tear down'. In this sense, vernietigen implies a complete annihilation whereas there may be some pieces left of the original object in the case ofvernielen. It is easy to see that this paradigmatic point of view, which can be found among others in Pluim (1894), is connected with the previous, syn-
Where does prototypicality come from?
39
tagmatic one: it is precisely because vernielen relates to material things that the notion of remaining debris comes to the fore. Likewise, a complete annihilation (in which the original objects disappear completely) is less likely in the material world of concrete objects, so that the restriction of vernietigen to abstract objects will tend to be related to the notion of complete annihilation. This is in fact done by Weiland and Landre (1825), though not all proponents of the paradigmatic distinction adhere to the syntagmatic distinction. For instance, De Flines (1810) mentions that verniefigen can in fact be used with regard to material objects, but that there is a difference with vernielen in the degree of damage achieved. By and large, these views faithfully reflect the insight into the prototypes of vernielen and vernietigen that we have gained by considering the actual facts of linguistic usage. Syntagmatically, it is recognized that the material context is more important for vernielen, whereas abstract objects are predominant in the case ofvernietigen. Paradigmatically, this is reflected by the fact that vernielen carries overtones of material destruction and damage (think of the relationship between the prototypical usage of vernielen with regard to buildings, and the definitions of that word that bring to the fore the act of smashing and demolishing things), whereas verniefigen calls forth the idea of complete annihilation (as it were, wiping something off the face of the earth). As such, the stubborn efforts of the compilers of synonym dictionaries to find semantic differences among near synonyms seem to be not entirely gratuitous. To the extent that they try to capture the characteristics of the most salient kinds of usage of both lexical concepts, they strengthen our own hypothesis about the differences in prototypical structure among the verbs.
5. The functional explanation of prototypicality There are a number of interesting conclusions to be derived from the above analysis of the near-synonyms vernielen and vernietigen. First, prototypicality is an interesting new point of view in the study of synonyms. It is traditionally well-known in lexical semantics that there are relatively few true synonyms in natural languages, and the ways in which near-synonyms differ can be very diverse. Our discussion of vernielen and vernietigen shows that there is one more factor to be added to the list of differentiating factors: near-synonyms may be distinct with regard to the prototypical structure imposed on an otherwise identical range of application. Once again, the
40
Prototypicality and salience
importance of prototype theory for the traditional concerns of lexical semantics becomes apparent (cp. Geeraerts 1983a, 1983c, 1984a, 1985a); prototype theory opens up new perspectives in the study of synonyms. Secondly, there are some indications that introspective judgments in lexical semantics relate to the prototypically salient instances of concepts rather than to the full range of actual usage possibilities. If this can be confirmed by additional comparisons between introspective perceptions of lexical meanings and actual usage patterns, more will be known about the value of both methodologies (introspective and corpus-based) in lexical semantics. Also, if we maintain the classical view of modem linguistics that it is one of the goals of linguistic theory to account for the introspective judgments of native speakers, and if these judgments appear to be influenced by prototypical phenomena, yet one more reason presents itself for incorporating prototype theory into lexical semantics. Thirdly, the fact that vernielen and vernietigen have the same conceptual and collocational range of application, and yet differ with regard to the core and the periphery of their categorial structure, indicates that there are at least some cases of prototypicality that cannot be explained by means of the referential model. Vernielen and vernietigen refer to the same set of acts and processes; as such, the differences in their prototypical structure cannot be the automatic consequence of their referential range, as is implied by the family resemblance hypothesis. In addition, the physiological and the statistical explanation will not be of much avail either. There is no particular organ or mechanism for the perception of processes of destruction, and even if there were, we would still need two different physiological structures to explain the distinction between both verbs, which is beyond all intuitive plausibility. The statistical explanation is inapplicable for the same reason as the referential hypothesis: since the range of application of both verbs is the same, the frequency of occurrence of the processes referred to is the same for both verbs. That is to say, the frequency with which the demolishing of buildings occurs in reality, relative to the frequency with which, say, agreements are cancelled in reality, has exactly the same effect on both verbs, since these refer to the same objective reality. Because they denote the same things in reality, the structure of reality (either with regard to the frequency of occurrence of its elements, or with regard to the mutual resemblances among those elements) cannot be invoked to explain the distinction in semantic structure between vernielen and vernietigen. In short, we can reject all materialistic explanations of the prototypicality effects observed in the verbs under consideration. Indeed, the physiologi-
Where does prototypicality come from.?
41
cal, the referential, and the statistical hypotheses have this in common: that they try to explain prototypicality on the basis of materialistic data, either the material structure of the human perceptual apparatus, or the material characteristics (statistical or otherwise) of the referential range of the concepts involved. Given that we have to reject these materialistic hypotheses, we can provisionally choose, by elimination, for the psychological, functional explanation of prototypicality. To support this choice, I would like to make clear that the functional hypothesis has some additional advantages, besides the fact that it avoids the problem of the materialistic hypotheses. First, however, three remarks have to be made. To begin with, it might be claimed that a statistical explanation of the prototypicality effects in vernielen and vernietigen can indeed be given, if we take into account, e.g., that the material sense occurs much more frequently with vernielen than with vernietigen, or that the abstract specification of the notion 'to destroy' is statistically much more prominent in the latter verb than in the fornler. However, the frequencies that are mentioned here are linguistic frequencies, not referential frequencies, i.e., they are frequencies of occurrence of words, not of the things those words refer to. Because the frequency at stake here is linguistic rather than referential, it can hardly be invoked to explain prototypicality; as an aspect of linguistic usage, it is one of the things we have to explain, not one of the things that are themselves part of the explanation. We can use linguistic frequencies to determine what instances of a concept are prototypical (that is what we did in section 3), but explaining prototypicality on the basis of linguistic frequency is putting the cart before the horse. Some kinds of usage are not prototypical because they are more frequent; they are more frequent because they are prototypical. The apple is not a prototypical fruit because we talk more about apples than about mangoes, but because we experience apples more often than we encounter mangoes (and this fact, in turn, may be the reason why we talk more about apples). Frequency of linguistic occurrence may be a heuristic tool in the pinpointing of prototypes, but it is not the source of prototypicality as meant in the statistical hypothesis. The second remark has to do with the fact that criticism with regard to the referential, family resemblance model of prototypicality has already been formulated elsewhere. This has been the case in the work of Pulman (1983) and - in more stringent fashion - in the well-known article by Armstrong, Gleitman, and Gleitman (1983). They argue that gradience can be observed in concepts with rigid boundaries (their examples relate to natural numbers), so that family resemblances cannot be invoked to explain the differences in
42
Proto~vpicalityand
salience
salience among numbers. There are two reasons, however, why their argumentation is less relevant than they assume. First of all, they more or less equate prototype theory and the family resemblance model of the sources of prototypicality, whereas it is quite clear that the family resemblance model is merely one of a number of hypotheses concerning the sources of prototypicality: ruling out one hypothesis does not mean that one can ignore the others. And also, I do not think they are successful in presenting a counterexample to the family resemblance model. Even if a concept has rigidly defining characteristics, family resemblances may exist among the nondefining characteristics of the instances of that category. Since Cognitive Semantics is basically encyclopedist in its approach, these non-defining, 'encyclopedic' attributes should be incorporated into the computation of degrees of shared attributes. As Lakoff (1982) has shown, such encyclopedic, experiential factors do indeed occur with regard to numbers, and they can be used to explain the prototypicality ratings found experimentally. My third remark is this: my criticism of the materialistic hypotheses should not be overgeneralized. The fact that they do not work in the case of vernielen and vernietigen clearly does not imply that they do not work in any case, but merely makes clear that next to the physiological, the referential, and the statistical model, there will have to be at least one other source of prototypicality. Let us now come back to the functional model of prototypicality and try to elaborate it. Remember that the psychological hypothesis involves requirements that the cognitive system is to comply with if it is to function efficiently: prototypicality exists because it is cognitively advantageous. As we have seen, Rosch has specified this functional advantage in terms of the economical effect of informational density~ prototypical categories enable one to reach the most information with the least cognitive effort. This functional line of reasoning can be supplemented with some additional (and perhaps even more fundamental) functional reasons for having prototypical categories. We can base the discussion on one of the fundamental insights of cognitive psychology, viz. that cognition should combine structural stability with flexible adaptability. On the one hand, cognition should have a tendency towards structural stability: the categorial system can only work efficiently if it can maintain its overall organization for some time, if it does not change fundamentally any time new information has to be incorporated. At the same time, however, it should be flexible enough to be easily adaptable to changing circumstances. To prevent it from becoming chaotic, it should have a built-in tendency towards structural stability, but this stability should not
Where
doesproto~ypicali~y
come from?
43
become rigidity, lest the system stops being able to adapt itself to new and unforeseen circumstances. This necessity of flexibility is one of the aspects of lexical semantics that was recognized by the prestructuralist tradition of historical semantics, but that has been more or less lost in the meantime, as a result of the structuralist attention for fixed synchronic structures. Be that as it may, it will be clear that prototypically organized categories are particularly well suited to fulfill the double demand for flexible adaptability and structural stability. On the one hand, the fact that slightly deviant nuances can be developed within a particular category indicates that categories have the dynamic ability to cope with changing conditions and changing expressive needs. On the other hand, the same fact (that marginally deviant concepts can be incorporated into existing categories as peripheral instances of the latter) proves that these categories have a tendency to maintain themselves as holistic entities, thus maintaining the overall structure of the categorial system. Prototypical categories maintain themselves by adapting themselves to changing circumstances and new expressive needs; at the same time, they function as expectational patterns with regard to reality: new facts are interpreted in terms of information that is already at the disposal of the individual. The flexibility of the cognitive system does not only show up in the fact that it ran adapt itself to new experiences, but this flexibility is supplemented with the fact that existing categories have a formative influence with regard to experience; new experiences are fitted into the expectational patterns provided by the existing categorial system. Along these lines, prototypicality appears to be the outcome of some fundamental, deep-seated principles of cognitive functioning. The form of the conceptual system appears to be determined by a set of basic functional requirements, and prototypically structured concepts admirably meet these requirements. If this is correct, the same basic principles should also have a role to play in other cognitive disciplines. That is to say, if prototypicality is an emanation of some basic characteristics of all cognition, we should be able to find analogies of the prototypical idea in other fields of cognitive science, next to lexical semantics. I have tried to prove at length elsewhere (1985a) that this is in fact the case: the importance of interpretative schemata mediating between experience and existing knowledge is an idea that can be traced in a number of cognitive disciplines. It is very much apparent in Artificial Intelligence (Minsky's frame notion); it can be found in cognitive psychology, particularly in the work of Bruner, and to some extent in that of Piaget; it can be related to some of the views of the early, Husserlian
44
Prototypicality and salience
phenomenological movement in philosophical epistemology; and it has some important similarities with the paradigmatic conception of scientific enquiry inaugurated by Thomas Kuhn. These are exciting parallels because they suggest that the functional, psychological hypothesis concerning the sources of prototypicality can at the same time be the basis for a truly integrated cognitive science in which the insights of linguistics, Artificial Intelligence, cognitive psychology, philosophical epistemology, and the philosophy of science can be brought together under a common denominator. In this respect, the functional model of prototypicality, even if it does not rule out the possible importance of the physiological, the referential, or the statistical explanation, does seem to be more general than the latter, not just because it is based on fundamental principles of cognition, but also because similar views have been put forward in other branches of cognitive science.
6.
Onomasiological and semasiological aspects of Cognitive Semantics
Unfortunately, the optimistic perspective of the previous paragraph does not solve everything. To round oft the discussion, I would like to show that a complete explanation of all questions to be raised with regard to vernielen and vernietigen is far from available. The picture we have reconstructed so far looks like this: apparently, the linguistic conmmnity at some point in its development finds it convenient to have two distinct categories for the concepts of material and abstract destruction. Thus, a pair of etymologically distinct words becomes available, originally vernielen and vernieten, later on vernielen and vernietigen; their compound character ensures that one of them signifies material destruction, the other abstract annihilation. Gradually, the flexibility that is inherent in all human categorization extends these concepts beyond their etymological usage; as a result, they have the same range of application in the nineteenth century. But now consider the original situation in which these flexible extensions have hardly begun taking place. Is it then not irrational to use vernielen to express abstract annihilation, when you al ready have vernietigen or vernielen to do so? The question can be put in terms of global and local efficiency. As we have argued, the global efficiency of the conceptual system commands its flexible, prototypical organization. But there is, in the case of vernielen and vernietigen, also a local efficiency principle that says that it is uneconomic to have two terms expressing the same things. We are then
Where does proto(vpicality come from?
45
forced to ask: why does not the local efficiency principle stop the application of the global principle? Why is not the prototypical extension of vernielen towards abstract forms of destruction checked or prevented by the consideration that you already have a lexical category expressing abstract destruction? There is yet another way of formulating the problem: prototype-based flexibility is necessary because of the expressive needs of the speaker: he may want to express concepts for which no specific term is available. But why then would he use these flexible mechanisms of semantic extension if such a specific term is in deed available? One kind of answer might simply be that the global principle is stronger than the local principle; the global principle simply supersedes the local principle to the extent that local inefficiencies are created. We are then saying that the global principle is so general that its strength overrules the local principle, and that it applies even where it is not strictly necessary. Still, this does not tell us why the local principle is weaker than the global principle. Also, it is rather awkward to explain a mechanism that is unfunctionally overproductive on the basis of functional considerations. Couldn't we therefore find a more rational explanation of the flexible extensions? The way out, as far as I can see, is to take into account other kinds of expressivity than the purely conceptual one. Using vernielen to express a concept that is commonly expressed by vernietigen may be conceptually superfluous, but that does not mean that doing so may not serve particular expressive purposes. On the level of the linguistic form, for example, it may be quite functional to use another word than the usual one. The varieties of such a formally expressive synonymy are well-known in traditional lexical semantics; near-synonynls may exhibit connotational and emotional differences (as in euphemisms), stylistic differences (as in popular words versus poetic terms), or sociolinguistic differences (as in learned words versus common words). Perhaps we can even say that speakers have an urge for stylistic variation as such, even if the fornlal variants do not carry specific overtones; variation may well be governed by a straightforward desire to avoid monotony, to create new ways of expressing oneself, to experiment with unexpected innovations as a way of stressing one's own individuality. Moreover , it may well be that the importance of metaphor in natural language is determined precisely by its stylistic expressivity; metaphorical expressions would then be created primarily to add expressive weight to the message one wants to convey. (See Rudzka-Ostyn 1988.) It is quite plausible, then, that factors such as these have governed the extension of vernielen and vernietigen beyond their original meanings and into
46
Proto~ypicali~yand salience
each other's etymological range of application. For instance, using vernielen to express a process of abstract cancellation may have been stylistically particularly expressive, because the process of material destruction normally denoted by vernielen carried overtones of physical violence that were less marked in the case of vernietigen. The extended use of vernielen would then have been a case of metaphorical hyperbole. It is, however, very difficult to pinpoint exactly which form of expressivity is the relevant one with regard to the two verbs that we are concerned with here; our historical material for the earliest (Middle Dutch) history of vernielen and vernieten, for instance, is very hard to interpret with regard to such questions. Still, some clear cases may in fact be found. For instance, the first quotation of (5) clearly carries more overtones of violence, force, and intensity than the second quotation in that pair of examples. (This is mainly made apparent by the presence of a simile, marked by alzoo.) This suggests that the verbs highlight slightly different aspects of the situation described, or rather , represent the situation from different points of view (determined by the prototypical core of each verb). (In Langacker' s terminology, the distinction between the two verbs, when used with regard to the same process, might then be characterized as a figure/ground-distinction: vernielen takes the violent process as figure, and vernietigen the destructive result.) This is not an altogether implausible hypothesis, but it is unfortunately hard to confirm for the simple reason that the historical texts used here do not give us enough clues to discern such subtle differences in stylistic or emotional overtones. On the whole, then, what can we conclude from our discussion of this additional problem? On the one hand, it inspires caution with regard to our attempts to explain prototypical phenomena: the linguistic materials at our disposal do not always allow completely satisfactory answers with regard to the questions at stake to be formulated. On the other hand (and this is, I think, the more important conclusion), the discussion suggests that prototype formation may be influenced by other factors than purely conceptual ones. Stretching the meaning of a lexical item may be motivated by the desire to use another form than the one that is usual to express the idea in question; stylistic, sociolinguistic, connotational expressivity rather than purely conceptual needs may determine the flexible use of a category. In such a case, the conceptual coherence of the prototypically structured category (i.e., the fact that the new, peripheral kinds of usage have to be accessible from the prototypical core) constitutes a limit to the desire for formal variation: you can use a particular lexical item to express an idea that is usually signified by another word, but only on the condition that the idea in question is part of
Where does proto~vpicali~v come from?
47
the prototypical potentialities of that lexical item. Basically, you stretch an item's meaning to express something conceptually new, but you can also stretch it to express something conceptually old in a formally new way. This is a very important suggestion, because it implies a warning against a tendency that is a natural characteristic of Cognitive Semantics: the tendency, in fact, to look for purely cognitive or conceptual explanations of the facts one encounters. Taking the cognitive, experiential, encyclopedic nature of linguistic signs seriously should not imply looking only for strictly conceptual explanations. Language is not just content: it is also fornl, and its fonllal side has an expressivity of its own, which does seem to create lexical configurations that can hardly be explained if we only take into account the conceptual expressivity of language. In the traditional terms of lexical semantics, this means that the explanation of prototypicality should not restrict itself to the semasiological perspective (in which each category is considered on its own), but that the onomasiological point of view (in which it is studied how several items may express similar or identical concepts) should be taken into account as well. (conceptual expressivity is basically a factor connected with the semasiological explanation of prototypicality, whereas the onomasiological influences on prototype formation seem to refer to other kinds of expressivity, as was suggested by our study of vernielen and vernietigen. The incorporation of the onomasiological approach does not mean that Cognitive Semantics moves away from the functional perspective advocated in the previous section; non-conceptual expressivity is just as much a functional principle as purely conceptual expressivity and cognitive efficiency. Rather, the incorporation of onomasiology implies that Cognitive Semantics moves much closer to the rich tradition of lexical semantics, in which onomasiological mechanisms and configurations have been thoroughly studied (see Geeraerts 1986a: Chapter 1). Such a link with traditional approaches can only strengthen the linguistic attractiveness of Cognitive Semantics. To summarize: I have tried to argue, on the basis of a case study involving the Dutch near-synonyms vernielen and vernietigen, that the functional point of view is the most encompassing, most promising one for studying prototype formation, though it should not be restricted to purely conceptual expressivity and efficiency, but should also take into account the kinds of functional mechanisms that have traditionally been studied by the onomasiological approach to lexical semantics. The fact, however, that prototypicality may come from a number of diverse sources, also implies that an adequate explanation of conceptual structures will not be easy.
Chapter 3 The semantic structure of Dutch over
Originally published in Leuvense Bijdragen. Leuven Contributions in Linguistics and Philology 1992, 81: 205-230. The preposition over plays a role in Cognitive Semantics that is somewhat comparable to that of bachelor in Katzian semantics: from Brugman (1981, 1988) over Vandeloise (1990), Cuyckens (1991), Deane (1992), and Dewell (1994), to Evans and Tyler (2001), Tyler and Evans (2003), it has been a rallying-point for comparing competing forms of semantic analysis. The present chapter reproduces my own go at the subject. Although 1 would now try to firmly base the analysis on corpus materials rather than on introspectively derived sample sentences, I believe that the basic point of the paper still holds: the semantic structure of lexical items like over has to be seen as a multidimensional structure, in which covariation of semantic shifts along different dimensions constitutes the essential backbone of a prototype-based network. Specifically, the semantic structure of over involves at least two dimensions: on the one hand, the spatially relational dimension in the strict sense (with regard to which three distinct spatial configurations have to be distinguished), and on the other, the motional dimension, with regard to which we have to distinguish between cases of real motion, fictive motion, and zero motion. A third dimension to be added involves the existence of actual contact between the prepositional object and the other entity involved in the spatial relation. This type of analysis contrasts with the more common type of radial set analysis in Cognitive Linguistics, which is much more atomistic: a classical radial set analysis focuses on individual readings and the links that connect pairs of such readings, somewhat neglecting the underlying structure of dimensions that link various readings among each other. Such a focus on the underlying multidimensionality may help to avoid the pitfalls of the polysemy/univocality debate. There are (as argued in Chapter 5) convincing methodological reasons for being careful with the question: the criteria for answering it are neither stable nor consistent. But a practical, descriptive exercise such as the analysis of over may inspire the same conclusion. If you think of semantic structure primarily in terms of individual readings, it is natural to ask the question 'How many meanings does a word have?' (cp. Taylor 1992), i.e. is this a monosemous or a polysemous, or a hyperpolysemous word? If, however, one recognizes that covariation of different dimensions is a crucial characteristic of the structure in question, the individual readings (and the number they have) become less important from a practical and theoretical point of view.
The semantic structure ofDutch over
49
1. Scope and purpose ofthe exercise
The English spatial preposition over occupies an important position within Cognitive Semantics. On the one hand, Brugman's analysis (1981), which also appears with slight modifications in Lakoff (1987), is one of the earliest examples of a radial set analysis of lexical semantic structures. On the other hand, Vandeloise's reaction against the Brugman/Lakoff-analysis (1990) raises a number of fundamental questions concerning the nature of polysemy and the description of polysemy within a Cognitive framework. In the following pages, a similar methodological point concerning lexical-semantic research in Cognitive Semantics will be dealt with on the basis of an analysis of over. However, by presenting an alternative for the treatment in Cuyckens (1991), the paper will focus on the Dutch item over rather than on its English counterpart. Also, the theoretical focus will not be on the possible univocality of over (as discussed by Vandeloise), but on the importance of a detailed definitional analysis oflexical items. Two points will be discussed with regard to which Cuyckens's analysis would seem to be subject to improvement: a moderate lack of comprehensiveness, and an insufficiently specific treatment of the links between the various senses of over. Both of these characteristics seem to follow from a tendency (not altogether surprising in the framework of a prototypetheoretical approach) to concentrate on the most salient instances of use of a category: on the one hand, there will be less attention for the non-typical cases; on the other, the salient cases will tend to be seen as self-contained entities. In contrast with such a synthetic conception of the salient instances of a category, an analytical alternative will be presented that not only tries to determine the conceptual relationship between the various instances of a category as precisely as possible (specifying the various attributes that link the instances to each other), but that also tries to make sense of less obvious cases. There are a number of restrictions on the following treatment of over that should be mentioned explicitly. Only the prepositional uses of over in the strictest sense will be taken into account; predicative and adverbial uses (de pUn is over, de pUn gaat over), or the use of over in compounds (overblijven, overdenken, overgaan etc.) are beyond the scope of the present treatment. The discussion will be limited to the spatial senses of the preposition; figurative and temporal uses (een disGussie over iets, een vergadering over drie dagen) will not be envisaged. Further, a lexical field approach (which would involve comparing over with related prepositions, specifically
50
Prototypicality and salience
with boven) will not be systematically adopted (although there will be incidental remarks about synonymy and near-synonymy). Likewise, there will be no contrastive analysis (with the BrugmanlLakoff-analysis of English over, or with German iiber, which combines the readings of over and boven). Since these are all legitimate topics for investigation that should really be taken up in a more extensive treatment of the preposition, it will be clear that the restrictions are a matter of convenience rather than principle. Like Cuyckens's analysis, the discussion focuses on the core of the semantic structure of over, which would be the basic step in a more detailed description at any rate. There is, however, one restriction on the present treatment of over that does involve matters of principle rather than convenience. In fact, the paper will remain agnostic about the question whether the various kinds of usage that will be described, constitute different meanings or not (in the theoretical sense in which 'meaning' is distinguished from 'vagueness'). How, in fact, to determine whether a particular instance in which a lexical item is used constitutes a different meaning of that item? As argued in Geeraerts (l993a), not only are there various kinds of semantic identity tests that might be invoked (and that have been invoked) to settle the question, but also, these various tests appear to yield mutually contradictory results. To say the least, linguistic semantics has not come up with a universally accepted, unequivocal operational test for distinguishing polysemy from mere referential vagueness. Against the background of this methodological indetenninacy, the present paper will not endeavor to establish the meanings of over (if any such set of clearly delimited semantic entities exists at all), but will rather pursue a more modest goal; it will try to describe the structure of the referential range of application of over in some detail. Which portions of the structure constitute a different meaning in the theoretical sense is a question that is beyond the intended scope of this paper - but conversely, it is a question that can probably only be answered properly on the basis of a referential analysis of the type to be demonstrated here. Finally, there is a restriction on the originality of the following analysis. Although it will try to go one step further than Cuyckens's analysis by presenting an alternative for the overall structure of the item, it also draws heavily on his work when individual applications are at stake (just like Cuyckens himself has drawn on the work of Brugman and Lakoft). The distinctions and correspondences will not be mentioned separately each time they appear, but a global comparison with Cuyckens's proposals will be made in paragraph 6.
The semantic structure ofDutch over
2.
51
The basic referential structure
The absence of a generally applicable operational criterion for meaning discrimination does not imply that the existing approaches could not be used as a starting-point for the kind of structural analysis to be illustrated here. For instance, lexical polysemies could at least in some context show up as a sentential ambiguity. To take over eruse's example (1982), the sentence (1), when uttered at the meeting of a political committee, may mean either that delegate Balder has taken another seat in the conference room, or that he has changed his opinion on the topic at hand. This ambiguity, then, testifies to the polysemy ofposition. (1)
(2)
Bafder has shifted his position! Wafdemar fletst over de fijn
'Waldemar rides his bicycle over the line' Now consider (2).2 Sentence (2) may refer to at least three different situations. First, Waldemar may be riding across the line (as when he moves from one lane to the other). Second, he may be riding on and along the line, staying roughly within its boundaries as he follows it. (He might be doing this as a game, or as a kind of test to enable the police to ascertain whether he is riding under the influence of alcohol). And third, Waldemar may be riding at the other side of the line (seen through the eyes of an implicit observer). For instance, imagine a group of children riding their bikes at a schoolyard, but being forbidden to cross a particular line. Sentence (2) could then be used to point out that Waldemar ignores the prohibition. Notice that the implicit observer turns this third reading into an example of subjectification in the sense of Langacker (1990b): the spatial relation that is expressed by over holds between the bicycle (or at least, between the act of riding the bicycle) and the line, but it is construed with an implicit reference to the subjective presence of an observer. In Figures la, 1b, and lc, the triple ambiguity of (2) is graphically represented. Accordingly, over can be assigned three distinct readings, as defined in [i]-[iii]. With regard to example (2), the definitions specify the spatial characteristics of the activity referred to by the verb fietsen: it is an activity that extends from one point to another (either within the spatial area defined by the prepositional object x, or by crossing that area), or an activity that is situated in a region construed as being to the other side ofx.
52
[i] [ii] [iii]
Prototypicality and salience
Extending from a point or region at one side ofx to a point or region at the other side of x, while crossing x Extending from one point or region within the space defined by x, to another point or region within that space (without going outside of it) Situated at the other side of x, relative to the position of an observer
1b
1a
~, .. _--
1c Figure 1. The basic semantic configurations of over
Now, it is crucial for the semantic structure of over that the three spatial relations as just defined can also occur in other syntagmatic contexts than the one represented by (2). First, the prepositional phrase can indicate the spatial extension of a fictive rather than a real motion. In (3), for instance, there is no real motion of the line; still, the line extends from a point on one side of the prepositional object x to a point to the other side of that object (which motivates the use of over), and further, this extension is construed, through the use of lopen, as a process of fictive motion. 3 Second, if over can be used to express stative spatial relations, it should be possible to combine it with stative verbs that do not express motion. An example is (4), where it is said that the blanket extends from one side of the fence to the other; it thus covers the fence, if one likes. Example (5) presents another syntagmatic context in which the 'zero motion' option is realized; if the prepositional phrase modifies a noun rather than a verb (and if the noun does not refer to a process or an action), the interpretation is automatically stative.
The semantic structure ofDutch over (3)
(4) (5) (6)
53
Er liep een rode streep over het woord verantwoort 'A red line ran across the word verantwoort' (which contains a spelling mistake) De deken hangt over de schutting 'The blanket hangs over the fence' De deken over de schutting is van Godelinde 'The blanket across the fence belongs to Godelinde' Het vliegtuig vliegt over de stad 'The airplane flies over the town'
It will be clear by now that the semantic structure of over involves at least two dimensions: on the one hand, the spatially relational dimension in the strict sense (with regard to which three distinct spatial configurations have to be distinguished), and on the other, the 'motional' dimension, with regard to which we have to distinguish between cases of real motion, fictive motion, and zero motion. A third dimension to be added involves the existence of actual contact between the prepositional object and the other entity involved in the spatial relation. As (6) exemplifies, the reading defined in [i] can also occur when there is no physical contact between trajector and landmark. (In fact, (6) could be shown to be ambiguous along the three readings [i]-[iii]). On the basis of these three dimensions, the referential range of application of over can be charted systematically. As Table I and the examples (7)-(22) show, almost all possible combinations of all values on the three dimensions actually occur. (7) (8) (9)
(10) (11) (12) (13) (14)
De haf roft over de fijn 'The ball rolls over the line' (to the other side of it) Het vliegtuig vliegt over de stad 'The plane flies over the town' (to the next one) Er liep een rode streep over het woord verantwoort 'A red line ran across the word verantwoort' De spoorlijn loopt over de rivier 'The railway line runs across the river' De deken hangt over de schutting 'The blanket hangs over the fence' Er hangen donkere wofken over de stad 'Dark clouds hang over the town' Engefram kuiert over de markt 'Engelram strolls over the market-place' O([ert rijdt over de fijn 'Olfert rides over (on and along) the line'
54
Prototypicality and salience
Table I. Combinations of values in the semantic structure of over FICTIVE MOTION
MOTION
ZERO MOTION
+contact
-contact
+contact
-contact
+contact
-contact
[i 1
(7)
(8)
(9)
(10)
(11)
(12)
[ii 1
(13)(14)
(15)
(16)
(17)
(18)
(19)
[iii 1 (15) (16)
(17) (18)
(19) (20)
(21) (22)
(23)
(20)
(21)
(22)
Het vliegtuig cirkelt over de stad 'The airplane circles above the town' Condensatiestrepen lopen kriskras over de blauwe lucht 'Condensation trails run haphazardly over the blue sky' Ermingards blik dwaalde over het blad 'Ermingard's eye wandered over the sheet of paper' Er figt een armoedig kleedje over de vloer 'A shabby rug lies on the floor' Een dunne streep mist hangt over de rivier 'A small trail offog hangs over (above and along) the river' Over de rivier exerceerde de vijandelijke mifitie 'Across the river, the enemy militia was exercising' Over de rivier loopt een rij bomen fangs de beek 'Across the river, a row of trees runs along the brook' Dordrecht figt over de Moerdijkbrug 'Dordrecht lies across the Moerdijk bridge' Over de brug hangt een ba//on 'Across the bridge, a balloon hangs in the air'
Three refinements are necessary to complete this picture of the basic range of application of over. First, it will be noted that the distinction between and <+contact> cases is basically irrelevant for the examples of [iii], because the position of the trajector x to the other side of the landmark (the prepositional object) automatically implies that there is no contact between them. It is, of course, possible to interpret the distinction between (22) and (23) as involving a <-contact> versus <+contact> situation, but because the contact in question does not involve the bridge (which is the landmark for the relation expressed by over) but rather the ground in general, the distinction does not affect the structure of over. In this sense, then, the gaps in the structure represented in Table 1 are systematic, not incidental ones.
The semantic structure ofDutch over
55
As a second remark, it should be established that the distinction between [i], [ii], and [iii] is not just one between, respectively, a 'directional' sense of over, a 'path' reading, and a 'locational' interpretation. Judging only on the basis of the leftmost column of Table 1, it would seem possible to say that over specifies either the direction of a particular movement, the path along which a movement takes place, or the place where something is located. However, it should be clear from Table 1 as a whole that the readings of [i] and [ii] are just as locational as [iii] as a whole. And third, let us notice that there is a particular complication with resultative readings of the verbs, when a location referred to by over is presented as the result of the movement referred to by the verb. The problem can best be introduced in connection with transitive causative constructions. In such constructions, the moving object is not the subject of the verb, but an entity on which the verbs exerts a direct or indirect force. To begin with, consider cases of the causative construction with what we have called the case of [i], as in (24)-(25). In these cases, the distinction between [i] and [iii] fades away, given that there are various possible paraphrases for the causative construction: while [i] characterizes the directional path of the movement of the affected entity, [iii] simultaneously characterizes the resultative final state of that movement - and both ways of reading the causative construction seem equally appropriate. Thus, (24) could be paraphrased as 'Lidewij causes the ball to move over the fence' (i.e. from one side of the fence to the other, which brings into play the directional reading [i] of over), but also as 'Nimrod causes the ball to reach a position over the fence' (i.e. at the other side of the fence, which activates the locational reading [iii]). If we call the latter interpretation the resultative one, the examples (24)-(25) would seem to suggest that the locational resultative case is epiphenomenal with regard to the directional reading of over. (24) (25) (26)
(27)
Lidewij gooit de bat over de schutting 'Lidewij throws the ball over the fence' Fiobert duwt de fiets over de markt 'Fiobert pushes the bicycle over the market place' Evermoed liet de deken over de schutting vallen 'Evennoed dropped the blanket over the fence' (in the sense that it hangs over the fence) De deken va!t over de schutting 'The blanket falls over the fence' (in the sense that it falls on top of it and subsequently hangs over it)
56
Proto~ypicali~vand salience
In (26), however, the movement of the affected entity cannot be described in terms of any of the cases charted in Table 1, although the resultant state is a simple case of the application of [i]. Consequently, the resultative cases cannot always be reduced to a by-product of a directional reading, and the label in Table 1 should be read so that it includes resultative constructions like in (26). At the same time, we can now also see from (27) that such resultative constructions are not restricted to transitive causatives. The 'resultative' construction undoubtedly deserves to receive a closer look than the one it has been subjected to here; in the context of the present paper, however, it will play only a minor role.
3. Sample syntagmatic restrictions The survey of the range of application of over given in the previous paragraph has to be further elaborated in two different ways. On the one hand, it has to be supplemented with some peripheral cases of semantic relations (which will be the subject matter of paragraph 4). On the other, it has to be refined by an investigation into the syntagmatic restrictions that the various applications of over are subject to. In fact, not any kind of trajector can be associated with any kind of landmark by means of over. Notice, to begin with, that the case of [i] can be used without restrictions. Or at least, objects, surfaces, and lines can move over each other in all possible trajector/landmark combinations, as shown in Table 2 and examples (28)-(45). (Within each cell of the figure, the first example is a <+contact> one, the second a <-contact> onel (28)
Berthilde jiefst over de heuvel naar huis
(29)
Fredegonde jietst over het plein naar het begin van de Agnietenstraat
'Berthilde bicycles home over the hill' 'Fredegonde bicycles over the square to the beginning of the Agnietenstraat' (30)
Dietbald jietst over de witte lijn
(31)
De trein stoomde over de heuvel
(32)
De trein stoomde over de vlakte
(33)
De trein stoomde over de grens
'Dietbald crosses the white line with his bike' 'The train steamed over the hill' 'The train crossed the plain' 'The train crossed the border'
The semantic structure ofDutch over
57
Table 2. Trajector-landmark combinations in the structure of over LM OBJECT
LM SURFACE
LM LINE
TR OBJECT
(28)(37)
(29)(38)
(30)(39)
TRSURFACE
(34)(40)
(35)(41)
(36)(42)
TR LINE
(31)(43)
(32)(44)
(33)(45)
(34)
Een Romeins legioen marcheerde over de heuvel
(35) (36) (37)
Een Romeins legioen marcheerde over het plein Een Romeins legioen marcheerde over de grens Het vliegtuig vloog over het monument
(38)
Het vliegtuig vloog over de vlakte naar de stad in het gebergte
(39)
Het vliegtuig vloog over de rivier
(40)
De wolk dreef over het monument
(41) (42) (43)
De wolk dreef over de vlakte De wolk dreef over de rivier Een rij vliegtuigen vloog over het monument
(44) (45) (46)
Een rz/ vliegtuigen vloog over de vlakte Een rij vliegtuigen vloog over de rivier Het vliegtuig cirkelt over de stad
,A Roman legion marched over the hill'
'The plane flew over the monument' 'The plane flew over the plain towards the mountain town' 'The plane flew over the river' 'The cloud drifted over the monument'
'A row of airplanes flew over the monument'
'The airplane circles over the town' In passing, it could be noted that the classification of an entity as either object, line, or surface can be a matter of perspectival construal rather than matter-of-fact objectivity. To give just one example, a town could be seen as a point-like object, but it can also be conceived of as a surface. In a sentence like (46), for instance, assigning [ii] to over automatically entails thinking of the town as a surface: because the movement of the plane has to remain within the (vertically projected) boundaries of the town, thinking of the town as a two-dimensional surface is the natural thing to do. Now, without systematically exploring the restrictions on all the possibilities mentioned in Table 1, let us consider two major examples of limitations on the combination of trajector and landmark. Importantly, the restric-
58
PrototypicaliZv and salience
tions will turn out to be an automatic consequence of the spatial relations defined by [i]-[iii]. (Later on, we shall come across an example of restrictions that are truly ad hoc.) To begin with, let us consider [ii]. Because the spatial relation referred to by over has to remain within the boundaries of the landmark, the prepositional object has to have a line- or surface-like extension: it should be possible to conceive of it as a line or a surface. This shows up quite clearly when considering three-dimensional objects. Whereas (47) and (48) are straightforward cases of lines and surfaces, (49) is an example of a threedimensional object whose third dimension (height) does not prevent it from having a bounded surface that allows for an over-relation. In (50), however, the three-dimensional object's vertical dimension is so important that no surface for aimless wandering can be imagined. As (51) shows, reading [i] of over does not impose this restriction. This does not mean, however, that an object such as schutting could not occur at all with reading [ii] of over. Example (52) shows that the relevant feature is not the objective horizontal or vertical orientation of the surface, but rather the functional characteristic of 'treadability': what is not a surface for free movement for human beings is a natural path for insects. Furthermore, (53) shows that (51) is actually ambiguous between [i] and [ii]5 Examples (51) and (53) are possible with [ii] if a situation is envisaged in which someone crawls along the top of the fence; in that case, the relevant character of the landmark 'fence' is not that of a surface, but that of a line. That is to say, any interpretation of (51) and (53) on the basis of[i] would have to link up with (47) rather than (48). (47)
(48) (49) (50) (51)
(52) (53) (54)
Trudofietste over het pad 'Tmdo rode his bicycle along the path' Sulfrida kuierde over de markt 'Sulfrida strolled over the market place' Kunigonde dwaalde over de berg 'Kunigonde wandered over the mountain' *Dankmar dwaalde over de schutting 'Dankmar wandered over the fence' Gradolfklom over de schutting 'Gradolf climbed over the fence' De spin kroop maar door over de schutting 'The spider kept on crawling over the fence' Odilde kroop maar door over de schuffing 'Odilde kept on crawling over (on and along) the fence' De trein glijdt over de monorail 'The train glides over the monorail'
The semantic structure ofDutch over (55) (56)
(57)
59
Een kolonne soldaten marcheerde over de heuvel 'A line of soldiers marched over the hill' Een rij pukkels over de wang van Adelheid is voor haar een bron van veelzorgen 'A row of pimples on Adelheid' s cheek is a source of much concern to her' Een rij bomen loopt schuin over het veld 'A row of trees mns diagonally over the field'
On the other hand, because the trajector of an over-relationship according to [ii] has to remain within the boundaries of the landmark, it will typically be smaller than the landmark - small enough, at least, to move about on the surface or along the line constituted by the landmark. Specifically in the case of line-like landmarks, however, this does not imply that the trajector has to be smaller than the landmark in all directions. Taking into account that real lines are never merely one-dimensional, but always have a certain width as their second dimension, the trajector does not have to be narrower (but it does have to be shorter). As example (54) shows, the moving object may be wider than the actual line. Two additional remarks have to be made at this point. First, the distinction between line-like and surface-like landmarks in combination with [ii] is not unimportant, because it correlates with a restriction on the kinds of movement that are possible with reading [ii]. Line-like landmarks, in fact, automatically impose an orientation on that movement. Because the movement goes from one point on the line to the other, and because the line restricts the sideward movement of the trajector, the movement follows the line. Automatically, the line defines the path of the movement. Surface-like landmarks do not define such a path: the movement from one place to another on the surface may be random (as in (49), where there is a suggestion of aimless wandering or loss of orientation), but it may also be a straight line. Because, in the latter case, the straight line will typically lead from one side of the surface to the other side, there is a really close relationship with the 'crossing' reading [i]: going over a surface according to [ii] is then the same thing as being in the middle of the process of going over that surface according to [i]. As a second remark, let us note that an example like (55) shows that a line-like trajector may be discontinuous, or, if one wishes, virtual. If we jump to the and applications of [i] for a moment, further examples may be found of such virtual/discontinuous trajectors, as in (56) and (57). Needless to say, examples such as these again
60
Proto(ypicali(y and salience
underline the importance of perspectival construal in natural language: spatially disconnected entities are conceived of as a single entity. Let us now turn to a second major example of syntagmatic restrictions. The restrictions that hold for the reading of [i] are to some extent the reverse of those just described. With regard to [ii], we noted that the landmark was necessarily line-like or surface-like, and that the trajector had to be smaller (in the relevant dimension) than the landmark. In the case that we are concerned with here, the trajector has to be construed as line-like or surface-like, and the landmark has to be smaller (in the relevant dimension) than the trajector. On the one hand, the trajector of the cases of [i] has to have a line-like or surface-like extension of its own. Because the spatial extension that is expressed by the prepositional phrase does not involve the path of a movement (of an entity of any dimensionality), but rather attaches to the entity itself, the trajector cannot be construed as pointlike. On the other hand, because the trajector has to extend from one side of the landmark to the other, it has to be larger than the latter. In the case of line-like trajectors, this means that the landmark is smaller than the trajector in the direction in which the trajector extends. (As the examples (58) and (59) illustrate, the landmark may very well be larger than the trajector in the direction orthogonal to the trajector's.6) (58) (59) (60) (61) (62) (63) (64) (65)
Er ligt een kabel over de weg 'There's a cable lying over the road' Er hangt een deken over de schutting 'There's a blanket lying over the fence' Een deken ligt over het bed 'A blanket is lying across the bed (covers the bed)' Heriswitha legt een vel papier over de vlek 'Heriswitha puts a sheet of paper over the blot' Het rode blad ligt over het gele blad 'The red sheet lies over the yellow one' Ricfridus heefl sproeten over heel z 'n gezicht 'Ricfridus has freckles all over his face' Er ligt een hoop zand over het emmertje 'A heap of sand covers the little bucket' De kaasstolp Iigt over de Camembert 'The cheese cover lies over (covers) the Camembert'
In the case of surface-like trajectors, the trajector may extend from one side of the landmark to the other in various directions, but it should do so in at least one direction. At this point, we come across a phenomenon of the
The semantic structure ofDutch over
61
same kind as the one we encountered when we discussed line-like landmarks in our first example of syntagmatic restrictions. Just like applications of [ii] with line-like trajectors yield the interpretation 'following along x', applications of [i] with surface-like trajectors may result in the interpretation 'covering x', as in (60) and (61).7 Another special case of surface-like trajectors and landmarks is when they are both the same size. When you have a yellow sheet of paper and a red one of precisely the same size, (62) is possible when one sheet lies precisely on top of the other. This is a special case because it is situated at the borderline between [i] and [ii] (both taken in the interpretation, of course): according to [i], the top page should be larger, whereas it should be smaller than the bottom page according to [ii]. Equal size is so to speak the upper limit of [ii] and the lower limit of [i]. However, because it would be natural to say that the top sheet covers the bottom one, and because the 'covering' interpretation is a usual one with [i], there would seem to be at least one indication that special cases of the 'equal size' kind are intuitively conceived of along the lines of [i] rather than [ii]. Note also that the equal size cases also occur with 'virtual' examples that are analogous to (56) and (57). In an example like (63), the virtual surface defined by the freckles is implied to have the same size as the face; characteristically, the freckles cover the face. To round off, let us have a look at (64) and (65). Because the trajectors in these cases are neither clear examples of lines or of surfaces, it might seem that they contradict the earlier statement that the trajectors of the interpretation of [i] are construed as line-like or surface-like. One way out of the problem would be to claim that the relevant 'surface' in these cases is the vertical projection of the maximal expansion of the trajectors; and another would be to point out that surface-like trajectors may involve curved and bended surface. However, rather than trying to salvage the letter of the original statement of the restriction, it would be more in line with the spirit of the original statement to accept that it has to be slightly modified: the basic point is to see that point-like trajectors are excluded (because they have a minimal extension), but this does not rule out object-like trajectors that are construed as masses or spheres rather than points.
4.
Some less systematic cases
Three cases have to be mentioned that fall outside the range of applications charted in Table 1. (Each time, the restrictions on the additional cases will
62
Prototypicali~yand salience
turn out to be ad hoc rather than an automatic consequence of the spatial relations expressed by over, as in the previous paragraph.) The most common case is illustrated by (66)-(69); the meaning expressed by over can be defined as in [iv]. This kind of usage can most easily be seen as a transformation of those that fall under [iii]; specifically, the prepositional object is now the starting-point of a virtual path rather than the path itself. In contrast with [iii], the path is necessarily an open area such as a street or a square (or even a dinner table) rather than just any obstacle that can be crossed: examples such as (66) are ruled out in a situation in which the museum is situated at the foot of a hill and it is your intention to make clear that Gosbert lives over the hill. This is to say that this kind of usage is subject to specific restrictions that do not follow automatically from the spatial relation that is being expressed. The reading of over at stake here is not just a metonymic transformation of [iii], but it is subject to additional restrictions that are absent in the case of [iii]. The restrictions can probably best be understood as involving two people that can see each other in the face (which explains the open space), or entities such as buildings, that are anthropomorphically modeled as looking out towards each other. In fact, the English word facing that can be used to translate over in these cases, expresses the same anthropomorphic metaphor. (66)
Gosbert woont over het museum
(67)
Aan ta/el zat JvJeinol( over JvJombert
(68)
Het museum bevindt zich over de rechtbank
(69)
In de erehaag stond Edmar over Notker
'Gosbert lives across the museum' At the table. Meinolf was sitting facing Mombert 'The museum is situated across the magistrates' court' 'In the double row of honor, Edmar stood facing Notker' [iv1
Situated at the end of a virtual path that extends from x in a straight line across an open space
A second additional case to be mentioned is illustrated by (70)-(72). The meaning of over in these examples can be defined as in [v]. Clearly, this type of usage links up with the <-contact> cases of [ii]. As in the latter, an entity is situated spatially within the boundaries (through vertical projection) of x, but above x. The distinction between the cases of [ii] and [v] is that the former presupposes that the trajector has a certain extension (as of a line or a surface), whereas the latter specifically allows for point-like trajectors (that are so to speak without extension). This kind of usage, then, is a relaxation (in fact, a generalization) of the
The semantic structure ofDutch over
63
reading described under [iii]. As shown by (73) and (74), the relaxation is subject to the ad hoc restriction that only <no contact> cases are allowed. It should be remarked that the presence or absence of [v] is particularly sensitive to relatively minor situational differences and differences of perspectival construal. If, for instance, (75) applies to a situation in which the dangling movement of the spider covers a sizeable area, (75) would not be interpreted along the lines of (72), but along those of(15). And whether (76) is a case of [v] or [ii] would seem to depend on the size of the kettle with regard to the fire (or on the extent to which these are taken into account by the observer). Methodologically speaking, the ease with which possible cases of [v] shift towards one of the other readings testifies to the peripheral character of [v]. (70)
De lamp hangt over de tafel
(71)
Het ongeluk gebeurde over de zee
(72)
Het spinnetje hangt sti! over de vensterbank
(73)
*Het spinnetje figt sti! over de vensterbank
(74)
*Het boek rust over de tafel
(75)
Het spinnetje bungelt over de vensterbank
(76)
De ketel hangt over het vuur
'The lamp hangs over the table' 'The accident happened over the sea' 'Motionlessly, the little spider hangs over the window sill' 'Motionlessly, the little spider lies over (on) the window sill' 'The book rests over (on) the table' 'The little spider dangles over the window sill' 'The kettle hangs over the fire' [v 1
Situated above x, within the spatial boundaries of x
A third and final special case is illustrated by examples (77)-(80). In all of these, the trajector has a circular structure, while the landmark is like a pole or stick. While (80), which features a horizontal landmark, falls under the <+contact> application of [i] in roughly the same way that (11) does, the other cases are problematic because the landmark is in vertical position. A straightforward allocation to the <+contact> application of [i] is out of the question, because the trajector does not cross or cover the landmark: it does extend from one side of the landmark to the other, but it does so sideways, by going round the landmark rather than by crossing and covering it. According to this line of thought, the examples (77)-(79) illustrate a separate reading [vi], which can be considered a relaxation of the cases of [i]; specifically, the relaxation involves discontinuous trajectors (roughly, with a hole in them) that
64
Prototypicali(v and salience
extend from one side of the landmark to the other by going round it horizontally rather than by going over it vertically. It is important to notice, however, that the syntagmatic restrictions on this type of usage still reflect something of the original meaning [i] from which [vi] appears to be derived. In fact, as the ungrammaticality of (81) illustrates, the landmark cannot be so tall that the trajector could not have reached its position by being pushed downward from above; a trajector that has reached its position of sideways encapsulation by a roundabout motion cannot be said to go over the landmark. Apparently, only the same kinds of covering movement that lead to the configurations illustrated by cases like (60), (61), and (65) can lead to the spatial configuration captured by over in its reading [vi]. And because (60), (61), and (65) are unproblematic examples of [i], the link between [i] and [vi] shows up in the syntagmatic restrictions on over in reading [vi]. At the same time, it will also be clear that these restrictions cannot be predicted from the spatial configuration [vi] itself. They can be derived from the semantic history of the reading [vi] within the structure of over, but from the point of view of the configuration [vi] as such, the restriction is an arbitrary one. (77)
Er hangt een krans over het standbeeld
(78)
De ring ligt over de kegel
'A wreath hangs over the statue'
'The rings lies over the cone' (79)
Wigbert schuift de rol toiletpapier over de pleeboy
'Wigbert slides the toilet roll over the toilet roll stand' (80)
De zwembandjes hangen over een lange haak
(81)
*Een geellint was over de boom gebonden ,A yellow ribbon was tied over the tree'
[vi]
Circularly encapsulating x, extending from one side of x to the other by going round x
'The lifebelts are hanging on a long peg'
To summarize, we have found three special cases that can each be considered a relaxation or generalization of one of the three basic readings discussed in section 2: [iv] is linked to [iii], [v] to [ii], and [vi] to [i]. On the one hand, then, [iv]-[vi] correlate with and testify to the systematicity of the basic range of application of over as discussed in section 2. On the other hand, the less systematic features of [iv]-[vi] show up in the fact that the syntagmatic restrictions that they are subject to CaImot be derived automatically from the semantic content that they express (in contrast with the syntagmatic restrictions on the major readings that were treated in section 3).
The semantic structure ofDutch over
65
5. Prototypicality effects The analysis so far might seem to imply that all applications of over have the same status as far as structural weight is concerned. This is hardly the case, though: there are unmistakable prototypicality effects at work. Intuitively at least, there can be no doubt that (for instance) the cases discussed in the previous section are less preponderant than [i] or [ii]. Given that we do not have psycholinguistic experiments at our disposal to corroborate this intuition, what criteria do we have to establish differences in structural weight among the various instances of over? In the first place, the semantic relations between [i],[ii], and [iii] establish the central role of [i], at least if it can be accepted that that reading is central from which the others can be most economically derived. In particular (restricting, for the sake of the argument, the discussion to cases of real motion), [i] specifies a path across an obstacle while [ii] refers to a part of that path, and while [iii] refers to the endpoint of the path. Both [ii] and [iii] are therefore relatively straightforward metonymical extensions from [i]. Taking into account, however, that metonymy often works both ways, is there any reason why, for instance, [ii] instead of [i] could not be the logical centre along this axis of the structure of over? Notice that the transition from [i] to either [ii] or [iii] can be taken in one step: there is a straightforward type of metonymy (basically of the synecdochal kind) to justify the development. Going from [ii] or [iii] to [i], however, is less straightforward. For instance, while [ii] merely specifies that a certain movement is situated within the boundaries of the landmark, there is no compelling reason why that movement should be intended at crossing the landmark; that is to say, because movement across x always implies movement within the boundaries of x, but not the other way round, deriving [ii] from [i] is more plausible than the alternative. Similarly, while crossing x normally implies that one reaches the other side of x, being located at the other side of x does not imply that one has reached that position by crossing x. In the second place, the syntagmatic restrictions on the trajector/landmark combinations discussed in section 3 point towards the application of [i] as the unmarked case within the structure of over. This is the only case, it will be remembered, that can be used with trajectors and landmarks of all types, and that therefore has the widest applicability. The restrictions on the other applications inevitably mark them off as special cases. At this point, another particular kind of restriction (one we have not discussed before) has to be pointed out in connection with the
66
Prototypicali~y and salience
tion> and cases. Examples like (82)-(85) show that a particular situation that does not involve motion can more easily be expressed as a fictive motion case than as a zero motion case, i.e., using a verb that is originally a motion verb enhances the grammaticality of the sentences. ?
(82)
Er bevindt zich een streep over het bord
'There is a line over the blackboard' (83) (84)
Er loopt een streep over het bord
;A line runs over the blackboard'
Er zijn condensatiestrepen over de lucht
'There are condensation trails over the sky' (85)
Er lopen condensatiestrepen over de lucht
'Condensation trails run over the sky' The phenomenon would have to be investigated more systematically; for instance, the applications of [ii] seem to be more susceptible to the restriction than those of [i]. Even so, the restriction as such seems to support the claim that cases are somehow more marked within the structure of over than cases. In the third place, the existence of alternative expressions for the same referential situation may point to the peripheral status of a particular kind of usage of over, specifically when the alternative expression is intuitively more natural than over in that particular context. The underlying idea is that the basic, central instances of over will also be those that are less like to have strong competitors for the expression of a particular concept. This is, in fact, a particular version of the idea that form and meaning have parallel characteristics: the whole idea boils down to the hypothesis that semasiological centrality (of instances with regard to a category) will be reflected by onomasiological dominance (of a form with regard to a referent). Now, if we take up again some of the examples from Table 1 (and some additional ones discussed in section 4), we can try to deternline whether over has strong competitors in those particular contexts. In examples (86)-(100), the alternative expressions are printed in italics at the end of each sentence. Brackets round the alternative prepositions indicate that they are less likely than over in that context. A plus-sign indicates that they are more natural. In general, the list supports the view arising from the previous criteria, viz. that the upper left corner of Table 1 represents the central set of instances of over. (86)
(87) (88)
(89)
De bal rolt over de lijn: (naar de andere kant van) Het vliegtuig vliegt over de stad: (boven) De deken hangt over de schutting: (van de ene kant naar de andere) Er hangen donkere wolken over de stad: boven
The semantic structure ofDutch over
67
(90)
Engelram kuiert over de markt: (op) Het vliegtuig cirkelt over de stad: boven Er figt een armoedig kleedje over de vloer: op Een dunne streep mist hangt over de rivier: boven (94) Over de rivier exerceerde de vijandelijke mifitie: +aan de andere kant van (95) Over de rivier loopt een rij bomen langs de beek: +aan de andere kant van (96) Dordrecht figt over de Moerdijkbrug: +aan de andere kant van (97) Over de brug hangt een ballon: +aan de andere kant van (98) Gosbert woont over het museum: Hegenover (99) De lamp hangt over de tafel: +boven (lOO) De ring figt over de kegel: rand
(91) (92) (93)
To summarize, we find that there are several indications for a prototypical structure of over pointing in the same direction. It should be mentioned, though, that neither of the criteria used above can be considered a foolproof test of prototypicality; their parallelism, however, does seem to warrant the conclusion we have drawn from them. It is obvious, at any rate, that additional research (of a basically psycholinguistic nature) could be invoked to complement the analysis given here.
6. A comparison with Cuyckens Having charted the range of application of over, let us try to compare our own treatment with the one presented by Cuyckens (1991). In Table 3, the various senses of over distinguished by Cuyckens (1991: 282) are charted within the same framework as the one presented in section 2 (supplemented with the three additional readings described in section 4). Figure 2 is a copy of Cuyckens's own overview of the polysemy of over. In this classification, over} represents the 'above and across' reading, assuming a process of real motion. Over2 refers to the related cases where the movement is conceptualized as being situated within the (vertically projected) spatial boundaries of the landmark. As can be seen in Figure 2, over} and over2 only refer to surface-like landmarks; over3, however, takes care of point-like landmarks in the case of a movement 'above and across'. Next to theses three readings referring to <-contact> cases, the possibility of contact is recognized in over4, avers and over6. Judging from the graphic representations, 4 and 6 refer to cases where a surface-like or object-like landmark (respectively) is crossed. Overs, on the other hand, considers cases where the movement stays
68
Prototypicali~v
OVER1D
and salience
."
OVER 9
~
."
cb
I--
1
•
OVER 3 "
~.
.v
•
. V 7\ V ~. :
•
t--
"
r~ f--
... :
.
.
0
OVER 2
OVER 1
11
I1
-E)- 8 r--
OVERS
OVER 4
OVER 6 OVER 7
r~
~of'
... . )
OVCR8
x•r~• '.V
....
~
q,",,~ OVCR13
lOVER 12
fI] .
'
: :
'.'
...
.. : '." .
..
'
...
-
Figure 2. The analysis of over in Cuyckens (1991)
@ff]
The semantic structure ofDutch over
69
within the confines of the landmark, but specifically envisages the trajector to pursue a 'fairly rectilinear path' (1991: 280). The addition is not without importance: over]3 refers to those cases where the movement within a surface has no dominant direction, but rather takes a random form. In the analysis presented by Cuyckens, over13 ranges with the 'covering' -readings over11 and over12 rather than with over5. Over11 and over12 themselves involve surfaces (not necessarily of equal size) of which one covers the other. Overl2 specifically takes into account what we have called the 'discontinuous' cases in section 3. Because the graphic representation suggests that the trajector of over11 is at least the same size as the landmark, it should be noted that the examples given by Cuyckens also provide for trajectors that are smaller than the landmark. These have been included as over 11' in Table 3. Other additions are over] and over2', which refer to the counterparts of over] and over2. They are mentioned in the text, but are not counted as separate readings by Cuyckens. The remaining over's represent fairly straightforward cases. Over? is the 'at the other side of, relative to an observer' reading (in all possible combinations); over8 is the 'across from, facing' reading; and overQ and over]O correspond with the 'above' cases that require no particular extension of the trajector. (The distinction between 9 and 10 depends on whether the landmark is a surface or a point.) Table 3. Mapping Cuyckens' analysis onto the multidimensional structure of over MOTION
FICTlVE MOTION
ZERO MOTION
+contact -contact +contact -contact +contact -contact [i]
4.6
1,3
1'
11,12
---->
[vi]
[ii]
5,13
2
2'
11'
---->
[v] 9,10
7
7
---->
[iv] 8
[iii]
7
In comparison with the analysis presented by Cuyckens, there are a number of characteristics that are specific to the one conducted in the previous pages. First and foremost, the present analysis paints a more detailed picture of the conceptual relations between the various applications of over. As can be seen in Table 3, the over's that Cuyckens identifies consist of bundles of co-occurring features (such as a particular type of spatial relation coupled with a particular value on the 'motion/contact' dimension). But while the analysis presented here specifies and isolates those features, Cuyckens treats the bundles that constitute the 'senses' in his description as largely unana-
70
Proto~ypicalityand
salience
lyzable wholes. As can be seen in Figure 2, he does identify a number of larger groupings of related over's, but these groupings do not constitute a systematic exploration of the semantic relations between the various possibilities of use of over. Specifically, the fact that the relevant features determine highly overlapping sets of uses is not prominent in his analysis. (The groupings overlap in the sense that all subsets within the range of application of over that are defined on the basis of the 'motion' dimension cross-classify with the subsets defined on the basis of the 'contact' dimension, and with those defined on the basis of the spatial relations referred to by [i], [ii], and [iii].) Conversely, the analysis presented here precisely tries to describe the semantic structure of over as a cluster of subsets; rather than focusing on the individual readings, it focuses on the structure of the possible groupings of those individual readings. Thus, it not only systematically explores the relations between the various subsets into which the range of application of over can be divided, but it also yields an insight into the question what various kinds of features enter into the conceptual structure of over. Notice, in fact, that there are three major kinds of features involved: spatial relations as defined by [i], [ii], and [iii]; characteristics that pertain to the action, state, or event involving that spatial relation; and characteristics of the trajector and the landmark (and their combination). As a fourth type of a structurally relevant feature, the process of semantic generalization that lies at the basis of [iv], [v], and [vi] could be mentioned. At the same time, it has also become clear that there is much more systematicity in the structure of over than the picture presented by Cuyckens (see Figure 2) suggests: it will be sufficient to compare Table 3 and Figure 2 to appreciate the fact. The difference in focus between the analysis presented by Cuyckens and the one presented here can be captured (and summarized) terminologically by calling his outlook a synthetic one (because the emphasis lies on the individual applications of over as a synthesis of co-occurring features), whereas ours is analytic. Now, the analytic focus leads to two further differences with the synthetic one. First, whereas Cuyckens is rather hesitant to answer the question whether over has a prototype, an exploration of (among other things) the syntagmatic restrictions that the various readings of over are subject to in terms of trajector/landmark combinations unmistakenly supports the suggestion that over has a prototype structure (see section 5). Second, a systematic exploration of the range of application of over leads to greater comprehensiveness than Cuyckens has achieved. Apart from the obvious empty slots in Table 3, it is unclear why some features that are con-
The semantic structure ofDutch over
71
sidered criterial by Cuyckens in one part of the entire range, are not given the same treatment elsewhere. For instance, it is unclear why the distinction between <+contact> and <-contact> is considered criterial for distinguishing between over4 and overj, but does not lead to a similar distinction in the case of point-like trajectors (why doesn't Figure 2 contain a <+contact> counterpart of over] ?).
7.
Methodological concluding remarks
The distinction between the 'synthetic' approach and the 'analytic' approach has to be reinterpreted on a more general methodological level. Prototype theory has (rightly) stressed the important role that the individual members of a category may play in the structure of that category. However, there is a danger that this kind of emphasis may lead to the tendency to restrict the analysis to the salient members of a category, thus underestimating the importance of a definitional analysis that not only tries to indicate the semantic relationships between the salient instances of a category, but that also tries to make sense of less central cases. The salient instances of a category take the form of bundles of co-occurring characteristics, but the foregoing analysis of over suggests that the true semantic structure of a category can only be revealed by systematically unraveling the features occurring in those bundles, their kinds, and their mutual relations. The tendency to concentrate on the salient instances of a category carries with it the danger of neglecting such an analytical approach. At this point, we naturally come back to a question that was raised in section I: how to determine what the readings of over are? Against the background of the analysis summarized in Figures 2 and 4, the same question can be restated: what should we consider to be the 'meanings' of over - higher level groupings such as [i], [ii], and [iii], or lower level applications such as those charted by Cuyckens? In order to make the structure of the problem more apparent, I may refer to a distinction that I introduced in Geeraerts (l987a). On the one hand, the question how to determine what the meanings of a lexical item are may be approached from a psychological angle, taking into account phenomena such as intuitive salience effects. On the other hand, the question may be tackled in a purely definitional, analytical vein; according to the traditional requirements, separate meanings will then be maximally general subsets of the extension of an item that can still be defined by means of necessary and sufficient characteristics 8 Now, the intuitive approach
72
Prototypicali(v and salience
seems to point specifically to the lower level applications of over as distinct meanings: the psychologically salient aspects of over (for instance, the semantic values that can easily be brought to mind by the native speaker) are synthetic entities, bundles of characteristics like the various over's discussed by Cuyckens. Conversely, an analytical, definitional approach automatically shifts the focus towards the more general groupings of the basic applications (as represented by [i], [ii], and [iii], and the like). Is there any way, then, to decide which approach (the intuitive or the definitional one) is most appropriate? I would say there is not. For one thing, as argued in Geeraerts (l993a), the existing criteria for distinguishing true meanings from mere cases of referential vagueness do not yield a clear-cut, unambiguous picture of what 'meanings' are. Rather, the question itself what the meanings of over are might well be suspect: it presupposes that meanings are like entities, stored as clearly separated objects in a word's contents, whereas the semantic flexibility that has been reintroduced into lexical semantics by prototype-theoretical studies could very well (when taken to its extreme) lead to the conclusion that 'meanings' are contextually flexible rather than stable and unaffected by context. Moreover, even if we would reach an agreement about what the concept meaning precisely covers, this would be basically a terminological decision. It would not detract from the fact that (at least from a cognitive perspective) both the psychological phenomena and the definitional analysis are indispensable aspects of any truly adequate study of the semantic structure oflexical items: up to a point, the lexical-semantic structure that Cognitive Semantics would like to describe, resides precisely in the interplay between both types of data. In this sense, my argument in favor of the 'analytic' approach to over does not imply that the 'synthetic' approach can be neglected; rather, the whole point is that both approaches have to be taken into account. To round off, let us note that the analytical type of prepositional description as represented by Figures 2 and 4 can be seen as an application and concretization of the ideas expressed by Leys (1988). Leys, in fact, has suggested that the basic conceptual structure of spatial prepositional relations takes the form [X S/M] R [Y], in which X is the trajector, Y the landmark, R the spatial relation as such, and S/M the stative or motional situation that determines the 'aspect' of the prepositional construction. It will be clear that what I have tried to do is precisely to specify the various values that R can take in the case of over by means of definitions like [i]-[iii], and to identify the various aspectual values of the construction by means of the distinction between <motion>, , and cases. Whether the
The semantic structure ofDutch over
73
attempt is as yet entirely successful is another matter, but I hope it has become sufficiently clear that taking the approach advocated by Leys seriously, enables us to make at least some progress in comparison with alternative treatments of over.
Notes 1.
2.
3. 4. 5. 6.
7.
8.
Actually, eruse's example features Charles rather than Balder. However, less commonplace proper names will be used throughout this article to draw the attention to the fact that Odo Leys, to whom this paper and this issue of Leuvense Bijdragen are dedicated, has made some major contributions to the study of proper names apart from his work on the central themes of this collection (and his manifold other research interests). Translations and paraphrases of the Dutch sentences are given in square brackets. Occasionally, round brackets within the square brackets contain explanatory glosses. A classification of the various kinds of fictive motion that could possibly be distinguished is beyond the scope of the present article. Given the system according to which the examples are construed, English translations are not always given. In fact, [iii] is possible as well, but it is irrelevant for the present discussion. The label 'line-like' should not be taken too narrowly in this description. The blanket in (57) could probably also be conceived of as a surface-like trajector; in that case, the requirement that the landmark should be smaller than a linelike trajector in the direction in which the latter extends would have to be modified to provide for oblong surface-like trajectors. It may be noted, however. that a 'covering' interpretation is not necessarily unavoidable in the case of multidimensional over-ness: when er ligt een kleedje over de tarel [a napkin lies over the table] is used to describe a situation in which the napkin is placed like a lozenge, leaving the four corners of the table free, it does not exactly cover the entire surface of the table, although it does lie over it in more than one dimension. (Of course, another way of dealing with this situation would be to accept a relaxed interpretation of 'covering'. so as to include almost complete coverage.) For more details, see the (1987) article, and the (1991) paper. in which the distinction between both types of polysemy criterion is further elaborated.
Chapter 4 Salience phenomena in the lexicon. A typology
Originally published in Liliana Albertazzi (ed.), 2000, Meaning and Cognition 125-136. Amsterdam/Philadelphia: John Benjamins. As argued in Chapter 1, the essence of prototype theory lies in the fact that it highlights the importance of flexibility (absence of clear demarcational boundaries) and salience (differences of structural weight) in the semantic structure of linguistic categories. But whereas prototypicality was initially described for the semasiological structure of lexical items, flexibility and salience may be observed in many other circumstances. In particular, as suggested in Chapter 2, it is crucial for cognitive studies to switch from a semasiological to an onomasiological perspective, i.e. from an analysis conducted on the basis of words to one conducted on the basis of concepts. The present chapter, then, explores the concept of onomasiological salience, and puts it in the context of a systematic analysis of salience phenomena in the lexicon. The article draws on (and partially repeats) a number of case studies included elsewhere in this volume, viz. the over study of Chapter 3 and the beer names study of Chapter 10. Also, like Chapter 6, it includes some materials taken from The Structure of Lexical Variation (Geeraerts, Grondelaers, and Bakema 1994). The detailed empirical investigation of lexical variation (both semasiological and onomasiological) carried through in that monograph constituted the first step in the development of an onomasiological line of research within our Leuven research group (again, see Grondelaers and Geeraerts 2003 for an overview, and compare Chapter 11).
1.
Salience as a lexicological variable
When one considers the question of the impact of Cognitive Semantics (in the sense of Langacker 1990a, Lakoff 1987, or Taylor 1989) on the field of lexicological research, it is probably correct to say that systematic attention to salience phenomena is one of the major innovations of the cognitive paradigm. As an inquiry into language as a system of categories, Cognitive Semantics has received major impetus from the prototype-theoretical insights into the structure of linguistic categories developed in psycholinguistics and anthropological linguistics. And prototype theory is basically concerned with
Salience phenomena in the lexicon
75
salience phenomena: it shows how the various semantic applications that exist within the boundaries of one particular category need not have the same structural weight within that category. The cognitive semantic attention to salience, however, has not yet been as systematic as it might have been. There are forms of salience phenomena that have not yet been investigated as thoroughly and completely as prototype-based salience. The purpose of this paper, therefore, is to provide a tentative overview of the various forms of salience effects that may be usefully distinguished. A typology of salience phenomena will be presented, based on a distinction between the following basic types: perspectival, semasiological, onomasiological, and structural salience. Although terminological hair-splitting and taxonomic overabundance are real dangers for an endeavor such as this one, there is one overriding reason why the present exercise may be useful: salience is the place where structure and use meet. Consider lexical semantics and prototype theory. Traditional semantic analysis is concerned with describing the various meanings of lexical items and the structural relations among those meanings. Katz and Fodor's bachelor example, for instance, describes the readings that 'bachelor' may possibly yield in actual text, and at the same time describes the conceptual relations among those readings, in the form of a hierarchical structure based on shared semantic features (Katz and Fodor 1963). Such analysis is basically a structured list of possibilities: it specifies the space of possible readings that language users may choose from in actual communication. The study of those communication processes, the actual choices made by the language users, is not envisaged. Prototype theory, on the other hand, does incorporate aspects of use in the structural description of the lexical categories: the fact that some readings are more salient than others in the structure of the category reflects the fact that they are more readily chosen when using that category. Admittedly, frequency of occurrence is not the only factor determining prototypicality, but next to conceptual coherence, it is certainly an important one. It should be clear that the combination of structure and use in the study of salience phenomena is essentially also a combination of semantics and pragmatics: semantics as the study of structure, and pragmatics as the study of use, combine when salience is seen as the structural reflection ofpragmatic phenomena. This interlocking combination of the semantic and the pragmatic perspectives implies a slightly different conception of the notion of 'structure' than is customary: whereas linguistic structures would tradi-
76
Prototypicality and salience
tionally be seen merely as ordered sets of possibilities, adding pragmaticsbased salience implies introducing probabilities rather than just possibilities.
2. Perspectival salience
The first step to take involves a definition of the concept of perspectival salience - a type that more or less stands apart from the other cases to be treated, and which will only be mentioned briefly here. Perspectival salience is represented in Cognitive Semantics by such terminological pairs as profile and base, or more generally, jigure and ground (for definitions and applications of these concepts, see Langacker 1990a or Talmy 1978, among many other publications). To mention just the simplest of examples, the lexical item hand evokes a particular body part, but it primarily does so against a particular conceptual background, namely, that of the arm. The arm is presupposed in the concept of the hand, but it is the hand itself that is highlighted against the backdrop of the arm. Crucially, this is a matter of perspectivization: a particular chunk of reality (the hand and the arm) is perspectivized in a particular way. The fundamental distinction between this perspectival type of lexical salience and the ones to be mentioned later, resides in the fact that the other types defined below invariably involve differences of salience among lexicological alternatives (where the alternatives may be meanings, syntagmatic contexts, lexical forms, lexical categories, or distinctive dimensions). In the perspectival type of salience, on the other hand, the salience phenomena involve just one of these lexicological alternatives (and specifically, one particular lexical meaning); within such a single meaning, particular aspects of the extra-linguistic situation referred to receive more weight than others. Perspectival salience, in short, involves differences of perspectival attention within one particular construal of reality, whereas the other type (which we will call variational salience) involves preference relations among different construals of reality. Perspectival salience is defined between the different aspects of the extra-linguistic situation (in the broadest possible sense of 'situation ') invoked by a linguistic expression: 'hand' invokes the entire arm as the background of the conceptualization, but it is only the terminal part of the arm that is saliently conceptualized by 'hand'. Variational salience, on the other hand, is defined between different linguistic expressions or properties of those expressions: the semantic property of 'hand' that it may mean 'the terminal part of a human arm' is more salient than the property that it
Salience phenomena in the lexicon
77
may also mean 'manual worker' (as in 'hired hand') - which is then the less prototypical reading. Confusion should be avoided in particular when the linguistic expressions to be compared from a variational perspective each embody a different perspectival construal of the same chunk of reality. Choosing to describe a commercial transaction by either the verb 'to buy' or the verb 'to sell' involves an onomasiological choice. If one of both terms is the more obvious alternative, there is a difference of variational salience between both verbs: we would then say, for instance, that 'to buy' is a more salient way of talking about a commercial transaction than 'to sell'. On the other hand, each of the two verbs separately attaches perspectival salience to different aspects of the commercial transaction scene. In the case of 'to buy', the person buying is more salient than the person selling, and the reverse is true in the case of 'to sell'. In examples such as these, comparing the differences in perspectival salience of two linguistic expressions is not the same thing as comparing the onomasiological salience of the linguistic expressions themselves, that is, the likelihood that one rather than the other will be used in situations where both are applicable.
3. Semasiological, onomasiological, and structural relations As a further step in the construction of the classification, a distinction must be drawn between semasiological salience and onomasiological salience. Now, the conceptual distinction between semasiology and onomasiology may require some background information, because it involves a terminological pair that has - unfortunately -not made its way into the standard set of linguistic terms to be found in introductory courses. It is a terminological distinction that was introduced at the beginning of this century, at the time of the birth of structural methods of semantic analysis. The tern1S were coined by the Swiss romanist Karl Vossler in 1919. Terminologically speaking, the distinction is still much alive in those branches of (continental) European lexicology that preserve the heritage of structuralist semantics (such as the Coseriu school in Germany). In the Anglo-Saxon world, however, the terminology is hardly known. Basically, the distinction involves two ways of perspectivizing the structuralist link between sign~fie and sign~fiant. This terminological pair itself, to be sure, is not without having problems of its own, but for present purposes it may be sufficient to equate sign~fiant with a word form, and sign~fie with the concepts associated with that word form.
78
Prototypicali(v and salience
A semasiological perspective takes the word as its starting-point and investigates how several concepts (meanings, if one likes) are associated with that word. This is, of course, the regular type of semantic analysis in which the meanings of a word are listed and the semantic relationship between them indicated. Onomasiological research, on the other hand, reverses the perspective, and takes the concepts expressed as its starting-point; it investigates, basically, which various words may express a given concept, and what the structural links between those words are. In actual historical practice, the onomasiological perspective has been realized primarily in the form of lexical field research. Bruno Quadri's bibliographical overview of onomasiological research, published in 1952, for instance, reads as an overview of various types of lexical field analysis (Quadri 1952). Kurt Baldinger (again a Swiss romanist) has succinctly described the difference in the following way: 'Semasiology [... ] considers the isolated word and the way its meanings are manifested, while Onomasiology looks at the designations of a particular concept, that is, at a multiplicity of expressions which form a whole.' (Baldinger 1980: 278). An even more succinct way of putting things would be to say that semasiology involves processes of meaning, whereas onomasiology involves naming. Baldinger's formulation, however, also makes clear that there is a certain ambiguity in the definition of onomasiology. On the one hand, when onomasiology is conceived of in tern1S of 'designation' (or naming), it involves the relationship between semantic entities and the lexical items that name them. If, on the other hand, onomasiology is conceived of in terms of 'a multiplicity of expressions which form a whole', it instead involves the mutual relations between related lexical items. In this conception, onomasiological research will tend to coincide with lexical field research as a purely structural endeavor - that is, as an approach in which the actual choices from among the set of alternatives present within the lexical field are barely envisaged, but in which most if not all of the attention is directed towards an analysis of the relations between those expressions. In actual practice, in fact, so-called onomasiological research has been basically restricted to lexical field research of the kind just described. It may be useful, therefore, to impose a terminological distinction between both kinds of onomasiological research, because it will help us to distinguish between the purely structural and the pragmaticized conception of onomasiology: if the term onomasiology is reserved for the pragmatic approach in which the relationship with actual referents is explicitly envisaged, the other approach may then be called structural.
Salience phenomena in the lexicon
c
79
c
~-ii
R
w w
R
Figure 1. Semiotic relations between the major groups of salience phenomena in
the lexicon The differences between the three perspectives defined in this way, may be presented schematically by starting from two semiotic triangles as in Figure 1. If we maintain a distinction between Word (or lexical item, W in the figure), concept (C in the figure), and referent (R in the figure), then semasiology investigates how a particular lexical item may express various concepts, or alternatively (putting things extensionally rather than intensionally), how it may designate various types of referents. This is expressed by the arrows in the left hand triangle. Onomasiology in the restricted sense defined a moment ago describes how referents or concepts may be designated by various expressions; this is indicated in the right hand triangle. A structural analysis, finally, studies the semantic relationship between expressions. In the figure, this is indicated by the arrow in between both triangles. The arrow actually connects the lines going from C to W rather than just both W's or both C's alone, because the structural analysis involves expressions as combinations of forms and meanings, rather than just word forms or semantic concepts alone. (The double directionality of the arrow between the triangles, therefore, does not indicate a one-to-one correspondence; it merely represents the fact that the structural analysis is a reciprocal one: it looks, so to speak, at one item from the perspective of the other, and vice versa.) Given the distinction between semasiology and onomasiology, and the double interpretation of onomasiology, concepts of semasiological salience, onomasiological salience, and structural salience may now be defined.
4.
Semasiological salience
Semasiological salience is a relationship among the various semantic possibilities of a given lexical item. A lexeme/morpheme may express various semantic values, as when in expresses containment ('the apple in the bowl ') or inclusion/surrounding ('the village in the desert'). Some of the values
80
Proto~ypicali~y and salience
expressed by the lexical element may be more central than others, for instance because they occur more frequently within the range of application of the lexical element, or because they are a semantic centre in a network or family resemblance structure from which the other values are extensions. For example, containment seems to be more central in the structure of in than inclusion (see Vandeloise 1994). Practically speaking, then, semasiological salience coincides with prototypicality. As prototypicality effects constitute the best known type of lexicological salience, it is not necessary to recapitulate the literature here (see Geeraerts 1989c for an introduction, and Taylor 1989 for an extensive overview.) Two things require clarification, though, in preparation for the following discussion of onomasiological salience. First, prototypicality may be defined both as a relationship between a lexical item and its meanings, and as a relationship between a lexical item in one of its meanings and the referents corresponding with that meaning (compare Kleiber 1990). In the former case, it is pointed out for instance that the 'biological species' sense of 'bird' is more common than the slang sense in which 'bird' refers to a girl. In the latter case, it may be pointed out that robins and sparrows are more central members of the 'biological species' sense than, for instance, ostriches and penguins. This distinction between the two possible definitions of prototypicality hinges crucially on the distinction between polysemy and vagueness: of all the multiple applications of a lexical item, some constitute true meaning differences (and hence involve polysemy), while others should merely be considered referential specifications (and hence involve vagueness rather than polysemy). It has recently been suggested (Taylor 1992, Geeraerts 1993a) that the distinction between polysemy and vagueness may well be less stable than is traditionally assumed; if this is correct, the distinction between two different levels of prototypicality (the semantic one and the referential one) marks the endpoints of a continuum, rather than a clear-cut dichotomy. Second, it may be useful to distinguish between the paradigmatic and the syntagmatic aspects of lexical meanings. To begin with, let us note that each of the values expressed by a relational lexical element may occur in a specific syntagmatic frame. For relational spatial concepts, Cognitive Semantics describes such a frame in terms of a 'trajector' and a 'landmark' (see Langacker 1990a). For instance, the landmark of a containment relation has to be a recipient with, say, 'containing potential', whereas the trajector has to be something 'containable' in relation to the landmark. The containment relation itself is a paradigmatic value, whereas the containing potential of the
Salience phenomena in the lexicon
81
landmark is a syntagmatic phenomenon. These restrictions on syntagmatic contexts resemble the selection restrictions of old, except that they now explicitly involve (groups of) referents and not just words. For a more extended example, we may turn to a word that more or less functions as a classic example in Cognitive Semantics, namely the preposition 'over' (see Lakoff 1987). In Geeraerts (l992a) it is argued that there are basically three spatial relations to be distinguished in the semantic structure of the Dutch preposition over. They are defined a, b, and c below and illustrated by the examples (1 )-(3) (lrn stands for 'landmark', tr for 'trajector'). The range of application of over is not restricted to these cases, amongst other things because the spatial relations may show up in situations involving real motion, in situations involving fictive motion, and in examples involving stative relations. (a)
(c)
Extending from a point or region at one side of Im to a point or region at the other side of Im, while crossing hn Extending from one point or region within the space defined by Im, to another point or region within that space (without going outside of it) Situated at the other side ofhn, relative to the position of an observer
(1)
De bal rolt over de lijn
(2)
De agent kuiert over het marktplein
(3)
Over de rivier deed de atleet zijn oefeningen
(b)
'The ball rolls over the line (to the other side of it)' 'The policeman strolls over the market-place' 'Across the river, the athlete was exercising' The relevance of a separate analysis of the syntagmatic restrictions on trajector and landmark becomes clear, for instance, when one compares the selectional restrictions holding for these three basic relations. Without going into all the details, let us briefly note that there are hardly any restrictions on the trajector and landmark of semantic value (a): both trajector and landmark may be one-dimensional (points), two-dimensional (lines), or threedimensional (objects), in any combination. In the case of (b), however, both trajector and landmark are subject to restrictions. On the one hand, because the spatial relation referred to by over must remain within the boundaries of the landmark, the prepositional object must have a line- or surface-like extension: it should be possible to conceive of it as a line or a surface. This shows up quite clearly when considering threedimensional objects. Whereas (4) and (5) are straightforward cases of lines and surfaces, (6) is an example of a three-dimensional object whose third dimension (height) does not prevent it from having a bounded surface that
82
Prototypicality and salience
allows for an over-relation. In (7), however, the three-dimensional object's vertical dimension is so important that no surface for aimless or random wandering can be imagined. As (8) shows, reading (a) of over does not impose this restriction. This does not mean, however, that an object such as schutting 'fence' could not occur at all with reading (b) of over. Example (9) shows that the relevant feature is not the objective horizontal or vertical orientation of the surface, but rather the functional characteristic of 'treadability': something that could hardly be a surface for free movement for human beings may be a natural path for insects. On the other hand, because the trajector of an over-relationship according to (b) must remain within the boundaries of the landmark, it will typically be smaller than the landmark - small enough, at least, to move about on the surface or along the line constituted by the landmark. Specifically in the case of line-like landmarks, however, this does not imply that the trajector has to be smaller than the landmark in all directions. Taking into account that real lines are never merely one-dimensional, but always have a certain width as their second dimension, the trajector does not have to be narrower (but it does have to be shorter). As example (l0) shows, the moving object may be wider than the actual line. (The train is wider than the rail.) (4)
Janjietste over het pad
(5)
Jeanne kuierde over de markt
(6)
Mark dwaalde over de berg
(7)
*Jvfarie dwaalde over de schutting
(8)
Karel klom over de schutting
(9)
De spin kroop maar door over de schutting
(10)
De trein glijdt over de monorail
'John rode his bicycle along the path' 'Jane strolled over the market place' 'Mark wandered over the mountain' 'Mary wandered over the fence' 'Carl climbed over the fence' 'The spider kept on crawling over the fence' 'The train glides over the monorail' Now, the syntagmatic frame of each semantic value may have typical and less typical instantiations. For instance, 'the apple is in the bowl' would probably refer to a typical containment relation, whereas 'the jelly pie is in the suitcase' is an untypical one (even though the containment relation itself is the same as in the typical case). Such typicality relations illustrate the concept of syntagmatic salience. In the over example, those instances of
Salience phenomena in the lexicon
83
reading (b) that more readily invoke a line-like or surface-like landmark will be syntagmatically more salient than the others. In the sample sentences, for instance, (4) and (5) are more typical than (6) or (9). Along the same lines, examples in which the trajector is not entirely contained within the boundaries of the landmark - as in sentence (10) - are syntagmatically less canonical.
5.
Onomasiological salience
Turning now to onomasiological salience, there is a possible misconception to be discarded first. From a very general point of view, salience on the onomasiological level involves a preference for one expression over another as the name for a particular referent or type of referent. The existence of such alternative names would therefore seem to coincide with the kind of formal variation that is customarily studied in sociolinguistics and dialectology. Consider Table 1, which specifies a number oflexical patterns from the clothing terms project described in Geeraerts, Grondelaers, and Bakema (1994). In each of the cases in Table 1, there is a different naming pattern in Belgian Dutch and in Netherlandic Dutch; the clothing types themselves, however, are the same. The choice between legging and caler;on as a designation for leggings, for instance, therefore acquires the same value as the distinction between, for instance, different pronunciations of the same phonological form in different geographical regions or social strata of a linguistic community. This type of onomasiological variation may be called 'formal' or 'synonymy-based', but it is useful to point out that these terms are not ideal in all respects. Terms like legging and caler;on are denotationally synonymous, but they are not absolute synonyms, because they have a different geographical distribution. The two terms have the same meaning in a narrow, denotational sense of the word, but they clearly have a different sociolinguistic value (in a broad sense of 'sociolinguistic '). In a similar way, the term 'formal' variation (to which I shall adhere in the rest of the paper) might suggest that the variation in question is trivial or irrelevant, in the sense that it could be taken to involve mere 'notional variants'. Again, however, the alternation between legging and caler;on is not a question of free variation, but signals a specific stratification of the linguistic community.
84
Proto(vpicali(y and salience
Table J. Differences of naming patterns between Netherlandic and Belgian Dutch ITEMS
BELGIAN SOURCES
NETHERLANDIC SOURCES
calec;on legging leggings
40 (38%) 26 (24.7%) 39 (37.3%)
91 (100%)
blouson jack
13 (46.4%) 15 (53.6%)
85 (100%)
jeans spi}kerbroek
64 (97%) 2 (3%)
38 (70.4%) 16 (29.6%)
The misconception, then, consists of restricting onomasiological variation to this type of 'formal' variation. Next to the synonymy-based variation meant here, there are, obviously, cases where the choice of an alternative name carries a semantic load. Consider the following distinction between Belgian Dutch and Netherlandic Dutch. (11)
Hi} kwam langs de deur naar binnen
(12)
Hi} kwam door de deur naar binnen
'He came in through the door' (Belgian Dutch) 'He came in through the door' (Netherlandic Dutch) In Belgian Dutch, the usual preposition is langs 'via', where the semantic focus is on the path as such. In Netherlandic Dutch, the usual preposition is door 'through', which adds the notion of a passage that is penetrated and crossed. Although the preference for one or the other construction is conventional, it is not conceptually neutral: there is a different perspective involved in the choice of one over the other. This implies that next to the formal salience of synonyms that is at stake in the leggings/cale90n example, we have to distinguish a form of onomasiological variation that is semantic or conceptual in kind: it involves the choice of one semantic category (and not just one lexical item) rather than the other for talking about a particular realworld situation (such as coming through the door). In contrast with the 'formal' kind of onomasiological variation, we will call this type the' categorial' type of onomasiological variation, because it involves the choice of a different category. To be sure, the existence of such a preferred categorial choice is not an entirely novel idea, as it is part of the foundation of the basic level model of taxonomical structure formulated by Berlin and colleagues (see Berlin 1978). There are, however, various points with regard to which the basic
Salience phenomena in the lexicon
85
level model of onomasiological salience may be refined. At least the following two points may be considered. To begin with, the basic level model attaches onomasiological salience to hierarchical levels in the taxonomical tree, rather than to individual words. However, it may be argued (Geeraerts, Grondelaers, and Bakema 1994) that the notion of onomasiological salience should not be restricted to levels, but should be extended - generalized, if one wishes - to individual categories. It may not be necessary to repeat all the arguments here, but Table 2 gives a sample of relevant observations. Table 2. Differences in onomasiological salience among co-hyponyms ITEM
ONOMASIOLOGICAL SALIENCE
broek short-shorts bermuda legging-Ieggings-caler;on jeans-jeansbroek-spijkerbroek
46.47 45.61 50.88 45.50 81.66
t-shirt blouse-bloeze-bloes overhemd topje shirt hemd overhemdblouse
70.61 61.52 31.45 29.62 29.06 22.31 12.74
The data in the figure are based on the following operational definition of onomasiological salience: the onomasiological salience of a lexical category is the frequency of the lexical element naming the category divided by the cumulative token frequency in the database of the semantic values expressed by that lexical element. The rationale behind this definition is as follows. For each semantic value (concept, or type of referents) expressed by a lexical element. there may exist alternative terms (co-hyponyms, hyponyms, hyperonyms). A lexical category is onomasiologically highly salient if it is a likely choice for the semantic values it expresses, that is, if it is stronger than the alternatives. Thus, given a corpus of language use, the onomasiological salience of an item like 'skirt' can be calculated by counting how many times skirts (i.e. the potential referents of 'skirt') are named in the corpus, and then checking how many times these are actually referred to with the lexeme
86
Prototypicality and salience
'skirt', rather than alternative ones. Such alternative terms could be hyponyms like 'miniskirt' or 'wrap-around skirt', or hyperonyms like 'garment' (although the choice of the latter term seems quite unlikely in the case of 'skirt'). The definition of onomasiological salience also takes cases of synonymy into account, but this aspect of the matter need not be spelled out here. The point about Table 2, then, is simply this: the onomasiological salience of different categories on the same taxonomicallevel may differ considerably. For instance, in the upper part of the figure, 'short', 'bermuda', 'legging', and 'jeans' are co-hyponyms, as they all fall under the hyperonymous category broek 'two-legged garment covering the lower part of the body, pants'. However, the onomasiological salience of the different categories differs considerably: that of the jeans-category, for instance, doubles that of the legging-category. This means that a potential member of the category jeans is twice as likely to be designated by an expression that names the category jeans than a member of the legging-category would be likely to be designated by an expression that names the category legging. The second point about the conceptual kind of onomasiological salience involves the recognition that the referential starting-point for the determination of onomasiological salience values is subject to variation. The salience figures given so far represent the likelihood that one specific category, such as legging or t-shirt will be chosen as the name for a particular referent, in those cases where that referent is a potential member of the category. The calculation, to be precise, is made for the category as a whole, that is, for the overall set of referents of the category. However, the calculation may also take its starting-point in a particular subset of the entire range of application of a lexical category. The question asked so far boils down to questions of the type: what is the probability for the set of leggings as a whole that its members will be named with an expression that precisely identifies the category 'legging'? The modified question now reads: given a specific subset of the set of leggings as a whole (for instance, leggings that do not reach down to the ankles but are not longer than the calves), what is the likelihood that it will be named with the item legging or one of its synonyms? Obviously, the answer to the question will be determined by the overall salience of the category legging as calculated a moment ago, but it may also depend on the structural position of the subset under investigation within the semasiological structure of legging. The 'local' onomasiological salience introduced here may be called the 'cue validity' of a particular subset of the range of application of an item: it indicates how good a cue that subset is for
Salience phenomena in the lexicon
87
triggering the name of the category. The suggestion, then, is clearly that this onomasiological cue validity and the semasiological properties of the subset exhibit a positive correlation. Put more simply: an expression will be used more often for naming a particular referent when that referent is a member of the prototypical core of that expression's range of application. When, for instance, a particular referent belongs to the core of item x but to the periphery of y, it is to be expected that x will be a more likely name for that referent than y. (At the same time, of course, the choice for x will be a function of the global salience of the category represented by x, but that is a factor we have already dealt with). A brief and rather impressionistic example may suffice to illustrate the idea (for more detailed, quantitatively elaborated examples, see again Geeraerts, Grondelaers, and Bakema 1994). The kinds of usage that are structurally peripheral in the semasiological structure of Dutch over are more readily replaced by alternative ternlS than the central cases. For instance, in the examples illustrating reading (c) of over, the expression aan de andere kant van 'to the other side of is more natural. In the same way, there is an extension of (c), defined and illustrated below sub (d), for which the alternative expression tegenover 'across, facing' is by far the more natural one. (d)
Situated at the end of a virtual path that extends from lm in a straight line across an open space
(13)
]v/artha woont over het museum
(14)
Aan tarel zat Marianne over Ariane
(15)
Het museum bevindt zich over de rechtbank
(16)
In de erehaag stand Samuel over David
'Martha lives across the museum' .At the table, Marianne sat facing Ariane' 'The museum is situated across the magistrates' court' 'In the double row of honor, Samuel stood facing David' The notion of local onomasiological salience may be subject to two different interpretations, involving the same distinction between polysemy and vagueness that was discussed above in cOlmection with the concept of prototypicality. The referential subset from the range of application of a lexical item that is taken as the starting-point for a calculation of local onomasiological salience mayor may not correspond to a meaning, in the traditional sense, of that item. There is no reason to restrict the investigation of onomasiological salience to subsets that actually correspond with 'meanings'. This is not only because (as mentioned before) the distinction between meanings (underlying polysemy) and mere referential subsets (representing vagueness)
88
PrototypicaliZv and salience
may well be unstable, but also because prototype-theoretical research has established beyond doubt that even referents that do not constitute a 'meaning' in the strict sense (for instance, central members of the category), may play an important role in the semantic structure of a lexical item. The meanings whose local onomasiological salience is calculated need not necessarily constitute denotational meanings: any lexically expressed semantic concept (whether it be of a denotational, emotive, stylistic, or discursive kind) can be subjected to the approach.
6. Structural salience We may now turn to the third major perspective distinguished above. Remember that a purely structural conception of onomasiological lexical relations involves the mutual relationship between lexical categories. Salience effects along this perspective, then, will have something to do with the weight of the distinctive dimensions that distinguish various categories from each other. A semantic feature or a semantic dimension will be structurally salient when it often occurs in the structure of the lexicon; the structure itself, in fact, coincides with the totality of distinctive relations in the lexicon. For an example, let us consider the lexical field of names for beer current in Belgium (see Geeraerts 1999a). The methodology can be summarized as follows: the more a particular feature is expressed in the names of beers, the more it is cognitively salient. When, for instance, the presence of fruit flavors never surfaces in the names of the beers that have such a flavor, it is unlikely that fruit flavor is a preponderant feature of beers. Of course, we have to be more precise about what it means for a distinctive feature or a distinctive dimension 'to be expressed in the names of beers'. In most cases, there is no problem: when you find a reference to kriek in a name, and when you know that kriek is the name for a type of cherry, there is no difficulty in concluding that the addition of cherry juice is mentioned in the name of the beer. In other cases, the reference is more indirect. A striking example is the following: in the French speaking part of Belgium, a peach-flavored beer is produced with the name La pecheresse, which translates as 'the female sinner, the sinning woman'. When you realize, however, that peche is the French for 'peach', it will be clear that the name contains a pun with an indirect reference to the presence of peach flavor. In general, a feature is expressed by a name (or part of a name) when that name (or the relevant part of it) only occurs in connection with that feature.
Salience phenomena in the lexicon
89
Some more examples may further illustrate the point. The fact that a particular beer is low on alcohol may be expressed directly by expressions like alcoholarm 'low on alcohol' or alcoholvrij 'without alcohol'. Indirectly, however, the item talelbier 'table beer' also signals the light character of beer, to the extent that it only occurs within the group of very light beers. Similarly, lambiek and geuze can be taken as expressions of the 'spontaneous fermentation' character of beers, to the extent that they only occur in names for beers with that characteristic. Table 3. Salience characteristics of sample referential features in the field of Belgian beer names REFERENTIAL FEATURE
REFERENTIAL FREQUENCY
CHARACTERISTIC NAMES
wheat (instead of rye) very light low alcohol seasonal beers added herbs raspberry flavor cherry flavor trappist beer
139 110 86 59 40 19 16
91 (65.5%) 25 (22.7%) 86 (100%) 4 (6.8%) 38 (95%) 18 (94.7%) 16 (100%)
Following these guidelines, we can show that not all features of beers are equally important. There are clear differences in the extent to which specific characteristics are expressed in the names. Table 3 presents a sample to demonstrate the kind of differences in salience that occur. The table should be read as follows. The first column of figures indicates the number of beer brands in the data set that has the feature mentioned to the left. The rightmost column indicates how many of those beers actually carry a name that refers directly or indirectly to the feature in question. The relationship among the figures mentioned to the right gives an indication of which features carry distinctive weight. Trappist origin, for instance, seems to be highly valued, while the addition of herbs is not an especially individuating feature. In the spatial realm, the featural salience that is at stake here would involve, for instance, the distinctive weight of dimensions like shape, or movement, or geometry, or the role of the observer across items (or even across languages). (See Taylor 1988 for an overview of potentially relevant dimensions).
Prototypicali~y and salience
90
7. A typology of lexicological salience
In the following overview, each of the seven major cases of lexicological salience as presented before is defined. Also, a shorter name is suggested for each of the various types; of course, these names are just suggestions that should be treated with care as long as there is no uniform lexicological terminology. 1. Perspectival salience: highlighting = the differences of perspectival attention attached to different parts of the overall chunk of extralinguistic reality evoked by a particular concept. 2. Paradigmatic semasiological salience: proto~ypicality = the preponderant structural weight of specific senses or members within the semasiological range of application of a lexical category 3. Syntagmatic semasiological salience: canonicality = the dominant occurrence of specific selection-restrictional frames within the syntagmatic range of a paradigmatic reading of a lexical category 4. Formal onomasiological salience: sociolinguistic prevalence = the preference for a lexical item, in comparison with its synonyms, in a specific language variety or a specific pragmatic context 5. Global categorial onomasiological salience: entrenchment = the preference for a specific lexical category as a designation for its range of application, taken as a whole 6. Local categorial onomasiological salience: cue validity = the preference for a specific lexical category as a designation for a subset of its range of application 7. Structural salience: distinctiveness the preponderant recurrence of a semantic dimension or a semantic feature as a distinctive dimension or feature within the semantic structure of the lexicon
=
By way of summary, the seven types of lexicological salience are brought together in their mutual taxonomical relations in Figure 2. At this point, we may try to chart the seven types against the background of the semiotic triangle used in Section 3. However, the inadequacy of the semiotic triangle for dealing with more complex linguistic relations is wellknown; see Hofinann (1993) for a brief history of the matter, and compare Geeraerts (1983b).
Salience phenomena in the lexicon
91
lexicological salience variational salience
perspectival salience semasiological salience
~
paradigmatic
syntagmatic
structural salience
onomasiological salience
~
formal
categorial
global
~ local
Figure 2. Taxonomical relations among the various types of lexicological salience
An attempt (but certainly no more than attempt) to graphically represent the six variational types of salience may be based on Figure 3. In Figure 3, three levels of analysis are distinguished: the level of words W, the conceptual level of meanings C of those words, and the level of the referents R corresponding to those meanings. The word W2 is polysemous; Wl and W2 are synonymous with regard to meaning Cl. The referential range of words may overlap in yet another case besides synonymy: R4 may be conceptualized either as C3 or as C4 (as when a short wrap-around skirt might be called either a 'miniskirt' or a 'wrap-around skirt'). The referential range of CS is represented in a special way to highlight the syntagmatic contexts S in which the referents of CS occur. The horizontal arrows represent the features that distinguish between the concepts C. Salience effects may now be represented by using bolder lines; the differences between onomasiological and semasiological salience are expressed by having bottom-up and topdown arrows respectively. In Figures 7 to 12, the six types of variational salience are represented graphically against the background of Figure 3. (The parts of the figure that are irrelevant are represented by dotted lines.) Figure 4 indicates that C2 (rather than Cl or C3) is the prototypical meaning of W2, and that R3 (rather than R2 or R4) is the prototypical type of referent of C2. Figure S shows that R-Sl is the canonical syntagmatic context for CS. Figure 6
92
Prototypicality and salience
represents a sociostylistic context in which W2 is the prevalent synonym for Cl. Figure 7 indicates that W4 is a more highly entrenched category than W 1, in the sense that W4 is more likely to be used (in those cases where it can in principle be used) than Wl is likely to be used (in those cases where it is a relevant category). Figure 7 does not contain cases in which Wl and W4 overlap, but obviously a comparison of entrenchment values could just as well be applied to such a situation. Figure 8 expresses that R5 has a higher cue validity with regard to C4 than with regard to C3, that is to say, W3 is more likely to be used to categorize R5 than W2. Figure 9 indicates that the dimensions or features that distinguish C3 from C4 are structurally more important than the others (for instance, because they recur elsewhere in the lexicon).
Wl
W3
W2
Rl
I
I
I Cl
W4
_----_~
R2
C2
R3
.4--_~
R4
C3 _ _ C4 _
C5
V/\ R-Sl
R5
R-S2
Figure 3. Sample network for representing salience phenomena
Wl
W2
Cl
C2
Rl
R2
R3
C3
R4
R5
W3
W4
C4
C5
R-Sl
R-S2
Figure 4. Schematic representation of paradigmatic semasiological salience
Salience phenomena in the lexicon
Wl
W2
Cl
C2
Rl
R2
R3
C3
R4
W3
W4
C4
C5
93
1\
R5
R-S2
R-Sl
Figure 5. Schematic representation of syntagmatic semasiological salience
Wl
W2
W3
W4
C4
C5
1________ C2
Cl
Rl
R2
R3
C3
R4
R5
R-Sl
R-S2
Figure 6. Schematic representation of formal onomasiological salience
Wl
W2
W3
W4
C4
C5
i
I.······· Cl
................
C2
. .... C3
...
1
Rl
R2
R3
R4
R5
1\
R-Sl
R-S2
Figure 7. Schematic representation of global categorial onomasiological salience
94
Prototypicality and salience
Wl
W3
W2
.~ Cl
Rl
C2
R2
R3
,
...........
R4
W4
t
C4
C3
\/ R5
C5
R-S2
R-Sl
Figure 8. Schematic representation of local categorial onomasiological salience
Wl
Cl
Rl
W3
W2
W4
_ - - - -__ C2 . 4 - -__ C3 ++C4 _ _ C5
R2
R3
R4
R5
R-Sl
R-S2
Figure 9. Schematic representation of structural salience
8. Salience and snow The classification presented in the previous pages is undoubtedly only a first step towards an exhaustive and systematic treatment of lexicological salience. This first step was taken for the purpose of clarification, but it may be interesting to round off by adopting a slightly different perspective: could the classification throw new light on other linguistic phenomena? As one example, let us consider the question of conceptual universals. If semantic universals are likely to be statistical universals rather than absolute ones, the notion of semantic salience will be crucial in establishing the statistical universality of a particular phenomenon: semantic phenomena that are cognitively basic are likely to be salient across genetically unrelated languages. Questions of cultural specificity (the logical counterpart of questions of universality) likewise involve salience phenomena. Consider, to address one perennial member of the lexicological stock, the alleged existence of various words for snow in Eskimo. Let us suppose for a moment that there
Salience phenomena in the lexicon
95
are no empirical problems with Boas's statement (1911) that the Eskimo have four words for snow, whereas English and Aztec have only one (but see Martin 1986). Boas mentions the following four: aput 'snow on the ground' gana 'falling snow' piqsirpoq 'scurrying snow' qimuqsuq 'snowstorm, blizzard'.
It has been repeatedly pointed out that this does not imply that people camlOt notice differences in snow merely because they speak English. Clark and Clark (1977: 555), for instance, draw attention to the fact that Englishspeaking expert skiers may have specific words for snow such as 'powder', 'corn', and 'ice'. More generally, there is always the possibility of ad hoc compounds and word groups: 'snow glue', 'drizzle snow', 'icicle flakes', or the translations of the Eskimo words given above ('falling snow' etc.). The possibilities of expression of a language are not restricted to the vocabulary, and neither are the possibilities of conceptualization of the speakers of the language restricted to what is included in the mental lexicon. This attenuated point of view, however, does not imply that the cultural differences between the Eskimo and the Aztec disappear altogether. The cognitive salience of a concept like 'falling snow' may well be higher for the Eskimo than for the Aztec: it is plausible that they will more easily and more readily conceptualize a particular white substance as 'falling snow' rather than just 'snow'. In the terms of the classification presented in this paper, this expectation can be translated into the hypothesis that the entrenchment (the global categorial onomasiological salience) of 'falling snow' will be higher for the Eskimo than for the Aztec. The relationship between salience and universals may be specified from yet another point of view. In her work on semantic universals, Wierzbicka (1992) demonstrates that such abstract concepts as emotions and moral notions may be of a culturally specific nature~ at the same time, she insists on the possibility of expressing such concepts in any given language by means of a language-independent metalanguage: no matter how culture-specific a term might be, it can be translated in the 'alphabet of human thoughts' that she has been researching for decades. The foregoing pages, then, suggest one way of looking for such an 'alphabet of human thought' that might usefully supplement Wierzbicka' s analytic approach: if these concepts really have the universal status attributed to them, they should in one form or another have a high degree of lexicological salience in a large number of natural languages.
96
Proto~ypicali~v and
salience
The universality question also invokes the problem of pre-linguistic forms of salience, namely the objective correlates of the linguistic phenomena. To what extent is linguistic categorization determined by the structure of the world itself, or at least by the structure of our perceptual apparatus? Ifnaming is often grounded in an act of perception, how do the independent properties of such perceptual acts influence the linguistic phenomena? The typology presented here has steered clear of these perceptual forms of salience, but it certainly does not mean to exclude the possibility of a pre-linguistic level of salience. Rather, a cognitive approach to the study of language will inevitably have to consider the possible impact of such a pre-linguistic type of salience. Tackling that question, however, is beyond the scope of this article: before one starts looking for explanations of the linguistic phenomena, one has to be clear about those phenomena themselves.
Section 2 Polysemy
Chapter 5 Vagueness's puzzles, polysemy's vagaries
Originally published in Cognitive Linguistics 1993,4: 223-272. Polysemy is the basic phenomenon in lexical meaning structures, and arguably the basic contribution of Cognitive Linguistics to lexicology is the renewed attention for the structure of polysemy that it generated. But how exactly should polysemy be defined? What does it mean for a reading of a word to be a different meaning? In this paper, I argue that the distinction between vagueness and polysemy (which is crucial to traditional conceptions oflexical-semantic structure) is unstable: what appear to be distinct meanings from one point of view turn out to be instances of vagueness from another. In the present chapter, an argumentation for this thesis is developed in three steps. First, three basic kinds of polysemy criteria are distinguished: a logical, a linguistic, and a definitional test. Second, it is shown that these three kinds of tests may yield mutually contradictory results. Third, it is argued that such inconsistencies also occur within each criterion, in the sense that there are contextual variations in the polysemy results that they yield. As a consequence, Cognitive Linguistics will have to start from a highly flexible conception of meaning. In the period after the publication of the original article, polysemy research has moved in various directions relevant to the issues addressed in this chapter. First, Cognitive Linguists have explored ways for incorporating the flexibility and instability of meaning into the models of semantic structure that they work with. Levels of schematicity in the Langackerian sense (i.e. the idea that the meaning of a linguistic unit can be described as a schematic network in which more specific and more abstract levels of description co-occur) play a crucial role in such attempts: see Tuggy (1993), Dunbar (2001), Taylor (2003), Zlatev (2003). Second, attempts have been made to link polysemy research in Cognitive Linguistics with psycholinguistic studies of polysemy (Sandra and Rice 1995), but this has somewhat surprisingly, given the very cognitive orientation of Cognitive Linguistics - not developed into a dominant trend; see the discussion between Croft (1998), Sandra (1998), and Tuggy (1999). Third, the interest in polysemy issues has spread outside the field of Cognitive Linguistics: see Ravin and Leacock (2000), Peeters (2000) and Nerlich, Todd, Hennan and Clarke (2003) for samples. In particular, a major new trend in polysemy research has arisen in the domain of computational linguistics, in the form of so-called Word Sense Disambiguation algorithms: computational and statistical analyses of large corpora that take the
100
Polysemy
co-occurrence behavior of lexical items as indications of their semantic range (see Stevenson 2001, Agirre and Edmonds 2006 for overviews). Together with detailed corpus-based studies as will be illustrated in the next chapter, psycholinguistic and computational studies of the type just mentioned will, I think, be of crucial importance for the further development of polysemy research in Cognitive Linguistics: the theoretical issues have been greatly clarified, but there is still plenty of room for methodological progress.
1.
Prototypicality and polysemy
In my paper 'On necessary and sufficient conditions' (1988a), I pointed out that there is an intimate connection between the proper evaluation of the novelty of the prototype-oriented movement in lexical semantics, and the long-standing issue how to determine whether a word has one or more senses. In fact, while the revolutionary nature of prototype theory is often said to reside in the fact that it rejects the classical view that lexical categories can be defined by means of a sufficient set of necessary conditions for category membership, the possibility that some lexical items cannot receive a definition in terms of necessary-and-sufficient attributes is not exactly absent from the classical tradition. Rather, it is traditionally taken as an indication of the polysemy of the item in question. Therefore, if prototype theory is to maintain its revolutionary claims, it will have to redefine them, or at least make clear that the absence of necessary and sufficient definitions that it reveals does not simply signal good old-fashioned polysemy. In his recent monograph La semantique du prototype (1990), Georges Kleiber has taken up this suggestion, incorporating it into a discussion of the various trends that exist within linguistic prototype theory. However, as I already briefly noted in my review of Kleiber's thought-provoking book (1992), his treatment of the consequences of polysemy for prototypical theorizing seems to assume that the concept of polysemy is methodologically unproblematic; at least, his book does not deal explicitly with the question how to identifY the polysemous readings of a lexical item. The present paper is devoted precisely to that issue. It will compare various criteria for distinguishing between vagueness and polysemy, and explore the consequences of this comparison both for Kleiber's views and for prototype-oriented lexical semantics at large. In particular, it will point out two sets of problems for the usual polysemy criteria: first, the various types of criterion will appear to be mutually inconsistent, and second, any single criterion yields internally inconsistent results. That is to say, I will try to show that the distinction
Vagueness's puzzles, polysemy's vagaries
101
between vagueness and polysemy is not stable, in the sense that what ap-
pears to be a distinct meaning in one context is reduced to a mere case of vagueness in another. The paper is organized as follows. First, the importance of the polysemy issue for prototype-oriented theories is discussed in more detail. Second, a classification of types of criteria for distinguishing between vagueness and polysemy is presented. As a third step, the problems with the criteria will be presented, first by comparing the criteria among each other, and second by discussing each criterion separately. In contradistinction with existing treatments (such as Zwicky and Sadock 1975 and Cruse 1982), the emphasis will not just lie on the evaluation of each criterion in isolation, but also on their mutual incompatibilities; another point of contrast with existing work is that the distinction between vagueness and polysemy will not simply be assumed, but will itself turn out to be somewhat doubtful. After a complementary discussion of some additional problems, the consequences of these findings will be explored; it will be argued that the instability of the distinction between vagueness and polysemy not only suggests a reorientation of our conception of lexical meaning, but also raises important questions about the objectivity of the methodology of lexical semantics. Although the paper will throughout assume some familiarity with the concepts of linguistic prototype theory (such as they may be found in Taylor 1989), the paper will proceed at a relatively slow pace. The reduced speed is, to be sure, not a sign of leisure, but rather reflects the difficult and often uncharted character of the territory that we will be venturing into.
2.
Intercategorial and intracategorial prototypicality
In the course of the development of prototype studies in linguistics, various definitions of prototypicality have been suggested. Apart from the absence of necessary and sufficient definitions, prototypicality has been said to involve (among others) categorization on the basis of the salient members of a category, vague category boundaries, and a radial set relationship among the various applications of a category. (Compare Lakoff 1987 and Geeraerts 1989c for attempts to chart the different kinds of prototypicality.) It is now generally accepted that these characteristics need not coincide; they rather represent prototypicality effects that a single category may exhibit in various degrees, and that may arise from various sources. Precisely because the various prototypicality effects need not be co-extensive, some of them might
102
Polysemy
be typical of single meanings (the separate senses of a lexical item), whereas others could be typical of the meaning of lexical items as a whole (i.e. of structured sets of senses). Specifically, the difficulty noted in the introductory paragraph may be resolved by making clear that the absence of definitions characterized by necessity-cum-sufficiency is in fact a true novelty when it applies to the single senses of a lexical item (whereas it would only be a traditional indication of polysemy when it applies to the lexical item as whole). For instance, if it turns out to be impossible to define bird by means of a single sufficient set of necessary attributes, this is hardly surprising if the biological sense of bird is taken together with a metaphorical reading as in A silver bird took the happy couple from Chicago to Santa Cruz. However, if the biological sense taken by itself (given that it is a single, separate sense of bird) cannot be defined in the traditional way, this is a result contradicting traditional beliefs about the definability of word senses. Kleiber (1990) has pushed this kind of reasoning one step further by bringing other prototypicality effects into the picture. If we distinguish between the lexical level (consisting of lexical items as forms, i.e. of signifiants), the conceptual level (consisting of cognitive categories that are at the same time the senses oflexical items), and the referential level (consisting of real- or possible-world entities), various types of prototypicality effects can be contrasted. On the one hand (considering the relation between the lexical and the conceptual level), there is prototypicality in the radial sets or family resemblance sense, defined as a relationship among the various senses of a lexical item (each of those senses being a category on the conceptual level). On the other hand (considering the relation between the conceptual and the referential level), there is prototypicality in the original sense of Rosch's psycholinguistic research, involving typicality ratings and degrees of membership (and also, in the sense of an absence of classical definitions). Here, the internal structure of one of the distinct meanings of a lexical item is involved (like the biological sense of bird, in contrast with a metaphorical reading as when an airplane is called a silver bird). Prototypicality is, then, either a model for the polysemy of lexical items (involving the relationship between the lexical and the conceptual level), or a model of categorization (involving the relationship between the conceptual and the referential level). In the latter case, the prototypicality effects are intracategorial; in the former, they are intercategorial. I This model of prototypicality (which is as attractive as it is simple) potentially considerably diminishes the importance of linguistic prototype theory. On the one hand, it could be argued that the 'categorization' relation-
Vagueness's puzzles, polysemy's vagaries
103
ship between the conceptual and the referential level is the proper domain of psychology rather than linguistics; linguistics is primarily concerned (in the structuralist line of thought) with the relationship between signifi£s and signifiants, rather than with the relationship between language and the world. On the other hand, the importance of the 'polysemy' relationship between the lexical and the conceptual level is only a relative novelty in the history of linguistics; both the distinction between central and peripheral meanings, and the idea that those meanings are related by motivated links (like metaphor and metonymy) are far from new in the history of lexical semantics (see Geeraerts 1988a). However, the second part of this interpretation is hardly disconcerting. Even if prototype theory as a model of the polysemy of lexical items is a rediscovery of older concerns rather than a complete novelty, it is important as a rediscovery. In the article that I just referred to, I have argued that prototype theory links up primarily with prestructuralist historical semantics. Granting, then, that the structuralist approach in lexical semantics has (in its mainstream development) downplayed the importance of polysemy, the revivification of the polysemic concerns of prestructuralist semantics is an important step towards a more adequate conception of lexical-semantic structure. Also, there can be no doubt that prototype theory is more than a rediscovery of older ideas; the importance of typical individual members of a category for the structure of that category is, for instance, hardly to be found in the prestructuralist conceptions of lexical structure. The other point is potentially more damaging. If the' categorization' relationship as defined above does not belong to the proper domain of linguistics, the claim of Cognitive Linguistics that it is a full-fledged theory of categorization will be undermined, because it would have to leave that aspect of cognitive structure in which the relationship between the conceptual level and the referential level is studied, to psychology. That is to say, either Cognitive Linguistics would indeed be studying the 'categorization' relationship together with the 'polysemy' relationship (but then it would be trespassing on psychological territory), or it would forsake the study of 'categorization' (but then its cognitive importance would largely evaporate, because the relationship between the referential level and the conceptual level is the basic one in the formation of cognitive categories). This argumentation rests, among other things, on the debatable assumption that existing disciplinary demarcations should be maintained. The fact that linguistics and psychology now exist as separate disciplines should not preclude the formation of an interdisciplinary cognitive science in which
I04
Po~ysemy
intercategorial and intracategorial prototypicality effects are studied together. But then again, independent arguments should be adduced for explaining why they should be studied together. It may well be interdisciplinarily legitimate in principle to study them together, but what is the positive motivation for actually doing so? The most drastic reply possible would be to challenge the distinction between intercategorial and intracategorial polysemy itself. Methodologically speaking, the model that we have derived from Kleiber's views requires a coherent way of establishing polysemy, i.e. of making a distinction between referential differences that do and those that do not correspond with conceptual differences. If, however, what is merely a conceptually neutralized referential distinction at one point turns into a case of polysemy at another, the categorial structure is not stable, and it would be natural (nay, unavoidable) to study intracategorial prototypicality together with intercategorial prototypicality. In fact, the very distinction between both types would have to be rejected. What is at stake, in other words, is not just the classification of prototypicality effects into intercategorial and intracategorial ones, but also the unity of prototype theory, that is to say, the legitimacy and the relevance of an interdisciplinary study of cognitive structures, in which the 'linguistic' study of polysemy and the 'psychological' study of categorization are inescapably intertwined. Proponents of a prototype-oriented perspective in lexical semantics therefore have every reason to have a closer look at the concept of polysemy.
3. Criteria for polysemy - A classification The distinction between vagueness and polysemy involves the question whether a particular piece of semantic information is part of the underlying semantic structure of the item, or is the result of a contextual (and hence pragmatic) specification. For instance, neighbor is not polysemous between the readings 'male dweller next door' and 'female dweller next door', in the sense that the utterance my neighbor is a civil servant will not be recognized as requiring disambiguation in the way that she is a plain girl ('ugly' or 'unsophisticated'?) does. The semantic information that is associated with the item neighbor in the lexicon does not, in other words, contain a specification regarding sex; neighbor is vague (or general, or unspecified) as to the dimension of sex.
Vagueness's puzzles, pozvsemy's vagaries
105
The existence of altemative terms (see Norrick 1981: Ill) for the phenomenon that we are concemed with is related to the fact that the term vagueness has various applications that do not coincide with the kind of vagueness meant here. Specifically, it is important to distinguish the categorical lack of specification illustrated above from the referential indeterminacy (or 'vagueness') that may characterize the individual members of a category or the category as a whole. The former, individual type of referential indeterminacy is illustrated by knee: it is impossible to indicate precisely where the knee ends and the rest of the leg begins. The class type is illustrated by any color term: it is impossible to draw a line within the spectrum between those hues that are a member of the category red and those that are not. In order to bring some structure into this set of interpretations of the notion 'vagueness', it should be noted that they involve indeterminacy on three distinct, hierarchically related levels. The knee example is situated on the level of individual entities; the question where a particular entity (such as a knee) ends is specified as the question whether a potential part of the entity (say, a specific area of the leg) is part of the entity or not. The red example is situated on the level of groups of entities, i.e. conceptual categories; the question conceming the boundaries of a particular category is specified as the question whether a particular entity (such as a particular hue) is part of the category or not. The neighbor example of vagueness, finally, is situated on the level of polysemous lexical items, considered as groups of related conceptual categories; the question where exactly the boundaries of a specific polysemous item lie is specified as the question whether a potential conceptual category (such 'male person living next door') is indeed a member of the set of meanings of a particular item (such as neighbor). While we concentrate here on the third kind of vagueness, it is mainly the second kind that has received a lot of attention in the recent literature on the subject of vagueness. 2,3 The difficulties created by the terminological divergences are enhanced by the existence of various kinds of tests for distinguishing between vagueness and polysemy. A definitive treatment of this subject does not yet exist; despite its importance for theories of semantic structure, it has received relatively little systematic and continuing attention. Without discussing all specific tests that have been suggested, three types of criteria will be distinguished here. (The most comprehensive treatment of tests for lexical polysemy that I am aware of is to be found in eruse 1982, which will furnish a basic reference point for the following discussion 4 .5 Other relevant works will be mentioned in passing in the course of the following pages.)
106
Polysemy
First, from the truth-theoretical point of view taken by Quine (1960: 129), a lexical item is polysemous if it can simultaneously be true and false of the same referent. Considering the readings 'harbor' and 'fortified sweet wine from Portugal' of port, the polysemy of that item is established by sentences such as Sandeman is a port (in a bottle), but not a port (with ships). Or, in the words of Quine: 'An ambiguous term such as light may be at once clearly true of various objects (such as dark feathers) and clearly false of them' (1960: 129). Second, linguistic tests involve semantic restrictions on sentences that contain two related occurrences of the item under consideration (one of which may be implicit or deep-structural); if the granmlatical relationship between both occurrences requires their semantic identity, the readings that can be attributed to the resulting sentence may be an indication for the polysemy of the item. For instance, the identity test described by Zwicky and Sadock (1975) applies to constructions that were assumed in the Chomskyan 'standard theory' to involve transfomlations such as conjunction reduction and so-reduction, which require the semantic identity of the items involved in the reduction. (It may be noted that constructions such as these are nowadays no longer discussed in transformational terms. However, as the current term 'identity-of-sense anaphora' indicates, the idea that there are semantic restrictions on the construction remains intact. Note also that the type of test mentioned here did not originate with Zwicky and Sadock; in particular, see Lakoff 1970.) Thus, at midnight the ship passed the port, and so did the bartender is awkward if the two lexical meanings of port are at stake; disregarding puns, it can only mean that the ship and the bartender alike passed the harbor (or, perhaps, that both moved a particular kind of wine from one place to another). A 'crossed' reading in which the first occurrence of port refers to the harbor, and the second to wine, is normally excluded. Conversely, the fact that the notions 'vintage sweet wine from Portugal' and 'blended sweet wine from Portugal' can be crossed in Vintage Naval is a port, and so is blended Sandeman indicates that port is vague rather than polysemous with regard to the distinction between blended and vintage wmes. Third, the definitional criterion (as infomlally stated by Aristotle in the Posterior Analytics II.xiii) says that an item has more than one lexical meaning if there is no minimally specific definition covering the extension of the item as a whole, and that it has no more lexical meanings than there are maximally general definitions necessary to describe its extension. Definitions of lexical items should be maximally general in the sense that they should
Vagueness's puzzles,
po~vsemy 's
vagaries
107
cover as large a subset of the extension of an item as possible. Thus, separate definitions for 'blended sweet fortified wine from Portugal' and 'vintage sweet fortified wine from Portugal' could not be considered definitions of lexical meanings, because they can be brought together under the definition 'sweet fortified wine from Portugal'. On the other hand, definitions should be minimally specific in the sense that they should be sufficient to distinguish the item from other non-synonymous items. A maximally general definition covering both port 'harbor' and port 'kind of wine' under the definition 'thing, entity' is excluded because it does not capture the specificity of port as distinct from other items.
4.
Criteria for polysemy - Remarks and refinements
There are three important (sets of) remarks to be made with regard to this classification of types of criteria for distinguishing between polysemy and vagueness. First, a number of tests have to be briefly mentioned that have not been included in the classification because their insufficiency is generally accepted. Second, the basic types enumerated above may be supplemented with variants (a possibility that will turn out to be of particular importance for the logical test). And third, there are some remarks about the theoretical background of the tests (and in particular, the definitional test) that are worthwhile mentioning. 4.1. Omissions For the sake of completeness, let it be noted that a number of tests whose insufficiency or impracticability has been sufficiently established elsewhere have not been included in the classification presented above. In particular, these involve the 'indirect' criteria distinguished by Cruse (1982), which refer to the relations that an item may contract with other items in the language (or in other languages, such as when the existence of different translations for a lexical item is taken as proof for the polysemy of that item). (The criteria discussed and rejected by Herringer 1981: 117-118 can be included in the same class.) As an illustration, consider the view according to which the existence of distinct antonyms for specific readings of a lexical item serves as proof of its polysemous nature. In the case of light, the reading 'not having much weight' contrasts with heavy, and the reading 'having a bright color' contrasts with dark. Although, in other words, the antonymy
108
Polysemy
test seems to work for light, there are various counter-examples. On the one hand, there are items with identical antonyms for distinct readings, as in the case of old, which has new as its opposite both in the reading 'former' and in the reading 'advanced in age'. On the other hand, there are items with distinct antonyms for what are merely distinct referential specifications rather than distinct meanings. For instance, when the Dutch word vers 'fresh' is applied to bread, it contracts a relationship of anton;my with oud 'old' but also with oudbakken 'stale'; when vers applies to meat, the antonyms would be oud and adellijk 'high, gamy'. Intuitively and definitionally, however, the two applications of vers constitute the single reading 'recent, and as such, optimal for consumption'. Examples such as these lead eruse to the evaluation that the 'indirect' criteria are less satisfactory than the 'direct' ones, which focus directly on an item's behavior in various contexts 6 4.2. Variants A second major point concerning the classification of polysemy criteria involves the presence, within each type, of other tests than the ones mentioned above. The three types of criteria are meant precisely as types: classes of tests that have something crucial in conunon. If, then, it is relevant to group polysemy criteria into the three classes identified above, we should be able to place specific tests that have not yet been discussed into one of the three classes. Because this point is specifically important for the first criterion, let us now have a look at the distinct varieties of the truth-functional approach, trying to bring some order into the scattered suggestions made by other authors. The starting-point for distinguishing between subtypes of the truthfunctional test is the recognition that Quine' s proposal rests on two features: first, polysemy is signaled by truth-conditional differences between the two alleged meanings, and second, these truth-conditional differences are revealed by applying the two readings to a single situation. Truth-functional criteria for polysemy that do not strictly coincide with the Quinean test as defined above, can now be introduced by taking these two features as a guide-line; basically, they contain variations on those features as involving the questions 'What kind of truth-conditional differences are we looking for?', and 'How do we diagnose them?', respectively. With regard to the first feature, Quine himself already noted that next to the oppositeness of truth value implied by the 'p and not p' criterion, a third truth value 'nonsensical' or 'irrelevant' may have to be taken into account. Questions and chairs can both be hard, but it is difficult to find a situation in
Vagueness's puzzles,
po~vsel1lY 's
vagaries
109
which both readings of hard are relevant (but in which one would be applicable and the other not). Although Quine seems to suggest that such a situation could be treated as a case of vagueness rather than polysemy, it is much more in line with the essence of the truth-functional approach to meaning discrimination, to assume (with Norrick 1981: 113) that a single sentence such as This book is sad may signal the polysemy of sad. The fact that the reading 'evoking sadness' of sad may render the sentence true or false according to the circumstances, whereas the reading 'experiencing sadness' yields a nonsensical interpretation, implies that there is a crucial distinction in the truth-functional properties of This book is sad under both interpretations. Given that the distinction depends on a distinction in the meanings we attribute to sad, this should be as good evidence for the polysemy of sad as a straightforward 'p and not p' case. The truth-functional differences involved in the test should, in other words, be taken broadly so as to include the difference between 'either true or false' and 'nonsensical'. Because the reading 'this book is experiencing sadness' turns This book is sad into a semantically ill-formed construction, it is difficult to have a single well-formed sentence of the 'p and not p' type in which the relevant interpretation of sad and the nonsensical interpretation are combined. This means that (as far as the second feature of the basic Quinean test is concerned) the differences in truth-functional properties that we are looking for will not be signaled by the spontaneous acceptability of a sentence of the 'p and not p' type, but will involve an active consideration of the properties of potentially distinct readings. The 'p and not p' test does not initially require a lot of conscious thinking about different meanings: if it is spontaneously clear that a feather can be light and not light, the polysemy of light is probably firmly established, psychologically speaking; it springs to mind so easily that the non-contradictory nature of the 'p and not p' construction is automatically recognized. In this sense, the 'p and not p' test has the same attractiveness that grammaticality judgments have in syntax: the judgment is unmediated. In the case of This book is sad, however, there is a lot of conscious mediation and active introspection involved, mainly because one has to actively conceptualize the different potential readings of the item and evaluate their truth~conditional consequences. Once this more active and more mediated approach is accepted (and it will have to be accepted if it is correct that the 'p and not p' frame cannot be applied to cases involving a nonsensical reading), varieties of this active search for truth-conditional differences may be explored. I shall mention two, but I do not want to claim that these are the only possibilities. We shall see later (in section 6.2) that
110
Polysemy
the variants of the logical criterion presented here run into trouble, but for the moment, we shall merely present them. To begin with, if a non-contradictory use of the 'p and not p' frame indicates the presence of polysemy, so does a non-redundant use of the 'p and p' frame. If p has two meanings, it should be possible to assert 'p and p' without raising feelings of redundancy or awkwardness. That is to say, while the main 'p and not p' version of the logical test considers cases in which a sentence can be used in the same situation with opposite truth values, the 'p and p' variant considers cases in which p is used in the same situation in two readings with the same truth value. The obvious question then is how we can know whether p indeed has two distinct interpretations. Apart from the possibility of asserting 'p and p' without redundancy or awkwardness (a possibility that is seldom noted in the literature), there is the introspective psychological phenomenon that in the process of interpreting the sentences in question, our attention seems to be switching continuously between one reading and the other. Take the sentence Charles has changed his position, in which Charles is an official delegate at a political summit meeting (the example is Cruse's ; cp. Cruse 1982: 70). The readings involved can be paraphrased as 'Charles has moved to another place in the conference hall' and 'Charles has changed his opinion concerning the topic discussed at the conference'. Now, even in those cases where both readings are true, it is difficult to mentally conceive of them together, i.e. to fuse them into a single reading. Rather, at anyone time, we seem to be able to have only one of the readings at the same time in mind. In the words of Herringer (1981: 95): Wichtig ist also dieses 'entweder/oder', und das, was man das Kippen zwischen den Bedeutungen genannt hat: Es gibt keine fliessenden Dbergange zwischen den Bedeutungen, sondern einen Aspektwechsel, der sich pl6tzhch und oft iiberraschend einsteUt. [What is important, then, is this either/or situation, and what has been called the tilting between meanings: There is no fluent transition between the meanings, but an aspectual change that imposes itself suddenly and by surprise.] Herringer (1981) also discusses a second variant of the basic logical criterion: the truth-conditional differences between sentences can be revealed by taking into account the implications of potentially distinct meanings: if these are indeed distinct, they should have different implications. For instance, the French sentence in (1) contains the polysemous word louer, which may mean either 'to rent' or 'to praise'. In one reading, (1) entails (2), but in
Vagueness's puzzles,
po~vsemy's
vagaries
III
another, it entails (3). However, if we combine these facts into the single implication in (4), and then compare that sentence with (5), where an unsuspected case of vagueness is involved, it becomes clear that the format of the entailment as such does not always distinguish between vagueness and polysemy: the necessity to express the entailments of a sentence in disjunctive form does not sufficiently distinguish between both phenomena. (1) (2) (3) (4)
(5) (6) (7)
11 a loue son appartement. 11 s 'est oblige de payer son appartement. 'He has taken on the obligation to pay the apartment.' 11 a dit du bien de I 'appartement. 'He has spoken well of the apartment.' S'il a loue I 'appartement, il s 'est oblige de payer ou il en a dit du bien 'If he has loue the apartment, either he has taken on the obligation to pay for it, or he has spoken well of it. ' Si c 'est un enfant, c 'est un garr;on ou unefille. 'If it is a child, it is a boy or a girl. ' lfit is not a child, it is not a boy and it is not a girl. If they did not reach the bank in time, either they did not reach the money-bank in time, or they did not reach the river-bank in time.
This problem can be avoided, however, when negation is added to the test (cp. Kempson and Cormack 1981, Cruse 1982). When a vague predicate is negated, the specifications of the predicate are both negated as well. In (6), for instance, the distinction between male children and female children (an example of vagueness with regard to child) lies at the basis of a conjunction of negations. On the other hand, in the case of polysemy, the negation need not extend to both readings at the same time. Thus, in (7), the negated ambiguous readings occur as a disjunction. Notice, in this respect, that the disjunction is necessary if the 'p and not p' frame is to be applicable in the case of polysemy: if PI and P2 are the readings involved, and if the negated form 'not p' would automatically entail both 'not PI' and 'not P2', an interpretation of 'p and not p' as 'PI and not P2' or 'P2 and not PI' would be excluded. 4.3. Specifications To round off the discussion of the classification of polysemy criteria, something more has to be said about the third criterion, if only because it is absent in the works we have been referring to. The criterion is of particular importance because it so to speak looks into the lexical items to determine
112
Polysemy
how the various meanings of the item are constituted. Whereas the other two criteria are at best diagnostic tools that signal the presence of polysemous readings, the third criterion is an analytical one that examines the internal constitution of the meanings of the item. In order to be better able to appreciate the impact of this definitional approach, let us have a closer look at Aristotle's description of it. We should look at what are similar and undifferentiated, and seek, first, what they all have that is the same: next, we should do this again for other things which are of the same genus as the first set and of the same species as one another but of a different species from those. And when we have grasped what all these have that is the same, and similarly for the others, then we must again inquire if what we have grasped have anything that is the same - until we come to a single account; for this will be the definition of the object. And if we come not to one but two or more accounts. it is clear that what we are seeking is not a single thing but several. I mean, e.g., if we were to seek what pride is we should inquire, in the case of some proud men we know, what one thing they all have as such. E.g. if Alcibiades is proud, and Achilles and Ajax, what one thing do they all have? Intolerance of insults; for one made war, one waxed wroth, and the other killed himself. Again in the case of others, e.g. Lysander and Socrates. WelL if here it is being indifferent to good and bad fortune, I take these two things and inquire what both indifference to fortune and not brooking dishonor have that is the same. And if there is nothing, then there will be two sorts of pride (Barnes 1984: 161). [An alternative, more literal reading for the Greek megalopsychia is 'greatness of the soul' rather than 'pride'.]
Granting that Aristotle does not seem to make a distinction between the definition of the word pride and the definition of the thing 'pride' (granting, in other and more traditional words, that his approach to definition is a realist rather than a nominalist one), the quotation helps to see the close connection between the definitional criterion of polysemy and the classical conception of definitions as sufficient sets of necessary criteria for class membership. The whole criterion could, in fact, be rendered in simplified form as the statement that an item is polysemous (or, alternatively, that what was thought to be a single category in fact consists of several categories) if there is no single set of necessary and sufficient attributes to describe the extension of the item or the alleged category. The full criterion does not, however, coincide entirely with this simplified version, because the latter only specifies the lower limit of an item's polysemy: if there is no single set of necessary and sufficient attributes that is minimally specific, the item has at least two meanings. But the definitional criterion of polysemy also sets an
Vagueness's puzzles, polvsel1ly 's vagaries
113
upper limit to the polysemy of an item, by specifying that an item has no more meanings than the number of maximally general definitions that are needed to cover its entire range of application. If, for instance, the two definitions 'indifference to fortune' and 'intolerance of dishonor' are sufficient to cover all examples of proud men that you know of, mega/opsychia does not represent more than two meanings. Notice that this 'upper limit' specification of the simple 'lower limit' rule that an item is polysemous if it cannot be defined by means of a single definition, is an entirely natural one: does not the very fact that we seem to be looking for single definitions first suggest that the crucial thing about definition is generality? In popular parlance, the definitional criterion might then read as follows: 'if you cannot achieve maximal generality, settle for the next degree of generality that you can attain'. After Aristotle, the definitional criterion for polysemy became a standard part of treatises on logic as the requirement that definitions should be neither too broad nor too narrow (see, for instance, Copi 1972: 138). The transition is not a straightforward one, however, to the extent that a requirement on definitions is not necessarily a criterion for polysemy: imposing restrictions on definitions would seem to imply that the definer already knows for certain what the definiendum is, i.e. what the lexical meanings are that are to be defined. However, if the definiendum is taken to be a lexical item rather than an independently discerned lexical meaning, the requirement actually turns into a criterion for the polysemy of the item. In that case, the requirement that a definition should not be too narrow embodies the principle that maximal generality should be aspired at, whereas the requirement that a definition should not be too broad reflects the necessity of minimal specificity as defined above. In actual practice, the fonnulations of the requirement that may be found in logical treatises waver between an interpretation of the target of the definition as a word and an interpretation as a word meaning. For instance, Copi's example ofa definition that is too narrow (1972: 138) is said to involve 'the word shoe', but practically, it seems that only the meaning 'outer foot-covering for a person' is at stake, and not (for instance) the meanings 'horseshoe' or 'thing that is like a shoe in shape or use (such as wheel-drags, sockets, ferrules etc.)'. It would seem, in other words, that the crucial distinction between both interpretations of the standard requirement on definitions is not always made. 7 It should be emphasized that the definitional polysemy test is particularly important in comparison with the other two: it is the only one that embodies a hypothesis about the principles of categorization that human beings em-
114
Polysemy
ploy. It suggests, in fact, that one of the basic characteristics of naturallanguage categorization is a tendency towards generality (or, in other words, towards maximal abstractions). Thus, while the other two classes of polysemy criteria merely contain diagnostic tools for establishing whether interpretations of lexical items constitute distinct meanings, the definitional test gives us an explanatory surplus, because it is based on an assumption about why particular senses are distinguished. It is all the more regrettable, then, that the definitional test has hardly received any explicit attention in the recent linguistic literature on the subject of lexical polysemy.
5. Problems with the polysemy tests - First series Given the existence of three major types of tests for lexical polysemy, we can now try to determine their mutual relationship. We will concentrate primarily on the tests that were earlier presented as the basic representative of each type, rather than bringing all possible variants of each type (as discussed in the previous paragraph) into the comparison. 5.1. Pairwise divergences among the criteria Just a few lexical items may suffice to show that the three tests, taken pairwise, may yield different results. First, consider an autohyponymous word such as dog (i.e. a word one of whose senses is a superordinate category with regard to one of the other senses). In the reading 'Canis familiaris', dog is a hyperonym of dog in the reading 'male Canis familiaris'; in the hyperonymous reading, dog contrasts with (for instance) cat, whereas the proper antonym is bitch in the hyponymous reading. But are these readings really distinct meanings? Applying the tests, it will be easily verified that dog is not polysemous according to the definitional criterion, but does exhibit polysemy according to the logical criterion. Taking Lady (from Disney's Lady and the Tramp) as a prime example of a female dog, it is possible to assert Lady is a dog alright. but she is not a dog. On the other hand, according to the definitional criterion, the hyponymous 'male Canis familiaris' reading is always a proper subset of the superordinate reading, so that no definition of that subset could ever claim the maximal generality that is required. Now, what about to the linguistic test? According to the linguistic test, a 'crossed' reading of sentences containing identity of sense anaphors (or similar constructions) should come out grammatically awkward. That is to say, interpreta-
Vagueness's puzzles, po~vsemy 's vagaries
115
tions of such a sentence in which the full occurrence of the item under consideration receives meaning x, and the anaphorical element receives meaning y, are ungrammatical. In the case of dog, however, it could be doubted whether such an ungrammaticality would ever surface as positive evidence in favor of the polysemy of dog, given that the anaphorical sentence would always automatically be interpreted as a grammatical construction involving just a single one of the two readings of dog. For instance (assuming that Fido is a male dog), Fido is a dog and so is Lady is spontaneously understood as the correct assertion 'Lady and Fido are both members of the species Canis familiaris', or conversely, as the false assertion that both animals are male members of that species. In neither case is the sentence spontaneously felt to be odd in the sense in which At midnight the ship passed the port, and so did the bartender is. Of course, it could be argued that the very fact that Fido is a dog and so is Lady can be either true or false (depending on the reading we attach to dog) in the same situation, testifies to the polysemy of dog. This is correct, but the invocation of truth-functional differences indicates that we are then dealing with a variant of the logical test, rather than with the linguistic grammaticality test that we are concerned with. Let us accept, therefore, that the linguistic test does not testify to the polysemy of dog. We will then have established a distinction between the definitional test and the linguistic test on the one hand, and the logical test on the other. Next, let us consider a case in which the linguistic and the definitional test are at odds. Examples are easy to find (in particular, see Nunberg 1979 and Deane 1988). In the sentence The newspaper has decided to reduce its size, the item newspaper refers first to the management or the board of directors of the paper, and then to the paper as a material object. The perfect grammaticality of the sentence does not allow us to spot any polysemy on linguistic grounds. Definitionally, on the other hand, it does not seem possible to find a description that fits the 'board of directors' usage and the 'actual. material publication' usage into a single well-defined category. This implies that there is a definitional polysemy in the concept newspaper that is obscured in the sentence The newspaper has decided to reduce its size. And this in turn implies that we now have examples establishing for each of the three pairs of tests that can be construed out of the basic set of three criteria, that the members of those pairs may be at odds with each other: while the dog example dealt with the definitional/logical and the linguistic/logical pairs, the newspaper example takes care of the definitionalllinguistic couple.
116
Polysemy
To complete the analysis of the newspaper example, let us notice that applying the logical test also yields polysemy. Speaking of the overpowering presence of a certain editor-in-chief, we could easily utter something like That man is the newspaper, meaning that he is the driving force of the whole business. Because the utterance would be nonsensical if newspaper were interpreted in the 'material publication' sense, the variant of the logical test that we discussed earlier in connection with This book is sad reveals the polysemy of newspaper. It can now also be appreciated that our earlier difficulties with the application of the linguistic test to dog are of relative importance. If we were to reject our earlier views on the matter and accept that, by some kind of mental aerobatics, a crossed reading of Fido is a dog and so is Lady would not only be possible but would also be recognized as odd, 8 the dog example could no longer be invoked to establish that the linguistic test and the logical test are not co-extensive. Even then, however, we could rely on the newspaper example to bring home the point. 5.2. Three-way divergence in a single item For further illustration, let us now consider a category that is a typical example of prototypicality. The category fruit is, in fact, among the categories originally studied by Rosch (see, in particular, Rosch 1975 and Rosch and Mervis 1975 for experimental results). An immediate difficulty withfruit is the vagueness of membership in the category (a problem that is less outspoken, if not absent, in the case of newspaper and dog): is a coconut or an olive a fruit? Notice, first, that we are not concemed with the technical, biological sense offruit, but with our folk model of fruit as a certain category of edible things. Technically, any seed-containing part of a plant is the fruit of that plant; as such, nuts in general are fruit. In ordinary language, on the other hand, nuts and fruit are basically distinct categories (regardless of the possible boundary status of the coconut): nuts are dry and hard, while fruits are soft, sweet, and juicy; also, the situations in which nuts and fruits are eaten are typically different. Second, remember that category membership is not the same thing as typicality: a penguin is undoubtedly an uncharacteristic kind of bird, but it is a bird nonetheless; as to the olive, the question is not just whether it is a typical fruit, but rather whether it is a fruit at all. This indeterminacy has an inm1ediate bearing on the applicability of the various polysemy tests. To begin with, the borderline status of olives shows up in the following exchange: Daddy, is an olive a fruit? - Well, it is and it is not (that is to
Vagueness's puzzles,
po~vsemy 's
vagaries
117
say, it has some of the characteristics of the things that we typically call fruits, but it also differs from the clear cases on a number of important attributes). The 'logical' polysemy that shows up here is similar to the use of hedges: 'up to a point', or perhaps even 'strictly speaking', an olive is a fruit, but seen from another angle, it is not. Applying the linguistic polysemy test, on the other hand, does not seem to yield indications of polysemy. Depending on one's conception of the category fruit, the sentence An orange is frUit and so is an olive would either seem to be simply true or false, but not ungramrnatical (as the linguistic test would have it if there is polysemy). Further, a definitional analysis is initially hampered by the uncertainty surrounding the extensional boundaries of the category. If there is consensus that olives are not fruit, we should not include the olive in an analysis of frUit. Conversely, if an olive is considered to be a fruit (however peripheral and uncharacteristic), it will have to be included. (We might then expect the definitional analysis to reveal the same polysemy that is illustrated by Daddy, is an olive afruit? - Well, it is and it is not.) To circumvent the problem with olives and their likes, let us restrict the definitional analysis to clear cases of fruit, i.e. cases for which doubts about membership do not play a role. Even for these clear cases, it can be shown that ji'uit cannot be defined by means of a single minimally specific set of necessary attributes. Because such absences are often assumed but seldom demonstrated in the literature, it may be useful to be a bit more specific on this point than is usual. A starting-point for a consideration offruit can be found in Wierzbicka's (characteristically lengthy) definition of the category (1985: 299-300). In order to show that this is not a classical, necessary-andsufficient definition, it has to be established, on the one hand, that not all attributes of fruit as mentioned by Wierzbicka are general (even within the set constituted by the examples of fruit that are high on Rosch's typicality ratings), and on the other hand, that the remaining set of general attributes is not minimally specific, i.e. does not suffice to distinguish fruit from, for instance, vegetables. The following characteristics mentioned by Wierzbicka are not general. (I repeat Wierzbicka's formulations, though not in the order in which she presents them.) (a) They have a skin harder than the parts inside. (b) They have some small parts inside, separate from the other parts, not good to eat. These parts put into the ground could grow into new things of the same kind growing out of the ground.
118
Polysemy
(c) They are good to eat without being cooked, without having anything done to them, without any other things, and people can eat them for pleasure. (d) Eating them uncooked makes one feel good. (e) Before they are good to eat they can be sour. (f) They have a lot ofjuice. (g) Their juice is good to drink. (h) They are also good to eat dried. Characteristic (a) is contradicted by the strawberry, which has no skin worthy of that name. Strawberries likewise do not have the seeds mentioned in (b); bananas are another case in point. Attributes (c) and (d) indicate that fruit can be eaten (with pleasant results) without further preparation, but this does not seem to hold for the lemon, whose sour taste generally requires sugaring. (Even if this counter-example were not accepted, adding (c) and (d) to the list of attributes that are general for fruit would not solve the problem that that list does not suffice to distinguish fruits from some vegetables and nuts.) Attributes (e) and (t) are not valid for the banana: first, an unripe banana is bitter rather than sour, and second, there is no juice in a banana. Because the generality of (g) depends on the generality of (t), it may likewise be discarded. Finally, as far as (h) is concerned, it is difficult to imagine a dried melon as being good to eat. Next, there is a set of characteristics whose non-generality seems to be accepted (or at least, implied) by Wierzbicka herself. (i) Wanting to imagine such things, people would imagine them as growing on trees. (j) They can be small enough for a person to be able to put easily more than one thing of this kind into the mouth and eat them all at the same time, or too big for a person to be expected to eat a whole one, bit by bit, at one time, but wanting to imagine such things, people would imagine them as too big for a person to put a whole one easily into the mouth and eat it, and not too big for a person to be expected to eat a whole one, bit by bit, at one time, holding it in one hand. (k) After they have become good to eat they are sweet, or slightly sweet, or sour but good to eat with something sweet. (I) Wanting to imagine such things after they have become good to eat, people would imagine things which are slightly sweet. (m) Things on which such things can grow can also grow in some places where people don't cause them to grow, but wanting to imagine such things, people would imagine them as growing on things growing out of the ground in places where people cause them to grow.
Vagueness's puzzles,
po~ysemy 's
vagaries
119
While (k) is a disjunctively defined attribute (i.e., is a superficial combination of two characteristics that are each not general when taken separately), the other features are introduced by the formula wanting to imagine such things, people would imagine them as; this would seem to indicate that the attribute is merely typically associated with the concept, rather than being general. For instance, the sweetness mentioned in (I) does not hold for lemons, and berries do not grow on trees, in contradistinction with the feature involved in (i)9 The set of general characteristics that is left over after the elimination of the previous sets contains the following features. (n) They grow as parts of certain things growing out of the ground. (0) They don't grow in the ground. (p) They become good to eat after they have grown long enough on the things growing out of the ground. (q) Before they are good to eat they are green or greenish outside. (r) People cause things of this kind to grow in many places because they want to have those things for people to eat. (s) They are good to eat cooked with sugar, or cooked as part of some things which have sugar in them. Is this set minimally specific? Up to characteristic (r), the set applies not only to fruit, but also to nuts, herbs, and large collections of vegetables (though not to the ones that grow in the ground, like carrots), so that the crucially distinctive attribute would be (s). However, if one takes into consideration the use of almonds and other nuts in certain types of pastry, the use of herbs (such as tansy) in pancakes, and the habit of cooking rhubarb with sugar, it soon becomes clear that there are counter-examples with regard to (s) in each of the three categories (nuts, herbs, and vegetables). All in all, most of the attributes mentioned by Wierzbicka are not general, whereas those that are, taken together, apparently do not suffice to exclude nonfruits. Given, then, that we cannot define the uncontroversial core applications of fruit in a classical, necessary-and-sufficient fashion, let us have another look at the other two polysemy tests. The linguistic test still does not yield indications of polysemy: An orange is a fruit and so is a banana is straightforwardly grammatical. Furthermore, the logical test does not seem to signal polysemy either: a sentence like A strawberry is a fruit and not a fruit is simply false, in spite of the fact that the strawberry does not belong to all the maximal subsets that we can distinguish in the structure offruit at the same time. That is to say, if each of those maximal subsets is a distinct meaning
120
Polysemy
according to the definitional approach, it should be possible to reveal that polysemy by applying the logical criterion; but apparently, such an interpretation is impossible for A strawberry is a fruit and not a fruit. However, consider the following exchange: 1<; a penguin a bird? - Well, it is and it isn't (that is to say, strictly speaking it is a bird, but it is not a typical bird). Taking into account that there is no doubt about the membership status of penguins in the category of birds, the 'p and not p' construction here involves differences in typicality: it sets off the prototypical application from peripheral cases. We now see that we could also have exchanges like the following: Is a mango a fruit? - Well, it is and it isn't (strictly speaking it is, but it does not have the typical characteristics of typical fruits like oranges and apples). If it is correct that the logical test does reveal a polysemy, and if this polysemy precisely involves the distinction between the typical and the non-typical cases of a category, it is also understandable that the 'p and not p' frame does not reveal any polysemy when applied to a highly typical case such as strawberry (even though it is perhaps not the single most typical case). But although the logical criterion does, in this case, yield polysemy, it nevertheless still does not lead to precisely the same kind of polysemy as the definitional test: if at all, the polysemy test can at most distinguish between core and periphery, but it cannot single out any and each of the definitionally maximal subsets that constitute the senses of fruit according to the definitional approach. In other words, the fruit example presents us with a case in which the application of all three criteria for polysemy yields different results. What are distinct meanings from one point of view (that is, according to one test) are reduced to cases of vagueness from another point of view.
6. Problems with the polysemy tests - Second Series At this point, two possibilities for further developments open up. In the first place, the recognition that the various criteria lead to different answers to the question what the meanings of a word are, may lead to the question what exactly those criteria are criteria of. If they result in contradictory statements as to what the meanings of a word are, not all of them can actually be criteria for meaning discrimination; in fact, only one could be the true test of polysemy. Pursuing this line of analysis, then, would mean eliminating the wrong criteria until the only true one is left over. Such an approach would still presuppose that words have a stable set of meanings, though. Hence, the
Vagueness's puzzles, polysemy 's vagaries
121
second road to follow would be to undennine this assumption even more than we have done already. More specifically, the belief in the stability of meanings will turn out to be ultimately spurious if the vagueness of the boundary between vagueness and polysemy can also be established when we restrict the analysis to just one of the three basic types of polysemy test (and not just when we compare the three among each other). Because this is a more radical and more fundamental departure from our traditional conceptions than the view that there is only one true kind of polysemy test, it is worthwhile pursuing this line of research first. The hypothesis, then, will be this: for each of the polysemy criteria, examples can be found in which what is a distinct meaning in one context, is reduced to a case of vagueness in another context (and vice versa). 6. I. The linguistic criterion Contextual influences on the linguistic test have been (implicitly or explicitly) noted by several authors. In fact, the recognition occurs relatively early in the literature on the subject. When Lakoff (1970) introduced the and soconstruction as a criterion for polysemy, he argued that hit is ambiguous between an intentional and an unintentional reading, because John hit the wall and so did Fred would constitute an anomalous utterance in situations in which John hit the wall intentionally but Fred only did so by accident (or the other way round). Catlin and Catlin (1972), however, noted that the sentence could easily be uttered in a context involving imitation. A situation in which John hits his head against the wall after stumbling over his vacuum cleaner and is then comically imitated by Fred, might very well be described by the sentence in question. Nunberg (1979) further drew the attention to sentences such as The newspaper has decided to change its size, which features intuitively distinct senses of newspaper ('management, board of directors' and 'material publication'). Of course, the point would be to show that the distinction between both senses is not just a matter of intuition, but that the linguistic test itself also supports the polysemy of the item under consideration. This can be done by concentrating on the asymmetries pointed out by Deane (1988). On the one hand, there are cases like the well-known The ham sandwich is sitting at table 5 and (with anaphora) The ham sandwich i tasted lousy, so hei demanded the check. These cases show neutralization of the distinction between the literal interpretation of ham sandwich and the customer metonymically connected with it. On the other hand, The ham sandwich i demanded the check because iti tasted lousy is ruled out.
122
Polysemy
While the examples presented so far involve anaphoric references, similar cases can be found involving coordination. For instance, Norrick (1981: 115) contrasted the (decidedly odd) sentence Judy's dissertation is thought provoking and yellowed with age with the (perfectly natural) construction Judy's dissertation is still thought provoking though yellowed with age. If the coordination generally requires that dissertation be used in the same sense with regard to both elements of the coordinated predicate, the sentences show that the distinction between the dissertation as a material product and its contents mayor may not play a role. Cruse (1982), in his turn, noted that none of the following series of sentences containing coordination produces feelings of oddity: (8) (9) (10) (11) (12)
(13)
John John John John John John
likes blondes and racehorses. likes racehorses and fast cars. likes cars and elegant clothes. likes elegant clothes and expensive after-shave. likes expensive after-shave and vintage port. likes vintage port and marshmallows.
Coordinating the first item in the series with the last, however, does produce an awkward sentence. So, while the awkwardness of John likes blondes and marshmallows would normally be taken as evidence for the polysemy of like, the pairings mentioned above suggest that there is a continuum of meaning rather than a dichotomy. Cruse concludes 'that readings which are close together can be coordinated without zeugma, but if they are sufficiently far apart, they are incompatible. If this picture is correct, it does not make sense to ask how many senses of like there are: there is just a seamless fabric of meaning-potential' (1982: 79).10 The previous examples sufficiently establish the context-dependent nature of the linguistic test. These contextual effects are fairly strong, in the sense that some contexts appear to be able to bridge the gap between widely divergent meanings. Deane (1988: 345) points out that explicit comparison is such a context; it can even combine actual homonyms. Examples (14) and (15) reproduce his examples. They can be contrasted with cases such as (16), in which the same meanings cannot be combined. (14) (15)
(16)
The arm ofa giant resembles that ofan ocean. Financial banks resemble those that you find by rivers: they control. respectively, the flow ofmoney and o[water. *The giant put his arm into that o[the ocean.
Vagueness's puzzles, po~vsemy 's vagaries (17)
(18) (19)
(20)
123
When Von Mi1nchhausen said a gigantic arm appeared before him when he came out ofthe jungle in Xanadu, he actually did mean the arm ofa giant, not that ofan ocean. The bank John had trouble finding again was not the one where he had cashed the cheque, but the one where he had moored his boat. Daddy, what exact!.v do you call a bank: the place where we moor the boat or the place where I bring my savings? - Well, son, the place where we moor the boat is a bank, but so is the place where you bring your savings. Daddy, what precisely is a race: a group ofpeople with same ~vpe of body, or a competition in which you try to be the fastest.? - Well, darling, competition in speed is a race, and so is a group ofpeople with the same type ofbody.
Taking Deane's observations one step further, let us note that there are other, comparison-like contexts that have similar effects. First, there are contexts of explicit contrast. For instance, in my earlier description of the difficulty of combining an intentional with a non-intentional reading, I used a sentence that did precisely that: I talked about situations in which John hit the wall intentionally but Fred only did so by accident. The case can be generalized, as illustrated by (17) and (18). A second type of context that exhibits the same strong neutralizing effect involves definition and explanation. Consider the exchanges in (19) and (20). Although it would be correct to remark that this is at least to some extent a case of metalinguistic use, this is hardly consequential for our interpretation of the possibility of identity-ofsense anaphora: the metalinguistic explanation or definitional comment is not realized as a metalinguistic statement about the words bank and race, but rather takes the foml of a categorial statement about category membership.ll At this point, two important remarks suggest themselves. First, we now have a basis for an evaluation of the position taken by Bosch (1979). His views are important, because they are the only ones (to my knowledge, at least) that explicitly take into account the contextual focusing effects that we have been considering here. Bosch's suggestion boils down to replacing the dichotomy between vagueness and polysemy by a threefold distinction, based on the presence or absence of contexts that bridge the gap between two interpretations of a predicate. Concentrating on do so ellipsis, Bosch proposes to talk of polysemy when there is no focusing context that could bring together the two distinct interpretations, of vagueness when there is no context in which they can be kept apart, and of contextually determined meanings when there is at least one context in which the contextual focus allows for a combination of both readings. But as the foregoing suggests that there is
124
Polysemy
always such a context (viz. when we are focusing on comparison, contrast, or explanation), the tripartite distinction is undermined. And even if the neutralizing effects of these contexts were not universal, they seem to be so strong that stable polysemy in the sense outlined by Bosch would turn out to be a relatively marginal case in the lexicon as a whole. Second, the kind of examples that Nunberg and Deane concentrate on raises the question whether the linguistic test (in its identity-of-sense anaphora form) is a test of polysemy at all. Is it correct to talk about such anaphoric constructions as involving 'identity of sense', when they bring together such blatantly different things as newspaper directors and paper copies of the newspaper, or even a river bank and a financial bank? Using them as evidence in a semantic analysis means treating them as categorial statements - but are they? Although Deane (1988) continues to treat these constructions as examples of polysemy, his own explanatory analysis suggests an alternative. On the basis of relevance-theoretic concepts and an explicit theory of attention, he arrives at a theory that predicts which anaphors will be acceptable. What appears to be at stake, however, is less sense identity than what might be called 'referential accessibility'. Crucial for the acceptability of constructions like The newspaper decided to reduce its size is neither the semantic (categorial) identity of the newspaper and it, nor their coreferentiality, but the question whether the referent of it can be accessed on the basis of the interpretation of the newspaper (and given certain restrictions on the flow of attention). This reinterpretation of 'identity of sense' anaphora as 'accessibility of reference' anaphora is an example of the reanalysis of the semantic tests that was suggested in the introductory passage of this paragraph. As already indicated, we will not pursue this line of thinking systematically, but the example given here may help to clarify what is involved in such a reinterpretation process. 6.2. The logical criterion There are two types of internal inconsistencies within the logical approach to polysemy. On the one hand, readings that are indistinguishable in most contexts may be separated in specific circumstances. On the other hand, readings that are mostly kept apart may be combined into a single sense in specific contexts. We have already seen an example of the first type. Whereas the concept 'bird' is usually used univocally when the biological category is meant (in the sense in which a sparrow is a bird and not a bird is unintelligible), it is
Vagueness's puzzles,
po~vsemy's
vagaries
125
possible to come across exchanges such as (21), in which non-prototypical exemplars are implicitly contrasted with typical exemplars. Furthermore, hedges and definitional specifications provide contexts in which 'p and not p' constructions can be fitted easily: see the examples (22), (23) and (24). (21)
(22) (23)
(24)
Is a penguin a bird.? - Well, it's a bird and not a bird. Is a whale ajish? - Irell, technical~y speaking it isn't, but loosely speaking you might say it is. Is a kiwi a bird? - Well, strict~v speaking it is, but it certainly isn't a bird par excellence. 1 bought this really expensive jade necklace last year, and it is certainly worth its price for all its beau~y and srylish appearance. Still, chemically speaking, is this rea/~v jade? - Well, it is jade i{you mean nephrite, but not ifyou mean jadeite.
The final example may require some comment. The traditional precious stone category jade may refer to two chemically different substances: nephrite (silicate of calcium and magnesium), and jadeite (silicate of sodium and aluminium). In outward appearance (hard and green) and function (used as ornament or for implements), both substances are similar; as such, ordinary parlance maintains the traditional unitary, non-polysemous character of the category. However, in specific circumstances such as the one evoked in the example, the chemical differences between both substances may become important the conceptual distinction that is then profiled obviously influences the possibility of the 'p and not p' construction. For the second type of internal instability of the logical criterion, it is natural to turn to cases in which two conceptually distinct readings occur together in one context. First, let us have a closer look at the second major variant of the test that we introduced in section 4.2. It can be shown, in fact, that it runs into difficulties precisely in cases where the two readings whose relationship we are trying to establish, occur together. Following a suggestion made by Kempson and Cormack (1981), we noted in 4.2 that the negation of a vague predicate p leads to conjunctive implications of the 'not PI and not p/ kind, whereas the negation of a polysemous predicate entails a disjunctive implication of the 'not PI or not P2' kind. It is easy to see that the latter disjunction is not exclusive; the negative sentence They did not reach the bank in time would still be true if they neither reached the financial bank nor the river bank. To take this fact into account, the implication frame for polysemous predicates could be rendered as 'not PI or not P2 or not (PI and P2)'. The contextual transformation of polysemy into vagueness then takes place in those situations in which the single alternatives that feature in the
126
Polysemy
disjunction ('not PI but possibly P2' on the one hand, and 'not P2 but possibly PI' on the other) are discarded in favor of the third alternative ('neither PI nor P2')' This will be the case, for instance, when They did not reach the bank in time is uttered in a situation in which the financial bank to be reached is situated on the bank of a river. Because in this particular context, the implication would be of the 'not Pl and not P2' type, bank would have to be considered vague rather than polysemous in the context, given at least that an implicational frame of the 'not PI and not P2' type is taken to be an indication of vagueness. Further, let us consider the other variant of the logical test that we introduced in 4.2. Remember that we assume that there is polysemy if the 'p and p' construction is not considered awkward or redundant. However, there are indications that this criterion needs further refinement. Let us have a closer look at the Dutch word vers, which we already briefly encountered in section 4.1 (for a full analysis of vel'S, see Geeraerts 1989a). Consider the following (quoted) examples. (25) (26)
(27)
Wie eet er de beste kost, wie koopt er verse vis en wild? 'Who can eat the finest food, who can buy fresh fish and garneT Nu meende zij, dat zij aan alles zou kunnen wennen, behalve aan gemis van beweging en verse lucht. 'Now she thought that she would be able to get used to everything, except lack of exercise and fresh air. ' Voor een goede verpakking dienen de verse sigaren tot zolang in vochtige, weke toestand te blijven. 'In order to be packed properly, the newly produced cigars have to remain soft and moist until packing. '
Example (25) illustrates the most frequent reading of vers, the one that is invariably the first to be found in dictionaries and to be remembered when informants are asked what vers means. It is used with regard to foodstuffs, and indicates that the latter are recently produced (caught, bought, harvested etc.), and are therefore in optimal condition for human consumption; one typically speaks of fruit, meat, bread etc. as being vel's. Quotations (26) and (27) show that the two notions that are combined in (25) can occur separately and independently. In (27), the notion of recency occurs without the notion of optimality, and in (26), the reverse is the case. Intuitively speaking, the central applications of vers constitute a distinct meaning: there is no spontaneous recognition of an ambiguity when one mentions fresh food, to the extent that it would be absurd to ask' Vel'S in what sense 7' when there is
Vagueness's puzzles, po~vsemy 's vagaries
127
talk of vers voedsel. Is it possible, then, to apply the logical test in such a way that the non-polysemy of the central applications is recognized? In all the circumstances in which vers could be used in the central meaning, the maximal readings whose overlap actually constitutes the central application apply. So either the 'p and p' construction can be used without awkwardness, and then the singularity of the central applications is not recognized by the logical test, or the 'p and p' frame cannot be used, and then we have a case of contextual neutralization of the distinction between the 'recency' application and the 'optimality' application (a distinction that can be firmly established when the logical test is applied in other contexts). Now, it would seem that the first alternative applies. By uttering, for instance, Dat verse brood dat ik net van de bakker heb meegebracht smaakt lekker vers ('the fresh bread that I just brought home from the baker's tastes deliciously fresh') one can enforce a construction that combines the meaning 'newly produced' in the first occurrence ofvers, with the meaning 'optimal for consumption' in the second occurrence. However, in such a case, we could salvage the logical test by imposing a restriction on it: regardless of the possibility of uttering 'p and p' without awkwardness or redundancy, the simultaneous applicability in a particular context of two interpretations of p does not constitute a case of polysemy if the two interpretations are more closely connected than merely by juxtaposition. What we would like to recognize is the 'fusion' of the two interpretations into a single one, and such a fusion is signaled precisely by the fact that there is more to the combination of those interpretations than merely their simultaneous presence. In the case of vers, for instance, the two separate readings are fused in the central app lications, because there is a causal connection between them: fresh fish is not just recently caught and (by coincidence) optimal for consumption, it is optimal in taste and nutritional value precisely because it is recent. If this is accepted, the logical approach appears to be instable: in its 'p and not p' form, it establishes the distinction between the readings 'optimal' and 'recent' of vers, but when we look at the simultaneous applicability of these interpretations, the restriction just formulated leads to the conclusion that there are contexts in which the distinction between both interpretations is overruled. It may be noted that we have now arrived a point where the logical test leads to difficulties no matter how we look at it. Suppose, on the one hand, that we simply stick to the 'p and not p' form of the test (ignoring the 'p and p' version), or that we accept the test in its 'p and p' format, but without the revision just suggested. In these cases, the intuitive univocality of the central
128
Polysemy
cases ofvers is not recognized. On the other hand, a revision of the 'p and p' test leads to the conclusion that the central application of vers is a separate meaning, and not just an ambiguous combination of the other two. As such, the distinction between the 'optimal' reading and the 'recent' meaning is neutralized in the contexts constituted by the central applications of vers. So either we repeat the observation from paragraph 5 (viz., that the results of the logical test may contradict those of other polysemy criteria), or we have to accept the contextual instability of the test. Now, while such a neutralized reading might seem to be less than damaging if it is permanently available as a third major meaning ofvers, it is unfortunately the case that the neutralization can be entirely contextual, in the sense that it is incidental. As an example of such a contextual neutralization that is like the vers case in all respects except for being purely incidental, let us have a look at the following passage from P.c. Hooft's play Granida. (Hooft is one of the major seventeenth-century Dutch authors.) The shepherd Daifilo talks to his master about his love for the princess Granida. (28)
Mijn heer, den laetsten nacht passeerd'jck bijder straten Onder de venster van mijn vrouw, - (... ): Mijn Vrouwe stondt voor 't glas, en soo sij mij sach gaen, Sondt sij haer voester at, om mij te roepen aen. Mijn Vrouw vernederde'haer, en quam mij selve vraeghen, Wat oorsaack inder nacht mij daer ontrent mochtjaeghen. [My lord, last night I passed through the street Under the window of my lady: My Mistress was at the pane, and when she saw me going there, Sent her servant down to call me. My Mistress humbled herself, and came to ask me herself What reason might have driven me there at night.]
The word we are concerned with is vernederen. Literally, it means 'to bring down' (one may recognize the English nether in the Dutch neder), but already in the seventeenth century, the figurative meaning 'to humiliate, to humble' is the primary one. In the quotation, vernederen is used reflexively; both the literal and the figurative reading are relevant. On the one hand, Granida comes down from the upper storey to meet Daifilo, but on the other hand, she humbles herself precisely by taking this step: as a lady, it is below her status to go out to meet someone of humble standing like Daifilo. Because she abandons her figuratively superior position by leaving her literally superior position, the two readings of (zich) vernederen are fused, and we have a context in which basically separate senses are combined into a single
Vagueness's puzzles,
po~ysemy 's
vagaries
129
reading, in the same way in which the two basic meaning components of vers are fused in the central meaning of that item. And as in the case ofvers, the fusion involves a causal connection between the fused readings: Granida humbles herself precisely because she comes down to meet Daifilo. Although most studies on polysemy would probably ignore cases such as this one as deliberate literary puns, the remarkable character of the simultaneous applicability of both readings of vernederen need not be based on the semantics of the construction, but may simply be a consequence of the novelty of this kind of use. Admitting that there is indeed something punning in the passage quoted above (on this point of literary style, the authors of the Dutch Golden Age are no different from their English counterpart), I do not think that we are obliged to attribute the pun to the polysemy of the construction. If a punning effect may simply be the result of unexpectedness, the remarkable character of vernederen may be a straightforward result of the unusual character of the fusion. In contradistinction with the vers case, the fused combination is not well-entrenched in the mental lexicon - but that need not mean that it is less a distinct meaning, at least not at this point in the discussion. (In section 7.1, we will explore the consequences of restricting the epithet 'meaning' to readings that are pennanently stored in the mental lexicon.) To sununarize, let me repeat the steps in the argumentation. First, the logical test is not only applicable according to a 'p and not p frame', but should also take into account cases in which two readings of a predicate p are simultaneously applicable. Second, in the latter case, polysemy does not occur when the two interpretations of p fuse coherently into a single one, as in the case of vers. Third, the incidental character of the fusion that we find in the case of vernederen underlines the unstable character of the distinction between vagueness and polysemy. More generally, vernederen is an example of contexts that may join readings that are usually separated, in contrast with the jade example and similar ones, \vhich distinguish what is usually considered identical. 6.3. The definitional criterion Let us now turn towards the contextual variability of the definitional test. Though it might superficially seem that this polysemy criterion is entirely objective and decontextualized, this is hardly the case: because the criterion requires that all the definitions that we take into account in order to establish the polysemy of an item are themselves unambiguous, the test may yield
130
Polysemy
different results according to whether we are able to accept a definition as monosemic or not. The requirement in question again refers to a traditional one in the literature on definitions: 'if the definiens is itself ambiguous, the definition obviously fails to perform its function of explaining the definiendum' (Copi 1972: 138). If the definition of what is allegedly a single sense of a lexical item is itself polysemous, the definition may either leave in the dark which of the various polysemous alternatives actually applies to the definiendum, or it may cover up the actual polysemy of the latter. The first difficulty occurs, for instance, when dark is defined as 'not light': because not light is ambiguous between 'dark' and 'heavy', the definition of dark does not prevent the meaning 'heavy' to be attributed to that word. It is the second difficulty, however, that is of interest to us here. Suppose you are describing a language with a word darvy that may be used both with the meaning 'dark' and with the meaning 'heavy'. A definition of darvy as 'not light' might then be pragmatically useful in a dictionary, because it would make economical use of the polysemy of the English expression not light to describe the range of application of darvy. In the context of the definitional test of polysemy, however, defining darvy as 'not light' would be misleading: instead of establishing the monosemy of darvy, the superficially unitary character of the definition would merely hide the actual polysemy (or even homonymy) of darvy. A real example of this kind of hidden polysemy may be found in Wierzbicka (1990). It will be recalled that Fillmore (1982) showed that bachelor (an old favorite of lexical semanticists) can only be defined as 'unmarried male' if the definition is made relative to a normal model of the world in which there is a social institution of marriage as a union of two people of opposite sexes, in which people have to reach a certain age before they can get married, and so on. Prototypical fuzziness in the application of bachelor then arises when this background model is applied to cases where the background conditions fail. Is a member of a homosexual couple (or a male adolescent of 13 or 14) a bachelor? In the sense that he is unmarried, perhaps, but certainly not in the prototypical sense of the word. Wierzbicka then argues that it is not necessary to make a distinction between the actual definition of bachelor and the related set of background assumptions (the Idealized Cognitive Model in the sense of Lakoff 1987) to explain the fuzziness of the word: a simple monosemous definition of bachelor as 'an unmarried man thought of as someone who could marry' (1990: 349) suffices, because the additional phrase 'thought of etc.' takes care of the fuzziness. Notice, however, that this is a prime example of hidden polysemy, if only because it
Vagueness's puzzles, po~ysemy's vagaries
131
includes the verb can. If one utters Yes, John can swim, various things could be meant that clearly need not all be true at the same time. It might mean (among other things) that John is objectively speaking not in a situation that impedes his swimming (he is not lying on a hospital bed with two broken legs), that he knows how to swim (he has acquired the skill), that he has the permanent right to swim (he is a member of the sports club that owns the pool), or that he has the momentaneous permission to swim (the bath attendant has made it clear that the pool is no longer overcrowded and that John may now dive in). Now, if these are distinct meanings of can, the definition of bachelor as 'an unmarried man thought of as someone who could marry' copies the polysemy into bachelor, and hides it at the same time. Also, the distinction between the various ways in which a person could marry is not irrelevant for the internal structure of bachelor, in the sense that judgments about bachelorship are related to it. On the one hand, consider Tarzan, who cannot marry because the objective conditions for marrying (such as the presence of a judge, parson, priest or whatever) are not met. On the other hand, consider a young man who does not get his parents' permission to marry his beloved. Although both cases involve a person who cannot marry in one of the various senses of can, the epithet bachelor can be applied much more easily to the latter case than to the former. If the example can be generalized, this means that the 'permission' sense of can is less important for the structure of bachelor than the 'objective possibility' sense. 12 The crucial point to make is this: if the definitional test is applied unthinkingly, a decision on the polysemy of the lexical item bachelor will contextually depend on our ability to recognize the polysemy of can - and such a recognition is initially an intuitive, individual matter: Wierzbicka, for one, seems not to have noticed it. But couldn't the recognition of such ambiguities be made less intuitive? If we are using a definitional criterion for determining the polysemy of a lexical item but have to determine independently whether the definitions that we consider are not themselves polysemous, two basic possibilities are available. First, as in our example, we may resort to other types of polysemy tests: for instance, when we establish that Yes, John can swim is polysemous because it could be both, in a given situation, true and false, we are relying on the logical test. But this would mean that the reliability of the definitional test of polysemy could be reduced to the reliability of the other criteria. Now, not only do we have independent evidence for the contextual instability of these other tests, but also, at this point in our argumentation we are specifically concerned with evaluating the definitional test in itself and by itself.
132
PoZvsemy
Second, then, an iterative application of the definitional test would not really help either, since this would merely set us off on a series of endless repetitions. Or at least, at some point an intuitive recognition of the univocality of the words we are using in our definitions would have to intervene to stop the endless regression. But precisely because we are relying on intuitive judgments of ambiguity rather than on tests and operational criteria, the possibility of contextual and individual oversight is not excluded. Moreover, if decisions about polysemy ultimately depend on an intuitive recognition of univocality, why could we not simply apply that intuition directly to the lexical item whose polysemy we are trying to establish? The whole point about considering operational tests of polysemy is to get away from the immediacy of simple intuitive judgments about polysemy. So either we accept that the ultimate validity of the definitional test is based on intuition (but then we have not got much of a test left), or we invoke the other tests (but then the singularity of the definitional test disappears, and all the problems connected with the other tests loom as large as ever).13
7.
Additional perspectives - The representational escape hatch
Because the extreme contextuality of meaning that is implied by the internal and the external inconsistencies of the three polysemy tests implies a rather drastic departure from traditional conceptions of semantic structure, it is important to proceed carefully and investigate any remaining possibility of salvaging a stable notion of lexical meaning. Taking for granted, then, that the tests for polysemy that we have discussed so far are sensitive to contextual effects, an entirely different approach suggests itself. If the 'meanings' we have been investigating are contextual surface phenomena, what model of semantic memory do we have to assume to explain the occurrence of those meanings? If the model is to be flexible, it would have to contain rules to derive flexible interpretations, together with elements that the rules operate on. What surfaces in a particular utterance is then nothing else but either one of the permanently stored elements, or an interpretation derived from applying the rules for meaning extension to one of those stored elements. In such a model, the surface-structure interpretations could be distinguished from the deep-structural meanings,14 and the polysemy problem would boil down to the question how to delimit the proper domain of the storage component from that of the rule component. For instance, if literal senses are stored, figurative ones could be derived by rule and need not be stored. In this
Vagueness's puzzles, po~vsel7lY 's vagaries
133
framework, we might try to circumvent the problematic question 'What is a distinct surface interpretation?' by merely asking 'What is a distinct meaning on the deep level of storage?'. The answer to the question is, in fact, easy in principle: an interpretation represents a distinct meaning if it cannot be derived by rule. Before trying to evaluate this approach, let us note, first, that it can be seen as belonging in the class of representational criteria for meaning discrimination: what ultimately decides on the semantic status of an alleged meaning depends on our way of representing lexical-semantic phenomena. Instead of consisting of simple definitions (as in the type of representational criterion that we discussed before), the representation now takes a 'distributed' form, part of the representational burden resting on the storage component of the mental lexicon, the other part being taken care of by the rule component. Second, it may be noted that this representational approach is not uncommon in Cognitive Linguistic debates. For instance, Sweetser (1987) is an attempt to reduce the polysemy of lie as described in Coleman and Kay (1981) to a set of contextually derived specifications of a single Idealized Cognitive Model for lie. And Vandeloise (1990) confronts the analysis of over that is included in Lakoff (1987) (and that is itself based on Brugman 1981); against the multiplication of the polysemy of over that he spots in Lakoff's description, Vandeloise tries to give a monosemic description of prepositions. What we will be concerned with here is not the detail of these debates, but the general strategy that consists of, first, defining the polysemy issue as a question involving penllanently stored meanings rather than surface structure interpretations, and second, assuming that what can be derived procedurally from stored information need not itself be stored. It is important to stress that the following objections against this view do not involve the distinction between stored and derived information as such, but rather the idea that the strategy as just defined somehow solves the polysemy Issue. The two components of the strategy can be evaluated separately. With regard to the first component, one reason for being careful with the representational approach is the fact that it seems to be based on a false analogy with phonology. In the same way that allophones can be contextually derived from underlying phonemes, contextually different interpretations of a lexical item are thought to be derived from underlying, stored meanings. Furthermore, there seems to be a functional similarity between allophony and semantic vagueness: the distinction between allophones (either in the form of free variation or in the form of contextual determination) is not a functional
134
Polysemy
one, and similarly, the contextual distinction between a male and a female interpretation of neighbor is not semantically functional, in the sense, for instance, that no truth-conditional differences are considered to correlate with it. Not surprisingly, then, the contextually determined readings are sometimes called allosemes (among others, see Deane 1988). The analogy breaks down, however, when the functional status of the allo-entities is taken into closer consideration. Take the well-known case of conjunctions. In Horn (1985), it is argued that sentences such as 5J'he 's not happy, she's sad versus She's not happy, she's ecstatic do not feature a distinct meaning of not, because the differences of interpretation can be derived from the context. Prominent among Horn's reasons for merely calling not 'pragmatically ambiguous' rather than polysemous is the fact that its actual interpretation can be derived from pragmatic principles. Now, there is a clear truthconditional distinction between the two sentences that corresponds precisely with the 'pragmatic ambiguity' of not: as the addition of the sentences' tails indicates, different truth conditions hold for she's not happy according to whether not is taken in its ordinary reading or in its metalinguistic reading. We may conclude, in other words, that not all allosemes are functionally neutral; as such, alloseme is a somewhat misleading term. To be sure, this is not an ultimately compelling objection against the representational conception of polysemy. It does not show that it is impossible to proceed in the representational way, but rather that it may be less important to do so than we might think on the basis of the phonological analogy. If the allo-entities are not structurally functional, they should not receive most of the emphasis in linguistic descriptions. But if, in the lexical-semantic domain, we conceive of contextual interpretations on the model of allophones, we run the risk of underestimating their functional importance. A similar risk is inherent in the representational reinterpretation of the tenninological pair vagueness/polysemy: because vagueness is considered the non-functional phenomenon in comparison with meaning, redefining meaning in terms of storage may make us lose sight of the fact that some of the non-stored readings are just as functionally important as the stored ones. In other words, the question 'What is a distinct meaning?' is then no longer more linguistically relevant than the question 'What is a distinct interpretation?'. Or, to put it in yet a different way, if we think of meanings as the semantically functional phenomena, it would be wrong to equate meanings and stored readings. The storage question is not an unimportant one, but it is not necessarily a question about polysemy, if polysemy is defined in functional terms. Even regardless of the terminological matter whether meaning should be defined in
Vagueness's puzzles, po~vsemy 's vagaries
135
such a functional way or not, an investigation into the functions of utterance interpretations cannot be avoided. IS So, if the difficulties we have encountered so far are interpreted as difficulties concerning the question 'What is a different surface interpretation?' (rather than 'What is a different stored meaning?'), they remain unresolved. 16 Now, with regard to the second component of the representational strategy, let us note that the representational approach as defined above is an example of the 'rule/list fallacy' (Langacker 1987: 29). If the impossibility of deriving an interpretation by rule is the criterion for distinguishing polysemy, the crucial assumption is that the set of stored meanings and the set of derived interpretations do not overlap. For what would be the status of a reading that is both stored and derivable? It would contradictorily be both a meaning and not a meaning. So, if the rule/list distinction can be shown to be untenable in this lexical-semantic context, our problems with the question 'What is a distinct meaning?' remain unresolved. Consider, in this respect, the fact that lexical items may change their prototypical centre in the course of their diachronic development. The process typically involves a shift from the periphery to the centre of one of the applications that is initially an extension of the original central sense. (Detailed examples of such processes may be found in, among others, Winters 1989, Melis 1990, Geeraerts 1990b). At the same time, this shift typically involves frequency effects: even though dominant frequency of occurrence is neither a sufficient nor a necessary criterion for the centrality of an item's reading, rising frequencies tend to correlate with growing centrality (cp. Geeraerts 1990b). If a particular peripheral meaning gets to be used more often, and if, at the same time, the originally central meaning gradually lapses into disuse, a conceptual reorganization takes place resulting in a shift of that application towards the centre of the category. But if the frequency with which an application occurs influences its prototypical status, this implies that there is some mechanism in our conceptual memory for keeping a count (however roughly) of the application's frequency of occurrence - and this would be the case regardless of the application's rule-governed nature. If changes in the frequency with which we use an application (or hear it being used) may result in a process by which the application becomes more central in our semantic memory, we are somehow aware of that reading as an individual element in our mental lexicon: the account we apparently keep of the reading's frequency of occurrence presupposes a representation of that reading - even if it is entirely derivable from a given stored meaning, a particular context, and a specific procedure for semantic extension. In short, there are reasons to
136
Polysemy
doubt that a representational conception of univocality, founded on a dichotomous conception of the rule/list distinction, will yield satisfactory results. 17
8.
Consequences
The consequences of the argumentation in the foregoing pages can be traced on two different levels. First, we may have to change our (pre-theoretical or theoretical) conception about what meaning is. More specifically, we will have to determine the impact of the instability of the distinction between vagueness and polysemy on Kleiber's model of prototypicality effects. Second, there may be an influence on our model of what semantics is, i.e. on our conception of the methodology of lexical-semantic research. And here as well, an interesting consequence for prototype theory may be envisaged. 8.1. Consequences on the pre-theoretical and the theoretical level If it can indeed be accepted that the distinction between vagueness and polysemy is not well-defined, the primary conclusion to be drawn relates to the problem that initially engendered our discussion of polysemy. Because the instability of the distinction between vagueness and polysemy precludes a strict dichotomy between intercategorial and intracategorial semantic multiplicity, Kleiber's conception of prototype theory need not worry us too much for the time being. It will be appreciated, however, that this conclusion is considerably hedged in. On the one hand, the fact that we have not found a stable criterion for a distinction between vagueness and polysemy does not mean that such a criterion is in principle impossible to find. If we have not found a stable operational test yet, we have perhaps not looked hard enough. Of course, our pre-theoretical notion of what a distinct meaning is, is not entirely clear, and in that sense, any search for polysemy criteria will be hampered by the fact that we do not know precisely what it is we want a criterion for. But this is not an uncommon situation in linguistics, and so further research into (new and old) polysemy criteria retains its importance. Needless to say, research into polysemy criteria also implies an investigation into the question what the criteria we originally considered to be polysemy criteria are actually criteria of, if not criteria of distinctness of meaning. On the other hand, our rejection of Kleiber's attempt to bring some order in prototype theory does not mean
Vagueness's puzzles, polysemy 's vagaries
137
that all is well in prototype country. Here too, further research into the relationship between the various sorts of prototypicality effects and their sources remains necessary. The main point we can claim on the basis of the foregoing pages, is that Kleiber's attempt to bring order into the prototypical matter seems to meet with fundamental difficulties - not that prototypicality is in no need of further investigation. This is all the more so because our own findings suggest modifications in our conception of lexical structure. After all, the idea that meanings (as distinct from mere referential specifications) do not exist, is rather disconcertingly at odds with what we traditionally believe. So let us try to determine what other consequences this reversal (conditional though it may be) leads to. One major thing to be changed, I would claim, is our metaphorical, pretheoretical image of what meanings are. Come to think of it, the very wording of the previous sentence reveals something about the way we talk about meaning: the plural meanings suggests that meanings are entities or entitylike things, presumably stored somewhere in the big storage room that we conceive the mental lexicon to be, to be carried by word forms to an interpreter who unpacks and understands them. This metaphorical image has, of course, been identified and labeled before: it is both that of the conduit metaphor of verbal communication, first described by Reddy (1979), and of the lexemist conception of the mental lexicon, more recently analyzed by Hoffinann (1989). The common characteristic of both metaphorical complexes is that of reification: meanings are things, prepackaged chunks of information that are contained in and carried about by word bags. But what kind of image could replace this reificational picture? The tremendous flexibility that we observe in lexical semantics suggests a procedural (or perhaps 'processual') rather than a reified conception of meaning; instead of meanings as things, meaning as a process of sense creation would seem to become our primary focus of attention. The image I would like to propose to make this conception more graspable, is that of a floodlight: words are searchlights that highlight, upon each application, a particular subfield of their domain of application. If this domain of application is seen, in model-theoretical fashion, as a set, each time the word is used a particular subset is selected. But while our traditional view of the distinction between vagueness and polysemy entails that the number of subsets that can be lit is fixed and restricted, we now have evidence that the verbal searchlight has much more freedom. The freedom is not absolute, surely, and there will be preferential subsets for each word; even so, the distinction between what can and what cannot be lit up at the same time is not stable. IS
138
Polysemy
Although we should obviously try to move beyond a mere metaphorical description of our new conception of lexical- semantic structure, the actual development of a model of categorial structure that takes into account the shifting nature of meaning lies beyond the reach of this paper. It may nevertheless be noted that, within the broad range of prototype-oriented studies, the 'vantage point theory' presented by Robert MacLaury (1991) is perhaps the one that is most congenial to this enterprise (most explicitly so because it provides for perspectival shifts of the prototypical centre of a category). 8.2. Consequences on the methodological level Accepting that the distinction between vagueness and polysemy is subject to contextual variation highlights the interpretative nature of semantics: meanings are interpretations that arise out of a particular context. But if meaning is radically context-dependent, what is the methodological status of meaning - given that it is us, the interpreters, who choose or provide the contextual perspectives that lead to a particular interpretation? Are meanings phenomena that are objectively out there and that can be discovered, or are they subjectively construed by the interpreter? Are meanings found or are they made? The philosophical problem tumbling head over heels into the discussion here is that of the objectivity of lexical semantics - not just the objectivity of the object of semantic research, but the objectivity of its method. Cognitive Linguistics generally follows Lakoffs (1987) characterization of the enterprise as an anti-objectivist one, but it is seldom realized that this characterization can be taken in two different ways. Usually, the anti-objectivism is taken to be a characterization of the object of the enquiry. But even if linguistic meaning is conceived of as a subjective, experiential, psychological phenomenon, does this also mean that the method of linguistic semantics is non-objective? After all, even if the meanings that we investigate are subjective entities, they might still be there objectively: we could find the meanings as they are, and we would simply have to extract them from the messages in which they are transmitted to us. However, this picture smacks of the reified conduit metaphor and lexemist model that we rejected earlier. What, then, if we do not find objective meanings ready-made, but construe them subjectively? First, this would be naturally in line with the rejection of the reificational lexemist/conduit model: while we primarily rejected the model as a picture of how meanings are put into words, we can now extend this rejection to the way in which meanings are extracted from words. And second,
Vagueness's puzzles, polysemy 's vagaries
139
extending the anti-objectivist, anti-reiticational stance to the receiver's end of the communicative process makes good sense of the difficulties we encounter when trying to fix the meaning of utterances like the ones we have been discussing throughout the paper. What we started looking out for was a unique meaning, i.e. the meaning of the utterance, the authentic and original one that the author put in the message (or perhaps: highlighted) and that is objectively there for us to be unpacked (or alternatively: observed). But, disconcertingly, we have serious difficulties in pinning down the unique meaning of an utterance, not least because the various perspectives that we adopt yield different answers to our semantic question. However, if we abandon the vestiges of objectivism in our methodological self-conception, the presupposition that there is a unique meaning itself can be rejected. Rather than a single unique meaning, there would only be the interpretations that we impose on the material - and our interpretative activities need not yield a unique result. It is not the intention of this paper to settle this question, but just to further illustrate its importance, two points are worth mentioning. First, the interpretative nature of the meanings we attribute to utterances has a bearing on the status of the prototypical structure that we attribute to lexical categories. In those cases where we do not rely on 'hard' techniques such as statistical analysis to determine the core meaning(s) of an item, but rather use traditional linguistic methods of data gathering and description, the question often arises 'How exactly do you determine what the prototypical centre of an item is?'. Now, if the prototypical core meaning of an item is no more objectively traceable than the meaning of a word in an individual utterance, could it not be the case that we choose that meaning as prototypical that gives us the best starting-point for interpreting the various applications in which we encounter that item? Such a meaning should not only be applicable in a straightforward way to the majority of the applications in which we encounter the word, but should also be suited to arrive at an interpretation of novel applications by applying regular mechanisms of semantic extension to the prototype. In that case, the prototype is that meaning that best enables us to make sense (literally, as an interpretative process) of the various ways in which words are used. Or, in other words, the prototype would have to be thought of as an interpretative perspective that helps us to interpret the uses of a word. (Our analysis could then make claims to objectivity to the extent that it mimics the interpretative structure other language users impose on the category.)
140
Polysemy
Second, lurking round the corner here are three major philosophical strands in the development of modem thinking. First, there is the hermeneutical tradition as it stretches from Schleiermacher over Dilthey to Gadamer, Habemlas, and Ricoeur (for an English introduction to the tradition, see a.o. Bleicher 1980 or Ricoeur 1981: 43-62). A crucial aspect of the hermeneutical tradition (specifically as embodied in the Diltheyan distinction between Geisteswissenschaft and Naturwissenschaft) is the methodological characterization of the human sciences (explicitly including linguistics) as being based on processes of interpretation rather than on the objectivist methodology of the natural sciences. In the methodological self-conception of linguistics, this idea has hardly been more than an undercurrent in the major part of the twentieth century; see, however, Itkonen (1978, 1983) and Anttila (1985) for a defense of a hermeneutical conception of linguistics. Second, there is pragmatism, specifically in the form that our cognitive activities are guided by pragmatic interests. For if our interpretations depend on differing perspectives, the choice of a particular perspective may be guided by various interests. Third, the whole set of problems touched upon here extends towards the philosophy of science: the falsificationalistl paradigmatic theories of science (in the line of Kuhn 1970 and Lakatos 1970) precisely emphasize the idea that all scientific thinking starts from perspectivist assumptions. (See Geeraerts 1985a for a discussion of the relationship between prototype theory and this line of thinking in the methodology of science.) However briefly evoked, these philosophical connections highlight the fundamental issue raised by the polysemy problem. If the difficulties we have with the determination of lexical meanings are indeed indicative of the ultimately non-objectivist, perspective-bound, hermeneutically interpretative nature of linguistic semantics, a methodological reexamination of the enterprise imposes itself. If our understanding is perspectival, what perspectives do we actually use, and what perspectives should or could we use? And what pragmatic interests could lexical semantics serve? It is beyond the scope of this article to answer these questions, but their far-reaching character should be sufficient to establish the importance of a systematic exploration into the methodological ramifications of the polysemy issue.
Vagueness's puzzles, polysemy's vagaries
9.
141
An inconclusive overview
The preceding pages have taken us on a tour through the mysterious world of lexical meaning. An examination of three basic classes of criteria for distinguishing between polysemy and vagueness has made clear, first, that those three types of criteria are in mutual conflict (in the sense that they need not lead to the same conclusion in the same circumstances), and second, that each of them taken separately need not lead to a stable distinction between polysemy and vagueness (in the sense that what is a distinct meaning according to one of the tests in one context, may be reduced to a case of vagueness according to the same test in another context). Rather than leading to the assumption that only one of the three tests is the true criterion for distinctness of lexical meaning, these observations give rise to the suspicion that the distinction between vagueness and polysemy is indeed unstable. If so, lexical meanings are not to be thought of as prepackaged chunks of information, but as moving searchlights that may variously highlight subdomains of the range of application of the lexical item in question. Furthermore, the recognition that the distinction between vagueness and polysemy is not unproblematic is not just important for lexical semantics in a very general way, but has specific consequences for prototype-theoretical approaches to the study of word meaning. In particular, the suggestion (as put forward by Kleiber 1990), that various types of prototypicality effects should be kept apart on the basis of their intracategorial or intercategorial status, can now be seen to rely uncritically on the stability of the distinction between vagueness and polysemy.19 In general, in spite of its vital importance for any theory of word meaning, lexical semantics does not seem to have arrived at a consistent operational definition of polysemy yet. Within the non-objectivist framework of Cognitive Semantics, this observation leads to the suggestion that the methodological uncertainty surrounding the concept of polysemy is a side-effect of the interpretative nature of linguistic semantics. However, rather than trying to argue for this thesis as such, the paper has merely tried to evoke the methodological set of problems in the context of which it can arise as a hypothesis. By describing the operationally problematic nature of polysemy, it has tried to make clear that the methodological status of lexical semantics is in urgent need of further analysis.
142
Polysemy
Notes 1.
2.
3.
This formulation presupposes that category as used in the terms intercategorial and intracategorial is defined as 'lexical meaning, sense of a lexical item'. This is not terminologically unquestionable, because in actual practice, category may sometimes be used to refer to lexical items (whether polysemous or not) just as well as to the individual senses of such items; in the latter case, one might speak of 'semantic categories', in the fonner of 'lexical categories'. If such lexical categories are polysemous, they constitute a cohesive cluster of semantic categories. Although the il1ter-/intracategorial terminology equates category with 'semantic category', the distinction between both types of prototypicality that is at stake here, could also be expressed by distinguishing prototypicality within 'semantic categories' from prototypicality within 'lexical categories'. This is not just the case for prototype theory's investigation into degrees of membership. In the context of a logical approach to meaning, the vagueness of predicates like large is discussed in connection with the Sorites paradox: removing one grain from a large heap of sand does not imply that the heap stops being large, but after an indefinite number of steps, the process can no longer be iterated. How, in other words, should the meaning of large be described to accommodate this fact and avoid paradoxes? See Ballmer and Pinkal (1983) for a sample of this line of research. It may be noted that there is yet a fourth type of vagueness to be mentioned, viz. the interpretative indeterminacy that occurs when there is a contextually unresolved polysemy in an utterance. The types of vagueness mentioned in the text involve indeterminate relations between lexical items and their referents (the relation between language and the world, if you wish). But indeterminacy may also include the relationship between the observer (say, the linguist) and the language, in those cases where it is difficult to determine which reading of an item is realized in a particular context. If, in a particular context, it is difficult to make out whether a particular predicate p exhibits reading PI or pz, this could either be because the distinction between PI and P2 is semantically neutralized in that context, or because the context does not provide enough information to choose between both. Whereas the first case represents the kind of vagueness that we will be concerned with in the rest of the paper (the speaker or writer intended to be neutral with regard to the distinction between PI and P2), the second case represents unresolved ambiguity (the addressor of the message intended to communicate either PI or P2, but, for lack of sufficient clues, the addressee cannot decide which is the case). An examination of the relationship between both types (which is interesting primarily for the methodological issues raised in section 8.2) is beyond the scope of the present text. It might be added at this point that the tenninologi-
Vagueness's puzzles, polysemy's vagaries
4.
5.
143
cal difficulties that are mentioned with regard to the term vagueness also occur with regard to the term ambiguity: while it sometimes used as a synonym of polysemy, it may also occur with different readings (for instance, as a specific term for the type of unresolved polysemy discussed in this footnote.) Because of the comprehensiveness of Cruse's (1982) treatment, it may be useful to give a brief indication of the relationship between the tests distinguished by him and the major types distinguished here. While he does not devote attention to the definitional approach, Cruse's tests Band D correspond with our logical and linguistic test respectively. His criterion A will be incorporated here as a specific subtype of the logical test (viz.. involving negation; see section 4.2). As for Cruse's criterion C, I would argue that it as well is a variant of the logical test to the extent that it involves the possibility of spontaneously recognizing (and asserting), in the same situation, two readings with different truth-conditional properties for a given sentence. Specifically, this same situation here takes the form of a sentence conjoined to the one featuring the predicate whose polysemy we are trying to establish. Consider the anomaly of The children wore red shirts, and the boys, blue ones in comparison with The dogs wore red collars, and the bitches, blue ones. The latter sentence is true-or-false in the reading in which dogs means 'male Canis familiaris', but anomalous in the reading in which dog means 'Canis familiaris' in general; the anomaly derives from the fact that a general reading for dog entails that bitches wear red collars, which is in contradiction with the second conjunct, specifying that bitches wear blue collars. (This is, of course, merely an encyclopedic contradiction, not a logical one.) So, in the same situation (the one in which bitches wear blue collars), the sentence The dogs ~j.'ore red collars can be spontaneously recognized as either true or false, or anomalous. As explained in 4.2, this is one of the realizations of the logical test. In the case of The children wore red shirts, on the other hand, no two interpretations with distinct truth-conditional properties can be spontaneously recognized in the situation in which the boys wear blue. The classification of polysemy criteria presented in Cruse (1986: Chapter 3) differs on two fundamental points from the one in the article of 1982. First, the direct criteria are divided into three classes. One of these corresponds with our 'linguistic' class, while a second one includes our class of 'logical' and related approaches. The third criterion is a novel one, to the extent that it invokes a distinction between semantic specifications that are derivable from context, and those that are context-independent. We will come back to this criterion in paragraph 7: in particular, see note 16 below. Second, and more importantly, Cruse' s argumentation in the book of 1986 seems to be partly convergent with the argumentation in the present article. to the extent that his text contains some observations questioning the idea of a clear dichotomy
144
6.
7.
8. 9.
Polysemy
between polysemy and vagueness; more specifically, he describes his own view as a 'compromise position' (1986: 81). Note that the vers case has more drastic consequences than the old case. It could be argued, in fact, that we should not consider relations such as antonymy separately, but that the entire set of relations that an item contracts with other items should be taken into account. Thus, although old may have identical antonyms in the two readings involved, the relationship of synonymy might be invoked to disambiguate between both readings, among other things because the synonym former only applies in one of the two. Even if this global approach would enable us to circumvent the class of problems exemplified by old (which would, to be sure, have to established separately), it would not help against the type of difficulties illustrated by vers: taking into account the entire set of lexical relations that an item contracts might avoid relational underspecification of meanings (i.e. the situation that a relational approach is unable to spot semantic distinctions), but it could not help against relational overspecifications (i.e. seeing more meanings than is feasible). On the contrary, it would probably strengthen the danger of overspecification. In Cruse (1986: 57), thin is used as an example of the vers-type of problem. If a longitudinal analysis confirms that this confusion was indeed a regular feature of the history of logic, part of the explanation might be that the interest in the phenomenon of ambiguity - at least judging from Rosier (1988) has traditionally always gone more to 'amphiboly', i.e. the rhetorical use and abuse of sentential ambiguity, than to the operational definition of lexical polysemy. A good candidate is the following sentence from Cruse (1986: 64): ? Dogs become pregnant at J2 months, but mature later than bitches. But could it not be argued that attributes such as those involved in (1) and (i) are general precisely as typical characteristics? This is (if I read her correctly) the position defended by Wierzbicka (1990), where she argues for the inclusion of 'mental characteristics' in definitions. The reasoning can, I think, be rendered in the following way: although not all fmits grow on trees (and although, therefore, growing on trees is not a general characteristic of fmit) , it is a general characteristic of fruit that people usually think of it as growing on trees. Now clearly, including prototypical characteristics into definitions is not what we would be arguing against from a cognitive point of view. Still, how exactly should we constme Wierzbicka's argument? On the one hand, 'usually growing on trees' as such is not a general characteristic of fruit: if 'usually growing on trees' is a general characteristic of fruit, any fmit should usually grow on trees - but of course, one cannot say that a strawberry usually grows on trees. In other words, if 'usually growing on trees' is constmed as part of our knowledge about fruits, it clearly is not a general attrib-
Vagueness's puzzles, poZvsemy 's vagaries
10.
11. 12.
13.
145
ute. On the other hand. suppose we construe 'usually growing on trees' more radically in the mental sense intended by Wierzbicka: it is then part of our knowledge of fruit that (other) people tend to think of fruit as growing on trees. But again, this attribute (which involves not features of x, but knowledge of features of x) can only be general if, for any thing that belongs to the category fruit, other people tend to think of it as growing on trees - and we probably would not claim that other people tend to think of raspberries as growing on trees. In short. attributes that take into account prototypicality phenomena by incorporating operators like usually do not restore generality. It might perhaps be remarked that Crose's argumentation is less compelling than he suggests. If the dichotomy involves (roughly defined) a 'gustatory' and a 'non-gustatory' sense of like, it could be argued that there is no sentence in the set that actually combines both readings (as would be required to establish the contextual neutralization of the polysemy). The only candidate would be (12), but a crossed interpretation of this sentence is not the first one that comes to mind; rather, it would seem that John non-gustatorily likes both items as status symbols or life style indicators. However, in Cruse (1986: 72), examples of sense-spectra are given for which the specific difficulties with like do not occur. Moreover, examples may be found in which metalinguistic overtones are entirely absent; see David Tuggy's paint example (Tuggy 1993). Furthermore, observe that Wierzbicka' s definition misses the point that apart from ability, willingness to marry also plays a role in judgments about bachelorship: an unmarried male person living together with a woman on a permanent basis is not (if at all) a good example of a bachelor, although he could marry in all relevant respects. This discussion does not imply that there could be no other grounds for establishing the contextual variability of the definitional approach. McCawley (1981: 7), for instance, notes that one person may more easily come up with a general definition than another one: until someone pointed out to him that the two interpretations of queen ('female ruler' and 'wife of male ruler') could be covered under the general definition 'enthroned female', he had been unable to define the word non-enumeratively. In general, the failure to come up with a single definition may be the result of insufficient ingenuity on the part of the definer, and ingenuity is obviously a characteristic that can change from one person to another. Yet another possibility for establishing the contextual variability of the definitional approach is based on the potential co-existence of two distinct definitional analyses of the same item. Suppose, for instance, that Aristotle's cases of megalopsychia are analyzed in such a way that the two groups for which a general definition can be given are constituted by Alcibiades, Achilles, Ajax, and Lysander on the one hand (megalopsychia being 'military ambition in the service of one's country'),
146
Polysemy
and Socrates on the other (megalopsychia being 'the unwillingness to waver from the quest for truth under the pressure of others'). It will be obvious that what is a distinct meaning in the context of this analysis does not have a separate semantic status in the context of Aristotle's definitional analysis. 14. The distinction is not new in lexical semantics; see, for instance, Schmidt's (1963) distinction between lexikalische and aktuelle Bedeutung. 15. There is yet another reason for not forgetting the contextual interpretations. If the stored meanings are by that very fact unobserved entities, and if, conversely, the things that we can actually observe (however indeterminatedly) are surface structure interpretations, getting an idea of what the surface structure meanings are is epistemologically prior to determining the background entities from which they are derived. If representations plus rules are postulated to explain the occurrence of particular surface phenomena. these phenomena have to be determined first before one can start postulating explanatory entities. 16. At this point, we may have a look at the third criterion for polysemy introduced by Cruse (1986); see note 5 above. In essence, the criterion relies on a distinction between interpretations that are totally conditioned by context and ones that are context-independent. For instance, the reading 'female sovereign. queen' of monarch in The Ruritarian monarch is expecting her second baby does not represent a case of polysemy, because the sex of the monarch can be derived from the contextual information that she is having a baby. The argumentation can be reconstructed as follows. 'On the level of surface interpretations, monarch may be interpreted as either 'male sovereign' or 'female sovereign'. These are distinct surface interpretations, because there is at least one context in which they are not compatible, i.e. in which the logical criterion for semantic distinctness applies. (In fact, they are incompatible in any context). These distinct surface interpretations do not represent distinct meanings. because the question which interpretation applies always depends on contextual information (such as the presence of predicates that can only refer to women)'. Now, consider the question what are the constraints on the kind of contextual information that is envisaged in this argumentation. There seems to be no reason to restrict it to purely linguistic contextual information, if only because any relevant piece of extralinguistic. encyclopedic or situational information could be turned into a piece of linguistic context by expressing it in words. But then, the same argumentation that is used to defend the claim that monarch is vague rather than polysemous with regard to the distinction between the readings 'male sovereign' and 'female sovereign' can also be applied to cases of genuine polysemy: 'On the level of surface interpretations, position may be interpreted as either 'occupied place' or 'mental attitude'. These are distinct surface interpretations. because there is at least one context in which they are not compatible, i.e. in which the logical crite-
Vagueness's puzzles, polysemy's vagaries
147
rion for semantic distinctness applies. These distinct surface interpretations do not represent distinct meanings, because the question which interpretation applies always depends on contextual information (such as the referents involved)'. Therefore, if the reconstruction of the argumentation and the generalization of 'context' are justified, there are reasons for doubting whether the criterion is sufficient yet to distinguish vagueness from polysemy. 17. Another failing escape hatch that should be mentioned could be dubbed 'the ploy from infinity'. In fact, could it not be argued that the distinction between vagueness and polysemy is necessary in principle, lest lexical semantics become impossible? The traditional line here is to state that the distinction between polysemy and vagueness is necessary to avoid endless polysemy. If there is no stop to referential distinctions turning up as distinct meanings, a word would have an infinity of meanings. If, for instance, neighbor is polysemous between the readings 'male neighbor' and 'female neighbor' merely because there is a referential difference between the members of both subsets of the extension of neighbor, what is there to stop us from claiming that 'neighbor who owns a Volvo' and 'neighbor who owns a Chevrolet', or 'neighbor who is 1.73 m tall' and 'neighbor who is 1.74 m tall' are distinct meanings? After all, the referential distinctions involved define subsets just like the distinction between 'male' and 'female'. Furthermore, the fact that such distinctions could be multiplied ad infinitum would yield an endless polysemy. So, it might be concluded, even if we do not know yet how to determine the distinction between referential differences that do and those that do not turn up as separate senses, we are at least sure that such a distinction has to exist. However, the argument from infinity assumes that any referential difference would eo ipso turn out to be the basis of a semantic distinction: only in that case could there be infinity. What we have discovered, on the other hand, is the instability of the distinction between polysemy and vagueness, not its complete absence. The distinction remains valid in the sense that referential differences do not automatically surface as distinct meanings: there is always a preliminary process of categorization involved that picks out particular referential attributes as the basis of categorial similarities or contrasts. The attributes that are being picked out may not be entirely stable (in the sense that a distinction that is disregarded in one context is promoted to semantic distinctiveness in another), but there is no automatic mechanism that turns referential differences into semantic oppositions. 18. This is not the first time a 'searchlight' or 'highlighting' metaphor has been used in lexical semantics. The searchlight image has been used by Weisgerber (1954) to specify the way in which various items illuminate portions of lexical fields. The concept of highlighting is used by Crose (1986: 53) to indicate the contextual modulation of an item's meaning, in the way in which the car needs servicing and the car need~ washing focus on different parts of
148
Polysemy
the car. It will be clear in what way our own terminology differs from Crose's and Weisgerber's. While the latter uses the term as an onomasiological concept, ours is a semasiological one. Crose's way of using the term, on the other hand, is referential rather than semantic: what is highlighted is less a part of the total set of semantic possibilities of an item than an aspect of the referent of that item in one of its senses. (This referential conception is not entirely separated from the semantic conception, however. In particular, Crose's approach seems to rely on a distinction between contextual modulation and contextual selection of senses that may not be tenable in the light of our present argumentation. The central tendency of the present paper leads to the suggestion that what are variations of one sense at one point, may turn out to be distinct senses at another. For instance, while Crose's examples refer to a situation in which both aspects of cars ('carriage' and 'engine') are simultaneously applicable but in which one of them is backgrounded, there are contexts in which both aspects are incompatible: an abandoned and demolished car with the engine removed would seem to be a car in the sense of the car needs washing but not in the sense of the car needs servicing. That is to say, if Crose's conception of highlighting presupposes that the backgrounded and the highlighted reading are compatible in the relevant context, our notion of highlighting does not carry that presupposition; moreover, interpretations that stand in a relationship of compatibility-plus-focusing in one context. may turn out to be incompatible in another.) 19. Admittedly, by focusing on the problematic, unstable cases we may have exaggerated the instability. It could still be the case that in actual practice, the problematic cases constitute merely a peripheral phenomenon. Even so, they constitute a constitute a problem in principle for semantics, to the extent that they show that the definition of lexical meaning is not as clear as might be hoped.
Chapter 6 Classical definability and the monosemic bias
Originally published in Rivista di Linguistica 1994.6: 189-207. Lexical semantics cannot at the same time retain a monosemic bias and the ideal of classical definability: there exist cases where sticking to the intuitively plausible idea that a particular lexical item is univocal leads to the conclusion that it cannot be classically defined. In the present chapter (which is based on section 3.2 of Geeraerts, Grondelaers, and Bakema 1994), this claim is supported by means of a corpus-based case-study involving denotational data based on non-elicited actual usage. Within this methodological framework, a procedure for determining classical definability will be demonstrated that operationalizes the requirement that definitions be both general and distinctive. While the previous chapter, then, looked at problems of definability from a methodological point of view. with an eye to the criteria that may be used for establishing either polysemy or univocality, the present paper takes a more practical approach, illustrating definitional problems with detailed examples. Crucially, the paper argues for a rigorous procedure in answering definability questions. If we assume that monosemy can be established by providing a classical definition (a definition in terms of necessary and sufficient features that is both general and distinctive with regard to the item to be defined), there are two requirements that discussions of definability should strictly adhere to. First, the generality of classical definitions implies that no disjunctive features are included in the analytical definitions, and second, the distinctiveness of classical definitions implies that they should make the right predictions about the lexical relations among the items involved. As the examples in the present paper show. complying with these demands requires a careful. step by step procedure that stands in sharp contrast with what semanticians usually do when they posit univocality. The present version of the text is considerably expanded with regard to the originally published article. The discussion of legging, as an example of classical definability. was missing from the version in the journal.
1.
Polysemy and definability
Definability plays an important role in current discussions about lexical categories. if only because it is intimately related to the problem of polysemy. Lexical-semantic research is to a large extent guided by a mono-
150
Polysemy
semic bias: the idea that postulating a single meaning for a lexical item is, on the whole, preferable over postulating multiple meanings. Authors who have recently defended this position with varying degrees of explicitness include Charles Ruhl (1989b), Anna Wierzbicka (1985, 1991), and Claude Vandeloise (1990). As Chris Sinha has pointed out (p.c.), the monosemic bias is related to Grice's 'Modified Occam's Razor': the principle that senses are not to be multiplied beyond necessity (Grice 1989: 47). The monosemic bias is not universal, however. In my own work on lexical semantics, for instance, I have had a tendency to stress the presence of polysemy (for instance, see Geeraerts 1989a, 1990a) and it has been argued (Landheer 1991) that I have so to speak exhibited an undue polysemic bias. Now, the monosemy/polysemy issue is complicated by various factors, two of which stand out. First, there may be various criteria for deciding about the presence of monosemy. In Geeraerts (1993a), an overview of some of the most common criteria is given (and it is shown that they are mutually inconsistent). Second, there is variation as to the level of semantic analysis with regard to which the monosemy question is defined. Most linguists, in fact, would seem to be willing to accept a two-stage model in which a distinction is maintained between, roughly, permanently stored 'deep stmcture meanings', and 'surface stmcture meanings' that can be derived from them on the basis of cognitive and pragmatic principles of semantic extension interacting with situationally and textually contextual factors. If the monosemy issue is defined with regard to the stored readings, different solutions will be arrived at in comparison with a situation in which it involves surface stmcture interpretations. In the former case, for instance, one could say that something is not a different reading when it can be derived by mle from a stored reading (cp. Kempson 1977). In the latter case, on the other hand, one of the possible criteria for semantic difference rests on the possibility of capturing all surface readings of an item under a single definition. Without trying to solve the problem of monosemy as a whole, the present paper will try to shed some light on the latter test of polysemy, viz. the definitional test ofpolysemy that says that two of more instances of a word do not constitute a different reading when they can be defined together as a single meaning. In particular, it will be demonstrated, first, how the criterion can be applied to usage-based data about the referential range of application of lexical items, and second, that applying the criterion involves more comparative computation than is usually assumed. Third, this demonstration will support the claim that lexical semantics cannot at the same time preserve the monosemic bias and the ideal of classical definability: cases will be
Classical definability
151
presented where sticking to the intuitively plausible idea that a particular lexical item is univocal leads to the conclusion that it cannot be classically defined. This claim will be illustrated on the basis of a corpus consisting of 9000 Dutch clothing terms, taken from fashion magazines and general magazines published in 1991. A full description of the project is presented in Geeraerts, Grondelaers, and Bakema (1994); the present paper corresponds largely with section 3.2. of the monograph. In the database, each garment is present in the form of a componential description. The descriptions are based on photographs (or, occasionally, drawings) that appear in the magazines; only words referring to such pictures have been included in the database. The componential system of description differs somewhat according to the major subtypes of clothing. All types of trousers, for instance, are described by means of the following set of features. First dimension: LENGTH 1 The garment does not reach further down than the groins 2 The garment reaches down to the thighs 3 The garment reaches down to the knees 4 The garment reaches down to the calves 5 The garment reaches down to the ankles Second dimension: WIDTH AND CUT 1 The garment is tight-fitting 2 The garment has a straight cut, neither tight-fitting nor wide 3 The garment has a loose, wide cut 4 The garment is tight-fitting round the hips but has widening legs 5 The garment is loose-fitting round the hips but has straight or tightfitting legs 6 The garment has a regular straight cut as far down as the knees, but has widening legs below the knees 7 The garment has a loose cut from the hip to the knee, but has tight-fitting legs below the knees Third dimension: 1 The ends of the 2 The ends of the 3 The ends of the 4 The ends of the 5 The ends of the
END OF LEGS
legs exhibit no special features legs fit tightly by means of an elastic band legs fit tightly by means of tied laces legs fit tightly by means of buttons legs have an elastic band that fits under the feet
Fourth dimension: 1\1ATERIAL 1 The trousers are made of cotton, linen or a similar smooth cloth
152
Po~ysemy
2 The trousers are made of wool or a similar coarse or fluffY material
3 The trousers are 4 The trousers are 5 The trousers are 6 The trousers are 7 The trousers are 8 The trousers are
made of denim made of corduroy made of silk or a similar smooth and shining material made of smooth, shiny leather or an imitation of it made of rough, mat leather or an imitation of it made of stretch
Fifth dimension: DETAILS I Seams and/or pockets are strengthened by metal buttons 2 The waist part of the garment has folds, pleats or creases 3 The garment has a very low crotch 4 The garment is made up of several visible strips of cloth 5 The legs have sharply pressed creases 6 The garment has an elastic band in the waist A specific garment may then be described, for instance, by the featural configuration H51431, where H indicates that the garment belongs to the group of trousers, and where 51431 indicates that it is a pair of trousers with value 5 on the first dimension, value 1 on the second, 4 on the third, and so on. It should be clear from this presentation that the approach followed here is a referential, denotational one: lexical meanings are studied not primarily through introspection, but on the basis of their ranges of application as they appear in actual usage. The referential, usage-based investigation followed here is not accepted unconditionally by all lexical researchers. In actual practice, lexical studies with a Cognitive Semantic orientation exist both in the form of introspective analyses, and in the form of corpus-based research. Schmid (1993: 272) even considers the corpus-based approach in work such as that of Rudzka-Ostyn (1988, 1989), Schulze (1988, 1991), and Dirven (1985, 1990) to be typical of the European branch of the Cognitive Linguistic movement, in contrast with the more introspectively conducted studies of American researchers of a Cognitive Linguistic persuasion. From a more theoretical point of view, explicit attention for the way words are actually used would seem to follow straightforwardly from Langacker's characterization of Cognitive Linguistics as a usage-based model that rejects the Chomskyan neglect of linguistic performance (1987: 46). However, Wierzbicka (1985) has coupled a prototypeoriented form of lexical research with an explicit defense of the introspective method. What advantages, then, are there to a referential, usagebased approach that avoids relying exclusively on introspection? Next to the advantages of a usage-based investigation that are enumerated by Ruhl
Classical definability
153
(1989: 13-16), let us note that the introspective method may succeed rather well in pinning down the prototypical core of the items under investigation, but is hardly able to capture the peripheral uses to which the core meanings appear to give rise in actual usage. This is a point to which we will return in section 5.
2. Criteria for classical definability The definitional criterion only works properly when restrictions on definitions exist: there has to be a way of assessing what is a good definition before definability can be invoked as a measure of polysemy. Classical definability implies that a definition can be found that characterizes all the members of the category to be defined, and only those. The definition has to be general, in the sense that it applies to all the members of the category, and it has to be distinctive, in the sense that it adequately distinguishes the category from all others. For instance, let us assume that we are trying to define the category 'bird' (as a biological species). We will then have to list the attributes that all birds have in common, if there are any; further, we will have to make out whether this list of attributes (or any subset of it) suffices to distinguish birds from mammals, reptiles, and fishes, to say the least. As illustrated in Geeraerts (l987a), the attributes that one would be inclined to mention as general characteristics of birds, often do not have the required commonality. On the other hand, the attributes that do seem to be general among birds do not suffice to distinguish birds from other species; even when the features in question are taken together, the duck-billed platypus is a counter-example to the alleged definition. It may be useful to point out here that there are various other ways of terminologically indicating the classical nature of definitions. One is to say that classical definitions define all and only the members of the category, while another is to say that they uniquely define the category. More importantly, however, it has to be noted that applying the definition meets with particular problems in the case of our material. Before turning to actual examples, let us consider each of the two requirements in more methodological detail, with specific reference to the kind of material incorporated in the database described above. The first part of the joint requirement of generality and distinctiveness would seem to be easy to check: the componential description of the referents of each lexical item allows us to check whether there are any attributes
154
PoZVsemy
that these members have in common. There is an important reason, however, for rejecting such a straightforward and mechanical procedure. The descriptive features that define the various configurations in the referential range of a lexical item cannot be taken at face value, but have to be interpreted (as an automatic consequence of which, the issue of classical definability cannot be settled mechanically). There are basically two forms of interpretation to be taken into account. For ease of reference, they will be called the quantitative and the qualitative one. The quantitative interpretation involves numerical dimensions, i.e. dimensions whose values constitute a graded continuum. The crucial point here is to see that it is not the individual value of a specific referent with regard to that dimension that is definitionally important, but rather the range of values with which the dimension occurs. If, for instance, a dimension like WIDTH receives the values [2], [3], and [4] in the semasiological range of application of an item, we should not say that the item has no common feature on the dimension WIDTH, but we should rather say that the width of the referents of the item in question ranges from value [2] to [4]. Although the presence of the values [2], [3], and [4] would superficially suggest that the referents of the item do not have common characteristics as far as their width is concerned, they do upon closer inspection: all of them fall within the range defined by the interval [2]-[4]. On the other hand, a qualitative reinterpretation of the superficially given values involves hidden variables. In particular, whereas all the dimensions in the database are visual ones, there may be covert dimensions of a fimctional nature. For instance, if the MATERlAL dimension of an item features the values [silk] and [cotton], there is again, superficially speaking, no common characteristic. If, however, both silk and cotton are used as light materials serving the purpose of keeping the person cool in warm weather, the common functional feature [light and cool] reduces the original variation on the MATERlAL dimension to epiphenomenal status. The distinctiveness criterion for classical definability should be handled with equal care. To begin with, notice that the distinctiveness requirement crucially involves negative evidence. If a definition is to hold for all and only the members of a particular category, the definition should not apply to any specific thing that does not belong to the category. The distinctiveness of the definition is contradicted, in other words, if we can find a referent that falls within the scope of the definition but that falls outside the scope of the category. This does not mean, to be sure, that the items falling within the scope of the definition could never occur as members of other categories than the one to be defined. For instance, let us define the attributes plusquint and
Classical definability
155
deciminus of natural numbers. A natural number is plusquint if it is larger than five; it is deciminus if it is smaller than ten. Both definitions are classical: they are as mathematically precise as you can get. At the same time, both categories naturally overlap: the natural numbers 6, 7, 8, and 9 fall within both categories. This means that the number 7 may sometimes be called a plusquint and sometimes a deciminus. Suppose further that we have actually encountered both ways of speaking in our corpus of mathematical language, and that we are trying to define the word deciminus. We have noted that all deciminuses share the property of being smaller than ten, and therefore propose to define deciminus accordingly; we also notice, however, that 7 is sometimes called a plusquint. When confronted with such a plusquint instance of 7, could we then say (repeating the sentence introduced above) that 7 falls within the scope of the definition of deciminus, but that it falls outside the scope of the category (because it is not then called a plusquint)? Of course not: falling outside the scope of the category means never occurring within it. The number 7 is not a counter-example to the proposed definition of deciminus because it is occasionally called a plusquint, but it would be a counter-example if it were never called a deciminus. In the same way, the duck-billed platypus is only a counter-example to the classical definability of bird because it is never categorized as a bird. At the same time, the fact that not all deciminuses can be called plusquint implies that the definition of the latter category should not be so broad as to include the entire range of application of deciminus; it must include those deciminuses that are sometimes called plusquint, but it must exclude those that are never so called. As a practical consequence of this observation, we will have to check any alleged classical definition of a lexical item against the words with which the item referentially overlaps. In particular, the definition should not be overly general, in the sense that the entire overlapping category (rather than just the intersecting part) is drawn into the category to be defined. The relationship between two items that share referents may, however, take other forms than the kind of overlapping (partial co-referentiality, semi-synonymy) that is illustrated by the plusquintldeciminus case. Systematically, there are three other relations to be envisaged. When the items are synonymous, no problem arises when the definition of the definiendum covers the entire range of application of the second item. Similarly, when the definiendum is a hyperonym of the second item, the definition may (in fact, must) cover all the referents of the second item. But when the definiendum is a hyponym of the other word, a definition that exceeds the referential boundaries of the hyponymous
156
Polysemy
item will have to be rejected. To summarize, the distinctiveness criterion does not apply to the synonyms and the hyponyms of the definiendum; in the case of overlapping and hyperonymous categories, it should only be applied to the overlapping and hyperonymous categories as a whole, not to those subsets of the latter that they share with the definiendum.
3. Legging: a case of classical definability Now that we have a better idea of how the classical definability of lexical items can be established, actual examples can be considered. In the following pages, the lexical item legging will be considered in detail as an illustration of the methodology outlined above. The other lexical items that will be analyzed further on in this section will be treated more succinctly; rather than concentrating on the analytical procedure itself, we will then concentrate on the results of the analysis. Table 1. The semasiological range of legging H3118.v H4118.v H41186v H4211.v
[3] [36] [3] [ 1]
H5118.v H51186v H5128.v H5154.v
[58] [2] [1] [1]
H5211.v H52115v H52116v
[3] [ 1] [1]
The semasiological information that can be extracted from the database takes the form of a list of referential descriptions of the form illustrated in Table 1. The table lists all the referential configurations with which the item legging occurs, together with their respective frequencies, which are mentioned between square brackets. The numbers refer to the five dimensions illustrated in the previous section. The feature v specifies that the item is worn by a woman. And a dot instead of a bumber indicates that the relevant feature is not relevant; in this case, it means that the dimension Detail that was introduced above is not applicable to the item in question. Finding out whether legging can be classically defined would now seem to follow a straightforward procedure: first, it will have to be established whether there are any characteristics that are common to all referents of legging, and second, it will have to be investigated whether the resulting definition is sufficient to distinguish legging from all other categories that are neither hyponyms nor synonyms of legging. On the basis of this proce-
Classical definability
157
dure, legging would definitively turn out not to be classically definable. A glance at the table suffices to appreciate that the only truly general characteristics of all the listed instances of legging are the fact that they all involve trouser-like gannents, as represented by the feature [H], and the fact that they are worn by women, as represented by the feature [v]. At the same time, the database contains various trousers worn by women that are never called legging; specifically, wider and shorter types of trousers for women fall outside the category. But obviously, we have not yet subjected legging to the quantitative and qualitative reinterpretation process that we described above. Let us now try to establish whether we can salvage the classical definability of legging by using a more refined approach. A quantitative reinterpretation is important for the dimensions WIDTH and LENGTH. The width of the referents of legging varies between the values [1] and [2], which is to say that leggings are either tight or narrow. The dimension LENGTH has a range between values [3] and [5]; the referents of legging reach down at least to the knee, but they may also cover the entire leg down to the ankle. The impact of a qualitative reinterpretation can be appreciated when we have a look at the dimension MATERlAL. The fact that the predominant value on this dimension, viz. [8], refers to stretchy fabrics suggests that there is a causal connection with the dimension WIDTH: leggings are mostly tight or narrow precisely because they are made of elastic material. Could it be the case, then, that the feature [elastic] allows for a reduction of the variation on the dimension MATERlAL? A renewed consultation of the original pictures on which the database records were based, reveals that this is indeed the case. On the one hand, the records that contain the value [1] on the dimension MATERlAL appear to be made of a finely woven tricot that is at least moderately elastic. On the other hand, the single record that features the value [4] for MATERIAL involves a stretchy, very tight-fitting corduroy. In other words, although not all leggings have the same degree of elasticity, they do share a certain amount of stretchiness. The resulting picture of the common characteristics of all instances of legging may be summarized as follows: a legging is a two-legged outer garment for women covering the lower part of the body from the waist down, ranging in length from the knee to the ankle, and made from elastic materials such that the width of the legs ranges from tight to narrow. The next step involves checking whether this set of common features is sufficiently distinctive to act as a definition of legging. In order to get an idea of the lexical items that have to be included in an analysis of the distinctiveness
158
Polysemy
of the definition of legging that was given above, Table 2 lists the onomasiological alternatives with which legging-configurations occur in the database. For each of the various configurations that are situated within the semasiological range of legging, Table 2 specifies the other lexical items referring to that configuration, together with their frequency. Table 2. The onomasiological alternatives for legging H3118.v
broekje J, kniebroekje J, legging 3, piratenbroek J, wielrennersbroek J
H4118.v
broek 12, broekje I, calec;on J, denimbroek J, jeans J, kuitbroek J, legging 36, leggings 9, piratenbroek 3, stretchbroek J, tricotbroek J broek 2, calec;on J, legging 3, leggings 2
H41186v H4211.v
broek J 7, calec;on J, kuitbroek J, legging J, pantalon J, streepbroek J
H5118.v
broek J2, broekje J, legging 58, leggings 23, skibroek 3, stretchbroek J, stretchleggings J, tricotbroek 5
H51186v
broek 6, calec;on 2, legging 2, leggings 7, pantalon J
H5128.v
broek J, joggingbroek J, legging J, skibroek 2
H5154.v
legging J
H5211.v
bloemenbroek J, broek 74, calec;on J, jeans JJ, legging 3, pantalon 2
H52115v
broek 24, legging J, pantalon Uricotbroek J
H52116v
legging J, broek 3
The status of the onomasiological alternatives in terms of their lexical relationship with regard to legging is specified in Table 3. In accordance with the methodology set out above, we need not worry about a possible lack of distinctiveness of the definition of legging with regard to hyponyms like kabellegging ('cabled !egging', legging decorated with a cable pattern) and stretchlegging ('!egging in stretchy fabric'), or synonymous expressions like leggings and cale90n. (In passing, it may be noted that the latter term is geographically restricted to the Belgian sources, in contrast with the magazines published in The Netherlands.) The hyperonym broek does not present a real danger either, because the definition of legging specifies an actual subset of the range of application of broek (which may, in particular, refer
Classical definability
159
to two-legged gannents that are much shorter and much wider than the referents of legging). Table 3. The relations between legging and its onomasiological alternatives
Synonyms Hyperonym Hyponyms
leggings, caler;on broek kabelleggings, stretchleggings
Semi-synonyms
kuitbroek, kniebroekje, broekje jeans, denimbroek, stretchbroek, tricotbroek streepbroek, bloemenbroek piratenbroek, wielrennersbroek, skibroek, joggingbroek, pantalon
As suggested in Table 3, various subsets may be distinguished within the set of semi-synonyms. A first set comprises kuitbroek 'calf-long pair of trousers', kniebroekje 'knee-long pair of trousers', and broekje 'short pair of trousers, shorts', which each delimit a specific area within the range of application of broek on the basis of the length of the gannent. The second set consists of jeans, denimbroek, stretchbroek, and tricotbroek, which each refer to trousers made of a particular fabric. In the third set we find items that refer to a particular type of motif or decoration: streepbroek indicates the presence of stripes, and bloemenbroek signals the presence of flowers. The fourth subset contains piratenbroek 'pirate's trousers', wielrennersbroek 'cyclist's trousers', skibroek 'ski pants', joggingbroek 'jogging pants', and pantalon 'pair of trousers'. There is an interesting distinction between the fourth subset and the previous three, in the sense that the items in the latter may be adequately defined as a hyponym of broek by referring to a single dimension (length, material, and motif respectively). These dimensions detennine the morphological structure of the words, in the sense that the first member of the compound refers to the specific value on the relevant dimension that is crucial for the item. For the items in the fourth subset, however, various dimensions have to be mentioned at the same time in order to specify their proper position. In the cases involved in the first three sets, establishing the referential overlap with the semasiological range of legging is a relatively straightforward matter. The items are actual semi-synonyms to the extent that they may refer to pieces of clothing that are not legging-like on any dimension that is irrelevant for the item in question. For instance, referents of kuitbroek that
160
Polysemy
are not tight-fitting enough to fall within the definition of legging are not so called either. Kuitbroek is characterized on the basis of the dimension LENGTH, and its specific value on this dimension happens to fall within the range of lengths that is definitional for legging. Because of its 'onedimensional' nature, however, it is not definitionally specified with regard to other dimensions that are subject to restrictions in the case of legging (such as, in the example, WIDTH), and it may therefore refer to pieces of clothing that are definitely too wide for leggings. We have to make methodological allowances, though, for the fact that the number of records we have for the various items does not always suffice to establish their overlapping status beyond all doubt. That is to say, we may not have enough examples of an item like bloemenbroek to establish whether it does indeed occur with referents that are not leggings. In these cases, we have nevertheless listed the item as a semi-synonym on the basis of the assumption that the morphological structure of the item is a good indication of its semantics - on the basis of the assumption, for instance, that trousers with a flower motif may be called bloemenbroek regardless of their other characteristics. For the items of the fourth subset identified above, such an abductive underpinning of the classification on the joint basis of intuition and morphological structure is less obvious. Although most of the referents of skibroek are relatively tight-fitting, the widest ski pants are definitely wider than the widest referent of legging; skibroek, in other words, has a wider range on the dimension WIDTH than legging. The same criterion also distinguishes joggingbroek from legging. For wielrennersbroek, on the other hand, the most important distinctive dimension is LENGTH: whereas legging ranges from dimensional value [3] to [5], the referents of wielrennersbroek are never longer than the knee. Similarly, piratenbroek refers to garments that may be as long as the knees or the calves, but never as long as the ankles.
4.
Vest etc.: absence of classical definability
The discussion of legging shows that a careful analysis of semasiological ranges and lexical relations is necessary to determine the classical definability of lexical items. The legging example produces positive results, in the sense that the item in question appears to be classically definable. The next step will be to discuss a number of cases where the definitional analysis yields negative results, in the sense that the items in question cannot be defined on a classical basis. The discussion will be based on a subset of the
Classical definability
161
field of clothing terminology. In particular, we will consider pieces of clothing that cover the upper part of the body, that can be entirely opened at the front, and that are never worn as the first layer of clothing above the underwear. The items with the highest frequencies in this subset are jack, colbert, blazer, jasje, and vestNL . The distinction between vestNL and vests is necessary because there is a marked difference between the ways in which the item vest is used in the Belgian and the Netherlandic sources; we will come back to this point below. Table 4 gives an overview of the ranges of application of the items. The table is based on individual analyses of the items along the principles demonstrated in connection with legging. That is to say, the dimensions used in the table do not necessarily reproduce the information structure of the database in a straightforward maJmer, but may be the result of a reinterpretation of the stored data or even a reconsideration of the original pictures. Dimensions that are not distinctive within the subset have been left out. For instance, all the types of clothing included in the subset are worn by men and women alike; accordingly, the dimension SEX has not been retained in the table. A plus sign means that a particular dimensional value occurs within the range of application of the item; a minus sign indicates that it never occurs. Thus, plus signs on all values of a particular dimension mean that both values may occur. For instance, the referents ofjasje may either occur with a type of fastening that can be fastened up to the neck, or with a type of fastening that stops on the chest somewhat lower than the neck; by contrast, jack is never used as a name for garments that cannot be fastened entirely. The only dimensions in the overview for which the relevant values might have to be restated in terms of ranges are LENGTH and FASTENING, since both involve measures of length; for instance, the referents of jasje have a fastening whose length ranges from up to the chest to up to the neck. The other dimensions consist of discontinuous values. A first thing to note is that the referential ranges included in Table 4 suggest the existence of certain hyponymy relations. It appears, for instance, that all dimensional values that occur in the range of jack also occur in the range of jasje; at the same time, the latter item exhibits a number of dimensional values that are absent in the case of jack. In this particular case, the suggestion that jack is a hyponym of jasje (because the referential range of the latter word includes that of the former) is supported by the intuition that jasje is a cover-term for the entire set of items included in 4. However, the overview in the table is not really a good way of settling the hyponymy relations among the items, because the referential ranges are being considered in
162
Polysemy
tenus of separate dimensions rather than dimensions in combination. Consider a fictitious case in which an item A is represented by the referential types [ac] and [bd], and an item B by the types [ad] and [bc]. In both cases, the first dimension ranges over the values [a] and [b], and the second dimension over the values [cl and [d]. Judging on the basis of an overview of dimensional ranges, then, A and B would be synonymous, since they have the same dimensional ranges. Judging on the basis of the dimensional values as they occur in combination, however, it becomes clear that there is neither a relationship of synonymy nor hyponymy between both items. It is therefore necessary to establish hyponymy relations on another basis than Table 4 as such, Table .;, The semasiological ranges of jack, blazer, colbert, vestNv jasje, jack
colbert
blazer
vestNL
jasje
+
+ + +
+ + +
+ + +
+ + +
+ + +
+ + +
+
+ +
+ +
+ + +
+
+
+ +
+ +
LENGTH
shorter than the waist as long as the waist lower than the waist
+ +
CUT
blousing wide and straight narrow and straight waisted
+ +
+ +
MATERIAL
woven fabrics knitted leather
+ +
FASTENING
up to the neck lower than neck
+
A corpus-based approach for the recognition of hyponymous relations may be established as follows. If A is a hyponym of B, B may occur as an alternative name for all referents of A. Of course, B need not be as frequent as A for the referential set in question, because A may be more entrenched than B. Also, it may be expected for statistical reasons that the less common referential types of A may not occur in the corpus with B as an alternative
Classical definability
163
denomination; in actual practice, it may be sufficient to establish that B occurs as an alternative for the most common referents of A. Table 5. VestNL andjasje as onomasiological alternatives for blazer. blazer: configurations
B1l22v B1222v B2111m B2111v B2112v B2121m B2121v B2122v B2131v B2132v B2212v B2221m B2222v B2232v C2212v C1312v C2311v C2312v C2322v C2332v C2412v C3212v C3312v C3332v C3412v C3432v
frequency of configuration
vest as alternative
10 10
2 1 17 5
14 48 3 8 12 3 22 5
1 1
+
I 7
+ + + +
1 1 1 1 14 1 4 1
+ +
+
jasje as alternative
+ + + + + + + + + + + + + + + + + + + + + + + + + +
In Tables 5 and 6, such overviews of onomasiological alternatives are given for blazer and colbert. Given the statistical margin that was just mentioned, it can be deduced from the table that colbert is a hyponym of blazer and jasje, and that blazer is a hyponym of ja5je. At the same time, of course, it should be established that there are cases of B that are not named by means of A (lest a situation of synonymy rather than hyponymy obtains). This type of information, however, can be safely derived from overviews like
164
Polysemy
the one in Table 4: for instance, the plus sign on the 'knitted' value of the dimension MATERIAL for blazer as opposed to the minus sign for colbert indicates that knitted referents of blazer never occur with the name colbert, for the simple reason that the referential range of colbert does not include knitted garments of any kind. Table 6. Blazer, vestl\JL andjasje as alternatives for colbert. colbert: configurations B1121m B1122v B2111m B2121m B2121 v B2122v B2131m B2211m B2212m B2221m B2222m B2222v C3212m C3311 v C3312v
frequency of configuration 1 7 2 52 4
17 5 1 1 13 1
blazer as alternative
vest as alternative
jasje as alternative
+ + + + +
+ + + + + +
+
+ + +
4
+
1 3 1
+
+
+ +
Repeating the procedure illustrated in 5 and 6 for the items jack and vest leads to the hyponymy relations that are charted in Figure 1. (The label '+lh' indicates that there is a relationship of hyponymy between the items in question, given that a lexical test of hyponymy as illustrated in 5 and 6 is used. The label '-lh' signals the absence of hyponymy according to the lexical criterion.) The definitional question regarding the five items can now be made more precise: can the items be classically defined on the basis of the overview in 4 without obscuring the lexical relations summarized in 7? This involves reviewing all possible classical definitions of the items and checking whether they respect the relations in 7. Note, however, that only two of the four dimensions included in Table 4 can be used to establish classical definability. Whereas the values on the LENGTH and the FASTENlNG dimension can be expressed in terms of ranges, any reference to the other two dimensions automatically involves disjunctive
Classical definability
165
values; therefore, including these dimensions in the definition inevitably turns the definition into a non-classical one. Thus, it will be sufficient to investigate only the potential definitions that refer to the dimensions LENGTH and FASTENING. A further restriction follows from the specific position of jasje. Both from the point of view of the featural ranges given in 4 and from the point of view of the lexical relations specified in 7, jasje appears to act as a coverterm with regard to the other items: it is a hyperonym of the other four, and the row of plus-signs accompanying jasje in 4 indicates that it indeed includes all definitional possibilities that are relevant for the other items. In this sense, the search for classical, distinctive definitions may be confined to the other four items. In all of the potential definitions mentioned below, then, jasje 'jacket' will be used as a cover-term; roughly, it may be defined as referring to garments that cover the upper part of the body, that can be entirely opened at the front, and that are never worn as the first layer of clothing above the underwear. hyponym jack
co/bert
blazer
vestNL
hyperonym colbert
-Ih
blazer
-Ih
vestNL
-Ih
jasje
+Ih
Figure 1. Hyponymy relations between jack, blazer, colbert, vestNL,jasje.
The definitions that are to be taken into account may be systematically grouped together in three sets: definitions that refer only to the dimension LENGTH, definitions that refer only to the dimension FASTENING, and definitions that involve both dimensions at the same time. For ease of reference, we will call jack] the definition that involves LENGTH, jack2 the definition that involves FASTENING, etc. All in all, the following twelve definitions have to be considered.
166
Polysemy
Jack] Jasje whose length ranges from the region of the hip to the region of the upper part of the legs Jack2 Jasje that can always be fastened up to the neck Jack] Jasje whose length ranges from the region of the hip to the region of the upper part of the legs, and that can always be fastened up to the neck Colbert] Jasje that reaches down to the region of the upper part of the legs Colbert2 Jasje that can only be fastened as far as the chest (but not up to the neck) Colbert] Jasje that reaches down to the region of the upper part of the legs and that can only be fastened as far as the chest (but not up to the neck) Blazer] Ja.sje whose length ranges from the waist to the region of the upper part of the legs Blazer2 Jasje that can only be fastened as far as the chest (but not up to the neck) Blazer] Jasje whose length ranges from the waist to the region of the upper part of the legs and that can only be fastened as far as the chest (but not up to the neck) Vest] Jasje whose length ranges from the waist to the region of the upper part of the legs Vest2 Jasje with a fastening whose length ranges from the up to the chest to up to the neck Vest] Jasje whose length ranges from the waist to the region of the upper part of the legs and that has a fastening whose length ranges from up to the chest to up to the neck
These twelve definitions may occur in 81 (=3 4) combinations. That is to say, the general definability question boils down to 81 questions of the type: if jack is defined asjackj, calbert as calbert], blazer as blazer2, and vest as vest], do the lexical relations that follow from these definitions then conform to the actual relations that are summarized in figure I? Or, in other words, is there at least one combination of the twelve classical definitions that makes
Classical definability
167
the right predictions about the attested lexical relations? Rather than considering all 81 possibilities separately, it can be shown in the following way that the question has to be answered in the negative. First, consider all combinations of two elements from among the set of definitions that refer only to LENGTH. The co-occurrence of jack] and colbert] is to be excluded, because this would counterfactually imply that colbert is a hyponym of jack (as the range of LENGTH for jack as defined includes the range as defined for colbert). The co-occurrence of jack] and blazer] is to be excluded because it would imply that jack is a hyponym of blazer. The co-occurrence of jack] and vest] is to be excluded because it would imply that jack is a hyponym of vest. The co-occurrence of colbert] and vest] is to be excluded because it would imply that colbert is a hyponym of vest. And the co-occurrence of blazer] and vest] is to be excluded because it would imply that blazer and vest are synonymous. Second, consider all combinations of two elements from among the set of definitions that refer only to FASTENING. The co-occurrence ofjack2 and vest2 has to be excluded because it implies that jack is a hyponym of vest. The co-occurrence of colbert2 and blazer2 has to be excluded because it implies that colbert is a synonym of blazer. The co-occurrence of colbert2 and vest2 has to be excluded because it implies that colbert is a hyponym of vest. And the co-occurrence of blazer2 and vest2 has to be excluded because it implies that blazer is a hyponym of vest. Third, consider all combinations of two elements from among the set of definitions that refer to both LENGTH and FASTENING. The co-occurrence of jack] and vest3 has to be excluded because it implies that jack is a hyponym of vest. The co-occurrence of colbert] and vest] has to be excluded because it implies that colbert is a hyponym of vest. And the co-occurrence of blazer] and vest] has to be excluded because it implies that blazer is a hyponym of vest. The alternatives that remain at this point are summarized in Table 7. It is now immediately obvious that there can be no combination of four classical definitions that respects the existing restrictions, if only because all possible definitions of vestNL have already been ruled out. We may conclude, then, that there is no set of classical definitions for jack, colbert, blazer, and vest that sufficiently distinguishes the items among each other and that respects the lexical relations that appear to exist among them. Whether this is the dominant situation in the lexicon is difficult to say on the basis of our material; after all, we have only been able to examine a few lexical categories. One general conclusion, at least, is that indeed not all lexical categories can
168
Polysemy
be classically defined. In addition, it is worthwhile to point out that the definability issue seems to be strongly influenced by the specific subfield of the field of clothing terminology that is being considered. The 'skirts' -subfields, for instance, contains classically definable categories, whereas the subfield consisting of shirts, blouses, t-shirts, and their likes is as unclassical as the subfield analyzed in the previous pages. Table 7. Combinations of classical definitions for jack, colbert, blazer, vest. blazer
vest
2
3 3
5.
Methodological discussion
What we have tried to illustrate in the foregoing pages (apart from the fundamental fact that non-classical definability is a real phenomenon) is the importance of a rigorous procedure in answering definability questions. Apart from the importance of distinguishing between polysemy and univocality, there are two requirements that discussions of definability should strictly adhere to. First, the generality of classical definitions implies that no disjunctive features are included in the analytical definitions, and second, the distinctiveness of classical definitions implies that they should make the right predictions about the lexical relations among the items involved. As both the legging example and the jasje-subfield show, complying with these demands requires a careful, step by step procedure that stands in sharp contrast with two other approaches to the relationship between polysemy and definitionality. In fact, while the approach demonstrated here applies the classical criterion of dejinability to denotational data based on non-elicited actual usage, the two most outspoken recent defenders of monosemy appear to drop either the one or the other of these two features from their methodology On the one hand, Ruhl (l989b) accepts the importance of usage-based data, but gives up on the necessity of actually defining the presumed single meaning of an item. If all the instances of use ofa lexical item are claimed to
Classical definability
169
exhibit the same meaning, the least one should do to make this statement falsifiable is to define the semantic value in question. Ruhl, however, explicitly denies the possibility of doing this. At the end of his analysis of the English verb bear, for instance, he writes: So what does bear mean? It should be clear by now that this question cannot be answered in words: there is no single word or phrase that can comprehensively capture exactly what bear contributes. I hope, by trying to show the unity of bear's contexts, to have revealed a unified meaning; but such a conclusion is inferential (1989: 63). Clearly, such an approach gives up on precision in favor of impressionism. For general methodological reasons of comparability and falsifiability, such a strategy has to be rejected in favor of the attempt to achieve descriptively adequate definitions. If it is not even possible to definitionally identify the allegedly unitary meaning of a lexeme, how can its unitary status be tested at all? This is not, to be sure, a mere rhetorical question: see Geeraerts (1993a) for an overview of other approaches to polysemy apart from the definitional one. Ruhl, however, dismisses the definitional criterion without being explicit about an alternative that is methodologically sound in the sense of being precise enough to allow for falsifiablity. On the other hand, Wierzbicka' s approach retains the importance of precise definition, but rejects the denotational, usage-based approach. Such an explicitly introspective strategy runs the risk of being imprecise with regard to the actual range of application of an item. As an illustration, consider Wierzbicka's definition of dress (1985: 382): A KIND OF THING MADE BY PEOPLE FOR WOMEN AND GIRLS TO WEAR. IMAGINING THINGS OF THIS KIND PEOPLE COULD SAY THESE THINGS ABOUT THEM:
they are made to be worn on the body, below the head, to cover most of the body so that all the parts of a woman's body which people think should not normally be seen are covered with that one thing and to protect most of the body with undesirable contact with the environment and to cause the woman wearing it to look good they are made in such a way that when they are on the body the lower half surrounds the lower half of the woman's body from all sides so that the legs are not separated from one another and so that the genital area of the woman's body seems to be hidden and so that women wearing things of this kind look different from men
170
Polysemy
things of this kind are thought of as something suitable for women to wear in most kinds of places and in most kinds of circumstances. When we have a look at the actual garments that occur in our database as instances ofjurk (the Dutch equivalent of dress), we find cases in our material that do not conform to the description. If, for instance, 'covering most of the body' is interpreted as 'covering more than 50% of the body', then a number of very short summer dresses with open backs and low necklines do not display the feature in question. And if 'the parts of a woman's body which people think should not normally be seen' include the upper part of the thighs, then dresses with long side slits contradict the image. Furthermore, some dresses have such wide armholes and such a plunging decolletage that they could not normally be worn without exposure of the breasts (unless they are worn with an additional t-shirt or blouse underneath). The comparison shows, in other words, that the description proposed by Wierzbicka may well be adequate for the majority of cases in the range of dress, but does not really cover all possible instances. Admittedly, such a comparison is risky for at least two reasons. First, we start from the assumption that English dress and Dutch jurk are equivalent as far as their referential range of application is concerned. As long as we do not have a similar corpus-based analysis of dress as the one we have made for jurk, the comparison will have to remain a conditional one. Second (and more importantly), it is not even certain that Wierzbicka actually intends the definition to apply to all the cases in the extension of dress. By introducing the phrase 'imagining things of this kind people could say these things about them', the perspective is shifted from the objective features of the things that are being called dresses to the subjective image that people say they have about dresses when they are asked for it. In a sense, Wierzbicka defines dress by referring to what people think dresses are. And if what people think dresses are only involves the central cases of the category'dress', then, of course, it makes no sense to complain that the description of this mental image does not apply to non-prototypical dresses: it never intended to do so anyway. On this reading of Wierzbicka's view, its reference to subjective images could be construed as implying a conscious restriction of the description to the prototypical core of the category. And because introspection probably does work efficiently for retrieving such prototypical images, the introspective method may be salvaged. It is not quite clear, however, whether this interpretation ofWierzbicka's position is a valid one. On the one hand, she argues that 'a valid definition must be empirically adequate, that is, it must be phrased in such a way that
Classical definability
171
it covers the entire range of use of a given word, expression, or construction' (1989: 738). On the other hand, if a definition such as that of dress is to be applicable to all things that may be called by that name, people should be able to assert all the characteristics mentioned in the definition any time they see a dress. But surely, when imagining a less prototypical kind of dress than the kind whose features are included in the definition, people will not imagine it as a prototypical case. What people could say about dresses changes when peripheral members of the category are at stake: default dresses, for instance, may well cover most of the body, but that feature may be suppressed when a fancy type of sununer dress is involved. Even if, however, we accept that Wierzbickian definitions of the kind illustrated above are explicitly restricted to the prototypical core of the categories, a counterargument may be advanced in favor of a method based on observing actual usage. Note that it remains a matter to be settled empirically whether the lexical knowledge that people have in their in minds is indeed restricted to a mental image of the core of the category in question. It is not a priori given that the idea of a category that people may introspectively retrieve from memory is an adequate reflection of the extent of that person's actual knowledge of the category. On the contrary, if it is part of his knowledge to produce or accept an application of dress to nonprototypical cases, then he 'knows' more about the category than would be included in his introspectively retrieved idea of the category. That knowledge, to be sure, is not necessarily conscious knowledge; it is less 'knowledge that (lexical item x may refer to entities with such and such characteristics)' but rather 'knowledge how (lexical item x may be successfully used)'. In order, then, to get a better grasp on the lexical 'knowledge how', usagebased investigations of the type illustrated in this article are vital, precisely if it is suspected that conscious knowledge may only partially cover the full extent of a person's 'knowledge how' . In short, the approach demonstrated above is inspired by the desire to steer clear both of the danger of theoretical imprecision that is implicit in Ruhl's strategy, and of the danger of empirical incompleteness that is implicit in Wierzbicka's strategy. But if the approach followed here is indeed to be preferred for general methodological reasons, what are the findings that it leads to?
172
Po~ysemy
6. Conclusions If the methodological framework defended in section 5 is accepted, the investigation carried out in section 4 shows that not all lexical items can be classically defined in terms of necessary, jointly sufficient features. If classical definability is considered a criterion for monosemy, this conclusion further implies that not all lexical items are monosemous. At the same time, however, the investigation suggests that classical definability is not necessarily a psychologically acceptable criterion for monosemy. The items whose definitionally polysemous nature was determined in section 4 are not ones that are intuitively recognized as polysemous. In this sense, the investigation supports and further illustrates the point made in Geeraerts (l987a) and (l993a): the different criteria for polysemy that are quite plausible taken by themselves may be mutually contradictory; specifically, vest etc. are definitionally polysemous, but not intuitively. This implies, in other words, that lexical semanticians cannot at the same time maintain the monosemic bias and the ideal of classical definability. If, on the basis of the monosemic bias, it is accepted that vest's intuitive univocality is a theoretically viable reflection of an actual monosemy, the ideal that an individual meaning can always be classically defined has to be relaxed (because there is no classical definition for vest in the usage under investigation). Conversely, if one sticks to the idea that classical definability is an integral part of what it is to be a distinct meaning, the monosemic bias will have to be relaxed to the point where even an intuitively non-suspect case of univocality like vest is accepted to be polysemous.
Section 3 Constructions and idioms
Chapter 7 The semantic structure of the indirect object in Dutch
Originally published in Willy Van Langendonck and William Van Belle (eds.), 1998, The Dative 11. Theoretical Approaches 185-210. Amsterdam/Philadelphia: John Benjamins. To a considerable extent, Cognitive Linguistics is characterized by a strategy that extrapolates models developed in the study of the lexicon to other areas of linguistics. (Structuralism, by contrast. is characterized by a strategy that extrapolates models of phonological analysis, like the use of distinctive features and distributional analysis, to the rest of linguistic structure.) Specifically in the domain of semantics, productive models of meaning description like prototype theory matured in the field of lexical semantics and were then applied to grammatical phenomena. This may be clearly seen in the area of Construction Grammar as represented by Goldberg (1995), Croft (2001), Fried and Ostman (2004), Ostman and Fried (2005). Constructions are so to speak grammatical patterns treated as lexical items: more or less schematic entities that should be treated as symbolic units in their own right, with properties that cannot just be compositionally derived. But if that is indeed the case, competing models for semantic description such as may be found in lexical semantics are likely to play a role in Construction Grammar as well. In the present paper, I argue that the multidimensional form of prototype-based analysis that is illustrated in Chapter 3 is relevant for the description of constructions just as well as for the description of lexical items. Such a multidimensional model, in which co-variation of semantic shifts along different dimensions constitutes the grid of a prototype-based network, contrasts with a simpler type of radial network analysis (as represented by Goldberg's analysis of the dative construction) in which the crucial elements of the network are points (meanings, readings, senses) rather than dimensions. Methodologically speaking, the paper is based on introspectively derived examples rather than corpus-based materials. and that is clearly not the way in which I would proceed at the moment. Current work by Colleman (2005) shows however that the semantic structure for the Dutch dative construction suggested here may be fruitfully applied to corpus data.
176
Constructions and idioms
1. Aims and background ofthe exercise The purpose of the present article may be defined from two points of view: a more narrow descriptive one, and a broader theoretical one. Narrowly defmed, the paper intends to show that the semantic structure of the indirect object construction in Dutch can be fruitfully described, a) by starting from a spatial prototype (the indirect object as the recipient of a spatial transfer) b) by deriving further uses from this basis by taking into account four specific conceptual mechanisms of semantic change, viz. generalization, metaphor, metonymy, and subjectification c) by systematically exploring the various relations of these mechanisms, i.e. by describing the field of application of the indirect object as a multidimensional space. Broadly defined, the paper will try to bring home three general theoretical points about the semantic structure of syntactic constructions. The first point implies that such constructions can be semantically described by means of basically the same descriptive mechanisms that apply to lexical categories. The second point involves the view that the semantic structure of categories is best described as a multidimensional space. The third point consists of the recognition that the semantic description of syntactic constructions has be intimately linked to their syntactic structure. All three points require some additional discussion. The first point is a straightforward consequence of one of the major methodological strategies employed by Cognitive Linguistics, viz. to model grammatical research on the basis of lexical research: the study of lexical categories is taken as a methodological point of departure for the study of categorization in the grammar at large, and grammatical construction types are considered categories just like lexical items. Given that linguistic categorization is the major focus of attention for Cognitive Linguistics, studying the lexicon first is, in fact, a plausible step to take: the categorizing function of the lexicon has received more attention in the linguistic tradition (and is also, perhaps, easier to investigate) than that of grammatical constructs. The intention to study the grammar of the language along the same lines as its lexicon obviously leads to a specific kind of grammatical theory, viz. one in which the grammar, like the lexicon, is conceived of as an inventory of symbolic elements (i.e. meaningful units). At this point in time, this idea has received its clearest and most elaborate form of expression in Langacker's
The indirect object in Dutch
177
Cognitive Grammar (1987, 1990a, 1991), and in Fillmore's Construction Grammar (Fillmore, Kay and O'Connor 1988, Goldberg 1994). A major difficulty for such a research strategy involves the formal identification of the grammatical constructions. In the domain of lexical research, it is usually quite clear what separate entities have to be distinguished: in spite of (stubborn but relatively marginal) problems surrounding the exact demarcation of homonymy, the fact that lexical items have a constant phonological form by and large suffices to identify what the basic units of the analysis should be. In the domain of grammatical research, however, the formal side of the categories/constructions cannot receive such a straightforward definition. For one thing, the grammatical categories do not necessarily have a constant phonological form: word classes, for instance, may well exhibit a type of formal constancy in terms of their distributional and morphological properties, but these characteristics are already abstract properties that do not correspond in a one-to-one fashion with the presence of an identifiable and constant phonological form. For an other, we tend to think of the formal side of grammatical constructions as being subject to variations and transformations (in a non-technical sense), but it is not immediately obvious which changes and alternations are allowed, and which ones rather undermine the formal identity of the construction. The purpose of the present article is definitely not to resolve this question, but it should be pointed out from the start that the problem has immediate consequences for the task at hand. Traditionally, various construction types have been subsumed under the label 'indirect object': ditransitive constructions featuring both a direct and an indirect object, some intransitive constructions, and some prepositional constructions. Against the background of the methodological question that we just raised, it is not clear whether it is entirely justified to lump together these various constructions: does the indirect object exist at all in its traditional definition, incorporating , some cases of , and constructions of the type or ? The present article will not try to answer the question, but rather circumvent the problem by focusing on the ditransitive constructions (which will belong to the hard core of the construction in any event). There will be some remarks about intransitive indirect objects (that is, indirect objects in a V NP frame rather than a V NP NP frame), but none about the prepositional indirect object. As for the demarcation of the single NP indirect object constructions in contrast with single NP direct object constructions, the paper relies on the groundwork laid by Van Belle and Van Langendonck (1992).
178
Constructions and idioms
The second theoretical point mentioned above involves the question what kind of semantic structure best describes the semasiological range of grammatical constructions (or, as the case may be, lexical items). Although the representational formats used in Cognitive Semantics basically have the same representational power (see Geeraerts 1995a), it is nevertheless the case that some types of representation inadvertently favor a particular type of analysis. Radial set representations of the type promoted by Lakoff (1987), for instance, would seem to suggest a two-dimensional analysis of semantic structures: starting from a prototypical centre, new readings are derived directly or indirectly from the prototype in what ultimately constitutes a two-dimensional plane of interconnected senses. As against this impression of two-dimensionality, it has been argued that a more adequate picture of the semantic structure of lexical items has to take into account the multidimensional nature of semantic extensions: the structure of polysemous categories is characterized by covariation on multiple dimensions. A case in point is the reanalysis suggested in Geeraerts (1992a) of the radial set representation of the Dutch preposition over proposed by Cuyckens (1991); a related argument for a multidimensional approach is to be found in Deane (1993). The point to make, to be precise, is not one about representational formats: it is not impossible, for instance, to adapt a radial set representation in such a way that the multidimensional character of the semantic structure is revealed (see, for instance, the way in which Brugman 1989 enriches a basic radial set representation with a type of componential analysis that reveals the multiple links between the readings in the radial set). Rather, the point is an empirical one: the description of semasiological structures (of lexical items and grammatical constructions alike) requires specific attention for the multidimensional nature of the semantic links among the readings of the item or the construction. In the case that interests us here, a multidimensional analysis of the indirect object construction will be contrasted with two related analyses that do not pay as much attention to this multidimensional character, viz. Goldberg's radial set analysis of the English ditransitive construction (Goldberg 1992), and Rudzka-Ostyn' s hierarchical network analysis of the Polish dative (Rudzka-Ostyn 1996). The third theoretical point mentioned above might seem to be uncontroversial: it is a rather obvious idea that the semantic description of grammatical constructions had better be systematically linked to their formal structure. More concretely, a construction-based definition will have to take a relational form: if the indirect object constituent is explicitly defined as part
The indirect object in Dutch
179
of the constructional frame , then it should receive a definition that includes a reference (however indirectly) to the other constituents in the construction. Given the overall plausibility of this idea, the present paper will highlight (by way of example rather than through elaborate theoretical discussion) the connection between this point and the previous one: it will be shown that the relational character of the definition of the indirect object constituent plays a crucial role in the multidimensional nature of its semantic structure.
2. The overall structure of the analysis The basic goal of the analysis is to show that the indirect object construction is semantically cohesive: the various instantiations of the category are linked to each other by means of semantically plausible links, just like the senses of ordinary lexical items are linked to each other on through semantic associations. The links that will be envisaged here largely belong to the traditional repertoire of semasiology as established by the prestructuralist tradition of diachronic semantics: we will make use of processes of generalization, of metonymy, and of metaphor. In addition, we will invoke the notion of subjectification, which is a more recent addition to the inventory of mechanisms of semantic change. The starting-point of the analysis consists of those cases in which the indirect object refers to the recipient of a benefactive material transfer, as in hi} gal haar een roos 'he gave her a rose'. The prototypical indirect object, in other words, is defined in terms of a frame featuring a process of transfer, something being transferred, and a specific functional character of the transfer. The indirect object itself, of course, is the person to whom the transfer is directed and who benefits from the transfer. The choice of this particular construction as the prototype of the category is not an arbitrary one: it links up with traditional localist conceptions of case relations, in which spatial aspects of meaning are considered to be basic. Although this lies beyond the scope of the present paper, it should be noted that independent evidence for the identification of the prototype might in fact be looked for: a frequency analysis of corpus data, or data of a psychological type involving the ease with which informants produce particular examples of the indirect object construction, might very well confirm the central position of the 'spatial' kind of use selected here. In the context of the present paper, however, the choice for this particular prototypical center is a hypothesis that can be con-
180
Constructions and idioms
finned only indirectly. It will turn out to be a suitable hypothesis, in fact, if it indeed provides a good starting-point for deriving a plausibly coherent analysis of the indirect object construction. It is important for the overall structure of the analysis in the following pages, that the prototype as defined here contains three major components: if the prototypical indirect object constituent is the recipient of a benefactive material transfer, semantic changes may take their starting-point in the 'recipient' part of the definition, in the 'transfer' part of the definition, and in the 'material' or 'benefactive' nature of the transfer. The observation that the prototype may be analyzed into a number of conceptual components has interesting repercussions in the light of the third theoretical point mentioned in the previous section, but discussion of that point will be saved for the final section of the paper. At this point, it suffices to see that mechanisms of extension such as metaphor and generalization may attach to different components of the prototype. And at the same time, of course, each component may be subject to different types of semantic shifts.
3.
Extension 1: Generalization
Starting from the indirect object as recipient of a material transfer as prototype of the construction, the first two dimensions leading to extensions from the prototype involve processes of metaphorization and generalization. Table 1 charts a number of these extensions, in such a way that their mutual combinations become apparent. Generalization features along the vertical dimension in the table, and metaphorization is charted along the horizontal dimension in the table. (1)
Ik geefhem een boek
(2)
Haar werd een hart ingeplant
(3)
De wind blies hem de regen in het gezicht
(4)
Ik geefhet huis een nieuw laagje verf
(5)
Ik schenk u mijn koninkrijk
(6)
Ze hebben je verkeerde ideeen ingeprent
'I give him a book' 'Her was a heart implanted, she received a new heart' 'The wind blew the rain in his face' 'I give the house a new coat of paint' 'I donate my kingdom to you' 'They have impressed- wrong ideas upon you'
The indirect object in Dutch (7)
181
20 'n geluk overkomt je nooit meer
'Such a happiness will never befall you again' (8)
De jury heeft 'De ontdekking van de hemel' de eerste prijs toegekend
(9)
Ik zeg u dat hij komt
'The jury has awarded the first prize to De ontdekking van de hemel' 'I tell you that he comes'
Table 1. Generalization and metaphorization of the Beneficiary role RECIPIENT OF
RECIPIENT OF
RECIPIENT OF
MATERIAL
ABSTRACT
COMMUNICATIVE
TRANSFER
TRANSFER
TRANSFER
EXPERIENCE
(1) (2) (3)
(5) (6) (7)
(9)
AFFECTEDNESS
(4)
(8)
transfer into the sphere of CONTROL USE
Generalization as a traditional mechanism of semantic extension in lexical semantics involves all fonns of semantic shift in which the original reading refers to a proper subset of the extension of the new reading. Classical examples are the development from French arriver 'to arrive at the river"s edge, to land' to the less specific meaning 'to arrive in general', or the development of English moon 'earth"s satellite' to the more encompassing reading 'any planetary satellite. Alternative tenns featuring in the lexical semantic literature for the concept of generalization include 'extension', 'weakening', 'broadening', and 'schematization' (in the Langackerian sense). The process of generalization identified in Table 1 basically involves a weakening of the Beneficiary-role of the recipient of the prototypical transfer. In Table 2, tentative definitions are given of the four positions that are distinguished along the generalization dimension in Table 1: for each of the four labels distinguished along the vertical dimension of Table 1, a componential definition is presented that reveals their mutual relations. In accordance with the definition of generalization, each secondary generalized reading includes the original one as a special case. All cases of 'transfer into the sphere of control', for instance, imply 'transfer into the sphere of experience". The overall structure, then, of the meanings affected by the process of generalization is one of nested subsets. The broadest category distinguished along the generalization dimension merely implies that the indirect object constituent is affected by the transfer
182
Constructions and idioms
referred to by the construction as a whole: the transfer involves a change (in the broadest possible sense) in the indirect object. At the most generalized end of the scale, the indirect object merely passively undergoes the effect of the transfer, without in any way being able to control or manipulate the transferred entity. At the other end of the scale, the indirect object constituent is a beneficiary in the strong sense: a human being who is free to use the received entity according to his own intentions. In between these two extreme positions, a continuum may be discerned going from 'control' over 'use' and 'experience' to 'affectedness'. Precisely because the generalization dimension is conceived of as a continuum, borderline cases are not to be excluded. For instance, because the 'use'-category distinguished in the figure is probably not a very prominent one as far as frequency of occurrence is concerned, it might already be considered a more or less marginal borderline case between the more prominent poles involving active control and manipulation on the one hand, and mere sentient experience on the other hand. In general, it is not necessarily the case that the intermediary positions distinguished in Table 2 are the only ones to be retained on a closer analysis. Table 2. Distinguishing features for control, use, experience, affectedness
CONTROL USE EXPERIENCE AFFECTEDNESS
the transfer involves a change of attributes and characteristics in the indirect object constituent
the transfer involves a sentient, mostly human being who is conscious of the change
the transferred entity is used but not manipulated by the beneficiary
the beneficiary has manipulative freedom in his use of the transferred entity
+ + + +
+ + +
+ +
+
4. Extension 2: Metaphorization The metaphorization dimension charted in Table I features two relatively straightforward patterns of metaphorical change. On the one hand, there is a shift from cases involving material transferred entities to cases involving abstract transferred entities. The link with the non-metaphorical readings
The indirect object in Dutch
183
shows up in the fact that the different positions on the generalization dimension that we distinguished earlier, also appear to be relevant for the metaphorized readings. The various abstract entities that are being transferred, have effects on the recipient that fall into the same categories as when material objects are being transferred. On the other hand, there are metaphorical extensions into the domain of communicative actions; broadly speaking, these transfers seem to rely on the conduit metaphor (Reddy 1979): messages are conceived of as entities that are being transferred from one person to the other. Logically, the recipient of a material transfer may thus be metaphorized into the recipient of a communicative transfer. One question that immediately imposes itself against the background of these metaphorical extensions is the following: why is the distribution of the instances of 'communicative transfer' over the 'control' etc. series so defective? An immediate answer to the question involves the recognition that a concept such as 'control' (as defined along the generalization dimension) simply does not apply to the objects of communicative transfers: the recipient of the message is more passive with regard to the message received than, for instance, the recipient of a present with regard to the present. Similarly, communicative transfers presuppose human (or at least animate) recipients, which excludes the 'affectedness' category as defined along the generalization dimension (because it specifically involved non-human recipients). Going one step further, however, it may be suggested that the asymmetrical distribution of the examples in Table I derives from a shift of perspective with regard to the prototype. Each of both metaphorical dimensions would then capitalize on different aspects of the prototype, and the different distributional patterns of both metaphors would be a reflection of that shift of perspective. Pursuing this line of thought, let us first identify the two perspectives involved. If the prototype case of the indirect object construction involves the indirect object as the recipient of a material transfer, the focus may lie either on the functional aspects of the transfer (the transfer as a process with specific effects for the recipient), or on the spatial aspects of the transfer (the transfer as a material change of place of an object). The simultaneous presence of both perspectives is reflected by the fact that the indirect object constituent in the prototypical case may be either identified as a Beneficiary (according to the functional perspective) or as a Recipient (when the focus is shifted somewhat to the spatial point of view). According to the functional perspective, the generalization continuum described in Table 2 is of prime importance: the various positions on the scale identify dif-
184
Constructions and idioms
ferent forms of functional effects that the indirect object constituent may be subject to. Conversely, distinctions between 'sphere of control', 'sphere of experience', 'sphere of affectedness' etc. recede to the background when the more purely orientational, spatial perspective is taken. Given, then, that the conduit metaphor takes its starting-point in the orientational conception of the literal prototype case rather than in the functional conception, it is not surprising to find that the distinction between the different positions on the generalization dimension turns out to be largely irrelevant for the metaphorical shifts into the field of communicative transfers. It may be noted that comparative evidence for the multiple perspectivizability of the prototype case might be found in languages where the dative case (the formally marked indirect object) expresses not just the notion 'recipient of transfer', but also the concept 'goal, direction of movement', in a clear-cut spatial, orientational sense. Such an extension of the category's range is a generalization of the spatial aspect of the transfer prototype, rather than of its functional aspects: the orientation of the movement is entirely detached from the notion of transfer. Incidentally, if we consider the dative in languages like Latin a morphologically marked indirect object, and if we further take into account that this morphological dative in Indo-European seems to be a development of an original locative case, it becomes clear that the diachronic relationship between the goal-reading and the recipient-reading may well be the converse of the synchronic relationship sketched above: historically speaking, the spatial aspects may have been primary with regard to the functional aspects. Such prototype shifts are not a theoretical anomaly, to be sure (compare Geeraerts 1997a).
5.
Extension 3: Metonymy
Two kinds of metonymy may play a role in the structure of the indirect object. First, we find indirect identifications of the indirect object entity: in a pattern like hij schrijft {het pa/eis, het stadhuis, het Instituut, de Wetstraat} een brief 'he writes {the palace, the town hall, the Institute, Downing Street 10} a letter', the human recipient of the communicative transfer is indicated by means of a reference to something (a building or dwelling-place) that is characteristically associated with the recipient in question. This is a form of metonymy that is quite wide-spread, and that is in no way specific for the
The indirect object in Dutch
185
indirect object construction. It will not be envisaged any further in what follows. Second, and more importantly, we find metonymical ramifications of the transfer aspect of the core indirect object meaning ('recipient of functional transfer'): instead of naming the recipient in an actual process of transfer, the indirect object constituent identifies entities or persons involved in actions or states of affairs that are metonymically associated with processes of transfer. Two important pattems may be discemed. On the one hand, there are states of affairs following acts of transfer: resultant states, effects, effectuated changes. If one says, for instance, dit cadeau is mU een kostbaar bezit 'this present is a precious possession to me', one describes oneself as the beneficiary of an act of transfer, but focusing on the situation after the actual transfer: the offering of the present itself is implied, but it is not explicitly mentioned. In such cases, the indirect object may be broadly defined as 'the entity affected by an (explicitly mentioned) state resulting from an (unmentioned) functional transfer (such as would be named in a core example of the indirect object construction)'. On the other hand, we find references to acts and states of affairs preceding acts of transfer: conditions, preparations, preliminary actions, intentions. If, for instance, one promises something to someone (ik beloof hem het koninkrijk, 'I promise him the kingdom'), the transfer has not yet taken place, but one describes someone as the beneficiary of a preliminary action (the promise) that will normally lead to the actual transfer. In these cases, the indirect object may be roughly defined as 'the entity affected by an (explicitly mentioned) action that is preparatory with regard to an (unmentioned) functional transfer of which the entity is the recipient'. Table 3 features examples of these two metonymical pattems ('resultant state' metonymies and 'preceding conditions' metonymies) for most of the types of use that we distinguished previously. In order not to clutter the figure too much, however, some simplifications have been applied: along the generalization dimension, the 'use' category has been left out; likewise, no instances of communicative transfers are included. (10)
Hij schenkt haar ko/fie in
'He serves her coffee' Ik gun hem het boek '1 grant him the book' (11)
Ik be/aa/hem het koninkrijk '1 promise him the kingdom'
186
Constructions and idioms
(12)
Er stond hem een regenbui te wachten
(13)
Zij gunnen hem dat genoegen niet
(14)
De burgemeester wenste het nieuwe museum veel bezoekers toe
(15)
Ik wens deze nieuwe theorie niets dan mislukte experimenten toe
(16)
Dit cadeau is mij een kostbaar bezit
(17)
Het rijk behoort haar toe
(18)
De hagel was hem onaangenaam
(19) (20)
Het valt mij moeilijk dat toe te geven 'Tt is for me difficult to admit that' De nieuwe kleur staat het huis goed
(21)
De aanpassingen waren de theorie tot voordeel
'A shower of rain was awaiting him' 'They do not grant him that pleasure' 'The mayor wished the new museum many visitors' '1 wish this new theory nothing but failing experiments' 'This present is a precious possession to me' 'The empire belongs to her' 'The hail was unpleasant to him'
'The new colour becomes the house' 'The changes were to the advantage of the theory' Table 3. Metonymical extensions of the indirect object construction
CONTROL EXPERIENCE AFFECTEDNESS
material abstract material abstract material abstract
PRECEDENT ACTIONS
RESULTANT STATE
METONYMY
METONYMY
(10) (11) (12) (13) (14) (15)
(16) (17) (18) (19) (20) (21)
It may be noted that the 'resultant state' cases typically involve the intransitive indirect object. This is probably an automatic, semantically driven consequence of the shift of perspective involved in the 'resultant state' metonymies: when the attention is shifted to the outcome of the transfer, the initiator of the initial transfer recedes to the background of the attention. A second point to note involves the fact that on top of the cases charted in Table 3, an additional type of metonymical extensions has to be distinguished. Let us notice, to begin with, that metonymical extensions need not imply that the original meaning is always relevant in those contexts where
The indirect object in Dutch
187
the new metonymical reading is applied. On the one hand, we find cases like he drank a glass: what he drinks is a 'quantity of water in a glass', but there is always at the same time a glass 'recipient of specific form and substance' around. On the other hand, we get cases like barrel: a barrel is a 'specific quantity of oil contained in a barrel', even if it is, in the circumstances, contained in a tanker or a pipeline. In the same way, to fall signifies 'to be killed by a weapon' even if the person does not fall down ('the parachutist who got caught in the tree fell under enemy fire'). Although this is not traditionally done, developments such as those affecting barrel and to fall could be seen as a generalization based upon an initial metonymy. For instance, barrel would undergo a metonymic shift from the initial 'recipient' reading to the reading 'amount of oil in such a recipient', and then further by generalization to 'same amount of oil, even if not actually in the recipient'. Now, regardless of the exact kind of analysis that one may wish to apply to such cases, it is crucial not note that the kind of extension exemplified by barrel and to fall also obtains in the indirect object construction. When, in fact, the metonymical 'resultant state' cases contained in Table 3 occur without reference to processes of transfer, 'pure' cases of 'possessor', 'experiencer', and 'affected entity' readings arise. For instance, the prototypically active recipient with powers of control and manipulation that features in the central cases of benefactive transfer, turns metonymically into an active, controlling entity in a stative relationship resulting from such a transfer; by a further process of generalization, it becomes an active, controlling entity in a stative relationship (regardless of whether an initial transfer took place or not). Table 4 contains a number of examples of the shift intended here. The examples are restricted to cases that link up with the 'resultant state' metonymies, but it should be noted that similar examples may be found with 'precedent conditions' metonymies. For instance, Ik gun hem zijn overwinning 'I grant him his victory, I do not begrudge his victory' may be said by someone who has no power at all to bring about (or to hamper) the victory. (22) (23) (24) (25)
Van al zijn lichaamsdelen was zijn linkerhand hem het dierbaarst 'Of all the parts of his body, his left hand was most dearest to him' Pannekoeken zijnmij het lekkerst 'Pancakes are the most tasteful to me' Jungles zijn tropische landen eigen 'Jungles are typical for tropical countries' Je vermogen om zeer snel te lezen zal je goed van pas komen 'Your capacity for reading very fast will be an advantage to you'
188
Constructions and idioms
(26)
AlIe vriendschap leek hem nutteloos
(27)
Zwaarmoedigheid is alle romantische muziek eigen
'All friendship seemed useless to him' 'Melancholy is common to all romantic music' Table 4. Secondary metonylllical extensions
CONTROL EXPERIENCE AFFECTEDNESS
6.
MATERIAL
ABSTRACT
(22) (23) (24)
(25) (26) (27)
Extension 4: Subjectification
On the basis of the 'experiencer' and 'affected entity' readings in the stative relations described at the end of the previous paragraph, a further extension leads to the so-called ethical dative, as in the idiomatic expression het is me wat 'that's something [to me, as far as I am concerned]!': the infonnation contained in the utterance is part of the sphere of experience of the indirect object (which is identical with the speaking subject), who is concerned, involved, interested with regard to the objective situation or event described by the utterance. Characteristically, the verb phrase need not refer to a process of transfer, which is an indication of the fact that the ethical dative is indeed a further development from the 'stative' cases described at the end of the previous paragraph. Against the background of the framework sketched so far, the ethical dative may be explained as a shift from the referential domain of the utterance to the discursive context. The interested party is not part of the referential domain in which the action takes place, but of the pragmatic domain in which the utterance is made. This shift is an example of the subjectification processes as described by Langacker (1990b) or Traugott (1989). (There are some differences between Langacker's and Traugott's conception of subjectification, but the differences need not detain our attention here). In general, subjectification involves a change in the conceptual construal of a particular relationship: from an objective construal (in which the relationship is construed without reference to the speech situation), a change is made to a subjective construal, in which the speaking subject (or another element from the speech situation) features as one of the relata in the relationship in question.
The indirect object in Dutch
189
In our case, the relationship could be, for instance, 'affectedness'. In the more common cases of the indirect object construction, the affected entity is part of an objective referential situation that is conceived of as being independent of the speech situation. In the case of the ethical dative, on the other hand, the affected entity is the speaking subject, as a basic element of the speech situation.
7. Extension 5: Perspectival switching A final extension to be noted involves cases in which the transfers that feature in the prototypical cases change polarity, so to speak, and are transformed into cases of blocked transfers or deprivations. Rather than receiving something, the indirect object entity is deprived of something: something is taken away from him that he already possessed, or he is denied something that he wishes for or that he expects to receive. The nuances that this reading occurs with are largely similar to the ones that were identified before. Without trying to be exhaustive about the possibilities, Table 5 lists examples involving actual transfers (where 'negative transfer' equals deprivation), together with a few examples involving stative relations; in each case, both material and abstract instantiations are included. (28)
Hij nam haar het speeltje a(
(29)
(30)
Een wolk ontnam ons het laatste straaltje zon ,A cloud took the last ray of sunshine away from us' Dit bericht ontnam haar de laatste kleur die haar gelaat nog sierde
(31)
Het ontbreekt hem aan het nodige geld om dat huis te kopen
(32)
Ze ontnamen hem aUe rechten
(33)
Het bericht ontnam Hemline haar gevoel van rust
(34)
De nieuwe snelweg ontneemt het landschap zijn laatste aantrekkelijkheid
'He took the toy from her'
'This message took the last shade of colour away that adorned her face' 'He lacks the necessary money to buy that house' 'All his rights were taken away from him' 'That message took Hermione's quiet feelings away'
'The new highway deprives this landscape from its last point of attractiveness' (35)
Het ontgaat haar dat het hoogtepunt van het generativisme achter de rug is
'She fails to see that the heyday of generativism is over'
190
Constructions and idioms
Table 5. Perspectival switches in the indirect object construction
negative transfer from the sphere of control negative transfer from the sphere of experience negative transfer from the sphere of affectedness stative relation of affectedness
MATERIAL
ABSTRACT
(28) (29) (30) (31)
(32) (33) (34) (35)
The explanation of the shift of perspective (or, if one wishes, the shift of polarity) that is illustrated by these examples is less straightforward than in the previous cases. Probably the most plausible analysis is to see the shift as the result of a generalization process affecting the 'transfer' component of the prototype case. Earlier on, we saw that the transfer component was subject to a metonymical shift in cases like Ik gun hem het back 'I grant him the book'. In the cases that concern us here, the transfer component is subject to a different type of change: it is generalized by dropping the 'benefactive' aspect that was associated with the prototype. The transfers envisaged here, in fact, do not add something to the sphere of control (etc.) of the indirect object constituent, but rather take something away: instead of being transferred in the direction of the indirect object constituent, the transferred entity moves away from the indirect object. Obviously, the indirect object can then no longer be characterized as a 'recipient', but had better be characterized as 'affected entity' or 'person concerned'. In the cases intended here, in other words, the generalization of the transfer component correlates with a shift in the 'recipient' component. Incidentally, this discussion shows that the label 'benefactive' that was used initially to characterize the prototype case may be slightly misleading, in the sense that it combines (in typically prototypical fashion) two different features that co-occur in the prototype case, but that are nevertheless separable. A benefactive transfer is typically one in which something good is given to someone: a positive for the indirect object constituent combines with a directionality oriented towards the indirect object. At the other extreme, the cases of deprivation illustrated in Table 5 involve transfers in which something good is taken away from someone. Notice, however, that giving something bad to someone (hi) gaf hem een klap 'he hit him') combines negative functionality with an orientation towards the indirect object. In the same way, taking something bad away from someone (haar antwoord ontnam hem zi)n somberheid 'her answer took away his melancholy mood') combines a positive, beneficial effect with an orientation of the transfer away
The indirect object in Dutch
191
from the indirect object constituent. The occurrence of all four combinations of directionality and effect shows, in short, that the generalizing extension of the 'benefactive transfer' aspect of the prototype actually covers two different processes. On the one hand, there is a perspectival change involving the directionality of the transfer. On the other, the functional aspect of the prototypical transfer may similarly (but independently) undergo a perspectival switch. Needless to say, this discussion as a whole once more establishes the importance of distinguishing between the functional ('benefactive') and the spatial (directional, orientational) aspects of the prototype case - a point that was already stressed in our discussion of the metaphorical extensions based on the conduit metaphor.
8. The resulting picture
If the prototype case of the indirect object construction is defined as 'the active recipient (with controlling power) of a benefactive transfer of material entities', various semantic extensions appear to start from each of the aspects of the prototype. An overview of the various extensions is presented in Figure 1. (For reasons of graphical economy, the extension towards the ethical dative is not included in the figure. Neither have the nuances of the perspectival switch discussed at the end of the previous section been explicitly represented. ) Crucially, the points in the network sketched in Figure 1 are not separate 'senses' of the indirect object construction, but conceptual components that co-occur in various combinations: Figure 1 only indicates the dimensions along which the indirect object construction may vary, but it does not chart all the individual readings that arise from the combination of the different positions that occur along the variational dimensions. In this sense, the representational fonnat of Figure 1 is fundamentally different from the 'radial sets' representations that are common in cognitive linguistic research: while the nodes in radial networks a la Lakoff (1987) constitute individual readings of the category in question, the representation of Figure 1 gives an overview of the conceptual components that may enter into such individual readings. For instance, a sentence like Zo 'n ge/uk overkomt je nooit meer (see Table 1) combines a shift from active recipient to passive experiencer along the generalization dimension (represented by the line going from the prototype to the lower right hand corner of Figure 1) with a shift from mate-
192
Constructions and idioms
rial to abstract transfer along the metaphorization dimension (the axis drawn vertically downward in Figure I).
affected entity instead of acti ve recipient
deprivation and blocked transfer instead of beneficial transfer
expenencer instead of active recipient
ill
communicative transfer instead of benefactive functional transfer
ACTIVE RECIPlENT OF BENEFICIAL TRANSFER OF MATERlAL ENTITY
preconditions and resultant states instead of transfers
stative relations instead of transfers
abstract entities instead of material entities
ill
GENERALIZATION
~
METAPHOR
Cl)
METONYMY
Figure I. The multidimensional structure of the indirect object construction
The indirect object in Dutch
193
Conversely, it is not necessarily the case that any point in the network that results from the combination of the dimensions in Figure 1, should be considered a different 'sense' (in the technical reading of that word). Already in the field of lexical semantics, distinguishing polysemy from vagueness is not always an easy task (compare Geeraerts 1993a). In syntax, then, the difficulties are even greater, because the criteria to apply are even less clear than in the field of lexical studies. Broadly labeled categories like Beneficiary or Affected Entity (Patiens), that occupy specific regions of the overall network, could certainly be considered candidates, but then again, according to what methodology should these categories be distinguished and demarcated? The road taken in the present paper has not been to try to answer the polysemy question, but rather to sketch the network of different uses with regard to which such a polysemy would have to be defined.
9.
Further steps to be taken
The analysis of the Dutch indirect object construction presented here does not pretend to be the final word on the subject, but merely tries to make a number of points that may prove stimulating for further research. In principle, there are two possible areas for a further elaboration of the model presented here. In the first place, the analysis as such will have to be refined. Specifically, by confronting the analyses given here with corpus data illustrating actual language use, it may be checked whether any additional semantic nuances have to be added to the picture drawn in the previous pages. Also, corpus data will yield frequency data that may confirm the prototypetheoretical structure assumed here. Is it the case, for instance, that the shades of meaning that are considered peripheral according to the present analysis, do indeed occur less frequently? In the second place, the model may be elaborated beyond the boundaries of the analysis given here. For one thing, the question arises whether the model has any typological and comparative validity: can the same structure be used to describe developments in other languages? Note that this question has two aspects to it: what languages share a similar construction with the same prototype as the one identified for the Dutch indirect object (see, for instance, Rudzka-Ostyn's identification of the prototype of the Polish dative: Rudzka-Ostyn 1996)? And - given such a prototypical similarity - can the extensions from the prototype that occur in the different languages be modeled along lines that are similar to the ones set out for Dutch? Another area
194
Constructions and idioms
for elaboration concerns the whole field of onomasiological research. The present analysis is a semasiological one: it describes the semantic structure of a single construction. At least for a number of the senses within the semasiological range of application of the indirect object construction, alternative modes of expression exist. First among these are the 'prepositional indirect objects', i.e. constructions of the type or . Assuming that these may indeed be considered different constructions (see the discussion in the first section of the article), the question arises whether the exact relationship among these onomasiological alternatives might be deternlined. In particular, would it be possible to analyze the choice between the ditransitive construction and the prepositional indirect object along the same quantitative lines set out in Geeraerts, Grondelaers, and Bakema (1994)? In that monograph, it is shown that the choice among lexical alternatives is determined simultaneously by semasiological and onomasiological salience effects (which may be calculated, given a sufficiently large corpus). Now, if the basic research strategy of Cognitive Linguistics (viz. of modeling the semantic analysis of grammatical constructions on the example provided by lexicological analyses) is carried through systematically, an extension of the analysis towards the onomasiological domain will have to be envisaged sooner or later.
10. Conclusion and discussion The major conclusion to be drawn from the present analysis involves the recognition that it is indeed possible to systematically and coherently describe the semantic structure of the (ditransitive) indirect object construction in Dutch on a 'lexical' basis, that is to say, by using models for semantic description developed in the lexicological domain. In the present analysis, this takes the form of starting from a prototypical meaning and deriving the other readings of the construction on the basis of well-known mechanisms of semantic extension such as metaphor and metonymy. As it stands, this general conclusion is not entirely original: the same idea lies at the basis of other recent analyses of datives and ditransitives, such as Goldberg (1992) and Rudzka-Ostyn (1996). On specific points of the analysis as well, there are many parallels with existing studies in a functional or cognitive linguistic framework. (These parallels have not been systematically indicated here, but see Rudzka-Ostyn 1996 for relevant references.) In what sense, then, does our analysis differ from these other proposals - in particular, from Goldberg
The indirect object in Dutch
195
(1992) and Rudzka-Ostyn (1996)? Restricting the comparison to these two articles, and focusing on fundamental matters rather than questions of detail, three points should be mentioned. First, instead of restricting the analysis to (or at least focusing it on) metaphorical links, as in Goldberg's (1992) radial set analysis of English ditransitives, other types of semantic extensions have to be considered. All the classical and less classical mechanisms of semantic extension (such metonymy, generalization, and subjectification next to metaphor) have to be envisaged in the analysis. The 'lexical' structure of the semantics of the indirect object construction cannot be adequately described by starting from an impoverished conception of semantic structure. This is an approach that the present analysis shares with Rudzka-Ostyn's (1996). Second, in contrast with Rudzka-Ostyn's proposal, the present analysis attaches no special structural position to extensions based on processes of generalization (which appear in Rudzka-Ostyn's proposal in the form of 'schematizations'). Just like Goldberg's use of a radial set representation seems to favor focusing on metaphorical links (given that the radial set model is predominantly used in the Lakovian tradition of metaphor-based analyses), Rudzka-Ostyn's adoption of a Langackerian schematic network representation seems to have predisposed her analysis towards overstressing the role of generalizations ('schematic' readings). Generalization, however, is but one of the many mechanisms in which the semantic coherence of a category may be grounded, and there does not seem to be any specific reason for making it more important than others. In particular, a category need not receive a single schematic reading to remain coherent. The obligatory search for such schematic readings easily leads to forced conclusions, and RudzkaOstyn's analysis is a case in point. Her identification of the topmost schematic reading of the Polish dative as a 'landmark' is, in fact, too schematic: it insufficiently distinguishes the dative from other constructions that also function as landmarks. The whole point, of course, mirrors the discussion in lexical semantics about the possibility of finding single definitions for lexical categories. Such definitions have to be maximally general and minimally specific: maximally general in the sense of applying to all the cases falling within the range of the category, and minimally specific in the sense of distinguishing the category from other, non-synonymous categories. A definition of the dative as a 'landmark', then, does not conform to the requirement of minimal specificity. On a more fundamental level, the point may be formulated as follows: given that lexical semantic research has established beyond doubt that monosemic, 'schematic' definitions of lexical categories
196
Constructions and idioms
need not always be possible (see Geeraerts, Grondelaers, and Bakema 1994), one has to beware of a forced search for such definitions in the case of grammatical constructions as well. Third (and foremost), instead of linking extended individual meanings directly to the prototype or to each other, a multidimensional structure has to indicate how non-prototypical readings are linked among each other by the co-occurrence of semantic shifts along various dimensions. Both Goldberg's radial set analysis and Rudzka-Ostyn's schematic network analysis interconnect individual readings of the construction in question, but insufficiently succeed in indicating the multidimensional structure underlying the interconnections between the individual readings. To appreciate the importance of these underlying dimensions, we may have another look at Figure 1. The representation used there visualizes an observation that was already brought to the fore in the second section of the article: the dimensions along which the semantic extensions take place, relate to specific components of the prototypical reading. For instance, assuming that the prototype case may indeed be defined as the recipient of a benefactive material transfer, one dimension of change pertains to the 'recipient' component, another to the 'transfer' component, and so on. We should now go one step further, however, and recognize that the major components in the definition link up with specific components of the syntactic frame in which the indirect object construction occurs. Focusing once more on the ditransitive -frame, it will be readily appreciated that the 'transfer' component of the definition links up with the verbal part of the formal construction, and the 'material' aspect of the definition with the direct object noun phrase (there is a material transfer when the direct object noun phrase refers to a material object). The 'recipient' itself, of course, is the second noun phrase, the indirect object as such. The definition that we have used, in other words, is a relational one that explicitly takes into account the syntagmatic dimension of the indirect object constituent. One could also say, alternatively, that the description is aframe-based one, in which the indirect object is defined as part of a larger conceptual structure. (Incidentally, while it might be suggested that this type of analysis characteristically distinguishes the description of verbal lexical items and grammatical constructions from that of nominal ones, this is not entirely adequate. Any relational concept, in fact, including relational nouns, will be most adequately described by incorporating the relational aspects into the definition.) What does not explicitly appear in Goldberg's and Rudzka-Ostyn's approaches, then, is certainly not the idea that the indirect object or the dative have to be de-
The indirect object in Dutch
197
scribed as (part of) a grammatical construction, but rather the insight that it is precisely this relational description that underlies the multidimensional nature of the semantic structure of the indirect object or dative. To summarize, what we have done, and what largely constitutes the specificity of the present approach, is a) to define the indirect object constituent as part of a construction: the components of the definition systematically evoke the other constituents that feature in the grammatical construction; b) to reveal the multidimensional semantic structure that correlates with this relational definition: the polysemy of the indirect object is determined by covariation of semantic extensions affecting the various components of the relational, construction-based definition; c) to stress the lexical nature of the semantic organization of the indirect object construction: an adequate picture of the changes affecting the components of the definition can only be acquired by envisaging all the mechanisms of semantic extension that have traditionally been identified in lexical semantics.
Chapter 8 The interaction of metaphor and metonymy in composite expressions
Originally published in Rene Dirven and Ralf Parings (eds.), 2002, Metaphor and Metonymy in Comparison and Contrast 435-465. BerlinlNew York: Mouton de Gruyter. Idioms and compounds have similar semantic characteristics: both types of expressions are semantically composite, in the sense that their meaning is (at least in principle) composed of elementary building blocks, i.e. the constituent parts of the expressions. Idioms have received considerable attention from Cognitive Linguistics (see e.g. Gibbs 1986, 1994, 1995, Fillmore, Kay and O'Connor 1988, Dobrovol'skij 1995, Feyaerts 1999), compounds much less so (see e.g. Ryder 1994, Rudzka-Ostyn 2003). The mechanisms of semantic compositionality as well have received only limited attention, with the exception of work in the line of mental space and blending theory, as in Sweetser (1999) or (Coulson 2001). The present paper takes a closer look at the semantics of composite expressions, without however attempting to provide a systematic comparison with the blending analysis: drawing on Geeraerts and Bakema (1993) and Geeraerts (l995b), it describes the interaction between the syntagmatic and the paradigmatic axes in the meaning of idioms and compounds, and then charts the various ways in which metaphor and metonymy can interact along these axes. The crucial idea is this: when we study compositional processes in the framework of a semantically oriented grammar like Cognitive Linguistics, we should not only describe the way in which the constituent elements work together to produce a composite meaning, but we should also take into account the semantic shifts that each of the elements (and their combination) may undergo in its own right.
1.
Metaphor and metonymy in compounds and idioms
The semantic architecture of idioms and compounds is identical: in both types of expressions, meaning is compositional to the extent that it is built up out of the constituent parts of the expression; at the at same time, this compositional meaning is usually but a first semantic step, to the extent that processes of meaning extension produce a figurative reading from the compositional one. This paper will explore the various ways in which metaphor
The interaction ofmetaphor and metonymy
199
and metonymy may interact in expressions with this type of semantic architecture. The first part of the paper describes the architectural characteristics, i.e. the paradigmatic and the syntagmatic axes of composite expressions and the way in which they interact. This part of the paper reproduces the major part of Geeraerts (1995 b). The phenomena in question will be introduced with regard to idioms in particular, but the extrapolation to the semantics of compounds is a straightforward one. The second part of the paper charts the various ways in which metaphor and metonymy combine and interact within the structure defined by the syntagmatic and paradigmatic axes. In particular, it will be shown that metaphor and metonymy may occur either in a consecutive or in a parallel sequence. This part of the paper is loosely based on Geeraerts and Bakema (1993). All the examples illustrating the argument are taken from Dutch.
2. The prismatic architecture of composite expressions
2.1. Isomorphism and motivation The paradigmatic and the syntagmatic dimension 1 of idioms are both twofold, in the sense that both can be considered with regard to the original, literal meaning of the idiom, and with regard to the derived, figurative meaning. The paradigmatic dimension of idioms primarily involves the relationship between the original meaning of the idiomatic expression as a whole and its derived meaning. Secondarily, it involves the relationship between the original, literal meaning of the constituent parts of the idiomatic expression, and the interpretation that those parts receive within the derived reading of the expression as a whole. The syntagmatic dimension of idiomatic expressions involves the relationship between the interpretation of the constituent parts of the expression on the one hand, and the interpretation of the expression as a whole on the other, but clearly, this syntagmatic dimension can be envisaged both with regard to the original meaning and with regard to the derived meaning. Systematically, then, the semantic relations in idioms might be charted by means of a prismatic structure as in Figure 1. (The figure is intended to serve purposes of clarification and reference only; it is obviously not a formal representation as envisaged in formal grammar. To keep matters simple, the expression is assumed to contain only two lexical items.)
200
Constructions and idioms
1 Expression as a whole in its literal reading 2 First constituent item in its literal reading 3 Second constituent item in its literal reading 4 Expression as a whole in its derived, idiomatic reading 5 First constituent item in its derived reading 6 Second constituent item in its derived reading Figure 1. The prismatic structure of idioms
In addition, the notion of compositionality may either receive a dynamic or a static interpretation. Within the dynamic interpretation, compositionality is thought of as a syntagmatic derivational process in the course of which the meaning of a compound expression is computed on the basis of the meanings of the constituent parts of the expression'" In contrast with this dynamic, bottom-up conception, a static interpretation can be envisaged that merely notes that a one-to-one correspondence between the parts of the semantic value of the expression as a whole and the meanings of the constituent parts of the expression can be detected, regardless of the question whether this correspondence has come about through a process of bottom-up derivation or through a top-dO\vn interpretative process. Examples of such interpretative processes will be given later on; at this point, it suffices to see that a 'non-directional' conception of compositionality is not excluded. Now, because the term compositionality, through its processual connotation, strongly calls up the idea of bottom-up derivation, a different term might be helpful for the neutral, non-directional interpretation. I propose to use the
The interaction ofmetaphor and metonymy
201
term isomorphism here: what is ultimately at stake, is a one-to-one correspondence between the formal structure of the expression and the structure of its semantic interpretation, in the sense that there exists a systematic correlation between the parts of the semantic value of the expression as a whole and the constituent parts of that expression. A third preparatory step can be made by refining the notion of derivation. Although statements to the effect that idiomatic meanings are not compositional, that they are specialized, and that they cannot be derived straightforwardly would seem to be interchangeable, it should now be clear that we have to distinguish in principle between the syntagmatic underivability of the meaning of an expression (which is, of course, its non-compositionality), and its paradigmatic non-derivability. The latter involves the transparency of the semantic extension that leads from the original meaning of an expression to its transferred reading. It is illustrated by the interpretative difficulty that we noted with regard to (3): along the top line of the prism in Figure 1, the transition from the literal meaning of iemand iets op de mouw spelden to its idiomatic meaning is opaque. The latter cannot be derived on the basis of the former, because the motivating image is lost. I propose to call this type of derivability motivation (in contrast with dynamic compositionality as a syntagmatic kind of derivability). To round off the preparations, let us note that isomorphism and motivation as defined here 3 share a common characteristic: both involve the transparency of some of the links indicated in Figure 1. More specifically, isomorphism coincides with syntagmatic transparency, whereas motivation can be defined as paradigmatic transparency. 2.2. Combinations of isomorphism and motivation We can now arrive at a basic classification of the specialized nature of idioms when we consider the various combinations that isomorphism and motivation can occur in. Isomorphism and motivation can, of course, each be considered at two points in the relational structure sketched in Figure 1. On the one hand, motivation may refer both to the paradigmatic top line of the prism and to the two paradigmatic bottom lines. On the other hand, syntagmatic isomorphism may involve the front triangle or the back triangle of the figure. Because the entity that we are primarily interested in is the idiomatic meaning of the expression as a whole (the top backside corner of the figure), we will for now only consider motivation and isomorphism as they directly relate to that part of the structure, i.e., we shall consider isomorphism within
202
Constructions and idioms
the figurative plane, and motivation on the global level. Extensions towards other parts of the structure will follow in the next sections. Consider, then, the following expressions. (1)
De koe bij de horens vatten 'To take the cow by the horns> to take the bull by the horns, to grasp the nettle' Aan de weg timmeren 'To practice carpentry at the roadside, to work in public> to attract attention by one's activities, to be in the limelight'
(2)
Met spek schieten 'To shoot with bacon> to tell a tall story, to boast' Een wit voetje bij iemand hebben 'To have a white small foot with
someone> to be in someone's good books, enjoy someone's favours' (3)
JI/fet de handen in het haar zitten 'To sit with one's hands in one's hair
> to be at one's wit's end, to be in trouble' Dat heeft niet veel om het lijTThat does not have much around the
body> there's nothing to it it does not mean very much' (4)
De kat de bel aanbinden 'To tie the bell to the cat> to bell the cat, to
take the lead in a dangerous activity' Als puntje bij paaltje komt 'When point reaches pole> when it comes
to the crunch, when all is said and done, when you get down to the nitty-gritty, Systematically, the idiomatic readings in (1) are both isomorphic and motivated. Those in (2) are isomorphic but not motivated, those in (3) motivated but not isomorphic, and those in (4) neither isomorphic nor motivated. The isomorphic nature of de kae bi} de harens vatten follows from the fact that a consistent one-to-one mapping can be defined between the elements of the global meaning and the meanings of the constituent parts of the expression. If we paraphrase the idiomatic meaning as 'to tackle a problem or a difficulty at the central, most dangerous or difficult point', it becomes clear that the cow maps onto the problem in its entirety, while the horns represent the most tricky part of it; taking hold of the horns further symbolizes tackling the core of the problematic situation. Similarly, met spek schieten is isomorphic because the tall tales that are told can be seen to correspond with spek, while the telling of the tales corresponds with schieten. Conversely, it is difficult to identify those aspects within a situation of being at one's wit's end that could map isomorphically onto the various aspects of the situation described by met de handen in het haar zitten: what would be the hands, and what would be the hair, for instance? As far as motivation is concerned, however, it can be readily appreciated that met de handen in het haar zitten is a metonymic expression for a situation of being in trouble; taking one's
The interaction ofmetaphor and metonymy
203
head between one's hands and pondering the situation is precisely what one does in the circumstances. In the same vein, it is easy to see that the literal situation described by de koe bi) de horens vatten is a metaphorical image for tackling a problem at its most difficult spot. But it is unclear why shooting with bacon should come to indicate boasting, or why a point reaching a pole should have anything to do with things becoming serious. For further clarification of this basic classification, four remarks have to be made. In the first place, the isomorphic relations identified above should not be confused with the question whether the isomorphically mapped readings of the constituent elements of the idiomatic expressions are themselves motivated. For instance, while the lexical item koe in de koe bi) de horens vatten maps onto the 'problem' part of the global figurative reading 'tackle a problem by its most difficult aspect', there is no independent motivation for extending the semantic range of koe towards the meaning 'problem'; a semantic shift from 'cow' to 'problem' is not a conventional aspect of the meaning of koe, nor is there a readily conceivable independent metaphor that leads from 'cow' to 'problem'. In other words, the bottom paradigmatic lines in Figure 1 are not present in the case of de koe bi} de horens vatten. 4 It should now also be clearer why I suggested to take into account isomorphism as a non-directional concept of compositionality. Although de koe bi} de horens vatten exhibits isomorphism, the idiomatic meaning 'tackle the problem by its most difficult aspect' could never be arrived at by means of a bottom-up compositional process, because the building blocks for that process (for instance, an interpretation 'problem' for koe) cannot be reached independently. For lack of an independent paradigmatic motivation at the bottom of the prism, the input for a possible compositional process can only be retrieved when the output of the process (the global figurative meaning of the idiom as a whole) is already available. In the second place, motivation and isomorphism may be partial. Given a paraphrase 'to give the orders' of de lakens uitdelen (mentioned under (5) below), it is isomorphically possible to map lakens onto 'orders' and uitdelen onto 'give'. At the same time, it is possible to imagine a situation in which the person responsible for distributing the sheets is generally in charge; as such, the idiomatic meaning is motivated. But the motivational link is weak: distributing sheets is not the kind of situation that is typically associated with being in charge (or at least, not any more: apparently, the image derives from the dominant position of the lady of the house, whose control over housekeeping is symbolized by her control over the Iinencupboard). In de kogel is door de kerk, an event (such as the making of a
204
Constructions and idioms
decision) whose coming about has been delayed or hindered by some kind of obstacle, has finally occurred. The general picture is fairly clear: a material obstacle (the church) obstructs and slows down the movement of the bullet, in the same way that various difficulties obstruct and slow down the materialization of the long-awaited event. But why a church and a bullet? Again, the transparency of the motivating image is only partial. (5)
De lakens uitdelen 'To hand out the sheets> to run the show, be the boss, play the first fiddle' De kogel is door de kerk 'The bullet has gone through the church> the dice has been cast, a decision has been reached, things have taken a final turn'
(6)
Met spek schieten 'To shoot with bacon> to tell a tall story, to boast' Abraham gezien hebben 'To have seen Abraham > to be over fifty' Als puntje bij paaltje komt 'When point reaches pole> when it comes to the crunch, when all is said and done. when you get down to the nitty-gritty'
(7)
VU de bol gaan 'To go out of one's head> to blow one's top, to go out of one's mind with excitement' Het hoofd verliezen 'To loose one's head' Niet goed bij z'n hoofd zijn 'Not to be well in the head> to be soft in the head' Het hoofd loopt mij om 'My head is going round' Z'n hoofd ergens bijhouden 'To keep one's head to something> to remain attentive, to keep one's mind on something' Buiten zichze!fzijn 'To be beside oneself Vit z 'n vel springen 'To jump out of one's skin> to be beside oneself (with rage)' Vitbarsten 'To burst out, explode' Exploderen/ontplojJen 'To explode' In de wolken zijn 'To be in the clouds> to be overjoyed' In de zevende hemel zijn 'To be in the seventh heaven> to be on cloud nine' In de put zitten 'To sit in the pit> to be downhearted, be in the dumps, feel down' Door een dal gaan 'To go through a valley> to go through an abyss, to suffer a depression' Erbovenop zijn 'To be on top of it> to have overcome one's troubles'
In the third place, loss or weakening of motivation often results from cultural changes. More often than not, the background image that motivates the figurative shift is an aspect of the material or the immaterial culture of a
The interaction ofmetaphor and metonymy
205
language conununity - and when the culture changes, the imagistic motivation may lose its force. A clear example is met spek schieten (repeated under (6». Apparently, enemy ships were shot at with bacon (and similar fat substances) to facilitate setting them afire; the interpretation 'to boast' can then be reached through the intermediary of an interpretation 'to subject someone to verbal aggression, to overpower someone verbally'. In this case, the relevant knowledge belongs to the material culture of old-time warfare at sea. In Abraham gezien hebben, the relevant piece of knowledge belongs to the realm of the inunaterial rather than the material culture. Because a meaning 'to be no longer young or inexperienced' can be derived to the extent that the idiom is interpreted as a hyperbolic expression with the reading 'to have seen someone from a long time ago', the motivation is only partial: why is the borderline set at 50? In fact, it takes a good knowledge of the Bible to recognize the background of the expression, which derives from the gospel of John 8: 57. It should also be clear by now, that the motivated nature of an expression is subject to considerable individual variation (depending, among other things, on individual differences in one's familiarity with the historically motivating context). This is not to say, however, that a vast encyclopedic knowledge will always suffice to recover the motivation behind an expression. The latter may indeed be near to irretrievable, which is typically the case when professional etymologists disagree on the origin of an expression. In als puntje bij paaltje komt, for instance, some think of a transformation of an older expression als putje bij paaltje komt 'when the pit comes to the pole, when it comes to putting the pole into the pit', while others think of a variation on de puntjes op de i zetten 'to dot the i > to be meticulous about the details' . In the fourth place, the motivating image need not be specific for the expression in question; moreover, the motivating image may be complex. At this point, we can link up with the generalized metaphor research in the line of Lakoff and Johnson (1980). Without going too deep into the matter, consider uit de bol gaan in (7). The expression seems to be motivated by a combination of at least three images that are each generalized ones, in the sense that they provide a general motivation for various specific expressions. First, THE HEAD IS THE LOCUS OF ONE'S SELF-CONTROL underlies het hoofd verliezen, niet goed bij z 'n hoofd zijn, het hoofd loopt mij om, z'n hoofd ergens bijhouden. Second, LOSING ONE'S SELF-CONTROL IS LEAVING THE BODY is to be found in buiten zichze(fzijn, uit z'n vel springen, uitbarsten, exploderen, and ontplojfen. And third, UP IS POSITIVE / DOWN IS NEGATIVE
206
Constructions and idioms
is present in in de wolken zijn, in de zevende hemel zijn, in de put zitten, door een dal gaan, erbovenop zijn. 2.3. Local motivation and absence ofliteral isomorphism In the previous section, only part of the original representation in Figure 1 was envisaged. Enlarging our perspective to Figure I as a whole, there are two additional phenomena to be dealt with. On the one hand, the global motivation of an expression can be contrasted with the 'local' motivation of each of the elements in the expression, i.e., motivation along the top paradigmatic line of the prism has to be supplemented with motivation along the bottom line. On the other hand, isomorphism at the figurative back of the prism can be contrasted with isomorphism at the literal front of the prismatic representation. Pare Is voor de zwijnen in (8) below is an example of an idiom in which Figure 1 is fully realized. The global image is motivated (it is easy to appreciate what it means to throw valuable things at the feet of unworthy beings), and the figurative meaning is isomorphic (parels maps onto the valuable things that feature in the idiomatic meaning, and zwijnen maps onto the lowly beings that they are surrendered to). At the same time (and this a major contrast with the de koe bij de horens vatten example that we discussed earlier), both the transition from pare! to 'valuable thing' and from zwijn to 'unworthy person' is motivated on the basis of an evaluative metaphor. Up to a certain point, the metaphorical transition may even be said to be lexicalized (in the sense of being conventional): zwijn is a regular term of invective in the same way that pig is, and een pareltje is a term of praise in the same way that a gem, a jewel, a pearl are. (This is not to say, however, that the metaphorically motivated readings of parel and zwijn in the idiom coincide exactly with the lexicalized metaphorical meanings. For instance, as a term of abuse, zwijn normally implies that the person in question lives an immoral life, whereas no such implication need be present in parels voor de zwijnen. The important point to see is that the lexicalized existence of zwijn in a reading like 'unworthy person, specifically because of his immoral behavior' strengthens the motivated nature of the related reading that the word receives in the context of the idiom.) (8)
Parels voor de zwijnen gooien 'To cast pearls before swine'
(9)
lemand de loe!a(steken 'To take the wind out of someone's sails> to
get the better of someone, to deprive someone of an advantage' Tegen heug en meug 'Against heug and meug > against one's will, re-
The interaction ofmetaphor and metonymy
207
luctantly' T/an hot noch haar weten 'To know neither hot nor haar > to be totally
ignorant' (l0)
lets aan de kaak stellen 'To put something at the jaw> to expose, de-
nounce something' lets op touw zetten 'To put something on the rope> to organize, plan,
start, launch something' Given this analysis of pareIs voor de zwijnen as a fully motivated and fully isomorphic idiom, two additional remarks can be made. To begin with, it will be appreciated that the figurative reading of fully motivated and fully isomorphic idioms can be arrived at along two interpretative routes: either the global literal meaning is derived first and then transferred into the figurative realm, or the shift from the literal meaning of the individual words to their transferred meaning is effectuated first, to be followed by their compositional combination into the idiomatic meaning of the expression as a whole. In terms of Figure 1, the interpretation can go from the bottom to the top first, and then to the back, or it can go to the back along the bottom first, and then move up. Or, in still other words, either the syntagmatic dimension is gone through first, or the paradigmatic one takes precedence (starting, each time, at the lower front of the prism). Which of both interpretative paths is psychologically real (or merely preponderant) is another matter, but it is important to see that both are at least in principle possible. Further, the 'local', lexical motivation may be partial, i.e., it need not involve all the items constituting the expression. When iemand de loef afsteken is interpreted as 'to deprive someone of an advantage (like an initial superior position)', loef maps onto 'the advantage', while afsteken can be associated with the notion of deprivation. In the latter case, the association is motivated: the metaphorical transfer from the literal meaning 'to cut off of afsteken to 'to deprive' is a transparent one. In contrast, no such motivating link is possible in the case of loef, because the latter does not have a literal meaning for the average speaker of Dutch. Notice also that there are gradations in the degree of motivation,S Conventionalized shifts of meaning like the ones we mentioned in the case of pareI and zwijn are stronger than the shift from 'to cut off to 'to deprive' in the case of a.f.~teken: the latter extension is possible and perhaps even plausible, but it is not a conventional one. Together with iemand de loefafsteken, the other expressions under (9) illustrate the second major extension of the basic classification presented in section 3. Because loe/. heug, meug, hot, and haar are cranberry morphs, they illustrate the case in which there is no isomorphism on the literal level: a
208
Constructions and idioms
global literal meaning cannot be computed because one or more of the lexical building blocks lacks a meaning of its own. (Incidentally, the earlier discussion of de loef afsteken has made clear that the absence of isomorphism on the literal level can co-occur with isomorphism on the figurative level.) In addition, it may be useful to mention the existence of 'hidden cranberry morphs' such as kaak and touw in (l0). In the context that motivates the idiom, the words have their older meanings 'pillory' and 'loom' respectively; at present, these meanings are no longer common usage. The homonyms kaak 'jaw' and touw 'rope', on the other hand, are high-frequency words. The expressions aan de kaak stellen and op touw zetten, then, can receive an interpretation incorporating the readings 'jaw' and 'rope' (as suggested in (l0», but this is obviously not the kind of literal reading that could motivate (on the level of the expression as a whole) the figurative interpretation of the idiom. To the extent that kaak 'pillory' and touw 'loom' survive only in the expressions aan de kaak stellen and op touw zetten, they are like ordinary cranberry morphs; to the extent, however, that they formally coincide with the homonyms kaak 'jaw' and touw 'rope', they can be called 'hidden cranberry morphs' . It appears, in short, that the associative links presented in Figure 1 may be present in various combinations. A full-fledged investigation into the semantics of idioms, then, will have to include an overview of the various ways in which the model sketched here may be partially realized. (See Geeraerts and Bakema 1993 for an example of what such an overview may look like.) 2.4. Reinterpretation processes The examples of isomorphism on the figurative level that were mentioned in section 2.2 yield secondary, non-original interpretations for the items involved in the process. For instance, the isomorphic association between lakens and 'orders' in de lakens uitdelen yields a contextually determined interpretation for lakens, but because the transition from laken 'sheet' to laken 'order' is neither conventional nor motivated (in the sense that the latter reading is a plausible and transparent semantic extension of the former), the contextually isomorphic interpretation of laken as 'order' is not likely to acquire much structural weight in the lexicon. The process of contextual reinterpretation within the idiom is not always, however, without structural importance. 6
The interaction ofmetaphor and metonymy (11)
209
Spek'Verkoper 'Person who sells bacon> person who boasts' Van heinde en verre 'From far and wide, from everywhere' Met zijn talenten woekeren 'To make the most of one's talents> to ex-
ploit one's gifts' In fact, the reinterpretation process can be shown to be real when the item in question (in its secondary reading) comes to be used in isolation from its original idiomatic context. This has happened with spek in met spek schieten, which has led to the compound spekverkoper, as in (11) below. The formation of the compound noun can only be explained if it is accepted that spek in its idiomatically contextual reading 'boasting speech, tall tale' has been isolated from the original idiom. S'pekverkoper as such cannot be directly explained as an original metaphor of its O\\Il1: there is no way in which selling bacon can be associated with boasting except through the intem1ediary of met spek schieten. This type of 'semantic back-fommtion' is also quite visible in the case of cranberry morphs. When people are asked for an interpretation of heinde in van heinde en verre, it appears that a majority understands the item as a synonym of near-synonym ofverre (more or less like the relationship between far and wide in the English counterpart of the expression). Etymologically speaking, however, heinde and verre are antonyms rather than synonyms; heinde is related to hand and basically means 'what is near, what is at hand, what can be found in the immediate neighborhood'. When the etymological relationship with hand and the semantic relationship with nearness is lost, however, the overall meaning 'from everywhere' of the idiomatic expression enables heinde to be reinterpreted as a synony111 of verre. Met zijn talenten woekeren is an even clearer case. Whereas talent in its original biblical context referred to a particular coin, the reading 'personal ability, capacity, aptitude' that it received in the figurative interpretation of the expression is now the major one; it occurs freely in separation from the original expression. From a very general perspective, reinterpretation processes such as these? indicate that a search for isomorphism (defined as syl1tagmatic transparency) is an active force in the mind of the language user. Apparently, isomorphism on the figurative level of the idioms is not just real when it is given on the basis of the literal meanings of the constituent elements of the expression, but it is also real in the sense of being sought for when it is not given. If this can be accepted, it also means that interpretation processes are not always bottom-up, but that they can also be top-down: the overall meaning of met zijn talenten woekeren determines the specific meaning of talent that has become the item's major meaning.
210
Constructions and idioms
(12)
Vi/en naar Athene dragen 'Carry owls to Athens/Athena > do something irrelevant, useless. superfluous' De kat de bel aanbinden 'Tie the bell to the cat> bring something out into the open, make something public, ring a bell about something'
Similarly, evidence can be found that the search for paradigmatic motivation too is a real one. Of course, the search for motivation is quite outspoken when etymological researchers disagree on the actual motivation of an expression, as in uilen naar Athene dragen, which is usually interpreted in terms of the owl as the symbol of the goddess Athena and the city Athens, but which some see as being motivated by the simple fact that there were many owls in Athens. But etymologists are professional searchers for motivation, so their interpretative creativity does not tell us too much about the spontaneous occurrence of such interpretative activities when ordinary language users use idioms. More important in this respect are, first, psycholinguistic investigations of the type reported on in Gibbs (1990) and later work, which show that motivating images for idioms are psychologically real. And second, evidence for actual reinterpretations along the paradigmatic axis may be found, i.e. evidence for new meanings coming about through the search for motivation. For instance, the older idiomatic meaning of de kat de bel aanbinden (viz. 'to bell the cat, to take the lead in a dangerous activity') refers to the old fable of the cat and the rats. Nowadays, however, it seems to be shifting towards the interpretation 'to bring something into the open, to make something public, to ring a bell about something': on the one hand, the older association with taking the responsibility in a dangerous action (in favor of other people) disappears into the background; on the other hand, the notion of drawing the public attention to something (in particular, something scandalous or negative) is foregrounded. Given that de kat de bel aanbinden is largely unmotivated for most speakers, the association between the bell referred to in the expression and the notion of making something public (of making it heard, that is) enhances the motivated character of the idiom. The search for greater motivation leads to a shift in the interpretation. 2.5. Summing up What I have tried to indicate in the previous pages can be summarized in three points. First, an adequate description of the various forms of semantic specialization that occur in composite expressions requires that a number of distinctions are taken into account: the distinction between the syntagmatic and the paradigmatic aspects of meaning (which can both be found on the
The interaction o.f metaphor and metonymy
211
level of the literal meaning and on that of the figurative meaning of the idiom), and the distinction between bottom-up and top-down semantic processes. Specifically, the concepts of isomorphism (defined as syntagmatic transparency) and motivation (defined as paradigmatic transparency) are of primary importance for describing the semantics of composite expressions. Second, semantic specialization is a matter of degree. More specifically, the classificatory framework defined on the basis of the conceptual distinctions just mentioned allows for a ranking of the degree of specialization involved. Least specialized are fully isomorphic and fully motivated cases like parels voor de zwijnen gooien. Somewhat less specialized are cases like de koe bij de horens grijpen and met de handen in het haar zitfen, which may not be derivable on a word-per-word basis, but which are entirely transparent along the upper side of the prismatic structure. Still further down the line, we find cases like met spek schieten and de kat de bel aanbinden, that lack the global motivation of the previous examples. Most specialized, finally, are cases where the literal meaning of the expression cannot even be recovered, such as iemand de loeI afsteken and other idioms containing cranberry morphs. In each of these cases, matters are further nuanced by the existence of degrees of motivation. And third, semantic interpretation is not just a question of bottom-up compositionality or literal-to-figurative transfer. The reinterpretation processes that can be observed point to the existence of top-down and figurativeto-literal interpretations. It is not just the case that literal meanings determine figurative ones; figurative meanings also determine literal ones. And it is not just the case that the meaning of the parts determines the meaning of the whole; the meaning of the whole also determines the meaning of the parts.
3, Interactions between metaphor and metonymy in composite expressions
3.1. The metaphor/metonymy continuum If there is a continuum between metonymy and metaphor, this implies that there are in-between cases between expressions that are fully metonymical and expressions that are fully metaphorical. Composite expressions as well can be fully metaphorical or fully metonymical, when the motivational links that are present within the semantic architecture of the expression are only
212
Constructions and idioms
metonymical or only metaphorical. Pare!s voor de zwijnen gooien, as discussed above, is fully metaphorical: the top level shift from 'to throw pearls at swine' to 'to present unworthy people with valuable things' is a metaphorical one, and so are the bottom level shifts from pare! 'pearl' to 'valuable thing',8 and from zwijn 'pig' to 'unworthy person'. Conversely, compounds of the bahuvrihi type, like roodhuid 'redskin', are well-known cases of metonymical compounds: the link between the initial, compositional meaning ('red skin') and the derived reading ('Indian, seen as one with a red skin') is metonymical link of the possessed/possessor type. In the roodhuid case, to be sure, motivational links at the bottom level fail, because the derived reading 'redskin' cannot be considered isomorphic. So, given that composite expressions can be either metaphorical or metonymical, how can we chart the in-between cases? How do metaphor and metonymy occur in mutual combination in compounds and idioms? I will argue that there are three basic cases to be distinguished: cases in which metaphor and metonymy occur consecutively, cases in which they occur in parallel, and cases in which they occur interchangeably. In the following sections, each of these cases is presented separately. (More, and more intricate, examples of the interaction between metaphor and metonymy in expressions may be found in Gevaert 1994 and Feyaerts 1997.) 3.2. Consecutive interaction of metaphor and metonymy A consecutive interaction between metaphor and metonymy occurs when one of the motivational links in the semantics of the composite expression involves a sequence of two semantic extensions. A first example is presented in Figure 2, which contains an analysis of the compound schapenkop. Literally, the word means 'sheep's head' (and the word could actually be used in this sense, in contrast with some of the other compounds that we will analyze presently, in which the literal reading is not conventionalized). The derived reading of schapenkop is 'dumb person', and this reading seems to involve two steps: first, 'sheep's head' is metaphorically extended towards the reading 'a (human) head like that of a sheep, a stupid head', and second, a metonymical step leads to 'a person with a head like that of a sheep, a stupid person'. (The representation 9 in Figure 2 can be completed on the bottom level of the prism, but that is a step that will be taken in section 3.3.)
The interaction ofmetaphor and metonymy
METAPHOR
--8---
213
METONYMY
1 Sheep's head 2 Sheep 3 Head 4 (Human) head like that of a sheep 5 Stupid person
Figure 2. The prismatic structure of schapenkop
A similar consecutive combination occurs in an idiomatic expression such as groen achter de oren zien. Literally, the reading is 'to be green behind the ears', which is then metaphorically interpreted as 'to be young, (as if people are like fruit that have a green color in the first stage of their existence, before they reach maturity). As with so many other expressions indicating young age, the expression next receives a further extension to the reading 'inexperienced, naive'. We may also note that the consecutive sequence need not always involve an alternation of metaphor and metonymy. In an example like hanglip, for instance, two consecutive metonymical steps may be identified. The literal reading is composed of the noun lip 'lip', and the verbal stem hang 'to hang'; the literal reading can therefore be paraphrased as 'hanging lip'. A first metonymical extension (involving the metonymical relationship between a specific feature and the bearer of that feature) leads to 'a person with a hanging or protruding (lower) lip'. A second metonymical extension (involving the metonymical relationship between a typical effect and the usual cause of that effect) leads to 'an unhappy, sulky, pouting person' .
214
Constructions and idioms
3.3. Parallel presence of metaphor and metonymy A parallel presence between metaphor and metonymy occurs when there is a difference in type among the different motivational links that occur in the semantics of a composite expression. If, for instance, the motivational link at the top level of the prismatic structure is metaphorical and one of the links at the lower level is a metonymical one, metaphor and metonymy work in parallel to produce the derived reading of the expression, or at least, both a metaphorical and a metonymical path could be reconstructed to arrive at the derived reading.
1 Sheep's head 4 Sheep-like 7 Stupid
2 Sheep 5 Head 8 Person
3 Head 6 (Human) head like a sheep 9 Stupid person
Figure 3. The prismatic structure of schapenkop completed
As an initial step, let us complete the picture for schapenkop. At the bottom level of the prismatic structure, the intermediate reading 'a (human) head like that of a sheep, a stupid head' can be considered isomorphic: the kop-part does not undergo a semantic change, and the 'sheep-like' reading is related by metaphorical similarity to the original reading of schaap 'sheep'. The ultimate reading 'stupid person' maintains the isomorphism: there is a metaphorical link from 'sheep-like' to 'stupid', and there is a conventional part/whole-metonymy linking 'head' to 'person'. (For instance, de koppen
The interaction a/metaphor and metonymy
215
tellen 'to count the heads' is a conventional expression for 'counting the individuals in a group'.) The example shows that the motivating links at the bottom level of the prismatic structure need not be the same as the ones at the top level: the identity link is absent at the upper level. In some cases, then, we get an alternation between metaphorical and metonymical links. A case in point is the expression in de stront zitten, which may be analyzed as in Figure 4.
1 To sit in the shit 4 To be in great trouble
2 To sit (in) 5 To be situated (in)
3 Shit 6 Trouble, unpleasantness
Figure -I. The prismatic structure of in de stront zitten
The top level shift is a metonymical one of the cause/effect-type: if you are literally surrounded by excrements, you are typically in an unpleasant, troublesome situation. At the bottom level, however, the motivational links are of a metaphorical type. Zitten 'to sit' is a conventional metaphor for 'to be characterized by, to experience': dat zit goed 'that sits well' means as much as 'that is okay', and in moeilijkheden zitten 'to sit in difficulties' equals 'to have, to experience difficulties'. Stront is likewise a conventional expression for anything extremely nasty. As a slightly more complicated example, let us consider droogkloot 'boring person, bore', which can be analyzed as in Figure 5. The compositional literal reading 'dry testicle' is the basis for a roodhuid-type extension, yielding the possessive compound 'person with dry testicles'. This reading, however, is itself the input for a further metaphorical extension, leading to the 'boring person' sense. At the same time, the derived reading is isomorphic:
216
Constructions and idioms
the aspect 'boring' correlates with the 'adjective droog, and the aspect 'man, person' correlates with the noun kloot. Such an isomorphic analysis is further supported by the motivational link between the initial and the derived readings of the constituent parts. Droog is in fact conventionalized in the reading 'boring, dull, dreary' (een droge klaas is 'a tedious fellow', where klaas is originally a proper name), and kloot is a conventional derogatory expression for 'man'.
1 Dry testicle 4 Person with dry testicles 7 Unworthy man
2 Dry 5 Boring person
3 Testicle 6 Lifeless, dull
Figure 5. The prismatic structure of droogkloot
3.4. Interchangeability of metaphor/metonymy analyses The prismatic model 10 implies that the meaning of composite expressions may sometimes be construed in two different ways: from bottom to top and then from front to back at top level (which is undoubtedly the standard pathway) or conversely from front to back at bottom level, and then from bottom to top. The 'non-uniqueness of semantic solutions' (to borrow the words introduced by Nunberg 1979 in a slightly different context) goes even further, if we can show that different prismatic analyses may - with equal or near-equal plausibility - be construed for one and the same expression. To
The interaction ofmetaphor and metonymy
217
the extent that such alternatives involve different configurations of metaphorical and metonymical motivational links, we shall say that we have interchangeable metaphor/metonymy. Compounds pointing in this direction are the following. Badmuts literally means 'swimming cap' but is also used jocularly for a bald person. The shift can be analyzed in two ways. Either 'swimming cap' leads metonymically to 'a person with a swimming cap' and from there by metaphorical similarity to 'a person who looks as ifhe was wearing a swimming cap, a bald person'. Or 'swimming cap' is directly metaphorized as 'a head that looks as if it is covered by a swimming cap, a bald head' and from there metonymically to 'a bald-headed person'. The reconstruction of the semantic process can go either way, and there is no principled way to favor one analysis over the other. An example with an idiom rather than a compound is over de rooie gaan 'to go over the red one', in which 'the red one' is a red mark on a gauge indicating the point of maximal pressure. The derived reading 'to explode with anger' may be arrived at in either of the following ways. To go beyond the point of maximal pressure is the metonymical cause of a material, literal explosion (as when a boiler explodes), and the literal explosion can then be metaphorically used to conceptualize an emotional outburst. Conversely, the notion of crossing the point of maximal pressure may be directly metaphorized into the psychological domain ('go beyond the point of maximal emotional strain'), and this event may then causally (i.e. metonymically) lead to an outburst. It should be clear that the alternatives need not always involve different sequences of metaphors and metonymies. It may also be the case, for instance, that two metonymies occur in alternative orders. Zultkop provides us with an example. Literally, it means 'head filled with or made from brawn'; the derived reading is again 'stupid person'. The consecutive steps could be from the literal reading to 'stupid head' via a metonymicallink (the presence of brawn rather than brain is the cause of the idiocy), and from there to 'stupid person' via another metonymical link of the part/whole-type. Or the sequence might involve an initial part/whole-metonymy producing 'a person with a head full of brawn' and hence to 'stupid person' through the intermediary of the effect/cause-metonymy.
218
4.
Constructions and idioms
Metaphtonymy and prismatic semantics
Within the context of recent metonymy studies, there is an obvious link between the phenomena described above and the notion of metaphtonymy introduced by Louis Goossens in 1990. The two types of metaphtonymy (i.e. interaction between metaphor and metonymy) distinguished by Goossens resemble the two basic interaction types identified above: what Goossens calls metaphor from metonymy refers to a sequential operation of the two mechanisms that can be linked to the consecutive type described in section 3.2, and what Goossens calls metonymy within metaphor/ metaphor within metonymy can be associated with the simultaneous, parallel type of interaction described in section 3.3 of the present article. But how far exactly does the correspondence go? Goossens introduces metaphor from metonymy by referring to the polysemy of giggle. The verb initially means 'to laugh in a nervous way', but this meaning can be used metonymically in a context like 'Oh dear', she giggled, 'I'd quite forgotten', in which giggle comes to mean 'say while giggling'. A further extension towards 'to say as if giggling' then constitutes the 'metaphor from metonymy' reading. Whereas the consecutive operation of a metonymical and a metaphorical shift links up with the cases discussed in section 3.2, it will also be clear that the approach in the present paper has a wider scope than Goossens's. We have identified not just successions of metonymies followed by metaphors, but we have illustrated a larger variety of sequences: metaphors followed by metonymies, metonymies followed by metonymies, etc. From a broader point of view, it is important to realize that neither our 'consecutive interaction of metaphor and metonymy' nor Goossens's 'metaphor from metonymy' can be considered real innovations in the context of lexical semantics. The recognition that mechanisms of semantic extension such as metaphor and metonymy may operate in succession (and in fact, in series with multiple steps) is a natural and time-honored one in diachronic semantics (cp. Geeraerts 1997a). What is being added to that idea in the prismatic model described above, is precisely the importance of a second dimension for an adequate description of composite expressions. As to Goossens's 'metonymy within metaphor', it involves cases like catch someone's ear 'ensure someone's attention'. Such examples (which invariably involve idiomatic expressions rather than single lexemes) receive a straightforward interpretation in the context of the model sketched in the present paper, as can be gathered from the analysis in Figure 6.
The interaction ofmetaphor and metonymy
219
1 Take hold of someone's organ of hearing 2 Take hold of, capture 3 Organ of hearing 4 Force to listen, obtain the attention 5 Attain, obtain 6 Attention Figure 6. The prismatic structure of to catch someone's ear
The literal meaning 'take hold of someone's organ of hearing' is metonymically extended to 'to obtain someone's attention'. Goossens basically sees a metaphorical shift at the level of the expression as a whole, but at the same time allows for a metonymic interpretation. In the context of the prismatic model, the metonymic interpretation would seem to be more plausible: materially taking hold of someone's ear is metonymically conceptualized as a cause (or at least, a contributing factor) for getting someone's attention. At the same time, there is indeed a metaphorical aspect to the expression, but it involves the development of catch at the bottom level of the two-dimensional structure: the verb undergoes a metaphorical shift from a material to an immaterial reading. To be sure, non-uniqueness surfaces again, to the extent that it could also be said that obtaining something is the result of taking hold of it (and of course, the relationship between action and result, or cause and effect, is a metonymical one). Ear, finally, is metonymically linked to the notion of attention: the hearing organ is one of the media for channeling a person's attention. In the light of this analysis, the advantage of the prismatic model can be defined as follows: it draws the attention to the fact that the more specific
220
Constructions and idioms
semantic development is not restricted to one of the constituent parts (as might be suggested by Goossens's examples) but actually has to be determined for all of them. At least for composite expressions, then, the prismatic model appears to have a wider scope than the notions introduced by Goossens, specifically because it allows for other sequences than just metaphor from metonymy. More importantly, the model combines the intuitions behind 'metaphor from metonymy' and 'metonymy within metaphor' / 'metaphor within metonymy' by bringing the two relevant aspects of the development of composite expressions together: on the one hand, the semantic development of the expression as a whole, on the other, the role of the constituent parts of the expression and their independent development. The model thus allows for a uniform and more detailed description of the semantics of composite expressIOns.
Notes 1.
2.
In the Saussurean tradition, syntagmatic relations involve associations between linguistic expressions that exist in presentia, whereas paradigmatic relations involve associations that exist in absentia. In an expression like dames en heren, for instance, the association between dames and heren is realized in the expression dames en heren itself. The semantic association between dames 'ladies' and vrouwen 'women', on the other hand, exists even if it does not show up explicitly in the expression being used as such. Paradigmatic relations may be of various sorts; they do not just include semantic associations of the type just mentioned, but also morphological relations between a lexical base and the derivates or compounds in which it features. Among the semantic paradigmatic associations, metaphor traditionally features prominently (see among others Jakobson 1971: 74); note that in this case, the association exists not beween to different words, but between two readings of the same word. The paradigmatic relations that will be envisaged in this paper are precisely of the kind illustrated by metaphor: semantic associations between different readings of one linguistic expression. Apart from metaphor, the relevant associations involve semantic relations like metonymy, generalization, and specialization. The bottom-up interpretation is present in quotations like the following: 'Whatever linguistic meaning is, there must be some sort of compositional account of the interpretation of complex expressions as composed from the interpretations of their parts and thus ultimately from the interpretations of
The interaction ofmetaphor and metonymy
3.
4.
221
the (finitely many) simple expressions contained in them and of the syntactic structures in which they occur' (Chierchia and McConnell-Ginet 1990: 6). In most work in the tradition of Cognitive Semantics, the concept of motivation is used in a slightly broader way than the way in which it is defined here. In Lakoff (1987) and related work, for instance, 'motivation' involves the principles that explain (or make plausible) why a particular linguistic expression means what it does. The concept is explicitly introduced as an alternative to the more traditional notion of predictability: even if meanings are not entirely predictable, they may be motivated by existing tendencies and schemata. The distinction that is drawn here between 'motivation' and 'isomorphism' tries to be more specific about the general concept of motivation by distinguishing between its syntagmatic and its paradigmatic form. The concept of 'isomorphism', on the other hand, links up with existing work within the Cognitive tradition relating to the iconicity of grammar (see e.g. Haiman 1980b). Isomorphism as used here is a form of iconicity to the extent that features of meaning (in particular, its complex nature) are reflected by features of the linguistic form (viz., its composite nature). The question might be asked how we can put koe as interpreted in the figurative context into correspondence at all with koe as interpreted in the literal plane. On the one hand, we would still be willing to say that koe 'problem' corresponds with koe 'cow'. But on the other hand, there is no associative semantic link from 'cow' to 'problem'. So can we say at all that koe maps onto 'problem'? What is there to stop us from mapping bi} de horens vatten onto the 'problem' part of the idiomatic reading of the expression? In this particular instance, of course, the link between vatten 'to seize (literally)' and the figurative reading 'to tackle' is not unmotivated, and hence, byelimination. koe is easily mapped onto 'problem'. But even if this paradigmatic link between both interpretations of vatten were to be just as non-transparant as that between 'cow' and 'problem', the syntactic structure of the expression (as interpreted literally) would favor a figurative interpretation of koe as a noun, and one of vatten as a verb. This would seem to lead to the conclusion that there is always some paradigmatic link at the bottom of the prism between the literal readings of the constituent items and their figurative interpretation: at the very least, the literal reading would motivate the figurative reading because the latter is consistent with the word class of the former. In principle, such a weak form of motivation can be accounted for by accepting degrees of motivation: it will be made clear further on in the text that this is a useful step to take in any case. Empirically speaking, however, it remains to be seen whether figurative readings are always consistent with the word class of the constituent elements of the literal expression. (Notice that at least in the realm of morphology, reinterpretative processes may violate the initial
222
5.
6.
7.
8.
Constructions and idioms
syntagmatic structure: the isomorphically metanalytic reanalysis of hamburger as ham + burger violates the initial formal structure.) The representation in Figure 1 might be adapted to take into account degrees of motivation by drawing thicker or thinner lines, or by attaching a numeric weight to them. The reinterpretation processes mentioned in this paragraph constitute one kind of proof of the cognitive reality of the semantic structure embodied in the prismatic model of Figure 1. Other kinds of support for the validity of the model will have to be explored in further research. Two main alternatives have to be envisaged. First, psychological investigations (involving on-line processing tasks, or on the basis of questionnaires) may be invoked to establish the psychological reality of a specific analysis. Second, synchronic linguistic phenomena (rather than the diachronic reinterpretation processes mentioned here) may point to the structural importance of the model. Consider, for instance, the possibility of incorporating anaphoric elements in the expression. As a working hypothesis, it would seem that only those idioms that are isomorphic on the figurative level allow for the introduction of anaphoric demonstrative pronouns referring to a previous instantiation of the figuratively interpreted concepts. In this way, it is quite plausible to have a sequence like the following: 'Then came the problem of formatting the text according to the style sheet. To take this bull by the horns appeared to be much more difficult than applying the revisions required by the editors'. Because bull maps onto 'the problem' that is to be tackled, this may be introduced to refer to a previous identification of that problem. In the case of met de handen in het haar zitten 'to sit with one's hands in one's hair> to be at one's wit's end', however, the absence of a clear interpretation for handen in the idiomatic context makes sequences like the following: 'Toen moesten er camera-ready kopieen van de figuren gemaakt worden. Met deze handen in het haar te zitten bleek veel erger dan het schrijven van het oorspronkelijke artikel geweest was (Then came the problem of producing camera-ready figures. To sit with these hands in one's hair appeared to be much more taxing than writing the original paper had been)' rather implausible. This is, to be sure, just an example of the type of phenomena to be studied, but it illustrates how additional evidence for the linguistic reality of the prismatic model may be sought. An example of a reinterpretation of compounds is provided by the element scharrel-, the verbal stem of scharrelen 'rummage about, scratch, scrape'. From the compound scharrelkip 'free-ranging chicken' it is extrapolated to compounds like scharrelei 'an egg ofa free-ranging chicken': scharrel- is reinterpreted as 'produced by biological farming'. There might be some discussion with regard to this specific case: it could perhaps also be considered a generalization. Examples of competing and in-
The interaction ofmetaphor and metonymy
223
terchangeable motivations will be discussed explicitly further on in the article. 9. In Geeraerts and Bakema (1993), the prismatic representation of compounds looks slightly more complicated than in the representation used here. For instance, the compositional reading 'sheep's head' would be analyzed as a specialization of a more general reading 'head having something to do with sheep'. The motivation behind this approach is the underdetennination of compounds. Schapenkop could in fact mean many things: a sheep's head, but also 'a head with a talent for or a specific interest in sheep' (just like a studiekop is 'a bright head, a head with a talent for of a specific interest in study'). The construction of nominal compounds in Dutch does not formally differentiate between the possessive reading and the alternatives; by contrast, the syntactical construction of idiomatic expressions is much more specific as to the semantic role of the constituent parts. This underdetermination of a compound XY can be expressed by merely defining the initial compositional reading as 'a Y that has something or other to do with an X'. The next step is then invariably a semantic specialization yielding specific readings like 'a sheep's head' or 'a head with a talent for or a specific interest in sheep'. In the present article, this complication at the front end of the prismatic diagrams for compounds has not been included, because it does not add very much to the line of thought that is relevant in this text. 10. It may be useful to point out that the 'prismatic model' is a model precisely because it involves a certain degree of abstraction: the semantics of the composite expressions is rendered in a schematic way (in the sense, for instance, that the number of constituent elements is systematically reduced to two).
Section 4 Meaning and culture
Chapter 9 Looking back at anger. Cultural traditions and metaphorical patterns
Originally published as Dirk Geeraerts and Stefan Grondelaers, 1995, in John Taylor and Robert E. MacLaury (eds.), Language and the Construal o[the World 153-180. BerlinJNew York: Mouton de Gruyter. This paper addresses the question of the influence of the Medieval doctrine of the four humors and the four temperaments on our contemporary vocabulary. Specifically, given the psychological part of the humoral theory, is there any way in which the influence of the theory can be felt in the way emotions are being talked about today? Concentrating on the concept of anger, the paper shows that there is. By taking a closer historical look at Kovecses' analysis of expressions for the emotional concept 'anger' (1989), it is argued that his a-historical method obscures the possible role of cultural traditions as a source of emotion concepts. In the context of the development of Cognitive Linguistics, the paper helped to reveal the possible tension between a universalist and a culture-specific interpretation of the 'experientialist' nature of Cognitive Linguistics. Cognitive Linguistics, by its very cognitive nature, has a tendency to look at language from a psychological point of view, i.e., at language as (part of) the organization of knowledge in the individual mind. But if cognition is embodied, and if our bodies are basically the same. then underlying cognitive patterns will be largely the same for all. In Kovecses' analysis of anger metaphors in terms of the ANGER IS HEAT pattern, the universalist assumption is that the pattern is physiologically motivated. However, if the experientialist nature of language and cognition includes the cultural and historical 'situatedness' of human experience, cultural and historical factors are likely to influence our cognitive patterns. That is precisely what the present paper tries to show. In more recent publications, Kovecses (2000. 2002) accepts the point that cultural determination may play a role in shaping cognitive patterns, and investigating the universality (or the cultural specificity) of metaphorical patterns has now become a standard topic in metaphor oriented cognitive linguistic research. Such studies focusing on anger include Taylor and Mbense (1998), Yu (1998), Mikolajczuk (1998), Forceville (1999), Barcelona (2001), Gevaert (2001), Harkins and Wierzbicka (2001), Soriano (2003), and Maalej (2004). More generally, there is a growing interest in Cognitive Linguistics in the social aspects of meaning and cognition. A number of researchers (Palmer 1996, Sinha and Jensen de L6pez 2000, Harder 2003. Itkonen 2003. Tomasello 2003. Geer-
228
Aleaning and culture
aerts 2005 and others: see Kristiansen and DilVen, forthcoming) emphasize that the experientialist nature of Cognitive Linguistics does not only refer to material factors (taking a notion like 'embodiment' in a physical and physiological sense) but that the cultural environment and the socially interactive nature of language should be recognized as primary elements of a cognitive approach. Further, laying the foundations for a true cognitive sociolinguistics, variational phenomena are being studied empirically in work such as Kristiansen on phonetic variation (2003), Berthele (2004) on differences in syntactic construal between dialects, and Grondelaers (Grondelaers 2000) on grammatical phenomena whose distribution is determined by a combination of internal (structural or semantic) and external (contextual or sociolinguistic) factors. (See also the work mentioned in the introduction to Chapter 11.)
1.
The mysteries of masturbation
In the course of 1989, the Belgian Department of Education started a school campaign against truancy. The major slogan of the campaign read Van spijbelen word je doof 'Playing truant makes you deaf - a jocular (but probably inefficient) reference to the old belief that excessive masturbation could cause deafness. It is not likely that this belief itself is still very much alive in our post-sexual revolution, sex education era, but the very fact that knowledge of it was assumed in the campaign seems to indicate that it is still around, and that it was being handed down from educators to pupils not too long ago. But what was the origin of that belief (which, incidentally, came in a number of variants, in the sense that next to deafness, blindness and deterioration of the spinal marrow were cited as the sinful results of promiscuous self-indulgence)? Was it just a conspiratorial invention of priests and parents, intended to keep personal frustration and public morality up? Or was there an actual basis for it? Let us turn to a specialist for an answer. In 1772, the honorable doctor Tissot (member of learned societies in London, Basel, Bern, and Rotterdam) published a lengthy treatise entitled L 'Onanisme. Dissertation sur les Maladies produites par la Masturbation (Grasset, Lausanne), in which all is revealed about this 'crime obscene'. After an extensive treatment of the detrimental influence of masturbation, he asks the question 'Col11l11ent une trop grande emission de semence produitelle tous les maux que je viens de decrire?' [How does an excessive emission of sperm produce all the evils that I have just described?]. On page 69, he begins his answer with a reference to the father of medicine, Hippocrates of
Looking back at anger
229
Kos (approx. 460-377 BC). (The relevant passages are in the Hippocratic treatise known as De Genitura.) Hippocrate a cru qu'elle [la semence] se separoit de tout le corps, mais surtout de la tete. La semence de l'homme vient, dit-il, de toutes les humeurs de son corps, elle en est la partie la plus importante.... Il Ya des veines & des nerfs qui de toutes les parties du corps vont se rendre aux parties genitales; quand celles-ci se trouvent remplies & echauffees, elles eprouvent un prurit, qui se communiquant dant tout le corps, y porte une impression de chaleur & de plaisir; les humeurs entrent dans une espece de fermentation, qui en separe ce qu'il y a de plus precieux & de plus balsamique, & cette partie, ainsi separee du reste, est portee par la moele de l'epine aux organes genitaux. [Hippocrates thought that semen secreted itself from the entire body, but specifically from the head. A man's semen, he says, comes from all the humors of his body, of which it is the most important component. ... There are veins and nerves that go towards the genital organs from all over the body, and when these organs are filled up and wanned up, they experience an urge that communicates itself through the entire body, producing an impression of warmth and pleasure. The humors then enter into some sort of fermentation that separates out the most precious and balsamic substance they contain, and this part, when it is separated from the rest, is carried to the genital organs by the spinal marrow.] If, in other words, semen is produced by a process of fermentation and distillation of the 'humors', it is plausible that an overproduction of sperm weakens the body, given the vital importance of those 'humors'. Specifically, the role of the head and the spinal cord in this process explains why, in particular, the functions of the head (such as seeing and hearing), and the spinal marrow may suffer the detrimental effects of excessive sperm production. In this sense, it all falls it into place - but it only does so against the background of the doctrine of the four humors. In Classical and Medieval physiology, the human body was thought to contain four kinds of fluids, or humors, which regulated the body's functioning and whose disproportionate presence could cause illness. Admittedly, conflicting views were expressed within the humoral framework about the origins of sperm, and the humoral doctrine as a whole was no longer valid in its original Hippocratic form by the end of the eighteenth century. But, although Tissot hardly follows Hippocrates in detail, he quotes him approvingly, and like most of his contemporaries does retain the basic idea that the production of semen (and the harm-
230
Meaning and culture
ful effects of overproduction) involves the extraction of the seminal substance from the vital bodily humors. Against the background of this historical link, extending from classical antiquity well into the modem era, the lingering belief in the negative effects of masturbation appears to be a recently deceased (or at least moribund) renmant of what was once solid science. And to be sure, it is not the only relic. In the Dutch-speaking part of Belgium, for instance, the concept 'cold, inflammation of the mucous membrane of nose and throat' is expressed in a majority of the dialects by the word valling. While valling is morphologically complex (being a nominalization of the verb vallen 'to fall '), it is not semantically transparent to the majority of speakers. Historically speaking, however, its formal complexity makes perfect sense in the framework of the theory of humors: when the nose runs or sputum is expectorated, what 'falls' is nothing else than phlegm, one of the four humors. Given that the head was considered to be the major locus of phlegm (and taking into account that phlegm was described as a cold humor, in contrast with, for instance, blood as one of the warm humors), a valling or a cold is nothing but a precipitation of the brain's fluid. And of course, in non-historical parlance, the English word phlegm now refers precisely to the thick semi-fluid secretion of the mucous membranes of the respiratory passages. Similarly, we speak of catarrh (derived from the Greek katarrheo 'to flow down'), and - in the case of another disease attributed to an excess of phlegm - of rheumatism, in which the Greek verb rheo 'to flow' can be discerned (cp. Siegell968: 323). Faced with examples such as these, we would like to address the question of the influence of the humoral doctrine on our contemporary vocabulary more systematically. What other relics of the old beliefs can we find? Specifically, given the psychological part of the humoral theory, is there any way in which the influence of the theory can still be felt in the way we talk about emotions? Concentrating on the concept of anger, we will try to show that there is. By taking a closer historical look at Kovecses' analysis of emotional expressions in terms of generalized metaphors (1989), we shall argue that his a-historical method obscures the possible role of cultural traditions as a source of emotion concepts. Our purpose, in other words, is factual to the extent that we will try to establish the importance of the old humoral theory for our contemporary emotional vocabulary, critical to the extent that we will try to qualifY Kovecses' analysis, and methodological to the extent that we will stress the methodological importance of a diachronic perspective for linguistic studies with a cultural orientation. Before dealing with the
Looking back at anger
231
specific linguistic part of the investigation, we will first give some more information on the theory of humors and its historical importance.
2. The history of the humors In this section, we will briefly (and simplifyingly) present the humoral doctrine, and sketch its historical development. On various aspects of the history of medicine at large and the humoral theory in particular, more information may be found in among others Lindeboom (1985), Godderis (1988), Beek (1969), Irwin (1947), Siegel (1968), Major (1954), Schafer (1966), and Diepgen (1955). Klibansky, Panofsky and Saxl (1964) deserve to be mentioned separately for their detailed history of the humoral doctrine up to the seventeenth century (with special emphasis on the concept of melancholy). The foundations of the humoral doctrine were laid by Hippocrates of Kos. Three aspects of his approach should be mentioned: the physiological, the psychological, and the medical. Physiologically, the four humoral fluids regulate the vital processes within the human body; the secretion of the humors underlies the dynamical operation of our anatomy. Psychologically, on the other hand, they define four prototypical temperaments, i.e. a person's character is thought to be determined by the preponderance of one of the four vital fluids in his body. Thus, the choleric temperament (given to anger and irascibility) is determined by a preponderance of the yellow bile, while the melancholic, gloomy and fearful, suffers from a constitutional excess of black bile. The phlegmatic personality is typically placid and unmoved, while the sanguine temperament (defined in correlation with blood, the fourth humor) is passionate, optimistic, and brave. The singular combination of physiological and psychological concepts that characterizes the theory of humors also shows up in the fact that a disequilibrium of the fluids does not only characterize constitutional temperaments, but also causes temporary diseases - which are then typically described in bodily, biological terms as well as in psychic terms. For instance, an overproduction of yellow bile may be signaled by the patient's vomiting bile, but also by his dreaming of fire. In the same line, an excess of blood shows up in the redness of the skin and swollen veins, but also in carelessness and a certain degree of recalcitrance. In this sense, the humoral theory is a medical doctrine: it identifies diseases and their symptoms, and defines a therapy. Obviously, the basic therapeutic rule will be to restore the balance of the humors, given that a disturbance of their well-balanced proportion is the basic cause of the pathological situa-
232
Meaning and culture
tion. The long-lasting popularity of blood-letting, for instance (a standard medical practice that continued well into the nineteenth century) has its historical origins in the theory ofhumors. The connection between yellow bile and fire that was mentioned a moment ago is not accidental. It is part of a systematic correlation between the human, anatomical microcosm and the macrocosm, thought to be built up from four basic elements. Thus, yellow bile, black bile, phlegm, and blood corresponded with fire, earth, water, and air respectively. In the Aristotelian elaboration of the Hippocratic doctrine, a 'componential analysis' was added to these correlating sets of microcosmical and macrocosmical basic elements. They were defined, in fact, as combinations of four basic features: cold, warm, wet, and dry. (Needless to say, these four features are themselves related along two dimensions.) Blood was thought to be warm and wet, phlegm cold and wet, yellow bile warm and dry, and black bile cold and dry. The classical humoral doctrine received the form in which it was to dominate the Middle Ages in the work of Galen (129-199). His incorporation of the humoral approach into an encompassing theory of the human digestive system is of particular interest. Galen distinguishes between three successive 'digestions'. In the first digestive process, food is transformed into chyle in the stomach; the residue of this first digestion is faeces. In the second step, the humoral fluids enter the picture. For instance, by the transformation of chyle in the spleen, black bile is produced, while the liver refines chyle into blood; the residue of the second digestion is urine. The third step takes the blood and carries it through the body, sustaining the growth of the body; the residue of this third digestion is perspiration. But while the substance that ensures the growth and maintenance of the body is known as the nutrimental spirit, there are also two other spirits to be taken into account in this third step. In a continuing and cumulative refining process, the heart produces the vital spirit (which regulates the temperature of the body and controls the passions), and the brain produces the animal spirit (Aristotle's pneuma psychicon, which commands the movement of the body, but also feeling and the workings of the mind). Further, Galen's digestive anatomy leads to a dietary pharmacology. All plants (and foodstuffs in general) could be characterized by one of four degrees of warmth, cold, wetness, and dryness. Given that diseases are caused by an excess of one of the four humors, and given that these are themselves characterized by the four features just mentioned, the basic therapeutic rule is to put the patient on a diet that will ensure a decrease of the superfluous
Looking back at anger
233
humor. For instance, because yellow bile is hot and dry, patients suffering from choleric diseases should avoid plants such as garlic and ginger, which are both warm in the fourth degree and dry in the fourth degree. Rather, they should resort to plants such as opium (wet in the fourth degree) and henbane (cold in the fourth degree). Table 1. A system of humoral correspondences
CHARACTERISTIC ELEMENT TEMPERAMENT ORGAN COLOR TASTE SEASON WIND PLANET ANIMAL
PHLEGM
BLACK BILE
YELLOW BILE
BLOOD
cold & moist water phlegmatic brain/bladder white salty winter North moon turtle
cold & dry earth melancholic spleen black sour autumn West Saturn sparrow
warm & dry fire choleric liver/stomach yellow bitter summer South Mars lion
warm & moist air sanguine heart red sweet spring East Jupiter goat
In the course of the Middle Ages, the Galenic framework was further developed into a large-scale system of signs and symbols. In a typically medieval analogical way of thinking, widely divergent phenomena (ranging from the ages of man to astrological notions such as the system of the planets and the signs of the zodiac) were fitted into the fourfold schema presented by the medical theory. In Table I, an overview is given ofa number of those correlations. It should be mentioned, however, that the system was not entirely without unclarities (which is not surprising for a system that was to a large extent devised independently of empirical observation). For instance, while there was general agreement on the core of the system, authors would differ as to the more peripheral elements (such as the question which planet correlates with which humor; in particular, the associated animals are highly unstable across authors). Also, the system so to speak contained its own sources of confusion. There is, for instance, a marked ambiguity in the use of the concept 'blood', which was not only considered to be one of the four basic fluids, but which was also thought to transport the other humors, and which could hence also be used to refer to the mixture of the humors that was carried through the body (cp. Schiifer 1966: 4). And while the basic
234
Meaning and culture
color associated with yellow bile is obviously yellow, it was believed that yellow bile turned red when heated (and black when it was entirely burned up), so that the color 'red' could receive multiple interpretations within the system. The humoral edifice began to be undermined as soon as the Renaissance introduced renewed empirical medical investigations. Harvey's discovery of the circulation of the blood, for instance, was in direct contradiction with the traditional position of the blood in the Galenic 'digestive' system. However, the disappearance of the theory from the medical scene was only very gradual, and it took approximately another three centuries before the last vestiges of the humoral framework were finally removed. The standard view of the historians of medicine is, in fact, that only in the middle of the nineteenth century (and more particularly, with the publication of Rudolf Virchow's Die Cellularpathologie of 1858) did the humoral pathological conception receive its final blow. This 'final' character only holds, of course, for the official medical science: we have already seen in the introductory section about masturbation that traces of the old doctrine continued to exist much longer in popular belief. Along the same lines, it could probably be shown that the contemporary revival of herbalist medicine at the fringe of official medicine has direct links with the Galenic dietary pharmacology.
3. Anger in art As we have seen, the humoral doctrine had developed into a full-fledged semiotic system in the course of the Middle Ages: an ordered set of signs for medical and psychological interpretation. As a first indication of the fact that this semiotic system was not confined to the field of medicine, let us see how it influenced the artistic production of the Renaissance. We shall give two examples, one from the pictorial arts, and one from the dramatic arts. Cesare Ripa's Iconologia of 1593 was undoubtedly one of the major reference works for the seventeenth-century graphic artist. It contained a thematic inventory of the emblematic subject-matter of art, that is to say, of the topics, motifs, and symbols that could be used in paintings, drawings, engravings, and the like. There is a separate section in Ripa devoted to the four temperaments, with a detailed enumeration of the iconography associated with each of the four types. This is how Ripa introduces the choleric temperament.
Looking back at anger
235
Vn giouane magro di color gialliccio, & con sguardo fiero, che essendo quasi nudo tenghi con la destra mano vna spada nuda, stando con prontezza di voler combattere. Da vn lato (cioe per terra) sar. .. vno scudo in mezo del qual sia dipinta vna gran fiamma di fuoco, & dall'altro lato vn feroce leone. [A thin young man in yellow, with a ferocious face, almost naked, holding a drawn weapon in his right hand, ready to fight. From one side, a shield will be placed on the ground. with a flame of fire painted in the middle, and from the other side, a ferocious lion.]
Each of these characteristics is then further explained and elaborated in the course of Ripa' s expose, which is interspersed with references to and quotations from authorities such as Oalen, Ovid, Seneca, and Avicenna. In general, the attributes mentioned in Ripa' s description can be easily related to the characteristics mentioned in Table 1. Basically, the irascibility of the choleric person is symbolized by depicting him as a battle-prone warrior. Note that each of the details subtly contributes to the meaning of the whole; in particular, the fact that the young man is naked, and the fact he is not carrying his shield but that it is merely lying on the ground, indicate the impulsiveness of his hot-tempered nature: in his fits of rage, he does not even think about his own protection. If this impulsiveness is the negative side of his personality, the braveness symbolized by the lion is its positive side. Further elements that can be traced easily are the fire, and the yellow calor (corresponding, of course, with the yellow bile that is the physiological basis of this type). Less clear perhaps is the leanness of the young man's body, but this is an expression of the consuming character of the dry heat that is typical of the choleric physiology. It is worthwhile noticing that Ripa's description contains only the basic iconography of the four temperaments. It suffices to have a look at Klibansky, Panofsky and Saxl's (1964) magisterial monograph on Diirer's well-known wood-cut Melancolia I to get an idea of the intricacies and subtleties that arise when the humoral iconology is used and transformed by a truly creative artist. But the influence of the humoral semiotic system was not confined to the graphic arts. For instance, it has been described by various authors (Campbell 1930, Cruttwell 1951, Draper 1965, Schafer 1966, Pope 1985, Kail 1986) how the psychology of Shakespeare's dramatic characters unmistakenly refers to the theory of humors. Just a few quotations from The Taming of the Shrew suffice to demonstrate this.
236
Meaning and culture
(1) (2) (3)
Were I not a little pot and soon hot [IV: 1:5] Is she so hot a shrew [IV: 1: 17] I tell thee, Kate, 't was burnt and dried away, and I expressly am forbid to touch it, for it engenders choler, planteth anger; and better it were that both of us did fast, since, of ourselves, ourselves are choleric [IV: 1: 156] Gm. What say you to a neat's foot? Kath. 'Tis passing good. I prithee let me have it. Gm. I fear it is too choleric a meat. How say you to a fat tripe finely broil'd? Kath. I like it well. Good Gmmio, fetch it me. Gm. I cannot tell. I fear 'tis choleric. What say you to a piece of beef and mustard? Kath. A dish that I do love to feed upon. Gm. Ay, but the mustard is too hot a little [IV:3:25]
(4)
The conceptualization of anger in these quotations conforms to the model furnished by the theory ofhumors: anger is caused by choler (3), the production of which may be stimulated by certain kinds of food (3,4); while a choleric temperament is a permanent personality trait (3), the main attribute of the choleric personality is hotness (1,2). (This is not to say, by the way, that Shakespeare's use of the humoral doctrine is unoriginal: see Pope 1985 on the vivid and original way in which he handles the humoral concepts.) The fact that passages such as the ones quoted above can be multiplied from the work of Webster, Marlowe, or Jonson, leads Schafer (1966) to the conclusion that the humoral conception of physiology and psychology is something of a true fashion in Elizabethan drama. He attributes this to the fact that it is only in the middle of the sixteenth century that the doctrine became known to a wider audience that that of learned men who could read the medical authorities in their Latin and Greek originals. It is only, in other words, after the invention of printing that works such as Thomas Elyot's Castel of Helthe (1539), Andrew Boorde's A Breuyary of Helth (c. 1542) and A Compendyous Regyment or A Dyetary of Helth (c. 1542), or Thomas Vicary's A Profitable Treatise o.lthe Anatomie ofMans Body (1548) could be widely distributed, and that they could contribute to the spreading of the humoral doctrine to the community at large. (In Chapman 1979: 277, the wide distribution of almanacs is mentioned as a specific factor contributing to its popularity.) But if this dissemination of the doctrine of humors from the realm of learned knowledge to that of popular belief implies that it is technically a piece of gesunkenes Kulturgut, the question arises how far it
Looking back at anger
237
actually sank. In particular. how deep did it become entrenched in the language itself?
4. The lexical legacy We have already seen, in the case ofvalling, catarrh, rheumatism, and even cold, that single lexical items that are current today may be traced back to the humoral doctrine. These items are not isolated cases. In Table 2, we have systematically brought together a number of items and expressions in three European languages (English, French, and Dutch) that can be considered a part of the legacy of the theory ofhumors. Table 2. Lexical relics of the humoral doctrine
PHLEGM
ENGLISH
FRENCH
DurCH
phlegmatic
avoir un jlegme
valling
imperturbable calm, cool, apathetic, undisturbed to be imperturbable BLACK BILE
YELLOW BILE
BLOOD
(dialect) cold, infection of the nose
spleen
melancolie
zwartgallig
organ filtering the blood; sadness
sadness, sorrow, moroseness
sad, depressed (lit. black-bilious)
bilious
colere
Z 'n
angry, irascible, hot-tempered
anger, bad temper, wrath, rage
to vent (lit. to spit out) one's gall
full-blooded
avoir du sang dans les veines
warmbloedig
vigorous, hearty, sensual
to have spirit, luck
gal spuwen
passionate (lit. warm-blooded)
It will be noticed that the items exhibit various kinds of etymological or semantic relationships with regard to the older medical vocabulary. To begin with, there are items like melancolie, colere, andjlegme that refer directly to the original Latin denominations of the four basic fluids or types of personality. Next, there are items such as bilious and zwartgallig, which are based on a synonym (bile) or a translation in the vernacular (gal) of the technical term for the humor in question. Finally, there are items that have a more indirect relationship with the humors, in the sense of being metonymically related with them. Thus, spleen and valling are not formed on the basis of
238
Meaning and culture
the name of one of the basic fluids, but rather refer, respectively, to the organ typically associated with black bile (and hence to the associated temperament), and to a physiological effect thought to involve phlegm. Not included in the figure but equally revealing are items such as French humeur 'temperament; mood', that involve the generic term for the four fluids. Ifwe zoom in on one of the cells of Table 2, still further examples may be found. According to Roget's Thesaurus, the items listed under (5) all refer to anger or related concepts (the glosses are our own). (5)
choler 'anger' gall 'anger' rouse one's choler 'to elicit anger' stir one's bile 'to elicit anger' galling, 'vexing, causing anger' choleric 'irascible' liverish 'irascible' splenetic 'irascible' hot-blooded 'irascible'
.fiery 'irascible' hot-headed 'irascible' Although we will return to methodological problems more systematically further on in the paper, a brief methodological remark may be useful at this point. The basis for quoting a particular item as evidence for the influence of the humoral doctrine is the degree of etymological or semantic motivation that may be attributed to the item in question when it is interpreted in that historical light. Because items such as gall, liverish, choler(ic), and to stir one 's bile would simply remain etymological puzzles if the historical medical background were not taken into account, a humoral interpretation has explanatory value for them. But not all of the cases mentioned under (5) are equally clear. Two kinds of more or less problematic cases can be distinguished. In the first place, there are items whose global motivation in terms of the theory of humors is plausible, but whose local motivation within the theory is not entirely clear. Take the case of hot-blooded: if the typical fluid associated with anger is yellow bile, how come this expression contains a reference to the warming up of blood rather than bile? However, we have already seen that blood had a highly specific position in the whole doctrine: it is not only a humor in itself, but also carries the other humors through the body. If, then, blood can also refer to the mixture of the four humors as it circulates
Looking back at anger
239
through the body, it is not surprising that the warming up that causes anger may be metonymically said to involve the entire mixture. A similar but less easily explainable case of a possible local lack of motivation is splenetic: although this expression supports our general point that the influence of the humoral theory on our contemporary emotional vocabulary can be demonstrated, there seems to be a contradiction in the fact that items referring to the spleen may be either related to the melancholic temperament (see Table 2) or to the choleric temperament, as in (5). Given that the link between the spleen and melancholy is the orthodox one in the framework of the humoral approach, there are at least two ways in which the association between the spleen and anger could be explained. First, the association between the spleen and anger could be an effect of the confusion about aspects of the humoral doctrine that we have already drawn the attention to. Second, the association could be motivated by specific subtleties and refinements of the theory that we have not yet dealt with. In fact, Klibansky, Panofsky and Saxl (1964: 88) draw the attention to a passage in Avicenna where a distinction is made between the natural, primary form of melancholy, caused by an overproduction of black bile, and a secondary form of melancholy caused by a combustion of one of the other humors; thus, there is a specifically 'choleric' form of melancholy, which typically expresses itself as a state of frenzy (compare Starobinsky 1962, Jackson 1986 for the history of melancholy). At stake here is the notion of 'adust melancholy', which was thought to lead to more aggressive behavior and less fearfulness and sorrow than the natural melancholy that was engendered in a straightforward manner by an overabundance of black bile. (On the distinction between natural melancholy, adust melancholy, and choler - and on the confusion it leads to among scholars - see Soufas's 1990 argumentation that Don Quixote is an adust melancholic rather than the choleric type he has been made out to be in earlier humoral interpretations of Cervantes's work.) Our intention here is not to choose between these alternatives, but to make the methodological point that settling the question requires a detailed diachronic analysis of the development of the humoral theory and of its influence on our emotional vocabulary. If it is in general clear that the historical motivation behind the meaning 'irascible' of splenetic has to be sought in the older physiological-psychological conceptions of the theory ofhumors, a closer historical look at the development of that theory would yield valuable information about the specific history of splenetic. If, on the one hand, an explanation is sought in the less central aspects of the humoral theory, the historical analysis should be able to show how, for instance, the dissemina-
240
Meaning and culture
tion of Avicenna's view that was mentioned above led to the lexical association between the spleen and anger. If, on the other hand, that association is the result of a confusion, it is probably a later development, caused by impurities in the dissemination of the theory from its learned origins to the common people (or, perhaps, by the fact that the theory became less transparent when it gradually lost its scientific and medical respectability). If, then, the cases that are characterized by an apparent local lack of motivation merely establish the need for more detailed historical research, the items that raise global motivational questions are potentially more damaging for the humoral hypothesis. Consider an example like fiery: the (metaphorical) reference to heat could be attributed to the lingering influence of the humoral doctrine, but it could also be motivated on entirely different grounds. Suppose, in fact, that increased body heat is a physiological effect of being in a state of anger. and that anger is metonymically conceptualized in terms of its physiological effects. Rather than an historical motivation as a relic of a now abandoned medical theory, an expression such as fiery would then have an a-historical physiological motivation. At this point, we can include another set of expressions for the concept 'anger' into the discussion. In an analysis that has been published in several places (Kovecses 1986, Lakoff and Kovecses 1987, Lakoff 1987, Kovecses 1989), conventionalized phrases such as those in (6) have been subsumed by Kovecses and Lakoff under the general metaphor ANGER IS HEAT, which is further specified into ANGER IS THE HEAT OF A FLUID IN A CONTAINER when the heat applies to fluids, and into ANGER IS FIRE when the heat is applied to solids. (We will base our discussion on Kovecses 1989; there are only minimal differences in any case between the four published versions of the analysis). (6)
I had reached the boiling point. She was seething with rage. He lost his
cool. You make my blood boil. He wasfoaming at the mouth. He'sjust letting ofsteam. Don't get hot under the collar. Bi/~v's a hothead. They were having a heated argument. When I found out, I almost burst a blood vessel. He got red with anger. She was scarlet with rage. I was fuming. When I told him, he just exploded. Smoke was pouring out of his ears. He was breathing fire. Those are inflammatory remarks. That kindled my ire. He was consumed by his anger.
At a still lower level of analysis, these and many similar expressions are grouped together under labels such as when the intensity of anger increases, the fluid rises (his pent-up anger welled up inside him), intense anger produces steam (I was fUming), and when anger becomes too intense, the person explodes (when I told him, he just exploded). Next to the basic general
Looking back at anger
241
metaphor ANGER IS HEAT, less elaborate metaphorical patterns such as ANGER IS lNSANITY, ANGER IS AN OPPONENT, ANGER IS A DANGEROUS ANI-
and CAUSING ANGER IS TRESPASSlNG are identified. It will be obvious that the general metaphor ANGER IS THE HEAT OF A FLUID IN A CONTAlNER neatly fits into the humoral views: the body is the container of the four cardinal fluids, and anger involves the heating up of specific fluids (either yellow bile as the direct source of ire, or blood as the mixture of the four humors). However, given the alternative explanation of the general metaphor in terms of a physiological metonymy, is there any way in which we can say that the humoral hypothesis provides a better explanation of the motivation behind this particular subset of our emotional vocabulary? Apart from the general methodological point that a humoral explanation achieves greater generality by being able to combine an explanation of the cases under (5) with an explanation of those under (6), we have to consider two specific reasons for preferring it over a purely physiological explanation. First, it seems better able to motivate the reference to fluids in the expressions. Kovecses explains these references in the following terms: 'The fluid version [of the basic metaphor] is much more highly elaborated. The reason for this, we surmise, is that in our overall conceptual system we have the general metaphor the body is a container for the emotions' (1989: 53). The latter is illustrated by expressions such as he was filled with anger and she could not contain her joy. However, it is not clear how this metaphor combines with the basic ANGER IS HEAT metaphor to yield the application to fluids (as Kovecses claims it does): the fact that the body is a container for the emotions does not predispose the interpretation towards a conception of the emotions as fluids; after all, the contained emotions could just as well be solids or gases as far as the container metaphor is concerned. We will presently have more to say about the 'solids' version of the basic metaphor, but it can already be remarked here that in Kovecses's view of the matter, one would not expect the fluid version to be more elaborate than the solid version, because the container metaphor that is invoked as an explanation does not seem to favor the one over the other. Second, the humoral interpretation may help us to make sense of cases that are beyond the reach of a physiological explanation. In general, one is tempted to argue that a physiological interpretation entails that like physiological effects lead to like patterns of lexicalization. In this sense, it would be a counterargument for the physiological approach that an emotion such as shame, which is no less characterized by redness in the face (flushing) and a
MAL,
242
Meaning and culture
subjective impression of increased body temperature than anger, is not lexicalized by the same set of expressions as anger. It would make no sense, for instance, to say that one's blood boils with shame, or that someone is fuming with shame. However, Kovecses has rightly pointed out that there need not be a simple correlation between physiological effects and linguistic patterns, and that motivation does not equal prediction (1989: 85). The physiological effects of anger motivate our anger vocabulary, but because they do not predict the linguistic situation, emotions with similar physiological effects may be differently conceptualized. This element of caution does not, however, work as easily in the other direction: similar physiological effects need not have similar lexical reflections, yet similar patterns of lexicalization had better correlate with similar physiological effects if the physiological explanation is to have any generality. To take up an example, there exists a rather hackneyed set of expressions to the effect that love is a fire: you can let the flame of your love die out, you can have a steadily burning devotion for someone, and you can feel warm towards that person. On the one hand, this accords well with the humoral belief that love is one of the 'hot' emotions. On the other hand, it is physiologically unlikely that persons in love have a permanently raised skin temperature (we, at least, are not aware of physiological research to that effect). Granting, in other words, that the methodological key element is motivation rather than prediction, it does seem to be the case that taking into account the historical humoral background may lead to better motivational success. (A related point that will only be mentioned in passing here concerns the subjective experiential prominence of the physiological effects thought to underlie our emotional vocabulary. Kovecses refers to the experimental results of Ekman, Levenson and Friesen (1983) to prove that anger indeed correlates with higher skin temperature, whereas fear correlates with a decrease. However, independent evidence is needed to show that these objective increases and decreases correlate with subjective experiences of warm and cold. This question is relevant because the changes that were measured were rather small: an increase of O. 15 degrees in the case of anger,