SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Edit...
141 downloads
1508 Views
23MB 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
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (University of Ottawa)
Series IV - CURRENT ISSUES IN LINGUISTIC THEORY
Advisory Editorial Board Henning Andersen (Copenhagen);Raimo Anttila (Los Angeles) Tomaz V.Gamkrelidze (Tiflis);Hans-Heinrich Lieb (Berlin) J.Peter Maher (Chicago);Ernst Pulgram (Ann Arbor, Mich.) E.Wyn Roberts (Vancouver, B.C.);Danny Steinberg (Honolulu)
Volume 20 Neal R. Norrick Semiotic Principles in Semantic Theory
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY NEAL R. NORRICK Gesamthochschule Kassel
AMSTERDAM / JOHN BENJAMINS B.V. 1981
FOR M Y PARENTS ROSEMARY AND ROBERT NORRICK
© Copyright 1981 - John Benjamins B.V. ISSN 0304 0763 / ISBN 90 272 3513 9 No part of this book may be reproduced in any form, by print, photoprint, microfilm or any other means, without written permission from the publisher.
PREFACE
I am not yet so lost in lexicography, as to forget that words are the daughters of earth, and that things are the sons of heaven. Samuel A. Johnson (1755): Dictionary of the English Language, Preface This investigation concerns regular semantic relations. I now believe, and hope to demonstrate convincingly, that regular semantic relations depend on general semiotic principles linking signs of all types and that they motivation in morphologically complex units and creative meaning extensions and transfers as well. But originally my interest in regular semantic relations stemmed from my work on polysemy.
describ
I wanted to clearly differentiate polysemy and homophon semantic principles led me to studies of historical meaning changes tropes. Certain principles recurred in both past, documented transfers in everyday language as well as in poetry. The same principles turned up in treatments of motivation in complex units such as compounds, idioms, proverbs etc. What is more, my concurrent research in semioticsreve It gradually dawned on me that a relatively small number of semiotic motivational principles suffice to describe
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
the regular types of relations in all the linguistic cases.
After (1) working out a set of general semiotic principles, them, and (3) Investigated the effects of Including these principles in semantic theory. These three steps, their consequences and ramifications make up the body of this book. Studies preparatory to this volume have been presented at conferences and/or appeared in journals or collections of papers. I develop aspects of the underlying semiotic theory in the papers: "Transfer-Ikone: Indirekt motivierte Zeichen" read at the 2nd Semiotics Colloquium in Regensburg, at the 2nd Congress of the International Association for Semiotic Studies (IASS) in Vienna, 1979 (1979c). The preliminary in: "Regular Multiplicity of Meaning," in: H.-J. Diller et al., eds. anglistik & englischunterricht 8: Semantik, Trier: Wissenschaftliche Verlagsgesellschaft, 1979: 103-116 (1979a), and "Regular Meaning Multiplicity in Nouns," in T. Pettersson, ed. Papers from the Fifth Scandinavian Co Section Papers. Stockholm: Almqvist & Wiksell, 1979: 221-232 (1979b). Special aspects of the semantic theory were discussed in: "The lexicalization of pragmaticfunct Relations and Motivation in Idioms" read at the 14th Linguistics Co E. Weigand, eds. Perspektive: textintern. Tübingen: Niemeyer, 1980 (1980). I want to thank participants in the conferences mentioned for comments and suggestions incorporated in this final, consolidated version.
Many people have influenced the content and form of this study since its inception; my sincere thanks go out to all of them. John Odmark, whose untimely death deprived the world of a productive scholar and me of a close friend, acted as a critical audience and a reliable adviser from the start. Rolf Breuer and Volkmar Lehmann helped shape my thinking on metaphor at various stages in my work. Philip A. Luelsdorff read and discussed at length various drafts of preparatory studies, rough and final drafts of the whole manuscript. Winfried Nöth, Daniel . O'Connell and W. Terrence Gordon read and commented extensively on the VI
PREFACE
final draft. I'm indebted to Konrad Koerner for his conscientious editorial work. Finally, I thank my wife, Petra, for her encouragement, advice and proof-reading. I dedicate this book to my mom and dad, Rosemary and Robert Norrick, for instilling in me the love of language and the spirit of inquiry. Würzburg, Germany
N. R. N.
February 1981
VII
CONTENTS
PREFACE . 1. 1.1 1.2
V INTRODUCTION REGULAR RELATIONS IN SEMANTIC THEORY
1 7
1.2.2 1.2.3 1.3
The Current State of Linguistic Semantics... The Need for Regular Relational Principles in Semantics The Interpretation of Lexical Items in Contexts The Lexicon Motivation in Morphologically Complex Units. An Integrated Theory of Semantic Relations..
10 13 15 17
2.
THE SEMIOTIC BASIS OF SEMANTIC REGULARITY...
21
2 .0 2.1 2.2 2.2.1 2.2.2 2.3 2.4
Introduction Motivation in Semiotics Principles of Motivation The Principles Testing the Principles.. The Iconic Code The Indexical Code
21 22 25 26 30 31 40
1.2.1
7 9
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
3.
REGULAR SEMANTIC RELATIONS
71
3.0
Introduction
71
3.1
Principles of Motivation and Semantic Regularity
71
3.2
Regular Semantic Relational Principles
79
3.2.1
Metaphoric Relational Principles
80
3.2.2 3.3
Metonymic
Relational Principles
86
Regular Relational Principles in Semantic Theory
101
3.3.1
Comments on Semantic Theory
101
3.3.2
The Role of Semantic Relational Principles.... 103
4.
THE PROBLEM OF POLYSEMY
109
4.0
Introduction
109
4. 1
Defining
'Polysemy'
4.1.1
Polysemy and Ambiguity
4.1.2
Polysemy versus Vagueness,
109 110
Generality and Depletion 4.1.3
Testing for Multiplicity of Meaning
4.1.4
Polysemy and Homophony
111 112 . . 116
4.1.4.1 Homophony, Homography, Homonymy
116
4.1.4.2 Polysemy versus Homophony
116
4.2
Lexical Polysemy and Derived Polysemy
118
4.3
Current Proposals concerning Polysemy
122
4.3.1
Proposals concerning Lexical Polysemy
122
4.3.2
Proposals concerning Derived Polysemy
129
4. 4
Comments on Feature Transfer
135
5.
DERIVED POLYSEMY.
139
5.0
Introduction
139 X
CONTENTS
5. 1
Anomaly and Selection
139
5.2
Lexical Readings
144
5.3
Regular Relational Principles in Interpretive Semantic Theory
146
6.
THE LEXICON
155
6.0
Introduction
155
6.1
Semantic Classes and Relations in the Lexicon
155
6.2
Semantic Classes and Lexical Readings
159
6.3
Regular Relational Principles and the Lexicon
161
6.4
Via-rules in the Lexicon
163
6.5
Lexical Polysemy
167
6.6
Remarks on Lexicalization
169
7.
MOTIVATION IN COMPLEX UNITS
173
7.0
Introduction
173
7.1
Complex Units
173
7.2
Complex Units and Motivation
175
7.3
Via-rules and Motivation in Complex Units...
179
7.3.1
Componential Motivation
179
7.3.1.1
Componential Motivation in Compounds
1 79
7.3.1.2
Componential Motivation in Stock Phrases....
181
7.3.2
Composite Motivation
183
7.3.2.1
Composite Motivation in Compounds
183
7.3.2.2
Composite Motivation in Stock Phrases
184
7.3.3
Classifying Complex Units
186
7.4
Stock Phrases in Semantic Interpretation....
190
XI
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
8.
EXTRA-LINGUISTIC
INFORMATION
IN
SEMANTIC RELATIONS
195
8.0
Introduction.
195
8.1
The Role of Extra-linguistic Information
8.2
in the Establishment of Semantic Relations..
195
Extra-linguistic Information
198
8.2.1
Etymological Information
198
8.2.2
Historical Information
201
8.2.3
Object Specific Information
202
8.2.4
Pragmatic Information
204
9.
POETIC FIGURES
209
9.0
Introduction
209
9.1
Simple Replacement
210
9.2
Pointing Formulae
211
9.3
The Copula Link
212
9.4
The Make Link
213
9.5
The Genetive Link
213
9.5.1
The Three-term Formula
214
9.5.2
The Two-term Formula
215
9.6
Verb Metaphor
216
9.7
Adjective Metaphor
219
9.7.1
Metaphorically Interpreted Adjectives
219
9.7.2
Transferred Adjectives
220
9. 8
Summary
224
CONCLUSION
22 7
NOTES
231
Notes to Chapter 1
231 XII
CONTENTS
Notes to Chapter 2
232
Notes to Chapter 3
.
232
Notes to Chapter 4
233
Notes to Chapter 5
233
Notes to Chapter 6
233
Notes to Chapter 9
233
BIBLIOGRAPHY
235
INDEX
245
XIII
.
INTRODUCTION
This study represents a contribution to the theory of meaning in natural language. It proposes a semantic theory containing a set of regular relational principles. Regular relational principles enable semantic theory to describe connections from the lexical reading of a word to its fig urative contextual reading, from one variant reading of a polysemous lexical item to another, from the idiomatic reading of a compound, collocation or other stock phrase to its literal reading or to the literal reading(s) of one more of its component lexical items. Semiotic theory (e.g. Eco 1976) provides a foundation for regular relational principles in semantic theory by supplying principles defining motivated expression-content relations for signs generally. I argue that regular seman tic relational principles must derive from semiotic princi ples of this type. This ensures the psychological reality and generality of the semantic relational principles. I elaborate a set of regular semantic relational prin ciples defining motivated expression-content relations, and then investigate the consequences of including this set of principles in semantic theory. I shall argue that seman tic theory requires regular relational principles, that cur rent semantic theories are too weak in failing to describe regular semantic relations, or too strong in failing to differentiate regular from irregular cases. Contemporary treatments of semantic theory have gener ally embraced one of two divergent approaches. (1) They have defined figurative meaning as beyond the domain of semantic theory proper (e.g. Katz and Postal 1964; Katz 1972), or (2) they have developed descriptive devices so powerful as the encompass figures of all kinds (e.g. Weinreich 1966; van Dijk 1972). Neither approach seeks to dis tinguish in a principled manner between regular cases of meaning extension and transfer, and spurious or poetic cases. A semantic theory enriched with regular relational principles represents an improvement on theories of the first type in defining standard sorts of figurative exten-
2
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
sions and transfers of meaning; it represents an improve ment on the second type of theory in constraining exten sions and transfers to truly regular cases. The theory proposed here is enriched relative to those semantic theories which define figurative meanings as out side their domain. Relative to those theories focused upon figurative extensions and transfers in poetic texts for the most part, it constitutes a rigorously constrained pro posal. In neutral terms, I hope to present a semantic theo ry accounting for regular types of meaning extension and transfer, but characterizing sporadic, spurious and genu inely poetic cases as irregular. In this theory, wine of love instance a regular metonymic transfer of the literal reading of wine to 'intoxication' (by means of Metonymic Principle 1 (cause-effect); cf. 3.2.2.1 below), whereas ship of the desert does not involve a regular extension or transfer of any kind from ship to the intended reading 'camel'. This theory thereby provides a basis for the different lar figurative constructions.
An enriched semantic theory will also distinguishpoly tinct lexical readings associated with a singlephonolo variant meanings of a polysemous lexical item if they m tic relation, such readings are assigned to separate, homophonous lex 'lumber' count as variant readings of a single polysemous lexical item (by Metonymie Principle 3 (naturalsource-n object' and ball 'dance' count as separate, homophonous lexical items. Traditional theories have attempted to up hold the distinction between polysemy and homophony without defining a set of regular relational principles, which would have provided a basis for this distinction. A seman tic theory enriched with a set of regular relational prin ciples can rectify this shortcoming. If a regular rela tional principle links two distinct readings associated with a single phonological representation, they belong to a single polysemous lexical item; if no regular relational principle links such readings, they must belong to separate homophonous lexical items.
Regularly semantically related readings associated with phonologically distinct but morphologically connected lexi cal items count as regularly morphologically related lexi cal items. In such cases, regular relational principles simply describe the semantic portion of the morphological relation between the two items. The relevant morphological rule describes the relation between their forms. Write and
INTRODUCTION
3
writer possess a regular semantic connection and hence ex hibit a regular morphological relation, while mite 'bit' and miter 'ornamental cap1 possess no semantic connection and manifest no regular morphological relation, in spite of their superficially similar forms. When regular relational principles capture semantic connections between otherwise unrelated lexical items, these simply instance regularly semantically related lexi cal items. Heal and doctor manifest the same regular mean ing connection as write and writer, and therefore deserve relation by the same semantic relational principle. Regu lar relational principles thus describe semantic connec tions between phonologically identical pairs, morphologi cally related and morphologically unrelated pairs of lexi cal entries. Motivation in morphologically complex units parallels semantic relations in other lexical items. Componential motivation has traditionally been attributed to complex units consisting of one or more lexical items which retain (essential portions of) their lexical readings. Shoe horn enjoys partial motivation, because shoe retains Its lexical pound 'instrument for putting on a shoe'. Chatterbox'loqu chatter retains its lexical reading partially within the idiomatic reading of the compound. A semantic theory enriched cally relate the lexical reading of chatter 'produce in of the compound (by Metaphoric Principles 1 (object-model) ; cf. 3.2.1.1 below). Composite motivation accrues to bury the hatchet due to its 'symbolic relation' (Chafe 1968) with its idiomatic reading 'make peace'. An enriched semantic regular metonymic principle of cause-effect(cf. Metonymic Principle in 3.2.2.1 below) linking the literal and idio matic readings of this complex unit. A semantic theory en riched with regular relational principles thus provides a basis for motivation of two types in morphologically com plex units.
In each of the areas mentioned, a semantic theory en riched with a set of regular relational principles appears to provide an Intuitively correct description not obtain able from a theory lacking regular relational principles. In the following I hope to clearly demonstrate the superi ority of a semantic theory containing regular relational principles over other theories. The chapters of this study will be organized as follows. Chapter 1 presents an evaluation of the current state
4
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
of linguistic semantics, and argues the need for regular relational principles In semantic theory. It concludes with an outline of the theory advanced in the remainer of the book. Chapter 2 elaborates the semiotic theory which provides the basis for the notions of motivation and regular rel ation. Principles describing motivated expression-content relations for signs generally are identified and shown to be of universal application in sign systems. They are cat egorized as iconic or as indexical principles and then further classified. We thereby obtain an inventory of reg ular motivating principles belonging to semiotic theory. In Chapter 3 I argue that semiotic theory provides a basis for semantic theory. Special cases of semiotic prin ciples are formulated as semantic relational principles. They predict lexicalized and derived relations, and moti vation in compound units. I illustrate the interaction of semantic relational principles with other semantic mech anisms and the lexicon. Chapter 4 concerns itself with the traditional problem of polysemy. It shows that no current semantic theory can handle polysemy in a unified manner. Various current pro posals and empirical findings are amalgamated and supple mented. I narrow and define anew the notion of polysemy: polysemy is distinguished from homophony with the help of the regular relational principles defined previously. I divide polysemy into lexicalized and derived cases. Then Chapter 5 takes over the discussion of derived polysemy. After clarifying my stand on the notion of anom aly, the status of selectional restrictions and the form of lexical readings, I describe the functioning of regular relational principles in an interpretive semantic theory of the type presented in Katz (1972). Along with feature transfer rules (cf. Weinreich 1966), regular relational principles take care of shifts in selectional properties as they introduce figurative interpretations for lexical readings. Allowing for simultaneous dual processing of figurative constructions, a wide range of possible (in cluding mixed) figurative interpretations can be obtained. Chapter 6 concerns the organization of the lexicon. In it I discuss the consequences of structuring the lexicon around semantic classes. I then introduce the device of the semantic via-rule, In order to establish links between lexical entries on the basis of regular relational prin ciples. In this framework I go on to explicate lexical polysemy and other regular lexical relations. Chapter 7 extends the treatment of simplex lexical items to complex units, i.e. to compounds and stock phrases. I distinguish various classes of complex units
INTRODUCTION
5
on the basis of their types and degrees of motivation. I differentiate componential and composite motivation in complex units. A complex unit receives componential moti vation through via-rule links from its idiomatic reading to the lexical readings of its components; it receives composite motivation through a via-rule link from its idiomatic reading to its literal reading as a single unit. I argue that the semantic interpretation of idiomatic stock phrases should be accomplished by referring to the surface structure level only. In Chapter 8 I examine the role of extra-linguistic in formation in the establishment of semantic relations. I extend via-rules to include extra-linguistic information of three basic types: etymological, historical, and object specific. I recognize pragmatic via-rules in addition to semantic via-rules. Via-rules of both sorts can appear in individual lexical entries without abolishing the div ision between the lexicon and the encyclopedia in linguis tic theory. Chapter 9 contains an attempt to apply the semantic theory developed in this study to lyric poetry. I note the strengths and weaknesses of a theory enriched with regular relational principles in dealing with complex figurative constructions as classified by Brooke-Rose (1958). In cases where the theory fails to yield satis factory figurative interpretations, I suggest alternative analyses. Finally, in Chapter 10 I review the consequences of the present approach for semantic theory and linguistic theory generally. In addition, I point out areas for further investigation.
1.
1.1
REGULAR RELATIONS IN SEMANTIC THEORY
The Current
State
of Linguistic
Semantics
Current theories of linguistic semantics define a set of entities on various levels of description and an in ventory of relations holding between them, both within and across levels. However, the theories as yet possess no sure means of deciding in crucial cases whether a particular relation should be regarded as spurious or regular, and hence included in semantic theory. This state of affairs has consequences for decisions concern ing what should count as an entity and what should not. In particular, the theories lack a principled means of distinguishing a contextual variant of a lexical item otherwise recognized from a unique lexical item which must be listed as such. Thus, cup 'container' and cup 'volume' will be treated as two separate entities in a semantic theory with no principle relating 'container' and 'content' readings generally; in a semantic theory stating such a principle, one reading for cup will suffice, since the other can be derived from it in regular fashion. The inverse relation between entities and relational principles may be exploited to attain maximum elegance in a theory, while ignoring empirical consequences. But the relation between entities and principles can itself be considered empirically significant and tested against appropriate evidence. If this relation is viewed as an empirical matter, some basis extrinsic to the theory tain a relational principle and one lexical reading for cup . or two readings for cup but no relational principle. Appeal to the simplicity metric in such cases belies an inability to unearth any more solid motivation for the decision. So long as statements with dubious claim to any sort of regularity can be freely concocted to estab lish connections from lexical readings to contextual oc currences, and between lexical readings, no improvement can be expected. The situation can be ameliorated only
itself
8
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
by searching for (natural) constraints on semantic rela tional principles, by requiring of them a high degree of regularity. In theories of syntax and phonology steps have been taken in the direction of constraining individual rules and circumscribing rule types. Indeed, much recent work, especially within the transformational paradigm, has concerned itself with precisely this problem1. Semantics, however, has barely begun to think in terms of constrain ing the devices it is developing or of narrowing the scope of 'regularities' so recently discovered and formu lated (but cf. Apresjan 1974). The present work, spurred by the dearth of solutions offered from other quarters, aims at getting clear about what counts as regularity in semantic theory, at determining the set of empirically founded relational principles. Once a well limned set of regular relational principles has been recognized, the arbitrary interchangeability of principles and en tities ceases. If a principle is available, it applies, and the entities in question are regularly related; if no regular principle is forthcoming, the entities remain unrelated. An empirically secured inventory of regular semantic relational principles can thus put an end to many traditionally disputed cases and nagging unclarities in semantic theory. While extrinsic motivation for semantic principles is clearly desirable, it has gone unsought for by con temporary semantic theories. Several reasons may be cited for this weakness. Perhaps most importantly, serious interest in developing linguistic semantics into a consistent syn chronic theory is quite recent. Researchers have justi fiably been more concerned with enlarging and refining their repertoire of tools for analyzing systems of mean ing than with imposing restrictions on their number or nature. Indeed, for the most part, semantic theories have erred in describing too little rather than too much (this has not been the case in syntactic and phonological theories). Since they abstracted away from most complex and/or creative types of meaning, they required only rather truncated systems of relational principles, which presented no danger of yielding too powerful a theory. Only as researchers grew confident enough to tackle the vagaries of creative language use (e.g. Weinreich 1966; Leech 1969; Abraham and Braunmüller 1971; van Dijk 1972; Gak 1976; Levin 1977; cf. Abraham 1975; van Dijk 19 75; Bergmann 1979 on figurative constructions) and complex lexical connections (e.g. Gruber 1976; Fillmore 19 71a; Apresjan 19 74; Lehrer 1974b; Jackendoff 1975;
REGULAR RELATIONS IN SEMANTIC THEORY
9
McCawley 19 78) did they begin enriching semantic theory to a point where the issue of constraints could meaning fully arise. It seems to the present writer, however, that the point has been reached and passed, that the issue is ripe for attention. Other factors working against recognition of the question of constraints on relational principles may be seen in the fact that many semanticists have been more concerned with sentence meaning than with word meanings; even studies apparently directed at relations between lexi cal items (e.g. Gruber 1976; Fillmore 1971b) have treated these only derivatively, their primary concern being (syntactic) relations between sentences. Interest in the problem of relating lexical entries in a manner in tended to mirror speakers' intuitions about word meanings as such rather than the exigencies of some syntactic theory have been pretty well limited to McCawley's (1968a; 196 8b) suggestion to collapse the lexical information for phonologically identical, morphologically related items assigned to distinct syntactic categories into a single, compound lexical entry (cf. Chomsky 1970), and Jackendoff's (1975) proposal to capture semantic connections between morphologically related items in entry redundancy rules, which characterize (some portion of) the semantic informa tion in a lexical entry as redundant on the basis of a link to some other entry (cf. Katz's 1972 feature redundancy rules), except for systems constructed around more or less traditionally understood semantic fields (e.g. Coseriu 1967; Lehrer 1974b).2 In addition, concern with entry redundancy rules in semantic research has consistently lagged behind concern with rules predicting redundancy between individual fea tures or feature complexes, including not only feature redundancy rules as such, but meaning postulates as well, in spite of psycholinguistic studies indicating the neces sity of describing connections between lexical entries in an account of human language competence (e.g. Fromkin 1971; Fillenbaum and Rapoport 1971; Lehrer 1974a; Forster 1976). Nevertheless, as increasing attention is paid to semantic relations among lexical items and from lexical entries to contextual (including figurative) readings, the question of which relational principles are to be recog nized and on what basis gains importance. 1.2 The Need for
Regular
Relational
Principles
in
Semantics
The weaknesses resulting to semantic theory from the failure to define a set of regular relational principles are manifold. They manifest themselves in three principal
10
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
areas: (1) the interpretation of lexical items in contexts, (2) the structure of the lexicon, (3) motivation in mor phologically complex units. I treat these three points in the order given. 1.2.1 The Interpretation
of Lexical
Items
in
Contexts
In any theory, the semantic interpretation of sentences (or any other units from the level of the lexical item up- . wards) will at some point involve the pairing of sentential constituents with items listed in an inventory which links each to some sort of semantic reading. I designate this inventory as the lexicon. In many cases, let us assume, the readings extracted from the lexicon may be hooked together in a manner reflecting the relations they contract with one another in the syntactic structure of the sentence to yield a semantic reading for that sentence (on Katz's 1972 definition). The rules relating sentence constituents with lexical entries are straight-forward, requiring simple non-dis tinctness for a match. Problems arise only when the sen tence constituent requires a semantic reading distinct from that supplied by the lexicon due to features of its con text. Except for truly anomalous cases, which are rare enough in real-life contexts, such sentences generally re ceive consistent interpretations from native speakers. In traditional terms, this means that the constituent in question is assigned a figurative interpretation. That is to say that the semantic contribution of the constituent in the reading for the sentence ends up distinct from but re lated (by regular principle) to its lexical reading. Since speakers tend to exhibit agreement in judgments of correct, possible and impossible figurative interpretations for sentences, semantic theory is faced with the problem of defining those relations which regularly obtain between contextual interpretations and lexical readings for lexical items, or at least those which may possibly obtain. Three sorts of solutions have been proposed to this problem, disregarding unformalized suggestions proffered by theories of poetic language from Aristotle down to the present. (1) The first (pseudo-)solution declares figurative meaning to be outside the realm of semantic theory proper, and labels all sentences involving figurative readings anomalous on these readings (e.g. Katz and Postal 1964; Katz 1972). This position fails in not distinguishing figurative meaning from true anomaly, and a fortiori in not describing the relation between figurative readings
REGULAR RELATIONS IN SEMANTIC THEORY
11
and connections within the lexicon, which derive histori cally from figurative extensions and transfers of literal readings for lexical items. Obviously this non-solution cannot distinguish the various sorts of figurative meaning. (2) The second attempt at solution allows for a trans fer of features from predicate terms to modified terms generally (Weinreich 1966; Baumgärtner 1969; van Dijk 1972; Levin 1977). Here some creativity of interpretation is recognized, but restricted to a single type. It does not do justice even to the traditional distinction between metaphor and metonymy found in accounts of rhetorical language (see e.g. Plett 1971; cf. Jakobson 1953; 1954). No attempt is made to link the clearly allied principles involved in both creative figurative extension and trans fer, and lexicalized semantic relations. In allowing feature transfers whenever necessary, such theories fail to recognize regularities and impossibilities within the domain of figurative interpretations. Processes of feature transfer must be constrained in some manner to reduce their power, but at the same time they are too weak in other respects. In particular, feature transfers take place only when selectional clashes occur; hence, they cannot account for figurative interpretations of structures without selectional violations, although these clearly occur. Sentence (1) may be assigned a consistent literal interpretation on which Jack became unconscious after being struck by a book. Consequently, feature trans fer cannot apply to yield the figurative reading on which Jack was intellectually deeply affected by the book. (1) This book knocked Jack out. Feature transfer, however, possesses the great advantage that it automatically accounts for the selectional reorgan ization of items whose meanings it transforms. I favor allowing transfers only from predicates to nominais, not in both directions. I agree with Levin (1977) in limiting transfers to selectional features only. Also feature transfers should be prohibited from creating or abolishing logical contradictions. (3) A third approach posits a set of semantic or prag matic rules describing various metaphoric extensions and metonymic transfers from lexical readings to contextual interpretations. The theories representing this approach can be further distinguished. One group of scholars treats figurative language as essentially rule breaking performance (e.g. Cohen and Margalit 1972; Price 1974; Davidson 1978).
12
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
A second group views figurative language as a rule governed performance phenomenon (e.g. Matthews 1971; Cohen 19 75; Abraham 1975; Horrocks 1976). Reddy (1969), Schofer and Rice (1977) and Nunberg (1978) attempt to handle figurative language in referen tial terms. A fourth and final group of researches seeks to elaborate semantic theory in such a manner as to enable it to describe much figurative language in terms of re gular semantic principles (e.g. Weinreich 1966; Leech 1969, 1974; Lehmann 1975; Gak 1976; and, to some extent. Droste 19 76). All the theories representing this third approach have in common a failure to establish any formal connection between figurative language and lexical con nections (but cf. Apresjan 1974; and Lehmann 1975). A differentiated account of speakers' ability to interpret figurative constructions is offered, but no constraints are proposed to distinguish regular from impossible cases. The problem of selection is generally ignored. Some method must be found to hook up meaning changes in duced by creative principles and the selectional prop erties of the items involved. All but the final group of proposals in this third group are doomed in this respect, because the others would require connections from per formance or reference rules to semantic mechanisms, which would hopelessly complicate the theory as a whole. Failure in this respect automatically entails an inability to distinguish figurative constructions from truly anomalous ones. Further, this failure brings with it the intuitively absurd conclusion that all figurative constructions are ill-formed, at least on their figurative interpretations. To summarize, all the solutions hitherto proposed fail to provide a battery of regular principles for relating lexical readings with their (possible) contextual inter pretations, or at least to define narrow constraints on admissible semantic principles. Nor do any current pro posals attempt to describe the connection between creative figurative language and established lexical relations (but cf. Hockett 1956, 1958; and Chafe 1970 for remarks on lexicalization). They thereby fail to capture an im portant semantic generalization, and to explain why original figures tend to mirror lexicalized 'dead' The present investigation will seek to amalgamate the semantically oriented proposals of the last group with the feature transfer theories in such a manner as to meet the objections raised to each here. In particular I shall con cern myself with relating the productive principles in volved in the creation and interpretation of original
metaphor
REGULAR RELATIONS IN SEMANTIC THEORY
13
figurative constructions with the passive relations link ing entries within the lexicon. This final consideration leads to the following discussion of the second primary weakness in contemporary semantic theories resulting from their failure to define a set of regular relational prin ciples. 1.2.2 The
Lexicon
In lexical semantics the problems ascribable to the lack of a systematic set of regular relational principles make themselves felt immediately. Inasmuch as the lexicon must list the words of a language, a theory of lexical semantics must have at its disposal a means of deciding what counts as a word in semantic terms3. The decision to treat two phonologically and syntactically identical units manifesting some semantic diversity as one word or as two depends upon whether a regular semantic relation obtains between the readings of the units. The precise sort of relational principle responsible for establishing a given semantic connection will determine whether the two units are considered alternative readings associated with a polysemous lexical item or simply permissible variants of the reading of a vague or depleted lexical item (see 4.2.4 below). Even phonologically identical pairs of words dif fering syntactically and semantically might appear in a single lexical entry, if they are viewed as regularly sem antically related (cf. McCawley 1968a, 1968b; Chomsky 1970). In both cases a judgment must rely upon a welldefined set of regular semantic relational rules. A theory of lexical semantics has essentially only two options in this situation: (1) to abdicate its respon sibility for taking correct decisions in crucial cases, declaring that they are in principle unresolvable (as do Lyons 1968, 1977; Bendix 1971; Hudson 1976b) or (2) to set about the task of identifying the regular relational prin ciples required (see Leech 1969, 1974; Apresjan 1974). Since speakers agree in judgments of relatedness in many cases (cf. Lehrer 1974a), an adequate semantic theory should account for this ability in terms of general prin ciples. The present work takes seriously the goal of identifying the relevant principles and securing them in a firm foundation outside the theory of semantics. The work of Gruber (Gruber 19 76) offers an interest ing example of an attempt to capture lexical relations of various sorts without appeal to relational principles. By accepting the hypotheses that lexical entries take the form of (sub-)trees of the same sort which occur in
14
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
syntactic-semantic structures, and that these trees may be inserted for segments of derivational structure at arbit rary points between the level of input to the component which delivers representations to the semantic-logical calculus and that yielding the input to the phonological component, Gruber can define semantic relations in previ ously unavailable ways. He handles relations of comple mentarity (e.g. buy and sell) as differences in direction of movement (here: of a transaction) described at the prelexical level. The polysemy of lexical items instances alternative readings dependent upon point of view (e.g. book as concrete object versus book as an abstract mental construction, text) is treated by allowing complex sub trees including elements of both readings in a single entry. The polysemy of large classes of items which predictably alternate between readings of two sorts (e.g. voll as a stative verb versus voll as a causative verb) is described in general rules applying to appropriate classes entire. This seems a lot of apparatus for relations which are of essentially the same type, viz. semantic connec tions between lexical entries. In the present study, re lations of complementarity and polysemy of both sorts find expression in connections defined by regular rela tional principles without affecting the syntactic compo nent, as Gruber's proposal does. The approach centering on relational principles clearly simplifies the theory as a whole, especially when we realize that Gruber's treatment of the second type of polysemy itself involves relational principles. Once the set of lexical entries has been circum scribed, a theory of the lexicon must determine the prin ciples according to which they will be ordered into some empirically motivated structure. Lexical theory has advanced far beyond the stage where the lexicon could be defined as the "complete list of irregularities" (Bloomfield 1933) or as "an unordered list of all lexical formatives" (Chomsky 1965). In semantic terms the lexicon must be structured in such a manner as to represent seman tic relations of various sorts holding between its compo nent entries4. One traditional approach has it that the lexicon should be structured around semantic fields (Lehrer 1974b). A recent suggestion pertaining to the arrange ment of semantic fields themselves places salient proto types at their center (cf. Fillmore 1975). Contemporary psycholinguistic research supports the notion of the sem antic field (cf. Fillenbaum and Rapoport 1971; Forster 1976; Lehrer 1978), and the conception of the lexicon presented in this study incorporates the notion. But the
REGULAR RELATIONS IN SEMANTIC THEORY
15
field approach, even when supplemented by additional sem antic principles, suffices only to represent the lexical relations of hyponymy, oppositeness, antonymy, and synonymy on a particular reading. Relations of polysemy and semantic connections across field boundaries, including those between morphologically linked pairs and those based upon principles of metaphoric extension or metonymic transfer, remain unexpressed within a lexical model relating entries with semantic fields alone. For these relations further mechanisms are required, mechanisms connecting entries across semantic fields. The entry redundancy rules proposed by Jackendoff (1975) go part of the way toward fulfilling thisrequirement.Jackendoff, items manifesting morphological connections. One might nevertheless formulate rules in keeping with the spirit of Jackendoff's proposal which allow the description of relations between lexical entries generally. The device of the semantic via-rule (cf. Norrick 1979a) to be developed below serves precisely this purpose. A semantic via-rule attached to a particular lexical entry links it to another by citing the semantic relational principle from the in ventory which describes their relation. Recall the example of cup 'container' and cup 'volume' from above. Assuming both are listed (one redundantly) in the lexicon, the entry for the 'volume' reading will contain a via-rule like (R 1), where Metonymic Principle 12 (containercontent) is one we define as belonging to the inventory of regular relational principles in 3.2.2.12 below. (R 1) Via-Rule 1 : Related to cup 'container' via (container-content).
Metonymie Principle 12
Via rules can thus make use of the same set of principles required to predict productive cases of metaphoric ex tension and metonymie transfer. Via-rules thereby offer a simple explanation of the relation between productive principles external to the lexicon and static lexicalized connections. At the same time, they capture a wider range of semantic connections and involve less complication of the lexicon than other sorts of entry redundancy rules through access to the lexicon-external inventory of regu lar relational principles. 1.2.3 Motivation
in Morphologically
Complex
Units
The notion of motivation bears various significances for various linguistic theories. Some semanticists have
16
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
been interested in semantic motivation of the type inhering in certain onomatopoetic words, whose so-called sound sym bolism has its basis in an iconic relation from the word qua phonological unit to the object it denotes, or to a sound customarily associated with this object (cf. Ullmann 1957; Wescott 1971). We shall not be concerned with this type of motivation in the present investigation. Since we aim primarily at describing relations between semantic readings, we are instead interested in the sort of motivation ascribed to morphologically complex units, viz. compounds and stock phrases such as phrasal verbs, collocations, tournures, irreversible binominals, clichés, proverbs etc., insofar as speakers store them as such. Linguistic theories of all kinds have treated motivation in compounds (e.g. Lees 1960; Marchand 1969; Brekle 1970; Lipka 1972; Shaw 1979). The notion of motivation, however, presents difficulties resulting from the lack of any welldefined set of motivating principles. This should not be surprising, since the principles cited in support of cer tain cases of motivation in compounds converge with those responsible for productive meaning extensions and transfers as well as for lexical relations. In particular, various metaphoric and metonymic (including synechdoche; cf. Jakobson 1953, 1954) principles are involved in all three cases. Hence, in elaborating a set of regular relational principles, we provide a basis for predicting certain types of motivation in compounds. The concept of motivation also occurs in discussions of stock phrases (e.g. Chafe 1968; Weinreich 1969; Makkai 1972; Newmeyer 1974). Stock phrases, like compounds and derived structures generally, manifest complex morphology. They are segmentable into recognizable morphemes, which often retain their non-idiomatic readings or can be de rived from them by regular processes. Clearly, the motiv ation assigned to an idiom should reflect the degree to which its components are recognizable occurences of in dependent lexical items. Yet no treatment of idioms to date has attempted to describe the relevant relation, although Chafe (1968) suggests an approach. Nor has any account been offered of the composite motivation stock idiomatic phrases derive through metaphoric or metonymic relations from their literal to their idiomatic readings as wholes. We will be attempting to remedy these two short comings in the theory of idiomatic phrases in the follow ing pages. In particular, the regular relational prin ciples required elsewhere in the theory will be extended to describe both componential and composite motivation in morphologically complex units.
REGULAR RELATIONS IN SEMANTIC THEORY 1.3
An Integrated
Theory
of Semantic
17
Relations
The theory to be described in detail below takes the following form. Semantic theory will contain regular relational prin ciples. The foundation for these principles must be sought outside the domain of semantics itself, in order to avoid the pitfalls of the petitio principii. Semiotics, concerned as it is-with relations between signs generally, offers the needed foundation. By eliciting confirmation for the relational principles proposed as part of semantic theory beyond the bounds of the theory proper, we secure motiv ation for one portion of the theory and, due to the inter connections described above, for the theory as a whole. Semiotics has progressed far enough to enable it to supply linguistics with a fairly complete framework for the dis covery, testing and classification of relational principles (e.g. in Eco 1976). We require of any relational principle proposed as part of semantic theory that it derive from a semiotic principle defining standard motivated connections between sign expressions and contents, that it exhibit re gular applications in various sign systems. This restric tion provides a psychologically founded basis for identify ing and circumscribing the set of relational principles in semantic theory. The principles function to define regular relations within the lexicon and from lexical entries to contextual occurrences. They apply optionally but without exception to entire classes of lexical items, thereby predicting standard sorts of contextual interpretations (e.g. generic readings for items lexicalized on their particular read ings) . They thus determine multiple readings for items which, if such principles were lacking, would have to be listed in separate entries or in composite entries in cluding both readings. As such, they represent a vast simplification of the lexicon. The principles also describe metaphoric and metonymic relations from lexical readings to contextual readings for lexical items. In order to avoid repeating the relevant principles as part of each lexical entry to which they apply, we in troduce the device of the via-rule. Semantic via-rules capture regular semantic connections between variant read ings of single, polysemous lexical items,which cannot be listed together in a composite entry, because this would violate the principle of ordering entries into semantic classes in the lexicon. In addition, semantic via-rules describe semantic connections between morphologically re lated pairs like g l a s s - g l a s s y , as well as entries for items exhibiting purely semantic relations such as buy-sell and
18
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
butcher-meat. Since the lexicon is structured around sem antic fields, no via-rules are required for connecting word pairs exhibiting relations of hyponymy like walkstagger, (close) synonymy like bucket-pail, complemen tarity (binary opposition) like female-male, antonymy (gradable opposition) like big-little, converseness like wife-husband, and incompatibility like violet-green (fol lowing Lyons' classification system: Lyons 1977; 270 ff. ) . Regular relational principles also apply during the process of semantic interpretation to relate items in figurative constructions with the appropriate lexical readings. Note that this statement and the ensuing dis cussion are cast in terms of interpretive semantic theory roughly along the lines of Katz (1972) for the most part, although attempts will be made at crucial points to describe specific proposals within a generative semantic framework and to indicate their consequences for it; our orientation toward Katz (1972) reflects current practice and a desire to facilitate comparison to a greater degree than personal conviction. The semantic principles apply optionally, interspersed with successive applications of the projection rule (Katz 1972: 113 f f . ) , whenever the conditions they impose are met. Inclusion of regular re lational principles in semantic theory rescues it from the absurdity of having to label as anomalous perfectly ac ceptable sentences involving figurative removals from lexi cal readings, and from the inadequacy of failing to derive all but a single non-figurative interpretation for a sen tence (often enough not even the most likely interpret ation; recall sentence (1) above). The principles account for motivation in complex units, both compounds and stock phrases. They relate idiomatic readings of individual items in complex units with their lexical readings,thus allowing determination of componential enjoys metaphoric componential motivation, because the literal lexical reading for leg is related by Metaphoric Principle 1 (object-model) (see 3.2.1.1 below) to the reading for the compound. They also relate idiomatic readings of complex units with their literal readings, thus allowing determination of their composite motivation.
Strike
while
the
iron
is
hot,
for instance, enjoys meta
phoric composite motivation, because its literal reading is related by Metaphoric Principle 3 (specific case-general rule) (see 3.2.1.3 below) to its idiomatic reading 'accomplish anything while auspicious circumstances obtain', A consequence of this analysis is that many stock phrases considered idiomatic in other theories turn out to be fully compositely motivated, i.e. interpretable with
motivat
REGULAR RELATIONS IN SEMANTIC THEORY
19
regular devices in the semantic theory. The inclusion of regular relational principles in sem antic theory also allows us to give substance to the ancient distinction between polysemy and homophony, where by polysemy requires phonological identity and a regular meaning relation between two elements, whereas homophony entails phonological identity alone. I assign two phonologically identical, semantically distinct lexical entries to a single polysemous lexical item if and only if they are linked by a semantic via-rule (because a via-rule by definition cites a regular relational principle). Phonologically identical, semantically distinct lexical entries unlinked by a semantic via-rule amount simply to a Even if all speakers share a common set of regular relational principles as part of their linguistic compet ence, they may still identify different pairs of elements as related or even as polysemous, because speakers differ in their extra-linguistic knowledge, which allows one speaker to recognize a regularity invisible to others. Some speakers will perceive a semantic connection between leech 'worm' and leech 'physician', and classify them as variant readings of a polysemous lexical item on the basis of the extra-linguistic information that physicians used to treat patients with leeches (by Metonymic Principle 8 (agent-instrument); cf. 3.2.2.8 below). By extending viarules to include extra-linguistic information, we can represent idiolectal connections without having to include extra-linguistic information in lexical readings as such. Semantic via-rules allow us to represent relations which can, but need not be part of the mental lexicon viewed as a model of simple competence. Truly poetic constructions often exceed the bounds of regular relational principles. Complex poetic figures typically involve not only relational principles, but feature transfer of the sort Weinreich (1966) describes, and even the over-extension of morphological and syntactic rules. The possible combinations of these devices in a single sentence account for much of the ambiguity inherent in lyric poetry. Allowing for simultaneous multiple pro cessing of figurative constructions, a wide range of possi ble (including mixed) interpretations can be obtained for them. Hence, determination of the set of regular rela tional principles provides a point of departure for ex plorations into poetic language and poetics generally.
homophono
2.
THE SEMIOTIC BASIS OF SEMANTIC REGULARITY
To me, there appear to be only three principles of connection among ideas, namely, resem blance, contiguity in
time
or
place, and cause or effect. That these principles serve to connect ideas will not, I believe be much doubted, A picture naturally leads our thoughts to the original : the mention of one apartment in a building naturally introduces an inquiry or discourse con cerning the others: and if we think of a wound, we can scarcely forbear reflecting on the pain which follows it. Hume (1748): An Enquiry Con cerning Human Understanding 2.0
Introduction
This chapter propounds the semiotic theory which pro vides the basis for the semantic analysis to follow. It begins with a general discussion of motivation in semiotics. I then elaborate the notion of the principle of motivation, and those of an iconic and an indexical code. I propose and evaluate criteria for recognizing a statement as a reg ular principle. Then statements are posited, tested for regularity and classified. To the next chapter I postpone my arguments that regular semantic relational principles instance special cases of the general semiotic principles of motivation at issue here, so that the discussion in this chapter can concentrate on establishing the semiotic prin ciples within a predominantly non-linguistic framework.
22
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
2.1 Motivation
in
Semiotics
Eco (1976) provides the basis for the following remarks. In terms of Eco's dyadic model of the sign relation, we can tentatively define 'motivation' as the property possessed by a sign expression exhibiting a non-arbitrary relation with its content. Eco's expression and content (of a sign) parallel Saus sure' s (1916) notions of signifiant and signifié. Saussure concerned himself with the traditional question of motivation and arbitrariness in linguistic signs. However, far from providing a basis for the study of motivation, Saussure elevated the "arbitraire du signe" to a fundamental prin ciple of his linguistic speculations. Saussure investigated signs and their relations within linguistic systems, rather than signs and their relations to extra-lingusitic reality. Consequently, his model of the sign relation is dyadic: it views the sign as a dichotomy of signifiant and signifié i. e. as functioning at once on the expression and content planes of language. For Saussure, then, motivation arises essentially from a non-arbitrary relation obtaining be tween a word's form and its meaning. Serious theoretical concern with the general phenomenon of motivation in sign relations begins with the work of Peirce (see Peirce's Collected Papers (1931-1961); I employ the standard form of citation to book and paragraph of this work in the following). Fundamental to Peirce's theory of semiotics, and in contrast to Saussure's dyadic sign model, is the triadic nature of the sign. For Peirce, a sign relation necessarily consists of an object in addition to an expression and its content, in his terminology: a representamen and its inter pretant. He defines the sign as "something which stands to somebody for something in some respect or capacity" (2.22 8), where the person indicated as somebody guarantees the sign its interpretant. This notion of the sign entails a mod ified interpretation of motivation, viz. as a non-arbitrary relation holding between a representamen and its object, or, in other terms, between an element on the expression plane and its referent. I prefer the dyadic sign model of Saussure and Eco here , but Peirce's discussion of types of motivation can be adapted as well. Peirce recognizes three distinct types of relations between representamen and object. First, a representamen may be "determined by its dynamic object by virtue of its own internal nature" (Letter to Lady Welby, in Hardwick, ed. 1977: 33); the representamen possesses qualities in common with its ob jects. Such signs Peirce terms icons or iconic signs (cf.2.2 76) 1 . "Anything whatever, ... is an Icon of any-
THE SEMIOTIC BASIS
23
thing, in sofar as it is like that thing and used as a sign of it" (2.247). Second, a representamen may be "determined by its dynamic object by virtue of being in a real relation to it" (Hardwick 1977: 33); the representamen is a part of or effect of its object. Such signs Peirce terms indices. An index "refers to its object not so much because of any similarity or analogy with it, nor because it is associated with general characteristics which that object happens to possess as with icons, as because it is in dynamical (including spatial) connection both with the individual object, on the one hand, and with the senses or memory of the person for whom it serves as a sign, on the other hand" (2.305). Third, a representamen may be "deter mined by its dynamic object only in the sense that it will be so interpreted" (Hardwick 1977: 33); the representamen is connected with its object by convention or rule. Such signs Peirce terms symbols. "The symbol is connected with its object by virtue of the idea of the symbol-using mind, without which no such connection would exist" (2.299). The distinction between motivated and unmotivated signs finds expression in Eco's classification of expressioncontent relations in signs into cases of either ratio dif ficilis or vatio faoilis (cf. below). I follow Eco (1976) and current practice generally in drawing a primary dis tinction between fully coded or arbitrary signs, on the one hand, and partially coded or motivated signs, on the other hand. We must not, however, lose sight of Peirce's distinction between icons and indices. In the type of psychologically oriented theory appropriate to linguistic investigation, one can distinguish interpretive processes based on the similarity of an iconic sign with its content from those based on the contiguity of an indexical sign with its content. They share the property of necessitating more interpretive effort {ratio diffioilis) than purely arbitrary connections {ratio faoilis), but differ in re quiring recognition of distinct types of connections. Within a predominantly linguistic framework, Jakobson (1953, 1954) identifies the similarity relation with meta phor (cf. Peirce 2.277), the contiguity relation with meto nymy. He goes on to correlate similarity-metaphor with paradigmatic (substitution) relations, and contiguitymetonymy with syntagmatic (combination) relations gener ally, along with the two basic types of pathological lan guage disturbances in aphasia, and discusses the crucial pervasiveness of these two complementary principles in natural language. Of importance for the present study is the psychological relevance Jakobson evidences for the distinction between similarity and contiguity relations. This relevance clearly extends beyond the range of lin-
24
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
guistic behavior to human competence in sign systems gen erally, particularly in light of the evidence Jakobson cites from psychological disturbances in aphasia. I return to the relation of iconicity and metaphor, indexicality and metonymy below. In Eco's (1976) theory, the primary concern of semiotics lies in the relation between the expression and content of a sign (1976: 60). For Peirce, the object is of importance only in iconic and indexical signs, since he conceives motivated relations as holding between repre sentamen and object; but, as Eco (19 76: 66) points out, even the object of a sign is an abstract entity subject to cultural conventions (witness such objects as: dog, pig, whiskey, cross, justice); see Black (1962) on the 'associated objects. So long as a theory restricts its attention to intensional (as opposed to extensional) meaning, only content matters. If one discusses the code or system as such, one needs only the notion of intensional meaning (defined with in the system itself), and can ignore referential objects. Thus, Eco defines a sign function as a correlation between expression and content based upon a conventionally established code (48-49). This returns us to the Saussurean dichotomy of signifiant and signifié. Indeed, Eco, like Saussure, makes motivation a relation between a sign's ex pression and its content. The relations between representamens and their objects established by the human con sciousness during the process of sign interpretation in Peirce's theory, Eco places in the code(s) connecting ex pressions with their contents. For Eco (191), motivated signs, like unmotivated signs, involve cultural conventions (codes); the two differ operationally, viz. as cases of ratio difficilis or ratio facilis. "There is a case of ratio facilis when an ex pression-token is accorded to an expression-type, duly re corded by an expression system and, as such, forseen by a given code. There is a case of ratio difficilis when an expression-token is directly accorded to its content, whether because the corresponding expression-type does not exist as yet or because the expression-type is identical with the content-type" (183). The recognition of motivated sign relations involves a degree of 'intrinsic coding' and hence, inference on the basis of past observations, along with extrinsic (conventional) coding. "The elements (principles) of motivation exist, but they can only work when they have been conventionally accepted and coded" (199). Hence, we shall be interested in discovering both the standard principles of conventional coding used in
commo
THE SEMIOTIC BASIS
25
interpreting motivated signs and standard types of infer ence. Eco's insistence on cultural conventions connecting ex pressions with contents of signs does not contradict Peirce's theory, except in the emphasis on content rather than object as discussed above; Peirce's inclusion of the human (and, hence, cultural) interpreter introduces essen tially the same element. Thus, Eco argues that Peirce's inclusion of the interpreter is simply a methodological (rather than an empirical) guarantee of a signification, i.e. of a sign-function established by a code (Eco 1976: 16). Hence also, Eco's definition of motivation as a re lation from content to expression rather than from object to expression. Now, if the interpretation of both arbitrary and motivated signs depends upon culturally defined con ventions, semiotic theory must contain iconic and indexical codes, part of cultural sign-manipulating competence, which allow for people's recognition and decoding of motivated signs. Note that by claiming motivated signs, like arbitrary signs, are governed by culturally determined codes, Eco achieves a unified definition of sign as the correlation of an expression with a content based upon con vention, but, at the same time, he abolishes the tradi tionally upheld identification of arbitrariness and con ventionality. For Eco, all sign relations depend upon con ventions, but motivated signs require conventions of a different sort (psychologically) than arbitrary signs. The iconic code establishes a correlation between a sign expression and an already coded perceptual unit, or between a unit in a system of representation and one in a semantic system depending upon previous codification of conceptual experience (1976: 208). The iconic code must account for people's recognition of expression-content re lations based upon similarity, analogy, shared properties and "programmed stimulation" (1976: 191 ff.; cf. Eco 1972). In a parallel fashion, the indexical code accounts for the recognition of relations based upon contiguity and infer ence (cf. Norrick 1979c). Both codes consist of sets of principles, which we can conceive of and formulate as rules individuals within a given culture apply during pro cesses of sign interpretation involving a ratio difficilis. We now turn to the determination of these principles of motivation. 2.2 Principies
of
Motivation
As we have seen, Peirce explains iconicity primarily in terms of similarity relations, but he has precious little to offer on the problem of what counts as similarity and
26
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
how we recognize it. Eco (1972, 1976) sets out to clar ify iconicity by getting clear about our perception of similarity. He begins developing the notion of an iconic code, which establishes essentially conventional relations of similarity between iconic signs and standard modes of perception. But Eco does not explicitly exemplify any of the principles which would be contained in such an iconic code, if we understand this code as a set of statements their perceptual contents. Nor does he set down any con ditions on the form or functioning of such principles. Peirce's notion of contiguity, upon which he founds his definition of indexicality, also remains largely un explored. His reduction of association by contiguity to a compulsion to connect two elements (1.389) does not take us very far. We should like an identification of the re gular dynamical connections, of the principles responsible for creating this compulsion to connect, just as we re quire a statement of the principles of motivation making up the iconic code. It is my present task to provide these principles. The empirical procedure to follow seems clear enough. Observ ation of a sufficiently large sample of motivated sign re lations, those traditionally defined or accepted by a significant number of informants as such, will reveal connections of various sorts. These can then be formulated as more or less general principles of motivation. I must proceed to test the statements to ensure that they repre sent truly regular principles. The danger of arriving at too restricted a list of principles need not trouble us; the crucial problem consists in establishing the regular ity and psychological validity of the principles proposed. 2.2.1 The
Principles
The principles proposed below represent an attempt to classify and formalize intuitions about regularly moti vated sign relations gleaned from a wide range of sources. As early as 1748 Hume set forth "three principles of connection among ideas, namely, resemblance , contiguity in time or place, and cause or effect" (Hume 1748: Section III). These three principles and others derived from them have been recognized in various sign systems including natural languages. After consulting properly semiotic treatments of motivated sign relations such as Peirce (1931-1961) and Eco (1972, 1976), I attempted to collate regular principles extracted from works dealing with motivated expression-content connections of specific types.
THE SEMIOTIC BASIS
27
In particular, I turned to works on rhetorical devices such as Kayser (1948), Dubois et al. (1970), Plett (1971, 1975) for regular types of tropes, and compared their findings with a large selection of lyric poetry. I re viewed linguistically oriented discussions of motivation in idiomatic phrases like Chafe (1968), Makkai (1972) for regular types of relations and tested their findings against collections of idioms and proverbs such as Müller (1960), Cowie and Mackin (1975), Koller (1977), Wilson (1970), as well as those of motivation in compounds such as Marchand (1969), Brekle (1970), Shaw (1979) for re current patterns of motivation. Studies of morphological systems like Ljung (1970), Apresjan (1974) provided reg ular types of semantic relations among morphologically linked words. I then turned to treatments of natural language semantic systems such as Stern (1931), Ullmann (1957, 1967), Leisi (1961), Wells (1977), for regular types of (historical) meaning extension and transfer. Re sults of studies of these last two sorts were tested against meaning relations of appropriate types indicated in current dictionaries. The taxonomy proposed is not the only one possible; if reflects my desire to chart a course midway the Scylla of too delicate a differentiation and the Charybdis of too gross a generalization. The principles seem to account for all the 'regular' (we define this term below) cases of motivated relations in the corpus considered, perhaps with some occasional stretching of individual principles at times. Certain principles may overlap to some extent; even the line between principles of similarity and of contiguity is at times fuzzy. But one must attempt to maintain the distinctions, even where difficulties arise. I feel some lack of precision is preferable to an un controlled proliferation of classes. Principles based upon the similarity of sign expression and content are reckoned to the iconic.code, those based upon contiguity of sign expression and content to the indexical code. All these principles will be defined and tested below.
28
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
ICONIC CODE
General
Class:
Model and Object
Principle 1 model-object: model and object differ by a single fea ture (complex) deleted from one of them by a transform ation; e.g. a drawing of a building and the building. General
Class:
Classes, Class Members and Features
Principle 2 specific-generic: class member and its class viewed extensionally; e.g. a dog and all dogs. 3 specific case-general rule: class member and its class viewed intensionally; e.g. a dog and the set of fea tures defining dogs. 4 object-feature : class member and a feature distinguish ing its class x from any class y containing x, or from any class z incompatible with x and also contained in y; e.g. a fish and the feature (Aquatic). 5 class member x-class member y: defective with respect to the criterion of inferential validity; e.g. any fish and any other fish. INDEXICAL CODE General
Class:
Cause and Effect
Principles 1 2 3 4
cause-effect; e.g. fire and heat. producer-artifact; e.g. an author and a book. natural source-natural product; e.g. a hen and an egg. instrument-product; e.g. a carving knife and a wood carving.
THE SEMIOTIC BASIS General
Class:
29
Act and Major Participants
Principles 5 6 7 8
object-act; e.g. a nail and the act of hammering. instrument-act; e.g. a pen and the act of writing. agent-act; e.g. a baker and the act of baking. agent-instrument; e.g. a driver and a car.
General
Class:
Part und Whole
Principles 9 part-whole; e.g. a sail und a ship. 10 act-complex act; e.g. an act of heating and an act of distilling. 11 central factor-institution; e.g. stage and the theater as an institution. General
Class:
Container and Content
Principles 12 container-content; e.g. a beer mug and beer. 13 locality-occupant; e.g. a city and its residents. 14 costume-wearer; e.g. an army uniform and an army soldier. General
Class:
Experience and Convention
Principles 15 experience-convention; e.g. a sensation of cold and a low thermometer reading. 16 manifestation-definition; e.g. a car moving fast and a car commonly considered as capable of high speed. General
Class:
Possessor and Possession
Principles 17 possessor-possession; e.g. a rich person and wealth. 18 office holder-office; e.g. a person serving as president and the office of the president as an institution.
30
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
2.2.2 Testing
the
Principles
I propose the four following, of necessity mutually interdependent criteria for regularity in principles defin ing motivated relations between sign expressions and their contents. (1) Multiple Application. The principles included in the iconic and indexical codes must apply in different sign systems and on different levels within these systems. This criterion guarantees that the principles cannot be artifacts of any particular system or sub-system. A principle which applies on various levels in various cultures runs a good chance of being universal. (2) Productivity. The principles assigned to the codes must be productive. A principle is productive if it enables one person to choose a previously uncoded motivated sign for a content, and enables another to decode the sign and re cognize its motivation. (3) Systematization. As proof of the acceptability of a principle's output, we require that items of it have found their way into the stock of coded elements. In natural language this amounts to lexicalization or integration into morphological or syntactic system. The distinction between, say, animate and inanimate correlates with dis tinctions systematized in the lexicons, morphological and syntactic rules of many languages. In other sign systems we will look for similar correlations between potentially productive principles and systematized relations. (4) Validity of Inference. Conclusions conforming to the rules of valid inference are acceptable in scientific in quiry or courts of law. If a sign could count as admiss ible evidence of its content (or object), it must be re gularly motivated. The principles we propose must then describe inferences which would be judged valid in some thing such as a courtroom situation. Eco (19 76:17) grants that signification amounts to inference, but only when the relevant association is culturally recognized and coded. For Eco, smoke functions as a sign of fire or, equivalently, fire may be inferred on the basis of smoke, only because of a cultural conven tion positing a coded correlation between the two. More clearly, for the first physician to discover the connec tion between measles and red blotches on the victim's skin, an inference was involved; but the connection has since become a matter of recognizing a coded relationship. It may be objected that the criterion of validity of inference is too strong, that many cases of what we would like to call motivation are based upon connections much weaker than valid inference. We readily grant the existence
THE SEMIOTIC BASIS
31
of instances of motivation not corresponding to valid inference and inferential motivation based upon more solid found ations. This seems a perfectly legitimate distinction, i.e. we simply require a relatively high degree of the re levant characteristics for inclusion among the truly re gular principles. Since we aim at proposing principles of a general and hopefully universal nature, validity of in ference represents an appropriate criterion. The four criteria proposed establish the properties we view as necessary for principles of motivation. We define as a regular principle of motivation any statement de scribing an expression-content relation in a sign system and possessing the properties set down by the four cri teria. Note that I am not identifying principles of motivation with specifically linguistic relations or rules at this point. I wish only to point out functions of natural lan guages qua sign systems which conform to general semiotic principles. Our formal proposals regarding the relation between these semiotic principles and semantic rules in natural languages will be presented and argued in the following chapter. 2.3 The Iconic
Code
Eco's (1976) discussion of iconicity gives us several indications of what sorts of principles of iconic motiv ation ought to be included in an iconic code of the sort envisioned here. 2.3.1 Model and
Object
Eco (199 ff.) explains our perception of certain types of similarity and analogy between expressions and their contents as a recognition of relationships established by coded transformations (cf. Jackendoff 1978:210 ff. on images and pictures). These transformations connect (our perceptions of) objects sharing a single set of features but manifested in differing mediums. We thus recognize a small wooden imitation as a model of a large concrete and steel building, if only certain forms and proportions are common to the two. Other transformations allow us to recognize as similar or analogous two objects agreeing in the possession of certain crucial properties, but di verging in their composition otherwise. Such transform-
32
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
ations are at work when we recognize a zebra as similar to a horse. Clearly also, transformations of both sorts might apply to motivate relations between objects differing in both medium and certain properties. Such will be the case when we correlate a painting adhering to artistic conven tions of some kind with a person, object or event inhabit ing some particular universe of experience. Since simi larity and analogy provide a primary basis of iconic moti vation, the same transformations will apply in cases where we judge an expression to be iconically motivated by its content. We might attempt a provisional statement of a first principle of iconic motivation as follows. Iconic Principle 1 (provisional): A sign expression x is iconically motivated by its content y, if x and y are related by a transformation. Of course, (1) remains without significance until we have fleshed out the notion of transformation and determined what sorts of operations they should perform in the iconic code. Since we are ultimately interested in restricting as far as possible the set of regular motivating principles, we will want to constrain the transformations defining re lations between expressions and their contents as severely as the data allow. To this purpose we might begin by con straining transformations to operate on a single feature of either expression or content. If a non-human, primate organism appears in a scientific experiment for the purpose of modeling human behavior, the motivation of the relation depends upon our abstracting away from the 'humanness' of the intended content. For a single match to serve as a motivated sign of a conflagration, we must be prepared to ignore its size. Note that accepting this constraint does not necessarily entail that an expression cannot be corre lated with a content from which it differs by more than a single feature, since successive applications of transform ations connect a series of expression-content pairs. We can go on to restrict the operations of transformations to deletion of features or their transposition into a dif ferent medium, the two functions seen to be necessary so far. In light of the fact that it appears to be irrelevant which medium is chosen for a model as a replacement for that of the original, we might proceed further to limit transformations to describing only deletion processes; thus, in choosing a model one ignores (deletes) the orig inal medium. The choice of deletion is not arbitrary. Deletion turns out to be inherently simpler than other operations with
THE SEMIOTIC BASIS
33
features. Reordering and substitution are more complicated than either deletion or addition, in that both involve two or more features rather than but a single one. Deletion is further simpler than addition in operating only on fea tures present in its input; addition must refer to some separate inventory from which it draws the elements to be added. Accepting the hypothesis that transformations only perform deletion operations and only affect a single fea ture of either the expression or content term, we can for mulate our 'proposal for the first principle of iconic moti vation. Iconic Principle 1 : A sign expression x is iconically motivated by its content y, if x and y differ by a single conflicting feature transformationally deleted from x or y.
Principle 1 describes the sort of relation to be ob served between a model and its object. The interpretation of a model-object relation typically involves abstracting away from certain features of the model or its content which are in conflict with each other or incidental to the relation. Admittedly, most models differ from their orig inals by more than a single feature, but difference in a single feature clearly instances the limiting case for models, unless we claim that any object serves as a model of itself, in which case Principle 1 could be easily altered to include identical feature composition in ex pression and content. This would entail the interesting consequence that identity thereby becomes a special case of similarity and/or analogy. We must now determine whether model-original relations exhibit the properties required by our four criteria. The model-object relation appears to enjoy universal application. People in all cultures create models. They scratch or draw figures, paint portraits, carve or sculpt objects with the intention of representing aspects of what they experience as reality, and these images are reco of the same original in reality (as perceived by the cr lation is evidenced by the fact that it functions between originals and models of all kinds: scientific models of physical, chemical, biological substances and relations; mathematical and logical symbolisms representingquanti etchings, reliefs; metaphor in natural languages (cf. Black 1962; also Peirce 2.277, 2.281 on the relation of icons, models and metaphor; and below).
34
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
The productivity of Principle 1 is amply demonstrated by the endless creativity of scholars and artists in the fields just enumerated. That model-object relations become systematized shows itself in the acceptance and further use of scientific models, mathematical and logical sym bolisms, the establishment of certain mannerisms and styles in the plastic arts, and the existence of specifically poetic diction, hackneyed phrases and dead metaphor in na tural language. The validity of inference from a model to its original seems indisputable on the surface. But clearly many cases of what we have termed models cannot be assigned any realworld object as original. This is as it should be, since we accepted Eco's (1976: 208) definition of motivation as a relation between a sign expression and its (cultural, mental) content, and consequently also cast Principle 1 in terms of expression and content. Our understanding of the term original includes the assumption that real-world ob jects, insofar as they become referents of signs, necess arily become (mentally processed) contents (cf. Eco 1976: 66). While Principle 1 might seemingly be reversible - if one thing x serves as a model of another y, then y should serve as a model of x - this is not the case, on the basis of the criterion of validity of inference. While a model clearly entails the existence of its original, at least in the mental world of its creator, the existence of any given entity in no way implies the existence of a model of it. Hence, Principle 1 should be characterized as irrevers ible. 2.3.2
Classes,
Class
Membership
and
Features
Eco (19 76:193) accounts for much of our recognition of similarity between sign expressions and their contents on the basis of their sharing identical or comparable coded perceptual structures. Narrowly construed, this makes one sort of similarity and a corresponding sort of iconic moti vation a matter of membership in a single (perceptual) class. Clearly, any entity serves as a sign of the class of which it is a member. Since a class may be understood either extensionally as the totality of its members or intensionally as the totality of its defining features, an entity shares identical perceptual structures both with all the members of its class and with the composite feature specification of its class. The perceptual structures by which we recognize a class are the same as those by which we recognize a particular entity as a member of that class
THE SEMIOTIC BASIS
35
or, indeed, by which we group entities into the class at all. We must then accept three fundamental similarity re lations based upon the recognition of classes and class membership, and ipso facto three types of iconic motivation, viz. (1) from an entity to its class taken as the collec tion of entities comprising it; (2) from an entity to its class taken as the set of features defining it; (3) from an entity to any other individual entity in the same class. 2.3.2.1
Specific-Generic
Formulating the first relation noted as a principle of iconic motivation, we get: Iconic Principle 2 : A sign expression x is iconically motivated by its content y, if x is a member of a class z and y is the set of members of class z. Principle 2 is of such a general and pervasive nature that we scarcely need argue its validity. It expresses essen tially the specific-generic relation. Any specific in stantiation of a class calls forth the whole class, and consequently serves as a motivated sign of it. The specific-generic relation is amply illustrated in natural languages; it apparently plays a role in all languages. The alternative In a language system would involve a double listing of, at least, all denotative terms for both specific and generic readings or, just conceivably, distinct lexical items with specific and generic meanings respectively for each denotation. In other sign systems, the specific-generic relation holds as well. As Eco (1976:255) points out, my showing a cigarette package to someone I know is going shopping may function as a sign that I wish him to get cigarettes of any kind or more of the particular type I display. In either case I use a specific item as a sign of its generic class. The criterion of multiple application is fulfilled in that the specific-generic relation holds on various levels and in different systems. A single violin may stand for the class of violins, that of bowed instruments, that of string instruments or even of instruments generally. Similarly a beaver may signify beavers, rodents, mammals, animals. The productivity of the specific-generic relation can be observed any time an object serves as a sign of its class. Any time someone points to a single flower, community, law etc. intending the class of flowers, communities, laws,
36
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
and he is understood, Principle 2 is productively employed. This principle shows a clear tendency to become system atized in sign systems of all types: a number stands not only for a quantity as such but for all sets with the designated number of elements; a musical note stands not only for a particular sound (to be) produced by a particu lar instrument at a particular point in some work but for that area of sound in the musical system as such. Examples of this type can easily be multiplied. The validity of inference from a member to its class, from a specific entity to all other entities sharing its generic type relies perhaps too much on definition or intuition to be amenable to proof as such. Clearly, in a court of law or anywhere else, the existence of a given entity provides the necessary evidence for the existence of the class containing it as a member (including the limiting case where the class contains only this single member), no matter whether we understand the class extensionally as the set of its members or intensionally as its defining characteristics. This latter case argues the validity of the following principle as well, to which we turn shortly. Before ending our discussion of Principle 2, however, we should note that, on the basis of our validity of inference criterion, it, like Principle 1, does not bear reversal. While any particular member of a class necess arily requires the existence of the class, the fact of a certain class's existence need not entail the existence of any particular member, especially when we allow for the existence of empty classes. Pertinent counter-examples occur any time a single element requires postulation of a unit class. Hence, Principle 2 must be labeled irreversible along with our first principle. 2.3.2.2 Specific
Case-General
Rule
Principle 3, based upon the same member-class found ation as the foregoing principle, enjoys essentially the same support it does. It differs from Principle 2 in view ing the class intensionally as the set of properties common to the members of the class. Iconic Principle 3: A sign expression x is iconically motivated by its content y, if x is a member of a class z and is the set of features defining z.
THE SEMIOTIC BASIS
37
We could, of course, state this principle in the following manner as well. Iconic Principle 3 (alternative): A sign expression x is iconically motivated by its content y, if x bears the features f 1 , f 2 . . f n and y is the set of features f 1 ,f 2 ...f n A set as a collection of features is an abstract entity which must be instantiated in some substance, viz. one of the set elements it defines. For this reason, Principle 3, like the two preceding principles, is irreversible. It ex presses the sort of relation we recognize between a speci fic instance of a thing or rule and the definition of the thing or the rule itself. Principle 3 holds between laws and their concrete instantiations generally, e.g. the laws of motion, gravity, genetics etc., between definitions and their definiens, e.g. definitions of chemical elements, phylogenetic species, word classes etc. Hence, Principle 3 defines universally accepted relations. The examples cited also point up the functioning of this principle in various sign systems and on different levels. They ensure adherence to the criterion of multiple application. The productivity of Principle 3 manifests itself in our recognition of concrete individual statements as ex pressions of general truths, specific entities or phenomena as instance of the application of general laws and defini tions. Principle 3 is systematized in the lexicons of na tural languages in the form of the assignment of particular words to semantic fields beneath the appropriate superordinate terms, the assignment of physical entities to taxonomic classes, organisms to species etc. Principle 3 satifies the criterion of validity of inference, since, as described above, a member of a class, by convention or definition, necessarily presupposes the existence of its containing class. 2.3.2.3
Object-Feature
We can derive as a sub-case of Principle 3 a further re lation holding between a class member and a single feature or feature complex characteristic of the relevant class. A natural constraint on this principle would be its restric tion to a feature (complex) necessary to distinguish the class in question from any class containing it or from any incompatible class contained in any of its super-ordinate classes. We thus state:
38
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY Iconic Principle 4: A sign expression x is iconically motivated by its content y, if x is a member of a class z contained in a class w, and y is the feature of z differenti ating it from w or any other incompatible class contained in w.
This principle expresses the relation we recognize between a specific member of a class and one of its prominent fea tures. When a porcupine calls forth thoughts of spines, or an office building (as opposed to a family domicile) in cites visions of business (as opposed to family life), we have examples of the effect of Principle 4. Any time a word such as fox is predicated of some subject with the intention of assigning the feature (Sly) to it (ignoring for the moment the problem of dead metaphor) or mountain with that of assigning the feature (Large), Principle 4 must apply to the expression term of the predication. The fact that slyness represents a characteristic attached to foxes on the basis of cultural norms, while largeness must appear as part of any definition of mountains, need not trouble us here, since, as Eco (1976, 66) points out, not real-world objects are decisive in the determination of motivated sign relations, but rather culturally coded perceptual models. In non-linguistic sign systems, Prin ciple 4 accounts for the interpretation of a racing car as an expression of the content '(capable of) high speed', or a glass of wine as an expression of 'alcoholic bever age'. The examples cited argue the multiple applicability of the principle. The productivity of Principle 4 ex hibits itself whenever anyone associates a property with an entity by predicating of it a term possessing that property as a characteristic quality. Whenever someone points to a red object with the intention of indicating that the person speaking is a leftist or sympathizes with certain leftist views, whenever someone uses Einstein as a sign of mathematical genius or Bach as a sign of musical genius, Principle 4 rears its head. The evidence that its effects have become systematized exhibits itself in the series of linguistic and nonlinguistic metaphors so far described. They constantly show themselves in the human tendency to concretize gen eral predicates such as 'rich', 'silly', 'heavy smoker' in particular individuals, animals or objects like Midas, goose, chimney. The validity of inference from an expression to a/the characteristic feature of its class is evidenced by the fact that any entity commonly serves as a sign of the characteristic features differentiating it from other
THE SEMIOTIC BASIS
39
entities. The existence of a porcupine guarantees the ex istence of the class of spiny animals and of spininess as a property; the existence of wine guarantees that of the class of alcoholic beverages and of the property of con taining alcohol. Like the other principles accepted so far, Principles 4 is irreversible. We cannot infer the ex istence of porcupines from that of spiny things, nor the existence of wine from that of alcoholic beverages. 2.3.2.4 Class
Member
-Class
Member
y
The third major principle we can derive on the found ation of Eco's definition of analogy is based upon the re lation of any two entities representing members of a single class. We formulate this relation as our fifth principle of iconic motivation. Iconic Principle 5: A sign expression x is iconically motivated by its content y, if x and y are both members of a single class. Any given entity can serve not only as a sign of its class, but also as a sign of any other particular entity within the class. Principle 5 plays a crucial role in sign systems of all types. To use Pike's (1967) terminol ogy, two etically distinct signs assigned to the same emic class must be recognized as emically non-distinct for the sign system to function. Dictinct and, in accurate physical terms, different spheroids count as the official basket-ball stances of a particular ceremony count as official mar riages according to particular laws and conventions. In each case, the etic form serves as a sign not only of its emic class, but of any other entity in the same class. By virtue of their mutual definition as belonging to a single class, co-members should always be able to function as motivated signs of one another. The examples cited illustrate the systematization of Principle 5. Its multiplicity of application exhibits particular stop sign calls up the image of any other stop sign,one maple tree that of any other. The productivity of Principle 5 is amply demonstrated any time someone uses one instance of an entity to refer to any other instance. A guest is disturbed by a noise from an adjacent room and I extract the cork from a wine bottle on the table in
itsel
40
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
front of him producing a popping sound to identify for him the disturbing noise: one member of the class of acts of uncorking a wine bottle serves as a sign of another member of this class of acts. The criterion of validity of inference presents prob lems in the case of Principle 5. As we mentioned above, the existence of a particular member of a class does not necessarily entail that of any other member. The existence of President Reagen does not entail the existence of any other American president, and the existence of President Washington surely did not. We readily grant that any class member does in fact provide a sign of another, but their relation is established by means of their common class membership, or, as Peirce would have it, by a law. Conse quently, their relation is not inferentially motivated in the strict sense laid down as a criterion here. This does not mean that Principle 5 plays no role in sign systems; on the contrary, as we have seen, it enjoys both productivity and systematization. It finds expres sion in the lexicon of natural languages in the classific ation of words denoting entities inhabiting the same class into the same semantic field, beneath the same superordinate term and, in cases of true synonymy, under the same set of distinctive features. Note that Principle 5, in contrast to the principles discussed so far, can be re versed. Clearly, if a member x of a class z serves as a sign of a member y, then y stands just as surely for x. Thus, Principle 5 is defective in two respects; we include it in our iconic code but characterize it as defective in these respects. Under the strict interpretation of the criteria pro posed then, we recognize four central principles of iconic motivation and a fifth partially defective one. We could, of course, derive further principles on the basis of those derived so far, just as we might develop others directly from similarity and/or analogy relations. Nevertheless, the principles proposed above appear to cover the standard cases of truly regular iconic motivation. Further, they suffice for the purposes of the semantic analysis to fol low, so that we have no reason to expand upon them. We now proceed to a discussion of the indexical code. 2.4 The Indexical
Code
As Peirce pointed out, indices derive their motivation from contiguity with their objects. This contiguity may be objectively present in the observable world or inherent in the process of perception and interpretation (Peirce 2.3o5).
THE SEMIOTIC BASIS
41
The ultimate source of association by contiguity within perception and interpretation lies in inference. We build up a stock of associations on the basis of experience. The set of regularly recurrent principles underlying these associations should amount to an indexical code. We must, of course, test these principles against the criteria dis cussed above, especially that of validity of inference. The principles in our indexical code will then mirror the set of general rules for valid inference, derived as they are from regularly successful patterns of association. The same principles establish relations of metonymy in Jakobson's sense (1953, 1954). Traditional accounts of rules for valid inference, of metonymy in literary texts, treatments of metonymical motivation in morphology and lexical relations (Apresjan 1974; Norrick 1979a and b) in dicate the all-pervasive influence of contiguity relations in sign manipulating behavior. Our present task thus en tails discovering which of the principles defining such re lations have a claim of regularity, as determined by the four criteria set forth above. Those principles satisfying the four criteria will be included in the indexical code. 2.4.1 Cause and
Effect
Probably the most important principle of indexical mo tivation is that relating cause and effect. Causality stands as the central principle of metonymy for Schofer and Rice (1977), who analyse synechdoche (part-whole and associated relations) as a category separate from metonymy. We regularly infer both causes from effects and effects from causes. Dark clouds just as frequently serve as a sign of rain as do wet streets in everyday situations. While our validity of inference criterion may force us to reject one or both of these particular inferences, we can say that, in general, principles of indexical motivation will be re versible. This property contrasts them with principles of iconic motivation, and may be considered a distinguishing characteristic for principles in the two different codes. Given the contiguity and fuzziness of semiotic cate gories generally, and the breadth and inclusive nature of the cause-effect relation in particular, it is difficult to decide where to draw the boundary between cause-effect and certain other relations. Some scholars (e.g. Schofer and Rice 19 77) have advocated the comprehension of such re lations as that from producer to product, instrument to product, agent to action into a generalized category of cause-effect relations. On a level of abstraction corre sponding to the psychological tendency to amalgamate simi-
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
42
lar processes, this comprehension has some validity, and I am in general agreement with this position. Clearly, one who produces an object can be said to have caused that ob ject to exist; slightly more abstractly, a tool or method crucially involved in the production of the object might also be reckoned a cause, albeit indirect, of the object's existence. Nevertheless, when confronted with the concrete task of accounting for the motivation inherent in using a shoe to signify a cobbler, one is hardly likely to cite the indexical principle relating cause and effect. In descriptions of the semantics of natural languages, we will require rules which apply to particular classes of lexical items or all items charaterized by a certain fea ture (complex). The formulation of such rules presupposes a relatively delicate differentiation of motivating prin ciples. For instance, the rule describing the semantic connection from a noun denoting an individual pro fessionally engaged in producing some object to that denoting the object itself must be restricted in its application to items containing the feature (Human) and, redundantly, (Animate). The indexical principle relating cause and effect need not, indeed it must not be restricted in this way. In light.of such examples it appears expedient to treat indexical principles like that describing the transfer from product to producer as separate from, albeit ultimately related to the general indexical principle of cause-effect. 2.4.1.1
Cause-Effect
I take up the discussion of these principles below, but we require a general principle relating causes and effects proper in any case. We can set down as our first principle of indexical motivation: Indexical Principle 1 : A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a cause and its effect. (reversible) Acceptance of Eco's (1976) definition of the sign rela tion as one involving an expression and its content alone without considering referential objects as such dictates the formulation "are conceptualized as" rather than simply "are."2 while the former is more cumbersome, I stick by it in the interest of simplifying the analysis as a whole. I label this first indexical principle reversible in light of the discussion above. In the intended sense, the reverse
THE SEMIOTIC BASIS
43
of the principle can be stated as: Indexical Principle 1 (reverse): A sign expression x is indexically motivated by its content y, if x and y are conceptualized as an effect and its cause. Principle 1 defines such relations as those holding be tween the concepts of fire and heat, just as its reverse defines that between those of heat and fire. Further ex amples of the application of Principle 1 (and its reverse; I eschew explicit mention of the direction, forward or reverse, as well as that of "the concepts" in individual examples, although I retain them in the statements of the principles proper) can be observed in the motivated re lations between the states of having p, knowing q and the actions of obtaining p, learning q. Principle 1 describes the relation between the literal meaning of the poetic figure dispatch to Hades and its figurative interpretation 'kill', insofar as killing a person is conceptualized as sending him to Hades, viz. in the classical Greek Weltan schauung. In the same way, the meaning relation exhibited by the morphologically related pair pleasure-please can be captured by Principle 1, insofar as please x is defined as 'cause x pleasure', viz. in the semantic system of Eng lish. The validity of the principle of cause-effect need scarcely be demonstrated. It defines the relation from red blotches on the skin of a patient to the conclusion that he has measles. It provides for the postulation of gravity on the basis of observations of falling objects. The multiple application of Principle 1 shows itself in its functioning within scientific theories, everyday induc tions, and in natural languages on various levels. We argue from observations of falling objects to the ex istence of gravity, from the law of gravity to the attrac tion between bodies generally, from this to the laws of planetary motion, and even to the hypothesis that light possesses mass. In language the principle of cause-effect establishes conceptual connections between the lexical meanings of morphologically related words (e.g. blackenblack, please- p l e a s u r e , warm 'temperature'-warm 'to a wearer'), between contextually conditioned interpretations for single (polysemous) words (e.g. l i g h t 1 'visible energy-light2 'source of light; lamp', s u r p r i s e 1 'mental state ' -surprise 2 'act or object inducing surprise1']' ) , and between metaphorical and literal uses of expressions (e.g. make tracks meaning 'leave', dispatch to Hades meaning 'kill').
44
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
The productivity of Principle 1 is illustrated by the creativity of scientists in abstracting general laws from series of individual observations, by that of writers and readers in offering and interpreting figures based upon effect to cause inference. The validity of inference from effects to causes is in general unproblematical. In a court of law, bruises will be taken as evidence that their bearer has been physically assaulted (if only by himself), a person's possession of an object as evidence that he obtained it, a person's experience of pleasure as evidence that he has been pleased, and smoke as evi dence of fire. In scientific experimentation, a change of state counts as evidence of a force exerted, a precipitate as evidence of a chemical reaction, and tracks in a cloud chamber as evidence of atoms. Of course, smoke or similar phenomena may be produced by means other than fire. But to the extent that these means are intended to call up the impression of fire, they rely crucially on Principle 1 and as such argue its general validity. Since I have chosen to describe motivation as a re lation from an expression to its content, rather than its object, I am committed to accepting as valid those in ferences accurately reflecting standard modes of cognition, even if these do not necessarily hold in some particular case. If someone testifies that there was a fire in the fireplace of a particular room, because the room was full of smoke, and he heard crackling noises and felt warmth coming from the fireplace, his testimony will be accepted by any reasonable jury, unless some other witness can bring forth evidence based upon more than his own sensory experience that, say, the fireplace was fitted with an electric heater so constructed as to crackle at intervals and produced smoke. In the same way, physically assault ing someone should count as evidence that he will bear bruises as a result, just as we predict that an unsup ported object with a mass exceeding that of air will fall. The failure of our principles to make correct predictions in particular cases does not affect their validity qua principles. So long as we bear in mind that the principles are intended to reflect psychological rather than physical reality, there exists no reason whatsoever to amend Indexical Principle 1 in such a manner as to prevent its assigning a motivated status to relations such as that
bet
THE SEMIOTIC BASIS 2.4.1.2
45
Producer-Product
A second principle of indexical motivation connects artifacts as sign expressions with their producers as sign contents. As noted above, this principle could, in a freehanded derived from the principle relating cause and effect. But I prefer to define it as a separate principle, to some ex tent with an eye to the necessity of recognizing a dis tinction between the principle of producer-product and that of cause-effect in their reinterpretation in the sem antic theory to be developed below on the basis of the present semiotic theory. Indexical Principle 2: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a product and its producer, (reversible) This principle defines the relation between any arti fact and the person who produced it, and vice versa, or the producer-product relation. This principle predicts that we will regularly accept a play as the sign of the playwright, a painting as that of the painter, wine as that of a wine-grower, bread as that of a baker. Similarly, the playwright serves as a sign of his play, the painter as that of his painting etc. These examples show firstly the reversibility of Prin ciple 2. We witness the multiple application of the pro duct-producer relation in the foregoing examples, but also in any case where a particular work serves as a sign of its creator. Principle 2 plays a role any time anyone says that Donatello is a great sculptor, Bernini a great archi tect, or Leonardo a great painter with reference to their respective works. Its productivity shows itself in these ex amples as well. When the butcher, baker and tailor refer to each other as meat, bread and clothes respectively, Principle 2 establishes the necessary relations between ex pression and content. The systematization of Principle 2 is amply evidenced in the morphology of languages where terms for products clearly signal terms for their pro ducers. German contains among myriad others such pairs as: Sattel-Sattler, Schuh-Schuster, Gedicht- Dichter, KarteKartograph; compare the English: clothes-clothier, hathatter , wheel-wheelwright, drama-dramatist, theory-theore tician . The inference from a product to its producer apparently counts as valid. Juries are commonly expected to infer the presence or existence of a brewery from beer, a painter
46
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
from his painting, a particular tailor from a garment bearing his label or characteristic style. We can thus accept Principle 2 as a fully motivated part of the indexical code. 2.4.1.3 Natural
Source-Natural
Product
Just as an artifact counts as a sign of its producer, a natural product serves as a sign of its natural source. I will interpret natural product to include the fruit of plants, progeny of animals, by-products of the natural functioning and growing of living beings and natural ob jects and processes generally. Indexical Principle 3: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a natural product and its natural source, (reversible) This indexical principle establishes such standard con nections as those between an acorn and an oak, a child and (either of) its parents, a butterfly and its cocoon, a paw print and some animal, holes in wood and termites, heat and oxidation etc. Conversely, oaks inferentially argue the presence of acorns, cocoons that of butterflies etc. The universal character of Principle 3 is attested to by scientific and pre-scientific statements about regu larities in the physical/biological world. Even in a cul ture where holes in wood are assigned to the workings of of an evil spirit instead of termites, it remains some thing such as Principle 3 which establishes the connection between the two. The functioning of Principle 3 in these distinct systems of thought argues its multiple applica bility; also, it applies on different levels within sys tems: we infer the presence of steam from the whistling of the tea kettle, the boiling of the water from the pres ence of steam, the presence of heat in the kettle from the boiling of the water, a process of oxidation (or electrical resistance) from the presence of heat etc. The productivity of Principle 3 is evidenced any time we infer natural sources for natural products or processes, unless their connection is already coded. The discovery of a particular microbe as the source of some disease can only take place after the investigator has gone through the thought process reflected by Principle 3 to interpret the disease as the sign of some microbe; our understanding of the figure adult cocoon to mean 'butterfly' similarly pre supposes the mental process Principle 3 is intended to ex press.
THE SEMIOTIC BASIS
47
The systematization of this principle manifests itself in the acceptance and further use of connections of the types cited above: termites and holes in wood, oxidation and heat, butterflies and cocoons etc. It may be further seen on the level of lexis in natural languages in the ambiguity of words which designate both a natural product and its natural source: flower (bloom, or plant), tomato (fruit or plant), chicken (meat or animal) etc. Examples of these types demonstrate the validity of inference com monly associated with Principle 3. Certainly, tomatoes will be accepted as evidence of tomato plants, beef as evidence of steers, red blotches as evidence of measles. We thus number Principle 3 among those in the indexical códe. 2.4.1.4
Instrument-Prduct
If an artifact serves as a motivated sign of its pro ducer, as defined in Principle 2, then it should also serve as a sign of an instrument or process crucially in volved in its production. Bread signifies not only the baker but the oven, a painting not only the painter but the paint. By Principle 2 wine signifies the wine-grower; by Principle 3 it signifies the grapes and, transitively, the grape vines; by the principle to be here proposed it signifies the wine press, the wine cask or even the process of fermentation. We express the relevant principle as: Indexical Principle 4: A sign expression x is indexically motivated by its content y, if x is conceptualized as a product and as an instrument or process crucially involved in its production.(reversible) The reversibility of this principle is clear from the possibility of reversing any of the relations mentioned above. Principle 4 has essentially the same claim to significance as the preceding indexical principles. Indeed, in our present mechanized age, artifacts seem to signify the instruments and/or processes involved in their produc tion more immediately than their human producers. Thus, cloth suggests a loom rather than a weaver; a photograph suggests a camera rather than a photographer; a synthetic material suggests a chemical process rather than the chemist who developed it. Principle 4 enjoys multiple applicability: it applies to relations between objets d'art and artists' materials or processes; between indus trial products and industrial tools, machines, processes;
48
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
even between a scientific theory and the instruments, axioms and procedures essential to its development (e.g. the current model of our solar system and the telescope or mathematical formulas). Principle 4 claims productivity on the basis of our ability to assign instruments to artifacts generally. A piece rounded with respect to an axis, even if one is unfamiliar with it, will signify a lathe; a musical note with certain characteristics will be recognized as issuing from a woodwind, another from a brass instrument and so on. Principle 4 receives systematization any time an arti fact is named after the instrument or process producing it, for instance when we distinguish between finger, brush and roller painting; between cubist, abstract and pointillist painting; or between atomic, coal and oil turbine energy. The validity of inference from an artifact to the in strument or process crucial in its production becomes in creasingly clear as the range of instruments and processes available for its production shrinks. Since a camera is the only instrument available to make photographs, a photograph clearly allows the inference to the camera. Since a basket requires weaving, it stands as a potential sign of this process. 2.4.2 Acts
and Major
Participants
Altering our focus from cause-effect to acts and their major participants, we can define principles describing re lations from actions to their characteristic objects, in struments and agents. The emphasis here lies on acts not on artifacts. Principle 2 related a product with its pro ducer; the next principle to be introduced relates an ob ject with an act characteristically performed on it. The latter relation may seem more abstract than the former, but it is no less a perceived contiguity relation. 2.4.2.1
Object-Act
The object-act principle will describe such relations as those between a nail and an act of hammering, between a book and an act of writing. I formulate this principle as follows:
THE SEMIOTIC BASIS
49
Indexical Principle 5: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as an act and its object, (reversible) The reversibility of this principle receives ample justification in the examples cited above: the nail sig nifies the act of nailing or hammering; the drink signifies applicability of this principle is seen in its functioning in everyday connections such as that between driving and automobiles, eating and food; in artistic connections such as those between acts of painting and sculpting, and the painting or sculpture; in scientific connections such as those between acts of weighing and weight, acts of classification and taxonomies. The examples from the domain of artistic expression illustrate the productivity of Principle 5, since acts of painting and sculpting can involve a number of different objects as their medium. By the same token, Principle 5 proves productive in our recognition of people's opinions as objects of acts of polling, of speech sounds as the objects of acts of classification on the basis of sound spectrographs. The relations described by Principle 5 are system atized in the conventional acceptance of such connec tions as those just mentioned. Once the sound spectro graph's output becomes directly associated in the minds of its users with speech sounds, once polls are associated directly with public opinion, these relations are coded and systematized. The systematization of Principle 5 is evident in the morphologies of natural languages in which the names of acts are related in form to the names of their objects (e.g. n a i l V - n a i l N k n i t - k n i t t i n g , integrateintegral,,) because in each case the lexical reading of the former term identifies an act conceptualized as involving the object identified by the lexical reading of the latter. The validity of inference from an act to its object or conversely rests upon their close connection within our experience. Certainly, in a court of law or scientific study, the object of an act will serve as evidence of the act itself, and vice versa. An act of painting or sculpt ing counts as evidence of a painting or sculpture, and conversely.
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
50 2.4.2.2
Instrument-Act
Just as actions stand for the class of their objects, they serve as signs of their instruments. I differentiate the action-instrument relation from that of product-in strument described in Principle 4. The product-instrument relation holds between a wood carving and a carving knife, whereas the action-instrument relation holds between the act of carving and the carving knife. Like the other re lations described by principles in the indexical code so far, this principle too is reversible, since the carving knife as surely suggests acts of carving as acts of carv ing do the carving knife, even if some carving knives are used as letter openers and some acts of wood carving are accomplished with chisels or table knives. Indexical Principle 6: A sign expression x is indexically motivated by its content y, if x is conceptualized as an act and y as an instrument crucially involved in x. (reversible) I interpret instrument broadly to include materials as well. The validity of Principle appears to be beyond dispute. An act of piano playing, say the performance of a piano concerto, clearly entails the existence of a piano; the existence of chairs presupposes acts of sitting. Along with the examples cited so far, the diversity of the following relations within the compass of Principle 6 illustrates its multiple applicability; we find such re lations as those between ropes and tying, pens and writ ing, cups and drinking, languages and communication, sets of axioms and theorizing. The productivity of Principle 6 manifests itself any time someone uses a rope as a sign of tying, a pen as a sign of writing etc. The systematization of Principle 6 is in evidence in such standard tableaux as the pen and scroll standing for scholarship, the rope and saddle standing for rodeos, pots and pans standing for the culinary arts etc. It receives further exemplification in the lexicons and morphologies of natural languages where a single sound sequence serves as both a noun naming a particular in strument and the verb naming the act in which the instru ment finds crucial employment, e.g. saw, hammer, cover, filter, salt, soap, bridge, cart etc. In addition, a number of morphological rules form nouns naming instruments from verbs naming the corresponding acts, and vice versa, e.g. house N - houseV, seat-sit, opener- open, cleansercleanse, distillery-distill, oil can-oil etc.
THE SEMIOTIC BASIS
51
The validity of inference associated with relations defined by Principle 6 should be evident from the examples offered above. If an archeologist discovers remnants of a device for shucking corn among the implements possessed in an ancient culture, he will confidently assign acts of shucking corn to members of that culture; if he discovers that the culture attached great importance to acts of distilling beverages, he will similarly attribute to the culture the possession of distilling apparatus, even if none is discovered at the relevant sites of investigation. 2.4.2.3
Agent-Act
Related to both the foregoing principles, a third one defines relations between acts and their agents. I want to distinguish the agent-act relation clearly from that of artifact-producer defined by Principle 2. The latter holds between bread and the baker; the agent-act relation at issue here will be defined to hold between the act of bak ing and the baker. Both are necessary, and they cannot be collapsed into a single relation. As the seventh principle of indexical motivation I submit: Indexical Principle 7: A sign expression x is indexicallv motivated by its content y, if x is conceptualized as an act and y as the agent responsible for the act x. (reversible) Examples of the effects of Principle 7 proliferate. In addition to the relation between baker and baking, we find such obvious relations as those between boxers and boxing, tax collectors and tax collection, scientists and science etc. The inferential validity of this principle can hardly be questioned. It derives from the immediacy of the conti guity of the agent and his act. Since an individual be comes an agent only by acting and an act, in our sense, re quires an agent, the inferential validity of Principle 7 is essentially by definition. Further, it enjoys multiple applicability. In addition to the examples already cited, Principle 7 would predict a number of lexical-morphological relations found in lan guages. In any language one finds pairs of words naming agents and their acts built upon the same stem, e.g. baker-baking, ness, speaker-speech etc. These same examples help illustrate the systematization of Principle 7, but it is clearly systematized in non-lin guistic sign systems as well. In order to create univer-
52
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
sally decodable signs for the various disciplines repre sented in the Olympic Games, the organizers chose schema tized drawings of a swimmer for swimming, wrestlers for wrestling, a horse and rider for riding etc. Any time anyone wants to mark specific areas or rooms for the activities they are intended for in a cross-culturally un derstandable manner, the obvious choice of signs involves picturing people engaged in acts to stand for those acts. Principle 7 also plays a role in our interpretation of figures such as the baker needs flour to mean acts of baking require flour. Our ability to create and interpret figures like this last one testify to the productivity of Principle 7. Since it fulfills the criteria we set down above, Principle 7 will be included in our indexical code. 2.4.2.4
Agent-Instrument
I have been discussing principles relating acts with the major participants in them. Before that I described relations holding between the major participants in acts of creation or production. But I have not yet treated the relation between an agent or producer and the instrument crucially employed in a particular act or process (of crea tion or production). It seems these may be collapsed into a single principle relating agents and instruments gener ally. Indexical Principle 8: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as the agent of an act z and an instrument crucially employed in z. (reversible) Since this principle defines relations like those holding between pipe wrenches and plumbers, cars and drivers, boxing gloves and boxers, ships and sailors, tobacco and smokers etc., all of which are reversible, the principle itself deserves to be labeled as such. These same examples strongly urge the validity of Principle 8. As in the case of Principle 7, its inferential validity derives from the immediacy of the contiguous relationship between agents and instruments. We infer with complete confidence the violinist from the violin, the potter from the potter's wheel, the fencer from the foil, and conversely. The multiple applicability of Principle 8 seems clear from these examples as well. Its applicability on different levels within a single system may be seen in the fact that,
53
THE SEMIOTIC BASIS
for instance, a steering wheel, a driver's seat, or a whole automobile stands for the driver. Principle 8 would predict the possibility of figures of speech involving the name of an instrument to designate its user, e.g. the celli sit across from the violins , and the interpretation of idioms cast in terms of instruments, e.g. the pen is mightier than the sword meaning 'the writer is mightier than the soldier'. The productivity of Principle 8 allows the endless creation and understanding of figures like those cited and the establishment of further foil-fencer type connections. Principle 8 therewith passes our tests and should be in cluded in the indexical code. This completes my consider ation of relations involving acts, agents, instruments, ob jects. 2.4.3 Part
and
Whole
The third major category of indexical principles re volves around the contiguity relation of part-whole. We regularly infer the whole from the part, and conversely, on the basis of our experience of them as contiguous. The rhetorical trope of synechdoche, traditionally illustrated by such figures as sail to mean 'ship' and lade to mean 'sword', instances the part-whole {pars pro toto) relation. Many stock idiomatic phrases, especially those involving euphemism, literally designate only a part of an action, thus leaving the construal of the whole to the relevant indexical principle and the hearer's imagination; witness such phrases as go to one .
the
b athroom
and go to bed
with
some
The indexical principle of part-whole motivates re lations of two different types at the word level. First, a word like jet may designate an entire airplane powered by a jet engine or the engine alone; here a single ex pression serves for two distinct but contiguous objects. Second, letters or sound sequences of a word or phrase may be taken as an abbreviation, acronym or clipped form stand ing for the word or phrase as a whole, thus NATO /neytow/ stands for 'North Atlantic Treaty Organization', U.N. for 'United Nations', and flu for 'influenza'. The first pro cess leads to polysemy, while the latter creates synonyms for existent expressions. In non-verbal sign systems, the part-whole principle also plays a pervasive role. Central to traditional paint ing and sculpture stood the idea that an entire event could be evoked by portraying a single representative moment of the whole. Nor is even the total reality of the single
54
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
moment portrayed, but only pertinent aspects of it, and from some particular perspective, which necessarily con centrates the attention on a single part of the whole scene. One might view truly abstract painting of the sort advocated and practiced by Kandinsky as an attempt to re duce the real whole to its essential part(s) (cf. Kandinsky 1963). The part-whole principle thus clearly be longs in the indexical code along with the others proposed so far. The examples cited illustrate its universality, multiplicity of application, productivity, systematization (in language and the pictoral arts) and validity of in ference, so that we need not argue them point by point. 2.4.3.1 Part-Whole I express the basic part-whole principle as: Indexical Principle 9 : A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a part and its whole, (reversible) One might view the principle relating a particular class member with its class (see Iconic Principle 2) as associated with or dependent upon the part-whole relation, as Levin (19 77) does. But claiming that the relation of the core to the apple parallels that of a particular Grime's Golden to the class of apples represents a category mistake. The part-whole relation is founded upon the objective con tiguity of parts with their wholes, whereas the member-class relation rests upon the definitions of class and member as well as conventions of classification based upon various taxonomies. Generalizing the notion of apple, one arrives at fruit; generalizing the notion of apple core, one arrives at fruit cores and pits, not at whole apples. Hence, Principle 9 defines a relation of a type clearly different from that of Iconic Principle 2, and it will be included in the indexical code. 2.4.3.2 Act- Complex
Act
Parallel or perhaps subordinate to Principle 9, one can recognize two further principles defining relations in the part-whole domain. The first relates a complex act or event with a particular act or event crucial to its character or success. A real act of cooking almost never consists entirely of cooking in its strictest sense of 'heating in
THE SEMIOTIC BASIS
55
water'. Instead, it can typically be composed of acts of cleaning, slicing, activating a source of heat, filling a container with water, adding the prepared food, waiting etc. Nevertheless, since the process of cooking in the strict sense is the sine qua non of the whole series of acts, it serves as the sign of them all. This receives recognition and systematization in our use of the verb cook in both the narrow and broad senses. Of course, once the broad sense of the verb cook was systematized, the word experienced a further removal from its strict meaning, so that we commonly use cook to mean 'prepare food' gener ally, even when the actual process of heating in water is not involved. I can say I am cooking even when I am pre paring a roast and potatoes in the oven and tossing a salad. Similar examples are easy to come by. On a narrow interpretation eating involves only chewing and swallow ing, but this makes out the crucial element in the complex of serving, passing dishes, distributing portions, seasoning/ cutting etc. Consequently, we view the partial and total actions as mutual signs of each other. This fact is systematized in all languages where a single word such as e a t , e s s e n , manger covers both. This illustrates the systematization of the relevant principle, which I formulate as: Indexical Principle 10: A sign expression x is indexically motivated by its content y, if x is conceptualized as a complex act or process and y as a crucial act or process in x. (reversible) This principle establishes the relation between the concrete act of sewing as the crucial one and producing a garment as a whole process; it underlies our use of the phrase sewing a garment to mean 'the whole process of measuring, cutting, sewing etc. a garment'. The examples cited so far support the multiple applicability of Prin ciple 10, as well as arguing its systematization in na tural languages. In the non-linguistic domain, Principle 10 is equally pervasive. Children conceptualize turning a wheel at arm's length to be the crucial act in driving; consequently they require only a miniature steering wheel to signify the entire act of driving. This relation receives systematiza tion at the hands of the toy industry, which produces children's car seats with steering wheels attached to them; apparently no need is felt to add a shift lever, turn signals, foot pedals etc. We often use a similar gesture, even when no wheel is present, to indicate the
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
56
act of driving. In the same way, swinging a golf club or simply imitating this action without a club will stand for the whole complex of actions involved in playing golf, swinging a baseball bat stands for the whole game of illustrates the productivity of Principle 10. It is further illustrated any time a new process or event receives its name from that of a partial process or event crucial to it. A person who describes placing an order for heating oil as heating the house does so through productive use of Principle 10. Since one's job may simply be viewed as a sub-process in the entire complex of actions leading to providing food, we hear such figurative phrases as bring home
the
b acon
and put
meat
on the
table
in reference to
working. The validity of inference associated with Prin ciple 10 can be judged from the obvious acceptability of claiming one was cooking even though one was still at the slicing stage, or claiming one was eating even though still in the process of carving the roast. We can thus count Principle 10 among those in the indexical code.
2.4.3.3 Central
Factor-Institution
The third principle defining a relation associated with the basic part-whole connection relates a particular product or process to an institution, industry or profes sion as a whole, when this product or process represents its crucial or central concern. Conversely, an institution, industry of profession as a whole represents a product or process central to it. We can capture these relations in the form of a semiotic principle of indexical motivation as follows. Indexical Principle 11 : A sign expression x is indexically motivated by its content y, if x is conceptualized as an institution, industry or profession and y as a central factor in x. (reversible) This principle parallels that defining the relation be tween artifacts and their producers adopted above as In dexical Principle 2 in some measure. But the latter defines the connection between a human producer and his product, whereas Principle 11 extends to whole industries or pro fessions and their central features generally, whether they be products, instruments, processes, edifices etc. Further, Principle 11 describes the conceptualization of the central feature as that of a part of this institution
base
THE SEMIOTIC BASIS
57
as a whole, while Principle 2 was cast in terms of a lishes the relation from an actor to his role, Principle 11 that from the acting profession to the stage or theater.
causa
The effects of connections established by Principle 11 are reflected in natural languages by the conventional use of such terms as the stage or theater and the press for the acting and publishing professions or industries re spectively. One can more easily interpret a television set as a sign of the technical communications network than as that of the person who was responsible for the production of its picture tube, since the former involves an immediate transfer from the integral and determining part to its whole as allowed by Principle 11, while the latter involves a producer-product transfer along with abstractions due to the mechanization of the production process in cases of artifacts such as televisions. Examples like the use of the theater as a sign of acting, the press as that of publish ing, a rocket as that of the aero-space industry, trees as that of forestry and a church building as that of the church as an institution offer strong evidence of the mul tiple applicability of Principle 11, which also functions on various levels within a single system. Thus, one views the mayor as a sign of local authority, the governor of state authority, the president of national authority in the United States.
In natural language, as pointed out above, the systematizat of words denoting objects to designate the institutions, in dustries or professions in which they play crucial roles, e.g. stage, press, church, television, medicine,entertain these same lines illustrate the productivity of Principle 11; we understand statements such as Harry works in food, toys, furniture, clothing to mean that he works in the in dustry producing the respective items. Principle 11 reflects the inferential validity of a physical part-whole connection fairly directly, and as such claims validity itself. theater as an institution, which, for its part, counts as evidence of (at least open-air) theaters. I thus include Principle 11 among the others in our indexical code.
2.4.4 Container
and
Content
The fourth major category of indexical principles de rives from the contiguity relation between containers and their contents. On a narrow interpretation of this rela-
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
58
tion, we require a principle predicting that a coffee cup, a wine bottle, a corn crib stand for coffee, wine, corn respectively, even when they are empty of these materials. One can point to a one liter beer mug as a sign of beer or as a sign of the content capacity 'one liter'. In both cases a container stands for its content, once on its in terpretation as a physical object and once on its inter pretation as a quantity. In its broadest sense, the con tainer-content relation may be thought of as applying to such diverse cases as connections between edifice and in stitution, location and occupants, costume and wearer or form and content generally. 2.4.4.1
Container-Content
I shall return to these relations below, and restrict my present attention to the narrower, physical interpreta tion of the container-content relation. I formulate the corresponding indexical principle as: Indexical Principle 12: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a con tainer and its content. (reversible) This principle is reversible, since coffee and wine serve as signs of coffee cups and wine bottles (or barrels), just as the cup or bottle signifies coffee or wine. This principle's inferential validity lies in the cultureindependent understanding of a cup, bottle or bucket as a sign of liquid. The existence of wine bottles con vincingly argues that of wine, and the existence of wine that of wine bottles (or containers of some sort intended for wine). The multiple applicability of Principle 12 is ex emplified by the examples cited, as well as by its signifi cance for natural languages on various levels. Principle 12 would predict that languages allow sentences like the milk tipped over meaning 'the milk container tipped over' and this bottle is sour meaning 'the contents of this bottle are sour', whose literal interpretations are anomalous, since liquids cannot tip over (unless frozen) and bottles cannot become sour. Something like Principle 12 is thus necessary to predict the intended interpreta tions of such sentences. The same principle allows us to interpret figurative stock phrases such as one bad apple spoils the barrel as meaning 'one bad apple spoils the other apples in the barrel'. The multiple applicability of
THE SEMIOTIC BASIS
59
Principle 12 can further be seen in the polysemy of words such as teaspoon, cup, pint, bushel, barrel, which may signify either containers themselves or their volumes as quantities. The productivity of Principle 12 presents itself any time we use or interpret a physical container as a sign of its volume, for instance in referring to a thimble, can or hat of some material. Recipes written for large groups of guests are sometimes (humorously) given in terms of bathtubs, wagon loads, buckets etc. The relation between physical containers and their volumes deserves some further comment, for it might appear to hold between a container viewed as, say, a drinking vessel and the very same con tainer viewed as a standard measure. As such, it may appear to some not so much a relation as an identity. No tice however that our conception of a cup qua drinking vessel involves our perception of it as a physical object in the class of drinking vessels with all the appertaining associations of liquids, drinking etc. By contrast, our conception of a cup qua standard volume measure involves our perception of it as a unit in the conventional system of measurements with all the appertaining associations of counting and measuring, perhaps recipes, cooking and baking as well. The former is defined by physical prop erties and physical laws, the latter by established con ventions. The relevant relation is thus one between two distinct functions of an object valid in distinct system, not one between the object as such and itself from another point of view. Seen as a relation between an object and a definition, the case of Principle 12 here at issue is of the same formal type as Iconic Principle 3, and no less justifiable than the latter. Now, on a different interpretation, Prin ciple 12 might be seen as extending naturally to define the relation between a container qua physical manifestation and its content qua mental content, i.e. the relation usually termed form and content. The relation of form and content holds between a book in its physical form and the book as a text, between a record in its physical mani festation as a plastic disk and the record as a recording of a particular (musical) text, or even between a text and its (literary)interpretation. But I would prefer to ex clude this extension of Principle 12. First, because it appears to define an arbitrary rather than a motivated re lation in some cases, viz. that holding between any token of a type and the coded type itself; and second, because truly motivated cases can be handled as instances of modeloriginal relations by our Iconic Principle 1. That is to say that cases of the mental construct-physical manifesta-
60
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
tion relation may be dissolved into expressions and con tents differing by but a single feature, viz. that ex pressing the difference between a concept or text and its physical occurrence. 2.4.4.2
Locality-Occupant
A second indexical principle based upon or parallel to that of container-content defines the relation between a locality and its occupants. Locality should be under stood to include geographically defined areas like con tinents and valleys, politically defined areas like countries and towns, and architectonically defined areas like museums and kitchens. We count as occupants not only individuals, but socially or politically defined groups and institutions. The rationale behind these broad des criptions lies in the desire to formulate a single prin ciple like the following rather than several closely re lated principles. Indexical Principle' 13: A sign expression x is indexically motivated by its content y, if x is conceptualized as a locality and y as the occupant(s) of x. (reversible) Based upon the definitions offered above for the notions of locality and occupants, Principle 13 describes a whole range of relations such as those holding between continents, countries, houses and their inhabitants, citizens, occupants respectively, but also between countries, towns, offices and the national groups, local governments, organizations (political, social, economic) inhabiting them. Since an organization, government, na tional group, occupant, social group etc. also serves to represent the geographical area, country, building, room it occupies, Principle 13 must be characterized as re versible. The same set of examples persuasively illustrates the multiple applicability of this principle both in different systems and on various levels within a single system. The identification of various kinds of groups with the lo calities they inhabit and their identification of them selves in the same terms presumably represents a universal tendency. Furthermore, because this identification pro ceeds from the entire solar system right down to the in dividual room, Principle 13 clearly enjoys multiplicity of application in both senses.
THE SEMIOTIC BASIS
61
The productivity of Principle 13 appears whenever an individual or group identifies itself with or becomes identified with a locality of any sort after inhabiting it for some period of time, whenever a locality of any sort comes to stand for its inhabitants. World history is rampant with examples of the application of Principle 13, and its unfortuante consequences have often enough in cluded wars and migrations. The systematization of the effects of this principle are seen in the French tout le monde meaning 'everybody' and the English city hall meaning 'the local goverment' in the phrase you can't fight city hall. Its systematiza tion also evidences itself in the high incidence of morphological connections between the names of national or local groups and the names of the localities they occupy in the languages of the world; witness: ParisP a r i s i e n , Torino-Torineso, New York-New Yorker; TirolT i r o l e r , Puerto Rico- Puerto Ricano, EnglandEnglishman among myriad others. Note too the lexicalization of readings for the names of towns or buildings on which they designate the government resident in them: Madrid for 'government of Spain', Stockholm for 'government of Sweden', The Kremlin for 'government of the U.S.S.R.', The White House for 'government of the U.S.'. The validity of inference from locality to occupant and conversely seems clear in most cases. Of course, in particular instances, a person from Thailand might not be a Thai and an Australian need not neccessarily be from Australia, since national laws concerning citizenship and place of residence as well as ethnic groupings within and across national-political boundaries tend to confuse 'natural' relationships. But as a general rule - and a psychologically real association - a person or group stands for its/his locality and vice versa. 2.4.4.3 Costume-
Wearer
Another indexical principle based upon the general con tainer-content connection defines the relation between a costume and its wearer, where costume may be construed as national costume, uniform or any mode of dress recog nizable as belonging to or characteristic of a particular individual or group. We can state the corresponding prin ciple in the indexical code as:
62
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY Indexical Principle 14: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as a costume and its wearer. (reversible)
The reversibility of this principle is ensured by the fact that a recognizable member of a group characterized by a certain costume just as certainly acts as a sign of the group, and hence the costume, as the costume does of the group, and hence of the person as a member of the group. The importance of this principle is also attested to by the cross-cultural recognition of costumes and groups as mutual signs of one another. In fact, it appears historically accurate to claim that costumes arise as a result of the efforts of certain groups to set themselves off from others, i.e. to be recognized as such and, con sequently, as different from other groups. The productivity of Principle 14 receives support whenever any newly organized group attempts to establish its identity by prescribing a certain sort of costume for its members or, indeed, when one group attempts to identify members of another group on the basis of a particular fea ture of the costumes of its members. Again, world history offers all too many examples of such practices. Of course, theater and cinema trade on our productive application of' Principle 14 to identify their characters as members of certain national, historic, professional and other groups. The multiple applicability of Principle 14 is evidenced by its significance in establishing group distinctions of various kinds in various constellations of groups. Its application on different levels within a single system shows itself in sets of examples in which one national group distinguishes itself from another on the basis of its customary costume, areal groups within this nation distinguish themselves from each other in the same way, and further, for the outsider negligible, differences in costume serve to set one ethnic or local group off from another within a particular area. The relations predicted by Principle 14 were histori cally systematized in cultures where each particular guild, class and religious group was identifiable on the basis of its costume. Contemporary examples proliferate in cultures where one class is distinguished from another, one age group from another, one sex from the other on the basis of modes of dress. We have no trouble identifying clergymen, soldiers (of different ranks), bakers, trapeze artists etc. on the basis of their individual costumes. The validity of inference from costume to wearer is in disputable. The effectiveness of disguise upholds rather
THE SEMIOTIC BASIS
63
than detracts from its validity. Despite the (fairly recent) acceptability of public transvestitism, a court of law will judge the presence of certain articles of wearing apparel at the scene of a crime to be significant evidence of the presence of a member of one or the other sex at that location. Similarly, various items of clothing, jewelry etc. will be accepted as evidence of the presence of a member of a particular ethnic, religious or age group, even (or perhaps especial ly) in societies which pride themselves on the integration of all groups or the leveling of all classes into a mono lithic culture. This ends the discussion of indexical prin ciples parallel to or parasitic upon the general relation of container-content. 2.4.5 Experience
and
Convention
I now turn to indexical principles defining relations between objects and states as we experience them in the world and objects and states as they are objectified, de fined, conventionalized in certain systems. 2.4.5.1 Experience-
Convention
One such principle helps persuade us that a reduction in the length of the column of mercury in a thermometer should directly correlate with our feeling physically colder, and that our physically feeling colder will be duly registered by the thermometer, or that an object which registers a low weight when placed on the scale will be easier to carry than one registering a higher weight. This principle would also establish the connection between things considered old by virtue of our relationships with them and objectively old things. We might formalize the relevant principle as follows. Indexical Principle 15: A sign expression x is indexically motivated by its content y, if x is conceptualized as the measure or registered value of a phenomenon z and y as the ex perience of z. (reversible) The characterization of this principle as reversible derives from examples like the foregoing. Our common ac ceptance of readings indicated by more or less technical instruments as signs of various phenomena in both everyday life and scientific experiment demonstrates the validity
64
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
of inference we attach to this principle. Under normal con ditions, we accept the direction a compass needle points as north and the fact that both hands of a clock are point ing straight up as evidence that it is twelve o'clock. But at the same time, we recognize a difference between what our calendars define as winter and what weather conditions count as winter; and anyone who has experienced jet lag will readily admit that what the clock indicates may significantly diverge from what he would like to recognize as the time of day. Such discrepancies find their way into the language system; witness the ambiguity of words like old and neu) : as noted above, an old friend can be an ob jectively old person or a person who has been a friend for a (subjectively) long period - even young people can have old friends. We can also use words like winter, noon, cold, heavy to mean either these entities as conventionally defined or these entities as typically perceived. The in troduction into weather reports of the so-called wind-chill factor, for example, represents a concrete attempt to bring meteorologists' definition of the ambient temperature into closer harmony with people's perception of it. The coded ambiguity of the words cited and further parallel examples exemplify the systematization of the effects of Principle 15, the diversity of the examples exemplify its multiplicity of application. Its productiv ity can be seen in the invention of ever new apparatus to test for the presence of factors in the environment and our interpretation and acceptance of the results their scales register as significant for our behaviour in this environment. The calculation of time from centuries down to milliseconds and weights from megatons down to milli grams exhibits the applicability of Principle 15 on dif ferent levels within a single system. Since the inferential validity of Principle 15 provides the basis for our recog nition of the results of measurements generally, it remains beyond question for present purposes (no matter what doubts Heisenberg might raise), so that I include it among the indexical principles. 2.4.5.2 Manifestation-
Definition
Another indexical principle concerning the relation be tween experiential phenomena and conventionalized know ledge describes connections such as that between a particu lar observable stage of an object not constant in time and the conventional definition of this object as the selfsame entity despite changes through time. Quine (1960) discus ses this relation as the stage-thing ambiguity. He points
THE SEMIOTIC BASIS
65
out that adherence to personal observation alone without appeal to some sort of relational principle leads to a view like that Heraclitus allegedly championed, according to which one cannot step into the same river twice, because of the constant flowing of the water. Of course, most of us naturally apply a principle of inference in such cases which allows us to conclude from our observation of a stage of a river that the river somehow remains the same through time and that we can ignore accidents and momentary changes. We can thus formulate as our sixteenth principle of indexical motivation the following, which represents the inferential step Heraclitus felt compelled to call into question. Indexical Principle 16: A sign expression x is indexically motivated by its content y, if x is conceptualized as a conventionally defined entity z and y as the experiential manifesta tion of z. (reversible) This principle is intended to define not only the stage-thing relation, but that between observed entities or states and entities or states accepted on the basis of con vention or definition. A person observed to be engaged in the act of skating stands in a particular relation with a person who by custom or profession skates, whether he be skating at the given moment or not. Natural language lexi cal systems reflect this relation in allowing application of the word skater to people in both situations. Black (1952) discusses the case of skater as an example of his action-custom ambiguity. Clearly, present action yields to direct observation, whereas customary action must be de duced on the basis of various observations. The situation is similar in cases where a property may be attributed to an individual or object due to present observations or because of knowledge of the potential of the individual or object in question. Thus, we commonly classify as fast cars both those actually moving rapidly at the moment of classification and those believed capable of rapid move ment on the basis of various observations or conventions. We see a similar connection in the relation between lunch viewed as a collection of food and lunch viewed as a cul turally determined habit or institution. Since all the relations described clearly function in both directions, Principle 16 is correctly characterized as reversible. A person golfing serves as a sign of the customary golfer and vice versa. The range of examples cited so far indicates the principle's multiplicity of application. That it functions on different levels within
66
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
a single system can be seen in its application in cases of nouns such as river or skater and adjectives like f a s t , but also in adjectivals generally, whether they consist of single items like b r a v e , funny, powerful, or whole phrases such as like a warrior, in a funny way* so long as their interpretation can vary between actual and potential assignment of the predicate in question. The lexicalized items cited provide ample evidence of the systematization of Principle 16. Its systematization is deeply ingrained in our non-linguistic perception as well. Any description of a person, ontologically an entity variable in time, will be understood as a description of the same person at a later point in time, apart from fea tures determined by age, accident etc., in which cases, however, we could still say that these adhere to a person through the power of memory; a description of a culture and/or society will be understood as a description of the same culture or society at later points in time. In both cases Principle 16 applies to provide the link between an observed state of an entity and the entity as an (unchang ing) convention. Its productivity demonstrates itself in our propensity to generalize from any single observation of an object, person, institution to an entity persistent in time. Any time a new word enters the language and automatically ex tends to cases of both present and potential characteri zation of the property in question, the productivity of Principle 16 is exemplified. The validity of inference as sociated with this principle has already been evidenced. Under normal conditions, we are prepared to accept an (ob jective) observation of one stage of an object or state as evidence of the (objective) reality of the object or state. The converse obtains by definition: anything existing through time necessarily exists at various times. We can thus include Principle 16 in the indexical code. This ends my discussion of principles of motivation based upon the relation between entities as perceived and entities as registered, measured, defined. 2.4.6 Possessor
and
Possession
Two final indexical principles are that relating people with their possessions, and that relating people with their offices or roles.
THE SEMIOTIC BASIS 2.4.6.1
67
Possessor-Possession
The first of these
principles I express as:
Indexical Principle 17: A sign expression x is indexically motivated by its content y, if x is conceptualized as a possession of y's. (reversible) Here y can represent either an individual person, a corporation, an institution or anything conceptualized as possessing something. In addition, possession will be in terpreted quite broadly to include not only physical ob jects, but such entities as an excellent memory, physical beauty, charisma, authority, political power. Nor do I re strict possession to cases of inalienable or life-long ownership, but rather include short term claim to an en tity. Examples of possessions standing for their owners or conversely proliferate. A group or nation stands for its language and customs, and conversely. A persons stands for (in both senses of the term) a particular set of beliefs and opinions, but he also calls forth thoughts of certain habits and personality traits. If upon hearing a descrip tion of some curious set of habits or beliefs I exclaim, "That's Harry!" I have applied Principle 17 in the intended sense. Of course, any physical object recognizable as the property of someone will serve as a sign of that person: we intend our homes, cars, clothing and accessories to stand for ourselves to all observers. Possessions clearly count as evidence of their owners in courts of law. If a weapon known to belong to a particular person is found at the scene of a murder, any jury will consider that person a suspect. Fingerprints are acceptable evidence only be cause they correlate with (at least short term) possession. The examples cited should suffice to exhibit the in ferential validity of Principle 17. Its multiple appli cability can be seen in the above examples and the fact that it applies from an object to an individual, from the object to a family, club, corporation, multinational etc. The productivity of Principle 17 appears any time someone uses or interprets any object as a sign of its owner. If I see Harry's car pull up outside, I will assume Harry has arrived. The systematization of Principle 17 takes several forms. Apparently all languages can designate an object as someone's possession by appending to its name or descrip tion some form of the owner's name, perhaps along with certain other morphemes. But we also commonly use people's names without morphological change to refer to their possessions. If I am carrying a tray of drinks, I can in-
68
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
dicate whom each is intended for by saying, "This is Harry, that's Judy etc." At the same time, one can refer to people by naming their possessions. If everyone at a bridge table has a different drink, I can review the bidding by saying something like, "White wine bid two spades, beer doubled, red wine said three hearts, and I passed." In light of these and the other examples cited, we are justified in in cluding Principle 17 in the indexical code. 2.4.6.2 Office
Bolder-
Office
The second principle based on possession of a sort re lates individuals with the offices or roles they occupy. Indexical Principles 18: A sign expression x is indexically motivated by its content y, if x and y are conceptualized as an office or role and the office holder.(reversible) Principle 18 relates the office of the president, secretary, treasurer or member with the individuals occupy ing these positions in a group. It relates individuals with the titles, ranks, departments or academic chairs they hold or belong to. We commonly refer to people either by their proper names or their titles, positions, roles etc. (and in some cases a combination of them: Department Chairman, Prof. Harry Harrison). A person stands for the office or position he occupies, at least within the in stitution in which it is defined, and often outside it as well. All groups recognize people as occupying roles within them, and necessarily associate certain roles with certain people. Indeed, children generally learn the designations for people's roles before they learn proper names: mommy, daddy, grampa, gramma denote the roles people play in the baby's world. Interestingly, many personal names derive from designations for social roles: in English we find many names such as: Baker, Carpenter, Farmer, Householder. This fact speaks for the systematization of Principle 18. Its systematization is further evidenced by the regular double meaning of any term denoting a role or office: such terms can be interpreted as referring to the role or office within the institutional system of which it represents a part or to the person occupying the role or office. We can say that the King of Sweden (as office) praised the civil servants and also that the King of Sweden (as person) has a cold. The multiple applicability of Principle 18 is evident from its functioning on all levels from the family up to
THE SEMIOTIC BASIS
69
and including the nation or international institution (e.g. The Pope opposes contraception versus The Pope is a Pole). It applies on all levels within a particular organization from Its productivity follows from the potentially endless use of position designations referring to individuals and personal names referring to occupants of positions. Harry invited the director to d i n n e r , and Judy accused management of bargaining dishonestly against us involve standard uses of a position designation for an individual and a personal name for a particular position in an institution. The validity of inference from an individual to his role or position and conversely is beyond question for most purposes. In a court of law, proof that the butler of the house committed the murder will more than suffice for the conviction of the person who fills this position and vice versa. I thus include Principle 18 among the others accepted into the indexical code, and thereby conclude the discussion of individual indexical principles of motivation. Further more or less plausible indexical principles could be shown to fail to meet the criteria we set down, to depend upon or be predictable from principles already accepted. For instance, the relation between an object and the material of which it consists falls under the domain of Iconic Principle 4, that holding between a dance and its music demonstrates an application of Indexical Principle 11, since the appropriate music represents a crucial part of the whole event of dancing. I shall then end this discussion of the indexical code, and of moti vation in sign relations generally.
REGULAR SEMANTIC RELATIONS
3.
3.0
Introduction
This chapter opens with a presentation of arguments for treating regular semantic relations as predictable on the basis of general semiotic principles of motivation. I argue that linguistic competence represents a special case of a more inclusive semiotic or sign-manipulating compe tence. Then, if semantic relational principles manifest the same properties as semiotic principles of motivation, the simplicity metric dictates the derivation of the for mer from the latter. Given the semiotic principles de scribed above, one can specify the semantic relational principles as redundant. I formulate the sets of metaphoric and metonymic relational principles to be included in sem antic theory; then I outline their functions and integra tion into the semantic component of a grammar. 3.1 Principles
of
Motivation
and Semantic
Regularity
Linguistic competence is embedded in semiotic or signmanipulating competence. This holds true in various ways. Just as we can say that grammatical competence represents one aspect of acceptable performance (cf. Chomsky 1965) , we can view both as aspects of communicative competence, in the sense of Hymes (1962, 1971; cf. Habermas 1971). The first concerns the ability to produce and interpret the grammatical sentences of one's native language, and to distinguish them from all other strings. The second con cerns the utterance and understanding of all the accept able sentences in the language, and recognition of unac ceptable sentences. Communicative competence concerns one's ability to behave linguistically in a manner appropriate to any given cultural situation, and to recognize inappro priate linguistic behavior as such. At this communicative and socially relevant level, we can define a further sort of competence which includes and
72
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
transcends the purely linguistic: let us call it semiotic competence. It consists in the ability to manipulate signs (including linguistic signs) in a manner appropriate to a given cultural situation, and, of course, to recognize as such inappropriate behavior with respect to signs gener ally. In any culture, semiotic competence will include, beyond linguistic competence, the ability to recognize and manipulate all sorts of signs, whether they are convention ally coded or not. Hence, it encompasses our familiarity with the significance of body movements and positions, types of dress, ceremonies and rituals, street signs and shop signs, diagrams and graphs, works of art, scientific and popular taxonomies and theories. It accounts for our recognition of the appropriateness of a scale as a symbol (in the normal, non-Peircean sense) of the idea of justice (cf. Saussure 1916), and the inappropriateness of sending a box of cigars along with a condolence card to someone in mourning, as well as determining the type of card to be sent and the sorts of messages appropriate to include in it. Defined In this manner, semiotic competence clearly provides the basis for and transcends communicative compet ence on Hymes' definition. There are, however, other senses in which it appears correct to view linguistic competence as embedded in sem-• iotic competence. In language learning, understood as the assimilation of communicative competence in general, it seems clear that a child must first grasp (or innately 'know') that signification is possible, that one thing or act can stand for another thing or act, before it can even begin to react to or experiment with language. Very early, children apparently associate big people with food and attention, and crying with the arrival of big people. These are already instances of semiosis; they begin to es tablish recognition of relations of signification or standing for. Later the child learns to spit out food it does not like or to throw it on the floor long before It learns to express its displeasure verbally; it will learn to point at or reach out toward the things it wants before it can begin to verbalize its desire for them. In general, command of the non-linguistic forms of signification appro priate to any given situation precedes command of the ap propriate linguistic forms. Nöth (1977: 19) writes that the process of learning to deal with signs begins with the index, then procedes via the icon to the symbol. Semiotic competence is ontogenetically prior to linguistic compet ence (cf. Nöth 1977: 16 ff. and references there). So in this sense too we can maintain that linguistic competence is embedded in and dependent upon semiotic competence.
REGULAR SEMANTIC RELATIONS
73
Many stages and processes in language acquisition are mirrored in inverse form in cases of pathological language loss (cf. Jakobson 1969, 1953; Luria 1973). Just as semiotic sition, the latter disappears before the former in cases of aphasia. Nevertheless, the loss of a particular portion of linguistic competence may sometimes be paralleled by a loss of a corresponding portion of more general sign-ma nipulating competence. Victims of severe agrammaticism, who experience great difficulty in correctly ordering morphemes into words and words into sentences, also ex perience difficulty in ordering objects of any kind into specific patterns or relations with one another (cf. Luria 1972, 1974). But concrete symbolic ability always remains intact to a greater degree than more abstract symbolic ability (cf. Goldstein 1948). Data from aphasia thus also corroborate the contention that linguistic competence is encompassed by and more highly specialized than semiotic competence. Linguistic behavior is embedded in more general sig nifying behavior in still another way. In all normal cases of face-to-face communication, linguistic acts are accom panied by other simultaneous significant behavior. Declar ations of affection are most appropriately accompanied by special types of bodily contact and/or facial expressions. Verbal abuse is commonly delivered in a loud voice, with an angry facial expression, while shaking a fist or even along with physical violence directed at the victim's person. Exophoric expressions such as this, that, him, her generally receive specification in the form of pointing with fingers, hands, the direction of the speaker's gaze etc. But gestures and facial expressions do not always simply correlate with or intensify linguistic messages; indeed, they may even contradict the linguistic informa tion offered. And it is significant that, when kinesic and linguistic Information are in conflict, we typically tion to be an outward show with the intention to deceive. If I say to someone, "I really like your suit," while pre tending to look ill, he will interpret the complex of word and act to mean I do not like the suit. Similarly, If Î say, "Uh, is that an ugly tie" with a pleasant facial expression and a so-called gleam in my eye, I will be in terpreted as having paid its wearer a compliment. If some one in tears over the loss of a game says, "I'm glad I lost," we interpret the truly present emotion to be one of sorrow at losing, despite what was said. Thus, gestures and non-linguistic signification generally are more basic
presume
74
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
and persuasive than language; they are convincing without the aid of and even in spite of the influence of language. Again we see that linguistic competence and behavior are embedded in and dependent upon more general modes of sig nification. Notice also that non-linguistic modes of signification commonly take over in situations where the linguistic mode appears insufficient, is subject to constraints or fails. When we are overcome by emotion or simply wish to express strong emotion, we often forsake language for more direct methods of signification: we sigh, caress, pound the table, throw up our hands, scream, physically assault our inter locutor(s). When we are at a loss for a word, we resort to pointing, drawing, onomatopoeic imitation, charades. In order to avoid taboo expressions, we draw from a large coded repertoire of gestures: drawing a finger across the throat to indicate that someone has been killed; holding a hand some distance from the belly to indicate that some one is pregnant; making certain 'lewd' gestures or move ments to mimic various sexual acts or body functions. When called upon to sooth or amuse a baby, we similarly with draw from language into babbling noises and bodily con tact. Finally, when two individuals, each quite capable of communicating in one or more languages, share no common language, they too resort to gestures, charades and sketches they hope bear cross-cultural significance.Sem provides the foundation for linguistic competence, it sur rounds the latter and encroaches upon its territory any time language reveals local weaknesses. From a number of different points of view one can thus correctly maintain that linguistic competence Is embedded in and dependent upon semiotic competence, that linguistic behavior mirrors or derives from more general sign manipu lating competence. If we can now show that a particular element in linguistic systems fulfills the same function on the relevant level as a corresponding element (i.e. one sharing a significant number of crucial properties) in semiotic systems generally, then considerations of simplic ity and common sense dictate that the linguistic element be extracted from linguistic theory proper and be stated only as part of the more general semiotic theory. Alterna tively, the element might be included in the linguistic theory, but characterized in some manner as redundant on the basis of the more general semiotic element. This mode of argument is common. To the extent that the syntactic distribution of certain elements on that level can be predicted on the basis of semantic properties
REGULAR SEMANTIC RELATIONS
75
assigned these elements, they are extracted from the syn tactic representations of the elements as such or at least specified as redundant. The law of gravity holding for ob jects near the surface of the earth can be derived from, and is redundant in light of the general law of gravita tion pertaining to bodies of all types. Heisenberg's Prin ciple of Uncertainty enjoys a special interpretation and significance in atomic theory, but holds in a general way of all observed phenomena. Fire represents a special case of the general chemical process of oxidation. In each case, a particular event or property is analysed as a special instance of a more generally occurring phenomenon on the basis that the event or property occurs within the domain of the more general phenomenon and shares crucial proper ties with it. Each particular theory would be complicated by viewing any of the cases cited as processes or facts independent of the more general theory. The simplicity metric dictates that the particular theory in each case make reference to the more general theory for prediction of the item or law in question. I employ precisely this reasoning in the argument to follow. In particular, I submit that regular semanticrelat semiotic principles of motivation. Since we have already seen that linguistic competence and ipso facto semantic competence are embedded in semiotic competence, I need only exhibit that regular semantic relational principles share crucial properties and functions with those of sem iotic motivation. The arguments are concretized by the pre sentation of the relevant principles of motivation and their characteristic properties in the preceding chapter. Hence, in specific terms, I will be attempting to demon strate that statements expressing regular semantic rela tions possess all the crucial properties seen to be charac teristic of the principles of motivation described above, the only significant differences being those directly at tributable to their definition over elements on the seman tic level of linguistic description as opposed to sign sys tems generally. Compare Jackendoff's (1978) comments on language as a special case of the general process of con ceptualization . First, semantic relational principles and semiotic principles of motivation both describe relations between signs; semantic principles are limited to describing rela tions between elements defined in semantic theory rather than between signs generally. In particular, semantic rela tional principles describe relations between semantic representations, which take the form of lexical readings in the lexical entries for words and larger stored units, and
76
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
(portions of) sentence interpretations. Thus, semantic relational principles appear to represent a special case of semiotic principles of motivation, viz. those restrict ed in their application to semantic representations. Viewed psychologically, principles of both types express people's tendency to recognize connections between cer tain entities.
Second, principles of both types are based upon inference concept of fire from that of smoke. Since the lexical entities for the words smoke and fire include readings intended to express our concepts of these entities, the lexical entry for smoke should call forth the entry for fire inferentially. That is, if the perception of smoke allows us to infer fire, then the concept of smoke allows us to infer the concept of fire, and, consequently, the word smoke by virtue of its lexical reading also allows us to infer the concept of fire. The semantic relation speakers recognize between (the meanings of) violin and violinist appears to be based upon the same inference from instrument to agent which our Indexical Principle 8 de scribes. Certainly, Principle 8 would predict a mutual in ferential relation between the meanings of these words, even in the absence of a special semantic principle. Simi lar arguments may be constructed for such word pairs as
baker-hake , bigot-bigotry,black-blacken
, open-opener
and
others cited in the previous chapter. Thus, semantic rela tions are based upon inference, just as we saw motivated sign relations to be. Third, principles of both types can be used produc tively. The statements we accepted as iconic and indexical principles above all evidenced productivity. The same can be shown for semantic principles describing regular rela tions. The principle which predicts the semantic relation between violin and violinist would also productively ex tend to describe that between flep and flepist, if flep designated an instrument and flepist the person who played it. Just as one finds golfer and golf, a productive appli cation of the principle here involved would also capture the relation between a noun bolfer and a verb bolf, if the latter designates an act and the former a person engaged in it. Obviously, such examples could be easily multiplied. We can thus maintain that both semiotic principles of mo tivation and semantic relational principles exhibit pro ductivity. Fourth, the effects of both sorts of principles become systematized. In the case of semantic principles,systemat lexicalization of relations predictable on the basis of
REGULAR SEMANTIC RELATIONS
77
semiotic principles were cited in chapter 2. They can all be interpreted as examples of the lexicalization of re sults of the corresponding semantic principles, so we need not belabor the point here. Fifth, both semiotic principles and semantic prin ciples can be divided into those based upon similarity relations and those based upon contiguity relations. In the case of semiotic motivation, similarity relations correspond to iconic principles, and contiguity relations to indexical principles. Semantic relations have tradi tionally been divided along the same lines into metaphoric and metonymic relations respectively (cf. Jakobson 1953, 1954). The principle relating an object with a feature characteristic of a class including it was defined as an iconic principle (number 4), because the relation between an object and its class is a matter of similarity. The principle relating part and whole, however, clearly de pends on contiguity and thereby counts as an indexical principle (number 9). In the domain of semantics, the fig urative occurrence of laugh meaning 'produce pleasant sounds' must be classified as metaphor, since the relation between laughing and making pleasant sounds is one of similarity, i.e. laughing belongs to the class of acts involving production of pleasant sounds. The figurative occurrence of boxing gloves meaning 'the act/sport of boxing1 in there's money in boxing gloves, on the other hand, counts as metonymy, since the relation between box ing gloves and boxing is the contiguous connection from instrument to act (cf. Indexical Principle 8). Sixth, not only do semantic principles divide along the same parameter as semiotic principles, they also ex hibit the same distribution with respect to the property of reversibility we saw to be significant in the latter. That is to say, like semiotic motivational principles, semantic relational principles show reversibility only in continguity relations (metonymy), and irreversibility in similarity relations (metaphor). While laugh can occur metaphorically meaning 'produce pleasant sounds', the phrase produce pleasant sounds cannot metaphorically mean 'laugh'. Firstly, because the waterfall produced pleasant sounds is not metaphoric: the projection rule (Katz 1972) will assign it a consistent reading. Secondly, because even a figurative occurrence of the phrase such as the idea produced pleasant sounds will not be interpreted to mean 'the idea laughed'. On the other hand, one might re fer to a pair of boxing gloves by saying that's boxing, say to differentiate them from hockey gloves. Metaphoric semantic principles are irreversible like iconic motiva tional principles, whereas metonymic semantic principles
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
78
are reversible like indexical motivation principles. Seventh, principles of both kinds lay claim to multiple ap chapter for semiotic principles and often illustrated with linguistic examples, which convey an idea of the multiple application of semantic principles as well. A semantic relational principle, say that connecting cause and effect, typically captures relations on different levels. We find morphologically related pairs such as openV-openA, blackenblack, ptease-pteasure, where the lexical reading of the first identifies an act conceptualized as the cause of the state identified by that of the second. The cause-effect relation also holds between the readings of otherwise unlinked also hold between two derived readings like those for At opened the door and the door is open or At kitted Fido and Fido died. We also find constructions whose figurative and literal readings are related by the principle of cause-effect. Make tracks literally describes an effect of its intended figurative reading 'leave'. In relating read ings of lexical items with and without morphological con nections, those of sentences, and literal readings for sentences with their figurative counterparts, semantic relational principles manifest a multiplicity of applica tion at least as great as their semiotic correlates. Eighth, semantic relational principles can be formu lated exactly as semiotic principles of motivation, making only those changes necessary to limit their scope of appli cation to semantic readings. Given a semiotic motivational principle like Indexical Principle 1, we can convert it to a semantic principle of cause-effect describing the rela tions outlined above as follows. (1)
A semantic reading x is indexically motivated as a sign of a semantic reading y, if x and y are conceptualized as identifying a cause and its effect. (reversible)
Insofar as motivation was seen to be a regular relation between an expression term and its content, (1) states a regular motivated relation between one semantic reading (viewed as the expression term of a sign) and another (as its content). So (1) predicts a regular relation between the lexical readings for blacken and b l a c k , and the other pairs cited. We have thus seen that semantic relational principles represent special cases of semiotic principles of motiva tion. Since the former must be limited in their applica tion to semantic entities, they should be stated in
REGULAR SEMANTIC RELATIONS
79
semantic theory as such, but characterized as redundant on the basis of the corresponding semiotic principles. Whether the entire set of semantic relational principles holding in any given natural language can be derived from the set of semiotic principles making up the Iconic and indexical codes remains unproven; to be accurate, I have shown only that all the regular semantic relational principles are derivable from semiotic principles and, on this basis, must be characterized as redundant if stated as such. I intend the predicate regular to bear the same relevance for semantic principles as I assigned it for semiotic principles above. Thus the semantic principles derived will conform to the same criteria I set down for semiotic principles. This result is in consonance with my objec tives, since I hope to determine a set of semantic prin ciples as narrowly constrained as will be consistent with the data, i.e. I seek the set of truly regular semantic relational principles. 3.2 Regular
Semantic
Relational
Principles
As I showed with the example of (1) in the preceding section, semantic relational principles can be derived di rectly from semiotic principles of motivation simply by introducing into the latter such changes as are necessary to restrict their operation to semantic elements. Since motivation consists in a regular, inferentially valid re lation, as defined here, one could cast all the semantic principles in terms of motivated sign relationships like that in (1). Indeed these principles serve on one of their functions to establish the motivation of morphologically compound units by relating them to their component lexical items, so that they can correctly be characterized as mo tivational principles. Nevertheless, for the purpose of formulating principles which apply to the readings of lexical items in particular classes or, equivalently, those associated with particular features, It makes better sense to formulate semantic principles in terms of defin ing regular relations between semantic representations and features than in terms of motivation. In any case we must state at some point in a total description of semiotic cum linguistic competence that semantic regularity derives from motivation in sign relations generally.
80
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
3.2.1
Metaphorie Relational
Principles
Because I want to maintain the distinction established above between similarity and contiguity relations even in the realm of semantic principles, I differentiate between metaphoric and metonymic relations respectively. In the framework of semantic theory, we must reinter pret the phrase "is conceptualized as" In a manner slightly different than above for general sign theory. Since lexi cal entries for words will be characterized in terms of truly distinctive features, i.e. only those features re quired to set a word off from every other word, the notion of conceptualization is required to ensure that cultural/ encyclopedic knowledge enters into the speaker-hearer's understanding of a particular lexical item. A word such as fox will be specified in the lexicon for the features (Wild), (Canine) etc., but is 'conceptualized' in certain cultures as involving features like (Sly), (Game (in hunt ing) ). The inclusion of "is conceptualized as" also takes the fuzziness of conceptual categories into account (see 2.2.2 below). In addition, in stating semantic principles, I shall employ the following conventions: x, y, z stand for semantic representations; these consist of sets of features f 1, f2. . . fn, where f m designates some feature beyond fn; capital , , stand for semantically defined classes. 3.2.1.1 Metaphorie Principle
1
I can now formulate the semantic statement correspond ing to Iconic Principle 1 as follows. Metaphorie Principle 1 : x is in regular metaphorie relation to y, if x is con ceptualized as specified by f1 , f2. . . fn and y as f1, f2. . . fm, where f m was removed from x by a sem antic transformation. The sort of transformation outlined for sign relations generally can be expressed in semantic theory as: Semantic Transformation 1 : Delete a feature f n from x in a class A, if fn class in and incompatible with A.
distingui
REGULAR SEMANTIC RELATIONS
81
Taken together with Metaphoric Principle 1, this trans formation describes typical examples of meaning extension. It differs little from the rules for metaphor proposed by such writers as Leech (1969, 1972), Abraham and Braun müller (1971), van Dijk (1972) and Levin(1977), except in being explicitly derived from a general semiotic principle. It works like this. The verb laugh will be classified among those predicates in the lexicon which describe human actions: laugh bears the feature (Human). For a sentence like the waterfall laughed we require a dehumanized inter pretation of laugh. In the formal description of this case, the transformation above applies to the feature (Human) of the semantic representation of laugh, since, in consonance with the transformation's description, the feature (Human) distinguishes laugh from such items as purr and coo, which will be characterized as (Non-Human). Our Metaphorical Principle 1 establishes a regular relation between the contextual interpretation of laugh and its lexical read ing. That is to say: one regular process of meaning ex tension involves applying a word as a predicate when one of its significant features does not match the context. To cite a further example, read must be characterized as a mental process involving interpretation of visual linguistic signs; its occurrence in the sentence Al read Sue 's face can be predicted by the application of Semantic Transformation 1 to remove the feature (Linguistic) from the lexical representation for read. Metaphoric Principle 1 guarantees that this will result in a regular relation between the two. The verb read also naturally extends to tactile interpretation in cases of Braille. Here the fea ture (Linguistic) remains, but the feature (Visual) must be deleted as in: Judy reads Braille with her left hand. The examples offered should suffice to illustrate the pro ductivity and multiple applicability of the first prin ciple of metaphoric extension. The final example also pro vides a case of its lexicalization. Others are instanced by Sue swallowed the story, Harry harked at the others, Vera unwound with a cocktail etc. Metaphoric Principle 1 or something like it must be included in any semantic the ory intended to capture lexical relations and productive processes of the type illustrated. We can accept It as the first semantic relational principle in the theory to be developed here, since we wish to describe the regular re lations and processes recognized by competent speakerhearers.
82
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
3.2.1.2 Metaphorie
Principle
2
The second iconic principle links class members with their classes taken extensionally as the complete set of class members; it defines the specific-generic relation. The corresponding semantic principle takes the following form. Metaphoric Principle 2 : x is in regular metaphoric relation to y, if x is con ceptualized as a member of A and y as the set of mem bers of A. This principle applies to any representation in a particu lar semantic class, i.e. to any content word in the lexi con, but also to entire phrases insofar as they define entities conceptualized as belonging to identifiable classes. For example, b e a v e r and run will both be asso ciated with lexical readings consisting of features identi fying any specific beaver or act of running, but our second principle predicts that they can designate the generic class of beavers or acts of running. Metaphorie Principle 2 relates the interpretation of the phrase rodent with a broad flat tail on which it picks out a particular rodent with that where it identifies the generic class of such rodents, and those of run the mile in Judy will run the mile from here to Harry's house, where it describes a spe cific act and a specific mile, with that in All who run the m i l e , where the generic mile is intended. The speci fic-generic principle must thus be productive. Clearly too, it enjoys multiplicity of application and systematization in natural languages generally. The primary func tion fulfilled by Metaphorie Principle 2 in semantic the ory is that of predicting the systematic specific-generic meaning relationship of lexical entries for nouns. The in clusion of this principle in semantic theory brings with it a radical simplification of the lexicon in making it unnecessary to list both specific and generic readings for items; it requires no other justification. 3.2.1.3 Metaphorie
Principle
3
The third metaphorie principle, like the third prin ciple of iconic motivation, relates class members to their classes taken intensionally as the set of their defining features. It will express the relation between any lexical entry and the lexical class to which it is assigned as
REGULAR SEMANTIC RELATIONS
83
well as to any hyponym in a higher, more inclusive class. On the productive side it predicts the connection from a concrete statement to a general proposition from which it follows. It thus enables us to interpret proverb or axiom type statements as expressions of general truths: by Metaphoric Principle 3, make hay while the sun shines can be assigned the figurative interpretation 'perform any act while propitious circumstances obtain'. I formulate the relevant semantic principle as: Metaphoric Principle 3: x is in regular metaphoric relation to y, if x is con ceptualized as a member of A and y as the feature spe cification of A,
As I pointed out above, this principle's systematization takes the form of hyponymy relations and lexical clas sification generally in a lexicon set up around semantically defined c in the present study. Since classes will necessarily be contained in more inclusive classes from the most particu lar to the most general in such a lexicon, this principle clearly claims not only systematization, but multiple ap plicability as well. Its productivity exhibits itself out side the lexicon, where it associates not only stored phrases like make hay while the sun shines and strike while the iron is hot with the general interpretation 'perform any act while propitious circumstances obtain', but original figures of the same kind, e.g. get out while the door is open or eat while the food is warm. We thus include Metaphoric Principle 3 in our semantic theory.
3.2.1.4 Metaphoric
Principle
4
Iconic Principle 4 of our iconic code defines the con nection between a class member and a single feature (com plex) crucial to the definition of its class. Since indi vidual lexical entries are associated with feature speci fications as well as semantic classes in the lexicon, we can state the semantic correlate of this iconic principle as: Metaphoric Principle 4: x is in regular metaphoric relation toy, if x is con ceptualized as specified by f1 , f2. . . fn defining its membership in A, and y as fx, a feature distin guishing A from any class containing A, or from any
84
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY class in and incompatible with A.
Notice first that this principle stands in an interest ing relation with Metaphoric Principle 1. The latter holds between two semantic representations differing by but a single crucial feature, the present principle holds be tween two semantic representations exhibiting but a single crucial feature in common. We might bring out the paral lelism more strongly by formulating Metaphoric Principle 4 as holding only when all but the single feature are de leted by an appropriate transformation. We would then have a transformation deleting all but a single feature paral lel to the transformation defined above to delete only a single crucial feature. Notice also that, unlike the iconic principle upon which it is based, Metaphoric Principle 4 appears to de fine a reversible process in natural languages.While the existence of spiny things does not ensure that of porcu pines in the strict inferential sense, the feature (Spiny) suffices to differentiate porcupines from all other ro dents, at least for purposes of the taxonomy reflected in natural language. Similarly, the feature (Female) suffices to pick out the class of mothers, when that of parents is given; (Liquid) sets cooking oil off from lard, shortening and margarine in the domain of fats used for cooking etc. So the statement of Metaphoric Principle 4 could be re versed to relate the feature f x as expression with x as a member of class A. I therefore reformulate as follows. Metaphorical Principle 4 (revised): x is in regular metaphoric relation to y, if x is con ceptualized as a feature f x present in the specifica tion of y, f1 , f2. . . fn/ where all others were transformationally deleted. The relevant transformation can be given as: Semantic Transformation 2: Delete all but a single feature fn from x in a class A, if fn distinguishes A from any class containing A, or from any class in and incompatible with A. The relation holds between a contextual occurrence of a word where a single feature essentially exhausts its sem antic contribution and the lexical entry for the word in question. Here the transformation describes the extension
REGULAR SEMANTIC RELATIONS
85
of the lexical representation of an item by deleting all but the feature relevant to the context. The former case illustrates the systematization of Metaphori Principle 4; its multiple applicability and productivity are exhibited outside the lexicon, except in cases where metaphors of the relevant type have been lexicalized. The lexicalized readings of 'sly' for fox, 'silly' for goose, 'strong' for iron, 'ordinary' for pedestrian are cases in point. Pro ductive application of Semantic Transformation 2 to yield relations described in the metaphoric principle is seen in sentences such as Judy is a factory where factory carries with it only the feature (Productive) and a glass excuse where glass contributes only the feature (Transparent). The description of such lexical relations and productive extensions presupposes something like Metaphoric Prin ciple 4 and the corresponding transformation, so I include both in the semantic theory. 3.2.1.5 Metaphoric
Principle
5
The fifth iconic principle we discussed above was seen to be defective in that it was reversible and not based upon a valid inference. It relates a member of a class with any other member of that class. The corresponding semantic principle seems clear enough. Metaphoric Principle 5: x is in metaphoric relation to y, if x and are con ceptualized as members of A. We consciously omit the word regular, since the source semiotic principle lacked inferential validity. As it stands, this principle predicts only that words in a single semantic class are related. In a lexicon of the sort envisioned here, this receives recognition in the ordering of synonyms into the same ceived lexical classes. For instance, a single lexical class characterized by the features (Human), (Action), (Locomotion), (Linear), (Rapid) etc. will contain run, sprint, race etc. In fact, all cases of identity or near identity in feature specifications for lexical entries will be expressed in their inclusion in some semantic class. Indeed, one could go beyond Metaphoric Principle 5 to say that any entries sharing major features are related, but this would entail no interesting consequences. It simply
lex
86
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
provides a reason for setting up the lexicon around seman tic classes, or in traditional terms, semantic fields (cf. Trier 1934; Weisgerber 1954; Coseriu 1967; Geckeler 1971; Lehrer 19 74b). On the productive side, Principle 5 seems to predict only the mutual substitutability of synonyms. Hence, unless we decide to include it redundantly and without consequences in semantic theory, the only expression this principle enjoys or requires consists in its relevance for the structure of the lexicon. I then recognize four metaphoric principles as part of semantic theory, as well as two semantic transform ations. Together they predict a range of lexical relations, semantic motivation in morphologically compound units, and productive processes. They cover many linguistic phenomena traditionally designated as metaphor (as opposed to metonymy, synechdoche, irony etc.) and certain pre dominantly lexical relations. I now direct my attention to metonymic principles, which account for further lexical relations, and semantic aspects of other linguistic figures, viz. metonymy and synechdoche. 3.2.2
Metonymic
3.2.2.1 Metonymie
Relational Principle
Principles 1
Above I made a first attempt at stating the semantic principle corresponding to Indexical Principle 1. I would now refine this statement as: Metonymie Principle 1 : x is in regular metonymie relation with y, if x and y are conceptualized as identifying an effect and its cause respectively, (reversible) Several examples of the effects of a principle relating effects with their causes were cited in the discussion of Iconic Principle 1 above. We conceptualize acts of ob taining, and of learning as leading to, or causing, states of possessing, and knowing respectively. My first metonymie principle defines such connections as regular metonymie relations. As such, they should be reflected in the lexical entries for obtain,
learn,
possess,
know
in same manner.
I shall present a concrete proposal for expressing such relations below. It suffices for present purposes to note that these relations illustrate the systematization of the effects of Principle 1. Further examples are to be found in morphologically linked pairs consisting of an adjective
REGULAR SEMANTIC RELATIONS
87
signifying a state and a verb signifying the act or process of inducing this state in some object or person, e.g. black - blacken, broken - break, chilly
- chill,
displeasure
Norrick 1978b for a method of relating such pairs out side the lexicon). This same principle extends to phrasal units literally expressing effects with idiomatic or figurative inter pretations identifying their causes, e.g. dispatch to Hades for 'kill', cut the rug for 'dance', and those ex pressing causes with idiomatic or figurative interpretations identifying their effects, e.g. take pen in hand for 'write', kill the fatted calf for 'prepare a meal from the meat of the fatted calf'. These examples exhibit the mul tiple applicability of Principle 1. Productive applications can easily be constructed; we can say empty a glass (effect) to mean 'drink a glass' (cause), unlock the prisons (cause) to mean 'let the prisoners free' (effect), take a picture (effect) for 'release the shutter (of a camera)' (cause), write some invitations for 'invite some guests'(effect). Metonymic Principle 1 thus clearly makes up a part of any semantic theory aspiring to provide an adequate description of semantic competence. 3.2.2.2 Metonymic Principle
2
The second metonymie principle, based on Indexical Principle 2, relates semantic readings characterizing artifacts with those characterizing their producers. I state it as follows. Metonymie Principle 2 : x is in regular metonymie relation to y, if x and y are conceptualized as identifying an artifact and its producer respectively.(reversible) In the realm of lexical relations, the readings con ceptualized as artifacts will in fact be found among the lexical entries classified as (Artifact), at least re dundantly. The lexicon here envisioned will distinguish within the class of objects between natural and man-made objects. The former will be further specified as (Organic) (Animate) etc. In productive applications of Metonymie Principle 2 as well, we could stipulate that phrases identifying artifacts must contain the corresponding sem antic feature (Artifact). At the same time, we might ex plicitly demand the presence of the feature (Human) in the
88
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
reading identifying the producer. This would render a more explicit principle like: Metonymic Principle 2 (revised): x is in regular metonymic relation to y, if x is specified as (Human) and y as (Artifact), and x is conceptualized as identifying the producer of y. (reversible) But this degree of specificity is neither necessary nor desirable.' First, because something conceptualized as an artifact or producer will also be specified as such semantically under normal circumstances. Second, because we may want to recognize as regular a relation obtaining between some agent not explicitly identified as (Human) and its product, or between a human producer and something not characterized as (Artifact). If an ape produces a artifact; if a botanist develops a new variety of tree, again the producer-artifact should hold, although we might not want to specify the tree as (Artifact). As noted above, it was with an eye to such cases that I adopted Eco's (1976) use of the expression "conceptualized as," and I shall continue to employ it for just this the less specific manner of the unrevised version of Prin ciple 2. Like the previous principle, it is systematized in lexical relations between words built on a common stem: drama - dramatist, pot - potter, wheel - w h e e l r i g h t . Its multiple applicability is shown by these examples and its function in linking morphologically separate pairs such as bread - baker, book - author, house - carpenter. In addition, Principle 2 predicts the figurative interpreta tions of bread requires patience as 'a baker requires patience' and the novel needs imagination as 'the novelist needs imagination'. These examples testify to its pro ductivity as well. 3.2.2.3 Metonymie
Principle
3
The third metonymie principle, based upon Indexical Principle 3, correlates semantic representations ident ifying natural products with their natural sources.
reason
REGULAR SEMANTIC RELATIONS
89
Metonymic Principle 3: x is in regular metonymic relation with y, if x and y are conceptualized as identifying a natural product and its natural source respectively. (reversible) As I pointed out in the earlier treatment of the general semiotic relation between natural products and their sources, the systematization of the relation can be a natural product or its natural source: flower (bloom or plant), chicken (meat or animal), wood (lumber or forest) etc. In addition, there exist meaning relations between pairs like acorn-oak, which support the multiple appli cability of this principle. Since grape will be lexically linked to grape vine, grape juice to grape, and win e to grape juice, the multiple applicability of Principle 3 is guarenteed. We see its productivity in the figurative use of the grape for 'wine', rye for the appropriate whiskey or bread, and in the sentences like Harry has ten acres of sugar Judy grows a lot of beef on her ranch, where the names of products must be interpreted as designating their corresponding natural sources. Thus, Metonymie Principle 3 should also be included in our inventory of semantic re lational principles.
3.2.2.4 Metonymie
Principle
4
The fourth metonymie principle derives in the same manner from Indexical Principle 4. It defines the relation between lexical entries identifying products and those identifying the instruments crucially involved in their production. Metonymie Principle 4: x is in regular metonymie relation with y, if x is con ceptualized as identifying a product and y as an in strument or process crucial in the production of x. (reversible) This principle exhibits multiple applicability in various systems and on various levels. We find, first, morpho logically linked terms naming instruments or processes and their products, e.g. copier-copy, heater-heat, toastertoast. Similar relations hold between morphologically un connected items: oven -bread, drill-hole, weaving basket. We say turn up the heat meaning 'turn up the
obser
90
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
heater', turn down the bread, meaning 'turn down the oven the bread is in'. We see here the systematization of Prin ciple 4 as well. Its productivity manifests itself in the transfer from mimeograph (machine) as instrument to mimeos as its products, and the rapid acceptance of xerox as the term designating the product of a xerox machine. Thus, Principle 4 also belongs in our inventory of semantic re lational principles. 3.2.2.5 Metonymic Principle
5
Remaining within the general class of semantic prin ciples derivative from the basic indexical relation of cause-effect, I define principles connecting semantic representations identifying acts with those identifying their objects, instruments and agents. The first derives its semiotic motivation from Indexical Principle 5. Metonymic Principle 5: x is in regular metonymie relation with y, if x is conceptualized as identifying an act and y as its object.(reversible) This principle exhibits the full range of lexicalized connections. One finds phonological forms representing both a verb identifying an act and a noun designating the corresponding object: drink, light, nails there exist many morphologically linked verb-noun pairs manifesting the same semantic relation: draw-drawing, integrate-integral, s i g n - s i g n a t u r e ; in addition, we will want to recognize the act-object relation between otherwise distinct words: try-case
(in court) , run-race,
take-pill.
These illustrate
the systematization of the principle and indicate its multiplicity of application. Productive transfers from verbs describing acts to form nouns designating their ob jects are common, e.g. eats 'food' from eat, smokes 'cigarettes' from smoke. The reverse process, whereby a noun designating an object yields a verb identifying the act affecting it, yields the verb collage from the noun collage,
the verb litho
from the noun lithograph
etc.
Thus, Principle 5 enjoys a certain productivity as well.
REGULAR SEMANTIC RELATIONS 3.2.2.6 Metonymic Principle
91
6
The next metonymic principle relates representations identifying acts with those identifying instruments crucially involved in them. It derives from Indexical Prin ciple 6. Metonymie Principle 6: x is in regular metonymie relation with y, if x is con ceptualized as identifying an instrument and y as the act in which x finds crucial employment, (reversible) Again we find examples where a single sound sequence serves as both a noun naming a particular instrument and a verb identifying the act it is involved in: hammer, cover, filter same semantic relation are morphologically linked: houseVhouse N sit- seats open- opener3 distill— distillery. Names of instruments regularly transfer to form verbs identi fying their characteristic employment: we obtain the verbs knife, gun (down), ice (tea), xerox from homophonous nouns. We interpret the violin is difficult to mean the act of playing is difficult, and the gun is loud to mean the act of shooting. Principle 6 thus meets the relevant criteria and will be numbered among our regular metonymie principles.
3.2.2.7 Metonymie
Principle
7
Metonymie Principle 7 establishes relations between the representations of acts and those of their agents. In any language one finds pairs of words designating agents and their acts built upon a common morphological stem, e.g. baker-baking, bigot-bigotry, fool- foolishness, of course, other pairs will exhibit the same semantic relation with no morphological connection, e.g. tailor-sewing, carpenterbuilding , barb er- cutting hair, but these pairs too will necessarily share a number of semantic features in their lexical representations. As mentioned above, the baker needs flour may be intended to convey that acts of baking r require flour, and Harry s sewing is poor may convey that his tailor lacks skill. I state the relevant semantic re lational principle as: Metonymie Principle 7: x is in regular metonymic relation with y, if x is conceptualized as identifying an act and y as its agent, (reversible)
92
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Since this principle is systematized, enjoys multiple appli cability and can be used productively, I include it in the inventory of semantic relational principles. 3.2.2.8 Metonymic Principle
8
The semantic relational principle deriving from Indexical Principle 8 connects representations signifying instruments used in acts with those signifying the agents of such acts. Metonymic Principle 8: x is in regular metonymie relation with y, if x is con ceptualized as identifying an instrument employed in an act and y as the agent of the act. (reversible) This principle establishes a regular semantic relation between the representations of such items as farmer and plow, sailor and ship, etc. It relates the readings of such items as smithy,
trawler
and pile
driver
on which they
signify a forge, fishing boat and a machine respectively with the readings on which they signify the people using these things. It establishes the same semantic connection between morphologically related pairs like truckertruck, optician-optics, diver-diving oard. Such examples illus trate this principle's systematization. They also indicate its multiplicity of application, which is further shown by its functioning in the interpretation of figures such as the raiding party consisted of ten rifles to mean it con sisted of ten people with rifles, and idiomatic phrases like the pen is mightier than the sword to mean writers wield more power than those who fight with weapons. Its productivity appears in the construction and interpretation of sentences like the celli sit across from the violins, where the name of an instrument is understood to designate the person using it. 3.2.2.9 Metonymie
Principle
9
I turn how to metonymie principles deriving from indexical principles based upon the contiguous connection of a part with its whole. I first define a general semantic relational principle linking semantic representations characterizing any part of an object with those character izing their respective wholes.
93
REGULAR SEMANTIC RELATIONS Metonymic Principle 9:
x is in regular metonymic relation with y, if x and y are conceptualized as identifying a part and its whole respectively.(reversible) This principle accounts for such traditional figures as sail for 'ship', blade for 'sword', head for 'animal' (head of cattle). Again we find instances where a single sound sequence is associated with two regularly related semantic representations, in this case one identifying a part, the other the whole: ¿et (engine or airplane), diesel (engine or truck), hand (body part or employee: farm hand), bar (serving counter or building), peak (top or whole mountain). I mentioned above phrases like go to the bathroom
and go
to bed
with
someone
which literally state
only part of a complex act, leaving the rest unstated, thus providing useful euphemisms. Expressions such as (set of) wheels for 'car' and (set of) threads for 'clothes' indicate in the use of tails for a whole formal suit or butt for a whole cigarette. We can thus count Principle 9 among those in our inventory of semantic relational principles.
3.2.2.10 Metonymie
Principle
10
Similar to the preceding principle describing partwhole relations generally, I recognize two further semantic principles defining relations in the part-whole domain. The first relates the semantic representations of terms describing complex acts or events with those of terms describing particular acts or events crucial to their character or success. As pointed out above, the verb cook extends from 'heat in water' to a whole range of operations associated with preparing food, just as eat goes far beyond 'chew and swallow' to encompass a broad set of activities involved in serving, cutting, distributing por tions etc. Similarly, write identifies not only acts in volving pens and paper, but extends to include acts of authoring generally, even if the text the author himself produces was spoken into a dictaphone. Conversely, we describe an act as sewing a coat, even though acts of measuring, cutting, designing etc. are involved. As dis cussed above, since one may view one's employment simply as an integral part of the total complex of acts intended to provide sustenance for his family, we hear figuratively meant phrases like bring
the
home
the
bacon
and put
meat
on
tab le as references to this employment. In the same
94
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
way we say we are going to town to imply the whole set of acts of shopping, delivering, collecting etc. involved. The semantic principle responsible for establishing the relevant connections becomes the tenth in our inventory of metonymic principles. Metonymic Principle 10: x is in regular metonymie relation with y, if x is conceptualized as identifying a crucial act or pro cess in a complex of acts or processes y. (reversible) 3.2.2.11 Metonymic Principle
11
The second semantic principle we should recognize on the basis of the general part-whole relation derives from Indexical Principle 11. Metonymie Principle 11 : x is in regular metonymie relation with y, if x is conceptualized as identifying the central factor in an institution y. (reversible) This principle experiences systematization in the use of the stage or theater, the press, steely television for the institutions, professions or industries for which these entities are central. These words must be associated with two distinct semantic representations (to some extent con text dependent), one identifying an object, the other an institution, industry etc. Metonymie Principle 11 estab lishes the relation between the two representations. Similar examples are seen in the interpretation of the church to mean an institution, medicine to mean doctors, hospitals etc., recording and entertaining to mean the artist, industry etc., education to mean schools, uni versities, teachers etc. These examples illustrate the systematization and multiplicity of application of Prin ciple 11. We see its productivity in further extensions of names of objects to designate institutions of which they represent a central factor: we say I'm in clothing, food, f u r n i t u r e , toys to mean employed in the industry producing the products named. We understand statements like the tobacco industry is unhealthy to mean their products are unhealthy, paper is doing well to mean the paper industry is doing well. Metonymie Principle 10 hence fulfills our criteria and belongs in the inventory of semantic prin ciples.
95
REGULAR SEMANTIC RELATIONS 3.2.2.12
Metonymic
Principle
12
We now turn to principles in a third major category, viz. that governed by the container-content relation. The first, and basic principle I state as: Metonymic Principle 12: x is in regular metonymie relation with y, if x and y are conceptualized as identifying a container and its content respectively, (reversible) The lexicalization of this principle exhibits itself in the systematic duality of meaning associated with items such as teaspoon,
cup,
bottle,
bushel,
barrel,
which may
signify either containers proper or their volumes as quantities. We also standardly interpret words designating containers figuratively as meaning their contents in cer tain contexts: this bottle is sour and this package is rancid are anomalous on the literal level, and will be in terpreted to mean 'the liquid in the bottle is sour' and 'the contents of the package (presumably animal fat) are rancid' respectively. The relation from the literal to the figurative meanings receives description in Principle 12. Conversely, words designating liquids are interpreted to mean the containers holding them in certain contexts: the milk tipped over for 'the milk container tipped over' or Al removed
the
label
from
the
wine
for 'Al removed the
label from the wine container'. Principle 12 thus claims productivity. As noted above, it allows us to interpret terms literally identifying any object which could be used as a container as meaning the volume it comprehends. We can use statements like add
a thimble,
can,
bucket,
bathtub of water intending to indicate a particular volume. These examples illustrate the productivity and multiple application of Principle 12. Note too its function in allowing the interpretation of one bad apple spoils the whole barrel to mean the other apples in the barrel. As a special case of Principle 12 we should also recognize the use of expressions identifying time periods to designate people living in them. Here Principle 12 transfers from the spatial to the time dimension. The Eighteenth Century hated Shakespeare can thus be interpreted to mean that people living during the Eighteenth Century hated him.
96
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
3.2.2.13 Metonymic Principle
13
A second semantic principle governed by the basic con tainer-content relation connects representations ident ifying localities with those identifying their occupants. It will predict the relation between alternative readings of proper nouns designating either a particular building or the institution it houses, e.g. The White House (building or government of the U.S.A.), The Kremlin (building or government of the U.S.S.R.). It will account for the common use of names of cities to designate the governments located in them. This principle is thus systematized in some cases; its effects are further syste matized in phrases like the French tout le monde meaning 'everybody' and the English city hall meaning 'the local government' in you can't fight city hall. We can state the relevant semantic principle as: Metonymic Principle 13: x is in regular metonymie relation with y, if x and y are conceptualized as identifying a locality and its occupants respectively, (reversible) We see productive applications of this principle in sen tences like the whole town voted for Harry, where the town's inhabitants are intended, or the illiterate countryside particular group lives or meets in Harry's kitchen, we can interpret Harry's kitchen approved the proposal as per taining to this group. The locality-occupants principle thus clearly claims a place in our inventory of semantic relational principles.
3.2.2.14 Metonymie
Principle
14
A related principle based upon Indexical Principle 14 connects representations for names of costumes with those of their wearers. Metonymie Principle 14: x is in regular metonymie relation to y, if x and y are conceptualized as identifying a costume and its wearer respectively, (reversible) This principle's systematization shows itself in the lexicalization of two semantic representations, one
despise
REGULAR SEMANTIC RELATIONS
97
designating an article of raiment, the other the person who wears it, e.g. cassock (garment or clergyman), crown (headwear or monarch), pantaloon (pants or old man). Further, we find terms literally describing articles of apparel lexicalized for readings which identify people, e.g. red cap 'porter', b lack b e l t 'person with a specified degree of skill in karate'. Traditional poetic figures like helmet for 'soldier' and epaulette for 'officer' are interpreted with the help of Principle 14. Such examples illustrate its multiple applicability as well. Productive instances occur in sports reports like the red jerseys are Lower Slobovia and red gives green a left jab. We interpret Harry chases skirts to mean he chases women on the same basis, just as an employer might say, "I don't want any jeans and sweatshirts working here." Thus, Metonymic Principle 14 is regular in the crucial respects. 3.2.2.15 Metonymic Principle
15
From Indexical Principle 15 I derive a metonymie prin ciple relating semantic representations identifying measured phenomena with those identifying phenomena as ex perienced. Metonymie Principle 15: x is in regular metonymie relation with y, if x and y are conceptualized as identifying a phenomenon as observed and as measured respectively (reversible) This principle defines the regular relation between the re presentation for warm according to which it indicates a state as registered by a thermometer and that on which it indicates a state felt by a person. Thus, this sweater is warm is ambiguous: it can be correctly maintained of a thin summer sweater lying on an oven or of a thick winter sweater lying in a refrigerator. All adjectives identifying temperatures manifest this meaning duality. Similarly, nouns identifying seasons or times of day exhibit dual re presentations reflecting the measured time versus felt time distinction predicted by Principle 15. Summer can designate either a time of year based upon the calender or a set of environmental conditions. Thus, it is summer in Sicily may be said in July with respect to the calendar or in October with respect to the weather, vegetation etc. If a friend arrives and finds me breakfasting in my bath robe at three P.M. , I might say, "It's still morning here," where morning indicates not measured clock time but a set
98
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
of acts associated with having been awake a short time. Even adjectives like heavy and lights which predicate degrees of weight, exhibit the measured-felt duality of meaning. Bulky objects are judged heavier than compact ones of the same measured weight, so heavy will be lexicalized for the reading 'registering a high value on a scale' and 'difficult to carry or lift'. Old and new are similarly indeterminate in pertaining to the period of time a thing has existed or the period of time a person has been involved with it. An old purse may be a new possession; a new friend may be an old person. These examples illus trate both the systematization and the multiplicity of application Principle 15 claims. It is productive in form ing new extensions from measurements to feelings and vice versa. Winter comes early in the Alps can be spoken intend ing to maintain that it gets cold there before in other places. We interpret without difficulty a statement such as everything is light here, as spoken by an astronaut on the moon, to mean 'easy to lift' rather than as entailing a loss of mass or weight as measured on a scale near the surface of the earth. So we can include Metonymic Prin ciple 15 in our inventory of semantic relational prin ciples. 3.2.2.16 Metonymic Principle
16
The next semantic principle to be proposed derives from Indexical Principle 16; it defines the relation be tween semantic representations identifying phenomena as they present themselves to direct observation and as they are conventionally understood. Metonymic Principle 16: x is in regular metonymie relation with y, if x and are conceptualized as identifying the manifestation of a phenomenon and the conventional definition of it respectively. (reversible) This principle defines the relation between alternative readings of words like angler, buyer, reader, skater, writer, which designate either a person observably engaged in the activity in question or a person by profession or habit engaged in the activity. In the first case, the identification results from direct observation; in the second case, it is based upon a conventional definition or image of the person in question. Similarly, adjectives such as strong, weak, fast, slow can be interpreted to
REGULAR SEMANTIC RELATIONS
99
mean properties manifested or presupposed of certain ob jects. We call a person strong either because he is ex hibiting strength at a particular time or because we as sociate him with strength as part of our total image of him. We apply the predicate fast to an automobile if we see it pass by rapidly or if we imagine it to be a fast car, perhaps because it is a race car or produced by a company we expect to produce fast cars, even if the car we so envision remains parked in a garage. Principle 16 also relates emotions with their manifestations: an evil smile is a smile we assume reveals evilness in the smiler. Further, nouns like r i v e r , tree, cloud naming objects which visibly change through time must be associated with two readings, one on which they designate a stage of an object, one on which they designate the object as con ventionally defined. We have seen the systematization and multiple applications of Principle 16. Its productivity appears in the extension of items such as a g i l e , cheerful kind to both observed and conventionally presumed cases. Thus, Harry's happy can be interpreted to mean he is ob servably happy at the moment or that he is considered happy generally. Principle 16 will also apply productively in cases where a whole phrase manifests the actual-potential ambiguity, like a w a r r i o r , in an athletic manner, with great gusto and similar phrases receive both actual and potential interpretations. We can thus include Metonymic Principle 16 among the regular semantic relational prin ciples in our inventory. 3.2.2.17
Metonymic
Principle
17
The following principle, derived from Indexical Prin ciple 17, relates semantic representations for items de scribing people with those for items describing their possessions. Metonymie Principle 17: x is in regular metonymie relation with y, if x and y are conceptualized as a possession and its owner re spectively (reversible) Names of ethnic or national groups are generally morpho logically related to those of their language or customs. Apparently all languages can designate an object as some one's possession by appending to its name or description some form of the owner's names, perhaps along with cer-
100
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
tain other morphemes. These facts illustrate the systematization of Principle 17. Its multiple applicability appears in our use of people's names directly for their possessions; pointing to a pile of coats, say, to indicate which belongs to whom, we say simply, "That's Harry, that's Judy, that's Ann" etc. The connection obtains on different levels, since we can refer to articles of cloth ing as belonging to individuals, couples, families, organizations, nations etc. The productivity of Principle 17 is seen each time we understand a person's name to designate one of his possessions and vice versa; this is Judy meaning 'this is Judy's gun', the long straw starts meaning 'the person with the long straw starts'. 3.2.2.18 Metonymic Principle
18
The final metonymic principle I wish to discuss de rives from Indexical Principle 18, and would define the relation between semantic representations identifying people and those identifying the offices or roles they occupy. However, while the indexical principle was seen to perform a necessary function in semiotic theory, the corresponding semantic principle appears to be unnecessary, This is so, first, because the theory of reference pre dicts that a single person or object can be referred to with a wide range of different names, descriptions, anaphoric and exophoric pronouns, and even gestures. In conjunction with Indexical Principle 18, it predicts that one can refer to a person by identifying his role and con versely. Terms like the director and my neighb or can be used to refer to appropriate people, and the speaker can go on to say something about them as people or in the capacity of director or neighbor: Second, the distinction between the president serves a four year term in which the office is intended and the president jogs where the person is intended can be treated by Metaphoric Principle 2, that describing the specific-generic relation. Hence, reference theory and the specific-generic distinction ex haust the range of examples which a semantic principle based upon Metonymie Principle 18 would be required to describe, so that we can close the inventory of metonymie principles without adding such a principle. This ends our discussion of the statements making up the inventory of regular semantic relational principles. Let us now turn to the task of showing how this inventory of principles functions in semantic theory as a whole.
101
REGULAR SEMANTIC RELATIONS 3.3 Regular
Relational
Principles
in Semantic
Theory
In the present section I want to outline the role re gular relational principles play in semantic theory. First, however, some comments are in order on semantic theory in general as it is to be understood here. 3.3.1 Comments on Semantic
Theory
Semantic theory "concerns the semantic structure of natural language in general. It is not a theory concerning the structure of any particular natural language or lan guages to the exclusion of others" (Katz 19 72: 11). I distinguish semantic theory from the semantic component of the description of any language. The semantic relational principles developed above are to be included in semantic theory proper; they need not be stated as part of any particular semantic component. So I have not been overly concerned with arriving at too narrowly defined a set of principles. Individual, language specific relational prin ciples can be stated non-redundantly in the description of the language in question. I was thus concerned with showing the relational principles to be of general application or at least to be derived from generally valid principles of sign motivation. Semantic theory concerns the elements of linguistic meaning and their interrelations. There is no a priori reason why semantic theory should connect linguistic mean ings with meanings in any other non-linguistic system; it need not provide translations of linguistic meanings into formulas of predicate calculus or into particular behav ioral patterns. We do not demand that theories of music, architecture or mathematical symbolism be transformable into one another or some other system. Musical elements and patterns can be analysed in terms of their relations to other musical elements and patterns without reference to the physical properties of vibrating bodies or the struc ture of the human ear, although theories of their relation to the latter are certainly of interest. In the same way, semantic theory can analyze relations within linguistic meaning systems without the aid of any logical calculus, although, again, a theory with such relations as its goal may reveal something about linguistic meaning systems. I therefore favor a feature approach over a truthfunctional theory of meaning. As Harman (1977) argues, the truth-functional theory of meaning is about truth, not about linguistic meaning; linguistic meaning makes possible and explicates truth, not conversely (cf. Katz 1977b).
102
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Wunderlich (1971), from the point of view of pragmatics, also argues the insufficiency of a simply true or false characterization of sentences. Even Lewis (1972: 14), who generally favors the truth-functional approach, notes the need for a fine (vs. coarse) meaning analysis to handle semantic differences among analytic sentences and to determine synonymy. Especially in the realm of lyric poetry, but for figurative meaning generally, truth-functional semantics lacks validity, while feature analysis remains necessary, as van Dijk (1970) insists. In showing that the literary text creates its own reality rather than imitating a given reality, Schmidt (1970: 66 ff. ) argues the irrel evance of truth and the importance of feature semantics, although he later (Schmidt 1976: 71) stresses the purely heuristic value of features, which he views as theoretical constructs. This relativization is due primarily to psy chological arguments of the sort advanced by Fodor, Fodor and Garrett (1975), arguments I feel are convincingly countered by Katz (1977a). I consequently agree with Kastovsky (1980: 6) that feature semantics remains the most promising {aussichtsreichste) alternative semantic theory offers at present. "Semantic theory must also contain a model of the sem antic component of a grammar which must describe the manner in which semantic interpretations are mapped onto" syntac tic structures (Katz 1972:33). This includes a model of the lexicon, rules for relating lexical elements with elements in syntactic structures, rules correlating lexi cally interpreted syntactic structures with semantic inter pretations, and rules relating one semantic interpretation with the other. The relational principles I propose as part of semantic theory affect each of these four functions of semantic theory in ways to be described below. The semantic theory to be presented below orients it self to a great extent on that developed in Katz (1972; cf. Katz and Fodor 1963; Katz and Postal 1964). There exist good reasons for this orientation. First, Katz's theory should be familiar to readers. Second, it is fairly com plete and quite explicit in many points. Third, much or even most recent work on various aspects of semantic theory is cast in Katzian terms and makes explicit reference to Katz's proposals. Many of the proposals argued for below can be viewed as additions to Katz's theory. Others re present genuine revisions of the Katzian framework, but will be compared with it and presented in its terminology as far as possible for the sake of clarity and compar ability.
REGULAR SEMANTIC RELATIONS 3.3.2
The Role
of Semantic
Relational
103
Principles
Semantic relational principles describe connections between semantic representations of lexical units, viz. not just simply items, but compounds, stock phrases, colloca tions, idioms, on the one hand, and between lexical re presentations and semantic interpretations, on the other. Since they can link a single lexically interpreted syn tactic structure with multiple semantic interpretations, they also capture relations between different semantic interpretations in some cases. The non-lexical or productive functions of relational principles can be roughly described as follows. In an in terpretive theory in which lexical items are inserted into syntactic structures by context free rules, lexical inter pretation (cf. Katz 1972) consists in paring symbols in a structure representing the syntactic relations between them (Katz's 'underlying phrase marker') with readings from the lexical entries for these symbols. Then the pro jection rule applies combining lexical readings into derived readings until they arrive at what we want to call the semantic interpretation for the whole structure. Assuming all the lexical symbols in the structure are assigned but a single reading in the lexicon, if all their readings fit together without clashes during applications of the projection rule, we obtain a single consistent de rived reading (or semantic interpretation) for the struc ture. If a feature clash occurs at some point, the deri vation blocks and the structure is labeled anomalous. In a theory enriched by the addition of regular re lational principles, applications of the projection rule will be optionally preceded in each case by application of any relational principle appropriate. This procedure enab les the theory to describe figurative meanings. If a sem antic derivation blocks at some point, relational prin ciples are tested in turn to yield a consistent inter pretation if possible. If one or more paths block in spite of the application of relational principles, the structure is labeled anomalous. In psychological terms, the addition of regular relational principles to semantic theory re flects our tendency to seek consistent interpretations for structures, no matter how anomalous they appear at first. Consider the following examples. In Katz's theory, sen tence (1) will be labeled anomalous, since milk bears the feature (Liquid), while tip over requires a subject nominal containing the feature (Solid). (1) The milk tipped over. Examples like this reveal a serious weakness in Katz's
104
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
theory, since speakers consistently and readily accept this sentence as fully correct2. In the theory proposed here, (1) will not be labeled anomalous. Instead, semantic relational principles from the inventory defined above are tested for application to both of the terms involved in the clash of features; if one functions, the projection rule reapplies. This procedure is repeated until a con sistent reading emerges or the inventory of relational principles has been exhausted. One relational principle which clearly applies in the case of sentence (1) is Metonymic Principle 12, which re lates semantic representations identifying containers with those identifying their contents. By this principle, the lexical reading for milk containing the feature (Liquid) is regularly related with a representation characterized as (Container of Liquid) which presupposes the feature (Solid) required to satisfy the selectional properties of tip over. We thus obtain a consistent reading for (1) equivalent to the literal reading assigned (2). (2) The container of milk tipped over. The relation established by Metonymic Principle 12 allows the substitution of (Container of X) for any semantic re presentation X identifying something we conceptualize as the content of some container. It allows us to derive the sort of semantic interpretation for sentences like (1) which speakers consider standard. Within the Katzian framework, only one interpretation for sentence (3) can be derived, viz. that on which the area Sue lives in consists entirely of white objects. (3) Sue's neighborhood is all white. The lexical reading for white on which it means 'Caucasian' is unavailable since it bears the feature (Human), which would clash with (Non-human) in the reading for neigh borhood. Given our Metonymie Principle 13, however, we can derive the interpretation speakers consider the most likely, viz. that on which Sue's neighborhood is inhabited only by people of the Caucasian race. Metonymie Principle 13 relates semantic representations characterized as (Locality) with those characterized as (Human), if the latter are conceptualized as the occupants of the former. These two examples show that the inclusion of relational principles in semantic theory enable it to come considerably closer to the goal of describing the full range of semantic interpretations speakers typically associate with sentences.
REGULAR SEMANTIC RELATIONS
105
Failure to include something like regular relational principles as an integral part of semantic theory results in a theory which blocks clearly normal interpretations, and labels as anomalous perfectly acceptable sentences. Both Leech (1969, 1974) and Gak (1976) propose semantic rules with essentially the same effects as the relational principles advocated here. But neither of these writers attempts to link the rules to lexical relations which mir ror the connection between the input and output of such rules. The present study seeks to capture this connection by hooking up regular relational principles to the lexicon in the manner described below. Semantic relational principles are significant for the structure of the lexicon in several ways. First, as I pointed out above in the discussions of metaphoric prin ciples, some of these support the view that the lexicon should be structured around semantic classes. Within the class of physical objects, we recognize the class of inorganic objects, within this class that of artifacts etc. Membership in the class of artifacts automatically entails being characterized by the feature (Artifact) and, redun dantly, by (Inorganic) (Object) (Physical). From a lexicon structured around semantic fields one can read off syn onymy, hyponymy and antonymy relations. Other lexical re- . lations predictable on the basis of regular relational principles must be expressed in some other fashion. For this purpose I introduce the notion of the semantic via-rule, which functions in semantic theory in essentially the same manner as phonological via-rules (first proposed by Vennemann 19 72; cf. Hooper 1976) in phonological theory (cf. Norrick 1979a, 1979d). Semantic via-rules connect lexical entries not included in or adjacent to each other. A semantic via-rule attached to one lexical entry identi fies another lexical entry regularly related to it and the relational principle from the inventory which estab lishes their connection. Via-rules refer to the same set of relational principles which predict productive trans fers and extensions of meaning. They thereby provide the required connection between productive principles external to the lexicon and lexicalized connections. The principle involved in the creation of a metaphor is cited as a part of the lexical entry for the figurative reading if it be comes lexicalized. Consider as an example the word tub . We can assume that tub is associated with a lexical reading containing the features (Artifact) (Container) (Open) (Round). Tub also occurs in figurative contexts such as Harry is a tub meaning Harry is round in shape, i.e. severely overweight.
106
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
The connection between the two interpretations is estab lished by Metaphoric Principle 4; Semantic Transformation 2 removes all features from the lexical reading for tub except (Round). Acceptance of this figure leads to a lexicalization of a reading for tub on which it designates a round or obese person. Unfortunately I have no method of predicting that tub will become lexicalized on the parti cular reading where it means 'round person', and no re gular relational principle in our inventory would charac terize the extension from 'round' to 'round person' as regular. Nevertheless, we can label the lexical reading for tub on which it means 'round person' as partially derivative from and regularly related to the original entry for tub where it designates a round container. The device I propose for accomplishing this is the sem antic via-rule. I attach to the lexical entry for tub 'round person' the via-rule: Via-rule I: Related to tub 1.'round container' via Metaphoric Prin ciple 4 and Semantic Transformation 2r which deletes all but the feature (Round) from tub 1. In this case the via-rule relates a lexicalized figurative reading with the reading from which it was derived. Since it defines a regular relation between two readings associ ated with a single phonological form, we obtain a. case of polysemy. Any time two distinct lexical entries for a single phonological form are linked by a via-rule, we can say they are alternate readings of a polysemous lexical item. Alternate readings for an item cannot be listed to gether in the lexicon, since it is structured in terms of semantic classes rather than phonological identity of entries. In addition to linking alternate readings of poly semous lexical items, via-rules hold between other pairs of words with common semantic content, when this content is redundant. They thus function as redundancy rules. No matter how the verb dance is described, the noun dancer will contain the same information as part of its reading. We can think of dancer as meaning 'one who dances pro fessionally'. To avoid restating the definition of dance as part of the reading for dancer, we could include in the latter a statement referring to the former. Or, as Jackendoff (1975) points out, we can go ahead and fill in all the redundant information, but then characterize it as such. I propose for this purpose adding to the entry for dancer a via-rule such as:
REGULAR SEMANTIC RELATIONS
107
Via-rule 2 : Related to dance V via Metonymic Principle 7 (con necting readings identifying acts with those ident ifying their agents). This via-rule characterizes the entire reading for dancer as redundant given its regular relation to the verb dance Via-rules will also define relations between morphological ly unconnected items. The lexical reading for author will contain that for book, since the former means essentially 'person who writes books'. Here too, for the same reasons outlined for dancer and dance' the reading for author should bear a via-rule citing its relation to that for book, in this case due to Metonymic Principle 2 (con necting readings identifying artifacts with those ident ifying their producers). We can thus say that author and book stand in the same semantic relation to each other as potter and p o t ; the former pair shows a purely semantic re lation, the latter pair a morphological relation in addi tion. Via-rules further describe cases of motivation in compound items. The reading for an item such as dining room tab le will bear via-rules relating it to the entries for table and for dining room, just as the latter will contain via-rules relating it to the entries for dine and room. All the non-idiosyncratic semantic information re quired in the reading for dining room tab le can be charac terized as redundant on the basis of the via-rules it manifests. Since each of its components retains its stan dard lexical reading, we shall say that this compound enjoys full componential motivation. Idiomatic stock phrases may also exhibit componential motivation. While the meaning of keep tabs on cannot be derived from the meanings of its components, keep appears to retain its complete lexical meaning of 'maintain' even in the context of the idiom reading 'maintain under sur veillance' (cf. keep smiling, keep in touch etc.). The via-rule attached to this idiom must therefore take a form like: Related to keep 'maintain' directly. This is a truncated sort of via-rule, of course, since it con tains no via-component at all, but I prefer to count it as a via-rule in spite of this fact. In addition to componential motivation of the types illustrated, stock phrases may manifest composite moti vation. i.e. the idiomatic meaning associated with a stock phrase may be regularly related with its literal meaning taken as a whole. The literal interpretation of the stock phrase b low off steam describes a special case of the more general act of releasing pressure. If pressure includes
108
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
emotional pressure, then the idiomatic reading for blow off steam stands in regular metaphoric relation to its literal interpretation by Metaphoric Principle 3, which links concrete examples of events to the events as general propositions. To summarize, we have seen that via-rules establish connections between lexicalized figurative readings for items and their (historically) original readings, thus relating alternate readings for polysemous lexical items. They also define relations between (redundant portions of) morphologically unconnected lexical entries. In addition, via-rules express regular relations between morphologi cally complex items and their component parts, as well as between literal and figurative interpretations for idio matic stock phrases. These last allow us to capture two types of motivation in morphologically complex lexicalized units. The device of the via-rule allows us to establish the psychologically real connection between creative pro cesses of meaning extension and transfer, on the one hand, and static lexical relations, on the other. The semantic relational principles defined in this chapter were seen to function as active rules in the process of interpreting figurative constructions as well. I shall elaborate on the function and interpretation of the inventory of re gular semantic relational principles in the context of the lexicon, stock phrases, literary figurative language, and semantic interpretation generally in ensuing chapters. For the present it suffices to have outlined the conse quences of including an inventory of regular semantic re lational principles in semantic theory.
THE PROBLEM OF POLYSEMY
4. 4.0
Introduction
This chapter concerns itself with polysemy, i.e. regu lar meaning multiplicity in lexical items. I intend all my comments on polysemy to refer only to cases where a single speaker views a particular item as bearing two different readings. As we shall see, an item polysemous for one speaker may be univocal for another for various reasons. A statement that some word is polysemous in a language can only mean that many, perhaps a majority of, speakers view it as such. Hence also the irrelevance of diachronic ap proaches in the matter of polysemy. Polysemy in lexical items takes two forms: either (1) a lexical item includes two distinct lexical entries related by a semantic viarule, or (2) it has a lexical entry which undergoes a sem antic change in some context(s) due to the productive ap plication of a regular relational principle. The former does not exclude the latter: individual entries associated with a single polysemous lexical item may also be affected by relational principles in their contextual occurrences. The former case I label 'lexical polysemy', the latter 'derived polysemy'. The two cases have often been separ ated; I deliberately term both 'polysemy' to underscore their close connection. After narrowing and defining the notion of polysemy, I proceed to argue that lexical and derived polysemy should be treated as a unified phenom enon. I then review and criticize current thought on poly semy of both lexical and derived varieties. 4.1 Defining
'Polysemy
'
In this section I draw a number of distinctions aimed at clarifying the notion of polysemy. We can tentatively accept a definition of a polysemous lexical item as one associated with multiple distinct but regularly related readings. Since I have already outlined a theory of
110
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
regular semantic relations, ray primary concern in the fol lowing will be with determining what counts as multipli city of meaning and when two meanings are distinct.
4.1.1 Polysemy
and
Ambiguity
It appears to be standard practice among linguists, but by no means among philosophers, to differentiate poly semy (or homophony) and ambiguity as follows. Polysemy is one type of meaning multiplicity in lexical items (the other being homophony), whereas ambiguity is meaning multi plicity in sentences (cf. Kooij 1971; Lyons 1977). Poly semy is common to all natural languages as a result of meaning extension and transfer, and of what Black (1952:
156) terms the "spread of meaning in linguistic signs," viz. that "things related tend to acquire identical names" (Black's italics). In the present theory, polysemy can be handled by a single set of relational principles outside the lexicon and via-rules within the lexicon. Ambiguity, by contrast, must be treated on a number of different levels. One obvious source of ambiguity, of course, lies in polysemy. Unless the sentential context filters out all but a single reading for a polysemous lexical item, its multiplicity of meaning will extend to the sentence as a whole. If the multiple readings of a lexical item exhibit no regular semantic relation, the lexical item is said to be homophonous rather than polysemous. Homophonous lexical items can cause sentential ambiguity just as readily as polysemous lexical items. The ambiguity of a sentence may also be syntactic in origin. The literature on transformational syntax contains many interesting examples of syntactically ambiguous sen tences. By recognizing two distinct deep structures and two transformational histories for sentences such as (1), transformational grammar can account for their ambiguity. (1) Flying planes can be dangerous. At the text or discourse level, sentences are often ambiguous due to indeterminacy in the reference of ana phoric pronouns. The second sentence in (2) allows two different interpretations, depending on whether he refers back to Harry or Fred. (2) Harry screamed at Fred. Then he left.
111
THE PROBLEM OF POLYSEMY
Finally, the source of ambiguity in an utterance may appear at the pragmatic level of analysis. This is the case when, for instance, a hearer is uncertain about the intended function of a speaker's utterance. Under certain circumstances a speaker may utter (3) as a real request for information or as an indication that he would like the window closed. (3) Is there a draft coming from the window? In the following, the term 'ambiguity' will be applied exclusively to sentences or utterances, while 'polysemy' will be reserved for lexical items. In general, we will be concerned with ambiguity only insofar as it has its source in the polysemy of a particular component lexical item. 4.1.2 Polysemy
versus
Vagueness,
Generality
and
Depletion
Polysemy must be carefully distinguished from vague ness. According to Chao (1959) a lexical item counts as vague rather than polysemous insofar as its range of appli cation is not precisely determined, insofar as borderline cases loom large in comparison with clear cases of its application. Color terms are by their very nature vague, but not for this reason polysemous. Orange designates an ill-defined portion of the visible spectrum rather than a number of different colors. Sorrow designates a wide range of emotional states to a degree overlapping with those de signated by sadnessr grief, regret etc., but its entire range of application can be predicted by a single lexical entry, so that it exhibits vagueness rather than polysemy. See also Lakoff (1970b) on vagueness and ambiguity. Chao (1959) also draws a distinction between polysemy and generality of meaning. A term counts as general rather than polysemous when it is applicable to any one of a num ber of objects whose differences are regarded as irrele vant in the particular context. In particular we should note that the existence of two or more words in one lan guage which together cover a concept exhausted by a single word in some other language does not Indicate that the single word in the latter is polysemous, but only that it is more general than either of the words in the former. The fact that German distinguishes Wand and Mauer where English has but a single word wall may make It difficult for English speakers to grasp the difference between the German pair or cause German speakers to feel English lacks an important distinction, but it does not argue for the
112
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
polysemy of wall . Instead, it simply shows that w a l l pos sesses more generality than either of the German words. Certainly, the normal speaker of English never misunder stands a sentence by interpreting wall in the sense of Mauer rather than in that of Wand. By the same token, a generally applicable item such as animal must not be con sidered polysemous on the basis that it can designate mon keys, elephants, dogs etc. Most common words identify a range of objects which can be further specified, either by means of some other word or by adding adjectivals to it. Thus, monkey Is itself general with respect to chimpanzee , and chimpanzee with respect to adult chimpanzee, to adult albino chimpanzee and so on without end. Weinreich (1963; cf. Zgusta 1971) differentiates poly semy from depletion. A lexical item becomes semantically depleted through extension to a wide range of contexts and/or occurrence in many collocations. Lexical items such as do, put and take exhibit severe depletion. We do work, the wash, arithmetic, time (in jail), good deeds, wood carvings and myriad other activities and objects. We put out animals, candles, newspapers, put on clothes and people, put off work and people, put up with people and states of affairs, in addition to putting things somewhere and putting the shot. Even if many of the occurrences of do and put are classified as idiomatic, we still have a very broad range of meaning in each case. Nevertheless, it would be a mistake to associate items such as do, put, take with multiple lexical readings (except those they occur in idiomatically, i.e. where their semantic contri bution to the whole remains unpredictable despite the con tent offered). As Zgusta (1971) points out, depleted items differ from polysemous items as follows. In cases of de pletion, the context specifies or adds to the meaning of the item, whereas, in cases of polysemy, the context elim inates unintended readings. Since depleted items make so little total semantic contribution to the constructions they appear in, it would be well nigh impossible to prove that they are associated with truly distinct readings in different contexts. Hence, depleted lexical items should not count as polysemous. See also Katz and Postal (1964) on feature transfer to depleted items like thing and pro nouns. 4.1.3 Testing
for
Multiplicity
of
Meaning
It should be obvious from the foregoing discussion that we require some objective means of distinguishing
THE PROBLEM OF POLYSEMY
113
cases of polysemy, or multiplicity of meaning in semantic representations generally, from cases of vagueness, gener ality and depletion of meaning. Polysemy must be set off from vagueness, generality and depletion by requiring truly distinct meanings for the former. On the other end of the spectrum, as indicated above, polysemy will be dif ferentiated from fortuitous homophony by requiring that the multiple meanings associated with polysemous items be regularly related by general semantic principles. The most promising approach to testing for distinct ness in meanings associated with a single item was sug gested by Quine (1960: 131). We can term the test Quine proposes the truth functional test for multiplicity of meaning. It states that a lexical item has distinct mul tiple meanings if and only if it can be clearly true and clearly false of the same object from one situation to the next. If one maintains (4) of a thin summer jacket lying on an oven, it will be true on the Interpretation of warm where it designates a measured temperature signi ficantly above that of the ambient, but false if warm is interpreted causally to mean the jacket keeps its wearer warm (cf. Metonymic Principle 15: experience-convention). (4)
That jacket is warm.
It may be necessary to recognize a third value 'non sense' in addition to 'true' and 'false' for purposes of the truth functional test. Sentence (5) is either true or false, depending upon Harry's emotional state. (5)
Harry is sad.
Sentence (6) will also be assigned the value 'true' or 'false', depending upon the particular book, if sad is interpreted to mean 'evoking sadness In a reader', i.e. in an experiencer (cf. Metonymic Principle 1 : cause-effect). (6)
This book is sad.
So it seems sad possesses two distinct readings; but it is difficult to find a single sentence where both are possible but would entail opposite truth-values. Perhaps (7) would fill this requirement. (7)
Harry is a sad author.
(7) might be considered true on the reading for sad where it predicates a particular emotional state of Harry, and
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
114
false where it is interpreted to mean Harry evokes sad ness in those who judge him as an author. Nevertheless, if we accept 'nonsense' as a third truth-value, a single sentence such as (6) will serve to illustrate the polysemy of sad, since it might be true or false on the reading for sad where it means 'evoking sadness in an experiencer' and must be assigned the value 'nonsense' where it means '(the emotional state) sadness', inasmuch as books cannot be reasonably conceptualized as experiencing emotional states. Another test for multiplicity of meaning is suggested by McCawley (1968a), albeit for somewhat different reasons. We dub this the deletion test for reasons which will be come apparent directly. Since deletion generally requires non-distinctness of lexical items1 , deleting one item under phonological identity with another which differs in meaning will result in anomaly. The traditional term for the resultant sort of anomaly was 'zeugma'. The anomaly, or zeugma, of (8) derives from the deletion of the 'second occurrence' of sad, which as we saw immediately above, differs semantically from the first 2 . (8)
Harry is as sad as the book he authored.
(8) can therefore be used to illustrate the multiplicity of meaning in the word sad. Comparative constructions like (8) function well for testing verbs and adjectives, but they become cumbersome when used for nouns. McCawley (196 8a) notes that words such as dissertation, novel, score etc., which designate scientific or artistic mental constructs, can also iden tify the physical manifestations of these (cf. Gruber 1976 and discussion of Indexical Principle 12). In order to de termine whether these readings amount to distinct meanings rather than simply vague variants, one must form an ex ample such as (9). (9)
Judy's dissertation is more thought provoking than yellowed with age.
Consequently, we may also employ simple coordinate constructions with and, which, if they do not involve de letion as such, at least behave in the relevant respects as if they did. In order to avoid the assumption that de letion is involved in such constructions, we can call this the coordination test. So instead of (9), we form a coor dinate conjunction as in (10), which exhibits the same anomaly, if not quite so strikingly.
THE PROBLEM OF POLYSEMY (10)
115
Judy's dissertation is thought provoking and yellowed with age.
Notice that we cannot require that all conjunctions in volving two distinct meanings of polysemous items be anom alous; a number of pragmatic factors conspire to render certain examples better than others or even quite accept able (cf. Nunberg 1978). Apparently (11) counts as fully acceptable. (11)
Judy's dissertation is still thought provoking though yellowed with age.
In doubtful cases, it seems best to rely upon the truth functional test. Alston (1971) proposes the substitution test in order to determine whether a word has multiple meanings. If one occurrence of a word can be replaced by some other word, while another occurrence cannot, then the word has two distinct readings. Consider the following sentences. (12) (13)
Winter comes early in the Alps. Winter begins on the shortest day of the year.
For the occurrence of winter (12) we can substitute salva veritate an expression like cold weather and snow, but not in (13). Sentence (13) is true by definition with winter, but contingent with cold weather and snow, in fact it will usually be false with this substituted phrase. Since the two separate meanings for winter are predictable by Meto nymic Principle 16 (manifestation-definition), it is poly semous. Like Quine's truth functional test, the substitu tion test appeals to judgments about truth-values of sen tences. It essentially just procedurizes the search for readings with alternant truth-values, which characterizes Quine's truth functional test. We might therefore just as well recognize both tests as valid for determining the multiplicity of word meaning; then we can define distinct ness of meaning as association with two or more readings capable of causing a change in the truth-value assigned a sentence containing the item in question. Polysemy thus re quires distinctness of meaning in this truth functional sense along with a regular semantic relation between the distinct readings as defined by our battery of regular re lational principles.
116
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
4.1.4 Polysemy 4.1.4.1
and
Homophony,
Homophony Homography,
Homonymy
In the present study the term 'homophony' designates the relation between two lexical entries associated with a single phonological representation and distinct semantic representations unrelated by any regular relational prin ciple in our inventory. It is complementary to (in Lyons' 19 77 sense) polysemy, not a hyponym of it. Homophony con cerns only lexical entries with identical phonological re presentations; words spelled the same but pronounced dif ferently play no role in this investigation. This latter relation, holding between, for instance wind /waynd/ 'turn' and wind /wInd/ 'air movement', is usually labeled 'homography' and need not interest us here. Both polysemy and homophony presuppose phonological identity. Homonymy is sometimes construed as synonymous with homophony; it sometimes serves as a cover term for cases of both homophony and homography. I avoid the term 'homo nymy' in this study all-together in order to prevent con fusion. I will be concerned exclusively with relations holding between lexical entries manifesting identical phonological representations, viz. with homophony and polysemy. 4.1.4.2 Polysemy
versus
Homophony
As outlined above, I want to recognize two types of polysemy: lexical and derived. A relation of lexical poly semy obtains when two or more lexical entries exhibiting the same phonological representations also bear via-rules connecting their semantic representations. By contrast, a relation of homophony obtains when two or more lexical en tries exhibiting the same phonological representation bear no via-rules connecting their semantic representations. Historically, cases of lexical polysemy arise in one of two different manners. (1) A particular lexical entry undergoes metaphoric extension or metonymic transfer and the resulting figurative interpretation, perhaps along with an additional feature relevant in its customary fig urative context, becomes lexicalized in an independent reading. Pertinent examples are found in dead metaphors of all kinds, e.g. the extension of mouth, leg, eye etc. from the class of human body parts to various classes of physi cal objects: bottle (mouth), table (leg), needle (eye).
THE PROBLEM OF POLYSEMY
117
See Stern (1931), Ullmann (1957, 1967) among others for myriad examples of such extensions, and discussion of sem antic change generally. (2) The phonological representations for two semantically linked entries fall together as a result of a sound change affecting one or both the entries. This process re sults in polysemy much less frequently than the foregoing one, of course. But the two events may reverse their order with the same result. That is, after the phonological re presentations of two items have fallen together, speakers may begin to recognize a relation between their semantic representations where none was noticed before, due to an observable tendency on the part of people to link histori cally unconnected meanings once their phonological forms converge or unite. This tendency results in what has traditionally been termed 'popular etymology'. On the basis of the knowledge that names of human body parts tend to extend into other areas of vocabulary, many speakers consider ear (of corn) to be derived from and thereby regularly related to ear (human), following Palmer (1976: 68). The two fell together phonologically when Middle English ere (from Old English eare) lost its final vowel, and ceased to differ from er (from Old English ær) . In another example, the process of clipping (cf. Mar chand 1969) results in phonological identity for tar (from tarpauline with the meaning 'sailor') and tar 'pitch'. On the basis of the resultant phonological identity, some speakers relate the 'sailor' reading of tar with the 'pitch' reading on the basis of Metonymic Principle 8 (instrumentagent) and the fact that sailors once waterproofed wooden ship hulls with pitch. Of course, recognition of a pre viously unnoticed semantic connection between two lexical entries with identical phonological representations need not necessarily lead to popular etymology. One might draw the etymologically correct link from tar 'sailor' to tar(li) on the basis of the realization that tar repre sents a clipped form of tarpauline and application of Metonymic Principle 8. Such links are expressed in the form of via-rules at tached to the related lexical entries, as outlined above. The relations thus expressed are included in the lexicon model for some individual, not in a description of the semantic component of the language in question, though they make reference to regular relational principles which are part of general semantic competence. If two words fall to gether phonologically but manifest no regular semantic re lation, they are simply homophones.
118
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Cases of derived polysemy arise through application of regular relational principles to contextual occurrences of lexical items. Since homophony arises fortuitously as a result of the collapse of two phonological representations into a single one, there can be no derived homophony cor responding to derived polysemy. 4.2 Lexical
Polysemy
and Derived
Polysemy
A number of observations speak for a unified treatment of figurative usage and semantic relations between lexical entries. First, the history of semantic change recorded in suc cessive versions of the lexicon of a language in progress is essentially that of creative figures and nonce exten sions and transfers from existing lexical readings which gain acceptance in the linguistic community, i.e. which become lexicalized generally. Historically the word speaker designated only humans. With the advent of artificial devices capable of repro ducing the human voice, speaker was extended (by Metaphoric Principle 1 (object-model) allowing for the dele tion (by Transformation 1) of the feature (Human)) to de signate the component of such devices we identify as re sponsible for reproducing the vibrations we recognize as those produced by humans when speaking. In this case, the derived reading achieved currency and found its way into the lexicon (for English speakers familiar with the elec tronic device). Any statement of the relation between the two lexical readings for speaker in the lexicon must per force mirror their historic (etymological) connection. The lexical relation is simply a record of the original figu rative meaning extension. The simplicity metric will de mand that reference be made to the same relational prin ciple (here Metaphoric Principle 1) in both cases. Parallel extensions from originally (Human) readings for items such as mouth to 'opening of river', 'opening of bottle', leg to 'support for table, chair, desk etc.', planter to 'planting machine' illustrate the same historical process and must receive essentially the same treatment, where the simplicity metric urges reference to a single relational principle predicting the original extension and describing the static relation once the derived reading has been lexi calized. Second, as a consequence of the process just described, the lexicon provides a veritable graveyard of dead meta phors. The verb pen 'to write with a pen' exhibits a
THE PROBLEM OF POLYSEMY
119
regular relation to the noun pen 'writing instrument', which is defined by our Metonymic Principle 6 (instrumentact) . This same principle must have applied productively in the original transfer from the noun reading to the verb reading. That it continues to be productive is shown by the above cited example x e r o x V derived from xerox {ma chine) . Technically, words like pen are not simply polysemous, but 'poly-categorial' as well, to coin a term in tended to convey the status of lexical items like pen which are associated with separate entries differing not only in their semantic representation but in their syntac tic categorization as well. Conversely to the previous argument, which showed that principles predicting semantic change are carried along into the lexicon with their out put, I intend to argue here that lexical relations already stored represent the principles which led to their crea tion. One can confidently assume that the principle re lating the two respective meanings of pen and xerox in the lexicon is the principle historically responsible for the derivation of one from the other. Again the simplicity metric demands a description of the historical process in terms of the same regular relational principle required to capture the lexicalized connection. Third, and partially determined by the foregoing argu ments, principles recognized to be regularly functional in lexical relations tend to experience productive applica tion outside the lexicon. Having noticed the lexical con nection between leg (human) and leg (furniture), face (human) and face (clock) etc., speakers are prepared to interpret ear (of corn) as derived from ear (human) or to extend knee (human) to a movable joint between two metal bars. So we can view the lexicon as a record and as a re pository of regular meaning relations, all of which offer themselves for productive application. In formulating reg ular semantic relational principles I essentially ex tracted the relevant relations from the lexicon. But since I intend the lexicon to model the competent speaker's knowledge of the stored units of his language, I continue to cite the extracted relations, in the form of relational principles, in via-rules connecting entries. At the same time, however, these relational principles find productive application. Again, considerations of simplicity should convince us to employ the same set of principles in de scribing lexical relations as we use in describing pro ductive figures. The foregoing discussion leads to a fourth relation between cases of lexical and derived polysemy. At any given point in time, some speakers will have lexicalized certain readings for lexical items which, other speakers
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
120
consider novel, odd or even inacceptable. Nevertheless, those speakers who have not lexicalized a particular read ing will understand an occurrence of the word in question requiring this reading. By the same token, a speaker might create what he considers a novel figure of speech only to find that his interlocutor(s) have already lexicalized the reading for the item he felt was original. Now if a speaker creates a figure others have already lexicalized or if a speaker employs what he considers a lexicalized reading for an item while his hearers Interpret it as fresh figure of speech, but have no difficulty Interpreting it, then it must be that relations between lexicalized Items and prin ciples allowing creative meaning extension or transfer are isomorphic or identical. At least cases of regular lexical relations and regular figures should be subject to a single set of relational principles. If someone says, "Cloth the table," in my hearing, I will interpret him to have in tended 'put a table cloth on the table' (by Metonymic Principle 5 allowing transfer from the name of the object of an act to the act itself). If he has stored a lexical reading for cloth as a verb meaning 'cover with cloth', its relation to the noun cloth will be expressed by the same metonymic principle. If, on the other hand, he acted creatively in transferring the standard reading for the noun cloth to the verb reading 'cover with cloth', the transfer will also be described by Metonymie Principle 5¿ The fifth and final argument for the identity of lexi cal relations and productively derived relations is based v upon selectional restrictions. In the process of metaphoric extension, features entailing selectional restrictions are deleted. Lexical readings related in similar fashion ex hibit the same selectional differences. The transitive verb swallow has one reading which identifies an act of in ternalizing x by way of the mouth and selects direct ob jects (to replace the variable x) with the feature (Con crete) ; it has another reading which identifies an act of mentally internalizing x and selects direct objects with the feature (Abstract). Consequently, both (14) and (15) are possible. (14) (15)
Harry swallowed the egg. Harry swallowed the idea.
On this model we could metaphorically extend the con crete reading for regurgitate to allow: (16)
Harry regurgitated the idea.
THE PROBLEM OF POLYSEMY
121
We derive the non-concrete reading for regurgitate by de leting the features (Concrete) (By way of mouth). The re sult is a change in the selectional properties parallel to that between the two readings of s wallow. If lexical and derived polysemy are treated as separate phenomena, there exists no natural way of capturing the generalization that meaning differences affect selectional restrictions in parallel fashion in both cases. The lexical item pantaloon has one reading identifying a garment and bearing the features (Artifact) and (Nonhuman) ; it has another reading identifying an old man and bearing the feature (Human). Thus, both (17) and (18) have acceptable non-figurative readings. (17) (18)
The pantaloon tore. The pantaloon swore.
The relation between the two readings is one of costumewearer (cf. Metonymic Principle 14). The former reading can be selected as subject by a verb like tear, while the latter can be selected by the verb swear . The same seman tic relation could come about as the result of a novel ap plication of Metonymic Principle 14 with the same conse quences for selection. On one standard lexical reading of uniform it meets the selectional requirements for a sub ject of tear. (19)
The uniform tore.
By metonymie transfer from costume to wearer, uniform meet the selectional restriction imposed by the verb viz. that it contain the feature (Human). (20)
can swear,
The uniform in the third row swore.
Certainly the parallel cases of differences in selectional properties should not be treated as independent in a sem antic theory intended to model speaker competence. We have observed a number of correlations between lexi cal and derived polysemy. The desire to describe a theory accurately reflecting the structure of language and the psychology of speakers as well as considerations of simpli city dictate the treatment of lexical and derived polysemy as a unified phenomenon. Cases of both sorts should be de scribed by a single homogeneous set of rules. The present theory accomplishes this by positing the inventory of reg ular relational principles predicting both derived inter pretations and lexical relations by way of via-rules.
122 4.3 Current
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY Proposals
concerning
Polysemy
Let us now review and criticize current semantic the ories as they concern lexical and/or derived polysemy. I hold my own contributions to a minimum here, preferring to present them in a cohesive theoretical framework in the two following chapters. 4.3.1 Proposals
concerning
Lexical
Polysemy
Katz and Fodor (1963), Katz and Postal (1964) make no attempt whatsoever to distinguish in the lexicon between regularly related readings associated with a single phono logical representation and other clearly fortuitous read ings resulting historically from the accidental coincidence of phonological representations. They propose a single lexical entry for a word b all with three distinct readings, viz. (i) social activity for dancing; (ii) physical ob ject of globular shape; (iii) physical object used as solid missile for projection by an engine of war. The only formal connection from reading (ii) to (iii) in this sort of representation is the presence of 'physical object' as a so-called marker in both, and neither exhibits any con nection with reading (i). Katz and Fodor, Katz and Postal have been criticized (with good reason) for treating such disparate readings as part of a single entry for ball. One such criticism is due to Weinreich (1966), who clearly considers the difference between reading (i), on the one hand, and (ii) and (iii) on the other, as sufficient reason for entering two sepa rate entries b a l l 1 and ball2 in the lexicon. But Weinreich does not offer any real justification for distinguishing related from unrelated readings, nor does he suggest how relatedness of meaning generally is to be captured in the lexicon; this shortcoming is serious, since Weinreich ad vocates purely univocal lexical entries. One could, of course, choose to list related readings like (ii) and (iii) for ball in a single entry, while placing reading (i) in an entry of its own. Nevertheless, phonological identity should not be considered of such great significance in the structure of the lexicon. A lexicon structured around groups of phonologically identical or related readings, irrespective of their semantic content or even whether they manifest regular relations, destroys the semantic class structure of the lexicon which was justified above. Katz (1972: 6 9 ff.) argues that polysemy and semantic relatedness generally are matters of degree, and that the
THE PROBLEM OF POLYSEMY
123
degree of relatedness between two readings can be deter mined by counting their common features. But speakers re port seeing clear relations between some pairs of readings and none at all between others (cf. Lehrer 1974a), even when the latter have some high level features like (Ob ject) , (Physical), (Concrete) in common. Therefore, one would have to designate the number of shared features re quired for relatedness, but this would return us to the problem of defining relatedness in a non-graded manner. In addition, if features are to be counted to determine re latedness of meaning, then relatedness should not be cited in arguments for particular feature specifications for lexical items. Some notion of relatedness must clearly be anchored outside semantic theory proper in order to avoid such circularity. Gruber (1976) differentiates two types of lexical polysemy. The first is illustrated by items like novel, which may designate either a mental construct or text in the abstract sense of a concrete object consisting of bound leaves of paper. Gruber expresses this polysemic re lation by formulating the latter reading in such a manner as to contain the former. In Gruber's tree-type lexical entries, the mental reading for novel looks like Figure 1, the physical reading like Figure 2.
The second type of polysemy is observed in large classes of items which predictably alternate between readings of two sorts, e.g. slide 'move' versus slide 'cause to move'. Here Gruber would enter only the non-causative readings for the whole class of items identifying movements, e.g. slide,
move,
roll,
stir,
sail
etc., while the availability
of the causative readings is predictable given exception less rules like Figure 3, into which items characterized as go (in this case) are inserted to become thereby causa tive in meaning.
124
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Gruber's reasons for recognizing two separate cases of lexical polysemy are unconvincing. The method of entry in clusion illustrated by Figures 1-2 could easily describe the causative cases as well. Given the presence of entry redundancy rules of the type Jackendoff (1975) proposes or like the via-rules described in the present investigation, the inclusion of all the causative readings in the lexicon need not count as additional complexity. Further, by de riving the causative readings for items like slide, Gruber's lexical proposals fly in the face of psychologi cal reality. It makes little sense in psychological terms to claim that speakers store two readings for items like novel, but only one for items like slide. But even grant ing that a consistent line can be drawn at some point be tween (partially) redundant stored readings and derived readings, a single device might serve for both cases. By contrast, the theory proposed here posits but a single set of relational principles; the distinction be tween lexicalized and derived polysemy is drawn along psy chological lines. One additional shortcoming in Gruber's theory should be noted. The fact that Figures 1 and 2 con tain a common component only becomes evident when both en tries are examined. I prefer a type of entry which con tains explicit reference to the other lexical entries which allow portions of it to be characterized as redun dant. Again, Gruber's proposals require complementation by some sort of redundancy rules. McCawley (1968a) also suggests that certain items be left out of the lexicon entirely, since their presence in a language can be predicted on the basis of other lexical items. McCawley recognizes principles predicting both sorts of polysemy defined by Gruber; there would be a causative principle for items like slide and a principle of reification predicting the physical readings for items such as novel on the basis of their mental readings. At the same time, McCawley (1968b) suggests complex lexical entries for phonologically identical and semantically re lated pairs like cookN-cookV, which I above labeled polysemous and polycategorial. In a manner similar to Gruber's, a complex lexical entry for the noun cook would take the
THE PROBLEM OF POLYSEMY
125
form: 'one who by custom/profession V' coupled directly with the entry for the verb cook: 'prepare (food) by heat ing' , adhering to the convention that the space labeled with the V in the semantic representation for the noun is to be filled in with the entry for the verb. My objections to McCawley will essentially repeat those offered to the other suggestions made so far: (i) the difference between the two methods of capturing polysemic relations does not reflect the psychological state of affairs, but simply the exigencies of the method of description; (ii) the presence of complex entries in the lexicon would disrupt its struc ture based on semantic classes, or even that based on a division of word categories; (iii) relations of both sorts could be captured in entry redundancy rules with an over all simplification of the lexicon. Chomsky (1970) proposes that lexical items be option ally left unspecified for syntactic category. Under this proposal, cook would be entered in the lexicon as cook V N , where the choice of one or the other category symbol is made anew each time information from the entry is inserted. The semantic differences associated with the different category choices would be predicted by general rules relat ing readings of the categories in question. Thus, some standard statement of the semantic relation between verbnoun pairs must be forthcoming. In the theory advocated here, such relations find expression in the inventory of regular relational principles. But if only those pairs with identical phonological representations are so related, the theory loses sight of the fact that other phonologically distinct or even unconnected pairs may exhibit the same semantic relation. At the same time, objections (ii) and (iii) offered for McCawley's proposal retain their validity for Chomsky's suggestion as well. Again I prefer Jackendoff's (1975) system whereby each entry receives a separate listing and predictable semantic information in any given entry is characterized as redundant by rules connecting entries. Jackendoff's solution allows one to retain the hypothesis that the lexicon is structured around semantic classes, rather than being governed by phonologi cal identity. Fillmore (1968, 1971a) offers a number of interesting proposals concerning the form of lexical entries pertinent to the discussion of polysemy. Fillmore writes (1971a: 385): "syntactically and semantically different uses of the same word type should be listed in the same lexical entry whenever their differences can be seen as reflect ing a general pattern in the lexical structure of the language." Hence, Fillmore would list the adjective open and the verb open in a single lexical entry
126
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
as alternate uses of the same word type, where open with the obligatory deep case Object noun phrase represents the adjective as in: The door is open, whereas open with the deep case Object and the optional case Agent represents the verb as in: Harry opened the door. Similarly, event descriptions shared in part by noun-verb pairs should occur only once in the lexicon. A single lexical entry such as: Verb Object Agent will suffice for both verb and noun readings of 'word types' such as pilot, if we write: Verb
= p i l o t V and Agent
=
pilotN
as part of the compound
lexical entry. In this fashion, Fillmore's deep case re lations mirror many of the principles we defined as part of our inventory. Nevertheless, using Fillmore's formalism, one can directly define only those regular semantic rela tions holding between alternate readings of 'word types', i.e. of phonological representations associated with mul tiple meanings. Here again semantic relatedness must take a back seat to phonological identity in lexical relations. Fillmore provides no way of accounting for the fact that productive meaning extensions and transfers are modeled on relations manifested by lexicalized readings. In addition, such lexical relations as hyponymy are ignored, since no provision is made for relations arising from intersecting and hierarchically ordered classes. Lehrer (1974b) represents the latest in a long line of scholars who argue for a lexicon structured around sem antic classes or so-called semantic fields. According to Lehrer, a lexicon built around semantic fields ignores polysemy, since each different sense of a lexical item must appear in the appropriate field. This conclusion, however, ignores the possibility that non-field relations are captured by some means other than interconnected sem antic fields. Lehrer argues that feature representations for lexical entries can be derived on the basis of their field membership and the taxonomies the fields represent, but there must be an essentially infinite number of dis tinct features to account for an infinite number of lexi cal entries. So a finite number of semantic fields must be (theoretically) infinitely divisible into individual areas of meaning. Lehrer also notes that the semantic field ap proach must be complemented in some way to account for the whole range of semantic competence. In particular, among other devices, Lehrer sees a necessity for relations be tween non-intersecting entries within non-intersecting semantic fields. Again, these could be handled by entry redundancy rules in the form of the via-rules proposed in the present theory. In a semantic field approach, polysemy can be handled as the appearance of a single lexical item
THE PROBLEM OF POLYSEMY
127
in various semantic classes, i.e. in multiple lexical en tries, where the readings are linked by rules connecting items in non-intersecting semantic classes. This amounts in essence to the proposal contained in the semantic the ory presented in.this study. I therefore support a supple mented field approach in the sense of Lehrer (1974b). Hudson (1976b) opposes any formal distinction between homophony and polysemy, but ignores the psycholinguistic evidence in its favor (e.g. Fromkin 1971; Fillenbaum and Rapoport 1971; Lehrer 1974a; Forster 1976). He hopes to do away with this distinction by disallowing all branching and disjunctions in lexical entries. However, as we have seen, forbidding branching and disjunctive lexical entries can only preclude the possibility of polysemy in a lexi con with no further means of relating entries. But we have clearly demonstrated the necessity of recognizing lexical relations beyond those defined or definable by compound or disjunct lexical entries. It is difficult to imagine that Hudson would be willing to accept a lexicon without such lexical relations, given speakers' recognition of semantic relations, e.g. between loan and borrow or learn and know, and of phonological identity, as evidenced by punning. But if speakers recognize both relations of sem antic connection and phonological identity, they must con sider certain pairs of words as polysemous. Consider as an example, German l e i h e n 1 'loan' and leihen2 'borrow', which for German speakers represent a single polysemous lexical item. One cannot simply define away the phenomenon of poly semy. Whether branching and disjunctive lexical entries are forbidden in the linguist's model of the mental lexi con or not, speakers will go right ahead extending lexical entries like learn 'come to know' to cases of causing to know, i.e. 'teach'. (21) (22)
Harry learned how to tie knots. Sue learned Harry how to tie knots.
Hudson's model of the lexicon fails every time speakers of a language extend or transfer existing lexical readings according to models they recognize from relations among the words they already know, i.e. among entries in the mental lexicon. I have already had occasion to refer to Jackendoff's (1975) proposals concerning the lexicon. Most importantly, he suggests that the simplest lexicon has fully specified entries and as many redundancy rules as possible. This is in line with Maenamara (1971), who equates parsimony in the lexicon with ease of accessibility rather than with a
L28
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
minimum of information. Jackendoff's redundancy rules do not fill in redundant information, they simply charac terize it as redundant on the basis of relations among en tries. The evaluation metric for the lexicon does not count information characterized as redundant in determining a lexicon's complexity. But Jackendoff goes on to limit en try redundancy rules to linking entries for morphologi cally related items. A separate set of semantic rules con nects entries sharing no morphological material. Since it seems that identical semantic relations may hold between pairs of words with and without morphological connections, I see no point in separating semantic relations along morphological lines. Just as branching and compound lexical entries make semantic relatedness depend upon phonological identity, Jackendoff's restriction makes it depend upon morphological connection. Yet it seems most reasonable to recognize semantic relations on semantic grounds alone; any other choice fails to capture generalizations about the range of various semantic relations and requires two sets of relational principles describing the same connec tions for different groups of items. If a regular semantic relation holds between phonologically identical entries, they belong to a single polysemous lexical item. If one holds between entries with morphological material in com mon, they are related by morphological rules. Viehweger et al. (1977) subscribe to a conception of the word as a sign function compounded of expression and content. They are therefore committed to treating each distinct meaning associated with a sound sequence as be longing to a separate lexical item, hence again losing track of the difference between polysemy and fortuitous homophony. This move obviates the necessity of deciding in each case of meaning multiplicity whether to posit one lexical item or two, but it does nothing toward solving the primary problem of determining which meanings are re lated and which not. Since speakers clearly recognize meaning relations and phonological identity, it makes no sense to deny the existence of polysemy. Since they can also distinguish relatedness from non-relatedness in lexi cal items and larger units, there clearly exists a basis for the distinction between polysemy and homophony. It is neither necessary nor by any means sufficient to claim, as e.g. Lyons (196 8) does, that precise boundaries between polysemy and homophony cannot be established. If Lyons simply means to say that no two speakers can be expected to agree in all cases, then he is right. But a model of any individual's lexicon must clearly allow for both semantically related and semantically unrelated readings
THE PROBLEM OF POLYSEMY
129
assigned the same phonological representation. According to our definitions, these two cases instance polysemy and homophony respectively. Different speakers recognize dif ferent pairs of items as related and consequently different pairs of polysemes. This represents a perfectly acceptable conclusion, especially because the lexicon is that portion of competence most susceptible to interpersonal variation. It is all the more true, since extra-linguistic knowledge plays a role in speakers' recognition of connections be tween certain pairs of items, as we saw above for examples such as tar 'sailor'; cf. Chapter 8 below. Via-rules al low us to describe the relations any given speaker recog nizes without thereby elevating this recognition to the status of a part of competence as such. 4.3.2 Proposals
concerning
Derived
Polysemy
We have seen that the semantic theory developed in Katz and Fodor (1963), Katz and Postal (1964), Katz (1972) and elsewhere labels as anomalous sentences containing selectional violations. Along with strings exhibiting gram matical violations, these structures make up the set of what Katz (1964) calls 'semi-sentences'. Semi-sentences are assigned readings on the basis of their relations to grammatical, non-anomalous sentences sharing structural properties and lexical formatives with them. Various ob jections to this theory have already been mentioned. We can summarize them as follows. (1) In this theory the search for non-literal interpretations begins only when selectional violations occur, but a single sentence may require both literal and figurative interpretations at the same time; witness: (23)
Judy tasted the book.,
which means literally that Judy put some of the book in her mouth, but figuratively that she read an excerpt or two from it. A semantic theory should be able to predict (possible) figurative interpretations even for sentences with consistent literal readings. Bergmann (1979) shows that the presence of a violation is neither sufficient nor necessary for figurative readings. (2) By treating only individual cases of anomaly as instances of figurative language, Katz's theory fails to recognize regular types of meaning extension and transfer, although speakers clear ly do recognize them. (3) The theory provides no explana tion of why or how figurative interpretations for items
130
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
become lexicalized. Weinreich (1966) extends a device Katz and Postal (1964) originally proposed for assigning features from pre dicates to proforms and depleted items so as to transfer features to any lexical item at all. Katz and Postal (1964) introduced feature transfers to avoid selectional viola tions in structures where semantically neutral items like it and thing occur in construction with predicates re quiring particular features. The verb pour out selects di rect object noun phrases containing the feature (Liquid). Since something cannot be characterized as (Liquid) in the' lexicon, it must be assigned this feature in sentences like (2 4) in order to avoid having to label them anomalous. (2 4)
Judy poured something out.
Weinreich (1966) suggests allowing feature transfers of this sort any time a predicate requires a particular feature in its subject or object noun phrase and this fea ture is not present. A 'construal rule' in the 'semantic calculator' deletes inherent features of the host repre sentation when they clash with transferred features. Thus, compromise would pick up the features (Object) (Physical) required by the verb hammer out in a sentence such as (25). (2 5)
We hammered out a compromise.
While Weinreich's proposal appears initially to possess great advantages, it becomes untenable upon closer inspec tion. As McCawley (196 8b) points out, feature transfers can account for only a small percentage of figurative readings. Since features are always transferred from predi cates to nouns, one can obtain only a single 'fairy tale' reading for constructions such as (26), on which the boats are animate and possess legs, but not the more usual meta phorical reading on which dance extends its meaning to something like 'move rhythmically'. (26)
The fishing boats danced in the harbor.
Further, feature transfer cannot account for the read ing of (2 7) on which it means 'the container of milk tipped over', although it would seemingly predict the reading on which the milk is rigid, say because it is frozen. (2 7)
The milk tipped over.
THE PROBLEM OF POLYSEMY
131
The objections raised against Katz's theory apply to Weinreich's proposals as well. Thus, (1) feature transfer only takes place when feature clashes occur; they too pro vide only a single reading for (23) above, viz. the literal one. (2) Feature transfers apply only when they are needed, so, again, the general principles of figurative language go unnoticed. (3) No explanation for the lexicalization of figures is offered. In addition, (4) feature transfers abolish the distinction between figurative language and logical contradiction, which Katz's theory upholds. If features are transferred whenever clashes occur, even sen tences like (28) and (29) will receive consistent inter pretations. (2 8) Harry ate the round square. (29) The colorless green idea slept. Obviously, the device of feature transfer must be con strained in some way to avoid yielding consistent inter pretations of logically contradictory sentences. What Weinreich's feature transfers do best is provide a means of obtaining what I termed 'fairy tale' readings for sen tences. Thus, while we do not want compromise to end up with the features (Object) (Physical) in the semantic reading assigned to (25), the (Animate) reading for boat in (26) could be significant in some textual contexts. Hence, Weinreich's proposal has enjoyed acceptance and ex tension by other writers. Baumgärtner (1969) suggests allowing for feature trans fers not only from predicates to nouns but in the opposite direction as well to account for a greater range of figu rative constructions. In addition to transferring the selectional features[ (Object) (Physical) ]from hammer out to compromise in sentence (25), Baumgärtner's proposal would allow transfer of a feature, say (Abstract), in the other direction. Then hammer out would be reinterpreted in an abstract sense. Certainly this proposal enables the fea ture transfer device to predict a wider range of interpre tations for sentences containing selectional violations, but it perpetuates the same errors pointed out in Wein reich's theory. Indeed, even if one allows transfers from nouns to verbs and adjectives, the container reading for (2 7) above cannot be obtained. Also, since we saw above that the feature transfer proposal is too powerful in cer tain ways, redoubling its range in Baumgärtner's manner seems ill-advised. Abraham and Braunmüller (1971) achieve essentially the same descriptive power as Baumgärtner (1969) with a set of
132
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
operations affecting selectional and inherent features during syntactic derivations. They define these operations in terms of post-cyclic optional transformational rules. This proposal errs in being too powerful, as do the fore going proposals. In addition, it seems misguided in using syntactic devices to describe essentially semantic opera tions. Again, in describing such a wide range of figurative usage, Abraham and Braunmüller fail to define regular types of figures. Van Dijk (1972) records several objections to Weinreich's theory of feature transfer. Most importantly, he notes that transfer as such cannot remove selectional clashes; it simply moves the feature responsible for the selectional clash between two items from one to the other. What is required instead are rules of feature extension, rules which copy a feature from one term of a construction into the feature specification for the other when clashes occur. This seems to accomplish what Weinreich (1966) pre sumably originally intended, so I follow van Dijk in view ing the relevant process as a feature extension (or copy ing) . Van Dijk goes on to widen Weinreich's analysis so as to allow features to be extended not only from predicate to nominal but in the reverse direction as well. There are problems with this revision of Weinreich's theory beyond the fact that it renders the device more powerful, and I touch on them in the discussion of Levin (1977) below. In addition, in most cases, feature extension from nominal to predicate yields the same result as direct feature dele tion from the predicate, a process van Dijk also allows, so that extension from nominal to predicate may be un necessary anyway. Despite these changes, van Dijk's fea ture extension continues insufficient in the same ways Weinreich's feature transfer was, viz. in failing to pro vide a means of constraining extensions, and In failing to capture any relevant generalization about the regular processes by proposing various 'metaphorization' rules. His feature deletion rule (rule (45), 1972: 261) parallels my Metaphoric Principle 1 described above; his personifi cation rule (rule (51): 264) assigns the feature (Human) where necessary to avoid clashes between predicates de scribing human actions and (Non-human) nominais. He also makes use of metonymic transfers (e.g. part-whole), com menting that the only problem facing a complete descrip tion of them lies in defining "the empirical contiguities" (26 7). The metonymic principles offered above, based as they are upon general semiotic principles of motivation, provide a solution to this problem. By allowing feature extension to operate along with such metaphorization rules,
THE PROBLEM OF POLYSEMY
133
van Dijk enables his theory to derive a wide range of figurative interpretations, and I am in general agreement with his method, especially as it applies to complex, literary/poetic figures (cf. van Dijk 1972: 258 ff.; also van Dijk 1975). Levin (1977) proposes further refinements on Weinreich's (1966) feature transfer theory. In addition to feature ex tension which adds a feature to the host specification, Levin allows extension which displaces a feature in the host specification. This opens the door to additional pos sibilities of construal. Like van Dijk (1972), Levin pro vides for transfer from both predicate to nominal and nom inal to predicate. He recognizes the problems surrounding feature extension from nominals to predicates, but seeks to neutralize them in two ways. First, he requires col lateralty of feature level for transfers, i.e. the fea ture transferred must occupy the same level in the class hierarchy as that it replaces. Second, he introduces the 'parallel verb device', viz. after a feature from a nom inal is introduced into the feature specification of a predicate, a predicate parallel to it but lacking its rightmost feature(s) must be substituted in its place, in order to ensure obtaining a modified reading. This is necessary because features transferred into predicate terms become selectional restrictions in their feature specifications rather than part of their meaning proper. Transfers from predicate to nominal terms present no prob lems in this respect, since selectional restrictions transferred from predicates become part of the inherent feature set for the nominal. Levin describes various con strual rules on the model provided by Weinreich (1966), which operate on feature complexes created by feature transfers to yield consistent feature specifications as input into the amalgamation (projection) rules. The con strual rules may delete neutralized and clashing features or combine the transferred feature(s) with the complex to which it was added. With these devices, Levin can obtain a greater number of apparently different readings for a single anomalous structure than can van Dijk (1972). Nevertheless, as I hope to demonstrate below, there is little point in attempting to distinguish more than three separate readings in a single construction, viz. those de riving from feature extension, feature deletion and their combination. Other writers have attempted to capture regular types of figurative constructions in principles or rules in cluded in the domain of semantic theory proper. To the ex tent that figurative language is truly regular, this seems
134
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
empirically sound. Also, it renders manageable the problem of correlating figurative meanings and selectional restric tions. Theories which treat meaning extensions and trans fers as rule breaking behavior or a matter of either per formance or reference outside the realm of semantic compet ence as such cannot handle this correlation in any satis factory manner. Clearly however, there comes a point where regular processes alone no longer account for truly cre ative figurative language. At this point, performance and referential theories or theories of rule breaking linguis tic behavior certainly have their place. I return to such theories below, but concentrate for the present on those proposals attempting to integrate regular rules of meaning extension and transfer into semantic theory. The theories just discussed which work with and expand the notion of transfer features as set down by Weinreich (1966) fall within this group, in addition to the following. Leech (1969, 1974) proposes rules in the semantic the ory of the basic form: "Replace A by If they are related in such and such a way." Leech considers metaphor an im plicit comparison, so the rule for deriving metaphors simply deletes the explicit expression of the similarity relation, i.e. converts Sue is like a lion to Sue is a lion; in parallel fashion, the interpretation of metaphor involves the insertion of some expression such as similar to. Leech analyses metonymy as an elliptical construction, so the rules for handling the various metonymies add appro priate phrases. Thus, a word or phrase identifying a local ity will be replaced by one identifying the residents of that locality; a word or phrase identifying a container will be replaced by one identifying the contents of that container, and conversely (cf. the metonymic principles above). Leech stresses the following characteristics of such semantic rules: (1) diversity: many may apply to a single form, (2) open-endedness: results are often incon clusive or ambiguous (because of (1)), (3) recursiveness: output of one rule may be input to another, (4) bidirectionality: all rules operate in both directions, (5) petri faction: output of the rules tends to get lexicalized. Ex cept for bidirectionality, which, as we saw above, applies for metonymic but not for metaphoric principles, all these characteristics hold for the relational principles pro posed in the present study as well. Gak (1976) proposes rules which he terms 'semantic transformations' as part of semantic theory. Semantic transformations transfer from the designation of one sem antic structure to that of another by adding or deleting semes (or features). They may thus narrow or extend the
THE PROBLEM OF POLYSEMY
135
lexical reading assigned an item. The relation between their input and output may be established logically, psycho logically or culture-historically. The first two should correspond closely to the metonymic principles presented here, while the latter parallel metaphoric principles to the extent that these involve encyclopedic information (cf. discussion above). Lehmann (1975) and Droste (1976) offer similar proposals, but do not go beyond Leech and Gak in terms of formalized suggestions. As noted in Chapter 1, none of these theories has much of value to say about the connection of figurative language and lexical relations or about the process of lexicalization as it pertains to figures. Nor do they undertake any attempt to provide a psychological founda tion for the principles posited, to distinguish regular from spurious, uniquely creative figures or to constrain the principles in any manner. Further, while van Dijk (1972) and Levin (1977) present relatively complete analy ses of a particular portion of figurative language, neither proposes a treatment of metonymic transfer, although it is fairly obvious how such a treatment could be integrated into either system. At the same time, while Leech (1969, 19 74) and Gak (1976) make provisions for theories inte grating metonymie principles with metaphoric principles, their systems remain fairly sketchy. We still require a complete theory integrating principles describing various types of figurative language. In addition, we still lack a model of the relation between creative figures and lexical links, and, particularly, a model of the process of lexi calization of figurative interpretations for lexical items. I address myself to these problems in the following two chapters. 4.4 Comments on Feature
Transfer
The following view of feature transfer rules can be distilled from the theories cited above and the comments on them. First, following van Dijk (1972), features are extended, i.e. copied, rather than transferred from one term to another. What is extended from a predicate term to a nominal is a selectional restriction holding for the predicate; it is copied into the feature specification for the nominal where it becomes an ad hoc inherent feature. If we accept something like Levin's (1977) construal rules, which may delete a feature in contradiction with another which has been extended into the complex it occurs in, it seems superfluous to allow feature extensions both to add to and to displace features of the host matrix. The same
136
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
effect as displacement can be achieved by addition and subsequent deletion of the conflicting feature which would have been displaced. So I suggest limiting feature exten sion to the addition of features. Further, while construal rules must be empowered to delete conflicting features, they must be prevented from resolving true contradictions. Thus, if the feature (Human) is copied into a matrix containing a feature such as (Bird), the construal rule in question will delete the latter feature; on the other hand, if the feature (Hairy) is copied into a matrix containing the feature (Bald), the construal rule should register this contradiction, not eliminate it. I thus uphold the distinction between anom aly and contradiction (cf. Katz 19 72; Levin 1977 and be low) . If a construal rule of this type is not included in semantic theory, feature extensions must themselves be prohibited from creating logical contradictions. By the same token, one must ensure that feature extension does not operate on an input sentence containing a logical con tradiction. So I stipulate that feature extensions can neither dissolve nor create logically contradictory sen tences. Since feature extensions lack any regular underlying motivation, in contrast to the relational principles de scribed above, I shall also specify that feature exten sions occur only when regular processes fail to yield a consistent reading for a string, i.e. for strings charac terized as anomalous even following attempted application of the regular relational principles. The lack of regular ity characteristic of feature extensions is reflected in the fact that readings derived by feature extension are never lexicalized. Consequently, we need not account for feature extension-type relations in the lexicon, and the question of hooking up the feature extension analysis with the lexicalization process does not arise. It appears that the regular relational principles proposed above suf fice to describe all the figurative readings derived from existent lexical items and lexicalized as such. Consider now the question of whether to allow feature extensions from nominals to predicates. As Levin (1977) points out, in order for feature transfer from nominals to predicates to eliminate selectional clashes, the feature from the nominals must become a selectional restriction in the predicate. This alters the set of frames the predicate may properly appear in, but it does not affect its mean ing as such. Hence Levin proposed the parallel verb device discussed above, where the verb substituted necessarily bears less total features than the one it replaces. This
THE PROBLEM OF POLYSEMY
137
latter step has the same effect as van Dijk's (1972) dele tion rule or my Metaphoric Principle 1 described above would, if applied to the predicate in question. So I see no reason to include the parallel verb notion in semantic theory, inasmuch as feature deletion rules will be re quired in any case. What feature extension from nominals to predicates does seem to offer in addition is the neu tralization of selectional clashes with no attendant change in the inherent feature complex of the predicate. Thus one possible effect of feature extension into a pre dicate consists in the neutralization of one of the pre dicate's selectional restrictions, just in case the fea ture received contradicts an already present selectional feature. But this effect too can be handled by standard deletion rules, if these are defined so as to operate on both selectional and inherent features. What other devices proposed to date cannot describe is the case in which a feature from a nominal encounters no contradictory selectional features in the matrix for the predicate to which it is extended as a selectional restric tion. This effects a narrowing of the environment imposed by the predicate in question. A moment's reflection, how ever, reveals that this case requires no special descrip tive device, nor even a special description. It is simply • the situation in which a predicate applies to a nominal possessing a greater degree of specificity than it imposes, and can be illustrated by the following sentence. (30)
The woman ate the jelly bean.
Here woman is specified as (Human) and redundantly as (Animate). The verb eat requires only that its subject nominal be specified (Animate), so that, on Levin's ana lysis, the feature (Human) can be extended from woman to the selectional restriction for eat. But nothing is gained by this operation; no information becomes available which was not already present, since the amalgamation of the readings for woman and eat by the projection rule will also have the effect of producing an interpretation on which (Human) combines with the other features of eat. Also, since woman bears the feature (Animate) redundantly, no feature extension is required to obtain a consistent reading. Obviously, no process of feature extension is necessary to account for constructions of this kind. Fur ther, since this was seen to be the only case in which the effects went beyond those attainable by deletion opera tions alone, there remains no motivation for feature ex tension from nominals to predicates. We can thus restrict
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
138
feature extension to copying selectional features of pre dicates into feature matrixes for nominals. In addition, feature extension takes place only in sentences which would otherwise be labeled anomalous, i.e. after attempted application of projection rules and regular relational principles. Feature extension does not dissolve contradictions, nor does it lead to contradictions. At least two construal rules are necessary for feature com plexes created by feature extensions: one deleting any fea ture (s) clashing with the newly introduced feature, and one combining the introduced feature into the feature matrix in such a manner as to create a new reading. Given a sentence such as (31)
The hammer shattered my wolf.,
where the selectional restriction[ (Brittle) ]required by shatter for its object is copied into the feature repre sentation for wolf, the latter type of construal rule should combine (Brittle) and the other features specifying wolf to yield a reading on which the wolf was frozen, pet rified or something similar. Since sentence (31) describes a possible, if rather unlikely, state of affairs, we cer tainly do not want to characterize it as logically incon sistent, nor even as a contradiction in the sense of Katz (1972). The feature specification for wolf will not con tain the feature (Non-brittle) nor should it presuppose this feature by way of feature redundancy rules, especially in a lexicon which classifies items in terms of distinctive features alone. Hence, the feature (Brittle) may be inte grated into the reading for wolf without creating a logi cally contradictory feature complex. A description of the actual process by which some in dividual interprets (31) as referring to a petrified or frozen wolf or something else clearly would have to take features of the context, aspects of the individual's sys tem of knowledge and beliefs about the world, and a range of additional information into account. We have been con cerned here with the much less ambitious task of modeling a portion of the semantic competence which, along with other factors, enables speakers to make sense of sentences like (31). To summarize this section, we have outlined the con straints on and functioning of feature extensions within the domain of semantic theory.
DERIVED POLYSEMY
5.
5.0
Introduction
This chapter describes the integration of the regular semantic relational principles derived previously into semantic theory. Before entering into the discussion of the relational principles as such, I address various issues surrounding anomaly, selection and readings for lexical items. I shall then describe the functioning of relational principles in an interpretive theory of the Katzian type. I am aware that many aspects of figurative language in volve pragmatic considerations like situation, speaker and addressee beliefs etc. But I believe that much can be ac complished in a purely semantic treatment. And some sen tences are figurative whenever and wherever uttered by whoever, as Black (1962: 29) notes. 5.1 Anomaly
and
Selection
If regular relational principles are to function in such a manner as to describe consistent interpretations for sentences which would otherwise be labeled anomalous on the relevant reading, some specification of the notion of ano maly must be provided. Selectional restrictions within the semantic representations of predicates are intended to fil ter out inconsistent readings for sentences. Pretheoretically, however, it is not clear what should count as sem antic inconsistency. At the one extreme, semantic theory must draw a line between truly semantic inconsistency arising from the meanings of the linguistic items and their relations'alone and inconsistencies recognized on the basis of our knowledge about the extra-linguistic world. At the other extreme, the theory must establish a boundary be tween semantic inconsistency and logical inconsistency. As regards the former case, we want semantic theory to characterize sentences like (1) as anomalous, i.e. semantically inconsistent, but not sentences like (2) which are
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
140
contingent. (1) (2)
The rock ate the idea. The rock weighed less than the same volume of helium.
Sentence (2) describes a state of affairs inconsistent with our knowledge of the physical world; most people would be prepared to label it false without further ado. But (2) is far from being non-sensical in the way (1) is. We under stand (2), and can easily imagine the sort of evidence ne cessary to test its truth-value. Sentence (1), however, can be assigned no immediately obvious interpretation. (1) clearly falls into the class of anomalous sentences, while (2) is presumably contingently false, but at any rate meaningful. Other cases are more difficult. Sentence (3), for in stance, might be considered anomalous. (3)
Tony ate my rocking chair.
If semantic theory foresees features such as (Non-edible) attached to lexical items naming artifacts like furniture, or if a feature redundancy rule automatically adds this feature to any matrix bearing the feature (Furniture), then (3) could be labeled anomalous. But note that it is indeed physically possible to consume a chair, say in the form of sawdust over a period of time. And if a person did eat a chair in this manner, (3) might represent a perfectly fine description of what had taken place. And if something can happen, a person must be able to describe it non-anomalously. So sentence (3) should not be labeled anomalous. Consequently, lexical items should not be characterized by the feature (Non-edible), either directly or redun dantly. By the same reasoning, the feature specification for eat ought not to contain a selectional restriction limit ing its objects to nouns with a feature like (Edible). Re call the scene from Chaplin's Gold Rush in which Charlie and his room-mate cook and eat Charlie's boot. Clearly semantic theory must be constructed in such a way as to allow a meaningful description of this scene. We thus ar rive at the following principle for determining the scope of selectional restrictions: Anything which can happen in the world as we perceive it must be describable without anomaly. This principle bars the door to a proliferation of features aimed at building extra-linguistic knowledge into semantic theory. It not only limits the range of
DERIVED POLYSEMY
141
features allowed as part of lexical entries proper, but also urges great care in the formulation of feature redun dancy rules and devices such as meaning postulates. This helps us maintain a clear distinction between system-internal information and information about the ex tra-linguistic world. If we are tempted to add a selectional restriction to the verb marry limiting its permis sible subjects and objects to nominals characterized as (Adult), (Non-infant) or even (Living), we are well ad vised to consider first whether a marriage is conceivable between children or infants, even if born dead, say in the case of a politically motivated union In the Middle Ages. Morgan (1969) advocates the consideration of such bizarre but possible situations in arguing for semantic analyses; Morgan's methodology is quite in keeping with the state ment above that anything which can happen must be describable without anomaly. Again, the stipulation that marry be restricted to subject-object pairs characterized as (Female) and (Male) respectively, or vice versa, besides being extremely cumbersome to express as a selectional re striction in any normal format, must be eschewed, because it would forbid meaningful description of homosexual mar riages. There exists further the problem of historical changes in the body of extra-linguistic knowledge. The semantic characterization of shark as (Fish) or atom as (Indivisible) illustrate the danger of including world knowledge in sem antic descriptions. We must rigorously avoid methods lead ing to the conclusion that a sentence such as (4) is anom alous. (4)
This cyclotron splits atoms.
Thus selectional restrictions on predicates limit them to occurrence with nominals of which they can be meaning fully applied to yield propositions1 about possible states and events in the world. Anomaly arises when the selec tional restrictions a predicate imposes are not met by a nominal in construction with it. An anomalous sentence ex presses neither a (logical) contradiction nor a contin gently false proposition. It expresses no proposition and counts as neither true nor false, but is assigned instead a third value 'non-sense' (cf. Katz 1972 and below). As pointed out above, it does not suffice to set sem antic information off from real world Information, one must also distinguish semantic information from logical content. One must, that is, differentiate anomaly from contradiction. Inasmuch as anomaly correlates with impos sible states in the world we know, it appears correct to
142
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
correlate contradiction with states which cannot obtain in any possible world. This view of contradiction coin cides with that Katz (1972) offers. Following Katz, a sentence like (5) is correctly characterized as a contra diction rather than an anomaly. (5) John has a hairy bald head. Hairy must not bear a selectional restriction limiting it to nominals specified as (Hairy). Hairy predicates hairi ness of objects rather than presupposing them hairy by way of selectional restrictions. In a similar vein, an adjective like square may bear a selectional restriction choosing nominals specified as (Geometrical), but not (Four sided); we want sentence (6) to be labeled a contradiction rather than anomalous. (6) Al observed the square circle. The two examples cited so far are relatively transparent, but they suffice to illustrate the principle that we must be on our guard when proposing selectional restrictions in order to preclude feature specifications which result in a confusion of anomaly and contradiction. Treatments of the verb read provide a case in point. Levin (1977: 38 f f . ) , following Katz and Postal (1964: 83), takes the interpretation of (7) as meaning that the some thing being read is a physical object with writing on it to be sufficient reason for assigning read the selectional restriction [(Physical Object with writing on it)] on its direct object nominal. (7) The man is reading something. Levin then proceeds to argue that feature transfer of this selectional restriction into the feature complexes for ob jects of read will account for the usual interpretation of sentences like (8), on which the manhole cover has writing on it. (8) John was reading a manhole cover. Notice that including this selectional restriction in the specification for read forces the conclusion that sentence (9) is anomalous, because the direct object of read (after the application of the projection rule to amalgamate the material following it) does not match its selectional re striction.
143
DERIVED POLYSEMY (9)
Judy read the page without seeing any writing on it.
But according to Katz (1972: 180 ff. ) , (9) should be char acterized as a contradiction; it meets his definition (4.122), just as his example (4.123e) does, here repeated as (10). (10) The towel soaked the floor without wetting it. The way out of this dilemma was indicated by the examples of hairy and square above. The information that writing is necessary to read must be a part of its lexical read ing and not a separate selectional restriction. So I would analyze read with a simplified selectional restriction and a slightly complicated feature description. If Figure 1 represents the lexical reading for read as viewed by Katz and Postal (1964) and Levin (1977) in the formalism of Katz (1972), then Figure 2 represents my modified version of the critical portion (which is all that would be affected). Figure 1 : read (((Action)
(((Visual) & (Mental)) [NP, VP, Pred-Phrase, S]
((Purpose)
((To interpret
X
)))))
[(Object) (Physical) (with letters on it)] [NP, S] X
)
[(Human)]
Figure 2 : [NP, VP, Pred-Phrase, S] (To interpret letters on
X
)
[ (Object) (Physical) ] If Figure 1 is revised as Figure 2, then sentence (9) turns out to be a contradiction, which intuition tells us it must be, rather than anomalous. The revision of Figure 1
144
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
also accounts for the interpretation on which the manhole in sentence (8) bears writing, since the verb read supplies this information directly as part of its content. It af fects but does not eliminate the feature transfer from read to the proform something in sentence (7); now the simpler feature [ (Object) (Physical) ] is transferred. Nevertheless, by eliminating the transfer in cases like sentence (8), the analysis of read offered here results in a sightly simpler total theory. The thrust of the pre sent discussion has been to warn against placing content features in selectional restrictions; doing so may lead to a confusion of anomaly and contradiction, and an over all complication of the theory. To summarize both the foregoing sections, we have seen the necessity of maintaining a clear distinction between anomaly, contingency and contradiction. The linguistic in formation determining anomaly must be kept separate from the world knowledge determining contingent falsity. Especial ly, we must avoid placing extra-linguistic information in selectional restrictions, just as logical content must not be allowed to migrate into selectional restrictions. Over all, the discussion points in the direction of less strin gent selectional restrictions with a consequent decrease in the set of anomalous sentences. 5.2 Lexical
Readings
In keeping with the objectives of distinguishing sys tem-internal from extra-linguistic information and simpli fying semantic description generally, I propose to char acterize lexical entries solely in terms of truly distinc tive features, i.e. features obtainable by investigating the set of relations any lexical entry contracts with all other entries in the lexical system of a language. Reli ance on purely distinctive features should tend to simpli fy lexical representation generally, and to discourage intrusion of extra-linguistic information into the lexi con, which will help preclude a needless proliferation of features. This methodological restriction does not, however, limit us to comparing lexical items pair by pair to deter mine the set of distinctive features. We can bring various sorts of system-internal evidence to bear on questionable cases. In the words of Katz (1972: 249), semantic theory "Interrelates semantic concepts and, thus, permits us to decide on the character of an unclear case by theoretical triangulation from clear cases of sentences having other semantic properties and relations." Consider an example
DERIVED POLYSEMY
145
Katz (1972: 2 49-2 50) employs for a somewhat different pur pose. If we wish to determine whether the nouns fish and whale should differ in distinctive feature representation, without recourse to ichthyology, and find speakers pre pared to assign both to the category of aquatic animals, we can construct an example of the following type. Since fishing may be characterized as the activity of trying to catch fish, and but-conjunction requires the conjoined expressions to contrast semantically, then (11) should be anomalous. (11)
Sue went fishing but caught a fish (bass, pike, etc.) instead.
To avoid anomaly, the object of the second clause must contrast with fish. Hence, (12) presents no anomaly. (12)
Sue went fishing but caught an octopus instead.
But notice that (13) excapes anomaly in the same way as (12). (13)
Sue went fishing but caught a whale instead.
Thus, the lexical reading for whale must contain a feature differentiating it from fish. Katz's "theoretical triangulation" permits the dis covery of distinctive features on the basis of information internal to the system; it should suffice to uncover dis tinctive differences in unclear cases. In cases where the oretical triangulation and the ingenuity of linguists fail to reveal distinctions, we will simply have to be satis fied with synonymy. If a theory which restricts meaning differences to distinctive features ends up with an in creased number of synonymic pairs, then it can only be be cause the differentiation of synonyms in other theories derives from their inclusion of information external to the language system proper. As an example, on the basis of an informal poll of speakers and investigation of texts and dictionaries, no consistent, significant, synchronic distinction can be drawn between hate and detest along semantic lines. Nor does an appeal to hated and detested, or hateful and d e t e s t a b le offer any useful possibility of triangulation. These two verbs must then receive identical lexical readings and selectional restrictions. Many other pairs will necessarily receive identical semantic represen tations. The semantic component of a grammar as such will characterize both (14) and (15) as contradictory sentences on Katz1s (1972: 178 ff.) definition. The resolution of
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
146
this contradiction must be left to more broadly based descriptions of linguistic behavior. (14) (15)
Sue does not just detest Sam, she hates him. Sue does more than hate Sam, she detests him.
Further, if we limit lexical readings and selectional re strictions to system-immanent information, the semantic component will not be in a position to characterize a sen tence like (16)
The cow jumped over the moon.,
as anomalous, since there exists no reason on purely lin guistic grounds to add a selectional restriction to jump over to prohibit its occurrence with objects characterized as celestial bodies. Hence, (16) will be assigned a read ing without application of any figuration rules. The de termination of (16) as contingently false is not incumbent upon the semantic theory as such.
5.3 Regular Theory
Relational
Principles
in Interpretive
Semantic
So far in this chapter we have discussed several prin ciples either modifying the interpretive semantic theory described in Katz (1972) in relatively minor respects or simply clarifying assumptions that theory leaves inexpli cit. They are as follows. (1) Lexical readings consist entirely of distinctive features. (2) These directly reflect the gap an entry occupies within the network of lexical semantic classes. (3) Selectional properties must be derived on the basis of evidence from the linguistic system proper. (4) Selectional restrictions should be simplified wherever possible in favor of more explicit lexical readings. (5) Anything which can happen in the world as we perceive it must be describable without anomaly. The major changes I would like to propose to the model Katz presents are the following. (1) Semantic theory includes a set of regular relational principles consisting of the five metaphoric and the seventeen metonymic principles defined in Chapter 3. (2) Semantic theory Includes a feature extension rule which copies selectional restrictions from predicates
DERIVED POLYSEMY
147
into the lexical readings of nouns they apply to in constructions which would otherwise be characterized as anomalous, subject to constraints proposed in 4.3.2. (3) Semantic theory includes two construal rules: one de leting from lexical readings any features clashing with the foreign features introduced by feature extension rule application, and one combining such foreign fea tures into the host reading when no clashes occur. (4) Strings classifiable as anomalous or contradictory sentences on the basis of the projection rule alone are not immediately filtered out of the grammar's output. (5) Strings of both kinds may receive consistent interpre tations due to the application of one or more regular relational principles preceding each successive appli cation of the projection rule. (6) Strings classifiable as anomalous on the basis of the projection rule alone may receive consistent inter pretations due to the application of the feature ex tension rule, one of the construal rules, and a reapplication of the projection rule. (7) Lexical entries are fully specified, including all re dundant information. (8) Redundancy rules do not fill in redundant features, they simply characterize certain portions of entries as redundant. (9) Since lexical entries are ordered into semantic classes, they are unitary, pairing one phonological with one semantic representation. These last three propositions pertain primarily to the structure of the lexicon and, as such, will not be dis cussed in detail until the following chapter. Consider first the functioning of the regular rela tional principles in this modified version of Katz's the ory. This theory foresees as an integral part of the gram mar of a language a semantic component which takes under lying phrase markers terminating in lexical items as its input and produces semantic interpretations for sentences as its output. Underlying phrase markers represent all the crucial structural relations holding between the lexical items they contain. Before a reading for such a structure can be formed, the lexical items in it must be assigned their particular lexical readings from the lexicon. Either these readings are carried along with each lexical item as it is inser ted, or readings matching previously inserted lexical items may be copied from the lexicon and attached to them.
148
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
In either case, once all the terminal lexical items in a base phrase marker have been assigned readings from the lexicon, we term it a lexically interpreted underlying phrase marker. Since our theory incorporates fully speci fied lexical readings containing the full set of redundant features, semantic redundancy rules do not apply product ively to expand lexical readings already inserted into phrase markers. Lexically interpreted underlying phrase markers con stitute the domain of the projection rule. The projection rule described in Katz (1972: 113 ff.) performs two func tions: it substitutes the reading for one item im place of the categorized variable in that of another item if the two bear a grammatical relation to one another, or simply forms their union if neither contains a categorized vari able. Given a predicate like die with a reading as in Fig ure 3, and a subject nominal like men with a reading like Figure 4, the projection rule substitutes the latter into the former to yield Figure 5. In schematic form: Figure 3:
die [NP, S] ( ( P r o c e s s ) , ( ) , . . . , ( . . . X)) [(Living)]
Figure 4:
man + Plural ((Living),(),...,())
Figure 5:
men die ((Process),(),..., (((Living), ( ) , . . . , ( ) ) ) )
To get the composite reading for men die daily, Figure 5 is simply coupled with the lexical reading for daily, since neither contains a categorized variable once men die has already been formed. The projection rule thus con tinues to reapply to successively larger constituents until a derived reading for the entire structure has been formed. We thereby arrive at what Katz terms a semantically interpreted underlying phrase marker. Katz then goes on to define the 'semantic interpre tation of a sentence' (1972: 47), which includes the set of semantic properties and relations associated with a sentence over and above Its semantically interpreted
DERIVED POLYSEMY
149
underlying phrase marker. But such relations need not con cern us here, and I shall employ the term 'interpretation of a sentence' to designate the final output of the sem antic component in cases where regular relational prin ciples or feature extension have applied along with the projection rule. Now, regular relational principles might apply at either of two points in a semantic derivation of this type: during lexical interpretation or during successive appli cations of the projection rule. If previously inserted lexical items are matched with lexical entries and as signed lexical readings from them, as in the second option described above, relational principles could apply direct ly to these lexical readings before they are assigned to items in phrase markers. They would apply optionally or, in cases where feature clash was imminent, obligatorily. Alternatively, they might apply obligatorily at points where clashes appear following lexical interpretation, and optionally at any level before the projection rule applies to it. This alternative presupposes that structures are not immediately filtered out whenever a selectional viola tion occurs in them, hence the inclusion of principle (4) among the major changes proposed to Katz's theory above. Either alternative could handle sentences such as (17), where the selectional violation must appear during the process of lexical interpretation as soon as the lexical reading for house is added to the phrase marker containing the reading for applaud, which must enforce a selectional restriction [ (Human) ] on its categorized variable. (17) The entire house applauded loudly. In either case, Metonymic Principle 13 (locality-occupant) can apply to the reading for house to add a complex like (Human) (Occupant), which renders the reading compatible with the selectional restriction. Problems arise for the first alternative, however, when sentences such as (18) are considered. (18) Harry knocked over what Tony had brewed. Following lexical interpretation, what is still a proform. It contains no features which could clash with the selec tional restriction imposed by knock over. Nor can any reg ular relational principle apply, since a proform empty of all feature content cannot meet any of their conditions of application. In order for a relational principle to become applicable to what , it must be imbued with some
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
150
content. Ignoring irrelevant details, the underlying phrase marker for (18) can be represented as Figure 6: Figure 6 : [Harry s
knocked over PRO.1 [s1Tony brewed P R 0 2 ] ] ,
where the two PROs are coreferential. First, the selectional feature [(Liquid)] from brew is assigned to P R 0 2 . Then successive applications of the projection rule amalgamate the readings in S1 into a single derived reading. Now a further application of the projection rule combines PRO 1 with the derived reading for S 1 , so that PRO1, is derivatively associated with the feature (Liquid) on the basis of its coreferentiality with PRO 2 . Finally, at this stage in the semantic derivation, Meto nymic Principle 12 (container-content) can apply to PRO-] adding features such as (Container for) (Solid) (Rigid), which renders PRO1 capable of meeting the selectional re striction defined by knock over. Failing this sort of de rivation, PRO-] will be assigned the feature (Liquid) and PRO 2 the feature (Rigid), which necessarily leads to a feature conflict once these two forms are amalgamated. Since the normal interpretation of (18) can be paraphrased as (19), (19)
Harry knocked over the container full of what Tony brewed.,
the semantic derivation must be essentially correct. This means that regular relational principles must ap ply to the output of the projection rule in some cases. Unless we are willing to accept both alternatives as co existent possibilities, it seems we must accept the second as the correct analysis. So regular principles will apply to lexically interpreted underlying phrase markers pre ceding applications of the projection rule to successively larger constituents. Consider one further example of the application of reg ular relational principles. Sentence (20) will be charac terized as a contradictory sentence, because the feature (Human) included in the lexical reading for mother contra dicts the feature (Geographical formation) included in the lexical reading for mountain. (20)
Sue's mother is a mountain.
Notice first that the feature extension rule cannot effect
DERIVED POLYSEMY
151
any sort of feature exchange in (20). Sentence (20) con tains no selectional violation; it is contradictory rather than anomalous. Nevertheless, Metaphoric Principle 4 (ob ject-feature) with Transformation 2 can operate on the lexical reading of mountain, which must contain the fea ture (Large) in order to differentiate it from h i l l , delete ing all but the feature (Large). Once only the feature (Large) remains, there is nothing to conflict with any feature on the noun mother. So application of Metaphoric Principle 4 allows deri vation of a consistent derived reading, or, in my termino logy, semantic interpretation for sentence (20). In fact, the semantic interpretation derived with the help of Meta phoric Principle 4 coincides with the interpretation speakers are generally prepared to offer for sentence (20). This fact supports not only the inclusion of Metaphoric Principle 4 among the regular relational principles to be included in semantic theory; it also provides a rationale for including principles (4) and (5) among the major changes proposed above to the theory elaborated in Katz (19 72). The agreement of the semantic interpretations of sentences (18) and (20) offered here with native speaker judgments concerning the interpretations of these sentences speaks strongly in favor of a semantic theory incorporating principle (4) among the major changes proposed to Katz's theory, and the respective metonymic and metaphoric prin ciples cited in the derivations outlined. In both cases cited, feature transfer or extension, no matter how defined, could not have delivered any alterna tive readings at all, let alone the readings substantiated by native speakers. In the case of sentence (18), feature extension cannot apply because no anomaly presents itself at the level of lexical interpretation; in the case of sentence (20), feature extension cannot apply because anomaly as such does not occur at all. Clearly then, a semantic theory allowing for only feature extension will fail to account for a wide range of commonly found figura tive interpretations. Required are regular relational principles which operate even in the absence of anomaly, and hence optionally in sentences which receive consistent readings otherwise. For instance Metonymic Principle 13 (locality-occu pant) can apply in the case of sentence (21) to get the interpretation on which people of the Negro race occupy seats in the rear end of the bus over and above the nonfigurative reading on which the rear end of the bus is painted black. (21) The back of this bus is black.
152
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Nevertheless, feature extension rules provide intuitively correct interpretations for certain sentences, so I prefer a theory containing them alongside regular relational principles. A semantic theory combining regular relational prin ciples with feature extension rules can account for the entire range of everyday figurative interpretations, and for much poetic figurative language as well. (22) The fishing boats danced in the harbor. Sentence (22) would be classified as anomalous by a sem antic theory relying entirely on the projection rule for deriving semantic interpretations, because boat does not contain the feature (Human) required by the selectional restriction dance imposes on its subject nominal. Metaphoric Principle 4 (object-feature) along with Transfor mation 2 might apply to the lexical reading for dance deleting all the features except (Rhythmic Movement), thus creating a semantic interpretation paraphrasable as (23). (2 3) The fishing boats moved rhythmically in the harbor. But there is a further reading for (22) on which the boats have legs and other human properties and dance retains its lexical reading; we might say that the selectional restric tion on dance causes an anthropomorphization of the sense of boat. Since interpretations of this sort typically occur in the text type 'fairy tale', I label them fairy tale readings (cf. Norrick 1979b). None of the regular relational principles proposed above can create fairy tale readings, but feature extensions naturally bring them about. The feature extension rule can copy the selectional feature [(Human)] from dance into the lexical reading for boat. A construal rule then integrates this feature into the reading, thereby deriving the anthropomorphic inter pretation of boat in (22). This appears to exhaust the range of meanings speakers normally report for (22), although some speakers identify a reading essentially midway between the metaphorical and the fairy tale readings outlined. That is, they interpret boat as partially anthropomorphized or animated, and dance as somewhat broadened in sense. The present theory will not describe such an interpretation explicitly, but simply make both distinct semantic interpretations available to a broader, more psychologically oriented theory of text interpretation which can take into account double pro cessing of a single structure.
DERIVED POLYSEMY
153
For the present we can assume that speakers derive double or multiple interpretations for a single structure, which they then weigh against each other in light of con text to develop what we may call a contextual interpreta tion. The process of textual interpretation may involve discarding one or more of the readings made simultaneously available by semantic theory or, as in the case just cited, may combine one or more readings into a composite inter pretation. It suffices here to allow for the double or multiple processing of sentences; in formal terms this simply means that the semantic component does not necess arily discontinue its processing of a sentence once a single consistent reading has been obtained. Actually, the psychological reality of multiple processing is indubitable, given the fact that the success of puns and so-called poetic ambiguity derives from people's ability to recog nize multiple interpretations for linguistic structures. Let us look at a further example where regular rela tional principles interact with the feature extension in the processing of a sentence. In Shelley's To Jane: The Recollection we find the line (24). (24)
The whispering waves were half asleep.
Observing first the unit whispering waves, we note a selectional violation arising from the fact that wave fails to manifest the feature (Human) required by whisper. We might here apply the feature extension rule to copy the selectional restriction [(Human)] from whisper into the lexical reading for wave, obtaining thereby an anthropo morphized interpretation for the latter. Once wave bears the feature (Human), its combination with asleep becomes possible without anomaly as well. Here a single applica tion of the feature extension rule and the construal rule integrating an extended feature into its host lexical reading suffices to yield a consistent fairy tale inter pretation for the whole sentence on which waves are viewed as possessing human properties. Alternatively, Metaphoric Principle 4 could delete all the features from whisper but (Quiet). This interpretation of the unit whispering waves will have no effect on the anomaly of combining waves with asleep. Again at this point we might invoke the feature extension rule to trans fer the selectional restriction [(Animate)] from asleep to the lexical reading for waves. This yields an interpreta tion In which quiet animate waves are half asleep. If Metaphoric Principle 4 applies to asleep as well as to whisper, deleting all but the feature (Inactive) from the
154
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
former, we obtain an interpretation paraphrasable as (25). (25) The quiet waves were nearly inactive. One further interpretation becomes available, if the feature extension rule applies to the unit whispering waves, while Metaphoric Principle 4 applies to asleep in the manner just described. The result is an interpretation on which the anthropomorphized waves are nearly inactive. Other less likely interpretations might also be de rived if other regular relational principles applied to either or both predicate terms. Given the possibility of multiple processing and the four alternative interpreta tions described, the semantic component outlined here de livers a wide range of possible interpretations for a single line of poetry. It appears to account for all the interpretations a reader of (2 4) might be prepared to offer. In any case, it derives all the interpretations a theory such as van Dijk's (1972) or Levin's (1977) would, in addition to being able to describe interpretations, especially those involving metonymy, which those theories cannot. Nevertheless, the foregoing example was not in tended to argue the adequacy of the theory presented here to provide interpretations for lines of poetry, but simply to demonstrate the functioning of feature extension rules and regular relational principles in an interpretive sem antic theory enriched to describe various sorts of figu rative interpretation.
THE LEXICON
6. 6.0
Introduction
Polysemy is not the only semantic relation to be ex pressed in the lexicon. This chapter presents a theory of the lexicon intended to capture regular semantic relations of all kinds. I discuss the role of semantic classes in the lexicon, first in their function of predicting certain lex ical relations directly, second as they relate to lexical readings. Then we turn to the consequences regular rela tional principles have for the lexicon. Semantic via-rules link entries in non-adjacent semantic classes on the basis of these principles. Their effects on the lexicon are also outlined. In this framework I go on to explicate the notion of lexical polysemy, and the problem of lexicalization.
6.1 Semantic
Classes
and Relations
in
the
Lexicon
We can envisage the lexicon as a structured network of intersecting, semantically defined classes. The classes reflect distinctions we make in segmenting our experience. They range from the broadest of distinctions, such as that between objects and events, to delicate taxonomies within classes like animals, professions, tools etc. Clearly then, we must allow for the possibility of one class's inclusion within another. Also, since one entry for bachelor, for instance, must occur in the class of humans to pair it with person, in the class of males to pair it with man, and in the class of unmarrieds to pair it with widow., crossclassification is necessary. This allows for cases of in tersection between classes where neither contains the other. There will be completely disjunctive classes as well, because we require features such as (Intentional) and (Durative) for classifying items describing actions, but not for items designating objects, just as (Opaque) and
156
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
(Fruit) apply to objects but not to actions. There are then relations of containment, intersection and disjunc tion between the semantic classes in the lexicon. In addi tion there exists the possibility of union, which from another point of view might be considered mutual contain ment. These four sorts of class relations translate directly into semantic relations between lexical entries. I define semantic relations between lexical entries rather than be tween lexical items, because a single lexical Item may be associated with multiple related entries. Since a single lexical entry contains only a single lexical reading in the present theory, the relations may be thought of as holding between readings as well. I prefer entries to read ings here simply for ease of formalization below. Let A1 , A 2 . . . A x represent semantic classes in the lexicon, where any A n properly contains any A m if n < m. Further, a lexical entry p is exhaustively contained in a class A, if and only if A contains p, and A neither con tains nor intersects with any other class containing p. We can define the lexical semantic relation of hyponymy as: Definition 1 : If a lexical entry p is exhaustively contained in the class A n , and a lexical entry q is contained in Am, where n < m, then p is a hyponym of q. Examples of hyponymy are pairs such as horse-palamino colored-green,
walk-stroll.
,
In each case the latter term
is contained in all the classes containing the former term and in at least one further class which is itself contained in the former class. Within the domain of what has been traditionally termed synonymy, one can recognize two separate relations. The first case I label complete synonymy, and define as follows. Definition 2: If a lexical entry p and a lexical entry q are both exhaustively contained in a class A, then p and q are complete ly synonymous. The second case may be termed partial synonymy. I define it in class relational terms as Definition 3. Definition 3: If a lexical entry p is exhaustively contained in A x , and a lexical entry q is exhaustively contained in A , and
THE LEXICON
157
A x overlaps with, but neither contains nor is contained by Ay, then p and q are partially synonymous. Complete synonymy is demonstrated by pairs like pocketbookpurse, hate-detest, big-large on the relevant readings, i.e. on the reading of pooketbook where it means 'handbag', and ignoring that where it means 'paperback book'. The de finitions (2) and (3) are cast in language which automati cally limits them to lexical entries and readings rather than lexical items. It makes no sense to state without further specification that the lexical item or word pocketbook is in complete or partial synonymy with purse, but only that some lexical entry or reading is in synonymy with another. We certainly do not want to say that pooket book is in partial synonymy with purse on the basis that one of the lexical entries associated with this word form shares its classification with purse. We see a relation of partial synonymy between the relevant entries for lamp and light, begin and initiate, again, ignoring the reading for light on which it designates a form of energy and that for initiate on which it designates a ritual act. The fourth relation, that of disjunction, can be de fined simply as the absence of any of the three foregoing relations (and of those which follow defined on the basis of class membership). Disjunction is therefore not a rela tion as such, but rather the lack of a relation. Any two lexical entries with no features in common on the basis of their semantic classification are disjunctive. Examples are non-pairs such as sincerity-ant and enter-nail. Dis junction must not be confused with distinctness. Distinct ness obtains any time two lexical readings differ by one or more features. Nor does disjunction equate with oppo siteness of meaning, to which we now direct our attention. Oppositeness of meaning can be of several types, but all involve lexical entries which agree in their member ship in all but a single class. Among the semantic rela tions exhibiting oppositeness In meaning, complementarity (binary opposition), antonymy (gradable opposition), converseness, and incompatibility, as they are defined in Lyons (1977, 270 ff.), can be usefully treated in terms of semantic classes. Lyons identifies complementarity with the binary opposition shown by the pair female-male . Here, the presence of either precludes that of the other; the negation of either Implies the positive presence of the other within the class of entities to which they apply. We can define complementarity on the basis of semantic classes as follows.
158
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Definition 4: If a lexical entry p is contained en tirely in the classes A-j , A2 . . . A n , A x , and a lexical entry q is contained en tirely in the classes A-] , A2 . . . A n , Ay, where all A n is either A x or Ay, then p and q are in complementarity. Lyons limits antonymy as such to gradable opposition. Unlike the ungradable or binary opposition shown by femalemale, gradable pairs such as b ig- little do not exhaust the whole range of entities in classes to which they apply. So not big does not necessarily imply l i t t l e , although little does imply not big. We can thus define antonymy in Lyons' sense as: Definition 5: If a lexical entry p is contained en tirely in the classes A-¡ , A2. . . A n , Ax, and a lexical entry q is contained en tirely in the classes A-] , A2 . . . A n , Ay, where A x and Ay represent diametrically opposed classes within A n , but all A n is not necessarily A x or Ay, then p and q are antonyms. Converseness of meaning Lyons exemplifies by pairs like wife-husband, doctor-patient. Here problems arise, because while wife and husband will occur in the semantic class of marriage partners, there is no natural semantic class which will contain both doctor and patient. Now if wife and hus band exhaust the content of the class of marriage partners, they turn out to be complementary on the basis of Defini tion 4. And there appears to be no good reason to reject this conclusion. The pair doctor-patient, however, fits neither Definition 4, nor any other definition in terms of semantic classes alone. The relevant relation can be cap tured by means of via-rules, and I postpone its discussion to a later section. The relation of incompatibility is that illustrated by any combinations of items from sets like Monday, Tuesday . . . Sunday, and red, orange. . . violet. The applica^ bility of any member of such sets precludes that of any other member, but the inapplicability of any one simply im plies that some other must be applicable (this latter only holds for the set of colors, when xohite and black are in cluded as well). I define incompatibility in semantic class terms as follows.
159
THE LEXICON
Definition 6: If a lexical entry p is contained en tirely in the classes A*] , A2. . . A n , Ax, and a lexical entry q is contained en tirely in the classes Ay, A2- . . A n , Ay, where A x and Ay represent any two of a set of three or more distinct, exclusive classes in A n , then p and q are incom patible. This exhausts the set of semantic relations which can be directly deduced from a lexicon structured around sem antic classes, given appropriate definitions of such rela tions. A complete semantic description of the vocabulary of a language will demand further relations, but it speaks in favor of a lexicon structured around semantic classes that it can handle even these. An unstructured lexicon will represent no semantic or any other relations whatso ever. Given the proper definitions, a lexicon of the type discussed here can also provide direct representation of so-called word fields or semantic fields, and certain nat ural classes of lexical entries (e.g. all cognitive pre dicates, all purposive verbs). The treatment of other reg ular semantic relations must await the description in terms of via-rules below. 6.2 Semantic
Classes
and Lexical
Readings
A lexicon structured around semantic classes, as shown in the previous section, is desirable in that it repre sents various semantic relations in a direct manner. But we have also seen in the foregoing chapter that lexical entries complete with readings, and selectional restric tions in the case of predicates, are required by the rules of the semantic component. In order to prevent duplicate representations of much semantic information, readings for lexical entries should derive from their lexical classifi cations. If this can be accomplished, either lexical read ings need not be explicitly expressed at all or they may be characterized as redundant on the basis of the classi fication an entry receives. For nouns there is no special problem. Lexical entries for nouns simply receive readings which consist of the set of features corresponding to the classes to which they belong. These features will be assigned in the order dic tated by the hierarchy to which they belong. If a noun ap pears in the ordered classes of objects, physical objects, animate objects, humans, females and adults, as does woman,
160
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
then it will receive the feature specification (Object) (Physical) (Animate) (Human) (Female) (Adult). All the structure required in the lexical reading is supplied by the hierarchical ordering of the features inherent in the classification of the entry. Verbs and adjectives present a problem, insofar as they must distinguish content features from selectional features. We must therefore recognize a distinction in kind between the sorts of classes to which verbs and ad jectives belong. The relevant distinction can be seen in the difference between classifications which narrow only the content of a predicate as such and those which also narrow the range of entities to which this content may meaningfully pertain. A verb such as theovize will be classified as designating an act which is intentional, purposive, mental etc. But all intentional, purposive acts are necessarily acts performed by animate entities, and mental acts must be performed by human entities. Therefore, theovize should be classified among the animate acts be fore it is classified as intentional or purposive, and among the human acts before it is classified as mental. We might simply proceed to classify acts as intentional and purposive without a previous classification as animate, if we specify that any predicate so classified automati cally receives the selectional restriction [(Animate)], and specify that predicates classified as mental receive the selectional restriction [ (Human) ]. But since I want to classify not only nouns but also predicates in such a man ner as to insure that all their relevant semantic proper ties are ascertainable from their classification alone, I prefer to include classes which further define acts as animate acts and human acts. At the same time, such classi fication will be recognized as introducing information beyond that required to narrow the content of predicates as such. In particular, they delimit their range of mean ingful application. So we should stipulate that the clas sification of a predicate as an animate act or human act, given its previous classification as an act, necessarily entails its inclusion of the selectional restrictions [(Animate)] and [(Human)] respectively. The same stipula tion can be extended to other classifications which narrow the range of applicability of predicates primarily or in addition to narrowing their content as such. Since predicates enforce selectional restrictions on their direct objects as well as on their subjects, we must also stipulate that classifications of predicates limiting their range in terms of objects bring with them selectional restrictions on their direct object nomináis.
161
THE LEXICON
Given a predicate like telephone , not only the classifica tion as a human act is necessary, but that as an act di rected at a human, call it human directed, is relevant. A lexical item such as elect is similarly human directed, just as kill is an act directed at animate entities. Since elect will be distinguished from choose by its classifica tion among human directed acts, we can count on this clas sification to supply the selectional restriction [(Human)] placed on its object nominal. In the same way, the classi fication of kill among animate directed acts, in order to differentiate it from destroy, will provide the selec tional restriction [(Animate)] on its object nominal. Since classifications like these are both necessary for a full semantic representation of predicates and suf ficient to predict their selectional properties, I con clude that complete lexical readings as well as selec tional restrictions for both nouns and predicates can be directly derived from their semantic classification in the structures of the lexicon. Nouns receive readings consist ing of the set of features corresponding to the classifi cations applying to them. Verbs and adjectives receive readings in the same manner; selectional restrictions de rive from their inclusion in classes specifying their directedness toward entities classified in certain ways. Let us consequently posit a lexicon in which entries are ordered into intersecting semantic classes which pro vide direct representation of the lexical semantic rela tions of hyponymy, complete and partial synonymy, dis junction, complementarity, antonymy, and incompatibility. The position a lexical entry occupies in the network of semantic classes determines its reading and selectional restrictions, which are redundantly expressed in it. 6.3 Regular
Relational
Principies
and the
Lexicon
Regular relational principles directly affect the model of the lexicon in two ways: (1) they support the thesis that the lexicon consists of a network of semantic classes, and (2) they predict additional readings for whole classes of lexical entries, which can therefore be excluded from the lexicon all-together or, at least, characterized as redundant. Regular relational principles affect the struc ture of the lexicon in a further, albeit indirect, manner, since they provide an inventory which via-rules may draw upon to connect entries. I return to this last point in the following section, and direct my attention first to the sort of support regular relational principles offer for a lexicon structured around semantic classes.
162
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Metaphoric Principle 3 (specific case-general rule) specifies the regular relation holding between an Item and the class(es) of which it is a member, where a class is viewed as the complete set of its members. In a lexicon which orders entries into appropriate semantic classes, this relation is directly expressed. Given any lexical entry, we can read off the set of classes of which it is a member. A lexicon which orders entries in some other fashion fails to explicitly represent this relation. Meta phoric Principle 5 describes the relation holding between any two members of the same class(es). Again, if the lexi con orders entries into semantic classes, this relation is directly expressed. The positions of entries in the system of lexical semantic classes alone suffices to determine their class relation. As we saw above, entries appearing in the same set of classes count as synonyms. The degree of semantic relationship two entries possess receives di rect expression in terms of the number of identical classi fications they share. Entries showing a high degree of cor relation necessarily inhabit the same area of the lexicon, so semantic fields correspond to areas in the lexicon. The second way in which regular relational principles affect the lexicon is through their prediction of addi tional readings for whole classes of lexical entries. Meta phoric Principle 2 (specific-general) for instance, will predict an optional generic reading for any lexical entry containing a reading for a count noun. Since Metaphoric Principle 2 applies to an entire broad class of entries without exceptions, the lexicon need contain no generic readings for lexical items at all. If such readings were included in separate entries in the lexicon, however, they could be characterized as totally redundant on the basis of Metaphoric Principle 2. In much the same way, Metonymie Principle 12 (con tainer-content) applies optionally without exception to any lexical entry in the class of containers predicting a reading for it on which its content is designated (or, in the opposite direction, to any lexical entry in the class of liquids predicting a reading on which its con tainer is designated). Consequently, a single lexical en try expressing the 'container' reading of items such as and bathtub suffices to predict the 'content' reading as well, given Metonymie Principle 12. Nevertheless cur rent English usage persuades us that we should recognize a difference in the status of the 'content' readings for and bathtub. Standard dictionaries of English list both 'container' and 'content' readings for cup, but only a 'container' reading for bathtub. Given the standardization
THE LEXICON
163
of the cup volume in the English system of measures, and its wide use in recipes, the 'content' reading of cup has achieved a currency which dictates its lexicalization in a separate entry. That is to say that we can justify the existence of a particular entry in the lexicon on psycho logical grounds, even if it is completely predictable on the basis of the semantic system. The existence of a sep arately lexicalized reading for an item is psychologically justified, if the reading has achieved equal or greater currency than its source reading, or if a child or foreign learner of a language might learn the derived reading as primary without awareness of its origin. So I include the 'content' reading for cup in a separate lexical entry. Metonymie Principle 12 may be thought of as applying vacu ously to the 'container' reading for cup to derive its already lexicalized 'content' reading. I express this state of affairs by means of a via-rule, as discussed below. There are apparently no grounds for claiming that generic readings for lexical items are ever lexicalized alongside their particular readings, so I assume Metaphoric Prin ciple 2 must always apply to derive them. I consequently envision metonymie principles such as Metonymie Principle 12 as applying optionally and exceptionlessly whenever their conditions are met. At the same time, there will be lexical readings derivable by such principles which re ceive lexical entries of their own in a lexicon intended to model speakers' lexical competence. Let us then postulate a lexicon in which entire classes of readings need not be listed, because they are predict able on the basis of regular relational principles. In other cases where such principles could predict entire classes of readings, certain of these readings receive lexical entries of their own, since usage indicates that they have achieved a status of independence from their source readings. Regular relational principles therefore simplify the lexicon by extracting perfectly regular deri vations affecting entire classes of elements, although particular readings derivable from them are lexicalized as such. 6.4 Via-rules
in
the
Lexicon
In addition to the lexical semantic relations dis cussed above which can be captured in terms of the class membership of lexical entries, there exist regular seman tic relations which speakers recognize between entries not adjacent to one another in the network of semantic classes
164
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
within the lexicon. I have already indicated that such relations will be captured by means of via-rules attached to individual lexical entries and linking them with other related entries. At the same time, via-rules bridge the gap between lexical and derived polysemy, since they cite regular relational principles in semantic theory to estab lish connections between items in the lexicon. Above we encountered situations which illustrate the need for a device such as the via-rule, so we can concentrate in this section on the consequences via-rules will have for our model of the lexicon. A via-rule of the following form is attached to one lexical entry in order to relate it to another lexical entry x, where y identifies the rule or principle estab lishing the relation. Via-rule 1 : Related to x via y. The rule or principle identified by y can be stated as part of semantic theory or within the semantic component. Hence via-rules can link regular relational principles in semantic theory to individual entries in the lexicon. As an example, the lexical entry for the verb wax roughly meaning 'cover x with wax', where x stands for a cate gorized variable, will contain a via-rule like: Via-rule 2: Related to the noun -wax 'substance pro duced by bees' via Metonymie Principle 6 (instrument-act). Via-rule 2 characterizes as redundant the entire reading for the verb wax, since it is predictable on the basis of the reading for the noun, given Metonymie Principle 6. That is, the verb reading could be derived by Metonymie Principle 6 from the noun reading. Once the verb reading is lexicalized, via-rule 2 records the history of its derivation, so we need not derive a form anew each time it occurs, merely in order to express its relation to some other form. This illustrates one effect via-rules will have on the lexicon: via-rules make it possible to capture relations between alternate readings of lexical items without complicating the process of semantic interpreta tion. This provides the basis for a definition of the tra ditional notion of polysemy and for its differentiation from homophony, which will be our topic in the following section. With via-rules to capture relations between alternate readings for lexical items, decisions as to which readings
THE LEXICON
165
should be lexicalized as such and which should be derived can be made on a psychological basis. We encountered an example illustrating this effect of the inclusion of viarules in the lexicon above in the case of the alternate readings for cup. In parallel fashion to the phonological via-rules Vennemann (1972) proposes, semantic via-rules enable us to construct a lexicon which more closely ap proaches a model of the native speaker's organization of the vocabulary of his language. Vennemann further points up the function phonological via-rules fulfill in expressing connections which play no direct role in derivations. In Vennemann's theory, a speaker need not recognize any morphophonological connec tion between items such as sane and sanity, write and writ etc. in order to derive the correct surface forms for them. But via-rules may express such connections. As such, they represent a form of linguistic knowledge which goes beyond that included in competence proper. Semantic via-rules parallel to Vennemann's phonologi cal via-rules will express relations from original source readings to derived readings which have been lexicalized for centuries. The lexical reading for fox on which it means 'sly creature' has been an integral part of the vocabulary of English for many years, but the lexical entry containing it will still be affixed with a via-rule like (3), which reflects its historical derivation^. Via-rule 3: Related to fox 'wild canine' via Metaphoric Principle 4 and Semantic Transformation 2, which deletes all but the features (Sly), (Animate) and (Object). Nevertheless, a speaker unaware of the historical link be tween the two readings of fox might easily use and inter pret both correctly, so that it would be wrong to say that his failure to recognize the information expressed in viarule 3 affected his competence in English as such. Viewed the other way around, his addition of the information ex pressed by via-rule 3 to his store of linguistic know ledge and his subsequent recognition of the relation be tween the two entries for fox would also fail to influence his competence proper. This points up two consequences via-rules entail for the lexicon. First, given via-rules, the lexicon relevant for any individual speaker can represent relations reco gnized by this speaker without attributing them to compe tence as such, and without affecting our model of the lexi con for any other speaker or for a whole linguistic commun ity. Second, via-rules allow us to represent relations
166
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
between items on the basis of information which exceeds that required for complete description of the linguistic system itself without implying that this information is an integral part of the system. We can thereby maintain the distinction between system-internal and system-external information, but still account for relations (some) speak ers recognize (see chapter 8 for more on extra-linguistic information in via-rules). Also, the device of the via-rule affects our model of the lexicon in that it offers a way to account for the process of lexicalization. Vennemann (1972) notes that underlying source forms and phonological derivations viewed as quite transparent by one generation of speakers will blur and finally become entirely opaque for succeeding generations. What once was recognized as a productive de rivation reduces in time to a static relation. Once a de rived form is accepted in its own right, it is entered into the lexicon, but it carries the record of its deriva tion along with it. That is, the lexical entry for such a form bears a reference to its source and the rule(s) in volved in its derivation from the source. Now this is pre cisely the information expressed in via-rules. We can view the process of lexicalization of a derived reading for a lexical item or even that of a new morphological form and categorization derived from a lexical item in essentially the same way. At the same time that via-rules establish a link be tween productive semantic processes and static lexical re lations and thereby provide the basis for an account of the process of lexicalization, they essentially abolish the significance of the distinction between derived and lexicalized polysemy, on any but the psychological basis discussed above. Since the regular relational principles in semantic theory continue to be productive and available to derive readings, all the regular types of figures which have been lexicalized as such could be rederived. Similar ly, since via-rules express relations of the same kind the relational principles can derive, any connection which can be productively derived can also be described by means of via-rules directly in the lexicon. There is thus nothing in the system itself to dictate whether any particular reading for an item should be lexicalized or not. In fact, as we saw above, we might lexicalize generic readings for all count nouns, and we might derive anew whenever needed such common readings as the 'measure' interpretation of cup. From the point of view of linguistic description it self in the theory presented here then, the question of what is lexicalized and what is derived is neither answer able nor of any particular importance. Consequently, it
THE LEXICON
167
can and, indeed, must be answered on the psychological grounds outlined above. Finally, via-rules capture not only relations between alternate readings of polysemous lexical items, but also those between other pairs of entries, whether they manifest any morphological relation or not. Thus, the pair doctorpatient, mentioned above as an example of the relation of converseness in Lyons' (1977) classificational system, will be related by at least one via-rule as follows. The lexical reading for patient will be essentially 'one under treatment by a doctor'. A via-rule will be attached to this lexical entry linking it with that for doctor, thus characterizing the relevant portion of this entry as re dundant. Also, we can posit via-rules parallel to (2) above, which linked the verb and noun readings for wax, for pairs such as open-opener and even lock-key. In each case the same semantic relation of act-instrument (Meto nymic Principle 6) can be observed to hold. In the case of the pair open-opener, the via-rule expresses the seman tic relation which came about historically through a mor phological derivation of one form from the other. In the case of the pair lock-key, neither form was historically derived from the other, so the relevant via-rule simply expresses a regular semantic relation between the two, which allows us to characterize as redundant any features the two share. Hence, if lock is to use a key in a certain manner, then the portion of the lexical reading for lock which overlaps with that of key will count as redundant. Further, since via-rules can link morphologically com plex units with their component parts, they also provide a tool for predicting the degree of semantic motivation possessed by such units. We shall have much more to say on this topic in the next chapter. 6.5 Lexical
Polysemy
The inclusion of via-rules in the lexicon provides the semantic basis we require for a definition of lexical poly semy. If via-rules connect all and only those lexical lexientries with regularly related readings, we must re quire of a pair of entries assigned to a single polysemous lexical item only that they be connected by a via-rule and have identical phonological representations. We can thus define a polysemous lexical item as a set of phonologically identical lexical entries connected by via-rules. Since we have been concerned with the lexicon from a purely semantic point of view, we have had no occasion to discuss the phonological portion of the lexical entries,
168
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
nor the question of how one might go about determining phonological identity in them. For our purposes here, it suffices to assume either that phonologically identical lexical entries will be directly linked to each other in the lexicon, or that the set of entries phonologically identical with any given entry can be elicited from the lexicon in some manner. Both polysemy and homophony rely on the phonological identity of the lexical entries involved, but only poly semy requires that the entries be related by means of viarules. We can define homophony as the relation holding between lexical entries with identical phonological repre sentations but no via-rule connections. Homophony is thus not a semantic relation at all; It is simply a phonologi cal relation. Notice that we can say that two lexical entries are homophonous, but not they are polysemous. Instead, we say that two lexical entries belong to a single polysemous lexical item. Polysemy is naturally expressed as a relation holding between a set of lexical entries, rather than a set of readings within a single entry, because distinct lexical readings must appear In separate lexical entries ordered into different semantic classes in a lexicon of the type described here. The lexicon thus contains no branching entries, not even for polysemous lexical items. Let us consider as an example the polysemous lexical item fox. As pointed out above, there are two regularly related readings for fox, both of which are syntactically categorized as nouns. The historically primary reading for fox will appear in a lexical entry within the class of wild canine animals, whereas that for the historically de rived reading will appear in a lexical entry within the class of sly animate objects. These two classes are not adjacent to one another. A via-rule like (3) above con nects the two entries, however, so that, given the ident ity of their phonological representations, they turn out to be alternate readings of a single polysemous lexical item. But this lexical item possesses a third reading as well. On this reading fox represents a verb with a mean ing much like that of trick. This reading must appear in a lexical entry classified among those verbs describing purposeful animate acts. Nevertheless, it is clearly semantically related to the second noun reading of fox, in fact, Metonymic Principle 7 (agent-act) will serve to link the two. I posit Via-rule 4 attached to the lexical entry for the verb fox Via-rule 4: Related to fox 'sly creature' via Metonymic Principle 7.
THE LEXICON
169
So three separate lexical entries in two different syn tactic categories and clearly distinct semantic classes are related by via-rules. Since they are phonologically identical, we are entitled to assign all three of them to a single polysemous lexical item fox. Note that fox is polycategorial as well, to employ the term coined for such cases above, because it includes lexical entries in two different syntactic categories. Notice that, while the 'sly creature' reading for fox is directly related to both its other noun reading and its verb reading, the latter are not directly related by via-rule to each other. Never theless, since the 'sly creature' reading provides the logical link between the two, we can consider all three as related for the purpose of polysemy. 6.6 Remarks
on
Lexicalization
The term 'lexicalization' is polysemous in that it may be interpreted as pertaining either to the subjective lexi con every speaker of a language has internalized, or to the linguist's model of this portion of competence. It mirrors the ambiguous use of 'lexicon' and 'grammar', as these terms are understood in recent linguistic literature (cf. Chomsky 1965). With respect to the subjective lexicon, lexicalization involves a speaker's storing a unit consist ing of phonological, semantic and syntactic information. With respect to the lexicon model, lexicalization amounts to the linguist's inclusion of such a unit In the network of entries he views as comprising the vocabulary of some language. I shall try in the following to correlate these two interpretations of lexicalization. Armed with the notion of the via-rule developed above, I propose to describe the process of lexicalization from the point of view of the speaker as follows. A speaker understands new (transferred or extended) meanings for items already present in his subjective lexicon by virtue of general principles. These same principles are cited in via-rules relating lexical entries. Lexicalization of a new meaning for a familiar form involves storing the mean ing as a separate entry in the appropriate semantic field and setting up a via-rule relating it to the original meaning via the semantic principle involved In the deriva tion of the new meaning. A speaker's subjective lexicon may contain two unrelated entries semantically distinct but phonologically identical. If the speaker, as a result of reflection on the meanings or the acquisition of a new piece of information, recognizes a relation between the
170
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
two entries, he labels both with a via-rule establishing the connection. Once a regular relation is recognized, the speaker perceives the two entries as separate meanings of a single word. I therefore count phonologically identical, semantically related entries as belonging to a single polysemous lexical item. Phonologically distinct entries re lated semantically by a via-rule but with no morphological connection simply belong to two semantically related lexi cal items. The readiness of a speaker-hearer to lexicalize a par ticular contextual meaning for some word and the ease with which he accomplishes this depend upon a number of factors including the frequency of the occurrence of the word with the meaning in question in his experience, its degree of relationship with already lexicalized meanings of the word and the availability of alternate means of expressing the same essential meaning in his subjective lexicon. It is, of course, thoroughly possible that some speakers have lexicalized a certain meaning for a word while others have not. In fact, this typifies the situation in matters of lexis, the most idiolect-dependent area of language. This becomes strikingly apparent when children are compared with adults. A child five years old controls the phonology, morphology and syntax of his native language as well as adults, but his subjective lexicon will be very poor in comparison, both as regards the number of individual items present and the rules interrelating them. If a particular, as yet unlexicalized meaning for some word occurs so often in a hearer's linguistic experience as to facilitate his storing it as such or if it fills a (more or less perceived) gap in his vocabulary, he is ob viously likely to lexicalize the meaning as such rather than to rehearse its derivation anew at each occurrence. He may also store the form as a separate word, not real izing its relationship to other lexicalized meanings for the sound sequence until later or not at all. Since so many metaphorical and situationally determined extensions of a central word meaning are possible on the basis of general principles alone, it behooves a speaker to lexi calize those which are commonly used as separate entries. Even if he has the relevant principles of extension at his disposal, it seems natural to assume that many such ex tensions will be lexicalized as such. Reflection on areas of interpretation where the pro cess of assigning meanings to signs is more consciously performed, say in the hermeneutic process of interpreting literary works, clearly indicates that many expression-con tent relations are stored as such in the mind of the in terpreter rather than being derived anew in each instance.
THE LEXICON
171
I take this to be symptomatic of the manner in which such relations become stored in the lexicon as well. This re presents an empirical hypothesis about how the subjective lexicon is organized, and evidence in its favor has al ready been presented above. On the basis of this hypothe sis, the subjective lexicon of any speaker will clearly contain many items requiring multiple entries related to each other in various ways. Especially in cases of multiple meanings of polysemous items and idiomatic interpretations for morphologically complex units, semantic relations may be too complicated to be stated in the simplex form of the via-rule. This is because multiple meanings derive historically from multi farious, complex processes, and because derived meanings themselves continue to develop and even bifurcate. The statement of semantic relations receives additional com plications when extra-linguistic knowledge plays a role in establishing the connection between two lexical entries. I will return to such problems in Chapter 8 below.
MOTIVATION IN COMPLEX UNITS
7. 7.0.
Introduction
This chapter argues the thesis that via-rules provide the means of accounting for two types of motivation in morphologically complex units. First, by relating complex units to the separately lexicalized readings of their com ponents, via-rules account for cases of componential moti vation. Second, by relating the idiomatic readings of com plex units to their literal readings, via-rules account for cases of composite motivation. I begin by discussing morphologically complex units and their motivation gener ally, distinguishing componential from composite motiv ation. Then I elaborate the functioning of via-rules in accounting for motivation of both types. Finally, I will outline the treatment of stock phrases in the process of semantic interpretation. 7.1. Complex
Units
I reckon among the complex units of a language any and all stored items analyzable into components some of which appear in lexical entries of their own. I want to avoid saying that all the analyzable components must appear in separate lexical entries, since I wish to allow for units containing morphemes with severely limited or unique con textual occurrences (e.g. oran- in cranberry , kith in kith and kin). Within this class one can further distinguish compounds (e.g. egghead, coffee pot) from stock phrases on a purely phonological basis, because compounds exhibit characteristic intonational patterns; other complex units receive normal intonation. I define 'stock phrase' as any group of words stored as a unit either because (1) it re sists certain syntactic operations, or (2) its semantic interpretation cannot be derived directly given its struc ture and components. Examples meeting the first condition are phrases not derivable at all by regular syntactic
174
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
means like by and large, idiomatic phrases like red her ring (cf. herring that is red)1, irreversible binominals like ham and eggs (cf. eggs and ham), clichés like to all intents and purposes (cf. to all purposes and intents) , and quotes like blessed are the meek (cf. it's the meek
people that are blessed). Examples meeting the second con dition are phrasal verbs like put up with, tournures like up a creek,
and proverbs like make
hay
while
the
sun
shines. The classes are not completely exclusive. Red her ring, for instance, meets both conditions, since its inter pretation is clearly not derivable on the basis of its structure and components; i.e. it is non-compositional. I want to extend the distinction between units inter pretable on the basis of their structure and components alone and those not interpretable on this basis to com pounds as well. In both groups of complex units then, I distinguish compositional from non-compositional items. Items in the latter group are often labeled idioms (see Shaw 1979; cf. Kooij 1968). The distinction between idioms and compositional terms thus cuts across the other div isions mentioned so far. The compound egghead 'intellec tual' should count as an idiom just as put up with does, because neither is compositionally interpretable. I dif ferentiate compound idioms and phrasal idioms where there appears need to do so. In keeping with the principle that all items with a degree of currency in the language should receive separate entries in the lexicon, even when they are morphologically and/or semantically predictable on the basis of regular rules, I propose that all but nonce compounds be listed in the lexicon. Idiomatic compounds must appear in the lexi con in any case, because their readings are not derivable. In this respect they are exactly like simplex lexical items; hence Hockett's (1956) extension of the term 'idiom' to cover even simplex lexical units. Compounds with wholly or partially predictable forms and/or readings will also re ceive individual lexical entries. Like entries for lexical items generally, these may contain via-rules relating them to other entries where appropriate. Theoretical opinions on where phrasal idioms and stock phrases in general should be inventorized in a language description have varied widely, reflecting differing con cerns with and views of their semantic interpretation and syntactic treatment. Katz and Postal (1964) placed their idiom inventory in a special part of the lexicon, while Weinreich (1969) and Newmeyer (1972) opted for a separate entity. Chafe's (1968) theoretical approach dictated that idioms should be included along with other semantic units,
MOTIVATION IN COMPLEX UNITS
175
whereas Fraser (19 70) would list idioms in the main lexi con, given certain differences between their entries and the entries for other lexical items. The present treatment of phrasal idioms and of stock phrases generally urges listing all stock phrases in a special inventory, which may (but need not) be considered a subcomponent of the lexicon. I shall motivate this decision in the section on the semantic interpretation of stock phrases. 7 . 2 . Complex
Units
and
Motivation
We concern ourselves here with the motivation complex units derive from the readings of their component parts taken separately or as a literally interpretable whole. The phonetic or onomatopoeic motivation some words derive from the fact that their phonological forms manifest simi larities with sounds associated with the concepts they designate (see Ullmann 1957, 1967; Wescott 1971) will not be considered. In spite of claims like that in Fraser (1970: 33) that "no part of the idiom actually contributes to the semantic interpretation of the expression, once the idiom has been formed," many theories of both word formation (e.g. Lees 1960; Marchand 1969; Brekle 1970; Shaw 1979) and idioms (e.g. Chafe 1968; Makkai 1972; Newmeyer 1974) have at tempted to recognize and build upon just such contributions of component parts to the idiomatic reading of the whole. Chafe's (1968) division of idioms proper and semi-idioms, Fraser's own (1970) recognition of idioms as occupying various positions along a scale of frozenness, i.e. of re sistance to deformations introduced by syntactic operations, the differentiation of fully unmotivated idioms and par tially motivated or one-sided idioms such as tall order, where order retains its non-idiomatic sense (cf. Roos 1979), and similar distinctions all correspond to differing relations between interpretations of idioms proper and those of their parts. Since idioms, whether compounds or stock phrases, mani fest complex morphology, they are segmentable into recog nizable morphemes. Furthermore, these morphemes often re tain all or part of their non-idiomatic readings or can be derived from them by regular processes. But this situation is typical of the sort for which we proposed the device of the via-rule above. In such cases, via-rules could connect idiomatic readings or portions of them with those of the items which go into making up the idiomatic compound or phrase. They could thus capture the relevant relations in
176
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
a natural manner. This would permit us to give formal substance to the distinctions mentioned above. A one-sided idiom, for instance, will be a complex unit one portion of whose idiomatic interpretation is re lated to the reading of the corresponding lexical item by means of a via-rule. So the entry for tall order, cited above, will contain a via-rule linking it directly to the appropriate lexical entry for order. Shaw (1979) lists table-leg as an example of a compound whose motivation de pends upon a metaphoric relation from its idiomatic inter pretation to the standard lexical entry for leg. The viarule attached to its idiomatic entry will cite Metaphoric Principle 1 in our inventory, which, along with Transfor mation 1 (model-object), removes the feature (Human) for the standard reading for leg, so that it can combine with table. Notice that via-rules of this type can define only that motivation in idioms which pertains to the presence of a particular morphological unit within them. Since it pertains directly to the readings of the individual compo nents which make up an idiomatic complex unit, I label this sort of motivation componential motivation. It dif fers, as we shall see, in.important ways from the composite motivation to be described next. There exists a second type of motivation in complex units. In spite of the fact that Weinreich (1969, 76) claims, "The relation between idiomatic and literal mean ing is so unsystematic as to deserve no place in the the ory," relations of precisely this sort have been advanced in treatments of both word formation and stock phrases. Ullmann (1957), Chafe (1968), Makkai (1972), Shaw (1979) and many others all recognize that the literal and idio matic senses attached to complex units are often related by "traditionally recognized rhetorical devices" (Makkai 1972: 169). The regular relational principles I proposed as part of semantic theory could thus be used to describe relations of this sort. Consider live wire on its idiomatic interpretation 'energetic person' as an example of a metaphorically moti vated compound, and hit the nail on the head meaning roughly 'do something exactly correctly' as a metaphori cally motivated stock phrase. The former receives motiva tion due to Metaphoric Principle 1, which, along with Transformation 1, removes the feature complex limiting its literal reading to physical objects and electrical energy, so that it can extend to energized objects generally, thereby linking its idiomatic and literal readings. The stock phrase receives motivation due to Metaphoric Prin ciple 3 (specific case-general rule), which establishes a
MOTIVATION IN COMPLEX UNITS
177
relation from its literal to its idiomatic reading on the basis that the former describes a special case of the general proposition expressed by the latter, i.e. hitting a nail on the head counts as a special case of doing any thing exactly correctly. As examples of complex units accorded motivation on the basis that their literal and figurative readings show a metonymic connection, consider the compound redbreast, which designates a type of bird with a red breast, and the stock phrase has the oat got your tongue? meaning 'can't you talk?'. The former receives motivation due to Meto nymic Principle 9 (part-whole), which allows for the transfer from the literally described part to the idiom atically designated whole. The stock phrase receives moti vation due to Metonymie Principle 1 (cause-effect), trans fer from the literally described state of affairs in which a cat is holding on to or has taken someone's tongue to the idiomatic reading, which represents one effect this state of affairs would bring about, viz. that the person affected would not be in a position to speak. Notice that here, as in the metaphoric examples just cited, the read ings for individual components in the complex unit are not directly related to the idiomatic reading. Clearly no por tion of 'can't you speak' can be linked to readings for items like oat and tongue. Instead, the whole literal reading is connected to the whole idiomatic reading. Roos (1979) and others have also mentioned rhetorical devices such as ellipsis, hyperbole, euphemism as sources of moti vation for idioms, since they too can establish regular relations from literal to idiomatic readings. Weinreich's (1969) claim about the relation between idiomatic and lit eral meanings therefore turns out to be clearly indefens ible in many cases, though certainly not in all. The work of other scholars indicates the necessity of recognizing various degrees and types of motivation in complex units. Motivation of the type I have been discussing here differs from that illustrated above; I label it composite motiva tion, in order to set it off from the componential moti vation defined above. An extension of the device of the via-rule as it has been defined up to now suffices to express regular rela tions between idiomatic and literal meanings of complex units. As we just saw, regular relational principles de fine such relations. Via-rules therefore represent a natu ral means of stating these relations, so long as idiomatic readings for complex units are listed along with them in the lexicon, as will be the case for compounds, or in the special inventory of stock phrases. I envision via-rules
178
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
which take the general form: Via-rule 1 : Related to the literal reading (of the complex unit in question) via X. Here, as previously, X identifies some particular rule or principle appearing at some point in the language descrip tion. The semantic component can be counted on to deliver the necessary literal readings, whereby we require that no appeal be made to the feature transfer rule or to regular relational principles. To return to an example cited above, the stock phrase hit the nail on the head will be entered in the stock phrase inventory along with its idiomatic reading 'do something exactly correctly' and Via-rule 2. Via-rule 2: Related to the literal reading via Metaphoric Principle 3 (special case-general rule). We might append to Via-rule 2 the extra-linguistic, tech nical information that, within the context of carpentry, hitting a nail squarely on the head instances an act per formed in exactly the correct manner. Given this additional background information, Via-rule 2 is able to capture the regular relation holding between the literal and idiomatic readings of the stock phrase, and thereby to account for its composite motivation. I will return to and elaborate on the inclusion of extra-linguistic information.in viarules in the following chapter. To summarize, recent research concerning complex units has identified two distinct sources of motivation for them, although the two have not always been clearly separated. I recognize cases of componential motivation, when the semantic contribution identifiable components of complex units make to the interpretation of the whole can be re lated to the readings these components receive in their own lexical entries either directly or by means of regular relational principles. I define composite motivation in complex units in contrast to componential motivation as follows. Composite motivation in complex units arises when the idiomatic reading assigned such a unit exhibits a reg ular semantic connection to its literal reading. Let us now turn to the functioning of via-rules in stating such relations.
MOTIVATION IN COMPLEX UNITS 7.3. Vía- rules
and Motivation
in Complex
179 Units
In this section I discuss the expression of relations leading to motivation in complex units with the device of the via-rule. I concentrate first on those via-rules asso ciated with componential motivation, then proceed to those involved in cases of composite motivation. 7.3.1. Componential
Motivation
Both compounds listed in the main lexicon and stock phrases listed in a special inventory within it can be linked to the entries of their components by means of viarules whenever semantic relations between them obtain. We find cases where component lexical items retain their lexi cal readings intact (e.g. keep in keep tabs on) as well as cases where the semantic contribution of a component lexi cal item represents a figurative variant of its lexical reading (e.g. leg in table- leg). Compounds and stock phrases exhibit relations of both kinds from one case to the next. Although the two types of complex units parallel each other with respect to motivation, I shall separate them in the following to simplify exposition. 7.3.1.1. Componential
Motivation
in
Compounds
In the simplest case, a two component compound pos sesses a semantic interpretation properly containing the readings of both its components. The portions of the in terpretation which are predictable on the basis of the readings of its components enjoy componential motivation. If the portions with componential motivation exhaust its semantic content, I say it enjoys full componential moti vation (not to be confused with composite motivation, which by its nature is always full rather than partial). A compound such as street sweeper exhibits full composite motivation, because both sweeper 'x sweeps' and street retain their standard lexical readings in the composite interpretation 'x sweeps streets', and they exhaust its content. I propose to express this state of affairs by inventorizing street sweeper in an entry like the follow ing, which contains two via-rules. Lexical entry 1: street sweeper: 'x sweeps streets' Related to street directly. Related to sweeper directly.
180
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Notice that the via-rules appropriate in this case contain no via-component at all. They serve merely as redundancy rules in the sense of Jackendoff (1975), and need not refer to regular relational principles to establish the relevant connections to other entries. They clearly represent the limiting case of a via-rule on one extreme, but I shall continue to consider them via-rules, albeit of a truncated sort, in order to minimize the number of different devices in the lexicon. At the same time, viewing them as a type of via-rule enables us to recognize a relation between di rect and indirect or figurative connections from readings for complex units to readings of their components. We also find compounds with partial componential moti vation. Crab grass will occur in an entry containing a via-rule relating it directly to the lexical entry for grass. No lexical entry for crab, however, correlates with the reading for the compound itself (but cf. crab apple). Hence, cra b grass receives one via-rule accounting for a portion of its interpretation on the basis of one of its components. It consequently counts as a case of partial componential motivation. Of course, some compounds enjoy no componential moti vation whatsoever. On a purely synchronic basis, the read ing for scapegoat 'person made to bear blame' has nothing in common with the individual readings of its component items. The entry for it will contain no via-rules, and it will consequently be accorded no componential motivation. Whether componential motivation is full or partial, it may be direct or figurative. We saw that street sweeper counts as fully componentially motivated. Since the viarules which characterize it as such establish direct rela tions to the readings for its components, i.e. they cite no regular relational principles, its motivation can be labeled direct. In other examples, however, regular rela tional principles must be cited by the relevant via-rules to account for relations. The compound table-leg mentioned above offers an example in point. The reading 'structure supporting table' bears a regular relation to that for leg, only if the human reading for leg undergoes an extension. We must posit a via-rule such as (3) to express the con nection. Via-rule 3: Related to leg 'limb supporting human' via Metaphoric Principle 1 and Transformation 1 deleting (Human) from this reading. Table -leg can be considered fully componentially motivated, given a second via-rule relating it directly to the lexi cal entry for table, but we must specify its motivation as
181
MOTIVATION IN COMPLEX UNITS
metaphorical due to the functioning of a metaphoric prin ciple in one of the via-rules. Componential motivation may also rely upon metonymic principles. Steel executive is accorded full componential motivation on the basis of two via-rules which, taken to gether, relate the whole of its composite Interpretation to the lexical readings of its components. The initial ele ment in 'executive in steel industry' is directly related by via-rule to the lexical reading for executive. The re maining portion of the composite interpretation cannot be directly related to the lexical reading for steel. But a via-rule like (4) suffices to establish the relation by citing the appropriate metonymic principle. Via-rule 4: Related to steel 'metal' via Metonymie Principle 11 (central factor-institution). We can thus say steel executive represents a fully componentially motivated compound, but specify its motivation as partially metonymic. In summary, componential motivation in compounds can be full or partial. In either case, it must be further qual ified either as direct, metaphoric or metonymic. 7.3.1.2. Componential
Motivation
in Stock
Phrases
Like compounds, stock phrases may enjoy full componen tial motivation. Many irreversible binominals, clichs and quotes which appear in the inventory of stock phrases be cause of their syntactic cohesion rather than because they are not compositionally interpretable offer cases in point. Ham and eggs, to all intents and purposes, and blessed are the meek, all of which were cited as examples of stock phrases above, are compositionally interpretable, and con sequently fully componentially motivated. Entries for them must contain via-rules relating their interpretations to those of their component parts directly. Partially componentially motivated stock phrases are also easy to come by. I mentioned above that the semantic contribution of keep in keep tabs on correlates directly with its lexical reading. The composite interpretations for that's a tough row to hoe, that's a hard pill to swal low, and that's a hard/tough nut to crack all involve the feature (Difficult), which can be directly related to the component lexical items tough and hard. The presence of via-rules capturing these relations dictates that we rec ognize these stock phrases are partially componentially motivated.
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
182
Stock phrases exhibiting no via-rule connections to readings of their component items lack componential moti vation entirely. They exemplify the class of complex units often labeled idioms in common parlance. Units like red herring, put up w ith and shoot the breeze are relevant ex amples.
Metaphoric componential motivation obtains for stock phrases as well as for compounds. The phrase pass the buck demonstrates this fact. The composite interpretation for this phrase, roughly expressed as 'transfer responsibil ity' , can be related to the concrete lexical reading of pass 'hand over to' or 'transfer by hand to' by means of Metaphoric Principle 1 (model-object) and Transformation 1, which deletes the concrete reference to 'hand' in the lat ter. Similarly, the composite interpretation for blow off steam 'release emotional pressure' correlates with the lexical reading for blow off again by Metaphoric Principle 1, which allows us to discount its concrete, physical fea tures. We also find examples of metonymic componential moti vation in stock phrases. Both hit the hay and hit the feathers, roughly 'go to bed' can be connected to the lexi cal readings for hay and feathers by means of Metonymic Principle 12 (container-contents), since both materials are, or historically have been, used as stuffings for bed ding, beds, pillows. To fully account for such cases, we again might extend the via-rule format to include, beyond the citation of a regular principle, an item of extralinguistic information as well. For instance, the entry for hit the hay would contain Via-rule 5. Via-rule 5: Related to hay 'dried plants' via Metonymie Principle 12 and the fact that hay is/was used as a material for stuffing beds. A further example of metonymie componential motivation in stock phrases is offered by the phrase town and gown mean ing 'townspeople and university staff'. This stock phrase is especially interesting, because it requires two differmetonymic principles to account for its fully motivated status. The relation between 'townspeople' and the lexical item town is established by Metonymie Principle 13 (local ity-occupants), and that between 'university staff' and gown is established by Metonymie Principle 14 (costumewearer) , along with the additional information that uni versity staff typically wear gowns in some societies.
MOTIVATION IN COMPLEX UNITS
183
To summarize this section, we have seen that stock phrases, like compounds, can enjoy either full or partial motivation. Both may be either direct, metaphoric or meto nymie. Via-rules, in some cases extended to include extralinguistic information, can account for the various sorts of motivation. 7.3.2. Composite
Motivation
Composite motivation derives from a regular relation holding between the idiomatic and literal readings of a single complex unit. Since a direct relation between these two readings would automatically amount to total compositionality, composite motivation must always be figurative. All composite motivation in complex units can be classified as either metaphoric or metonymic. We will thus be con cerned only with via-rules involving regular principles in the following two sections on compounds and stock phrases respectively. 7.3.2.1. Composite
Motivation
in
Compounds
If one literal reading of a compound can be linked to its idiomatic interpretation by means of a regular rela tional principle, it enjoys composite motivation. In addi tion to the compound live w i r e cited above, consider the example of bean pole '(Long) (Thin)' as an example of metaphoric composite motivation in compounds. The literal reading of bean pole identifies a particular physical ob ject, but the idiomatic reading designates a tall, thin person. Nevertheless, the literal reading contains the features (Long) and (Thin), distinguishing pole from peg and stake respectively. Consequently, Metaphoric Principle 4 (object-feature), along with Transformation 2, can de lete all but these features, thereby relating the literal and the idiomatic readings - the feature (Human) in the idiomatic reading arises historically as a result of ap plying bean pole often to human beings (cf. discussion of tub in 3.3.2 above). The relevant via-rule attached to the entry for bean pole is represented by: Via-rule 6: Related to the literal reading via Metaphoric Principle 4 and Transforma tion 2 deleting all but the features (Long) and (Thin).
184
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Bean pole therefore supplies an example of metaphoric com posite motivation in compounds. Metonymic composite motivation in compounds was illus trated by the example redbreast above in 7.2. We saw that it derives its motivation on the basis of Metonymic Prin ciple 9 (part-whole). We find a similar example of meto nymie composite motivation in the compound redcap 'porter'. On the literal level this compound describes an article of clothing, and has nothing to do with porters or carrying luggage. But given the extra-linguistic information that porters customarily wear red caps as an outward sign of their function in certain societies, we can correlate the literal reading for redcap with its idiomatic reading by means of Via-rule 7. Via-rule 7: Related to the literal reading via Metonymie Principle 14 (costume-wearer) and the fact that porters typically wear red caps as the badge of their trade. With the help of an additional piece of extra-linguistic information, then, we can link the literal and idiomatic readings for redcap, thereby predicting its metonymie com posite motivation. 7.3.2.2
Composite
Motivation
in Stock
Phrases
Composite motivation in stock phrases parallels that in compounds. In the class of stock phrases deriving com posite motivation metaphorically, I cited above hit the nail on the head nd proposed a via-rule accounting for this motivation. Many stock phrases derive metaphoric com posite motivation from relations between their literal and idiomatic readings established by Metaphoric Principle 3 (special case-general principle). This holds especially for proverbs. Proverbs exhibit a tendency to express gen eral truths in the concrete terms of a particular natural phenomenon or human activity. To the extent that they do this, they represent typical instances of the application of Metaphoric Principle 3, which relates special cases with the general principles to which they adhere. The gen eral tendency toward concretization of general truths and maxims has often been noted in work on proverbs (e.g. Bar ley 1972; Žolkovskij 1978 et al.). The apple doesn't fall far from the tree literally presents a concrete example of a general rule, viz. that offspring end up (at least, in appearance) close to their progenitors. Don't bite off more
MOTIVATION IN COMPLEX UNITS
185
than you can chew offers specific, concrete advice on the literal plane, which is generalized to 'don't take on more than you can handle' in its idiomatic reading. In both these cases, Via-rule 8 expresses the appropriate relation between the idiomatic and literal readings. Via-rule 8: Related to the literal reading via Metaphoric Principle 3 (special casegeneral principle). Comparing the literal reading don !t bite off more than you can chew with the idiomatic reading 'don't take on more than you can handle', one can easily recognize that this stock phrase would rate fairly high in terms of componential motivation, in addition to enjoying metaphorical composite motivation. But componential and composite moti vation cannot be treated as additive; otherwise we would be forced to recognize complex units with greater than full motivation. One of the two must be superordinate to the other. I choose to discount componential motivation in the face of composite motivation. Given alternate methods of accounting for the motivation a single item claims, I assume speakers will tend to see composite before compo nential relations or at least to rely on the former as a mnemic device. The concrete terms in which proverbs are cast often stem from activities not practiced by the majority of speakers or not practiced by anyone at all in the present, due to the persistance of such phrases in the language over extended periods of time. So we should not be sur prised to find that certain pieces of extra-linguistic information are required by a speaker, if he is to recog nize the motivation of such phrases. For instance, a stitch in time saves nine casts a general maxim in sewing terms; the relevant via-rule should include the objectspecific information that stitches are taken to repair garments. Make hay
while
the
sun
shines
will be assigned
a via-rule like (9). Via-rule 9: Related to the literal reading via Metaphoric Principle 3 (special casegeneral principle) and' the objectspecific fact that hay must be dried in the sun. Compare in this context the idiomatically synonymous pro
verb strike
while
the iron is
hot.
Stock phrases often exhibit metonymic composite moti vation as well. Our example has the cat got your tongue?
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
186
from above illustrated composite motivation deriving from Metonymic Principle 1 (cause-effect) in connecting the literally described cause with the idiomatic reading ident ifying an effect of it. Make tracks in the sense of 'leave' also requires Metonymic Principle 1 for its composite moti vation, but here the literal reading describes an effect of the action identified in the idiomatic reading. Chafe (1968) suggested bury the hatchet 'make peace' as an ex ample of a semi-idiom (as opposed to a true idiom, which exhibits no motivation whatsoever), saying that it lit erally described an action symbolic of what its idiomatic meaning intended. Chafe himself recognized the inadequacy of this attempted description, but he declined to offer any alternative treatment. In the present theory, bury the hatchet represents a compositely motivated stock phrase. The appropriate via-rule may be stated as follows: Via-rule 10: Related to the literal reading via Metonymie Principle 1 and the cultural/ historical information that the hatchet was the archetypal instrument of war for the American Indian. The device of the via-rule can here be seen to allow the formalization of the relation Chafe intuitively designated as 'symbolic'. The distinction Chafe wanted to recognize between true idioms and semi-idioms is formalized within the framework presented here as the lack or presence of via-rules in the entry for a stock phrase, where their presence, of course, indicates motivation of some variety. By distinguishing partial from full motivation, as well as direct from metaphoric and metonymie motivation, however, we arrive at a much more adequate differentiation of idiom types. To summarize the last two sections, we saw that complex units exhibit composite motivation of both metaphoric and metonymie varieties. We noted the necessity for and the possibility of including extra-linguistic information in via-rules in order to describe certain relations. We also saw that componential motivation is discounted in complex units enjoying composite motivation as well. 7.3.3
Classifying
Complex
Units
With via-rules to capture connections from idiomatic readings of complex units to both their literal readings and the lexical readings of their component items, we end
MOTIVATION IN COMPLEX UNITS
187
up with considerably fewer true idioms, in Chafe's (1968) sense, than other theories have. Both compounds and stock phrases enjoying composite motivation due to some regular relational principle differ from true idioms. Also those complex units with full componential motivation, whether of a direct or figurative sort, must be distinguished from true idioms, since their readings are compositionally de rivable from those of their components - given regular re lational principles for figurative cases. Even partially componentially motivated complex units should perhaps not count as true idioms, because their idiomatic interpreta tions are not totally unpredictable on the basis of their components. Many units classified as idioms in other the ories can be dealt with in the present framework as com positionally interpretable compounds and stock phrases lexicalized as units only because they hang together syn tactically. Given the distinctions drawn above, we arrive at the following taxonomy of complex units, based upon the mode of their interpretation and/or type of motivation.
188
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
1. Compositionally interpretable complex units 1.1 Complex units interpretable with the projection rule alone 1.1.1 Compounds with full direct componential motivation (literal compounds) e.g. street sweeper 1.1.2 Stock phrases with full direct componential motiva tion (literal stock phrases) e.g. ham and eggs, keep smiling 1.2 Complex units interpretable with the projection rule and regular relational principles 1.2.1 Complex units with composite motivation (figurative complex units) 1.2.1.1 Compounds with composite motivation (figurative compounds) 1.2.1.1.1 Compounds with metaphoric composite motivation (metaphoric compounds) e.g. live wire 1.2.1.1.2 Compounds with metonymic composite motivation (metonymic compounds) e.g. redcap 1.2.1.2 Stock phrases with composite motivation (figura tive stock phrases) 1.2.1.2.1 Stock phrases with metaphoric composite motiva tion (metaphoric stock phrases) e.g. the apple doesn't fall far from the tree 1.2.1.2.2 Stock phrases with metonymie composite motiva tion (metonymie stock phrases) e.g. bury the hatchet 1.2.2 Complex units with full figurative componential motivation (at least one via-rule citing a regular relational principle) 1.2.2.1 Compounds with full figurative componential moti vation 1.2.2.1.1 Compounds with full metaphoric componential motivation e. g. table- leg 1.2.2.1.2 Compounds with full metonymie componential moti vation e.g. town and gown 2. Compositionally non-interpretable complex units 2.1 Complex units with partial componential motivation 2.1.1 Compounds with partial componential motivation 2.1.1.1 Compounds with partial direct componential moti vation e.g. stone-broke 'completely moneyless' Related to broke 'moneyless' directly. 2.1.1.2 Compounds with partial figurative componential motivation
MOTIVATION IN COMPLEX UNITS
189
2.1.1.2.1
Compounds with partial metaphoric componential motivation e.g. chatterbox 'loquacious person' Related to chatter 'produce inhuman sounds rapidly' via Metaphoric Principle 1 and Trans formation 1 deleting (Inhuman). 2.1.1.2.2 Compounds with partial metonymic componential motivation e.g. bluestocking 'aristocratic person' Related to stocking 'garment' via Metonymic Principle 14 (costume-wearer). 2.1.2 Stock phrases with partial componential motivation 2.1.2.1 Stock phrases with partial direct componential motivation e.g. keep tabs on 2.1.2.2 Stock phrases with partial figurative componential motivation 2.1.2.2.1 Stock phrases with partial metaphoric component ial motivation e.g. pass the buck 2.1.2.2.2 Stock phrases with partial metonymie componential motivation e.g. hit the hay 2.2 Complex units with no motivation (no via-rule links) 2.2.1 Compounds with no motivation e.g. scapegoat 2.2.2 Stock phrases with no motivation e.g. shoot the breeze In addition, as mentioned above, there are complex units which are not only compositionally non-interpretable, but syntactically malformed as well (e.g. by and large). Our taxonomy reveals sixteen classes of complex units which all differ with respect to the number and type of via-rules contained in their entries. At the same time, our list presents a continuous scaling of complex units from complete compositionality to complete non-compositionality, or from complete transparency to complete opaqueness, when read from top to bottom. Even if concrete decisions about exactly where a complex unit should appear in our taxonomy prove difficult in some cases, it certainly offers a sys tem of differentiation more useful than that of distin guishing only idioms and semi-idioms. Since this taxonomy is based upon via-rule connections, we should count the formalization of such a classificational system for complex units among the consequences via-rules entail for the lexi con.
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
190 7.4 Stock
Phrases
in Semantic
Interpretation
Compounds behave exactly like simplex words in syntac tic and semantic respects. We consequently decided to place them in normal entries in the main lexicon. Stock phrases, however, differ from simplex words in one import ant respect. From a syntactic point of view one can de scribe the difference by saying that stock phrases resist to varying degrees deformations deriving from syntactic operations. In semantic terms one can say that stock phrases receive their usual idiomatic interpretations only when their surface forms adhere to their lexical forms in specifiable ways. Recall the example of red herring above, along with other phrases which cannot be freely reordered or inter rupted by other elements. Of the four sentences (1) - (4), only (1) admits interpretation of red and herring as an idiomatic unit. (1) (2) (3) (4)
That is a red herring. That herring is red. That red thing is a herring. We saw a herring which was red.
Given the close semantic connection between (1) - (4) in compositional terms, we must conclude that the stock phrase red herring is rigidly constrained as far as its syntactic form is concerned. Other stock phrases exhibit less rigid ity, or frozenness, to use Fraser's (1970) term. (5) - (7) are all susceptible to idiomatic interpretation apparently, although (8) seems not to be. (5) We shoot the breeze every day after lunch. (6) They have been shooting the old/proverbial breeze for an hour. (7) Lunch will be eaten, then the breeze will be shot for an hour. (8) The shooting of the breeze lasted an hour. Notice that (8) can be forced into the idiomatic mold by the addition of proverbial, a word commonly occurring in idioms, proverbial phrases, and proverbs proper. (9) The shooting of the proverbial breeze lasted an hour. The problem for the semantic component consists in determining which structurally connected occurrences of
MOTIVATION IN COMPLEX UNITS
191
the breeze and shoot count as variants of the idiom. The semantic component must somehow take final surface form into account, in addition to the base phrase marker into which a stock phrase is inserted. In fact, the surface structure level plays a two-fold role in the interpreta tion of idiomatic stock phrases. Not only does It deter mine whether the syntactic form of the phrase has remained within the bounds specified for It, surface structure may turn up strings corresponding to permitted variants of stock phrases, even when these did not originate as such. Speaking literally of a real pecan, one might say, "That's a very hard nut to crack." And, as Chafe (1968) pointed out, hearers will tend to interpret structures as idio matic, whenever possible. Consequently, the semantic com ponent will have to scan surface structures to determine whether they contain strings corresponding to stock phrases, no matter how stock phrases are treated otherwise. This observation led me in Norrick (1980) to propose that, in an interpretive semantic framework, all identifi cation of stock phrases occurs at the level of surface structure. Since some stock phrases may undergo various syntactic deformations during derivations, entries for them must be rich enough to predict all admissible surface forms. Short of listing in each entry all the permissible surface forms of the stock phrase, we can index these phrases for their positions along the scale of frozenness, as suggested by Fraser (1970). This permits prediction of possible surface forms without drastically complicating entries for stock phrases. Restricting identification of stock phrases to surface structure, and ordering it be fore applications of the projection rule in the process of semantic interpretation would automatically account for the tendency hearers show to interpret idiomatically even structures not so intended. So the process of semantic interpretation would be initiated by an operation of scan ning surface structure for matches with entries in the inventory of stock phrases (hence the decision to inven torize stock phrases separately from simplex and compound words). If a match occurs during this operation, the read ing for the phrase identified passes entire into the seman tic interpretation for the structure. Since, as we saw, surface structure must be scanned for spurious idioms, i.e. strings matching entries for stock phrases from which they did not originate, in any case, proposing that all identification of idiomatic stock phrase occurs at the surface structure level entails no complication of interpretive semantic theory. Nor is this proposal the only one involving the claim that
192
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
semantic interpretation must take surface structure into account. Chomsky (1970), Jackendoff (1971, 1972) and sub sequent research in the framework (cf. note 1 to chapter 1) they describe argue that various aspects of sentence meaning, including presuppositions, focus, and the scope of modals and negation, must be determined at the level of surface structure, since they are affected by transformational operations, just as the forms of idiomatic stock phrases are. Furthermore, the alternatives to surface structure identification of stock phrases in interpretive semantics bring with them even less palatable consequences. One alternative to surface structure identification of stock phrases would involve marking them as exceptions to certain transformations at the deep structure level when they are inserted. Of course, each phrase would re quire a separate marking reflecting its peculiar behavior with respect to syntactic operations. Fraser's (1970) at tempt to account for certain transformational deficiencies of idiomatic stock phrases on a semantic basis fails, as Quang Phuc Dong (1971) revels in demonstrating, so that all such markings will be of an essentially ad hoc nature. Insofar as markings of this type would be carried along from the point of insertion to the surface structure level, they constitute global derivational constraints, on the definition of Lakoff (1970a). Whether such devices are preferable to, more complicated than, or notational equi valents of surface structure interpretation has apparently not been definitively decided (see e.g. Katz 1971; McCawley 1971; Katz 1976). At least for the case at hand, it seems the two alternatives are in fact equivalent. For present purposes it suffices to note that interpretive semantic theory requires surface structure interpretation, and generative semantic theory requires global derivational constraints in order to account for the semantic interpre tation of stock phrases. A second alternative treatment of idiomatic stock phrases is proposed by Newmeyer (1972, 19 74). Newmeyer's proposals do not count as a true alternative, insofar as they also postulate a need for global rules. Newmeyer would list stock phrases in entries including both their idiomatic and literal readings. In addition, lexical items which occur in idioms will be characterized as such in their lexical entries. This extension ensures recognition of idioms in surface structure and governs the applicabili ty of transformational rules during syntactic derivations. If the rules Newmeyer proposed to connect idiomatic stock phrases to their component lexical items had also cap tured semantic relations between the two, they would be
MOTIVATION IN COMPLEX UNITS
193
essentially the same as our via-rules describing componential motivation. Newmeyer's proposal would then parallel that offered here, except that it could not handle com posite motivation directly, in spite of his suggestion that entries for idiomatic phrases contain both idiomatic and literal readings. In the final analysis, however, his solution explains less, while requiring the same amount of apparatus, so that we must prefer the analysis de scribed here. I uphold the analysis presented in Norrick (1980), whereby the semantic component scans surface structure for strings corresponding to those foreseen by entries in the inventory of stock phrases. If a match is detected, the reading from the appropriate entry is inserted entire into the semantic interpretation for the construction in ques tion. If no match can be detected, semantic interpretation proceeds by means of successive applications of the pro jection rule.
8.
8.0
EXTRA-LINGUISTIC INFORMATION IN SEMANTIC RELATIONS
Introduction
In this chapter I discuss semantic relations speakers establish with the help of extra-linguistic information. I consider first the role of extra-linguistic information in semantic interpretation generally. I identify four classes of extra-linguistic information relevant for seman tic relations, then demonstrate inclusion of items from each class in via-rules. 8.1 The Role of Extra-linguistic Establishment of Semantic
Information Relations
in
the
We have already observed several examples of semantic relations which speakers establish on the basis of extralinguistic information. Recall the metaphoric interpreta tion of fox as 'sly creature'. As pointed out in 2.3 and 6.4 (cf. note 1 to Chapter 6 as well), the feature (Sly) will not occur as part of the lexical reading for fox. Slyness is a characteristic attributed to foxes in certain cultures, but it is not one which serves as a feature dif ferentiating the lexical reading for fox from that of any other lexical entry. In a theory of the lexicon adhering to the principle of determining lexical readings on the basis of truly distinctive features, we can not admit (Sly) as a feature for fox. Instead, this feature must originate outside the lexicon proper. We can think of (Sly) as de riving from the entry which fox will receive in what we can call the 'encyclopedia', i.e. in the pool of knowledge and beliefs a particular culture, or, in some cases, a particular group within that culture, accepts. Establish ment of a regular semantic relation from fox 'wild canine' to fox 'sly creature' consequently entails extra-linguistic (or encyclopedic) information about the slyness of foxes. We can model this state of affairs by specifying that a speaker must assign the feature (Sly) from the encyclopedia
19 6
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
entry for fox to its lexical reading before metaphorically extending it to mean simply 'sly' by means of Metaphoric Principle 4 and Transformation 2 (the further expansion to 'sly creature', i.e. (Sly) (Animate) (Physical) (Object), on which this metaphorical extension becomes lexicalized, has its source outside the metaphorical process as such; cf. discussion of tub 'round person' in 3.3.2). In Inter preting the metaphorical reading of fox,one must first as sociate the contextual occurrence with the standard lexi cal entry (essentially Katz's 1972 'lexical interpreta tion'), then assign to this the feature (Sly) from the encyclopedia entry as before. This addition of information from a different level and subsequent continuation of sem antic processing closely parallels the device presented by Droste (1976). As in Droste's model, optional inclusion of extra-linguistic information at certain stages of the pro cess of semantic interpretation allows generation of read ings which transcend the level of purely semantic inter pretation in the relevant respect. So at least in certain cases, it appears correct to maintain, as many scholars have done (e.g. Abraham 19 75b; Horrocks 19 76; Davidson 1978), that metaphor requires descriptive mechanisms over and above those within linguistic theory proper. Note, however, that we could handle the metaphoric extension of fox to mean 'sly' system internally, if we relaxed the restriction on lexical readings, whereby they may contain only distinctive features. Once lexicalized the metaphoric reading of fox will be linked by via-rule to the source reading. The full viarule attached to the former should look something like this. Via-rule 1: Related to fox 'wild canine' via Metaphoric Principle 4 and Transformation 2, which deletes all but the feature (Sly) and the cultural fact that the fox is considered very sly, whence the transferal of the feature (Sly) into the lexical reading. This formulation has several advantages. First, it is simple. Second, it seems to model psychological reality correctly: the knowledge of Metaphoric Principle 4 and the associated Transformation 2 is part of competence, while that of foxes is not. Via-rule 1 separates the two sorts of information, but clearly states that the extra-linguis tic information is necessary in order to establish the connection between the two readings. Further, as Vennemann (1972) points out for via-rules in phonology, they allow regularities individual speakers
EXTRA-LINGUISTIC INFORMATION
197
perceive to be stated without forcing the obviously un tenable assumption that all speakers in a linguistic (and cultural) community recognize all the same lexical rela tions. For instance, a particular speaker may associate crux and crucial by means of a phonological via-rule, but this connection does not thereby become a part of the de scription of English phonology as such. Neither nor crucial is considered more basic in any sense, nor will either be derived from the other. Vennemann proposed via-rules in phonology primarily to avoid abstract underlying forms, especially those based upon historical relations. Via-rules capture the relevant historical connections without requiring abstract under lying forms and complex sets of rules describing deriva tions. Semantic via-rules can also describe historical relations between readings without ascribing these to com petence as such. They simplify lexical entries by removing from them object specific information which exceeds that required for distinctive definition and is needed only to establish semantic relations to other entries. Via-rules prevent 'abstract' semantic representations for items, viz. those containing multiple readings, overly vague readings and/or unrestricted amounts of non-distinctive, extra-linguistic information, by offering a simple means of relating multiple readings of items, even if they ap pear in separate entries. There are two very important reasons for distinguishing the kind of knowledge speakers have of the meaning of the words of their languages and that they have of certain (non-adjacency) relations between the meanings of these words. First, speakers often establish connections between meanings on the basis of personal experiences, felt con notations and even mistaken beliefs (as in popular etymo logy) ; nevertheless, the relational principles themselves belong to competence as such. Via-rules keep the two sep arate by citing both a regular relational principle and an item of extra-linguistic information. Second, the lexicon by its very nature reflects the area of language manifest ing the greatest lectal variation, not only from one speaker to the next, but within the various lects and registers of a single speaker. This necessitates a de scription of the semantic relations at issue here on the level of individual lexical entries rather than in general rules applicable to whole classes of items. Again viarules provide the needed device. We also noted that the recognition of composite moti vation in complex units sometimes depends on extra-linguis tic information. The via-rules which account for cases of
198
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
composite motivation parallel those required for semantic relations of other types. They differ only in relating readings of phrase or even sentence length to each other, whereas other via-rules link readings consisting of only one or two morphemes in most cases. Further, the extralinguistic information leading to the recognition of com posite motivation in a complex unit may relate to any por tion of the literal reading for the unit. For instance, the recognition that make hay while the sun shines relates to 'accomplish any task while auspicious conditions obtain' by Metaphoric Principle 3 (specific case-general principle) hinges on the fact that hay must be dried In the sun; the relevant extra-linguistic information pertains only to hay. By contrast, recognition of the relation between don't look
a gift
horse
in
the
mouth
and its idiomatic reading
'don't be critical of gifts' requires at least extra-lin guistic information concerning horses and the practice of looking into their mouths to determine their ages. We must consequently specify that the extra-linguistic Information of relevance to the relation of a pair of literal and idio matic readings may derive from encyclopedia entries for any of the items in the literal reading, and in some cases for combinations of them (e.g. look x in the mouth). 8.2 Extra-linguistic
Information
Extra-linguistic Information involved in establishing semantic relations between lexical entries (including those for complex units) can be discussed under four major headings: (1) etymological information, (2) historical in formation, (3) object specific information, (4) pragmatic information. These are rather rough-and-ready categories, and a relatively great degree of possible overlap and boundary fuzziness must be expected. Their interrelatedness and lack of specificity result from and reflect the varieties of semantic transfer and extension as well as the erraticism of lexicalization in natural languages. Con sequently, pure examples of the four types will be diffi cult to find. Nevertheless, this division should suffice to provide an adequate framework for the following dis cussion. 8.2.1 Etymological
Information
By 'etymological information' I intend information con cerning historical meanings, historical forms, their
EXTRA-LINGUISTIC INFORMATION
199
linguistic historical changes and morphological relations. Obviously, this represents a specialized type of informa tion which is irrelevant to the understanding of the mean ings assigned to lexical items as such and a fortiori to their correct use. The etymology of a particular word af fects neither the specialized, object specific knowledge of the real object(s) it designates nor its use in even technical jargon, except, of course, that of the science of etymology itself. Etymological knowledge may even per suade a speaker to set up a relation between entries in his subjective lexicon with no regular semantic connection in their present-day readings. Such relations are of no direct interest here; instead we are concerned with those lexical relations which cannot be recognized without a certain item of etymological information, but which follow from regular principles once this information is possessed by a speaker. This latter case must be treated in a syn chronic description, because the relations so established represent structures present in the subjective lexicons of contemporary speakers. Of course, we want to avoid classifying etymological information as belonging to the language system as such: hence the need for via-rules. Knowing that ωent derives historically from wend, say, certainly does not affect a speaker's use of these forms, nor does it increase his knowledge of their present meanings in any practical or measurable fashion. Because their historical connection has become so tenuous, no regular relational principle can capture it in any straightforward manner. It consequently makes no sense to claim that went and wend bear any seman tic connection in contemporary English. To the extent, how ever, that etymological information along with regular re lational principles serves to establish semantic relations between lexical entries, these should be linked by means of appropriate via-rules. Consequently, phonologically identical lexical entries which are related by etymologi cal information alone will count as homophonous lexical items (e.g. modal might and nominal might, which bear no regular semantic connection in synchronic terms); those related by via-rules involving etymological information count as multiple meanings of a single polysemous lexical item (e.g. score 'music', score 'point tally', score 'twenty', which all three derive from the verb score meaning 'mark' or 'scratch', since they involve recording notes, wins or points, and numbers respectively). Let us consider the example of tell and teller. Speak ers prepared to recognize a regular relation between the verb tell and the noun teller presumably do so on the
200
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
basis of etymological knowledge. If this knowledge is necessary in order to establish a relation between these two items, then it must be included as additional informa tion in the relevant via-rule. The lexical entry for teller» might, therefore, contain the following via-rule in the lexicons of those speakers who recognize the rela tion to the verb tell. Via-rule 2: Related to tell via Metonymic Principle 7 (agent-act) and the etymological informa tion that tell once meant 'count'. Mention of Metonymic Principle 7 in itself does not suffice to explicate the relation in this case, so etymological information must be explicitly stated in the via-rule if the relation is to be expressed at all. Recall now the relation between tav 'sailor' and tarpauline mentioned in 4.2.6.1 above. In order for a person to establish the correct semantic relationship between these two words, he requires not only Metonymie Principle 8 relating names of professions with objects or tools charac teristic of them, but also the etymological knowledge that tar (like tarp) was a standard clipped form of tarpauline especially in the jargon of sailors. I stress the word correct in the previous sentence, because more than one informant reported associating the 'sailor' meaning of tar with the fact that sailors used tar (pitch) to waterproof the hulls of wooden ships. This transfer follows from the same relational principle as the former, but requires no etymological knowledge; it is thus not surprising that many speakers establish this connection rather than the other. To represent the connection perceived by such speakers, we must set up a via-rule as well. Obviously, if via-rules express relations individual speakers perceive on the basis of extra-linguistic information, mistaken beliefs may lead to via-rules expressing incorrect rela tion or relating items with no historical connection. Re turning to the example of tar 'sailor', the entry for this meaning will contain the following via-rule in the lexi cons of speakers who recognize the relation in question. Via-rule 3: Related to tarpauline via Metonymie Principle 8 and the etymological informa tion that tar was a standard clipped form of tarpauline. Of course a speaker who relates tar 'sailor' with tarpaw line as above must also possess certain knowledge of
EXTRA-LINGUISTIC INFORMATION
201
sailors, viz. that they have much to do professionally with tarpaulines, or at least did at one time. Mention of historical and object specific knowledge of this kind leads to the discussion of the following two types of extra-linguistic knowledge playing a role in lexical se mantic relations. 8.2.2 Historical
Information
By 'historical information' I intend historical facts relating to the objects and events designated by words, but not to the forms or meanings of the words themselves. Like etymological information, such facts represent infor mation which goes beyond what is required to define the re lated lexical items in most cases. This information is hence clearly irrelevant to the correct use of these items. Historical information may, however, play a role in the meanings of certain items manifested in various technical uses or at least in archaic occurrences. Again we require that relations between lexical entries be established on the basis of regular relational principles, but here his torical facts will be shown necessary to allow application of the principles at all. Consider, for example, the relation between the entry for leech where it denotes a particular sort of worm and that where it denotes a physician. In order to relate these two entries, one must possess the historical knowledge that physicians customarily employed leeches in blood letting in former times. Without this knowledge, the prin ciple regularly relating professions with their customary objects or tools cannot apply. One can correctly apply the word leech to both the worm and physicians without recog nizing their connection; one might even possess the rele vant historical knowledge without explicitly recognizing the relation, but one cannot establish the semantic rela tion between the two meanings of leech without the his torical knowledge. For speakers who do recognize the rela tion, we need a via-rule containing the pertinent histori cal information appended to the lexical entry for leech 'physician', e.g. Via-rule 4: Related to leech 'worm' via Metonymic Principle 8 and the historical fact that physicians customarily employed leeches in bloodletting, a standard method in former times.
202
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Similarly, a via-rule relating the lexical entry for bar on which it denotes the legal profession with that on which it denotes an oblong object will have to include the historical fact that the railing in court rooms at which those in the legal profession spent much of their time and energy was termed the bar. Obviously, the kinds of histori cal information which must be mentioned in via-rules for speakers recognizing many relationships of this sort will be quite extensive. It is thus even more important that such historical information not be placed in individual lexical entries. Via-rules offer a natural method of des cribing semantic relations without burdening lexical en tries with much information irrelevant to both lexical meaning and correct use for most speakers and most pur poses. If a complete description of communicative compe tence includes something like an encyclopedia for each cul tural-linguistic community, via-rules could themselves be simplified to refer to the relating principle and the ap propriate encyclopedia entry, thereby further unburdening the lexicon. 8.2.3 Object
Specific
Information
By 'object specific information' I intend facts of a non-historical nature relating to the objects and events designated by words, but not to the forms or meanings of the words themselves. Particularly important are facts concerning the roles objects play in the prevailing cul tural matrix and standard judgements as to their value de riving from it. Much of this knowledge may be reflected in connotations associated with the lexical items themselves and/or restrictions on their appropriateness in various contexts, registers etc. So object specific knowledge, and especially the lack of it (which becomes painfully appar ent in utterances of speakers unfamiliar with the conven tions of the particular cultural matrix in which they find themselves), can influence the appropriateness and full contextual, and cultural meaning of utterances relating to the objects. Given the methodological principle of including only distinctive features in lexical entries, much factual and culturally relevant information concerning the sorts of entities lexical items designate must appear in what we have been calling the encyclopedia. Hence, we saw that the feature (Sly) required for the interpretation of fox on its standard metaphorical reading originates in the ency clopedia. Again since the lexical entry for lion 'large
EXTRA-LINGUISTIC INFORMATION
203
wild cat' eschews mention of such culturally relevant in formation as the common attribution of courage to the lion, this information must be included in the via-rule relating this entry with that for the metaphorical reading 'cou rageous person'. The latter will receive the following via-rule. Via-rule 5: Related to lion 'large wild cat' via Metaphoric Principle 4 and Transformation 2, which deletes all but the feature (Cou rageous) and the object specific fact that the lion is considered very courageous. Again here one might prefer a reference to the encyclopedia entry for lion. Consider now the case of an item such as wishbone. An adequate lexical entry for wishbone would characterize it as a noun denoting a particular bone in birds with a certain form. The information that this bone plays a cru cial role in a wishing game in some cultures will certainly not be included in the lexical entry as such. Yet the rela tion of this item to wishing and hence to the lexical item ω i s h cannot be recognized without this information. A speaker who identifies the initial morpheme of wishbone with the freely occurring wish instead of treating it as an otherwise non-occuring bound morpheme possesses know ledge we can represent in a via-rule attached to the lexi cal entry for wishbone. Via-rule 6: Related to the noun wish via Metonymic Principle 5 (object-act)and the object specific fact that this bone is used in a wishing game. This reference to a game defined in a particular culture illustrates the socio-cultural aspect of object specific information clearly. Object specific information may be of a technical na ture as well as related to cultural norms and practices. In order to recognize a semantic relation between the 'set' and 'electrical storage cell' meanings of battery or be tween differential (automotive) and different, one needs technical knowledge of rather special kinds. To the extent that this knowledge transcends the information required for defining purposes in the lexical entries for such items, it must appear in the via-rules establishing the relations between them, from the appropriate entries of an encyclopedia including technical information as well as
2 04
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
more general cultural knowledge and beliefs. Such technical knowledge is not in the possession of the language commun ity as a whole but in that of particular individuals who consequently recognize semantic connections not perceived by others. No matter where one decides to draw the line between lexical and non-lexical (encyclopedic) information, many cases of semantic relations will remain to be handled by via-rules containing object specific knowledge. 8.2.4 Pragmatic
Information
By'pragmatic information' I intend knowledge of the various types of speech acts and the standard ways of per forming them. This includes facts about the standard func tions of lexical items, their usual contexts and frequency of occurrence, at least. Unlike the previous types of in formation discussed, pragmatic information involves neither technical nor specialized knowledge. All speakers are con stantly confronted with pragmatic information of various sorts, and they presumably store similar amounts of it. Pragmatic information represents a portion of competence within the framework of those language theories which de fine this notion in broad communicative terms (e.g. Hymes 1962, 19 72; Habermas 1971). From the point of view of such theories, pragmatic information is not extra-linguistic at all; it does, however, exceed the information individual lexical entries will contain. Consequently, to be accurate, we should be speaking here of extra-lexical information. Let us now consider the example of sorry. The standard lexical meaning of sorry is emotive, but it occurs very often as an apology, either with a complement or as one word utterance. (7) a. b. d.
I am sorry that I kicked your dog. Sorry that I kicked your dog. Sorry about kicking your dog. Sorry.
As a result of its very frequent occurrence in apologies, it appears many speakers have assigned a secondary meaning to sorry which we can represent as 'I apologize'. I model this situation by entering a secondary 'apologize' meaning for sorry in the lexicon separate from the original entry. In doing so, I claim essentially that speakers do not re hearse the steps connecting the expression of remorse with the act of apologizing for every occurrence of sorry in any apology. Saying sorry "counts as" an apology (in
EXTRA-LINGUISTIC INFORMATION
205
Searle's (1969) terminology), because it expresses remorse on the part of the speaker for an act which he, In express ing his remorse, admits committing in the hope that the victim will believe him contrite and pardon him (cf.Norrick 19 78a). Nevertheless, we do not want to claim that a speaker develops a strategy of apologizing involving the foregoing pragmatic information every time an apologizing situation arises. Instead, he has sorry ready to hand as a standard formula for apologizing under certain circum stances. That is to say that this speaker has lexicalized the meaning 'I apologize' for sorry alongside its original meaning. Given these two separate lexical entries for sorry, we will also want to express the relation holding between them. The sort of via-rule appropriate for accomplishing this must contain pragmatic information, inasmuch as the lexical entries themselves do not. I suggest a via-rule attached to the 'apologize' entry for sorry like this. Via-rule 8: Related to sorry 'remorseful' via the pragmatic rule which states that expressing remorse counts as apologizing under cer tain conditions. If pragmatic rules occur at some point in our language description, via-rules like (8) can simply cite the ap propriate rule, which, of course, will state the conditions of its own application. Via-rule 8 could be simplified to the following, where pragmatic rule 419 identifies the principle in question. Via-rule 9: Related to sorry matic rule 419.
'remorseful' via prag
Notice that this no longer adheres to the form we laid down for semantic via-rules containing extra-linguistic information. Instead, it expresses a pragmatic relation between two lexical entries established on the basis of a pragmatic principle, rather than a semantic one. Indeed none of our regular relational principles could describe the connection between these two readings of sorry. Be cause via-rule 9 expresses the relation between the two meanings of sorry correctly, it appears we should recognize pragmatic via-rules alongside semantic via-rules in the lexicon. This presents no particular problem, however, since lexical entries will be related by means of phono logical via-rules as well on independent grounds. I there fore recognize three separate sorts of via-rules relating
2 06
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
lexical entries. It seems reasonable to posit pragmatic via-rules as a separate type, since all speakers have ac cess to pragmatic knowledge, which is hence appropriately included by citing a principle from pragmatic theory, whereas true extralinguistic (as opposed to simply extralexical) knowledge must be identified as such. Consider now a second example exhibiting a relation which will require a pragmatic via-rule. The lexical item hell must receive two separate lexical entries: one for its literal meaning, one for its use as an expletive indi cative of anger. The latter cannot be very satisfactorily represented by a reading in any direct manner. Instead, the derived use/meaning of hell will be categorized among those items in the lexicon with little or no cognitive meaning but strong connotative value. Clearly, the second ary lexicalization for hell arises as a result of its fre quent occurrence as an expression of exasperation. The pragmatic rule in effect here states that uttering certain lexical items with religious or sexual meanings and those identifying certain body parts or functions counts as (in some cultural matrices) expressing exasperation. In this case, the pragmatic via-rule will include not only the pragmatic rule just cited, but the object specific infor mation that hell is considered the abode of souls damned for all eternity, according to some religious communities. Parallel to semantic via-rules, then, pragmatic via-rules may express a general principle alone or in conjunction with some piece of extra-linguistic knowledge. Another example should clarify the point. please re ceives two separate entries in the lexicon: one on which it means 'give joy', and another for its use as a particle expressing politeness. These two entries should be related, of course, and their relation is pragmatic in nature. The relevant pragmatic principle states that offering choices and appealing to the desires of others counts as polite ness. But this principle alone still does not explain how the meaning 'give joy' relates to offering choices or ap pealing to others' desires. Here the etymological knowledge plays a role that please as a polite form derives from if you please by reduction. An adequate via-rule connecting the polite form of please with its original meaning 'give joy' would take the following form. Via-rule 10: Related to please 'give joy' via prag matic rule 26 (offering choices and ap pealing to others' desires counts as pol iteness) and the etymological fact that
it derives from if
you
please.
EXTRA-LINGUISTIC INFORMATION
207
To summarize, we have seen that via-rules can be ex panded to include extra-linguistic Information, in order to describe relations (some) speakers perceive between pairs of lexical entries which could not be linked on the basis of regular relational principles alone. The device of the via-rule makes it possible to capture such rela tions without complicating lexical readings by the addi tion of extra-linguistic information. We have further seen the necessity for some device linking alternate readings for lexical Items in terms of pragmatic principles. Prag matic via-rules accomplish this; they too can be formu lated to include extra-linguistic information in addition to the pragmatic rules they cite.
POETIC FIGURES
9. 9.0
Introduction
We have been developing a semantic theory intended to handle a range of figurative constructions. This theorycontains an inventory of regular relational principles describing metaphoric extensions and metonymic transfers of lexical readings, as well as a feature extension rule eliminating selectional clashes from anomalous sentences. The data cited in this investigation in support of en riching semantic theory with such devices has been drawn for the most part from everyday language. Our restriction of regular relational principles to those reflecting stan dard modes of sign interpretation was particularly aimed at constraining semantic theory to describe only generally accepted figures. Nevertheless, the foregoing discussion has touched upon figures not uncommon in literary texts at many points. This is, of course, particularly true for those figures involving feature extension, since these are not regular in our restricted sense. In this chapter I wish to apply this enriched semantic theory to poetic texts, in order to demonstrate its strengths and weaknesses in dealing with complex poetic figures. I shall discuss portions of lyric poems chosen to illustrate particular types of figures. Where the sem antic theory presented in this study fails to account for a particular figure, I attempt to sketch an analysis on some other basis. I shall proceed by citing examples fit ting the major classes of metaphoric constructions BrookeRose (1958) recognizes; this should help clarify the dif ference between types of metaphoric constructions, which were Brooke-Rose's concern, and types of meaning extension, which compose the central concern of the present investi gation. All examples are quoted from The Oxford Book of Nineteenth Century English Verse; citations are labeled with the poet's name Poem Title, line numbers; page number in that volume.
210
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
9.1 Simple
Replacement
Brooke-Rose (1953) differentiates various types of metaphoric constructions. Her first type she labels Simple Replacement. This type of construction arises when a meta phorically intended noun phrase replaces a literal one and has no immediately recognizable antecedent in the text, as in (the metaphorically intended term in this and following quotes are italicized): (1)
The blue
deep
thou wingest,
(Shelley To a Skylark,
9; 283),
where blue deep is intuitively interpreted as 'sky'. Given the verb wing, which imposes the selectional restriction [(Gas)] on its direct object nominal, the feature extension rule can copy this feature into the lexical reading for deep, thereby forcing the interpretation of the phrase as 'sky'. No regular relational principle alone would suffice to derive this reading. It therewith counts as an irreg ular metaphorical extension in the present framework. This will be the case for many metaphorical constructions of the Simple Replacement variety. A second example of Simple Replacement, this time without the article (Brooke-Rose separates the two cases on syntactic grounds, but this distinction has no import ant semantic consequences), we see in (2). (2)
Thy
vesper-bell
(Coleridge Youth
hath not yet tolled: -
and Age,
29; 174)
Thy vesper-bell cannot be identified with any other ele ment in the text. Given the context, we intuitively inter pret the phrase as meaning 'thou wilt/you will not die soon'. In this case, no constituent in construction with the metaphoric phrase allows the feature extension rule to assign a feature or features which would warrant its interpretation in this fashion. Here we must rely upon the interpreter's knowledge of the significance of the vesper-bell in relevant cultures, and his association of this knowledge with the existential situation of mortals generally. Coleridge's image is certainly understandable in a poetic context, but does not instance a regular type of meaning extension in the present theory.
POETIC FIGURES
9.2 Pointing
211
Formulae
Brooke-Rose's second type of metaphorical construction she labels Pointing Formulae. Pointing Formulae establish an anaphoric or cataphoric relation to some other constitu ent in the text by means of parallelism, apposition, de monstratives, or the vocative, I.e. by standard mechanisms of textual cohesion (e.g. as discussed In Halliday and Hasan 1976). Consider Wordsworth's (3) She dwelt among th'untrodden ways Beside the springs of Dove, A Maid whom there were none to praise And very few to love. A Violet by a mossy stone Half hidden from the Eye! (Wordsworth [Lucy II ] , 1 - 6; 6 2 ) , where we identify A Violet with A Maid, with She, and ul timately with Lucy in the preceding stanza. Anaphoric re lations in the text permit this identification, so that the figure amounts to an implicit comparison of Lucy with a violet. This granted, Metaphoric Principle 4 (class member-class feature) will relate the lexical reading for violet with a contextual interpretation consisting of any particular feature defining it, where Transformation 2 deletes all other features. The remaining feature may be one inherent in the lexical reading such as (Small), which differentiates violet from large blooms like chrysan themum and carnation, or (Woodland), which differentiates violet from meadow flowers such as rose and tulip. Or it may be a feature associated with violets in the culture, for instance: (Inobtrusive) or (Shy) (cf. shrinking violet). Interpretations involving more than a single one of these features derive from parallel processing, where Metaphoric Principle 4 (object-feature) and Transformation 2 yield a range of mutually compatible readings for the structure. Of course, this set of readings may be derived for violet independently of whether it can be anaphorically related to Lucy or not, but to get combinations like 'inobtrusive person' or 'woodland maiden' they must be combined with the features (Human) (Female) etc. from the antecedent nouns. In the following further example of a Pointing Formula, fair creature has no antecedent, but can be associated with another element in the text, this time cataphorically, viz. with thee addressed in the next line.
212
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
(4)
And when I feel, fair creature of an hour! That I shall never look upon thee more, (Keats [Sonnet], 9 - 10; 369)
Inasmuch as the identification of the presumably human thee with fair creature entails no feature conflict, the present case involves no metaphoric extension whatsoever. Instead, creature as applied to a human illustrates a type of generalizing synechdoche in traditional rhetorical terms. Because generalizing synechdoche can never result in a feature conflict, however, it presents no problem for semantic theory. Coleridge's (5)
This breathing house not built with hands, This body that does me grievous wrong,
(Coleridge Youth
and Age,
8 - 9 ; 173)
involves a cataphoric relation like (4) along with an implicit comparison as in (3). Yet even before house is identified with body in the following line, its colloca tion with the adjective breathing results in an animated interpretation due to the selectional feature [(Animate)] copied into its lexical reading from breathing. The re sultant interpretation is much the same, if Metaphoric Principle 1 (object-model) and Transformation 1 delete (Artifact) from the lexical reading for house, rendering it neutral with respect to the feature (Human).
9.3 The Copula
Link
A third type of metaphoric construction consists in two terms linked by a copula verb, as in (6). (6)
And what is love? It is a doll dress'd up For idleness to cosset, nurse, and dandle; (Keats Modern Love, 1 - 2; 36 8)
Since these lines equate love with doll dress 'd up by means of the copula b e, the feature extension rule can not apply. Instead, Metaphoric Principle 4 (object-feature) along with Transformation 2 will provide various possible interpretations for doll corresponding to individual fea tures of its lexical reading. As in the case of (3), par allel processing will yield a range of interpretations such as 'love is a toy/pretty object/imitation etc.'. Given a copula link between two terms, Metaphoric
POETIC FIGURES
213
Principle 4 will generally derive appropriate readings for the metaphorically intended one. This characteristic groups the Copula Link with Pointing Formulae over against Simple Replacement, which inhibits construal by means of regular relational principles, since it draws no direct comparison. 9.4 The Make Link A fourth type of metaphoric construction involves linkage of two terms with make and similar verbs. "The link with the verb to make (or equivalent) is as cate goric as the copula, but it states the actual process of changing the proper term into a metaphor, as well as the agent who performs or causes the change" (Brooke-Rose 1958: 132). We see a relevant example In (7). (7)
. . . the glow of the fire, Whose puny flaps and freaks the idling Spirit By its own moods interprets, every where Echo or mirror seeking of itself, And makes a toy of Thought. (Coleridge Frost at Midnight, 20 - 24; 162)
These lines do not directly equate thought with a toy, they simply assert that it becomes a toy for the idling spirit. Nevertheless, again by Metaphoric Principle 4 and Transformation 2, the significant feature(s) of toy may be associated with the reading for thought. As in the previous example, a range of readings will be made avail able by parallel processing of the structure. The link with make must be grouped with the Copula Link and Point ing Formulae as a type of metaphoric construction access ible to the device of-regular relational principles. 9.5 The Genetive
Link
There are various cases of the metaphorical construc tion she terms the Genetive Link. Brooke-Rose distinguishes two different cases on a semantic basis. (i) The three-term formula, in which the A of expression identifies a third term, e.g. the messenger of Spring for ' cuckoo'. (ii) The two-term formula, in which the A of expression identifies A with B, e.g. the cloak of death, which im plicitly compares death with a cloak.
214
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Both formulas take various syntactic forms, e.g.
messenger, messenger sent Death . . . his cloak. 9.5.1 The Three- term
by Spring,
Spring's
death-cloak,
Formula
We find a standard occurrence of the three-term formula In (8). (8)
Ah, Sun-flower, weary of time,
Who countest the steps
of
(Blake Ah!
1 - 2; 24)
Sun-flower,
the
Sun,
As Brooke-Rose (1958: 150) points out, the three-term formula in examples like this "is essentially a replacing relationship, and has this in common with Simple Replace ment, that the metaphor has to be fairly self-evident." Consequently, regular relational principles should account for the relevant metaphorical extensions in at least some cases. Since the lexical reading for step may be roughly paraphrased as 'movement of foot', Metaphoric Principle 1 (object-model) relates it to 'movement', while Transforma tion 1 accounts for the deletion of the concrete specifi cation limiting step to feet. The resultant reading 'move ments of the sun' approximates our intuitions about the interpretation of the crucial phrase in (8). If the third term of the three-term formula occurs explicitly in the metaphoric construction, we again have an implicit comparison of essentially the same kind as in Pointing Formulae, the Copula Link and the Make Link. (9)
England hath need of thee: she is a fen of stagnant waters: . . . (Wordsworth London: 1802, 2 - 3; 80)
As in similar examples above, Metaphoric Principle 4 will relate the lexical reading for fen with a contextual interpretation consisting of any particular feature clas sifying it, where Transformation 2 deletes all the other features. Parallel processing allows derivation of the whole range of intuitively assigned Interpretations. Thus England is viewed as stagnant, unhealthy, unproductive etc., or a combination of these.
215
POETIC FIGURES
9.5.2 The
Two-term
Formula
The two-term formula can be illustrated with Blake's: (10)
In what distant deeps of skies Burnt the fire
of
(Blake The Tyger,
thine
eyes?
5 - 6; 23),
where fire and eyes are identified. Again Metaphoric Principle 4 and Transformation 2 will derive various in terpretations for fire consisting of any of its defining or culturally attributed features, say: (Hot), (Yellow), (Dangerous). We thereby arrive at an interpretation paraphrasable as 'hot eyes', 'yellow eyes' or 'dangerous eyes'. The first reading 'hot eyes' might be further in terpreted to mean not that the eyes themselves are at a temperature significantly higher than that of the ambient, but by Metonymic Principle 15 (experience-convention) to mean eyes which someone experiences as hot, i.e. eyes which make someone feel hot. A somewhat more complex example of the two-term for mula can be seen in (11). (11) Pain had no sting, and pleasure flower: (Keats Ode on Indolence,
's wreath
no
18; 375)
Pleasure here is equated with a wreath which has no flowers. The features usually associated with wreath and predicated of pleasure in (11) like (Decorative) (Floral) are negated in Keats' image. We thus end up with an interpretation where pleasure has become the negation of itself. One sometimes finds examples of the Genetive Link ambiguous as to whether they should count as two-term or three-term formulas. The interpretation of Shelley's line: (12)
I fall upon the thorns (Shelley Ode to
the
of
West
life! Wind,
I bleed! 54; 282)
may identify the thorns with life, or seek a third term identified by the whole phrase thorns of life, say: sorrow. On the former two-term interpretation, the feature (Pain ful) or something similar associated with thorn by Meta phoric Principle 4 is predicated of life after Transfor mation 2 deletes the other features. This yields essential ly 'I fall upon (suffer) painful life'. The three-term interpretation 'I fall upon (suffer) sorrow' cannot be
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
216
derived directly. Again by Metaphoric Principle 4 and Transformation 2, the reading for thorn is reduced to 'painful (object)', but the interpreter of this line must still move from 'painful part of life' to 'sorrow'. Our regular relational principles relate lexical readings to contextual readings, but they cannot provide final inter pretations for metaphoric constructions which go beyond the derived readings of lexical items to establish com parisons of this type. 9.6 Verb
Metaphor
Brooke-Roses's sixth major type of metaphoric con struction involves extension of verb meaning. Verbs occur metaphorically in contexts where their literal readings would entail feature clashes. The clashes may arise as selectional violation or as contradictions. Let us first look at an example illustrating a selectional violation obtaining between a verb and its subject noun phrase. (13) The Pansy at my feet Doth the same tale repeat: (Wordsworth Ode: Intimations 54 - 56; 86)
of
Immortality,
The lexical reading for the verb repeat on which it takes a direct object like tale requires a subject noun phrase with the feature (Human). Pansy, of course, is character ized as (Plant) and, redundantly, as (Non-human). The feature extension rule can therefore copy the selectional restriction [(Human)]from repeat into the reading for pansy, triggering the construal rule to delete the feature (Non-human). This gives us the fairy tale reading for (13), on which pansy receives an anthropomorphic interpretation. Alternatively, Transformation 1 deletes the selectional feature (Human) from repeat to yield an interpretation, regularly related to the lexical reading for repeat by Metaphoric Principle 1 (object-model), which can be para phrased as 'signify'. Non-human entities, including plants, can signify something, including a tale, so long as a hu man interpreter is on hand to discern the significance. Metaphoric Principle 1 thereby allows derivation of a con sistent reading on which some interpreter, presumably the poet himself, sees the pansy as signifying what he terms a tale. At the same time, this latter interpretation re presents an instance of regular meaning extension in a
217
POETIC FIGURES verb, while the fairy tale reading cannot claim semantic regularity on our terms. A verb may also relate literally to its subject noun phrase and metaphorically to its object noun phrase. (14)
He who bends to himself a joy Does the winged life destroy; (Blake Eternity, 1 - 2; 26)
In Blake's figure, bend accepts without selectional viola tion the subject noun phrase he identifying a human, but requires in its direct object noun phrase the feature com plex (Object) (Physical), which joy cannot contain, be cause it identifies an emotional state. Metaphorlc Prin ciple 1 relates the lexical reading of bend to a contex tual reading on which it characterizes only an act di rected at changing something (i.e. ((Act) ((Purpose) (Change x)))), where the selectional restriction on x along with the physical aspects of the reading have been deleted by Transformation 1. The reading for (14) involv ing a regular metaphoric extension of the meaning of the verb is thus 'he who changes a joy to/for himself'. The feature extention rule could also copy the selectional feature complex [ (Object) (Physical) ]. from bend into the reading for joy, but this interpretation does not corres pond to intuitions concerning Blake's image. Passive constructions can require a metaphorical in terpretation of a predicate term with respect to a (deep structure) direct object noun phrase appearing as super ficial subject. (15)
My sleep had been embroider'd (Keats Ode to
Indolence,
with dim dreams;
42; 376)
In this line an unidentified agent has "embroidered" dreams on (the poet's) sleep. Embroider must be meta phorically extended to accomodate sleep as a direct ob ject, not to mention dim dreams in the prepositional phrase attached to it. If embroider receives a lexical reading like 'ornament x with needlework', then Transformation 1 must delete the 'with needlework' portion of the reading along with the selectional restriction limiting embroider to a choice of direct object noun phrases with the fea tures (Object) (Physical) (Thin) and (Penetrable). We arrive at a reading for embroider generalized to something such as 'embellish', and at an interpretation for the en tire construction like 'my sleep had been embellished with dim dreams'.
218
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
We also find cases where a verb relates metaphorically to both its subject and object noun phrases. Consider Wordsworth's line (16). (16)
A slumber did my spirit seal, (Wordsworth [Lucy V], 1; 65)
The relevant lexical reading of seal seems to be that paraphrasable as 'shut tightly', which should select a subject noun phrase with the semantic feature (Human) or at least (Animate), and an object noun phrase with the feature (Physical). Neither slumber, characterized as (State) (Bodily), nor spirit, characterized as (Abstract), bears a feature specification appropriate to its function in (16). Metaphorical Principle 1 (object-model) along with Trans formation 1 can account for an interpretation of seal on which its selectional restrictions are relaxed to allow a subject like slumber and an object like spirit, while simultaneously extending its reading to something such as 'close (off)'. This appears to correlate with intuitions about the interpretation of this line. Again, the sorts of readings for slumber and spirit which could be obtained by assigning to them the relevant selectional features from seal appear to play a rather secondary role here. Hopkins' The Wreck of the Deutschland offers a more striking example of the same construction. (17)
And frightful a nightfall folded rueful a day (Hopkins The Wreck of the Deutschland, 117; 857)
Fold imposes selectional restrictions both nightfall and day fail to meet as its subject and object noun phrases respectively. Here again Transformation 1 can delete the selectional features from fold along with other features correlating with them. This leaves fold with practically no content at all, nor does it help account for the intuit ive interpretation of fold as something like 'end'. Per haps, however, semantic theory proper can and should pro vide no more complete description of Hopkins' image. The presumably intended interpretation in any case transcends anything we should be willing to recognize as a regular extension of verb meaning. On semantic grounds, we can include among metaphoric constructions of this type verb phrases containing some copula verb and a prepositional phrase, for example: (18)
The slightest wind was in its nest, (Shelley To Jane: The Recollection , 11; 301)
where the prepositional phrase in its
nest
is predicated
219
POETIC FIGURES
of wind. The primary lexical reading for nest will con tain the features (Habitation) (of Bird); that for wind the features (Movement) (of Air). The semantic theory we have propounded can characterize (18) as contradictory on its literal reading on this basis, but it can not pre dict the metaphori interpretation on which the wind is at rest. The most a semantic theory can offer in this case is the information that the state of air movement is being implicitly compared with a (implied) bird in its nest. To arrive at a complete interpretation, a person must rea son from the extra-linguistic information that a bird in its nest will be at rest, in contrast to a bird in flight, to arrive at an interpretation where the wind is at rest. None of our regular relational principles provides a basis for this comparison. On this seemingly discouraging note, I conclude the present section, because I am as interested in the fail ures of the enriched semantic theory described here as in its successes in dealing with poetic metaphoric construc tions. A semantic theory capable of describing all kinds of poetic language would necessarily be too powerful as a model of ordinary linguistic competence, which is the aim of the present theory. 9.7 Adjective
Metaphor
Brooke-Rose (1958) treats constructions involving meta phorically extended adjective readings as the major type in a class including these along with extended adverbs, pronouns and prepositions. I restrict my attention here to adjective metaphor exclusively, because the theoretical deliberations so far have not extended to adverbs and pre positions, and because the scope of metaphoric extension is drastically restricted, if admitted to exist at all (cf. Brooke-Rose 1958: 253 ff. ) . 9.7.1 Metaphorically
Interpreted
Adjectives
Metaphoric adjective constructions, like literal ones, fall into two syntactic groups: those involving attribu tive adjectives and those involving predicate adjectives. (19) and (20) illustrate metaphorical attributive and meta phorical predicate adjective constructions respectively. (19)
The dense and the driven Passion, (Hopkins The Wreck of the Deutschland,
53; 854)
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
220 (20)
For the infinite air is unkind, (Hopkins The Wreck of the Deutschland,
100; 856)
As to (19), both dense and driven select noun phrases specified as (Object) (Physical), whereas passion will be classified as (State) (Emotion), so that anomaly arises on the literal level. Transformation 1 can delete the selectional restriction [ (Object) (Physical) ] from dense and driven along with whatever features correlate with them, just as the feature extension rule could copy this selectional restriction into the reading for passion. It goes beyond the scope of the present theory to make sense of dense and driven applied to an emotional state, and/or of a concretized passion. The anomaly in (20) derives from the fact that unkind selects noun phrases specified (Human), whereas air must be characterized redundantly as (Non-human). Here the feature extension rule would yield an anthropomorphic in terpretation for air. More in line with our intuitive in terpretation of the line, Metonymic Principle 15 (exper ience-convention) allows a transfer of the lexical reading for unkind to something like 'experience of unkindness'. Thus, the air is implicitly being compared with the ex perience of human unkindness. This approaches fairly close ly the interpretation one might intuitively offer for (20), viz. 'the air makes one feel as human unkindness does'. On semantic grounds, some adjective metaphors may parallel noun metaphors of the Pointing Formula or Copula Link varieties. We can paraphrase dense passion as passion which is a dense object, and the air is unkind as the air is an unkind person, thereby transforming both into noun metaphors. But adjective metaphors may parallel verb meta phor as well. The driven passion could be interpreted as de rivative from someone/something drives the passion. A per son might drive passion as he drives/steers a vehicle; a storm might drive passion as it drives/blows leaves. As Brooke-Rose would have it, "the adjective, in fact, hovers between the noun and the verb" (1958: 239), in its meta phoric possibilities. 9.7.2 Transferred
Adjectives
As it turns out, the adjective possesses metaphoric capabilities neither the noun nor the verb can match. A metaphoric adjective may involve a transfer from one (per haps implicit) constituent of a construction to another. This results in what the classical rhetoricians term
221
POETIC FIGURES
hypallage. I propose to call adjectives in such construc tions transferred adjectives. The lyric poetry of Keats contains many clear examples of constructions containing transferred adjectives, so that we may draw all our ex amples for the following discussion from this corpus. In Keats' (21) And silent was the flock in woolly (Keats The Eve of St,
Agnes,
4;
fold: 338)
not the fold is literally woolly, but rather the flock or its members. Similarly, Ruth stands amid not corn which is alien, but amid familiar enough corn in an alien land in: (22) Perhaps the self-same song that found a path Through the sad heart of Ruth, when, sick for home, She stood in tears amid the alien corn; (Keats Ode to a Nightingale
, 65 - 67; 354)
With respect to constructions of this type, the semantic theory here developed provides little enlightenment. A fold could in fact be woolly and corn might be alien to someone. (2 1) and (22) are not anomalous on the semantic plane; they simply describe rather unlikely states of af fairs, given our knowledge of the world. I propose to derive constructions like (21) and (22) from normally interpretable underlying structures by means of an overextension of an otherwise regular syntactic pro cess which preposes an adjective from within a relative clause or prepositional phrase modifying a noun to attrib utive position before it1. Given the phrase (2 3) creature which has a woolly hide, this syntactic process places woolly before creature and deletes the remainder of the clause. Woolly fold can be derived in parallel fashion from (24). (24) fold which contains woolly creatures. Without knowing the conditions syntactic and semantic the rule preposing adjectives imposes on its input in regular cases, we cannot determine why its application to (23), but not to (2 4) , produces an acceptable construction. We must simply assume that some condition has been relaxed in its application to (24). The source for alien corn, pre sumably something such as (25), apparently also fails to
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
222
exactly fulfill the input conditions for adjective preposing, since it exhibits an irregularity similar to that of woolly
(25)
fold,
corn (which is) in an alien land
Keats' (26)
Tasting of Flora and the country green, Dance, and Provencal song, and sunburnt mirth! (Keats Ode to
a Nightingale
, 13 - 14; 352)
differs from the foregoing examples in that mirth clearly fails to meet the selectional restriction[(Human) ]imposed by sunburnt. Regular relational principles or feature ex tension might apply to remove the feature clash, but again the intuitively correct reading seems to be one involving a transferred adjective, as in (27), (2 7)
the mirth of sunburnt people,
so that an overextension of the process of adjective preposing will account for the construction. Contrast Keats' sunburnt mirth with a phrase from everyday language like guilty smile. Superficially the two resemble each other. One might surmise that guilty smile derives from (28), in the same way sunburnt mirth does from (27). (2 8)
the smile of a guilty person
Notice, however, that guilty as applied to people can be regularly related to its interpretation in guilty smile by Metonymic Principle 16 (manifestation-definition), since the property of being guilty as it applies to a person naturally extends to his smile as the manifestation of this property. No such regular relation holds for the ex amples from Keats, so they must be treated as transferred adjectives. A somewhat different type of adjective construction also stems from Keats. (2 9)
The gradual sand that through an hour-glass runs, (Keats {Sonnet'], 13; 368)
The noun sand fails to fulfill the selectional restriction [ (Process) ] imposed by gradual. A consistent interpreta tion of (29) would associate gradual in adverb form with
POETIC FIGURES
the verb run, (30)
223
as in (30).
The sand that gradually runs through an hour glass
If the process of adjective preposing were further overex tended to extract an adverb from a relative clause modify ing a noun and to place it before this noun, where it takes on adjective form, then it would account for (29) on the basis of a source structure such as ( 3 0 ) . We find examples in everyday language which indicate that adjective pre posing can sometimes undergo an extension of this kind. For instance, a quick pipe in (31) is a pipe smoked quickly rather than a pipe capable of rapid acceleration. (31) Judy smoked a quick
pipe.
Similarly, a hot sax in (32) is a saxophone played hotly or in the hot jazz style, not an instrument with a tem perature significantly above that of the ambient. (32) Judy plays a hot
sax.
Further, the adjective fatherly in (33) is interpreted as applying to the manner in which Harry places his hand rather than to the hand itself. (33) Harry placed a fatherly
hand on Sue's shoulder.
I assume here that (33) derives from an underlying struc ture like (34). (34) Harry placed a hand on Sue's shoulder in a fatherly manner. Keats transfers adjectives in still another way. (35) Season of mists and mellow fruitfulness, Close bosom-friend of the maturing sun; (Keats To Autumn, 1 - 2; 359) In (35) maturing does not directly modify sun; the sun causes the maturing of fruits and vegetables, but is not itself in the process of maturing. One similarly interprets noon in (36) as causing drowsiness, rather than as being drowsy itself.
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
224 (36)
For Poesy! - no, - she has not a joy, At least for me, - so sweet as dro wsy noons, And evenings steep'd in honied indolence; (Keats Ode on Indolence, 35 - 37; 375)
Now the relation from the sun causes turing sun, and from the noon causes
x to mature to ma x to be drowsy to
dro wsy noon parallels a process of word formation more closely than one of adjective preposing. In particular, it approaches the process deriving melting pot from pot which causes x to melt, as opposed to pot which is melting, and happy hour from hour which causes x to be happyt as op posed to hour which is happy. Melting pot and happy hour
naturally exhibit the intonation contour of compounds, whereas maturing sun and drowsy noons do not, at least on the basis of the metric structures of the poems in which they occur. One might thus propose that the latter two be derived from causative underlying structures which act as input to a rule of word formation. This rule applies ir regularly in that it fails to assign its output the struc ture of a compound: maturing sun and drowsy noons end up as simple noun phrases containing attributive adjectives. Consequently they are assigned standard intonation con tours in the phonological component. 9.8
Summary
We have reviewed a wide range of metaphoric construc tions following the classificational scheme of Brooke-Rose (1958). In many metaphorical constructions, regular rela tional principles can account for the contextual interpre tations of lexical items on the basis of their lexical readings; they do not, nor were they intended to, describe the semantic relation holding between a metaphorically in terpreted expression and its extra-sentential context. The enriched semantic theory propounded here derives a set of figurative interpretations for a structure by predicting possible contextual readings for its component lexical items, which can then be amalgamated by the projection rule. Sometimes the failure of our semantic model to obtain an intuitively satisfactory interpretation for a construc tion reflects its syntactic irregularity, as in the case of transferred adjectives. Exhaustive syntactic analysis of such constructions may reveal well-formed underlying structures interpretable along normal lines. Other cases exhibit genuinely semantic irregularities, e.g. Hopkins'
POETIC FIGURES
225
And frightful a nightfall folded rueful a day and The dense and driven Passion. When the semantic model provides no complete interpretations for the lexical items in such constructions, they simply represent meaning extensions and/or transfers exceeding the scope of its regular prin ciples. In restricting the foregoing discussion to types of metaphorical constructions classified by Brooke-Rose (1958), we neglected cases in which lexical items undergo a change in their categorial syntactic classification. Weinreich (1966: 461 ff.) considers examples such as the following. (37)
Scientists study the if.
(38)
He trues the rumour.
He proposes a redistribution rule as part of his semantic calculator which would assign the deviant items appro priate syntactic features from their environments. But this amounts to an essentially syntactic treatment. McCawley (1968b) too argues that such sentences instance over-extensions of productive or semi-productive syntactic processes, just as transferred adjectives do (cf. note 1). Semantic theory proper therefore need not deal with ex amples of this sort. Our enriched semantic theory fails to provide (com plete) interpretations for a range of poetic figures, but this is no reason to discard or alter the theory. In fact, the theory's failure for certain constructions argues its value in establishing a boundary between truly poetic figures and those commonly found in everyday language. A figure not interpretable on the basis of regular prin ciples in an enriched semantic theory exemplifies the poetic function (Jakobson 1960) to a greater degree than one interpretable on this basis. Sentences interpretable in a non-enriched semantic theory like that of Katz (1972) exemplify the poetic function to an even lesser degree than those requiring application of one or more regular relational principles or the feature extension rule. The fact that the semantic theory fails in crucial cases thus renders it of greater interest for a theory of poetics.
CONCLUSION
Let us review briefly the main conclusions of the fore going investigation. First, I set forth general semiotic principles relating signs of different types in various sign systems. The prin ciples fell into two groups: those describing similarity relations (iconic principles) and those describing conti guity relations (indexical principles). These principles were tested for the properties of multiple application, productivity, systematization and validity of inference. We ended up with an iconic code consisting of four prin ciples along with two semantic transformations which delete features, and an indexical code consisting of seventeen indexical principles in six major classes. Second, I argued that regular semantic relations were special cases of motivated sign relations, and that regular principles in semantic theory could be derived from the semiotic principles already ascertained. Then I showed that linguistic semantic competence represents a special case of general semiotic, sign manipulating competence. So considerations of simplicity dictate that regular semantic principles be derived as special cases of general semiotic principles and be characterized as redundant in semantic theory. I formulated one set of metaphoric and another of metonymic relational principles; then I outlined their in tegration into and their functioning within semantic the ory. The rest of the study was devoted to investigating the consequences of enriching semantic theory with these sets of regular relational principles, especially with re spect to traditional problem areas for semantics. The first problem addressed was that of polysemy, which was split into the two closely related cases of lexical and derived polysemy. After defining polysemy, I went on to review contemporary treatments of both lexical and derived varieties. It was shown that no current theory can handle both lexical and derived polysemy in a consistent manner, nor even account for their common characteristics in a
228
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
revealing way. In addition, few scholars have made any attempt to distinguish truly regular relations (and deri vations) from non-regular, spurious cases. In the next two chapters I presented my own treatment of polysemy within an enriched semantic theory. The discussion of derived polysemy began with a con sideration of anomaly, contradiction and selection. Cur rent semantic theories have often mixed knowledge of the world with knowledge of word meaning, thereby blurring the lines between anomaly, contradiction and contingency. I formulated the principle that anything which could logi cally happen must be describable without anomaly. This principle dictates that selectional restrictions should be simplified and their content transferred to readings proper as far as possible. At the same time, lexical read ings should be constrained to consist solely of distinct ive features. The net effect is a decrease in the number of sentences marked anomalous and a recognition of their contingent nature. I contrasted my enriched semantic the ory with the standard Katzian interpretive theory, reveal ing the strengths of the former and the weaknesses of the latter especially in dealing with metaphoric meaning ex tension and metonymic meaning transfer. In particular the regular relational principles in the enriched theory allow it to obtain standard readings for figurative construc tions . The discussion of lexical polysemy was embedded within a general description of the lexicon; I presented a model of the lexicon structured around a network of semantic classes, or fields. The semantic classes provided the basis for definitions of hyponymy, synonymy, disjunction and vari ous types of oppositeness of meaning. We also saw how read ings and selectional restrictions for the lexical entries they contain can be directly generated from these semantic classes. I then explored the effects regular relational principles have on the lexicon just described. I introduced and motivated the notion of the semantic via-rule for cap turing semantic relations between lexical entries in nonadjacent fields. Via-rules allow us to define lexical poly semy in terms of the regular relational principles in sem antic theory; they thereby provide the link sought be tween lexical and derived polysemy. This link makes it possible to characterize the process of lexicalization in a consistent manner as well. The inclusion of regular relational principles in sem antic theory enable it to account for motivation in mor phologically complex units. I argued that via-rules nat urally account for (1) componential motivation, viz. by
CONCLUSION
229
relating idiomatic readings of complex units to the lexi cal readings of their component parts, and for (2) global motivation, viz. by relating idiomatic readings of complex units to their literal readings as wholes. I demonstrated how this can be done for compounds and for stock phrases of various kinds, and provided a classification of com plex units based on their type and degree of motivation. Finally, I described the process of semantic interpreta tion as it concerns complex units in texts. Speakers establish relations between lexical entries with the help of extra-lexical information in many cases. After considering the role of extra-lexical information in semantic relations generally, I differentiated and dis cussed four kinds of extra-lexical information, viz. ety mological, historical, object specific and pragmatic, which must be included in via-rules to capture the rela tions in question. I demonstrated how via-rules can be ex tended to include extra-lexical information. We further noted the need for and the form of pragmatic via-rules, as contrasted with semantic via-rules. In order to assess the applicability of my enriched semantic theory to sophisticated figurative language, I applied it to selected passages from lyric poems. Where the semantic theory failed to account for a particular type of figure, I attempted to sketch an alternative ana lysis. But the failures of the enriched semantic theory in treating poetic figures were important in themselves as evidence that the theory described truly regular types of figures. On the other hand, its successes demonstrate its value as part of a theory of poetics. I hope to have convincingly shown that semiotic theory can supply semantic theory with a set of regular rela tional principles for describing relations of meaning and motivation, and to have clearly illustrated the advantages of a semantic theory so enriched. I view the individual analyses, the particular rules and even the rule forms described as secondary to the spirit of the theory ad vanced. In principle, any interpretive or generative sem antic theory could be enriched with regular relational principles of the general type described here. The precise form the relational principles and other devices proposed above will take in any given semantic theory must be de termined by those working with the particular model. One problem for any adaptation of the semiotic and/or semantic principles will lie in the degree of delicacy desired. As stated above, I aimed for an intermediate de gree of delicacy. One semantic theory with goals different than those described here might find, say, my six basic
230
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
indexical/metonymic types sufficient even without the sub-classes I recognized within them, whereas another might require a more delicate classification of metaphor types for certain purposes. If the semantic theory presented above were to be integrated into a theory of poetics, for instance, additional kinds of figures should probably be recognized as regular, or at least as 'regular in lyric poetry', 'regular In Euphuistic Renaissance prose' etc. Further adjustments in the notion of regularity could be similarly tied to different genres, periods, styles, indi vidual writers etc. A related problem not given much attention above is the matter of scaling different principles or figures as to their relative degrees of regularity. While any given semantic connection should be considered regular so long as one of our principles describes it, clearly some of the principles and the relations they describe are more regu lar than others. Or perhaps certain types are simply more common, which leads to the question of the relation be tween frequency and (our perception of) regularity. As signing degrees of regularity or positions on a scale to semantic principles and/or the relations they describe, as well as the link between frequency and regularity are problems which must be addressed in future work. I have myself begun to investigate the frequency relations of different figures in proverbs. Under the not unreasonable assumption that proverbs exhibit only truly regular types of figures, a frequency scaling of proverbial figures should approximate a scaling of regular figures generally. Obviously there is much to be done in these areas.
NOTES
Notes
to Chapter
1
l)The search for constraints in transformational gram mar was headed by Ross's (1967) work on variables in syn tactic rules. Postal's (1971) discussion of cross-over phenomena as well as the successive revisions of the socalled Standard Theory (e.g. Chomsky 1972, 1973; Perlmutter 1971; Lasnik and Fiengo 1974; Emonds 1976, Fiengo 1977) represent further attempts to constrain rules and rule types. The arguments in favor of restricting the power of syntactic principles hold mutatis mutandis for constrain ing semantic principles. 2)This account ignores investigations which posit as yet unformalized rules linking entries to phonologically, mor phologically, semantically and encyclopedically related items of all varieties (e.g. Macnamara 1971; Fromkin 1971; Fay and Cutler 1977), since their concern is not with meaning relations as such, and because they offer no con crete proposals about the form of the linking rules. Apresjan, Mel'chuk and Zolkovsky (1969), on the other hand, make a number of explicit suggestions about the form a dictionary should take, but they apparently do not intend to model the mental lexicon, which is at issue here. 3)The notion word can, of course, be defined in other than semantic terms. Depending upon the particular goal a linguistic analysis sets itself, word or possible word may be defined as a phonological, morphological or syntactic unit. In the following, I use the expression word in its pre-theoretical sense. The crucial terms in this investiga tion are lexical entry and lexical item. A lexical entry consists of a phonological form associated with a semantic reading at a specified point in the lexicon. A lexical item
232
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
is an abstract entity represented by the set of all the lexical entries with a particular phonological form and connected by via-rules. A lexical item with a single lexi cal entry is unitary; a lexical item with multiple entries is polysemous. 4)This should not be construed as aplea for a onesidedly semantic lexicon. In fact, as we shall see below, phonological and morphological relations must play a role as well, even within a semantically based lexical theory. Nevertheless, since this study approaches the lexicon from the point of view of semantics, its primary focus will be on semantic aspects of lexical organization. Phonological identity or distinctness among lexical entries, as well as the presence or absence of morphological connections are presumed to be given in a complete model of the lexicon rather than built explicitly into the present framework.
Notes
to Chapter
2
1)I ignore in the following Peirce's distinction between Icon and iconic sign or hypoicon, as well as that between Index and subindex, distinctions which Peirce himself often disregarded. 2)This formulation was not strictly necessary in the case of the iconic principles, since, being cast in terms of features and classes, they naturally involved concepts rather than objects as such.
Notes
to Chapter
3
1)We will usually omit the expression 'the meaning of' or the 'semantic representation of' in the following, and write simply 'between word x and word y', where the context makes clear that we are concerned with semantic interpreta tions rather than with words as such. 2)Katz also developed a theory of semi-sentences for such cases; cf. Katz (1964) and below.
NOTES
Notes
to Chapter
233
4
l)The requirement of non-distinctness for deletion operations is not as strict as once thought. See Ross (1969) on the notion of 'sloppy identity' in deletion, and Nunberg (1978). 2)There has been considerable controversy about whether coordinate conjoined structures involve deletion opera tions, and about their derivation generally. See Koutsoudas (1971), Hudson (1976) and references there.
Notes
to Chapter
5
l)We assume here, following Searle (1969), that proposi tions are made not only by declarative sentences, but by questions and commands as well. In Searle's terms, speech acts of all kinds necessarily involve acts of propositionalizing, and, this being so, all contain recognizable prop ositions at one level of analysis.
Notes
to Chapter
6
l)To be really complete, via-rule 3 ought to include the additional, extra-linguistic information that the fox is considered a sly animal, since no feature (sly) will appear in the lexical reading of fox itself and this feature must consequently be supplied from the 'encyclopedia entry' for fox. I will return to the problem of including this sort of information in the via-rule notation in chapter 8.
Notes
to Chapter
9
l)This proposal ultimately derives from and represents an extension of a treatment McCawley (1968b) suggests for certain constructions originally discussed by Weinreich (1966). Weinreich's examples parallel those involving transferred adjectives in relevant respects; see 9.8 below.
BIBLIOGRAPHY
Abraham, Werner. 1975. A Linguistic Approach to Metaphor. Lisse: de Ridder. and Kurt Braunmüller. 1971. "Stil, Metapher und Pragmatik". Lingua 28.1-47. Alston, William P. 1971. "How Does One Tell Whether a Word has One, Several or Many Senses?". Semantics
disciplinary chology
Reader
in Philosophy,
Linguistics
: An
and
Inter
Psy
ed. by Danny D. Steinberg and Leon A. Jakobovits,
35-4 7. Cambridge: Cambridge University Press. Apresjan, Ju. D. 1974. "Regular Polysemy". Linguistics 142.5-32. , Igor A. Mel'chuk, and A. K. Žolkovsky. 1969. "Semantics and Lexicography: Towards a New Type of Unilingual Dictionary". Studies in Syntax and Semantics ed. by Ferenc Kiefer, 1-33. Dordrecht: Reidel. Barley, Nigel. 1972. "A Structural Approach to the Proverb with Special Reference to the Anglo-Saxon Corpus". Proverb ium 20.737-50. Baumgärtner, Klaus. 1969. "Der methodische Stand einer linguistischen Poetik". Jahrbuch für Internationale Germanistik 1:1.15-43. Bendix, Edward H. 1971. "The Data of Semantic Description". Semantics : An Interdisciplinary Reader in Philosophy, Linguistics
and
Psychology
ed. by Danny D. Steinberg
and Leon A. Jakobovits, 39 3-409. Cambridge: Cambridge University Press. Bergmann, Merrie. 19 79. "Metaphor and Formal Semantic The ory". Poetics 8.213-30. Black, Max. 1952. Critical Thinking. Englewood Cliffs: Prentice-Hall. . 1962. Models and Metaphors. Ithaca: Cornell University Press. Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart & Winston. Brekle, Herbert E. 1970. Generative Satzsemantik und Trans formationelle Syntax im System der englischen Nominal— komposition. Munich: Fink.
236
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Brooke-Rose, Christine. 1958. A Grammar of Metaphor. London: Secker & Warburg. Chafe, Wallace L. 1968. "Idiomaticity as an Anomaly in the Chomskyan Paradigm". Foundations of Language 4.109-27. . 1970. Meaning and the Structure of Language. Chicago: University of Chicago Press. Chao, Yuen Ren. 1959. "Ambiguity in Chinese". Studia serica Bernhard Karlgren dedicata ed. by SØren Egerod and Else Glahn, 1-13. Copenhagen: Munksgaard. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: M.I.T. Press. . 1970. "Remarks on Nominalization". Readings in English Transformational Grammar ed. by Roderick J. Jacobs and Peter S. Rosenbaum, 184-221. VJaltham, Mass.: Ginn. . 1972. Studies on Semantics in Generative Gram mar, The Hague: Mouton. . 1973. "Conditions on Transformations". A Fest schrift for Morris Halle ed. by Stephan R. Anderson and Paul Kiparsky, 2 32-86. New York: Holt, Rinehart & Winston. Cohen, L. Jonathan and Avishai Margalit. 19 72. "The Role of Inductive Reasoning in the Interpretation of Meta phor" . Semantics of Natural Language ed. by Donald Davidson and Gilbert Harman, 722-40. Dordrecht: Reidel. Cohen, Ted. 19 75. "Figurative Speech and Figurative Acts". The Journal of Philosophy 72.669-84. Coseriu, Eugenio. 1967. "Lexikalische Solidaritäten". Poetica 1.293-303. Cowie, A. P. and Ronald Mackin, eds. 1975. Oxford Diction ary of Current Idiomatic English: Verbs with Preposi tions and Particles. Oxford: Oxford University Press. Davidson, Donald. 19 80. "What Metaphors Mean". Critical Inquiry 5:1.31-47. Dijk, Teun A. van. 1970. "Neuere Entwicklungen in der literarischen Semantik". Text, Bedeutung, Ästhetik ed. by S. J. Schmidt, 43-79. Munich: Bayerischer Schul buchverlag. . 1972. Some Aspects of Text Grammars. The Hague: Mouton. . 19 75. "Formal Semantics of Metaphorical Dis course". Poetics 4.173-98. Di Pietro, Robert J. 19 76. "The Role of Metaphor in Lin guistics". Linguistic and Literary Studies in Honor of Archibald A. Hill. Volume 1: General and Theoretical Linguistics ed. by Mohammad Ali Jazayery, Edgar C. Polomé, and Werner Winter, 99-107. Lisse: de Ridder.
BIBLIOGRAPHY
237
Droste, F. G. 19 76. "Semantics as a Dynamic Device: Redun dancy Rules in the Lexicon". Linguistics 182.5-33. Dubois, Jacques et al. 1970. Rhétorique générale . Paris: Larousse. Eco, Umberto. 19 72. "Introduction to a Semiotics of Iconic Signs". Versus 2.1-15. . 1976. A Theory of Semiotics. Bloomington: Indiana University Press. Emonds, Joseph. 1976. A Transformational Approach to Eng lish Syntax: Root3 Structure Preserving3 and Local Transformations. New York: Academic Press. Fay, David and Anne Cutler. 1977. "Malapropisms and the Structure of the Mental Lexicon". Linguistic Inquiry 8.505-20. Fiengo, Robert. 19 77. "On Trace Theory". Linguistic In quiry 8.35-5 1 . Fillenbaum, Samuel and Amnon Rapoport. 1971. Structures in the Sub ¿ective Lexicon. New York: Academic Press. Fillmore, Charles J. 1968. "The Case for Case". Universale in Linguistic Theory ed. by Emmon Bach and Robert T. Harms, 1-90. New York: Holt, Rinehart & Winston. . 19 71a. "Types of Lexical Information". Seman tics: An Interdisciplinary Reader in Philosophy„ Lin guistics and Psychology ed. by Danny D. Steinberg and Leon A. Jakobovits, 370-92. Cambridge: Cambridge Uni versity Press. 1971b. "Verbs of Judging: An Exercise in Sem antic Description". Studies in Linguistic Semantics ed. by Charles J. Fillmore and D. Terence Langendoen, 273-89. New York: Holt, Rinehart & Winston. 19 75. "The Case for Case Reopened". Mimeo, Berkeley: University of California at Berkeley. Fodor, J. D., J. A. Fodor, and M. F. Garrett. 1975. "The Psychological Unreality of Semantic Representations". Linguistic Inquiry 6.515-31. Forster, Kenneth I. 1976. "accessing the Mental Lexicon". Neu Approaches to Language Mechanisms ed. by R. J. Wales and Edward Walker, 257-87. Amsterdam: NorthHolland. Fraser, Bruce. 1970. "Idioms Within a Transformational Grammar". Foundations of Language 6.22-42. Fromkin, Victoria A. 19 71. "The Non-anomalous Nature of Anomalous Utterances". Language 47.27-52. Gak, V. G. 19 76. "The Semantic Structure of the Word as a Semantic Component of the Utterance". Linguistics 180. 27-41. Geckeler, H. 1971. Strukturelle Semantik und Wortfeldrtheorie. Munich: Fink.
238
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Goldstein, Kurt. 1948. Language and Language Disturbances. New York: Grune & Stratton. Gruber, Jeffrey S. 1976. Lexical Structures in Syntax and Semantics. Amsterdam: North-Holland. Habermas, J. 1971. "Vorbereitende Bemerkungen zu einer Theorie der kommunikativen Kompetenz". Theorie der Gesellschaft oder Sozialtechnologie ed. by J. Habermas and N. Luhmann, 101-41. Frankfurt/Main: Suhrkamp. Halliday, Michael A. K. and Ruqaiya Hasan. 1976. Cohesion in English. London: Longman. Hardwick, Charles S., ed. 1977. Semiotic and Significs: The Correspondence between Charles S. Peirce and Victoria Lady Welb y. Bloomington and London: Indiana University Press. Harman, Gilbert. 1977. "Against Universal Semantic Repre sentation". Proceedings of the Texas Conference on Per formatives _, P r e s u p p o s i t i o n s , and Implicatures ed. by Andy Rogers, Bob Wall, and John P. Murphy, 1-11. Arlington, Va.: Center for Applied Linguistics. Hockett, Charles F. 1956. "Idiom Formation". For Roman Jakobson ed. by Morris Halle et al., 222-29. The Hague: Mouton. . 1958. A Course in Modern Linguistics. New York: Macmillan. Hooper, Joan B. 1976. An Introduction to Natural Generative Phonology. New York: Academic Press. Horrock, Roger. 19 76. "Metaphor in Competence and Perform ance". Papers in Linguistics 9.149-60. Hudson, Richard A. 1976a. "Conjunction-Reduction, Gapping, and Right-Node Raising". Language 52.535-62. . 1976b. "Regularities in the Lexicon". Lingua 40.115-30. . 1976c. Arguments for a Non-transformational Grammar. Chicago: Chicago University Press. Hume, David. 1748. "An Enquiry Concerning Human Under standing". The English Philosophers from Bacon to Mill ed. by Edwin A. Burtt. 1939,585-689. New York: Modern Library. Hymes, Dell. 1962. "The Ethnography of Speaking". Anthro pology and Human Behavior ed. by T. Gladwin and W. C. Sturtevant, 15-33. Washington, D.C.: Anthropological Society of Washington. . 1971. "Competence and Performance in Linguistic Theory" Language Acquisition : Models and Methods ed. by R. Huxley and E. Ingram, 3-2 3. London and New York: Academic Press. Jackendoff, Ray S. 1971. "Modal Structure in Semantic Representation". Linguistic Inquiry 2.479-514.
239
BIBLIOGRAPHY
Jackendoff, Ray S. 1972. Semantic Interpretation in Gener ative Grammar. Cambridge, Mass.: M.I.T. Press. . 19 75. "Morphological and Semantic Regularities in the Lexicon". Language 51.6 39-71. Jakobson, Roman. 1969. Kindersprache, Aphasie und allge meine
Lautgesetze
. Frankfurt/Main: Suhrkamp.
. 1953. "Aphasia as a Linguistic Topic". Selected Writings II ed. by Roman Jakobson, 1971, 229-38. The Hague: Mouton. . 1954. "Two Aspects of Language and Two Types of Aphasic Disturbances". Selected Writings II ed. by Roman Jakobson, 1971, 239-59. The Hague: Mouton. . 1960. "Linguistics and Poetics". Style in Language ed. by Thomas A. Sebeok, 350-377. Cambridge, Mass.: M.I.T. Press. Kandinsky, Wassily. 1963. Über das Geistige in der Kunst. Bern: Benteli. Kastovsky, Dieter. 1980. "Zur Situation der lexikalischen Semantik". Perspektiven der lexikalischen Semantik ed. by D. Kastovsky, 1-13. Bonn: Bouvier Verlag. Katz, Jerrold J. 1964. "Semi-sentences". The Structure of Language : Readings in the Philosophy of Language ed. by Jerry Fodor and Jerrold J. Katz, 400-416. Englewood Cliffs: Prentice-Hall. . 1966. The
Philosophy
of
Language
. New York:
Harper & Row. . 1971. "Generative Semantics is Interpretive Semantics". Linguistic Inquiry 2.313-31. . 1972. Semantic Theory. New York: Harper & Row. . 1976. "Global Rules and Surface Structure Interpretation". An Integrated Theory of Linguistic Ability ed. by Thomas G. Bever, Jerrold J. Katz and D. Terence Langendoen, 415-25. New York: Crowell. . 19 77a. "The Real Status of Semantic Representa tions". Linguistic Inquiry 8.559-84. . 1977b. Propositional Structure and Illocutionary Force. Hassocks, Sussex, England: Harvester. and Paul M. Postal. 1964. An Integrated Theory of Linguistic Descriptions. Cambridge, Mass.: M.I.T. Press. Kayser, Wolfgang. 1948. Das sprachliche Kunstwerk. Bern and Munich: Franke. Koller, Werner. 1977. Redensarten. Tübingen: Niemeyer. Kooij, Jan G. 1968. "Compounds and Idioms". Lingua 21. 250-68. . 1971. Ambiguity in Natural Language. Amsterdam: North-Holland.
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
240
Koutsoudas, Andreas. 1971. "Gapping, Conjunction Reduc tion, and Coordinate Deletion". Foundations of Lan guage 7.337-86. Lakoff, George. 1969. "On Derivational Constraints". Papers from the Fifth
Regional
Meeting of
the
Chicago
Linguistic Society ed. by Robert I. Binnick et al., 117-39. Chicago: Chicago Linguistic Society. . 1970a. "Global Rules". Language 46.627-39. . 1970b. "A Note on Vagueness and Ambiguity". Linguistic
Inquiry
1.357-59.
Lasnik, Howard and Robert Fiengo. 19 74. "Complement Object Deletion". Linguistic Inquiry 5.535-71. Leech, Geoffrey N. 1969. Towards a Semantic
English.
Description
of
London: Longmans. . 19 74. Semantics. Harmondsworth: Penguin.
Lees, Robert B. 1960. The Grammar of English
Nominaliza-
tions. Bloomington: Indiana University. Lehmann, Volkmar. 1975. Metapher und semantische Beschrei bung, Giessen: Schmitz. Lehrer, Adrienne. 19 74a. "Homonymy and Polysemy: Measuring Similarity of Meaning". Language Sciences 3.33-39. . 1974b. Semantic
Fields
and Lexical
Structure
.
Amsterdam: North-Holland. . 19 78. "Structures of the Lexicon and Transfer of Meaning". Lingua 45.95-123. Leisi, Ernst. 1961. Der Wortinhalt : Seine
Struktur
im
Deutschen und Englischen. Heidelberg: Quelle & Meyer. Levin, Samuel R. 1977. The Semantics of Metaphor. Baltimore and London: Johns Hopkins Press. Lewis, David. 19 72. "General Semantics". Semantics of Natural Language ed. by Donald Davidson and Gilbert Harman, 169-218. Dordrecht: Reidel. Lipka, Leonhard. 1972. Semantic Structure and Word Forma tion. Munich: Fink. Ljung, Magnus. 1970. English Denominal Adjectives .
Göteborg: Acta Universitatis Gothoburgenis. Luria, A. R. 1972. The Man with a Shattered World: The History of Brain Wound. New York: Basic Books.
guistics Current
. 1973. "Two Kinds of Aphasic Disorders". Lin 115.57-66. 1974. "Basic Problems of Neurolinguistics". Trends in Linguistics
Volume
12 ed. by Thomas
A.Sebeok, 2561-94. The Hague: Mouton. Lyons, John. 1968. Theoretical Linguistics. Cambridge: Cambridge University Press. . 19 77. Semantics. Cambridge: Cambridge Univer sity Press. Macnamara, John. 1971. "Parsimony and the Lexicon". Language 47.359-74.
BIBLIOGRAPHY
241
Makkai, Adam. 1972. Idiom Structure in English. The Hague: Mouton. Marchand, Hans. 1969. The Categories and Types of PresentDay English Word Formation. Munich: Beck. Matthews, Robert J. 19 71. "Concerning a'Linguistic Theory' of Metaphor". Foundations of Language 7.413-25. McCawley, James D. 196 8a. "The Role of Semantics in a Grammar". Universale in Linguistic Theory ed. by Emmon Bach and Robert T. Harms, 12 4-69. New York: Holt, Rinehart & Winston. . 1968b. "Review of Current Trends in Linguis tics , Volume 3 ed. by Thomas A. Sebeok. The Hague: Mouton, 1966". Language 44.556-93. . 1971. "Interpretive Semantics Meets Franken stein". Foundations of Language 7.285-96. . 1973. Grammar and Meaning. Tokyo: Taishukan. . 19 78. "Conversational Implicature and the Lexicon". Syntax and Semantics, Volume 9: Pragmatics ed. by Peter Cole, 245-59. New York: Academic Press. Morgan, Jerry. 19 69. "On Arguing about Semantics". Papers in Linguistics 1.49-70. Müller, Walter. 1960. Englische Idiomatik nach Sinngruppen, Berlin: de Gruyter. Newmeyer, Frederick J. 19 72. "The Insertion of Idioms". Papers from the Eighth Regional Meeting of the Chicago Linguistic Society ed. by Paul M. Peranteau et al., 294-302. Chicago: Chicago Linguistic Society. . 1974. "The Regularity of Idiom Behavior". Lingua 34.327-42. Nöth, Winfried. 1977. Dynamik semiotischer Systeme. Stuttgart: Metzler. Norrick, Neal R. 1978a. "Expressive Illocutionary Acts". Journal of Pragmatics 2.277-91. . 1978b. Factive Adjectives and the Theory of Factivity. Tübingen: Niemeyer. . 1978c. "Transfer-Ikone: Indirekt motivierte Zeichen". Read at the Second Semiotics Colloquium in Regensburg, 19 78. (appearing in the conference pro ceedings ed. by A. Lange-Seidl, 1981.) . 1979a. "Regular Multiplicity of Meaning". anglistik & englischunterricht, Volume 8: Semantik ed. by Hans-Jürgen Diller et al., 103-16. Trier: Wissen schaftliche Verlagsgesellschaft. . 1979b. "Regular Meaning Multiplicity in Nouns". Papers from the Fifth Scandinavian Conference of Lin guistics, Frostavallen, Part 1: Section Papers ed. by Thome Pettersson, 221-32. Stockholm: Almqvist & Wiksell.
242
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Norrick, Neal R. 1979c. "Principles of Motivation". Papers from the Second Congress of the International Associa tion for Semiotic Studies (IASS), Vienna, 1979 ed. by Tasso Borbé. . 1979d. "The Lexicalization of Pragmatic Func tions". Linguistics 17.671-85. . 19 80. "Semantic Relations and Motivation in Idioms". Perspektive : textintern ed. by G. Tschauder and E. Weigand, 51-60. Tübingen: Niemeyer. . "The Role of Extra-Linguistic Knowledge in Lexical Relations". Reproduced, Kassel: University of Kassel. Nunberg, Geoffrey D. 1978. The Pragmatics of Reference. Bloomington: Indiana University Linguistics Club. Palmer, F. R. 1976. Semantics : A New Outline. Cambridge: Cambridge University Press. Peirce, Charles S. 1931-1961. Collected Papers. Cambridge, Mass.: Harvard University Press. Perlmutter, David M. 1971. Deep and Surface Structure Constraints in Syntax. New York: Holt, Rinehart & Winston. Pike, Kenneth L. 1967. Language in Relation to a Unified Description of the Structure of Human Behavior. The Hague: Mouton. Plett, Heinrich. 1971. Einführung in die rhetorische Textanalyse. Hamburg: Buske. . 1975. Textwissenschaft und Textanalyse : Semiotik, Linguistik, Rhetorik. Heidelberg: Quelle & Meyer. Postal, Paul M. 1971. Cross-Over Phenomena. New York: Holt, Rinehart & Winston. Price, J. T. 19 74. "Linguistic Competence and Metaphorical Use". Foundations of Language 11.253-56. Quang Phuc Dong. 1971. "The Applicability of Transforma tions to Idioms". Papers from the Seventh Regional Meeting of the Chicago Linguistic Society ed. by Douglas Adams et al., 200-05. Chicago: Chicago Linguistic So ciety. Quine, Willard v. 0. 1960. Word and Object. Cambridge, Mass.: M.I.T. Press. Reddy, Michael J. 1969. "A Semantic Approach to Metaphor". Papers from the Fifth Regional Meeting of the Chicago Linguistic Society ed. by Robert I. Binnick et al., 240-51. Chicago: Chicago Linguistic Society. Roos, Eckhard. 19 79. "Semantische Aspekte englischer Idioms" anglistik & e n g l i s c h u n t e r r i c h t , Volume 8: Sem antik ed. by Hans-Jürgen Diller et al., 117-30. Trier: Wissenschaftliche Verlagsgesellschaft.
BIBLIOGRAPHY
243
Ross, John Robert. 1967. Constraints on Variables in Syn tax. M.I.T. Diss.: Reproduced. . 1969. "Guess Who?". Papers from the Fifth Regional Meeting of the Chicago Linguistic Society ed. by Robert I. Binnick et al., 252-86. Chicago: Chicago Linguistic Society. Saussure, Ferdinand de. 1916. Cours de linguistique générale . Paris: Payot. Schmidt, Siegfried J. 1970. "Text und Bedeutung". Text, Bedeutung, Ästhetik ed. by S. J. Schmidt, 43-79. Munich: Bayerischer Schulbuchverlag. . 1976. Texttheorie. Munich: Fink. Schofer, Peter and Donald Rice. 1977. "Metaphor, Metonymy, and Synechdoche Revis(it)ed". Semiotica 21:1/2.121-49. Searle, John R. 1969. Speech Acts: An Essay in the Philos ophy of Language. Cambridge: Cambridge University Press. Shaw, J. Howard. 1979. Motivierte Komposita in der deutschen und englischen Gegenwartssprache. Tübingen: Narr. Stern, Gustav. 1931. Meaning and Change of Meaning. Bloomington: Indiana University Press. Trier, Jost. 1934. Der deutsche Wortschatz im Sinnbezirk des Verstandes. Heidelberg: Winter. Ulimann, Stephan. 1957. The Principles of Semantics. Oxford: Blackwell. . 1967. Semantics : An Introduction to the Science of Meaning. Oxford: Blackwell & Mott. Vennemann, Theo. 1972. "Rule Inversion". Lingua 29.209-42. Viehweger, Dieter et al. 1977. Probleme der semantischen Analyse. Berlin: Akademie-Verlag. Weinreich, Uriel. 1963. "On the Semantic Structure of Language". Universals of Language ed. by J. H. Greenberg, 142-216. Cambridge, Mass.: M.I.T. Press. . 1966. "Explorations in Semantic Theory". Current Trends in Linguistics, Volume 3 ed. by Thomas A. Sebeok, 395-477. The Hague: Mouton. . 1969. "Problems in the Analysis of Idioms". Substance and Structure of Language ed. by Jaan Puhvel, 2 3-81. Berkeley and Los Angeles: University of Cali fornia Press. Weisgerber, Leo. 1954. Vom Weltbild der deutschen Sprache. Düsseldorf: Schwann. Wells, Rulon S. 19 77. "Metonymy and Misunderstanding: An Aspect of Language Change". Current Issues in Linguis tic Theory ed. by Roger W. Cole, 195-214. Bloomington: Indiana University Press. Wescott, Roger W. 1971. "Linguistic Iconism". Language 47.416-28.
244
SEMIOTIC PRINCIPLES IN SEMANTIC THEORY
Wilson, Frank P., ed. 1970. The Oxford Dictionary of English Proverbs. London: Oxford University Press. Wunderlich, Dieter. 1971. "Pragmatik, Sprechsituation, Deixis". Zeitschrift für Literatur wissenschaft und Linguistik 1.153-90. Zgusta, Ladislav. 1971. Manual of Lexicography. The Hague: Mouton. Zolkovsky, Alexandr . 1978. "At the Intersection of Linguistics, Paremiology and Poetics: On the Literary Structure of Proverbs". Poetics 7.309-32.
INDICES
A. INDEX OF AUTHORS Abraham, W. 8, 12, 81, 131-32, 196 Alston, W. P. 115 Apresjan, Ju. D. 8, 12-13, 27, 41, 231 Aristotle 10 Barley, N. 184 Baumgärtner, . 11, 131 Bendix, E. H. 13 Bergmann, M. 8, 129 Black, M. 33, 65, 110, 139 Blake, W. 214-15, 217 Bloomfield, L. 14 Braunmüller, K. 8, 81, 131-32 Brekle, H. E. 16, 27, 175 Brooke-Rose, 5, 209-11, 213-14, 216, 219-20, 224-25 Chafe, W. L. 3, 12, 16, 27, 174-76, 186-87, 191 Chao, R. Y. 111 Chomsky, N. 9, 13-14, 71, 125, 169, 192, 231
Cohen, L. 11 Cohen, T. 12 Coleridge, S. T. 210, 212 Coseriu, E. 9, 86 Cowie, A. P. 2 7 Cutler, A. 231 Davidson, D. 11, 196 van Dijk, T. 1, 8, 11, 81 , 102, 132-33, 135, 137, 154 Droste, F. 12, 196 Dubois, J. 2 7 Eco, U. 1, 17, 22-26, 30-31, 34-35, 38-39, 42, 88 Emonds, J. 231 Fay, D. 231 Fiengo, R. 231 Fillenbaum, S. 9, 14, 127 Fillmore, . J. 8-9, 14, 125-26 Fodor, Janet 102 Fodor, Jerry 102, 122, 129 Forster, . I. 9, 14, 127
246
INDICES
Fraser, . 175, 190-92 Fromkin, V. A. 9, 127, 231 Gak, V. G. 8, 12, 105, 134-35 Garrett, M. 102 Geckeier, H. 86 Golds te irr, . 73 Gruber, J. S. 8-9, 13-14, 114, 123-24 Habermas, J. 71, 204 Halliday, M. A. . 211 Hardwick, . S. 22-23 Harman, G. 101 Hasan, R. 211 Heisenberg, W. 64, 75 Heraclitus 65 Hockett, F. 12, 174 Hooper, J. B. 105 Hopkins, G. 218-20, 224 Horrocks, R. 12, 196 Hudson, R. A. 13, 127, 233 Hume, D. 21, 2 6 Hymes, D. 71, 204 Jackendoff, R. S. 8-9, 31, 75, 125, 127-28, 180, 192 Jakobson, R. 11, 16, 23, 41, 73, 77, 225 Kandinsky, W. 54 Kastovsky, D. 102 Katz, J. J. 1, 4, 9-10, 18, 77, 101-04, 112, 122, 129-31, 136, 138-39, 141-49, 151, 174, 192, 196, 225, 228, 232 Kayser, W. 2 7
Keats, J. 212, 215, 217, 221 Koller, W. 27 Kooij, J. G. 110, 174 Koutsoudas, A. 2 33 Lakoff, G. 192 Lasnik, H. 231 Leech, G. 8, 12-13, 81, 105, 134-35 Lees, R. B. 16, 175 Lehmann, V. 12, 135 Lehrer, A. 8-9, 13-14, 86, 126-27 Leisi, E. 27 Levin, S. R. 8, 11, 54, 81, 132-33, 135-36, 142-43, 154 Lewis, D. 102 Lipka, L. 16 Ljung, M. 27 Luria, A. R. 73 Lyons, J. 13, 18, 110, 116, 128, 157-58, 167 Mackin, R. 2 7 Macnamara, J. 127, 2 31 Makkai, A. 16, 27, 175-76 Marchand, H. 16, 27, 117, 175 Margalit, A. 11 Matthews, R. J. 12 McCawley, J. D. 9, 13, 114, 124-25, 130, 192, 225, 233 Mel'chuk, I. A. 231 Morgan, J. 141
INDEX OF AUTHORS Müller, W. 27 Newmeyer, F. J. 16, 174-75, 192 Nöth, W. 72 Norrick, N. R. 15, 25, 41, 87, 105, 152, 191, 193, 205 Nunberg, G. D. 12, 115, 2 33 Palmer, F. R. 117 Peirce, . S. 22-26, 33, 40, 72, 232 Perlmutter, D. 231 Pike, . L. 39 Plett, F. 11, 27 Postal, P. M. 1, 9-10, 102, 112, 122, 129-30, 142143, 174, 231 Price, J. T. 11 Quang Phuc Dong 192 Quine, W. v. . 64, 113, 115 Rapoport, A. 9, 14, 127 Reddy, M. J. 12 Rice, D. 12, 41 Roos, E. 175, 177 Ross, J. R. 231, 233 Saussure, F. de 22, 24, 72 Schmidt, S. J. 102
247
Schofer, P. 12, 41 Searle, J. R. 205, 233 Shaw, J. H. 16, 27, 174-76 Shelley, P.B. 153, 210, 215, 218 Stern, G. 2 7, 117 Trier, J. 86 Ullmann, S. 16, 27, 117, 175-76 Vennemann, T. 105, 165-66, 196-97 Viehweger, D. 128 Weinreich, U. 1, 4, 8, 11-12, 16, 19, 112, 122, 130-34, 174, 176177, 225, 233 Weisgerber, L. 86 Welby, Lady V. 22 Wells, H. G. 27 Wescott, R. W. 16, 175 Wilson, F. P. 27 Wordsworth, W. 211, 214, 216, 218 Wunderlich, D. 102 Zgusta, L. 112 Zolkovskij, A. K. 184, 231
248
INDICES
. INDEX OF SUBJECTS depletion/depleted 13, ambiguity 110-11 111-13, 130 anomaly/anomalous 10, 12, disjunction 155-57, 161, 18, 103-04, 129-30, 228 136, 138-47, 151-53, 161, 209, 220-21, 228 distinctive feature 40, antonymy 15, 18, 105, 80, 138, 144-46, 195157-58 196, 202, 228 double (multiple) process aphasia 23, 73 ing 152-54 arbitrary 2 22-2 3, 2 5 encyclopedia 195-96, 198, case, deep 125-26 202-04, 233 causative (verb) 14, 87, extra-linguistic (extra123-24 lexical) 5, 19, 129, 140-41, 144, 166, 178, code/coded 24-25 182, 184-85, 195-207, iconic 21, 25-28, 30-40, 219, 229, 233 79, 227 euphemisms 53, 93 indexical 21, 25, 2 7fairy tale reading 130-31, 30, 40-69, 79, 227 152-53, 216-17 competence 71-75, 87, 119, 121, 126, 129, 134, feature transfer (extension) 138, 163, 165, 169, 11-12, 130-38, 142, 196-97, 202, 204, 219, 144, 146-47, 151-54, 227 178, 209-10, 212, 216, 220 complementarity 14, 18, figuration/figurative 10157-58, 161 12, 18, 43, 77, 83, compositional(ity) 174 86-87, 95, 97, 102, 183, 187-90 105, 108, 116, 129-35, 139, 151-52, 154, 177, construal rule 130, 135-36, 180, 183, 187-89, 209, 138, 147, 152-53, 216 224, 228-30 contiguity 23, 2 5-2 7, 40fuzziness 41, 80, 198 69, 77, 80, 227 contradiction/contradictory generality 111, 113 11, 131, 136, 138, 141generic see Iconic-Meta147, 150-51, 219, 228 phoric Principle 2 converseness 18, 157-58, 167 global rules (derivational dead (lexicalized) metaphor constraints) 192 12, 34, 116, 118 homography 116
INDEX OF SUBJECTS homonyny 116 homophony 2, 4, 19, 109-10, 116, 118, 127-29, 164, 168, 199 hyponymy 15, 18, 83, 105, 116-17, 126, 156, 161, 228 icon(ic)(ity) 4, 22-26, 3133, 72, 77, 227, 232 (cf. code, iconic) idiom(atic)(ity) 174-79, 182-92, 198 incompatibility 18, 157-59, 161 index(ical)(ity) 4, 23-24, 26, 72, 77, 227, 232 (cf. code, indexical) inference/inferential 30-31, 34, 36-38, 4041, 44-49, 51-52, 54, 56-57, 61-62, 64-67, 69, 76, 79, 84-85, 227 irony 86 lexicalization 12, 30, 7677, 85, 95-96, 106, 118-20, 130-31, 135-36, 155, 163, 165-66, 169170, 196, 198, 205-06 lexicon 4, 10-11, 13-15, 17-19, 40, 80, 82-83, 85-87, 102-03, 105-06, 108, 117-19, 122, 124128, 130, 136, 138, 144, 147-48, 155-56, 159, 161-71, 174-75, 177, 180, 189-90, 195, 197, 199, 202, 204-06, 228, 231-32
249
metaphor(ic) 23, 33, 77, 80-81, 86, 105, 116, 134, 177, 181-86, 188-89, 195-96, 202203, 209-14, 216-20, 224-25, 227-28, 230 (cf. Iconic-Metaphoric Principles) metonymy/metonymie 23, 41, 77, 80, 86, 116, 134, 177, 181-86, 188-89, 209, 227-28, 230 (cf. Indexical-Metonymic Principles) morphologically complex units 3-5, 16, 173193, 197-98, 228-29 (cf. idiom, stock phrase) motivation (in signs) 21-23, 25, 30-69, 7879, 101, 227 motivation (in complex units) 3, 5, 10, 1516, 18, 27, 79, 86, 107, 167, 173, 175-89, 193, 197-98, 228-29 non-compositional(ity) 174, 189 (cf. idiom) onomatopoeia/onomatopoeic 16, 175 oppositeness 15, 157, 22 8 pars pro toto 53 (cf. Indexical-Metonymic Principle 9) polycategorial(ity) 119, 124, 169 polysemy 1, 2, 4, 13-15, 19, 43, 53, 59, 106, 109-19, 121-29, 155, 164, 166-71, 199, 227228, 232
INDICES
2 50
popular etymology
117, 197
pragmatic(s) 11, 111, 198, 204-07, 229
115,
stative (predicate) 14
projection rule 18, 77, 103-04, 133, 137-38, 142, 147-50, 152, 188, 191, 193, 224 ratio
difficilis
ratio
facilis
redundancy 167
23-25 23-24 125, 159,
161,
rule 9, 15, 106, 124, 126, 138, 140141, 147-48, 180 selection(al) 4, 11-12, 104, 120-21, 129-42, 144-46, 149-53, 159161, 209-10, 216-18, 220, 222, 228 semantic (word) field 9, 14-15, 18, 40, 86, 105, 126-27, 159, 162, 169, 228 signifiant 2 2 , 24 signifie
similarity 23, 25-27, 32-33, 77, 80, 227
22, 24
sign dyadic model 22 triadic model 22
stock phrase 173-79, 181-93, 229 surface structure inter pretation 190-93 symbol (Peirce)
23, 72
synechdoche 16, 41, 53, 86, 212 (cf. Indexical-Metonymic Prin ciple 9) synonym(y) 15, 18, 40, 53, 85-86, 102, 105, 145, 156-57, 161-62, 228 theoretical triangulation 144-45 transformation, semantic 3 1 - 3 3 , 8 4 , 8 6 , 134, 227 Semantic Transformation 1 8 0 - 8 1 , 118, 176, 180, 182, 189, 2 1 2 , 2 1 4 , 2 1 6 - 1 8 , 220 S e m a n t i c Transformation 2 8 4 - 8 5 , 106, 1 5 1 - 5 2 , 165, 1 8 3 , 196, 2 0 3 , 211-16
content 22-2 5, 30-34, 42, 44-45, 77
truth-functional semantics 101-02
expression 22-25, 3034, 42, 44-45, 77
vague(ness) 114
interpretant
via-rule 4-5, 15, 17-19, 105-10, 117, 119, 121, 124, 126, 129, 155, 158-59, 163-71, 173174, 176-86, 189, 193, 195-207, 228, 232
object
22-2 3
22-25, 30
representamen
22-2 4
simplicity (metric) 7, 71, 74-75, 118-19, 121, 227
zeugma 114
13, 111,
113-
INDEX OF REGULAR RELATIONAL PRINCIPLES
251
C. INDEX OF REGULAR RELATIONAL PRINCIPLES Iconic-
Metaphori 1
2 3 4 5
(model-object) 3, 18, 28, 31-34, 59, 80-81, 84 118, 132, 137, 176, 180, 182, 186, 189, 212, 214, 216-18 (specific-generic) 28, 34-35, 54, 82, 100, 162, 163 (specific case-general rule) 18, 28, 36-37, 59, 82-83, 162, 176, 178, 184-85 (object-feature) 28, 37-39, 69, 77, 83-85, 106, 151-54, 165, 183, 196, 203, 211-16 (class member x-class member y) 28, 39-40, 85-86, 162
In dexical1 2 3 4 5 6 7 8 9 10 11 12 13
Priniciples
Metonymic
Principles
(cause-effect) 2, 3, 28, 41-45, 78-79, 86-87, 114, 177 (producer-artifact) 28, 45-46, 51, 56-57, 87-88, 107 (natural source-natural product) 28, 46-47, 88-89 (instrument-product) 28, 47-48, 50, 89-90 (object-act) 29, 48-49, 90, 120, 203 (instrument-act) 29, 50-51, 91, 119, 164, 167 (agent-act) 29, 51-52, 91-92, 107, 168, 200 (agent-instrument) 19, 29, 52, 76-77, 92, 117, 200-01 (part-whole) 29, 53-54, 77, 92-93, 132, 177, 184 (act-complex act) 29, 54-56, 93-94 (central factor-institution) 29, 56-57, 69, 94, 181 (container-content) 15, 29, 57-60, 95, 104, 114, 150, 162-63, 182 (locality-occupant) 29, 60-61, 96, 104, 149, 151, 182
252
INDICES 14 15 16 17 18
(costume-wearer) 29, 61-63, 96-97, 121, 182, 184, 189 (experience-convention) 29, 63-64, 97-98, 114, 215, 220 (manifestation-definition) 29, 64-66, 98-99, 115, 222 (possessor-possession) 29, 66-68, 99-100 (office holder-office) 29, 68-69, 100