Historical Semantics and Cognition [1999. Reprint 2013 ed.] 9783110804195, 9783110166149

Citation preview

Historical Semantics and Cognition

1749

1999

Cognitive Linguistics Research

13

Editors

René Dirven Ronald W. Langacker John R. Taylor

Mouton de Gruyter Berlin · New York

Historical Semantics and Cognition Edited by

Andreas Blank Peter Koch

W G_ DE

Mouton de Gruyter Berlin · New York 1999

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

® Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Historical semantics and cognition / edited by Andreas Blank, Peter Koch p. cm. - (Cognitive linguistics research ; 13) Includes bibliographical references and index. ISBN 3-11-016614-3 (cloth ; alk. paper) 1. Semantics, Historical - Psychological aspects. 2. Cognition. I. Blank, Andreas. II. Koch, Peter, 1951- . III. Series. P325.5.H57H48 1999 40Γ.43—dc21 99-32695 CIP

Die Deutsche Bibliothek — Cataloging-in-Publication Data Historical semantics and cognition / ed. by Andreas Blank ; Peter Koch. - Berlin ; New York : Mouton de Gruyter, 1999 (Cognitive linguistics research ; 13) ISBN 3-11-016614-3

© Copyright 1999 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printing: Werner Hildebrand, Berlin Binding: Lüderitz & Bauer, Berlin Printed in Germany

Preface

The papers collected in this volume evolved from a symposium that was held September 19-21, 1996, at the "Clubhaus" of the Freie Universität Berlin. The symposium was organized with the double intention of providing a forum in which synchronically and diachronically oriented scholars would have to exchange their ideas and where American and European cognitive linguists would be confronted with representatives of different directions in European structural semantics. While the confrontation indeed happened as planned, the expected synergetic effects were perhaps not as intensive as we had hoped. However, we are convinced that some of the discussions we had will bring long-term results, thanks to the opponents' modified perception of each other generated by this encounter. We would like to express our gratitude to the "Außenamt" of the Freie Universität Berlin for all its various forms of support, and especially to the Volkswagen-Foundation, without whose grant this symposium would not have been possible. All the work, the preparations including the program and the schedule of meetings, the duplication and distribution of hand-outs and papers, as well as the organizing of coffee-breaks, restaurants, accomodations and transfer from airports to hotels, could not have been done without a devoted team of co-workers. We take this opportunity to thank once again Mary Coppie, Geneviève Gueug, Paul Gévaudan, Richard Waltereit and especially Sigrid Kretschmann whose experience and readiness were an enormous support and contributed to the success of the symposium. Ideas of how the proceedings could best be published were discussed during the Berlin symposium itself. Due to changes in both our academic affilations, some time went by until it was decided that a greater part of the papers read at the Clubhaus should be published in a volume rounded off with two articles that fit the volume's the-

vi Preface matic framework better than the papers originally presented in Berlin. A lot of work by Cinzia Cazzaro, Mary Coppie, Angela Dom, Cristina Fossaluzza, Keith Myrick, Eberhard Matt and Alexandra Twardy went into elaborating the decisive version of the book. Finally, we would like to express our gratitude to all of the contributors for their comprehensive cooperation, to the editors of the Cognitive Linguistics Research series and to Anke Beck of Mouton de Gruyter. Marburg / Tübingen

Andreas Blank / Peter Koch

Contents

Introduction: historical semantics and cognition Andreas Blank and Peter Koch

Section I:

Theories and Models

Cognitive semantics and structural semantics JohnR. Taylor Diachronic semantics: towards a unified theory of language change? Helmut Liidtke Why do new meanings occur? A cognitive typology of the motivations for lexical semantic change Andreas Blank Diachronic prototype semantics. A digest Dirk Geeraerts Cognitive semantics and diachronic semantics: the values and evolution of classes François Rastier

Section II:

Descriptive categories

Losing control: grammaticization, subjectification, and transparency Ronald W. Langacker

viii

Contents

The rhetoric of counter-expectation in semantic change: a study in subjectification Elizabeth Closs Traugott

177

Synecdoche as a cognitive and communicative strategy Brigitte Neriich and David D. Clarke

197

Laws of thought, knowledge and lexical change Beatrice Warren

215

Section III:

Case studies

Intensifiera as targets and sources of semantic change Ekkehard König and Peter Siemund

237

Cognitive ease and lexical borrowing: the recategorization of body parts in Romance Thomas Krefeld

259

Cognitive aspects of semantic change and polysemy: the semantic space HAVE/BE Peter Koch

279

List of contributors

307

Index

309

Introduction: historical semantics and cognition Andreas Blank and Peter Koch

1.

General survey

Cognitive linguistics has had considerable influence on the development of theories and methods of description in semantics (cf. Lakoff 1987; Langacker 1987/90; Taylor [1989] 1995; Kleiber 1990; Ungerer/Schmid 1996). Nowadays, even manuals of historical linguistics refer to issues in cognitive research relevant to problems of diachrony.1 Indeed, some of the favourite subjects of cognitive semantics (metaphor, metonymy, polysemy etc.) deal precisely with the synchrony/diachrony-interface. In our opinion, investigation of diachronic problems can, in turn, sharpen our view for fundamental semantic processes and should therefore be able to advance theorizing in cognitive linguistics. In this sense, historical semantics is an ideal testing ground for semantic models and theories, as cognition and our basic human conceptual system are highly involved in lexical and grammatical change. The authors of this volume approach the synchrony/diachrony-interface from various theoretical points of views and apply or develop different conceptions of cognitive linguistics. 1.1. The first group of articles deals with fundamental theoretical issues in synchronic and especially diachronic linguistic description. John Taylor discusses the foundations and basic issues of cognitive semantics in contrast with European structural semantics, as it is paradigmatically represented by the work of Eugenio Coseriu. The central point of this controversy is the question of whether it is useful and efficient to distinguish encyclopedic semantic structures from internal, language-specific semantic structures.

2 Andreas Blank and Peter Koch

On the ground of his more general model of linguistic change (cf. Lüdtke 1980; 1986), Helmut Liidtke studies a number of cases of semantic change in lexicon and grammar, in order to demonstrate the cognitive linkage of the different levels of language on which change can occur. Andreas Blank discusses traditional classifications of the motivations for lexical semantic change and develops a comprehensive typology of these motivations on the basis of recent issues in cognitive as well as in modern diachronic linguistics. Dirk Geeraerts focuses on two major topics in his "diachronic prototype semantics" (cf. Geeraerts 1997): i) the mapping of diachronic semantic processes for several aspects of the protypical structure of categories (e.g. typicality, family resemblance, blurred edges, importance of encyclopedical knowledge), and ii) the typology of motives for lexical change based on speaker-oriented or hearer-oriented strategies aimed at increasing either communicative efficiency or expressivity. François Rastier reflects upon the epistemological status of the definition of a prototype as the "best" representative of a category especially with regard to the valorization of the prototype by the speakers. He interprets certain types of semantic change as a displacement of "evaluative thresholds" dependent upon social values and practices. 1.2. The second group of contributions develops categories for the linguistic decription of diachronic processes. By analyzing examples taken from different word classes, Ronald Langacker describes several semantic processes whose common denominator is the gradual change from physical movement to a merely virtual movement in the speaker's mind (e.g. Engl. The mailbox is across the street·, I'm going to sing). The resulting attenuation of the semantic aspect [control] in the meaning of linguistic entities is what Langacker calls "subjectification". The same term is defined in quite a different way by Elizabeth Traugott in her study of the semantic development of Engl, in fact:

Introduction 3

"subjectification" in her understanding is the rise of a new sense from pragmatic inferences in typical discourses ("pragmatic strengthening"). In contrast with the older sense, the new one focuses on the subject of a discourse because either subjective valuations are emphasized or because the new sense has acquired a pragmatic function at the speech-act level itself. Brigitte Neriich and David Clarke elaborate a number of criteria to distinguish the traditional, but usually not well defined trope "synecdoque" from "metonymy" and "metaphor". They further explore the cognitive background of synecdoque, as they have defined it, as well as its rhetoric, pragmatic and semantic potential in synchrony and diachrony. Beatrice Warren introduces a model for the contextualization of word-meanings based on semantic and encyclopedic knowledge. On the ground of this model, she develops three major types of semantic innovation called "novel hyponymic senses", "non-literal senses" and "appended senses". 1.3. In the third group, theoretical options and categories related to cognitive approaches are applied to describe selected diachronic phenomena. Ekkehard König and Peter Siemund explore the main cognitive strategies for conceptualizing and verbalizing "intensifiers" in a great number of languages as well as the semantic development of intensifiers into genuine reflexive pronouns. Analyzing the changes in conceptualization of the human body and the limbs of the body from Latin to Romance, Thomas Krefeld retraces the passage from the Latin model with "overlapping" denominations to a clear-cut torso-extremities-model in the Romance languages. The latter seems more natural from a point of view of Gestalt theory. Starting from basic conceptual distinctions in the "semantic space" HAVE/BE, Peter Koch detects typical paths of change in this area. Certain patterns of metonymy, metaphor and semantic extension seem to occur polygenetically in different languages and thus

4 Andreas Blank and Peter Koch

reveal modes of how we can conceptualize fundamental relations like POSSESSION, EXISTENCE, and LOCATION. In each of the contributions to this volume, fundamental topics of cognitive linguistics (cf. section 2) are in some way connected to recent issues in diachronic linguistics or pragmatics (cf. section 3).

2.

Cognitive models and approaches

2.1. European structural semantics has pleaded for a strict theoretical separation of encyclopedic knowledge from language-specific semantic features and has determined the latter to be the only object of linguistic semantics. In contrast to this, cognitive linguistics has strongly emphasized the importance of encyclopedic knowledge for semantics. Indeed, certain phenomena that are relevant to linguistic theory and description cannot be explained on the level of intralinguistic regularities as, e.g., the "associative anaphor" in (1), which the hearer can only interpret against the background of his world knowledge: (1)

We arrived at the village. Unfortunately, the church was closed.

The papers brought together in this volume show that it is necessary to partially or even entirely anchor diachronic studies in encyclopedic knowledge. While some authors do not discuss this problem explicitely, others claim that semantic knowledge is exclusively extralinguistic (-» Langacker and esp. -» Taylor).2 -» Geeraerts clearly gives priority to the encyclopedic knowledge, but nevertheless recognizes the relevance of intralinguistic semantic facts. —> Blank emphasizes the overall importance of encyclopedic knowledge for semantic change, but also accounts for changes induced by intralinguistic constellations. On the one hand, -» Krefeld highlights diverging segmentations of the human body in different languages

Introduction

5

(Latin vs. Romance), on the other hand, his analysis is rooted in fundamental anthropological and gestaltist categories. While observing very accurately intralinguistic semantic factors, Rastier nevertheless stresses the importance of social values for semantic change. In our opinion, linguists should not renounce completely the distinction between encyclopedic aspects of meaning and intralinguistic semantic features. It is true that intralinguistic features are not substantially different from encyclopedic information, but they have acquired a categoriali}) different status, insofar as they reflect semantic oppositions that in some languages are expressed by a simple lexeme, while other languages either have recourse only to a complex word or a paraphrase or even simply cannot realize them at all.3 Divergent semantic structures of this kind must be interpreted as emanating from cognitive constellations, because the diversity of pragmatic and social relevance and the resulting differences in the profiling of a concept determine the linguistic strategies used by the speakers of one language. Thus, distinguishing intralinguistic from encyclopedic knowledge opens a new field of research to cognitive semantics, esp. with regard to cross-linguistic (and to "cross-cultural") studies. 2.2. The verbalizing of extralinguistic entities is always related to the problem of categorization. According to the framework of cognitive linguistics, categories have a prototypical internal structure and their external hierarchical relations show a different cognitive profiling (superordinate/basic/subordinate level). It now appears that diachronic semantic processes often involve questions of categorization and of prototipicality (cf. Geeraerts 1997; Koch 1995, 1996; Blank 1997). For example, the phenomenon of semantic change can be understood as the immediate corollary of the blurred boundaries of prototypically organized categories (-» Geeraerts). From a different perspective, prototypical conceptual constellations are viewed as necessary - but not sufficient - conditions for certain types of semantic changes (-> Blank). Indeed, as demonstrated by —» Koch, metonymies and metaphors operate on a prototypical view of source

6 Andreas Blank and Peter Koch

and/or target domains. While most work in cognitive linguistics takes prototypicality for granted, Rastier raises the question of how new prototypes in language emerge. 2.3. Information relevant to meaning organizes not only in categories - be they prototypically structured or not - , but also in conceptual networks, i.e., frames, scenarios, domains etc. This is another important point for historical semantics, because semantic change can derive from altered perspectivization, profiling or highlighting of concepts or conceptual aspects inside these cognitive networks. These processes play an important role in -» Langacker's work, esp. with regard to his conception of subjedification (cf. section 2.4.) as well as in the interpretation of changes based on contiguity in -» Blank, —> Traugott, —» König/Siemund and —» Koch. In contrast to frame and scenario, the notion of "domain" is rather blurred, as it is used indifferently to describe structures based on contiguity and taxonomic relations (cf. the terminological distinctions made in Taylor 1995: 83-87). This terminological inaccurateness can even lead to explicit rejection of the term "domain" for the description of semantic change (-> Warren). Many studies in cognitive linguistics have emphasized the role of the human body as a fundamental reference point of cognition. In a diachronic perspective, this frame has a double function. First, if we take the body as a target domain, we can find examples for innovative denominations of body parts, which are of great cognitive interest, and we can even find evidence for a change of the conceptualization of the body itself (-> Krefeld). Secondly, the body also serves as a source domain for diachronic processes and for grammaticalization, e.g., the creation of intensifying adverbs out of co-referential pronouns, which themselves derive from words for body parts (-» König/Siemund). 2.4. Specific properties of prototypically organized categories and particular conceptual structures build the cognitive background of semantic change. When it comes to a concrete semantic transfer,

Introduction 7

speakers intentionally or accidently perceive or reinterpret a given concept in relation to another concept. The question is which types of associative links can relate the source to the target domain and how the resulting linguistic processes of semantic transfer can be described systematically. This observation leads us directly to figures of speech like metaphor, metonymy, synecdoque or ellipsis and some others whose history goes back to antique rhetoric. On the basis of the work done in cognitive linguistics, these tropes are now considered as notions of theoretical linguistics, instead of tools of practical rhetoric, and have to be submitted to systematic analysis and definition. The contributions of -» Nerlich/Clarke, Warren and, partially, of Lüdtke are going in this direction; —» Blank shows some typical correlations between certain types of associations and the motivations for semantic change. A highly interesting aspect in a large number of papers is the great, if not to say, outstanding relevance of conceptual contiguity ("metonymy"). It is fundamental for the studies of -> König/Siemund, and - together with metaphor and semantic extension - it also plays a central role in -> Koch. Seen from this perspective, even "subjectification" (according to both -» Langacker's and -> Traugott's understanding) can be completely reduced to the profiling of concepts against a background that is constituted by the respective frames or contexts.

3.

Recent issues in diachronic linguistics

In the last two decades, diachronic linguistics have been strongly influenced by pragmatics, a tendency that has also marked the present volume. First of all, we note that linguists have "rediscovered" the importance of the speaking subject, but the hearer's role has also been reconsidered. Thus, language as a means of self-presentation and expression of subjectivity (-> Traugott) is coming into view. Speaker-

8 Andreas Blank and Peter Koch

and hearer-oriented linguistic strategies ( - » Geeraerts) and, in a more general way, the importance of expressivity and efficiency in lexical change are emphasized ( - » Lüdtke; Geeraerts; Blank; Nerlich/Clarke). Finally, one encounters the phenomenon of the "valorization" of words and concepts ( - » Rastier) and the process of "pragmatic strengthening" (—• Traugott; cf. also König in several other publications). The greatest progress in diachronic linguistic theory during the last years has been the conception of language change as an "invisible-hand process" (cf. detailedly Keller 1994). The theory of the "invisible hand" provides us with an explanation for language change that combines a framework taken from pragmatics (e.g., the speaker- or hearer-oriented strategies as mentionned above) with cognitive regularities of linguistic innovation (see section 2.), which are corroborated by the interpretation of empirical data (—» Lüdtke; - » König/Siemund; -> Koch). In this context, it has been discovered that many diachronic processes are unidirectional and therefore normally are not reversible.

4.

Grammaticalization

During the last two decades, grammaticalization has been a major line of study in diachronic linguistics. In as much as grammaticalization constitutes both a formal and a semantic process, linguists have inevitably resorted to concepts such as "semantic bleaching", subjectification, metaphor and metonymy (cf. Heine, Claudi and Hiinnemeyer 1991; Hopper and Traugott 1993). Therefore, exploring the "grammaticalization channels" and "scales" (Lehmann 1995: 25) opens a broad field of study to cognitive linguistics. Conceptual mechanisms involved in grammaticalization are explored by - » Langacker (who uses the term "grammaticization"), -> Traugott and -> König/Siemund. Insofar as grammaticalization is typically unidirectional (cf. Lüdtke, Traugott, -> König/Siemund), it serves a good example

Introduction

9

for invisible-hand processes. Once a word or a syntagmatic construction is conventionalized as a grammatical rule there seems to be no way back to the lexicon.

5.

Two perspectives in semantic investigation: semasiology and onomasiology

Traditional synchronic and diachronic semantics distinguish between two complementary perspectives on the objects of investigation: semasiology and onomasiology. The present volume includes studies in both directions (excepted Taylor's contribution, where general problems of semantic theory are discussed). The semasiological perspective prevails or is exclusively chosen in the following papers: Traugott investigates the "development of meanings associated with a form" (p. 181) on the example of Engl, in fact. In other words: the conception of "subjectification" that is developed and illustrated in her paper is semasiological in nature. Langacker considers "an expression's meaning" as "a function of both the content it evokes and the particular construal it imposes on that content" (p. 149). Thus, "subjectification", as Langacker defines it, is also a semasiological process. Investigating the relation of the types of lexical change with contextual factors, Warren starts with the following clearly semasiological question: "in what ways can dictionary meanings be modified to yield new meanings?" (p. 224). Nerlich and Clarke focus on synecdoque and define it as an autonomous, semasiologically described trope that is clearly distinct from metonymy and metaphor. Geeraerts sees "changes in the extension of a single sense of a lexical item ... as expansion of the prototypical centre of that extension" (p. 93) and thus makes use of a semasiological conception of prototypes.4 Other articles variously combine aspects of the semasiological and of the onomasiological approach.

10 Andreas Blank and Peter Koch

With regard to a "unified theory of language change", Liidtke presents some of his examples in a way that allows a semasiological as well as an onomasiological reading; e.g., figure 2 in his contribution shows the change of meaning of Lat. manducare 'to chew' > Fr. manger 'to eat' as well as the change of the expression for the concept EAT from Lat. edere to Fr. manger. Change of meaning, which is principally a semasiological process, is approached from an onomasiological perspective in Blank's study of the speakers' motivations for inventing new expressions for concepts. Rastier combines both perspectives in analyzing, dealing with the semantic evolution of Fr. face on the one hand, and with the history of the expressions for FACE in French on the other. König/Siemund first focus on the concept of INTENSIFIERS which is thus onomasiologically defined. Then, the particular semantic development of the corresponding expressions in their sample of languages is subject to a double semasiological study, retrospectively as "targets of semantic change" and prospectively as "sources of semantic change". Insofar as it investigates Latin and Romance (changes of) expressions for parts of the HUMAN BODY, Krefeld's study is onomasiologically oriented. But by discovering changes in the segmentation of the conceptual frame itself, it is essentially dependent on semasiological insights. In Koch'S article, POSSESSION, EXISTENCE, LOCATION, ASCRIPTION, and their subdivisions constitute onomasiologically defined target concepts. A retrospective (semasiological) view leads, then, to the source concepts that serve as cognitive reference points for expressing the target concepts. The semasiological approach not only gives us access to the history of particular linguistic phenomena, but, more importantly, it also focuses our understanding of the cognitive basis and interpretation of diachronic processes. The onomasiological approach shows the continuous change in the way we express concepts and conceptual domains while at the same time sharpening our view for recurrent types of expression and for their motivations.

Introduction 11

Combining the onomasiological approach with a well-founded semasiological typology of diachronic semantic processes will enable us to understand, in a sort of "panchronic" perspective, the basic cognitive patterns of how man conceives the world. We can hope to identify the source-concepts that serve as typical reference points for verbalizing a given target-concept. We can hope to describe accurately the semantic path from source- to target-concept.5 A diachronic approach applied to a large language sample should help us to "neutralize" historical idiosyncrasies and to make fundamental cognitive patterns transparent. According to this view, "Historical Semantics and Cognition" does not constitute a gratuitous side track of cognitive linguistics, but rather proves to be a central field of activity for what we could call "anthropological linguistics" or "linguistic anthropology", exploring the limits that the specific structure of human perception imposes upon linguistic creativity. The contributions to this volume lay some fundamental groundwork towards this promising project.

Notes 1. 2. 3. 4. 5.

Cf., e.g., Trask (1996); Posner (1997); Campbell (1998); Fritz (1998); still no reference to cognitive approaches is found in Hock (1991). Consider also typical statements in Haiman (1980); Langacker (1987: 63); Croft (1993: 336). For further discussion of this topic cf. Lüdi (1985, 91-94); Koch (1998: 118-120) and Blank (in press, section 11). Onomasiological case studies are found in Geeraerts/Grondelaers/Bakema (1994: 117-153). Two research projects at the university of Tübingen are attempting to make this twofold program a reality. They are studying the lexical and semantic evolution of the words for parts of the body (and for some related conceptual domains), in the Romance languages (project DECOLAR = Dictionnaire étymologique et cognitif des langues romanes) and in a representative sample of other languages of the world (project Lexical change - polygenesis — cognitive constants as part of the interdisciplinary Research Center 441 "Linguistic Data Structures"). Their goal is to discover the typical stra-

12 Andreas Blank and Peter Koch

tegies for verbalizing these concepts, and, moreover, to establish empirical evidence for their polygenetic origin, their areal distribution, and possibly their idiosyncratic nature. We hope to leam which concepts have relatively stable expressions and which are submitted to continuous change. Cf. Blank/Koch, in press; Blank/Koch/Gévaudan, in press.

References Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. (Beihefte zur Zeitschrift für Romanische Philologie 285.) Tübingen: Niemeyer. in press Neuere Entwicklungen der lexikalischen Semantik. In: Günter Holtus, Michael Metzeltin and Christian Schmitt (eds.), Lexikon der Romanistischen Linguistik, Volume I (art. 34b). Blank, Andreas and Peter Koch in press Onomasiologie et étymologie cognitive: l'exemple de la TÊTE. In: Mário Vilela (ed.), Atas do 1 Encontro de Linguistica Cognitiva, Porto 29 e 30 de Mayo 1998. Blank, Andreas, Peter Koch and Paul Gévaudan in press Onomasiologie, sémasiologie et l'étymologie des langues romanes: esquisse d'un projet. In: Actes du XXIIe Congrès International de Linguistique et Philologie Romanes. Campbell, Lyle 1998 Historical Linguistics. An Introduction. Edinburgh: Edinburgh University Press. Croft, William 1993 The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 4: 335-370. Fritz, Gerd 1998 Historische Semantik. Stuttgart / Weimar: Metzler. Geeraerts. Dirk 1997 Diachronic Prototype Semantics. Oxford: Clarendon Press. Geeraerts, Dirk, Stefan Grondelaers and Peter Bakema 1994 The Structure of Lexical Variation. Meaning, Naming, and Context. (Cognitive Linguistics Research 5.) Berlin / New York: Mouton de Gruyter.

Introduction 13

Haiman, John 1980 Dictionaries and encyclopedias. Lingua 50: 329-357. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization. A Conceptual Framework. Chicago, London: University of Chicago Press. Hock, Hans Henrich 2 1991 Principles of Historical Linguistics. Berlin / New York: Mouton de Gruyter. Hopper, Paul J. and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Keller, Rudi 2 1994 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. Kleiber, Georges 1990 La sémantique du prototype. Paris: Presses Universitaires de France. Koch, Peter 1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46. 1996 La sémantique du prototype: sémasiologie ou onomasiologie? Zeitschrift für französische Sprache und Literatur 106: 223-240. 1998 Saussures mouton und Hjemslevs tree: zwei Schulbeispiele zwischen Semstruktur und Polysemie. In: Edeltraud Werner, Ricarda Liver, Yvonne Stork and Martina Nicklaus (eds.), et multum et multa. Festschrift für Peter Wunderli zum 60. Geburtstag, 113-136. Tübingen: Narr. Lakoff, George 1987 Women, Fire and Dangerous Things. What Categories Reveal about the Mind. Chicago etc.: University of Chicago Press. Langacker, Ronald W. 1987/91 Foundations of Cognitive Grammar. 2 Volumes. Stanford: Stanford University Press. Lehmann, Christian 1995 Thoughts on grammaticalization. (LINCOM studies in theoretical linguistics 1.) München: Lincom Europa. Lüdi, Georges 1985 Zur Zerlegbarkeit von Wortbedeutungen. In: Christoph Schwarze and Dieter Wunderlich (eds.), Handbuch der Lexikologie, 64-102. Königstein/Ts.: Athenaeum.

14 Andreas Blank and Peter Koch

Lüdtke, Helmut 1980 Sprachwandel als universales Phänomen. In: Helmut Lüdtke (ed.), Kommunikationstheoretische Grundlagen des Sprachwandels, 1-19. Berlin / New York: Mouton de Gruyter. 1986 Esquisse d'une théorie du changement langagier. La Linguistique 22: 3-46. Posner, Rebecca 1997 Linguistic Change in French. Oxford: Clarendon. Taylor, John 2 1995 Linguistic Categorization. Oxford: Clarendon. Trask, Robert L. 1996 Historical Linguistics. London: Arnold. Ungerer, Friedrich and Hans-Jörg Schmid 1996 An Introduction to Cognitive Linguistics. London / New York: Longman.

Section I Theories and Models

Cognitive Semantics and Structural Semantics John R. Taylor

This contribution is not specifically about historical semantics. My focus, rather, is some basic issues in semantic theory, especially as these arise from a confrontation of Structural Semantics and Cognitive Semantics. By "Structural Semantics" I refer to the well-established continental European tradition, represented above all by Coseriu (e.g. 1977). Structuralist approaches are also evident in Lyons (1968) and Cruse (1986). Structuralism has profoundly influenced historical linguistic studies, especially as these have pertained to the familiar European languages (e.g. Coseriu 1974). By "Cognitive Semantics" I refer to the study of semantics within the framework of "Cognitive Grammar", as developed above all by Langacker (1987, 1991). The work of Lakoff (1987), Talmy (1988), and many others, is broadly compatible with Langacker's approach. Studies of grammaticalisation, e.g. Heine (1993, 1997), can also be assimilated to the Cognitive Grammar programme. For lexical historical semantics within the Cognitive Grammar framework, especially important is the work of Geeraerts (e.g. 1985,1997). Coseriu (1990) sees a profound gulf between Structural Semantics and the Cognitive Grammar approach to semantics. In recent years, however, others (e.g. Koch 1995, 1996) have been keen to incorporate insights of Cognitive Grammar into historical linguistic studies, without, however, wanting to give up some of the basic assumptions of Structuralism. In this connection, it should be noted that the aspect of Cognitive Semantics that has been most commonly seized upon, has been the idea of categorisation by prototype. Although "cognitive semantics" appears in the title of Coseriu (1990), the only aspect of Cognitive Semantics that is dealt with in any depth in the article is categorisation by prototype. This emphasis on proto-

18 John R. Taylor

types is unfortunate, in that there is much more to Cognitive Grammar than categorisation by prototype! I have several aims in this chapter. One is to dispel what appear to be some current misunderstandings about Cognitive Grammar. I also question some of the assumptions underlying Structural Semantics, arguing that some of the postulated distinctions may be unnecessary, and, to the extent that these distinctions do have validity, they can be incorporated unproblematically into the Cognitive Grammar model. I also tentatively point to some aspects of the Cognitive Semantics approach that I believe are likely to be especially relevant to historical semantic investigations. I begin by observing that although Cognitive Grammar and Structuralism have developed independently, with little mutual interaction, the two approaches can be seen to have a common origin. Both, namely, have developed, albeit in different ways, some basic insights of Saussure.

1.

The Saussurian heritage

Not the least of the achievements of Ferdinand de Saussure was to have established Linguistics as an autonomous academic discipline. Contrary to the naive view of the matter, it is not, according to Saussure (1916: 23), the existence of a certain subject matter (in our case, language, and languages) that justifies and circumscribes a discipline, but rather a "point of view", a distinctive way of treating the subject matter. Scholars with all manner of interests have had things to say about language. Yet if it is to achieve disciplinary autonomy, Linguistics cannot simply be the grand sum of whatever literature students, anthropologists, sociologists, psychologists, etc. say about language. Paradoxically, the very centrality of language to human existence sharpens the need to define the proper object of Linguistics. As is well known, Saussure's proposal was that the distinctively "linguistic" study of language had to treat language as a semiotic system. The proper object of linguistic inquiry is therefore the "lin-

Cognitive Semantics and Structural Semantics 19

guistic sign", the symbolic association of a signifier (an "acoustic image") and a signified (a "concept"). Cognitive Grammar is strongly committed to the symbolic nature of language, and in this respect is profoundly Saussurian in spirit.1 On the Cognitive Grammar view, a language is essentially a vast inventory of "symbolic units", each of which associates a phonological representation (analogous to Saussure's "acoustic image") with a semantic representation (Saussure's "concept"). But whereas Saussure had illustrated his notion of the linguistic sign mainly on the example of lexical items (such as arbor 'tree'), Cognitive Grammar takes patterns of word formation (morphology) and phrase formation (syntax) to be also inherently symbolic in nature. Although sentences, phrases, and words may differ in their degree of internal complexity, in their status as symbolic units sentences, phrases, and complex words form a continuum with the morphemes of a language. In order for the symbolic enterprise to be feasible, it is obviously necessary to allow phonological and semantic representations of considerable internal complexity, in a manner that was probably not foreseen by Saussure, also to postulate various kinds of relations between linguistic units, i.e. between signs, between their phonological poles, and between their semantic poles. After all, if the symbolic thesis is to be taken seriously, symbolic units and their properties have to bear the full weight of what in other linguistic theories is carried by various modules of the grammar (phonology, morphology, syntax, pragmatics, etc.). Thus, a crucial notion in Cognitive Semantics is that the meaning of an expression is not exhausted by the expression's designation. Designation (or "profiling", in Langacker's terminology) always takes place against a (more or less complex) network of background knowledge. (I return to this point in section 4.) Concerning the relations between linguistic units, three kinds need to be recognised. One is the "is-a" relation. One unit instantiates (can be regarded as a more fully specified instance of) a more schematically characterised unit. A second relation is the "part-of ' relation. One unit is part of a larger, more complex unit. A third relation is the "is-like" relation. One unit resembles another unit, in

20 John R. Taylor

some respect(s), and can thus be assimilated to it, as a marginal instance to a prototype. A special case of the "is-a" relation obtains between the use of an item on a specific occasion and the item as stored in a speaker's memory.2 When the word tree is uttered with reference to a specific tree, the semantic pole of the utterance (the specific tree referred to, or, to be more precise, the speaker's conceptualisation of the tree) is an instance of the more abstractly characterised tree-concept associated in the speaker's mind with the stored lexical item. In parallel manner, the pronunciation of the word on a specific occasion is an instance of the more abstractly characterised phonological representation stored in the speaker's mind. Note that the properties of the instance may not fully match the more abstract schema. (Suppose that the word tree is applied, not to a "prototypical" tree, but to a date palm, or that the word is pronounced in a non-standard way.) In such cases, the instances may still count as instances of the symbolic unit [TREE] in virtue of the "is-like" relation; the usage would count as "marginal", but would still be attracted to the stored unit in virtue of its similarity to it. A related point, is that there is no need to make a principled distinction between "linguistic meaning" and "encyclopedic knowledge". Consequently, Cognitive Grammar does not draw a distinction in principle between "sentence meaning" and "utterance meaning". Traditionally, sentence meaning is the meaning that a sentence has in virtue of the "linguistic meaning" of its parts, whereas utterance meaning is the meaning that an utterance acquires in a particular communicative context. Both kinds of meaning properly belong in the semantic representation associated with the symbolic unit, and both need to be characterised relevant to appropriate background knowledge. Naturally enough, "utterance meaning" may need to be characterised against a much richer array of background assumptions, which appeal to specific aspects of the speech situation. But this fact is fully consistent with the view that utterance meaning stands in an "is-a" relation to sentence meaning. Sentence meaning, to the extent that the notion is valid at all, is schematic for the range Λ

Cognitive Semantics and Structural Semantics 21

of utterance meanings that an expression may have on specific occasions of its use. I have emphasised the Saussurian roots of Cognitive Grammar, in order to better contextualise a comparison with Structural Semantics. Structural Semantics has also drawn its inspiration from Saussure, albeit with an emphasis on other aspects of Saussure's thought. Saussure, as we all know, asserted that the link between the signifier and the signified is arbitrary (Saussure 1916: 100-102). There are, to be sure, slightly different (though not incompatible) ways of understanding "the arbitrariness of the sign". In the first place, the sign is arbitrary in the sense that there is no inherent association between sound and meaning. On this view, arbitrariness contrasts with motivation. A sign may be motivated, to the extent that the language user can perceive some reason why the signifier should have the meaning that it does (and vice versa). Internally complex signs are usually motivated, to varying degrees. On a slightly different understanding of arbitrariness (and it is this understanding that is especially emphasised in Cognitive Grammar), the linguistic sign is arbitrary in the sense that it is the product of conventionalisation. Speakers act in the belief that the signifier-signified relation (at least for established units, cf. footnote 3) is shared by other members of a speech community. And, as Saussure (1916: 104) pointed out, a speaker is unable to single-handedly modify the established and shared conventions. Saussure (1916: 155-157) goes further, and maintains that it is not only the signifier-signified relation that is arbitrary, the signifiers and the signifieds in any given language are themselves arbitrary, in the sense that there is no intrinsic reason why just these meanings should receive symbolic expression, nor why just these phonological forms should serve as signifiers. Saussure emphasised that the signs that make up a language do not constitute a nomenclature, i.e. they are not labels for an independently given list of concepts. It is the language itself that structures cognition, thereby creating the concepts through the very process of symbolising them. Likewise, there is nothing intrinsic to a sound that renders it suitable to function as a linguistic signifier. Sounds have the status of speech sounds only in

22 John R. Taylor

virtue of the structuring of sound by the semiotic system that is a language. If pursued, Saussure's views on the arbitrariness of concepts and sound patterns must inevitably lead to a position of radical relativism, of a form that perhaps not even Whorf would have wanted to endorse!4 Research on semantic and phonological universale has, of course, revealed rather severe constraints on the concepts and sound patterns that may come together in symbolic association, while still allowing, within the boundaries set by these constraints, considerable cross-language diversity.5 But perhaps the most challenging component of Saussure's thought, and one that has fired the imagination of generations of his readers, lies in his thesis that units on the phonological and semantic levels have a status within the language only by virtue of the relations which they contract with other units on the same level (phonological or semantic). Saussure (1916: 158160) introduced the term "value" to refer to this aspect of the linguistic sign. A "concept" receives its "value", not in virtue of any intrinsic semantic content, but in virtue of the relations (syntagmatic and paradigmatic) which it contracts with other symbolised concepts. Likewise, sound units have a value in the language system in virtue of the sound units they contrast with, and combine with. Paradoxically, the value of a linguistic unit is determined, not by its intrinsic content, but by what it is not, by "ce qui existe en dehors de lui" [what exists outside of itself] (1916: 160). In proposing the notion of "value", Saussure is not denying the obvious fact that signs do have a positive content, alongside their contrastively defined value. Structuralist Semantics captures this distinction by the terms "signification" and "designation" (or, in German, "Bedeutung" and "Bezeichnung").6 The "signification" of a sign is the concept understood contrastively i.e. in terms of its relations to neighbouring concepts; the "designation" is the concept understood in its positive aspects, i.e. in terms of its potential to refer to actual states of affairs in the world. To give a simple example: the morphophonemic category [SINGULAR] has the same designation, both in a language which has a simple two-way contrast between singular and plural, and in a language which has a three-way contrast

Cognitive Semantics and Structural Semantics 23

between singular, dual, and plural. Yet the signification ("value") of [SINGULAR] is different in the two cases. In the one, [SINGULAR] enters into a simple binary contrast with [PLURAL], in the other, it enters into a ternary contrast with [DUAL] and [PLURAL]. We can make a further distinction, between "signification" and "designation" on the one hand, and "reference" on the other. Here, we need to return once again to Saussure. Saussure (1916: 98) made it very clear that for him, the linguistic sign was a mental entity. The linguistic sign [TREE] did not associate a particular tree growing in the yard with a specific utterance [tri:]. The semantic content of the sign was a "concept" in the mind of the language user. Saussure likewise insisted that the acoustic image was a mental representation, distinct from any physical manifestation, and thus neutral with respect to pronunciation and perception. But in any particular act of speech, there is, obviously, a specific articulatory/acoustic/perceptual event, probably (though not necessarily) associated with reference to a particular entity in the real world. Thus, it is easily possible for two signs to be used with the same reference, but having different designations (and eo ipso, different significations). To extend on the earlier example: The category [PLURAL], in the two kinds of languages mentioned above, has the same reference when used of a group of three entities. But the designation (and signification) of [PLURAL] is different in the language which only has the two-way contrast between singular and plural, from the designation that it has in a language which knows a three-way contrast between singular, dual, and plural. Saussure represented the sign as a simple bipartite entity (Fig. 1). More elaborate schémas have been proposed; the "semiotic pentagon" in Fig. 2 has been adapted from Koch (1996), who attributes it to Raible (1983). As mentioned, the sign for Saussure was a purely mental entity. In order to link the two elements of the sign to entities outside the mind (i.e. to an acoustic-phonetic event on the one hand, and, on the other, to a referent in the world), the bipartite sign is extended to include a "name" and a "referent". A fifth element is introduced, in order to capture the distinction between signification and designation.

24 John R. Taylor

Figure 2. The "semiotic pentagon" (after Raible 1983 / Koch 1996)

I have already hinted at what might be the Cognitive Grammar approach to the issues touched on above. The distinction between an acoustic-phonetic event and a signifier (Saussure's "acoustic image") is an "is-a" relation, i.e. the relation between a fully specified

Cognitive Semantics and Structural Semantics 25

instance and a more abstractly characterised schema. The same goes for the relation between a referent and the designation/signification of a linguistic unit. A further point is that the distinction between the extra-linguistic, extra-conceptual aspects of the sign as used in a specific communicative context and its linguistic-conceptual aspects (the distinction is represented by the broken line in Fig. 2), also falls away. It is an error to suppose that people use language in order to refer directly to "things in the world", i.e. to things outside of the mind. Language is used to refer to mental "projections" of the world (Jackendoff 1983: 29), or, to put it more generally, to elements in what Fauconnier (1985) has called "mental spaces". A mental space may purport to be a model of the world as it is. But equally, the model may be of a world that is imagined, dreamt, represented in a picture, novel, film, and so on. There is no linguistic difference between a fictional narrative and a narrative which purports to portray events that "really" happened. Concerning the phonological pole of the sign, it is also an error to suppose that acoustic-phonetic aspects of an utterance are any less "cognitive" than a phonological (or semantic) representation. Sounds, as categorised by a speaker/hearer, are also conceptual entities. What all this means, is that, from a Cognitive Grammar point of view, the bipartite structure of the linguistic sign, as depicted in Fig. 1, is perfectly adequate as it stands, it is in no need of further elaboration along the lines of Fig. 2. Given the austerity of Fig. 1, the challenge of Cognitive Grammar is to describe languages in all their complexity and variety - including aspects of their use and their variation over time - in terms of the essentially Saussurian notion of the bipartite "signe linguistique".

2.

Signification vs. Designation, or: Where are prototypes?

A major point of disagreement between Cognitive Semantics and Structural Semantics concerns the special status accorded in the latter to "signification" in contrast to "designation" and "reference". What is at issue here, essentially, is whether it is justified to postu-

26 JohnR. Taylor

late a level of purely "linguistic" meaning, in contradistinction to a level of encyclopedic knowledge, between "una semántica lingüística" [a linguistic semantics] and "una semántica de las cosas" [a semantics of things] (Coseriu 1990: 281).7 For Coseriu (1990: 267), Cognitive Semantics commits "el error más grave y más elemental que pueda cometerse en semántica" [the most serious and most error that it is possible to commit in semantics]; this is the capital error of confusing linguistically structured meaning with experientially derived knowledge about the states of affairs that linguistic expressions refer to. The confusion manifests itself, according to Coseriu, in the treatment of prototype effects. On the Structuralist view, prototype effects lie outside the language system proper; they have to do with difficulties a person may encounter in properly applying a word to a state of affairs, i.e. they are matters of designation and reference, not of signification. That it might be difficult to determine, at a given time and place, whether it is "night", or "day", in no way entails that the significations of the words night and day are "fuzzy" or indeterminate, or structured around a prototype. On the contrary, Coseriu argues, prototype effects arise precisely because the linguistic meanings of night and day are absolutely clear-cut; were this not the case, we could have no confidence in asserting that a certain state of affairs constitutes a good example, or a less good example, of the application of the word (Coseriu 1990: 258). Likewise, penguins and ostriches can only be recognised as "marginal" examples of the bird category if the category is already clearly defined. In order for a bird to be a "less good" example of the category, it must already have been categorised as a bird (1990: 279). And the clear-cut concept of what a bird is, is a matter of signification, not of designation, or of associating exemplars with a prototype. Whether birds, as entities in the world, constitute a clear-cut category or not, is a matter of biology, not of linguistics. Coseriu (1990: 268) observes that Cognitive Semanticists have selectively focussed on just those words (such as to lie, and names of natural kinds) whose real-world applications tend to give rise to prototype effects, whilst ignoring linguistic-semantic contrasts which

Cognitive Semantics and Structural Semantics 27

are clear-cut, not only on the level of signification, but also on the levels of designation and reference. He mentions the example of motion verbs. Spanish venir and ir contrast with respect to 'motion to the place of the 1st person' vs. 'motion to the place of the 2nd/3rd person'. In Italian and Catalan, the contrast between venire!andare, venir/anar, is drawn differently, between 'motion to the place of the lst/2nd person' vs. 'motion to the place of the 3rd person'. (Hence, Catalans, speaking Spanish over the telephone, will tend to make the error of saying * Mañana vengo a verte 'Tomorrow I come to see you', instead of the correct Mañana voy a verte 'Tomorrow I go to see you'.) Moreover, the notion of "structure", and of a "structured lexicon", suggests that a semantic contrast might serve to differentiate more than one word pair. And indeed, the above mentioned contrast in Spanish shows up with verbs of carrying: traer vs. llevar. (In Italian, though, the contrast is not made: portare serves for both senses.) It is difficult to imagine, Coseriu remarks, what "prototypes" could be associated with these clear-cut meanings, and what deviations therefrom could look like. There is a further point. This is that the distinctions in question are language-specific, and therefore cannot plausibly arise from any natural categorisation of non-linguistic reality. By focussing on the referential possibilities of lexical items, and on the naming of real-world (and therefore universally accessible) categories, Cognitive Semantics has ignored the structured, language-specific relations that exist between significations. In brief, Cognitive Semantics falls into the trap that Saussure warned us about, of viewing a language's lexicon as a nomenclature, a list of names for pre-existing categories. With respect to its allegedly onomasiological orientation, Coseriu brings in a third player, in addition to Structural Semantics and Cognitive Semantics, namely the theory of word definition by necessary and sufficient conditions. Although Fillmore (1975) presented prototype semantics in opposition to "check-list" theories of meaning, Coseriu groups both together as examples of onomasiologically oriented approaches, and both stand in contrast to Structuralism, which looks in the first place at relations of contrast between linguistic units, not at states of affairs in the world. Thus, for Coseriu (1990:

28 John R. Taylor

245), the Katz and Fodor (1963) analysis of bachelor, which defines the word as a conjunction of the features [HUMAN], [ADULT], [MALE], [NEVER MARRIED], suffers from the same fault as prototype theories, in that it defines the word in terms of the conjunction of (real-world) features of its potential referents (i.e. in terms of the word's designation), rather than in terms of the word's linguistic value. In evaluating Coseriu's critique, let us first consider the content of the proposed contrasts at the level of significations. Let us accept that day and night stand in a simple two-way contrast. The contrast has to do, presumably, with the presence vs. absence of sunlight (assuming an open-air environment). Note that the contrast appeals intrinsically to a real-world phenomenon, one that can only be apprehended empirically, through experience of the world. Coseriu, generally, is quite happy to give natural language glosses (in French, German, Spanish, or whatever) to the content of distinctive semantic features. Now, Jackendoff (1990: 33) has remarked that one cannot create a semantic feature simply by taking any old expression and putting a pair of square brackets around it. Behind Jackendoff s quip is the idea that if linguistic meanings are to be distinct from encyclopedic knowledge, the features that go into the linguistic definitions must be ontologically distinct from attributes of the real world. For if there is no such distinction between linguistic-semantic features, and attributes of extra-linguistic reality, the methodological basis of the distinction becomes vacuous. And indeed, a common strategy of many two-level theorists (see footnote 7) is to propose that semantic features have the special status of semantic primitives, presumably innate to human cognition, and that are independent of experience. Jackendoff, for example, postulates a set of "conceptual constituents", of the kind [THING], [PLACE], [GO], [STAY], [MOVE], [CAUSE], etc., which are combined in accordance with "conceptual well-formedness rules". These generate the general architecture of all possible concepts, whose substance is filled in by information derived from acquaintance with the world. Such an approach will tend to emphasise the universality of semantic structures, at least at a certain level of abstraction. Coseriu, on the other hand, makes no pretence that distinctive semantic fea-

Cognitive Semantics and Structural Semantics 29

tures ("semes") might be, or might be built up out of, universal semantic primitives.8 Distinctive semantic features have to be determined case by case, according to the structural relations obtaining in a given language, and are as simple or complex as the data requires. Furthermore, significations are not "built up" out of features; it is the features that emerge from the contrasts, not vice versa (Coseriu 1977: 17). Coseriu (1990: 261) cites with approval Pottier's (1964) well-known analysis of seating objects in French, which lists such real-world features as "avec pieds" [with feet], "avec bras" [with arms], "avec dossier" [with back]. Note here that the very notions of a "(chair)-leg", "(chair)-arm", and "(chair)-back" already presuppose (encyclopedic) familiarity with the domain of furniture, and with the conventional practice of naming parts of furniture metaphorically in terms of animal (or human) body parts.9 It would indeed be "patently ridiculous" (Jackendoff 1990: 33) to propose "avec dossier" as a universal semantic feature. But it is also difficult to imagine what the "linguistic" meanings of chaise, fauteuil, tabouret, etc., could be, if not knowledge of what these kinds of objects actually are, and how they are to be differentiated one from the other.10 Since, for Structuralist Semantics, the distinction between the linguistic and encyclopedic levels does not reside in the content of the distinctive features, we need to ask whether there are other characteristics of significations, which render this level of description ontologically distinct from designation and reference. Two aspects appear to be relevant for Coseriu. The first I have already mentioned. This is that significations (within a given semantic field) are clearly contrastive, and betray no "fuzziness" or prototype effects. A corollary of contrastiveness, which I shall address in the next section, is that significations are taken to be unitary entities, i.e. betray no polysemy. The second aspect is the possibility of neutralisation (Coseriu 1977: 17-18). The notion is familiar from phonology. In certain environments, the contrast between two otherwise contrastive phonemes is suspended. A well-known example concerns the neutralisation of the voicing contrast in word-final obstruents in German and Russian. Coseriu (1990: 260) views neutralisation as a specifically linguistic (not a conceptual) phenomenon; consequently, the possibi-

30 John R. Taylor

lity of neutralisation can serve as a diagnostic, as it were, of a contrast at the level of significations. Thus, the contrast between day and night can be neutralised, as when day is used, not in opposition to night, but to cover the 24 hour period comprising both day and night. (It does not follow, therefore, that day is polysemous between two meanings.) The possibility of neutralisation must be determined on a case-by-case basis. Whereas Spanish allows neutralisation of the gender contrast between hermano and hermana (in that hermanos can mean 'brothers and/or sisters'), no comparable neutralisation is posssible between brother and sister in English, and this in spite of the fact that English does allow (or at least, used to!) gender neutralisation between he and she, man and woman (as when he and man are used as gender neutral items). Let us take the second point first. I suspect that semantic neutralisation, as described by Coseriu, is in fact a multifaceted phenomenon, and which therefore cannot be explained in terms of a single mechanism. In many cases, established polysemy cannot be ruled out. The fact that Spanish hermanos can have a meaning which is not simply the plural of hermano, indeed suggests this. On the other hand, the use in English of he as a gender-neutral pronoun (as feminist critics never tire of reminding us!), arguably does represent a conceptual bias, which views "male" as the default value for human beings (females simply do not count); it is therefore not just a "structural" fact about the language system. Concerning the day and night example, this plausibly represents an instance of metonymy; the 24 hour period is designated by its (for most people) most salient component. (Hoteliers calculate the duration of a guest's stay in terms of so many nights.) The other aspect of significations that Coseriu emphasises, is their clearly contrastive character. The first point to make here, is that there is absolutely nothing in the Cognitive Grammar framework that precludes the proper characterisation of the clear-cut contrast between e.g. the motion verbs venir/ir in Spanish, or venire/andare in Italian. (What a Cognitive Semanticist would be inclined to look at, though, would be extended uses of these verbs, uses which do not literally denote motion to the place of a person, but which can never-

Cognitive Semantics and Structural Semantics 31

theless be conceptually related to the "basic" motion sense.) Furthermore, there is no inherent conflict between prototype categorisation and semantic contrast. On the contrary, Rosch (1978) argued that "basic level" categories achieve salience largely because their prototypes maximise the distinctiveness of the categories (cf. Taylor 1995: 50-51). Neither is it fair to charge Cognitive Semantics with undue concern with real-world, and hence "universal" (Coseriu 1990: 252) categories, for which a language merely supplies a list of names. From its very inception, Cognitive Grammar has emphasised the role of "construal" in semantics; linguistic expressions do not refer "directly" to states of affairs in the world, but to speakers' conceptualisations of these states of affairs (Langacker 1987: ch. 3). Furthermore, it is fully accepted that different languages may make available to their speakers different sets of "conventionalised" modes of construal. Secondly, it is not always the case that words contrast so clearly as in the examples that structuralists like to quote. This is most obvious in the case of (near) synonyms. Cruse (1986: 266) characterises (near) synonyms as items which have "a low degree of implicit contrastiveness". Thus, in declaring that a building is "high", one is not implicitly denying that it is "tall" (and vice versa). Although high and tall do not share exactly the same meaning, the difference can hardly be stated in terms of the presence vs. absence of some distinctive semantic feature.11 Cruse (1986: 285) also drew attention to what he called "plesionyms" - sets of words that are only weakly contrastive, and which stand mid-way, so to speak, between (near) synonyms on the one hand, and fully contrastive word sets on the other. Take Cruse's examples fog, mist, haze. Whereas other words for meteorological phenomena, such as rain, snow, hail, arguably do form a clearly contrastive set, this is certainly not the case with fog, mist, haze. Precisely because the words are only weakly contrastive, the boundaries of their meanings are not clearly defined - either conceptually, or referentially. Even so, I still have a fairly clear conception of what a prototypical fog etc. is like. Consequently, if I attempt to apply one of these words to a specific state of affairs, I can do no other than appeal to a conception of a prototypical fog, etc., and as-

32 John R. Taylor

sess how well the actual situation conforms to the prototype(s), and on this basis, decide which of the three words might be most appropriate. But if this is the case, there is no reason to suppose that a similar process does not apply when I use the words snow and hail, day and night. The only difference is, that in the latter case, the prototypes are clearly distinct, and characterisable in terms of the presence vs. absence of some easily statable attribute, whereas the prototypes o f f o g , mist, haze are not.

3.

The question of polysemy

For Structural Semantics, as for other "two-level" approaches, it is axiomatic that the linguistic meanings ("significations") of words are unitary entities, i.e. that at the level of significations, polysemy does not exist. It is not denied that a word may be used in a variety of ι 'y senses. But these senses only arise when uniquely specified values get filled out with semantic content, either as a matter of conventional usage, or in a specific discourse context. In this connection, we may refer to Coseriu's well-known distinction between "system", "norm", and "discourse". The "system" is specified at the level of the language-determined significations, the "norm" comprises established elaborations of significations, while "discourse" pertains to specific readings that emerge within a text. (The phonological analogy should be obvious. Phonemic contrasts pertain to the "system", while the "norm" comprises established allophonic realisations.) Coseriu's position is that Cognitive Semanticists are inclined to find polysemy everywhere, because of their fixation on "norm" (and even "discourse"), and their neglect of the "system". As with prototypes, there is unfortunately some misunderstanding of what the Cognitive Grammar position is with regard to polysemy. It is certainly true that Langacker (1988: 50) has asserted that the normal, expected state of affairs in lexical semantics is that a word (especially a word in frequent use) will be polysemous, i.e. will have a range of established senses. Thus, the semantic pole of a symbolic unit may need to be represented as a network of units, linked by rela-

Cognitive Semantics and Structural Semantics 33

tions of schematicity and resemblance to a prototype. Although the matter has not been as extensively studied in Cognitive Grammar, the phonological pole, to the extent that a unit may receive a variety of pronunciations, may also need to be represented as a network of 11 possibilities. A number of "case studies" of individual lexical items pursued within the Cognitive Grammar tradition, have portrayed these as sometimes highly polysemous, and this fact may well have contributed to the perception that Cognitive Grammar encourages the proliferation of polysemy.14 On the other hand, whether or not an item is to be regarded as polysemous is an empirical question, to be determined case by case. There is certainly nothing in the Cognitive Grammar framework which excludes the possibility that a linguistic unit may have a constant, invariant value. And some analyses have indeed emphasised the unitary value of some linguistic signs. This is especially the case with respect to grammatical categories. Thus, Langacker (1987) argued that the lexical categories [NOUN] and [VERB] can be associated with a single, highly abstract (schematic) value.15 With regard to many lexical items, however, polysemy is surely a brute fact, which simply cannot be argued away.16 Consider Fillmore's (1982) well-known analysis of climb. Fillmore, it will be recalled, postulated a prototypical sense, which involves the features "clambering (with the limbs)" and "ascending". Both are present in climb a tree. But in climb down a tree, the feature "ascending" is defeated. Coseriu (1990: 256-257), addressing Fillmore's analysis, observes that the very possibility of "climbing down" a tree demonstrates that Fillmore's analysis was incorrect; the proper characterisation should be, not "ascend", but "(move) in a vertical or inclined plane" (sobre un plano vertical o inclinado). Concerning the feature "clambering", given that monkeys, snails, and even plants can climb, the proper characterisation should be "keeping hold with the extremities" (agarrándose con las extremidades). The fact that, in the absence of specifications to the contrary, "climbing" is taken to be in an upward direction, is a default interpretation, associated with the "norm", not with the "system".17

34 John R. Taylor

Unfortunately, this proposal fails to cover some further uses of climb (which Coseriu does not address). The plane climbed to 30,000 feet is fine (even though a plane has no extremities with which to hold itself in place). But we can not say that the plane climbed down to 20,000 feet. With reference to an airplane, upward motion is paramount, contrary to the conclusion drawn with respect to "climbing down a tree". As I see it, there is simply no way in which these various senses can be brought under a single semantic formula. The only feature that all the uses of climb have in common, is probably "move". But at this level of abstraction, it would not be possible to differentiate climb from other verbs of motion in English (including wove).18 Neither is it plausible to claim that climb is homonymous. The various readings overlap, and are therefore not independent of each other. As mentioned, Coseriu is inclined to locate the specific readings of a lexical item on the level of "norm", while general meanings belong on the level of "system". It is not disputed that to be proficient in a language, a speaker needs to be familiar with the norms prevailing in that language (Coseriu 1990: 281). But if this is true - which it surely is! - the question arises, whether a person could be proficient in a language, knowing only the "norm", but remaining ignorant of the "system". Suppose a person has learned to use the verb climb (or any other word, for that matter) in its full range of established readings. Would not this fact, of itself, guarantee the speaker's full mastery of the word? Values and contrasts at the level of signification need play no role whatsoever in a speaker's performance. In Structural Semantics, however, the unity of meanings at the level of signification is a logical necessity, rather than an empirical matter. Coseriu (1977: 8-10) writes that meaning variants can be derived from meaning invariants (significations), but not vice versa; it is only on the basis of unitary meanings that meaning variants can be established at all (Coseriu 1990: 270). The very fact that different readings are recognised as such, rests on the prior knowledge of the invariant meaning. Furthermore, it is the unitary meaning that sets a limit on the extent of meaning variation; a word cannot end up meaning "n'importe quoi" [anything at all] (Coseriu 1977: 10).

Cognitive Semantics and Structural Semantics 35

These are curious arguments. We can surely recognise that climb has a range of different readings, and we can state them quite precisely, and point to their similarities (as Fillmore did), irrespective of whether there is (or whether we recognise that there is) a unitary meaning. However, to the extent that a speaker is creatively extending the usage range of a word, it may well be true that the speaker does need to recognise some commonality between an accepted usage A and the new usage situation B.19 (Still, the pertinent meaning invariant common to A and Β need not coincide with the invariant which justifies the extension of the word from A to another situation C, which is tantamount to claiming that there will be no invariant that unifies all three readings.) In Cognitive Grammar terms, these commonalities would be captured by means of low-level schémas that cover the relevant cases. But with respect to a range of already established (and conventionalised) uses, nothing excludes the possibility that these uses are simply learned, on the basis of input data. Indeed, without some such assumption, it would be difficult to explain how different readings of a word can drift so far apart over time. A speaker of modern English probably no longer perceives any relationship at all between type 'kind' and type 'printer's character' (cf. Geeraerts 1985), or between buff 'dull pale-brown' and buff 'amateur enthusiast'. Not only is the structuralist level of signification not strictly necessary in order to guarantee a person's adequate use of a language, it is difficult to imagine how significations, as understood by Coseriu, could be learned in the first place. Recall that significations do not emerge from usage events: "le relevé des procédés employés dans la production des phrases ne pourrait jamais amener à la délimitation du signifié" [listing the procedures employed in the production of sentences could never lead to the delimitation of significations] (1977: 12). A little further down, we read that from "des acceptions ou des variantes isolées", "on ne peut pas, en principe, déduire d'une façon immédiate le signifié" [from isolated readings or variants, it is not possible, in principle to directly deduce the signification]. Significations, in fact, appear to inhabit an idealist world, distinct from the world in which and of which language is used: "el mundo de los

36 John R. Taylor

significados es un mundo ordenado; no es el mundo caótico y continuo de las 'cosas'" [the word of significations is well-ordered; it is not the chaotic and continuous world of 'things'] (Coseriu 1990: 277). And even if we do succeed to bring some structure into the chaotic world of things, there is no assurance that the categories thus derived will match up with the categories provided by language, for "las clases de 'cosas' no coinciden con las categorías mentales" [the classes of 'things' do not coincide with mental categories] (1990: 262); Coseriu (1977: 12) doubts whether linguistic structures can be based at all on the "structures des contenus d'une pensée prélinguistique" [structures and contents of prelinguistics thought]. In other theories that postulate a special level of linguistic semantics, such as Jackendoff s, the problem of acquisition does not arise; if linguistic-conceptual categories (or at least, their basic building blocks and skeletal structure) are innate and universal, they do not have to be learned on the basis of experience. Coseriu, however, emphatically rejects the idea of the universal, or even the non-linguistic basis of linguistic-semantic structuring. He speaks merely of a person coming to recognise the "unidad intuitiva" (1990: 278) of a mental category, while the linguist's task is to "reveal" (revelar), to OCi "make manifest" (poner de manifesto) the intuitive unity. While it might make sense to suppose that a person does have (or may come to have) an intuition about the unity of, say, the bird-category, this probably has as much to do with beliefs about natural kinds as with the supposedly linguistic meaning of bird. But with respect to vast areas of basic vocabulary, it is surely a nonsense to claim that speakers become intuitively aware of the linguistic-semantic unity of the items in question, or even to suppose that they need to do so. Different uses of e.g. climb certainly stand in a family resemblance to each other, and speakers of English can readily generate mental images of a person "climbing (up) a tree", "climbing (down) a mountain", or a plane "climbing into the sky". But the only common denominator to these states of affairs is the fact that they are designated by the same phonological form, not that they elaborate a unique semantic content!

Cognitive Semantics and Structural Semantics 37

Any linguistic theory has to be evaluated, not only in terms of its theoretical postulates and inherent plausibility, but also in terms of the research results which it generates. In spite of the reservations expressed above, it might be objected that Structuralist Semantics has indeed proved an invaluable tool in explicating semantic change. The distinctive contribution of Structuralism is to have pointed to changes in the "system"; for example, contrasts at the level of signification are abandoned or created, a distinctive feature is lost or comes into being. If we take a long-term perspective, we can indeed note a drastic change in the "value" of Latin passer 'sparrow' as it evolves into Spanish pájaro '(small) bird'. Such changes are as clear-cut as the significations themselves are claimed to be. But precisely for this reason, focus on structural relations alone can say little about the mechanics of language change. For this, we need to focus again on the "norm" and on "discourse", i.e. on speakers' conceptualisations and categorisations. Coseriu (1990: 260) certainly allows the possibility of "categorización de emergencia" - the one-off application of a word to a novel situation. With increasing frequency, this designation can enter the "norm", and can even effect a change in the signification. But not, Coseriu insists, by adding a new nuance to the signification, or, even less, by introducing an element of polysemy. Rather, the change will effect "todo el significado" (original emphasis). But at what point in historical development does the change in signification occur, and on what basis can one state with confidence that the change has occurred? Coseriu (1990: 260) suggests that some residual problems with his analysis of English climb, e.g. the fact that the word can be used of a snail (which lacks "extremities"), might be "exceptional", or even metaphorical, and thus betray a designation "de emergencia". But given this loophole, the theory of invariant significations becomes vacuous.21

38 John R. Taylor

4.

Concepts

Although Saussure used the word concept to designate the semantic pole of the linguistic sign, many semanticists have been reluctant to appeal to concepts at all. Concepts, by definition, are private, mental entities; a person can have no access to another person's concepts except, of course, through the medium of language. But if language is defined as a means for symbolising concepts, there is no methodology for independently establishing the nature of another person's concepts. Appeal to concepts, therefore, could be circular (cf. Lyons 1977: 113). Interestingly, Lyons (1968: 443) favoured a structuralist approach to semantics precisely because it frees the linguist from the need to refer to "concepts". The meaning of a word becomes nothing other than the set of relations that the word contracts with other lexical items.22 I do not think that "concepts" need be such mysterious entities as Lyons and others make out (cf Taylor, in press a). A common view amongst psychologists is that a concept is a principle of categorisation (Komatsu 1992). To "have" a concept, is to have the means to categorise entities as examples of that concept. Put crudely, to have the concept TREE, is to have the ability to recognise a tree when one sees one. Understood as schémas for categorisation, concepts are by no means restricted to nominal entities. One of Langacker's major achievements is to have proposed a theoretical apparatus for the elucidation of the conceptual structure, not only of various relational units, such as verbs, prepositions, and adjectives, but also of "functional" morphemes such as the articles and case categories. What goes into a concept? Coseriu (1990: 261) - rightly - criticises the view that concepts might be "imágenes de las clases" mental representations ("pictures") of categories. A crucial notion of Cognitive Grammar (Langacker 1987: 183-185) is that the meaning of an expression involves the "profiling" (or designation) of an entity, against background assumptions. (These latter are referred to variously as domains, frames, idealised cognitive models, etc.) The (by now) classic example is the word hypotenuse (Fillmore 1985). The word designates a straight line, no more, no less. A hypotenuse,

Cognitive Semantics and Structural Semantics 39

however, is categorised (recognised) as such, in virtue of the fact that the straight line functions as part of a non-designated (non-profiled) entity, namely, a right-angled triangle. It is not the straight line as such, nor the right-angled triangle, that constitutes the concept HYPOTENUSE, but the profiling of the straight line against the notion of a triangle. The example of hypotenuse is relatively simple, in that the concept presupposes a fixed, and easily circumscribed domain of knowledge. Most concepts need to be characterised against multiple domains, of varying centrality, which may be selectively activated on particular occasions of their use. (Thus, read a book, print a book, drop a book, etc., construe BOOK slightly differently in each case, and highlight different background domains, even though the profiled entity arguably remains the same.) Concepts, therefore, turn out not to be fixed entities, but rather "emerge" in the act of conceptualisation. By the same token, complex expressions are rarely fully compositional, in the sense that their meaning can be computed from the fixed meanings of their component parts. Combining concepts is not just a matter of combining the profiles, it also involves the integration of background knowledge. Coseriu accuses Cognitive Linguists of exaggerated concern with "objective" categories; there are classes of things out there in the world, which the words of a language pick out. This view, I think, seriously misunderstands the Cognitive Grammar programme. Objectively speaking, I dare say there is much in common between writing (with a pen) on a piece of paper and drawing (with a pencil) on a piece of paper. Both involve a person holding a slim instrument and making marks on a surface. Why do we not categorise the two kinds of events in the same way? The Structuralist view would be that it is the language system itself that presents us with the structured opposition between write and draw. But there is surely more to it than this. Writing and drawing are understood against broader constellations of knowledge. Murphy and Medin (1985), not inappropriately, speak of "theories", which serve to give coherence to categories.23 Writing is understood against a theory of written linguistic communication, drawing against a theory of visual representation. It

40 John R. Taylor

is in virtue of the background theory, that writing (with a pen) and writing (with a word processor) - two very different kinds of activity, objectively speaking - are nevertheless both categorised as instances of "writing". Changes in word meaning are likely to have as much to do with changes in background assumptions, i.e. domain-based knowledge configurations, as with designation ("profiling"). Indeed, the aspect of Cognitive Semantics that promises the greatest scope for insightful studies of meaning change, could well be the importance attributed to background assumptions. The development of Latin scribere from 'make marks on a surface' to 'write', and the development of legere from 'pick out' to 'read', are not just a matter of "restriction" or "specialisation" of meaning (nor of the addition, or subtraction, of semantic features). In each case, the profiled activity remains much the same. What has changed are the background assumptions (the "theories") against which this activity is profiled.

5.

Conclusion

Let us return to Saussure's original insight that a "concept" needs to be characterised both positively (in terms of its actual content), and negatively (in terms of what it is not). Structuralist Semantics chose to separate out these two aspects, proposing a level of designation (the positive content of the signified), and a level of signification (the signified in contrast to other signifieds). This, I think, was an error. Just as the Saussurian sign resided in the integration of the signified and the signifier, so too the signified resides in the integration of designation and signification. Cognitive Grammar achieves this integration by means of the notion of profile and base. The profile is the concept in its positive aspects, i.e. the entity (or category) actually referred to. The base comprises background knowledge that is not specifically designated. But without the base, there can be no profile, and the base, without profiling, lacks structure. The major achievement of Structuralist Semantics is to have emphasised the semantic relations between lexical items. Some of the

Cognitive Semantics and Structural Semantics 41

earlier studies of lexical items within the Cognitive Semantics tradition (e.g. Brugman 1981; Coleman and Kay 1981) probably did tend to study words in isolation from other lexical items with which they stand in contrast. But it would certainly not be fair to say that Cognitive Semanticists have in general been insensitive to matters pertaining to lexical fields, and to implicit contrasts between lexical items. These implicit contrasts belong in the domain-based knowledge against which an entity is profiled. The background knowledge against which a concept is profiled may comprise not just "encyclopedic" knowledge pertaining to a conceptual domain, but equally, "linguistic" knowledge pertaining to the paradigmatic and syntagmatic relations that the linguistic unit contracts with other linguistic units.

Notes 1.

2.

3.

4.

Cf. Langacker (1987: 11): "Language is symbolic in nature. It makes available to the speaker - for either personal or communicative use - an open-ended set of linguistic signs or expressions, each of which associates a semantic representation of some kind with a phonological representation" [bold in original], Lakoff s (1987: 473) characterisation of constructions as "pairings of form and meaning" can also be taken as an endorsement of the symbolic nature of language. Whether or not an item is stored in memory is a function of "entrenchment" (in turn a function of frequency of successful use; Langacker 1987: 59). Obviously, entrenchment is a matter of degree. There is therefore no clean cut-off point between "stored" units and ad hoc constructed units. This is not to deny that certain facets may be more intrinsic to an expression's meaning, and relevant to just about all its uses; nevertheless, even highly central aspects can sometimes be defeated, and outweighed by other, more circumstantial aspects. See Langacker (1987: 158-161). As a matter of fact, Saussure (in the representation of his thought that has come down to us) appears to shy away from the full implications of his theory. Thus, he observes (1916: 160) that if, of the three "synonyms" redouter, craindre, and avoir peur, redouter did not exist, its meaning would be shared out amongst its competitors. Saussure, therefore, appears to presuppose the existence of a conceptual content, which is independent of language, and which has to be lexicalised, some way or other.

42 JohnR. Taylor

5.

6. 7.

8.

9.

10.

This was the principal theoretical import of Berlin and Kay's (1969) work on colour categories. Essentially, Berlin and Kay demonstrated for a semantic domain (colour) the same kinds of universal constraints that Jakobson (1968) had claimed for phonology. Terminology, however, is far from uniform. In maintaining this distinction, Structuralist Semantics aligns itself with a number of other "two-level" approaches to semantics (e.g. Searle 1980, Bierwisch 1981, Kirsner 1993, Wunderlich 1993). Although these approaches may differ in their details (especially concerning the manner in which "linguistic" meanings are represented and get projected onto encyclopedic meanings), a common theme is the assumption that linguistic meanings are unitary, clearly-defined entities, which lack the rich detail derived from experience of the world. For discussion, see Taylor (1994; 1995: ch. 14). For Coseriu, "universals" are to be found, if anywhere, in extralinguistic reality and its categorisation, not in significations. Thus the label "semántica 'universal'" is applied to both prototype theories and theories of necessary and sufficient conditions (Coseriu 1990: 252). Coseriu (1977: 10-11) even charges generative grammar with an exclusively onomasiological (and therefore universalist) perspective: "la grammaire générative part de la réalité extra-linguistique désignée, ou bien d'une pensée prélinguistique 'universelle' (c'est-à-dire non encore structurée par telle or telle langue), et passe pour ainsi dire à travers et par-dessus les langues pour aboutir à la parole." [generative grammar starts from designated extra-linguistic reality, or from 'universal' prelinguistic thought (i.e., from thought which is not yet structured by a particular language), and by-passes, so to speak, the language system, in order to arrive at the utterance.] The proverbial linguist from Mars, on learning that a fauteuil is an object for sitting on, which has arms, legs, and a back, could be excused for supposing that a man giving a piggy-back to his young son, is a fauteuil. The point of this flippant example, of course, is that word meanings are not the "minimalist" (cf. Coseriu 1990: 263) constructs envisaged by Structuralist Semantics, but are likely to be extremely rich in detail and background (encyclopedic) assumptions. E.g. knowledge of seating objects pertains not only to the parts of which they are composed, but also to how humans typically interact with these objects. Concerning Pottier's analysis, Coseriu and Geckeler (1981: 42) do indeed raise the question whether we are here dealing with "an analysis of linguistic content" or "a description of a series of... objects, which is to say, of a part of extralinguistic reality". The authors maintain that although Pottier begins his analysis by considering the objects as such, and the real-world

Cognitive Semantics and Structural Semantics 43

11. 12.

13. 14. 15.

16.

17.

features that distinguish them, he proceeds to eliminate the linguistically irrelevant features, thereby arriving at the (linguistically) "pertinent features". Still, it is legitimate to ask what these "pertinent features" are supposed to be, if not the necessary and sufficient features of check-list theories. For discussion of high and tall, see Dirven and Taylor (1988) and Taylor (in press b). Cf. Coseriu (1977: 10): "Poser l'existence des unités fonctionnelles ne signifie nullement qu'on n'admette dans chaque cas qu'une seule 'signification' (= acception), mais au contraire qu'on s'efforce justement de définir les limites, données par la langue, à l'intérieur desquelles une infinité d'acceptions peuvent se présenter." [To postulate the existence of functional units by no means entails that we allow, in each case, only one 'signification', or reading; rather, we attempt to circumscribe the limits, set by the language system, within which an infinity of readings are possible.] (Note that in this passage, "signification" appears to be used in the sense of "designation", while "unité fonctionnelle" corresponds to my "signification".) For some observations, see Taylor (1995: 223ss.). Particularly influential has been Brugman's (1981) analysis of over, subsequently elaborated by LakofF(1987). In Taylor (1996), I argued, within the Cognitive Grammar framework, for a unitary, schematic account of the possessive morpheme in English, and against the adequacy of prototype accounts. Similarly, for many grammatical categories, it would be fruitless to search for a unitary phonological representation. In this connection, it is noteworthy that while Jakobson (1936) insisted on the methodological necessity to assign a constant, albeit highly abstract, semantic value to each of the Russian cases (otherwise, he argued, the linguistic sign would fracture into numerous form-meaning relationships), he was quite unperturbed by the absence of a unique phonological representation for each of the cases. If the absence of a unique representation can be tolerated with regard to signifiers, one wonders why polysemy should be outlawed with signifieds? Coseriu notes that the phenomenon is not unknown in other languages, cf. German steigen. Or consider the English verb grow. His debts grow day by day would be understood to mean that his debts get bigger (i.e. that they grow "upwards"). (The example is mine, not Coseriu's.) But it is equally possible to defeat the default interpretation: His debts are growing smaller day by day. Coseriu would probably argue, therefore, that "upward motion" is not intrinsic to the semantics of grow - the word "really" means 'change in the vertical extent of an entity'. However, still other uses, e.g. The sound of the music grew less as the band marched away (LDCE) suggest an even more schematic sense, i.e. 'become', 'change in state'. But now, the seman-

44 JohnR. Taylor

18.

19.

20.

21.

22.

23.

tic content of grow has become so impoverished, that it is scarcely possible to differentiate the word from other change-of-state verbs, such as become. It is noteworthy that in addressing the linguistic value of climb, Coseriu proceeds by abstracting what is common to a range of specific uses. Curiously, he fails to implement what is surely the central idea of a structuralist semantics, namely, the possibility of semantic contrasts between climb and other items in the same lexical field. But note that métonymie extensions are not based on similarity at all, but on contiguity (within a conceptual domain). The development of English bead had nothing at all to do with the "similarity", at any level of abstraction, between a prayer and a spherical object on a string. Cf. Coseriu (1977: 17): "Les unités fonctionnelles correspondent d'une façon immédiate à des intuitions globales unitaires." [Functional units correspond immediately to global unitary intuitions.] No doubt, the use of mouse to refer to the computer gadget, was once a "categorización de emergencia". Now, however, mouse is the standard term. (What else is one to call the thing?) Do we therefore say that the "value" of mouse as a name for the small mammal has changed? Surely not. Mouse has simply acquired an additional meaning, and the two meanings (which are related in a fairly transparent way) happily coexist. Structuralism is not the only conceivable alternative to a conceptualist semantics. On a behaviourist semantics, knowledge of a word resides in following the rules for using the word correctly. This is the essence of Wittgenstein's (1978: §43) aphorism that "the meaning of a word is its use in the language". The notion of "theories" can help explain the conundrum touched on at the beginning of this section, viz., by what right can we base a theory of language on such irredeemably, subjective entities as "concepts"? One answer is, that each of us attributes to other people a mental life (replete with "concepts") which is very much of a kind with our own, precisely on the basis of a "theory" that other human beings function in much the same way as we do. Cf. Fodor (1980).

References Berlin, Brent and Kay, Paul 1969 Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press.

Cognitive Semantics and Structural Semantics 45

Bierwisch, Manfred 1981 Basic issues in the development of word meaning. In: Werner Deutsch (ed.), The Child's Construction of Language. London: Academic Press, 341-387. Brugman, Claudia 1981 The Story of Over. MA Thesis, University of California, Berkeley. Coleman, Linda and Kay, Paul 1981 Prototype semantics: The English word lie. Language 57: 26-44. Coseriu, Eugenio 1974 Diachronie, Synchronie und Geschichte: Das Problem des Sprachwandels. (Internationale Bibliothek für allgemeine Linguistik 3.) München: Fink. 1977 L'étude fonctionnelle du vocabulaire: Précis de lexématique. Cahiers de Lexicologie 29: 5-23. 1990 Semántica estructural y semántica "cognitiva". In: Manuel Alvar et al. (eds.), Profesor Francisco Marsà: Jornadas de Filología, 239282. (Col leció homenatges 4.) Barcelona: Publicacions Universität de Barcelona. Coseriu, Eugenio and Geckeier, Horst 1981 Trends in Structural Semantics. (Tübinger Beiträge zur Linguistik 158.) Tübingen: Narr. Cruse, D. Alan 1986 Lexical Semantics. Cambridge: Cambridge University Press. Dirven, René and Taylor, John 1988 The conceptualization of vertical space in English: The case of "tall". In: Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 207-229. (Current Issues in Linguistic Theory 50.) Amsterdam: Benjamins. Fauconnier, Gilles 1985 Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, MA: MIT Press. Fillmore, Charles 1975 An alternative to checklist theories of meaning. Proceedings of the Berkeley Linguistic Society 1: 123-131. 1982 Towards a descriptive framework for spatial deixis. In: Robert J. Jarvella and Wolfgang Klein (eds.), Speech, Place, & Action in Deixis and Related Topics. Chichester: John Wiley, 31-59. 1985 Semantic fields and semantic frames. Quaderni di semantica 6: 222254.

46 JohnR. Taylor

Fodor, Jerry 1980 The present status of the innateness controversy. In: id., Representations: Philosophical Essays on the Foundations of Cognitive Science, 257-316. (Harvester Studies in Cognitive Science 13.) Cambridge, MA: MIT Press. Geeraerts, Dirk 1985 Cognitive restrictions on the structure of semantic change. In: Jacek Fisiak (ed.), Historical Semantics - Historical Word Formation, 297-323. (Trends in Linguistics. Studies and Monographs 29.) Berlin: Mouton de Gruyter. 1997 Diachronic Prototype Semantics. Oxford: Clarendon Press. Heine, Bernd 1993 Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press. 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. (Cambridge Studies in Linguistics 83.) Cambridge: Cambridge University Press. Jackendoff, Ray 1983 Semantics and Cognition. (Current Studies in Linguistics Series 8.) Cambridge, MA: MIT Press. 1990 Semantic Structures. (Current Studies in Linguistics Series 18.) Cambridge, MA: MIT Press. Jakobson, Roman 1936 Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der russischen Kasus. In: id., Selected Writings, 2, 23-71. The Hague: Mouton. 1968 Child Language, Aphasia, and Phonological Universals. The Hague: Mouton. (Tr. by Allan R. Keiler of German version of 1941). Katz, Jerrold and Fodor, Jerry 1963 The structure of a semantic theory. Language 39:170-210. Kirsner, Robert 1993 From form to message in two theories. In: Richard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualizations and Mental Processing in Language, 81-114. (Cognitive Linguistics Research 3.) Berlin: Mouton de Gruyter. Koch, Peter 1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46. 1996 La sémantique du prototype: Sémasiologie ou onomasiologie? Zeitschrift für französische Sprache und Literatur 106: 223-240.

Cognitive Semantics and Structurai Semantics 47

Komatsu, Lloyd 1992 Recent views of conceptual structure. Psychological Bulletin 112: 500-526. Lakoff, George 1987 Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. Chicago: Chicago University Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol 1. Stanford: Stanford University Press. 1988 A view of linguistic semantics. In: Rudzka-Ostyn 1988,49-90. 1991 Foundations of Cognitive Grammar. Vol 2. Stanford: Stanford University Press. Lyons, John 1968 Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. 1977 Semantics. 2 Vols. Cambridge: Cambridge University Press. Murphy, Gregory and Medin, Douglas 1985 The role of theories in conceptual coherence. Psychological Review 92:289-316. Pottier, Bernard 1964 Vers une sémantique moderne. Traveaux de linguistique et de littérature2: 107-137. Raíble, Wolfgang 1983 Zur Einleitung. In: Helmut Stimm and Wolfgang Raíble (eds.), Zur Semantik des Französischen, 1-24. (Beihefte zur Zeitschrift für französische Sprache und Literatur, N.F. 9.) Wiesbaden. Rosch, Eleanor 1978 Principles of categorization. In: Eleanor Rosch and Barbara B. Lloyd (eds.), Cognition and Categorization. Hillsdale: Erlbaum, 2748. Rudzka-Ostyn, (ed.) 1988 Topics in Cognitive Linguistics. (Current Issues in Linguistic Theory 50.) Amsterdam: Benjamins. Saussure, Ferdinand 1916 Cours de linguistique générale. Paris: Payot. Searle, John 1980 The background of meaning. In: John R. Searle, Ferenc Kiefer and Manfred Bierwisch (eds.), Speech Act Theory and Pragmatics, 221232. (Synthese Language Library 10.) Dordrecht: Reidel.

48 John R. Taylor

Talmy, Leonard 1988 The relation of grammar to cognition. In: Rudzka-Ostyn 1988, 165205. Taylor, John R. 1994 The two-level approach to meaning. Linguistische Berichte 149: 326.

1995

Linguistic Categorization: Prototypes in Linguistic Theory. 2nd edition. Oxford: Clarendon Press. 1996 Possessives in English: An Exploration in Cognitive Grammar. Oxford: Clarendon Press. in press a Concepts and domains. in press b Synonymy. Wittgenstein, Ludwig 1978 Philosophical Investigations. Tr. by Gertrude Ε. M. Anscombe. Oxford: Blackwell. Wunderlich, Dieter 1993 On German urn: Semantics and conceptual aspects. Linguistics 31: 111-133.

Diachronic semantics: towards a unified theory of language change? Helmut Lüdtke

1.

Introduction

In linguistic description, phonology, morphology, syntax and vocabulary are more or less clearly separate compartments. This is true for both synchrony and diachrony, at least in mainstream linguistics. Hence sound-change as well as syntactic, lexical and semantic change would each seem to require separate diachronic theories. The same holds for morphological change, although here we might rather distinguish between two types of phenomena, viz. on the one hand, internal reshuffling within existing morphological (sub)systems and, on the other hand, what is called, or rather subsumed under, grammaticalization: the fact that some lexical items become in the course of time grammatical ones. It might therefore seem reasonable to envisage two separate theories for diachronic morphology. If we proceed in the fashion sketched above, the chapters of linguistic diachrony on the whole reflect the traditional headings of synchronic description. But why should they? The paths most often trodden are not always the best ones. Another approach (such as I have in mind), an alternative to "compartment-oriented" research, would aim at a unified theory of language change. The idea here is that speech performance is an organized whole, an engrenage or sequence of intertwined cerebral and outer-world processes whose essential qualities may no longer be correctly perceived if the processes are treated each in different chapters. According to ideas presented in several studies (Lüdtke 1980a; 1980b; 1986; 1988; 1996; 1997) it seems to me that a unified theory as mentioned above could easily accommodate sound-change as well as syntactic and lexical change; morphology (or accidence) is

50 Helmut Lüdtke

also likely to find a place in such a theory. A question, however, is bound to arise as regards semantics: will it be possible to fit changes of meaning convincingly into patterns - let alone laws - similar to those found in the other fields?

2.

Cognitive diachrony

Ever since Hermann Paul (1966: 32 [§16]) it has been stated that the main cause ("Ursache") of language change is normal human speech activity (or performance); this is a basic tenet of empirical linguistics. When treating phenomena of language change it is advisable to distinguish between an observation level relating to performance, and a description level relating to language systems (competence); while on the former we state what is happening or what is going on, on the latter we declare either what is the case now or what has happened in the past. The whole chain of events that leads from a given state of some language system to another may be conveniently divided into three stages: OUTSET - » INTERMEDIATE OUTCOME. The outset of change is human creativity in single speech acts; the intermediate is imitation and repetition ("following suit"), the outcome, difference in state. The latter is described as a property while the former two are observed (if possible) as phenomena. The above tripartite model: CREATIVITY ->· IMITATION - » DIFFERENCE

or, in other terms: INNOVATION - » DIFFUSION - > RESULT

is applicable especially to performance phenomena that require full consciousness on the part of both speaker and hearer, such as lexical borrowing and coinage of new words or locutions. In these cases, the outset may be a single act performed by one speaker. But it is equally possible, and has occurred several times in my own experience, that a number of persons at different places within a short period produce the same innovation without knowing about each

Diachronic semantics: towards a unified theory? 51

other. Even then we may speak of independent single events constituting the first stage in what may result in a change, viz. lexical enrichment. The second stage in the process will be a period when speaker/hearers recognise a word or locution as being commonly heard but still remember it coming into use during their lifetimes; hence they consider it as a loan or a neologism. For the next generation of speaker/hearers, who learn the item in question during childhood, it becomes part and parcel of ordinary language behaviour: this is the third stage. Another type of cognitive development in a language is the semantic expansion of existing lexical items through metaphor and/or metonymy. As regards the former, we may distinguish between first using a given word with a metaphoric meaning as an act of creativity (stage 1), a period of imitation (stage 2) characterized by the speaker/hearers' awareness of using a (well known) metaphor, and a state of affairs in which the hierarchical relationship between the basic and the extended meaning of the word has become blurred; if this has come about we have an example of polysemy. Such a cognitive shift occurred with Engl, law and its counterparts in other European languages: although - historically speaking - the facts are obvious (legislation came first, "laws" of nature were discovered later), to many people, especially intellectuals, the formerly metaphoric meaning of the word appears to be basic; therefore, law has become polysemous. Incidentally, there are linguists who question the appropriateness of my speaking of universal "laws" of language change1 on the ground that these "laws" are susceptible of being transgressed. They obviously no longer regard the term law of nature as being metaphoric. While leaving the eternal dispute as to where polysemy ends and homophony (or homography) begins to lexicographers, I would like to mention one very extreme case, viz. Fr. voler 'to fly' and 'to steal', with its latter meaning arising in the 16th century (and thereby ousting the rival verb rober, which had had that meaning earlier). At the beginning of etymological research linguists were not sure whether voler 'fly' and voler 'steal' were one and the same word. The missing link was to be found in the medieval practice of falcon-

52 Helmut Lüdtke

ry (ars venandi cum avibus): voler un oiseau was used with reference to birds of prey, and hence the verb could be used as a euphemistic metaphor implying swiftness, for 'theft'.

3.

Metonymy

Metonymy seems to be based on some sort of "nearness": from material contiguity (e.g. Lat. bucca 'cheek' > Span., Pg., Cat. boca, It. bocca, Fr. bouche 'mouth') 2 via pars pro toto (e.g. Lat. visus 'faculty of seeing' > It. viso, OFr. vis 'face') to the use of a word in a new context as, for example, German billig 'cheap' (discussed by Rudi Keller, forthcoming), which formerly meant something between 'just' and 'adequate'. The semantic shift undergone by this adjective results from the change in nouns it can qualify: from billiger Preis 'just (or adequate) price' towards billige Ware 'cheap merchandise'. The tertium comparationis of this shift is some sort of metalinguistic concept like BENEFITING THE CUSTOMER and, of course, reference to the market situation. Moreover, there is a special phenomenon which might be qualified as "semantic encroachment". It may be generally described with a term familiar from biogenetics: as a sort of cross-over (or recombination) between two signifiants (A; B) and their corresponding signifiés (α; β). The following example involves two (or rather three) forms of the Latin verb esse 'to be' - es, eris (est) - with their respective meanings: SIGNIFIANT

SIGNIFIÉ

A

eris

'future'

a

Β

es

'present'

ß

Figure 1: Latin > Spanisch 'thou art'

On the way to Spanish, eris 'thou wilt be' became eres 'thou art', while es disappeared as a result of homonymie clash with (Latin) est

Diachronic semantics: towards a unified theory? 53

'is' > (Spanish) es. One may wonder how a grammatical item could possibly, as it were, "exchange" its former future meaning (a) for 'present' (β). How can a meaning become detached from a form it innately belongs to, in order to cross over to another form? If we rule out purposeful language engineering, we have to look for an invisible-hand explanation, i.e. for some speaker/hearers' behaviour that will account for the above example (cf. Keller 1997: 422-427). We can imagine a scenario where es (< es and < est) had both meanings, '(thou) art' and '(he/she/it) is'. In some speech situations this would be tolerable if the overall context eliminated ambiguity; in other cases es would be misunderstood. It has often been observed in European languages that verb forms of the present are used with reference to future events (in connection with an adverb); there seems to be a general rule allowing the substitution of unmarked, or less marked forms for marked, or more marked ones. So there would be nothing strange about Latin es being used instead of eris; but what about the contrary? It would violate the rule. Only in a particular situation when es (2nd person) would clash with es(t) (3rd person), e.g. in slurred speech, would speakers, realizing that they had been misunderstood, possibly resort to full forms on repeating, such as *este (3rd person), with a paragogic vowel (according to Latin sandhi rules reconstructible from Romance dialect material). In order to emphasize that not the third but the second person was intended they might choose the form eris, reasoning that the hearers would certainly understand the present instead of the future meaning, given the particular situation.3 In a second stage of the process ("imitation"), speakers with such experience in their background would be tempted to use eris > eres prophylactically, i.e. in order to ward off the possibility of misunderstanding es as an allegro form of est. When additionally the old synthetic future tense forms came to be more and more often replaced by periphrasis with the auxiliary verb habere (esse-habes instead of eris > eres) the abnormality of eris being used with the present meaning dwindled, so that eres eventually acquired the meaning "(thou) art". Another example of semantic encroachment may be adduced. Latin manducare originally meant 'to chew' whereas its Romance

54 Helmut Liidtke descendants, i.e. Fr. manger, It. mangiare, Rum. mînca, have acquired the meaning 'to eat'. The semantic "nearness" is as obvious here as it was in the previous example. The "invisible-hand" explanation is also similar although this is an instance of lexical and not grammatical cross-over. There was homonymie clash between some very frequent forms of edere (the classical word for 'to eat') and those of esse 'to be', and the table illustrating the situation is analogous: SIGNIFIANT

SIGNIFIÉ

A

manducare

'chew'

α

Β

edere

'eat'

ß

Figure 2. Late Latin 'to eat'

In the history of Latin > Romance the case is a little more complicated since two different words came to be used instead of edere: manducare 'to chew' and comedere 'to eat up'. The latter survived in the Iberian Peninsula (> Span., Pg. comer). The Emperor Augustus is reported to have used manducare with the new meaning, while in Bible Latin both words occur along with classical edere (García de la Fuente 1994: 85-88 and 126-162). As time went by, many words that had rather been somewhat colloquial variants (like caballus 'horse' besides equus) came to be considered as vulgar (in the sense of 'not used in educated writing'). This happened also to manducare meaning 'to eat' so that it was a shock for educated Romans to read Jesus' words at the Holy Communion: "Accipite et manducate! Hoc est corpus meum" (Matthew 26,26 in the Vetus Latina, cf. Jülicher 1938. In the Vulgate manducate was corrected into comedite). 4.

Randomness or directionality?

Both the manducare and the eris case have in common that, as an outcome of the whole process, the more frequent signifié wins the

Diachronic semantics: towards a unified theory? 55

(in a pseudo-Darwinian sense) fitter signifiant over to its side: A goes with β leaving α to find some new partner, i.e. in the eris case to the periphrastic future tense form, in the case of manducare to a loan word masticare borrowed from the Greek μαστιχαω 'to gnash the teeth'. During the intermediate stage 2 there must have been a period of polysemy; but since the hearer would be more easily inclined to interpret the word in question giving it the more frequent meaning, this recombination eventually gained the upper hand. The general speaker-to-hearer strategy (at stage 1) that brings about the cross-over may be formulated as follows: if a frequently occurring meaning is attached to a precarious form (i.e. one liable to misunderstanding) don't worry! Use a makeshift, choosing another, semantically close, form and trust the hearer! He will surely be eager to make good sense of what you are saying. Once people have been successful with this maxim, both as speakers and as hearers, the process is in stage 2, when "collusion" becomes normal: communis error facit ius. The cross-over is accomplished when a new form is established for the less frequent meaning a. There are further instances of semantic encroachment and crossover. In contemporary French, the equivalents of Engl, day, evening, year are patterned couples: jour

soir

an

journée

soirée

année

Figure 3. Time units in French

The upper row stands for 'time' in a strictly chronological perspective, while the derived words in -ée refer to the respective periods as 'being lived through' (le temps vécu). This is the general rule. It is violated, however, in speech practice since an exception is quite regularly made for année, because its partner an is "precarious". This means that speakers shrink from using it in certain contexts and replace it by année even though, strictly speaking, the general semantic rule would require an. The expressions from day to day, from evening to evening, from year to year have as their French equi-

56 Helmut Liidtke

valents de jour en jour, de soir en soir, d'année en année, and not, as one might expect, *d'an en an (pronounced /dânânâ/). The reason, cacophony, is obvious. In a similar way, several hundred years ago, there happened another cross-over with French terms referring to time: Β

hui

aujourd'hui

A

ß

'today'

'nowadays'

α

Figure 4. French adverbs of time

The meaning β occurs, of course, more often than a. The form Β seems to have been precarious for its shortness or maybe for its phonetic similarity to oui 'yes'. Anyway, a cross-over β 'drinking container made out of glass' > 'glassful'

In my opinion, efficiency and expressivity are not directly comparable or complementary on the same level. If we bear in mind that speakers don't want to change their language, and that their principal goal is to communicate as successfully as possible and to reach this communicative goal with minimal linguistic effort, then communicative success is efficient in a rather abstract sense. I would like to call this the efficiency of communication: it is, in other words, the general purpose of communication and the general motivation behind language change. It is true, of course, that avoiding paraphrases or complex words and creating metaphors, metonymies or ellipses instead is efficient too, but, in this case, the efficiency operates on a more concrete level, and it is indeed a linguistic strategy speakers adopt more or less willingly. On the same level, expressivity is also a strategy that speakers can adopt for optimizing their communicative success when they want to impress their interlocutors, treat him or her gently, manifest emotions, show things under a different light etc.; in short, when they want to come out on top. Both strategies then contribute in their specific manner to the general efficiency of communication and of language change.7

66 Andreas Blank

To conclude this section, we can say that efficient communication is to maximize success by either reducing or increasing linguistic effort.8 Thus, a study of the motivations for lexical semantic change must be concerned with both developments, gather individual motivations for particular changes into types and try to identify the cognitive basis for these types. This will be the main issue of section 3. Section 2 will look back into the history of semantics and discuss some of the problems raised by traditional typologies of the motivations for lexical semantic change.

2.

Traditional approaches to the motivations for semantic change

2.1. Stephen Ullmann 's typology The essentials of the 19th and early 20th centuries' reflections on lexical semantic change are summarized in Stephen Ullmann's two books Principles of Semantics (1957) and Semantics (1962). In the latter, Ullmann distinguishes three aspects of semantic change: its "causes", "nature" and "consequences" (cf. the critique in Geeraerts 1983, 1997: 85-92; Blank 1997: 34-44). This has been for decades the most popular and important theory in this domain. As to the causes of semantic change, Ullmann gives a list of six types (1962: 197210), which integrates other, less comprehensive typologies, as e.g. those of Meillet ([1905] 1965) and Sperber ([1923] 1965). A closer look at Ullmann's typology in the light of what has been said in section 1 is rather deceiving: Ullmann's list lacks both a cognitive and an empirical background and is merely an eclectic collection of motivations, necessary conditions and accessory elements.9 From his six types only two are relatively unproblematic, i.e. the "social causes" and the "psychological causes". The first type concerns words that are often used by a group of speakers in a restricted context and, as a result, become semantically restricted to the actual sense they have in this context:

Why do new meanings occur? 67

(3)

Lat. cubare 'to lie' > Fr. couver 'to breed'

Vice versa, the meaning of a word can become generalized when this word is used outside its usual context: (4)

Engl, lure 'decoy' > 'anything that attracts'

The theory of the psychological grounding of semantic change was introduced by Hans Sperber (1965). According to him, emotionally marked concepts can serve as an onomasiological "center of attraction" for other words to verbalize the "attractive" concepts and, vice versa, serve as a cognitive basis, as a semasiological "center of expansion" for verbalizing other concepts (cf. also Blank 1998a: 1517). Concepts within emotionally marked fields (cf. section 3.7.) are frequently akin to new verbalization, be it in an expressive, "exaggerating" way, as in ex. (5), or in the manner of euphemistic "understatement", as in ex. (6): (5) (6)

MHGerm. sère 'wounded, sore' > 'very' VulgLat. male habitus 'in a bad state' > Fr. malade, It. malato 'ill'

While the relevance of these two types is convincing, this is not the case for the remaining four. According to Ullmann, a semantic change can be due to newly developed objects or ideas (need for a new name, cf. ex. 7), to technical, scientific, political or sociocultural developments which influence our conception of things, people, ideas etc. (historical causes, cf. ex. 8), as well as to changes that have already occurred in one language and that are subsequently copied in another (foreign influence, cf. ex. 9). Finally, change can happen when two words are habitually collocated in speech and the sense of one word is transferred to the other (linguistic causes, cf. ex. 10): (7)

Engl, torpedo 'electric ray' > 'self-propelled submarine explosive'

68 Andreas Blank

(8) (9)

Fr. plume 'goose feather for writing' > 'pen with metal nib' OGr. aggelos 'messenger' > 'angel', copying the polysemy of Hebr. ml'k 'messenger', 'angel' (10) Fr. pas 'step' > 'not' ( ne... pas, lit. 'not (a) step') First of all, I believe that Ullmann's historical causes and need for a new name are facets of one and the same type, insofar as in both cases there are new concepts - the submarine bomb, the modern pen - that need to be expressed.10 The same holds true for foreign influence, where it is not the adstrat situation itself that is the cause for semantic change, but the need to verbalize a new concept. Semantic loan is only a rather smart device adopted for that purpose, because it imitates a polysemy that seems to function well in another language and which thus promises to be successful. As to Ullmann's example for linguistic causes, it must be observed that in fact the sense of the whole collocation is transferred to the simple word and not only the sense of the part that is omitted. Lexical ellipsis, therefore, must be defined in a different way than it is interpreted by Ullmann (cf. Blank 1997: 288-292). In our context it is important that the collocation of French ne and pas is not the motivation for the semantic change which pas has undergone, but a necessary condition that makes a change possible. If it was the motivation, all collocations (and complex lexemes in general) would necessarily become elliptic, which is clearly not the case.

2.2. Other causes of semantic change We can thus reduce Ullmann's six types to three. Furhtermore, we find that his typology lacks traditional causes such as "irony" or what Nyrop called "connexion entre les choses" [relation between the things] (1913: 80). To my knowledge, Nyrop was the first to state that concepts in our mind are interconnected and that one concept can evoke those concepts related to it.11 Strong conceptual relations seem to induce semantic change, as happened in the following example:12

Why do new meanings occur? 69

(11) Lat.focus 'fireplace' > 'fire' Another traditional cause for semantic change that does not appear on Ullmann's list is homonymie clash: (12) Lat. vicarius 'village mayor' > Gasc. bigey 'rooster' Lat. gallus 'rooster' > Gasc. *gat\ LateLat. cattus 'cat' > Gascon gat) The traditional interpretation, which goes back to Gilliéron and Roques (1912), sees the semantic change of vicarius motivated by the homonymy of Latin gallus and gattus in Gascon, caused by the parallel sound change of intervocalic -II- > -t- and -tt- > -t-. In a medieval rural society this homonymy could indeed have given rise to confusion. The question is whether the metaphor for vicarius was only created when the homonymie clash had already occured. In agreement with Wartburg (1971), I would rather say that the already existing expressive "Trabantenwörter" [satellite words] vicarius/bigey were used more and more as the normal names for the 'rooster' when the homonymy arose, while the original word gat was falling into disuse. Thus, the homonymy of gat is not the motivation for the metaphor bigey but the cause of its status as the normal word for 'rooster' in Gascon. We can conclude that the avoidance of misunderstandings due to homonymy (as well as to polysemy) is situated on the level of motivations for adoptions of innovations, while the creation of the metaphor of Gascon bigey 'rooster' was due to expressivity, viz. "psychological causes" (for details cf. Blank 1997: 354-359).

70 Andreas Blank

3.

A new typology of the motivations for lexical semantic change

3.1. An empirical approach based on cognitive foundations Traditional historical semantics have not succeeded in conceiving a consistent theory of the motivations for semantic change, but rather they provide a mix of mechanisms, such as metaphor, necessary (but not sufficient) conditions, such as collocations, and actual causes.13 Furthermore, rarely any consideration is given to the different levels of motivations introduced in section 1.1. Considering the eclecticism of a typology like Ullmann's, which until the 1990s was reputed to represent the "state-of-the art" in historical semantics, one should favour an empirical approach to this matter in order to reduce the risk of omitting types that are less striking than others. Such a typology should concentrate mainly on the motivations for semantic innovation.14 It is self-evident that recent developments in pragmatics and semantics (esp. cognitive semantics) should be integrated whenever they support classifying concrete examples and further theorizing in this domain of historical semantics. A corpus of more than 600 particular semantic changes taken essentially from Romance languages and completed by examples from German and English as well as from some other languages was compiled and classified by Blank (1997: 497-533). The items were mainly collected in view of a classification of the mechanisms for lexical semantic change and not with regard to its motivations. This assures a certain randomness and objectivity of the corpus with respect to the motivations for semantic innovation. For each example, I tried to find the most plausible reason for its creation, and then similar cases were grouped together. I have to admit that some examples simply refuse classification, as the specific conditions and motivations for their creation have not yet been established.15 The great majority, however, fits into six main categories, some of which show subdivisions, which I would like to call linguistic (section 3.6.) or cognitive constellations (sections 3r5., 3.7.) favouring semantic change. The six main types of motivations

Why do new meanings occur? 71

are established empirically, but show significant correlations to theoretical frameworks of cognitive semantics, viz. prototype theory, basic-level theory and frames-and-scenes semantics. Furthermore, all six types are consistent with the general motivation for language change, insofar as their common denominator is to enhance communicative efficiency.

3.2. New concept (needfor anew name) Our first type has already been discussed in section 2. New concepts arise when we change the world around us or our way of conceiving it (ex. 13 and 14), but also when we leave our "habitat" and enter a new one (ex. 15 and 16). New concepts can of course be verbalized by paraphrase, but it is more efficient, and in most cases more pervasive, to express them by semantic change. The common background of all four examples is the confrontation with a new referent or concept: (13) Lat. pecunia 'cattle (used as a currency)' > 'money' (14) Engl, mouse 'small rodent' > 'small, hand-guided electronic device for executing commands in computer programs' (15) EurSp. léon 'lion' > AmerSp. 'puma' (16) Fr. lézard 'lizard' > Creole (Réunion) 'gecko'

3.3. Abstract concepts, distant and usually invisible referents The second type of motivations concerns conceptual domains whose referents are either abstract or usually distant or hard to see and thus rather difficult for us to seize intellectually. Abstraction explains the usual metaphorical verbalization of e.g. TIME, UNDERSTANDING, SENSE-PERCEPTION or EMOTIONS (cf. Sweetser 1990: 32-37; Blank 1998b: 19-23). Verbalizing abstract concepts by metonymy occurs more rarely:

72 Andreas Blank

(17) Lat. luna 'moon' > Rum. luna 'month' It can happen that, with time, metaphors (or metonymies) lose their concrete sense and become opaque, as happened in ex. (18): (18) It. capire 'to understand' (< Lat. capere 'to catch') In this case, the cognitive perception of the mental process by means of physical grasping is no longer possible. Here indeed, Italian speakers were quick to create and adopt a new metaphor that renews the old conceptual link between UNDERSTANDING and GRASPING:

(19) It. afferare 'to grip, to grasp' > 'to understand' Concretely seizable but distant concepts and things that are difficult to see can be brought "closer" into our view by metaphor, as in ex. (20) and (21): (20) Fr. gorge 'throat' > 'gorge, canyon' (21) Lat. pupilla 'little girl or doll' > 'pupil (of the eye)' This type of motivation once more illustrates the difficulty of applying the labels "expressivity" and "efficiency" on the level of lexical semantics: verbalizing an abstract or somehow difficult notion by metaphor certainly is an expressive issue that makes communication more convincing, but it is also - and mainly through its expressivity - a very efficient tool of lexical enrichment.

3.4. Sociocultural change Changes in our conception of the world can also lead to the transformation of an already existing complex conceptual system by the loss of one or more concepts, by shifting concepts or by introducing new ones. For example, a change in the legal system made the Latin dis-

Why do new meanings occur? 73

tinction between relatives on one's mother's side (avunculus/matertera) and relatives on one's father's side (patruus/amita) obsolete. This led, among other things, to the extensions of meaning in ex. (22) and (23), which are a part of a greater restructuration in the semantics of Latin kinship terms (cf. Coseriu 1978: 136-138): (22) Lat. avunculus 'uncle on one's mother's side' > Fr. oncle, Rum. unchiu 'uncle' (23) Lat. amita 'aunt on one's father's side' > OFr. ante, ModFr. tante, Occ. tanto, Engd. amda, Rum. mätu§ä 'aunt' Fascinating examples which demonstrate the multiple reorganisation of a lexical structure due to sociocultural change are the French denominations for the different meals of the day: until the 16th century people used to have the main meal (disner/desgeiiner, both from VulgLat. *disjejunare 'to defast') in the middle of the morning and a second, lighter meal (souper) in the afternoon. Changes in lifestyle of the noble society and in the (imitating) urban bourgeoisie shifted the main meal of the day to noon. The long gap between getting up and eating made it necessary to introduce a breakfast in the morning and led to the semantic différenciation of déjeuner 'breakfast' and dîner 'main meal'; souper now meant 'evening meal'. In the 19th century, it appeared more suitable to the members of urban professional societies to have the main meal in the evening, which made dîner to acquire the sense of 'dinner' and déjeuner 'lunch'. For 'breakfast', a new complex lexeme was created {petit déjeuner) and souper now serves to designate a late-evening meal. The binary system of the Middle Ages thus developed through a ternary to a fourpartite structure shifting the words along the temporal contiguity of meals (cf. DHLF, ss.w.):

74 Andreas Blank

MIDDLE AGES

16th - 19th c.

19th/20th c.

EARLY MORNING

desgeiiner/disner MID-DAY

souper EVENING

déjeuner

petit déjeuner

dîner

déjeuner

souper

dîner souper

LATE EVENING

Figure 1. The diachrony of French words for MEALS

3.5. Close conceptual or factual relation When we speak, it can happen that we use a word in a sense that is different from its usual one. Normally, our interlocutor understands what we mean because the context may help and because the word we have chosen usually refers to a concept that is somehow closely linked to the concept we have made it refer to in this concrete speech act. Close links between concepts make name transfers possible and, when they are considered to be efficient, they might become lexicalized, and the word that has undergone semantic change becomes polysemous. According to the nature of this conceptual relation, three types of cognitive constellations can be distinguished: 1. Frame relation: A strong and habitual relation between two concepts within a frame makes speakers express them by using only one word: the frame relation is "highlighted" (cf. Croft 1993: 348; Koch, in press b). This was the type of motivation Nyrop called "connexion entre les choses" (cf. section 2.2.). The examples of this type are mainly metonymies (24), but frame relations may also lead to autoconverse change, as in ex. (25): (24) Lat. testimonium 'testimony, witness' > Fr. témoin 'witnessed (25) It. noleggiare 'to lend' > 'to borrow'

Why do new meanings occur? 75

Completely different frames led Latin plicare 'to fold' to take opposite semantic directions in Ibero-Romance and Rumanian, producing a kind of "interlinguistic antonymy": (26) Lat. plicare 'to fold' > Rum. a pleca 'to leave' (27) Lat. plicare 'to fold' > Sp. llegar, Pg. chegar 'to arrive' The reason for this curious development is probably the following: in the shepherd society of Rumania folding the tents was associated with leaving, while in the marine society of Spain folding the sails was associated with arrival.16 2. Prototypical change: In contrast with Geeraerts (1983), I would like to restrict "prototypical change" to the following cases: In the first case, a word is constantly used to refer to the prototype of the usually designated category (cf. also Rastier, in this volume). This gives rise to restrictions of meaning, as in ex. (28) and (29), which rely on the fact that in France WHEAT is the prototypical CEREAL, and in patriarchal societies the prototypical HUMAN BEING is the MAN:

(28) Lat. frumentum 'cereal' > Fr. froment 'wheat' > 'special type of wheat', Fr. blé 'cereal' > 'wheat' (29) Lat. homo 'human being' > VulgLat. 'male human being' The opposite happens when the word which usually designates the prototype of a category is extended to refer to the whole category, as in the following examples: (30) VulgLat. hostis 'opposed, hostile [!] army' > Olt. oste, OFr. ost 'army' (> MEngl. (h)ost) (31) Lat. tenere 'to hold' > Sp. tener, Sard, teniri 'to have' Here we can say that ΤΟ HOLD (IN THE HANDS) is the clearest and most typical instance of POSSESSION mirroring an anthropologically motivated salience (cf. Koch 1991: 31, this volume), while the OP-

76 Andreas

Blank

POSED ARMY forms a kind of "moral prototype" because of its elementary threat. Note that the prototypical structure of a category is only a necessary condition. The sufficient condition that triggers this type of change seems to be the speakers' strong emotional or factual fixation on the prototype. In the first case, they then use the word usually linked to the whole category in a specific context where only prototypical instances can be referred to (ex. 28 and 29), e.g. in a public discourse when the orator addresses the whole community, but indeed speaks to an entirely male assembly. In the second case, the word for the prototype is expressively taken to refer, in fact, to another member of the category or the category itself (ex. 31). Both "abuses" will probably lead to a higher frequency of use of the semantically extended or restricted word which makes their new senses even more attractive. The question is why in one case we have semantic restriction and in another semantic extension. In my opinion, this depends essentially on what's on the left of the arrow: if this is a word whose referential class is prototypically structurable (like CEREALS or HUMAN BEINGS), then we can only have restriction to the prototype; if, on the other hand, the referential class of the word we examine shows no prototypical structure, but is itself a prototypical instance, then it can only be extended to the whole category. We must keep in mind that in both cases the fixation on the prototype triggers the semantic change.17 Sometimes we find extensions or restrictions of meaning which are difficult to be explained by fixation on a prototype, as ex. (32) and (33), where at first sight it is hard to imagine that one of the two referential classes involved could be a prototypical instance of the other:

(32) MEngl. hound 'dog' > 'dog trained to pursue game' (33) VulgLat. adripare 'to get on shore' > 'to arrive' In these cases, it helps to see if the transfer could have happened inside a specific frame, and, in fact, the typical dog in the frame

Why do new meanings occur? 77

is the HOUND, and in the frame typical arrival is GETTING ΤΟ SHORE. "HUNTING"

"SEAFARING"

the proto-

3. Blurred concepts: So far, we have discussed cases where a name transfer is considered to be efficient. In some cases, however, speakers make transfers without being aware of it, because their knowledge about the limits of these concepts and the respective categories is momentarily or permanently blurred. Confusions of this kind happen every day to almost everybody. If, however, the confusion between two concepts is a widespread matter of fact in a speech community, this can lead to a co-hyponymous transfer. A conceptual field where changes of this kind have happened several times in the history of Romance languages is constituted by RATS, MICE and similar small animals (cf. for details Blank 1998c): (34) LateLat. talpus 'mole' > It. topo, Sard, topi 'mouse' (35) Lat. sorex 'shrew-mouse' > It. sorcio, Fr. souris, Rum. §oarece 'mouse' (36) ? *ratt- 'rat' > Pg. rato, Fr. rat, It. (dial.) rat, rät 'mouse'

3.6. Complexity and irregularity in the lexicon A fundamental speaker-oriented strategy is to communicate at the "lowest possible costs". Consequently, speakers reduce irregularities or superfluous complexity in the lexicon, most of the time without being aware of it. Irregularity may, in fact, be a consequence of "expressive factors" and the efficiency principle pushing speakers to linguistic optimization (cf. Geeraerts 1997: 108), but for the following situations this is not necessarily the background. In any case, like in other instances of semantic change, "good" innovations are subsequently adopted by the community. Four different lexical constellations can be distinguished: 1. Lexical complexity: The more frequently a word is used, the more speakers tend to reduce its signifiant ("Zipfs law"; cf. Zipf 1945:

78 Andreas Blank

142-144). If the word in question is a compound or a syntagmatic construction (e.g. mother-in-law, matter of fact), the reduction may concern one part of the complex lexeme. This process is usually called "lexical ellipsis". Semantically, a simple lexeme also receives the meaning of a complex word of which it is formally a part, so that "absorption" or "incorporation" (of the meaning of the complex lexeme) would be more appropriate terms (cf. Koch 1991: 287; Blank 1997: 291): (37) It. portatile 'portable' > 'notebook-computer' ( ^ computer portatile) (38) Lat. separare 'to separate' > Fr. sevrer 'to wean' ( sovereign, under the influence of MEngl. reign). Where it occurs with semantic change, the formal similarity is backed up by a conceptual relation (usually contiguity), as in the following examples: (39) Fr. forain 'non-resident' > 'belonging to the fair' ( ClassLat. necromantia >

Why do new meanings occur? 79

LateLat. nigromantia 'black art', 'magic' ( 'to bore so.' (54) Engl, bad 'not good' > Engl, (slang) 'good, excellent' Euphemistic and expressive words are subject to a general tendency: their veiling or drastic-hyperbolic power weakens the more frequently they are used. As most of the examples above show, the expressivity or the euphemistic character totally wears away and new euphemisms or expressive words have to be created.

4.

Conclusion

The aim of this paper was to give a typology of the motivations for semantic innovation that integrates what we have learned about language and its function since the times of Ullmann. I am convinced that by drawing upon the background of pragmatic and cognitivist

Why do new meanings occur? 83

models a deeper insight into the reasons why speakers change their lexicon has been gained. The corpus analysis confirmed the general assumptions made in section 1 and the tripartite typology of motivations elaborated by Coseriu. There is, on one side, a particular, situative motivation for an individual to risk a semantic innovation; on the other side, innovation is generally motivated by the wish to communicate as efficiently as possible, i.e. to influence the interlocutor in the desired manner and to do this at the least possible expense. Finally, one finds that the individual motivations can be grouped into six types of general motivations or sufficient conditions for semantic innovation, as presented in section 3. These conditions derive either from our perception of the world and our way of structuring our concepts or from the structure and form of one language's lexicon. The several subtypes are to be understood as more specific frameworks that trigger specific mechanisms of semantic change.

Notes 1.

2.

3.

Instead of "causes", I prefer the term "motivation", because it emphasizes that semantic innovations (like any linguistic innovation) are speaker-based; they occur because speakers have a motive that makes them innovate. I emphasize that the metaphor itself is not, on any level, the cause of the change, but only the mechanism or "procedure" the speaker uses (cf. Blank 1997, ch. 5). As Geeraerts puts it, metaphors, metonymies etc. "indicate the associative mechanisms that define and delimit the set of possible (or plausible) semantic changes" (1997: 103). Coseriu uses the following comparison to underline this necessity: "Dar sólo una explicación genérica de un cambio históricamente determinado es como decir que una casa se ha incendiado 'porque el fuego quema la madera', lo qual es cierto desde el punto de vista genérico ..., pero no nos dice nada acerca de la causa histórica (particular) del incendio." (1958, 104106) [To give a general explanation of a historically occured change is like saying that a house caught fire 'because wood burns', which is correct from a general point of view, but doesn't tell us anything about the historical (particular) reason for the fire].

84 Andreas Blank

4.

5. 6. 7.

8.

9. 10.

11.

12.

13.

The existence of an individual motivation and a sufficient condition does not necessarily lead to a semantic innovation. Speakers can also take another pathway of verbalization, as e.g. word formation or borrowing. It is self-evident that the interlocutor doesn't need to be a concrete person, but can also be a fictive hearer/reader imagined by the author. This central hypothesis of Sperber/Wilson (1986) is not unanimously accepted; cf. e.g. Horn (1989, 194-195); Keller (1995,211-213). The efficiency of language change has not necessarily to be of panchronic validity; it suffices that a group of speakers is convinced that the innovation serves their communicative goals to make it appear efficient. A lexicalized innovation can turn out later to be inefficient, as e.g. in the case of homonymie clash as a consequence of - probably efficient - phonetic change or in the case of "polysémie clash" as a consequence of - certainly efficient - semantic change. This happened e.g. to the polysemous OSp. pregón 'message', 'messenger': the latter sense was later expressed by the derivation pregonero. Cf. Koch (in press a: section 6.3). Maybe the distinction between expressivity and efficiency is also misled because when linguists say "language change" they often mean "sound change" or, at the most, "morphological change". It is obvious that by investigating the diachrony of the lexicon one gets different results. For a detailed discussion cf. Blank (1997: 347-354). The difference is that in the case of torpedo we have an entirely new object and, consequently, a new word, while in the case of pen there already existed a word to designate a former stage of development. Obviously, the word is maintained by what could be called "linguistic conservatism" (Ullmann 1957: 211-212), but, to begin with, writing instruments with a metal nib must have been classified as being similar to traditional pens. To that extent, the metal-nibbed pen was once a new object in search of a name. Cf. Nyrop (1913, 80): "Quand on dit: j'attends le courrier, on pense moins à l'homme qui apporte les lettres, qu'à ces lettres elles-mêmes, et courrier prend ainsi le sens de 'lettres': tout le courrier est pour vous." [When one says: I'm waiting for the 'courrier' (lit. courier), rather than thinking of the man who brings the letters, one thinks of these letters themselves, and thus courrier acquires the meaning 'letters': all the 'courrier' is for you.] Nyrop's "connexion entre les choses" is the type of motivation that often stands behind the tendencies of semantic change detected by Elizabeth Traugott (cf. e.g. 1989, 1990, this volume), which usually run under the label of "subjectification". This critique includes the theories of semanticists not mentioned in section 2, as e.g. de la Grasserie (1908), Stern (1931), Gamillscheg (1951), Kron-

Why do new meanings occur? 85

14.

15.

16. 17.

asser (1952), Duchàòek (1967) and even the otherwise very rich and exemplary study by Nyrop (1913). The reader may notice that this objective is in slight contradiction to the heading of the present section: if we keep the title "motivations for semantic change" this is mainly due to tradition. But as we have already noticed in section 1.1., linguists usually only touch upon successful innovations, i.e. those that have led to semantic change. One example: Lat. collocare 'to dispose' was restricted to Fr. coucher, It. coricare 'to lay down' and to Sp., Pg. colgar 'to hang'. Both ways of restriction seem to be likewise fortuitous. Cf. Rohlfs (1971: 138); Tagliavini (1973: 123-124). A different interpretation is given by Klein (1997: 239-240). For other diachronic aspects of prototypes cf. Koch (1995, 39-41).

References Allan, Keith and Kate Burridge 1991 Euphemism and Dysphemism. Language Used as a Shield and a Weapon. New York etc.: Oxford University Press. Bally, Charles 4 1965 Linguistique générale et linguistique française. Bern: Francke. Blank, Andreas 1993 Das verwaiste Wort. Zum Bedeutungswandel durch Volksetymologie. In: Foltys, Christian and Thomas Kotschi (eds.), Berliner Romanistische Arbeiten. Für Horst Ochse, 43-61. Berlin: Freie Universität. 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Niemeyer. 1998a Der Kopf in der Romania und anderswo - Ein metaphorisches (und metonymisches) Expansions- und Attraktionszentrum. In: Gil, Alberto and Christian Schmitt (eds.), Kognitive und kommunikative Dimensionen der Metaphorik in den romanischen Sprachen, 11-32. Bonn: Romanistischer Verlag. 1998b Les adjectifs temporels du type long/court dans les langues romanes: un cas de 'métaphoricité étroite'. In: Dupuy-Engelhardt, Hiltraud and Marie-Jeanne Montibus (eds.), L'organisation lexicale et cognitive des dimensions spatiale et temporelle. Actes d'EUROSEM 96, 15-37. Reims: Presses Universitaires.

86 Andreas Blank

1998c

Topo et al. - Onomasiologie, Semasiologie und Kognition am Beispiel der Bezeichnungen von MAUS, RATTE und MAULWURF in der Italoromania. Zeitschrift für romanische Philologie 114: 505-531. in press Pathways of lexicalization. In: Haspelmath, Martin, Ekkehard König, Wulf Oesterreicher and Wolfgang Raíble (eds.), Language Typology and Language Universals. An International Handbook. Berlin / New York: De Gruyter. Brown, Penelope and Stephen Levinson 1978 Universals in language use: politeness phenomena. In: Goody, Esther N. (ed.), Questions and Politeness. Strategies in Social Interaction, 56-289, 295-310. Cambridge: Cambridge University Press. Coseriu, Eugenio 1958 Sincronia, diacronia e historia. El problema del cambio lingüístico. Montevideo: Universidad de la República. 1978 Für eine strukturelle diachrone Semantik. In: Geckeier, Horst (ed.), Strukturelle Bedeutungslehre, 90-163. Darmstadt: Wissenschaftliche Buchgesellschaft. Croft, William 1993 The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 4: 335-370. DHLF 1992 Dictionnaire historique de la langue française. Paris: Robert. Duchàòek, Otto 1967 Précis de sémantique française. Bfno: University J.E. Purkynë. Fiehler, Reinhard 1990 Kommunikation und Emotion. Theoretische und empirische Untersuchungen zur Rolle von Emotionen in der verbalen Interaktion. Berlin / New York: De Gruyter. Gamillscheg, Ernst 1951 Französische Bedeutungslehre. Tübingen: Niemeyer. Geeraerts, Dirk 1983 Reclassifying semantic change. Quaderni di semantica 4:217-240. 1997 Diachronic Prototype Semantics. A Contribution to Historical Lexicology. Oxford: Clarendon. Gilliéron, Jules and Mario Roques 1912 Études de géographie linguistique d'après l'Atlas Linguistique de la France. Paris: Champion. Grasserie, Raoul de la 1908 Essai d'une sémantique intégrale. 2 volumes. Paris: Leroux.

Why do new meanings occur? 87

Hopper, Paul J. and Sandra A. Thompson 1980 Transitivity in grammar and discourse. Language 56: 251 -299. 1985 The iconicity of the universal categories 'noun' and 'verbs'. In: Haiman, John (ed.), Iconicity in Syntax, 151-183. Amsterdam: Benjamins. Horn, Laurence R. 1989 A Natural History of Negation. Chicago: University of Chicago Press. Keller, Rudi 2 1994 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. 1995 Zeichentheorie. Zu einer Theorie semiotischen Wissens. Tübingen: Francke. Klein, Franz-Josef 1997 Bedeutungswandel und Sprachendifferenzierung. Die Entstehung der romanischen Sprachen aus wortsemantischer Sicht. Tübingen: Niemeyer. Koch, Peter 1991 Semantische Valenz, Polysemie und Bedeutungswandel bei romanischen Verben. In: Koch, Peter and Thomas Krefeld (eds.), Connexiones Romanicae. Dependerá und Valenz in romanischen Sprachen, 279-306. Tübingen: Niemeyer. 1995 Der Beitrag der Prototypentheorie zur Historische Semantik: Eine kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46. in press a Ein Blick auf die unsichtbare Hand: Kognitive Universalien und historische romanische Lexikologie. In: Stehl, Thomas (ed.), Unsichtbare Hand und Sprecherwahl. Typologie und Prozesse des Sprachwandels in der Romania. Tübingen: Narr. in press b Frame and contiguity. On the cognitive basis of metonymy and certain types of word formation. In: Radden, Günter and Klaus-Uwe Panther (eds.), The Metonymy in Thought and Language. Amsterdam: Benjamins. Koch, Peter and Wulf Oesterreicher 1996 Sprachwandel und expressive Mündlichkeit. Zeitschrift für Literaturwissenschaft und Linguistik 26: 64-96. Kronasser, Heinz 1952 Handbuch der Semasiologie. Heidelberg: Winter. Labov, William 1972 Sociolinguistic Pattems. Philadelphia: University of Pennsylvania Press.

88 Andreas Blank

Mair, Walter Ν. 1992 Expressivität und Sprachwandel. Studien zur Rolle der Subjektivität in der Entwicklung der romanischen Sprachen. Frankfurt: Lang. Meillet, Antoine 1965 Linguistique historique et linguistique générale. Paris: Champion. Nyrop, Kristoffer 1913 Grammaire historique de la langue française, Volume 4. Copenhague: Gyldendal. Rohlfs, Gerhard 1971 Romanische Sprachgeographie. München: Beck. Sperber, Dan and Deirdre Wilson 1986 Relevance. Communication and Cognition. Oxford: Blackwell. Sperber, Hans 2 1965 Einführung in die Bedeutungslehre. Bonn: Schroeder. Stempel, Wolf-Dieter 1983 'Ich vergesse alles'. Bemerkungen zur Hyperbolik in der Alltagssprache. In: M. Faust (ed.), Allgemeine Sprachwissenschaft, Sprachtypologie und Textlinguistik. Festschrift für Peter Hartmann, 87-98. Tübingen: Niemeyer. Stern, Gustaf 1931 Meaning and Change of Meaning. Bloomington, Ind.: Indiana University Press. Sweetser, Eve E. 1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tagliavini, Carlo 1973 Einführung in die romanische Philologie. München: Beck. Traugott, Elizabeth C. 1989 On the rise of epistemic meaning in English: an example of subjectification. Language 65: 31-55. 1990 From less to more situated in language: the unidirectionality of semantic change. In: Adamson, Silvia, Vivien Law, Nigel Vincent and Susan Wright (eds.), Papers from the Fifth International Conference on English Historical Linguistics, 496-517. Amsterdam: Benjamins. Ullmann, Stephen 2 1957 Principles of Semantics. Oxford: Blackwell. 1962 Semantics. An Introduction to the Science of Meaning. Oxford: Blackwell. Wartburg, Walther v. 3 1971 Einführung in die Problematik und Methodik der Sprachwissenschaft. Tübingen: Niemeyer.

Why do new meanings occur? 89

Zipf, George K. 1945 The repetition of words, time-perspective, and semantic balance. The Journal of General Psychology 32: 127-148. 1949 Human Behaviour and the Principle of Least Effort. Cambridge, Ma.: Addison-Wesley.

Diachronic prototype semantics. A digest Dirk Geeraerts

The purpose of this contribution is to present the major points of the argumentation contained in my book Diachronic Prototype Semantics (1997). The book is a systematic exploration of the way in which the notion of prototypicality can be made to bear fruit in historical semantic research; it draws on the various articles on the subject that I have written over the last fifteen years or so, and sketches a conceptual framework in which the different case studies that I have dealt with in the individual papers, fit together into an encompassing view of semasiological change, considered from the point of view of prototype theory. The present résumé (which could be seen as an updated version of the overview that was given in Geeraerts 1992) will not be concerned with examples and case studies, but will boil down the framework presented in the monograph to its essentials. For those who are familiar with the book, the present paper can function as a summary; for other readers, it might possibly serve as an appetizer.

1.

Prototypicality from a descriptive point of view

The crucial distinction to be kept in mind with regard to the use of prototypicality in diachronic semantics is the distinction between a descriptive and an explanatory approach. Prototypicality is basically a descriptive concept, but it fits perfectly well into an explanatory model that links up with the "Natural" models of the lexicon represented by Natural Phonology and Natural Morphology. While the present paragraph will be concerned with the use of prototypicality as a descriptive concept, the following one will deal with the explanatory aspects.

92 Dirk Geeraerts

If the prototypical view is accepted as an adequate model for the description of synchronic categories, specific characteristics of semantic change are explained as predictions following from that structure. It is useful to think of that synchronic prototype structure in terms of the following four features. First, prototypical categories exhibit degrees of typicality; not every member is equally representative for a category. Second, prototypical categories exhibit a family resemblance structure, or more generally, their semantic structure takes the form of a radial set of clustered and overlapping readings concentrating round one or more salient readings. Third, prototypical categories are blurred at the edges; there may be entities whose membership of the category is uncertain, or at least less clear-cut than that of the bona fide members. And fourth, prototypical categories cannot be defined by means of a single set of criterial (necessary and sufficient) attributes. Although these four characteristics do not necessarily co-occur, they are systematically related. The first and third look extensionally at the members of a category, whereas the second and fourth intensionally consider definitions rather than members. Characteristics one and two refer to salience effects and differences of structural weight, whereas three and four focus on flexibility and demarcation problems. (In what follows, I will sometimes use the notion "nonequality" with regard to features one and two, and "non-discreteness" for three and four.) There is obviously much more to say about the status of the four features and their relations (see Geeraerts 1989; Geeraerts, Grondelaers and Bakema 1994), but for present purposes, this brief overview will have to suffice. Turning to historical semantics, we can now turn each of the four characteristics of prototypicality into a statement on the structure of semantic change.

1.1.

Modulations of core cases

By stressing the extensional non-equality of lexical-semantic structure, prototype theory highlights the fact that changes in the referential range of one specific word meaning may take the form of modu-

Diachronic prototype semantics. A digest 93

lations on the core cases within that referential range. Changes in the extension of a single sense of a lexical item are likely to take the form of an expansion of the prototypical centre of that extension. If the referents that may be found in the range of application of a particular lexical meaning do not have equal status, the more salient members will probably be more stable (diachronically speaking) than the less salient ones. Changes will then take the form of modulations on the central cases: if a particular meaning starts off as a name for referents exhibiting the features ABCDE, the subsequent expansion of the category will consist of variations on that type of referent. The further the expansion extends, the less features the peripheral cases will have in common with the prototypical centre. A first layer of extensions, for instance, might consist of referents exhibiting features ABCD, BODE, or ACDE. A further growth of the peripheral area could then involve feature sets ABC, BCD, CDE, or ACD (to name just a few). In the monograph 1997, this hypothesis is supported by a case study involving the close inspection of the development of a recent Dutch neologism, viz. the clothing term legging. (For a separate publication of the case study, see Geeraerts, in press.)

1.2. The development of radial sets By stressing the intensional non-equality of lexical-semantic structure, prototype theory highlights the clustered set structure of changes of word meaning. This hypothesis shifts the attention from the extensional structure of an individual meaning of a lexical category, to the intensional structure of the lexical item as a whole, that is, to the overall configuration of the various readings of the word. The hypothesis suggests that the structure of semasiological change mirrors the synchronic semantic structure of lexical categories, given that the latter involves family resemblances, radial sets, and the distinction between central and peripheral readings. Semasiological change, then, involves the change of prototypically clustered concepts. This general statement can be broken down into two more

94 Dirk Geeraerts

specific ones. First, the structure of semasiological change as a whole is one of overlapping and interlocking readings; specifically, a novel use may have its starting-point in several existing meanings at the same time. Second, there are differences in structural weight among the readings of an item; specifically, there are peripheral meanings that do not survive for very long next to more important meanings that subsist through time. In the monograph 1997, this hypothesis is corroborated by means of an extended case study involving the Dutch item vergrijpen (originally published as Geeraerts 1983a).

1.3. Semantic poly genesis By stressing the extensional non-discreteness of lexical-semantic structure, prototype theory highlights the phenomenon of incidental, transient changes of word meaning. That is to say, the synchronic uncertainties regarding the delimitation of a category have a diachronic counterpart in the form of fluctuations at the boundaries of the item. In the monograph 1997, a specifically striking example of such fluctuations is discussed under the heading "semantic polygenesis" (originally published as Geeraerts 1985b). Semantic polygenesis involves the phenomenon that one and the same reading of a particular lexical item may come into existence more than once in the history of a word, each time on an independent basis. Such a situation involves what may be called extremely peripheral instances of a lexical item: readings that are so marginal that they seem to crop up only incidentally and that disappear as fast as they have come into existence. Specifically, when the same marginal meaning occurs at several points in time that are separated by a considerable period, we can conclude that the discontinuous presence of that meaning is not due to accidental gaps in the available textual sources, but that the meaning in question must actually have come into existence independently at the two moments. The theoretical importance of the phenomenon of semantic polygenesis resides in the fact that it illustrates the existence of transient applications in the diachronic de-

Diachronic prototype semantics. A digest 95

velopment of lexical categories. Since such transient cases could spring into existence at any moment in the history of a word, they are at the same time an illustration of the synchronically flexible character of word meanings: exactly what belongs to a category at one particular moment is not necessarily clear.

1.4.

Semantic change from subsets

By stressing the intensional non-discreteness of lexical-semantic structure, prototype theory highlights the encyclopedic nature of changes in word meaning. That is to say, diachronic semantics has little use for a strict theoretical distinction between the level of senses and the level of encyclopedic knowledge pertaining to the entities that fall within the referential range of such senses. In semantic change, the "encyclopedic" information is potentially just as important as the purely semantic "senses" (to the extent, that is, that the distinction is to be maintained at all). This view follows from a prototype-theoretical conception in general, and from the fourth feature mentioned above in particular, in the following way. If the meaning (or a meaning) of a lexical item cannot be defined by means of a single set of necessary features that are jointly sufficient to distinguish the category from others, the definition necessarily takes the form of a disjunction of clustered subsets. If, for instance, there is no feature or set of features covering ABCDE in its entirety, the category may be disjunctively defined as the overlapping cluster of, for instance, the sets ABC, BCD, and CDE (and, in fact, others). Similarly (turning from a description from an extensional perspective to a description from an intensional perspective), if no single combination of features yields a classical definition of a category, it can only be properly defined as a disjunction of various groupings of the features in question. The subsets within such a family resemblance structure need not themselves constitute different senses (in the theory-laden interpretation of the word); they will not necessarily be recognized as such by the language user. But even if they do not constitute separate

96 Dirk Geeraerts

senses, they are definitionally (and hence structurally) important: even if ABC does not constitute a separate meaning of the category ABCDE in any psychologically or intuitively important interpretation, it is a subset that has to be taken into account in the (disjunctive, clustered) definition of the category. Hence, it is impossible to maintain the view that only the senses of a category (and not those extensional subsets that do not constitute separate meanings, or those descriptive features that are supposed to have "encyclopedic" rather than "semantic" status) are structurally important. From a diachronic point of view, this means that semantic changes may take their starting-point on the extensional level just as well as on the intensional level, or in the domain of encyclopedic information just as well as in the realm of semantic information. Even where a classical definition is possible, extensional subsets or intensional features with an "encyclopedic" rather than a "semantic" status may play a crucial role in processes of semantic change. The examples discussed in the monograph (previously published in Geeraerts 1994) show precisely this: the semantic extensions through which new meanings arise may take their starting-point in extensional subsets that do not correspond with senses in the structuralist sense, even in those cases where the categories in question might be classically defined. To round off the overview, it should be stressed that the aspects of semantic change enumerated here are not necessarily new to diachronic semantics. What is indubitably new, however, is the fact that these more or less known aspects of change can now be incorporated into a global model of lexical-semantic structure. That is to say, from a descriptive point of view the importance of prototype theory probably resides less in the novelty of its observations, taken separately, than in the fact that it brings them together in an overall model of the structure of lexical meaning.

Diachronic prototype semantics. A digest 97

2.

Prototypicality from an explanatory point of view

In the cluster of approaches exploring the concept of naturalness in linguistic theory, some areas of linguistics are better represented than others. Best known are probably Natural Phonology (Stampe 1979, 1987; Donegan and Stampe 1979) and Natural Morphology (Mayerthaler 1981; Wurzel 1983; Dressier 1985), while Natural Syntax (Haiman 1980, 1985) and Natural Text Linguistics (Dressier 1989) are somewhat more recent additions. But what about Natural Lexicology? Elaborating the line of thought started in Geeraerts (1983b), I will argue that prototype theory is part and parcel of a "Natural" theory of the lexicon.

2.1.

The functional motivation for

prototypicality

Given a prototypical conception of semasiological structure, the first step on the road towards Natural Lexicology involves the question how to explain prototypicality. It would seem that the best way to do this is to explain the presence of a prototype-based type of conceptual organization on functional grounds. There are, in fact, at least three functional reasons for having a prototypical conceptual structure of word meanings, and all three are functional requirements that the conceptual system has to fulfil if it is to carry out optimally its task of storing categorial knowledge and making it accessible for cognitive and communicative purposes. The first of these requirements has been mentioned by Eleanor Rosch herself (see Rosch 1977): it is cognitively advantageous to lump as much information as possible into one's conceptual categories. Making conceptual categories as informatively dense as possible enables one to retrieve the most information with the least effort. Clearly, prototypically organized categories achieve such an informational density, because they are clusters of subconcepts and nuances. Further, the cognitive system should combine structural stability with flexibility. On the one hand, it should be flexible enough to

98 Dirk Geeraerts

adapt itself to the ever-changing circumstances of the outside world. On the other hand, the categorial system can only work efficiently if it does not change its overall structure any time it has to cope with new circumstances. Again, prototypical categories obviously fulfil the joint requirements of structural stability and flexible adaptability. On the one hand, the development of peripheral nuances within given categories indicates their dynamic ability to deal with changing conditions and changing cognitive requirements. On the other hand, the fact that marginally deviant concepts can be peripherally incorporated into existing categories indicates that the latter have a tendency to maintain themselves as particular entities, thus maintaining the overall structure of the system. Prototypical categories are cognitively efficient because they enable the subject to interpret new data in terms of existing concepts; as expectational patterns with regard to experience, prototypically organized categories maintain themselves by adapting themselves to changing circumstances. In short, the cognitive system favours prototypical categories because they enable it to fulfil the functional requirements of informational density, structural stability, and flexible adaptability as a pattern of expectations. This functional view of conceptual structure can be further specified in the following way. The flexibility that is inherent in prototypically organized concepts cannot work at random; there have to be a number of principles that restrict the flexible extendibility of concepts, or, to put it another way, that specify the principles according to which concepts can be used flexibly. These principles define what is an acceptable extension of a particular concept. The traditional associationist mechanisms of semantic change (such as metaphor and metonymy) have precisely that function; they restrict the set of acceptable conceptual extensions to those changes that are brought about by regular associationist mechanisms such as metaphor and metonymy. In this sense, then, the traditional classificatory categories of historical semantics can in fact be incorporated into a functional classification of the causes of semantic change. But prototypicality itself has a similar restrictive function: the constraint that new meanings be linked to existing ones prevents the semantic flexibility

Diachronic prototype semantics. A digest 99

of lexical items of deteriorating into communicatively inefficient arbitrariness. In this respect, the most profound reason for the adequacy of prototype theory for specifying the characteristics of semantic change, is most likely the dynamic nature of the synchronous notion of a prototypical conceptual organization. The recognition that conceptual categories are not rigidly defined, and that they combine a number of nuances through the centralising action of a conceptual kernel, implies the possibility of dynamically actualizing the central concept in new peripheral uses. This dynamic character of prototypes can be situated on an even more fundamental epistemological level: it then characterizes a basic trait of human cognition, viz. that of interpreting new facts through old knowledge. Incorporating slight deviations into flexibly interpreted existing concepts, is but a special example of the general characteristic of achieving conceptual efficiency through flexible constancy: the conceptual organization is not drastically altered any time a new concept crops up, but new facts are as much as possible integrated into the existing structure, which can thus remain largely unchanged. From this point of view, prototype theory in semantics is connected with the "cognitive" trend in psychology, stressing the mediating role of existing concepts in cognitive development (Bruner, Piaget); with the paradigmatic trend in the theory of science, stressing the role of existing scientific theories (or "research programmes") in the forging of new ones (Kuhn, Lakatos); and with the phenomenological trend in philosophy, in as far as it stresses the interactional nature of human knowledge and opposes the epistemological monism of idealism and realism (Husserl's theory of intentionality). (These parallels are studied in detail in Geeraerts 1985a. For the philosophical aspects, see also Geeraerts 1993.) The implications of prototype theory for the functioning of the human conceptual capacities make it an explanatory basis for diachronic semantics, because the dynamic nature of human thinking is recognized as one of the fundamental structural characteristics of conceptual categories. In this respect, accepting prototype theory is a question of explanatory adequacy rather than descriptive adequacy:

100 Dirk Geeraerts

prototype theory explains the observed prototypical characteristics of semantic change, because it relates them to general epistemological beliefs about the working of the human conceptual system, beliefs it shares with other cognitive theories. And at the same time, of course, the overall conception of a prototypical organization of conceptual categories can itself be explained on functional grounds.

2.2. Prototypicality in an explanatory framework As a next major step in the argumentation, we should recognize that the type of efficiency achieved by the prototype-based conception of lexical-semantic structure is part of broader range of efficiency phenomena in lexicology. Without trying to be exhaustive, two additional examples of efficiency may be mentioned. First, let us consider homonymie clashes. Gilliéron's famous example involves the collision of Latin cattus ('cat') and gallus ('cock') into Gascon gat (Gilliéron and Roques 1912). The tension is resolved by replacing gat ('cock') by bigey, a local equivalent of vicaire ('curate'), or by azan, the local equivalent of faisan ('pheasant'), or by the cognates of Latin pullus. The moral of the story is usually taken to be that homonymie ambiguities set off therapeutic diachronic changes towards their resolution. The rationale behind the avoidance of homonymy might be called a principle of formal efficiency, more particularly a "one form, one meaning" principle: formally disambiguated languages are functionally superior, because they avoid communicative misunderstandings. Second, popular etymology instantiates a tendency (at least in some cases) towards formal, morphological transparency. In Dutch, for instance, the loan-word hamac 'hammock' is changed into hangmat 'hanging carpet'. The semantic transparency of the latter expression (which is composed of the verbal stem hang- 'to hang' and the noun mat 'carpet, mat') is communicatively efficient; those who are not familiar with the foreign word may grasp (or at least get an idea of) what is referred to.

Diachronic prototype semantics. A digest 101

As a further step, let us now compare the approach developed so far with the functional principles at work in the theory of Natural Phonology. Natural Phonology assumes that research into phonological phenomena has to take into account whatever is known about the physiological structure of our auditory and articulatory organs: the physiological possibilities of and restrictions on those organs determine what is "natural" in phonetics and phonology. The fact, for instance, that sounds may be subject to a process of assimilatory voicing in a voiced environment is a natural phonological process, because our articulatory organs would generally encounter difficulties producing a sequence of a voiced, an unvoiced, and then again a voiced segment. Natural phonological processes do not have to be learned; they follow automatically from the physiological difficulties that occur in the production and perception of speech. Conversely, learning a language implies learning which natural processes have to be suppressed according to the phonological system of a particular language. Word-final devoicing, for instance, is a natural process, but children learning French or English have to learn not to give in to the natural tendency. Natural Phonology distinguishes between two major types of phonological processes: lenition and fortition. (There is a third type relating to suprasegmental, prosodie phenomena, but the present discussion will be restricted to processes involving single segments.) Fortition occurs when a sound segment is pronounced in a more outspoken manner, in a way, that is, in which it can be more clearly distinguished from its surroundings. Dissimilation, diphthongization, and epenthesis are frequently occurring examples of fortition processes. Lenition, on the other hand, occurs when the contrast between a segment and its surroundings is weakened, as in the case of assimilation, monophthongization, shortening, and deletion. Fortition and lenition processes have a tendency to occur preferentially in specific environments. Fortition is typical for "strong" positions, such as vowels in stressed syllables, or word-initial consonants. Lenition favours segments in "weak" positions, such as word-final segments or in unstressed syllables. In addition, there is a stylistic difference between both types: fortition is more likely to occur in

102 Dirk Geeraerts

slow, careful, and formal speech, whereas lenition occurs more readily in sloppy, fast, or familiar speech. Most crucially, both process types serve different aims. Fortition is a hearer-oriented process: it makes speech more distinctive and more clear. Lenition is speaker-oriented: it achieves an "ease of articulation" (in the traditional terminology) that allows the speaker to spend less energy. Both mechanisms, of course, are motivated by efficiency, and hence belong in a functional explanatory framework. But if the kind of phenomena involved implies a distinction between speaker-oriented and hearer-oriented efficiency, could not the same distinction be applied to lexical semantics? On the one hand, prototypical polysemization clearly derives from a speaker-oriented form of efficiency: the advantages achieved through a prototypical conceptual organization primarily involve the stability of the speaker's mental lexicon and his capacity for a flexible response with regard to changing circumstances. On the other hand, efficiency principles such as isomorphism and transparency primarily help the hearer. Structural adherence to the principle of "one form, one meaning" means that it is easier for the reader to decode a particular message: a particular formal cue will only lead to one specific meaning. Similarly, morphological transparency may help the hearer to understand the intended meaning even if he is not familiar with the word as such. Another major principle in this hearer-oriented class could be formal iconicity: sound symbolism, for instance, helps the reader to imagine what the referent of a word could be. This means, in other words, that the functional conception of phonology as developed in Natural Phonology, and the functional conception of lexicology developed here, can be brought together naturally if a distinction is maintained between hearer-oriented and speaker-oriented phenomena. The resulting picture is schematically represented in Figure 1.

Diachronic prototype semantics. A digest 103

SPEAKER-ORIENTED: OPTIMIZATION OF PRODUCTION

HEARER-ORIENTED: OPTIMIZATION OF PERCEPTION

CONCERNING PHONOLOGICAL FORM

ease of articulation: lenition processes

fortition processes

CONCERNING THE LEXICAL RELATION BETWEEN FORM AND MEANING

prototype-based flexibility and stability

• • •

isomorphism iconicity transparency and motivation

Figure 1. A classification of efficiency principles

We can see, in other words, that the prototype-theoretical approach to lexical-semantic structure could easily form the basis for a "Natural" theory of the lexicon, provided that the existence of other types of lexicological efficiency is taken into account. It should be noted that the distinction between the basic forms of the efficiency principle (speaker-based optimization of production versus hearerbased optimization of perception) features in various forms in the recent literature on linguistic change. In Langacker (1977) and Kemmer (1992), for instance, it appears in the form of a distinction between a Principle of Least Effort, and a Principle of Maximal Distinctiveness. Lewandowska-Tomasczcyk's classification of causes of change (1985) makes clear that a term such as "Principle of Least Effort" - although traditionally receiving a speaker-oriented interpretation - may also be interpreted from the perspective of the hearer: maximal coding distinctiveness on behalf of the speaker favours minimal decoding effort on the part of the hearer. In addition, note that the distinction is far from new: one of the oldest formulations is von der Gabelentz's distinction between Bequemlichkeitstrieb and Deutlichkeitstrieb (1891). The comparison with Natural Phonology suggests a specific problem for a functional explanatory theory of lexical change, and this recognition should help to avoid any exaggerated optimism. In general, it has to be recognized that the functional approach sug-

104 Dirk Geeraerts

gested here is far from answering all questions that arise in the context of historical semantics. To illustrate, let us go back to our discussion of popular etymology. It was suggested that a form such as Dutch hangmat (from hamac) may be explained by an economical tendency to have morphologically transparent word forms. Some cases of popular etymology are less clear, however. For instance, while Dutch cichorei 'chicory' is sometimes transformed into suikerij 'sugary', semantic transparency is far from achieved: chicory and sugar have nothing in common, chicory does not even taste sweet. Perhaps we might say that the functional principle at work here is a tendency to exploit the morphological possibilities of the lexicon (that is, to maximize the number of morphologically complex words at the expense of newly introduced base forms). This tendency in itself would then be an illustration of a more fundamental tendency towards an economical lexical organization (keeping the number of lexical base forms down is efficient because it diminishes the memory load of the system). However, even apart from the fact that the semantic opacity of suikerij increases rather than diminishes the strain on lexical memory (the language user has to remember that suikerij has nothing to do with sugar), the operation of the economic principle with regard to the number of lexical forms is unsuccessful, since cichorei actually continues to exist next to suikerij: the transparency principle creates a situation that is in conflict with the isomorphic principle. In short, the operation of the functional principles does not guarantee success: some changes seem to miss their probable goal, or at least yield results that are incompatible with other instantiations of the efficiency factor. Specifically, given that Natural Phonology accepts that hearer-oriented and speaker-oriented processes may be in conflict, the question arises how tensions between the hearer-oriented principle of isomorphism, and the speaker-oriented principle of prototypicality are resolved.

Diachronic prototype semantics. A digest 105

3.

Conclusion

To round off the discussion, it may be useful to specify how the present analysis of diachronic prototypicality from an explanatory point of view, relates to my earlier attempts to define a classification of functional causes of semantic change, as in Geeraerts (1983b). The present classification is a direct continuation of the older one, to the extent that it makes crucial use of the distinction between prototypicality on the one hand, and factors like isomorphism and iconicity on the other. However, I earlier connected this distinction with the difference between expressivity and efficiency, respectively, as fundamental factors shaping lexical changes. In the present framework, the two types of factors (prototypicality on the one hand, isomorphism etc. on the other) are rather seen as different forms of the efficiency principle - a speaker-oriented one, and a hearer-oriented one, respectively. This does not mean that expressivity no longer has a role to play in the classification of causes of lexical change. Expressivity (intended here in the sense of 'the need to express something verbally') remains the basic motivating force behind any form of lexical change, either because a new concept has to be put into words, or because an existing concept receives a different form of linguistic expression. The different (and competing) types of efficiency, on the other hand, suggest different formal ways in which this communicative, expressive intention may be realized. In this sense, the present framework makes explicit the asymmetrical relationship between expressivity and efficiency that was inherent in my original proposals (see Blank 1997: 362): expressivity is always a primary cause of change, whereas efficiency involves the choice of the linguistic means realizing the expressive intention. At the same time, by highlighting the distinction between speaker-related and hearer-related types of efficiency, the explanatory approach described here as part of a summary of my book Diachronic Prototype Semantics (1997), opens the way towards the development of a theory of "Natural Lexicology".

106 Dirk Geeraerts

References Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandel am Beispiel der romanischen Sprachen. Tübingen: Max Niemeyer Verlag. Donegan, Patrica and David Stampe 1979 The study of natural phonology. In: Daniel A. Dinnsen (ed.), Current Approaches to Phonological Theory, 127-173. Bloomington: Indiana University Press. Dressier, Wolfgang U. 1985 Morphonology. Ann Arbor: Karoma Press. 1989 Semiotische Parameter einer textlinguistischen Natürlichkeitstheorie. Vienna: Verlag der österreichischen Akademie der Wissenschaften. Gabelentz, Georg von der 1891 Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Tauchnitz. Geeraerts, Dirk 1983a Prototype theory and diachronic semantics. A case study. Indogermanische Forschungen 88: 1-32. 1983b Reclassifying semantic change. Quaderni di Semantica 4:217-240. 1985a Paradigm and Paradox. Explorations into a Paradigmatic Theory of Meaning and its Epistemological Background. Leuven: Universitaire Pers. 1985b Semantische polygenese: een bijzondere vorm van historische betekenisvariatie. Forum der Letteren 26: 120-130. 1989 Prospects and problems of prototype theory. Linguistics 27: 587612.

1993

1992

1994

1997 in press

Cognitive semantics and the history of philosophical epistemology. In: Richard A. Geiger and Brygida Rudzka-Ostyn (edsj. Conceptualizations and Mental Processing in Language, 53-80. (Cognitive Linguistics Research 3.) Berlin / New York: Mouton de Gruyter. Prototypical effects in diachronic semantics: A round-up. In: Günter Kellermann and Michael D. Morrissey (eds.), Diachrony within Synchrony: Language History and Cognition, 183-203. (Duisburger Arbeiten zur Sprach- und Kulturwissenschaft 14.) Frankfurt etc.: Lang. Subsets in semantic change: generalizing inductive generalization. In: Keith Carlon and Emma Vorlat (eds.), Perspectives on English, 128-139. (Orbis: Supplementa 2.) Leuven: Peeters. Diachronic Prototype Semantics. Oxford: Clarendon Press. Neologism at short range. In: Rainer Schulze (ed.), Making Meaningful Choices. Tübingen: Gunter Narr Verlag.

Diachronic prototype semantics. A digest 107

Geeraerts, Dirk, Stefan Grondelaers and Peter Bakema 1994 The Structure of Lexical Variation. Meaning, Naming, and Context. (Cognitive Linguistics Research 5.) Berlin / New York: Mouton de Gruyter. Gilliéron, Jules and Mario Roques 1912 Etudes de géographie linguistique. Paris: Champion. Haiman, John 1980 The iconicity of grammar: isomorphism and motivation. Language 56:515-540. 1985 Natural Syntax. (Cambridge Studies in Linguistics 44.) Cambridge: Cambridge University Press. Kemmer, Suzanne 1992 Grammatical prototypes and competing motivations in a theory of linguistic change. In: Garry W. Davis and Grant L. Iverson (eds.), Explanation in Historical Linguistics, 145-166. (Current Issues in Linguistic Theory 84.) Amsterdam: John Benjamins Publishing Company. Langacker, Ronald W. 1977 Syntactic reanalysis. In: Charles N. Li (ed.), Mechanisms of Syntactic Change, 57-139. Austin: University of Texas Press. Lewandowska-Tomaszczyk, Barbara 1985 On semantic change in a dynamic model of language. In: Jacek Fisiak (ed.), Historical Semantics - Historical Word Formation, 297-323. (Trends in Linguistics. Studies and Monographs 29.) Berlin: Mouton de Gruyter. Mayerthaler, Willy 1981 Morphologische Natürlichkeit. (Linguistische Forschungen 28.) Wiesbaden: Athenaion. Rosch, Eleanor 1977 Human categorization. In: Neil Warren (ed.), Studies in Cross-Cultural Psychology I, 3-49. London: Academic Press. Stampe, David 1979 A Dissertation on Natural Phonology. New York: Garland. 1987 On phonological representations. In: Wolfgang U. Dressler, Phonologica 1984,287-299. Cambridge: Cambridge University Press. Wurzel, Wolfgang U. 1983 Thesen zur morphologischen Natürlichkeit. Zeitschrift für Germanistik 4: 196-208.

Cognitive semantics and diachronic semantics: the values and evolution of classes François Rastier For the centennial of the publication of Bréal's Essai de sémantique.

The theory of prototypes is the best known contribution of cognitive semantics to lexical semantics. Yet, due to the very fact of its universalism, it does not succeed in defining the specific lexical classes of different languages, nor in accounting for diachronic evolution. We shall nevertheless take up the problem of qualitative inequalities among lexical classes. We shall thus distinguish two types of prototypes: (i) the terms for generic usage, which are neutral with respect to associated values, and (ii) the paragons, terms which usually carry a positive value. We shall then deal with the question of qualitative inequalities in diachrony by showing how the social evolution of norms of valuation leads to rehandling the structure of elementary lexical classes (taxemes); we shall finally give a detailed account of an example dealing with the history of denominations for face in French since the 15th century. Besides from the fact that structural semantics and cognitive semantics are not unified movements, they remain difficult to compare because their objects and objectives - none of which have been reached yet - are somehow different. In the field of diachrony, the contributions of structural semantics have been, to tell the truth, notorious for a long time (see Coseriu 1964) and if the cognitive issue managed to stimulate it (see the works of Koch 1995 and Blank 1997), it remains that diachronic cognitive semantics (Sweetser 1990, for instance) has not or not yet shown any visible theoretic or practical progress: most of the time, it does not do much more than retrace well known problems (see Nyckees 1997).

110 François Rastier

1.

The panchronic problem of qualitative inequalities

At the beginning of the thirties, Hjelmslev set the problem of qualitative inequalities among linguistic categories by distinguishing intense terms and extense terms.1 The main merit of the cognitive theory of prototypes will have been to restate this problem among lexical classes. Yet, the concept of prototype does not by itself account for these inequalities. Indeed, it introduces two types of fuzziness: one through its own definition which varies from one author to the other, and the other from its use to describe matters of graduality. Moreover, it must be founded. Why does a given term become prototypical or ceases to be such? This question is never asked and nothing in the theory of prototypes gives us a way of knowing how prototypes are born, grow and disappear. Hjelmslev's problem of qualitative inequality of lexical units inside the taxeme must be considered in synchrony as well as in diachrony in a panchronic perspective taking into account the structure of lexical classes. Whereas in a synchronic description, discrete representations are preferred, continuous representations with thresholds are necessary in diachrony. The panchronic perspective thus requires to articulate two kinds of representations: discrete and continuous. If in synchrony the relations inside lexical classes can be characterised by discrete semic oppositions, the gradual nature of diachronic evolution can be represented by dynamic models which, without contradicting the semic analysis, identify sememes as characteristic zones within dynamic evolutions (see infra, 2.2.).

1.1. Some weak points of cognitive semasiology 1.1.1. The problem of semantic classes The main shortcoming of cognitive semantics is the weakness of its theory of semantic or conceptual classes. This is firstly due to the vagueness of Rosch's concept of category, to the theorization of this vagueness (see the cue validity) and to the absence of a defined point

Cognitive semantics and diachronic semantics 111

of view on its constitution: from which point of view, according to which system of relevance can a specimen be considered deviant? Rosch's take-over consists in centering the inequalities of the category around the prototype, furthermore postulating that there is one and only prototype per category.3 Correlatively, one must recall the weakness of the concept of domain in cognitive semantics. Langacker for instance sustains that "semantic structures ... are relative to 'cognitive domains'", and that a domain "can be any sort of conceptualization: a perceptual experience, a concept, a conceptual complex, an elaborate knowledge system, and so forth" (1986: 4). How can one solve this vagueness? The use of the theory of typicality in diachronic lexical semantics would therefore request some re-elaborations, especially of the concept of category, an undefined concept which certainly is only an experimental artefact of cognitive psychology. The concepts of taxeme (Pottier 1974) and domain (Coseriu 1968) taken from structural semantics seem to us to be better defined and more operative. Especially, they are not founded on an ontology like Rosch's concept of category (Rosch 1978) nor on the illusion of a perceptive naturality,4 as for Berlin and Kay (1969); one can hence conceive their evolution inside a culture and a history. The question of lexical classes cannot be solved, nor even asked, in a semasiological perspective. Indeed, the inventory of all acceptations of a lexeme or grammatical morpheme is not a semantic class, since it has no other common principle than the identity of its signifiers (contingent criteria based on expression and not on content).

1.1.2. The semasiological method and onomasiology Cognitive psychology and cognitive semantics diverge precisely on a crucial point of methodology. Whereas for Rosch and the psychologists who came after her categories are classes of concepts (and their putative corresponding objects), the study of which should be submitted to an onomasiological method,5 for cognitive semantics they are classes of acceptations which are studied through a semasio-

112 François Rastier

logical method. Taking a signifier as invariant, it tries to cross-define its various acceptations, as formerly Katz and Fodor (1963) did with the word bachelor, cross-defining daringly male sea lions, knights, unmarried men and students. It thus neglects that these acceptations do not have the same history, are not generally found in the same discourses, nor the same genres, nor the same contexts and that they therefore are distinct linguistic units. This seems paradoxical: the onomasiological method of structural semantics (and, with due allowances, of Rosch's theory of concept), in so far that she starts from the signifier to discriminate the synchronic varieties and diachronic varieties of the signifier, is closer to a "conceptual" approach than cognitive semantics which nevertheless claims such an approach. But the semasiological method, the most traditional one can find, is only descriptive and not explicative. It can only rest upon the prelinguistic prejudice, born from the philosophy of language, that to one word corresponds one signified; and, as this is obviously not the case, one must find for it a preferential signified, or more precisely a basic conceptualization (counterpart to the literal sense in vericonditional semantics). Cognitive semantics therefore postulates an identity or a partial semantic equivalence between the different meanings of the word and tries to reduce polysemy by organising the acceptations around one acceptation considered as central, i.e. the prototype - without even thinking that polysemy is an artefact of a sign-based linguistics. It is, as always, a way to weaken semantics in order to reinforce ontology.6 Nevertheless, the semasiological approach, as it is usually applied in cognitive semantics, has three characteristics which make it hardly compatible with a diachronic approach: (i) the list of acceptations of a lexeme or grammatical morpheme is considered achronic. (ii) The structurally central acceptation or prototype is defined independently of any diachronic consideration (in Lakoff 1987: 419 for over or Langacker 1986: 3 for ring), (iii) The only temporality retained is the one, internal and ideal, of cognitive operations: it is abstract and unhistorical, as it is represented in a transcendental space.7

Cognitive semantics and diachronic semantics 113

By declaring that the history of language is a series of distributions, Bréal had introduced a diachronic structural principle,8 which Saussure transposes in synchrony in the theory of value. This differential principle will precisely enable the constitution of semic analysis and its use to account for the different phases of evolution. The semic analysis is then considered in context, since the new meanings are described from the new contexts (a semic analysis which would reduce itself to analysing words out of context would be illegitimate because normative). Yet, if the polysemy of acceptations, a contingent fact in synchrony, can be enlightened by diachronic studies which show how one managed to pass from one acceptation to another, this does not entail that it should be the preferential object of diachronic lexicology.

1.1.3. The factors of diachronic changes Factors are not causes, and the description generally limits itself, and wisely so, to the how of evolution; but we shall see further that one cannot avoid the question of why. a) The four operations. — Since the beginnings of lexical semantics, four operations have persistently qualified changes in meaning: extension and restriction, metaphor and metonymy. These very operations which are used to describe the diachronic relations (for instance in Darmesteter 1887) are also used to describe the synchronic relations between acceptations from Reisig (1839) to Clédat (1845) up to Robert Martin (1992). They are furthermore linked to a semasiological perspective, because they link two states (synchronic and diachronic) of a signified and they take the signifier as invariant. They can be grouped by two: extension and restriction are logical concepts which belong to the classical theory of ideas,9 metaphor and metonymy belong to the restricted rhetoric, theorized by the grammarians of the 18th century (Dumarsais 1730 in particular), precisely to save the literal sense and with it the own identity of concepts.

114 François Rastier

(i) Logico-referential operations. — Extension and restriction belong to referential semantics10 which is applied to objects according to an unverifiable principle of quantitative variation very wellknown under the name of law of Port-Royal. These operations rest upon (i) a discrete ontology, in such a way that the designated objects can be counted, and (ii) a logic of classes, in such a way that relations of inclusion can be described. (ii) Tropes. — It is known that the use of tropes in diachronic semantics, even when it claims to have its roots in Lakoff (Sweetser 1990), is the most traditional thing one can find: it pre-exists even to Bréal's semantics, since it takes an important place in Darmesteter (1877), and already earlier in Reisig (1839). The description through tropes supposes a deviation, or at least a variation between the literal sense and the figurative sense. Yet, except if we repeat Saint Paul, one cannot pretend that the ancient meaning is literal and the new one figurative, nor a fortiori that the evolution of languages departs from nature, as Dumarsais regretted in a famous page about catachresis.11 The use of rhetorical concepts in diachronic semantics is not selfevident because tropes are used there to name relations that must still be described and explained. Moreover, if one accepts that the trope is a textual form, the relations which are in this way decontextualized are not tropes, strictly speaking. We consider metaphor and metonymy as critical points of semantic forms, but it is not enough to name them to consider them described.12 What is the difference between the units they unite? In other words, what is the metaphorical or metonymical orientation? What motivates these paths? To answer these questions, we shall study hereunder the relative evaluations of the terms matched, especially their respective position in relation to the evaluative thresholds inside the taxeme. b) Two evolutions. — Studying the relative valuation of the units matched seems to be for us a means to set the problem of evolution as a whole without resorting to logics or rhetorics. In a study on paragons (1991: 198-202), we have extracted a law of panchronic valuation which accounts for two complementary evolutions:

Cognitive semantics and diachronic semantics

115

(i) The evolution through extension goes from the higher valued term to the lower ones: pecunia (Latin for 'cattle') has extended to mean 'richness'. (ii) In the same way, restriction goes from the less valued to the most one: frumentum, which in Latin meant 'cereal', became froment ('wheat') in French (the most valued cereal); a form of this valuation remains in the familiar acceptation of blé ('dough'). Viande (with the generic meaning of 'food' in Old and Middle French) "becomes" viande ('meat') to designate food par excellence.13 This applies also to synchrony (diatopical): in Marseille, for instance, you can hear J'ai un enfant et deux filles Ί have a child and two girls', enfant being restricted to 'boy', eminently valorized in this Mediterranean city. Qualitative inequalities between marked terms (or paragons) and unmarked terms are thus linked to two principles of distribution and panchronic summation. Extension is hence a distribution of the positive evaluation from the paragon, and restriction the summation towards the paragon. The law of panchronic valuation expresses in this way the relations between intense and extense zones, the restriction towards the highly valued consisting in a passage from the extense zone to the intense zone, and the extension from the highly valued bringing about the opposite movement.14

r

τ

Marked sememe

panchron restrictioi

panchronic extension Unmarked sememe

Unmarked sememe

V

Λ

Intense zone

Extense zone Figure 1. Law of panchronical valuation

J

116 François Rastier

Examples: synchronic restriction, see enfant·, diachronic restriction, see viande·, synchronic extension, see gagner son bifteck 'to earn one's living'; diachronic extension, see egeria, hoover. Let's return to our four diachronic operations. (i) Extension and restriction. — We can now reverse in terms of valuation the law of Port-Royal (which nevertheless is only a law of quantitative and not qualitative variation): the more an intension is valued, the more extensions it has. (ii) Tropes. — If we agree on this point, metonymy (and synecdoche which it is often confused) does not obey to other rules: it is from this point of view only a particular case of extension. It is the most priced part or associated unit which extends its name to the whole or the set. The transformation of proper names into "common nouns", extremely widespread and witnessed everywhere, goes in the same direction: egerias and Hercules, Hoovers and PCs are everyday examples. Admittedly, these valued terms become unmarked and get neutralized by their very use, and other words come and replace them in a never ending process without any direction - but not without showing some regularities. The usage of the concept of metaphor answers a totally different problem, the one of changes in semantic domains. But, since taxemes are included in domains, any change of domain leads to an alteration of the taxeme. There was a time when, in Old French ouailles (from the Latin oviculas) meant 'ewes' in the domain AGRICULTURE and 'faithful companions' in the domain RELIGION. In modern French, only the last acceptation survives. The evolution process can be described in the following way: the evangelical metaphor of the good shepherd introduces, among others, a picture of the domain AGRICULTURE into the domain RELIGION. As every metaphor, it shows two effects: in the classeme, it virtualizes the generic seme and actualizes a new generic and afferent seme; in the semanteme, it reshuffles the hierarchy of attributes (cf. Rastier 1987: ch. VII).

Cognitive semantics and diachronic semantics 117

2.

The taxeme as semantic form

We have used the concept of semantic form according to our hypothesis that semantic perception15 pertains to pattern recognition and not to computation.

2.1. Qualitative inequalities inside the taxeme That qualitative inequalities inside lexical classes can be described through intensity and extensity (Hjelmslev) is a fact that we shall relate with the hypothesis of fundamental aesthetics: languages do not articulate descriptions (as the objectivist tradition would like it to do) but evaluations. In particular, qualitative inequalities inside lexical classes seem to be linked to social evaluations which evolve in history. This is not surprising if we recall that in spite of their name natural languages are indeed cultural productions.1 Structural semantics, or at least the impoverished image given of it by most manuals, presents nevertheless one deficiency: the absence of qualitative inequalities. Here are two ways to overcome it: a) To distinguish evaluative zones inside a taxeme (by drawing one's inspiration from descriptions given formerly by Coseriu 1968). It is a means to break with denotation as no metrics can distinguish the big from the huge or the cold from the icy. b) To take account of diachronic variations of the signified: (i) by variations of thresholds inside the taxeme, (ii) by changes of taxeme (iii) and finally by changes of domain.

2.2. Elements of a morphodynamic model We shall use for this purpose a model taken from the theory of differential varieties and dynamic systems. From Thorn's and Zeeman's works in particular, it has been applied to the domain of speech and case-frames by Petitot (1983, 1985), to semasiological semantics (on

118 François Rastier

the polysemy of Fr. encore by Victorri and Fuchs 1996), to the struc11 turai analysis by Piotrowski (1997). Continuisi modelization presents one important advantage: the gradual and the discrete can be described as particular cases of the continuous and not the reverse. Hence, without suggesting any strong hypothesis on the continuous character of the "semantic space", neither on the spatial nature of cognitive schemes, we admit that the semantic discretization consists in isolating outstanding points or zones on dynamics. We make the assumption that semantic evolution can be represented on a gradual process with thresholds: the discrete elements (like semes and phemes) result from the capture of discontinuities (see the phenomenon of categorial perception discovered by Liberman). A morphodynamic model is characterized by the functional coupling of an external space (or substrate space or also space of control) with the internal states of a system S. The singularities of the internal space are projected as discontinuities on the external space. Pass

Attractor

Figure 2. Dynamic system S

Cognitive semantics and diachronic semantics

119

As we have chosen an onomasiological perspective, we consider as semantic space the set of the sememes of a taxeme,18 for instance the one of levels of temperature: 'icy', 'cold', 'cool', 'lukewarm', 'hot'. 19 acceptability threshold

acceptability threshold

Figure 3. Dynamic representation of a taxeme

Regrettably, this representation ignores the context, because determining the main attractor and the position of qualitative and acceptability thresholds depends on the context (cool is neutral for a beer, not for a bath20). The taxeme displays three main basins of attraction separated by two absolute maxima: we call doxal zone the big basin of attraction delimited by two absolute maxima and located below the acceptability thresholds; and paradoxal zones the two zones located beyond these thresholds. Each sememe corresponds then to one local attractor. All the various contexts attested for it define its basin of attraction (in synchrony). The bottom of the basin of a sememe corresponds to the zone of semantic stability, in other words to its meaning (prevailing acceptation, or in other words set of semes inherent to the sememe). The slopes of the basin, zones of instability, correspond to usages. The basin broadens or narrows as the number of contexts grows or diminishes. The marked or intense terms correspond to narrow basined

120 François Rastier

attractors and the unmarked or extense terms to wide basined attractors. The slope of the basins vary whether the sememe has more or less usages in the corpus. Two neighbouring sememes are separated by passes whose altitude is also variable: they are low when the sememes count many similar contexts, and high when they only count a few or none at all. Parasynonyms are separated by low passes which do not cross any qualitative threshold; the distinction, which is always possible, consists in lowering the closest qualitative threshold.

Discretization qualitative threshold

'visage'

•face'

2. 'visage'

'face'

Marking

Figure 4. Discretization and marking

The differential principle of structural semantics expresses itself by a covariance of the basins of attraction of sememes. Any local distortion may have consequences on the global form and neighbouring sections of local forms.21 Yet, as the global determines the local, even if an important local perturbation may modify the global

Cognitive semantics and diachronic semantics 121

organization, a taxeme shows a certain resistance to the distortions brought by the occurrence of new contexts. The local evolution of the sememes inside a taxeme can be described by the modification of their basin of attraction and of the passes which separate this basin from its neighbours. The global evolution of a taxeme can be described as a merging or separation of basins of attraction in a general dynamic. The proximity of sememes is expressed by the presence of their basins inside larger semantic zones (which we shall call evaluative zones: the main ones are the doxal and paradoxal zones). The semantic distance between sememes is expressed by high passes between their basins. When the basins are shallow and the passes low, the equivalence prevails on the opposition or, in other words, coactivation prevails on reciprocal inhibition; when the basins are deep and the passes high, inhibition prevails on coactivation. When the pass which separates two sememes is low, they share common contexts (as is the case for vis and face, which in Old French are present in the same contexts, as is still shown today by the partial equivalence of the phrases vis-à-vis and face-à-face). The high passes (local maxima) indicate qualitative thresholds: their slopes are in the same evaluative zone. The highest passes (absolute maxima) indicate acceptability thresholds: their slopes are located in different evaluative zones. The thresholds are located at points of singularity: local minima and absolute maxima (which are transition points between opposed evaluative zones).22 As far as diachronic or synchronical evolutions are concerned, two opposed movements can be considered: to the extension corresponds a broadening of the basin of attraction with a lowering of the point of stability, whereas to the restriction corresponds the downcoming of the highest qualitative threshold. When a semantic zone contains several attractors, two outstanding attractors may represent the class, because every form is recognized by its singularities:

122 François Rastier

• The attractor closest to the highest relative maximum or to the superior qualitative threshold (corresponding to the paragon, e.g. bifteck). • The attractor furthest to the absolute maximum, which is also the one whose basin of attraction is the most widespread and/or the deepest because it accepts the highest number of contexts (we call it generic neutral, e.g. rue). To these two attractors correspond the two main and contradictory definitions of the prototype as defined by Rosch's theory. acceptability threshold

acceptability threshold

generic neutral Figure 5. Thresholds

The relationships between part-whole inside the global basin of the taxeme can also be described as relationships between these two outstanding points, in a way which does not have recourse to the concept of inclusion, but to the one of relationships between minima and maxima. A "logical" representation can indeed norm and codify a taxeme but not represent the dynamic it has created nor the ones that make it evolve. According to the modelization we have put forward (1987), forces are inhibitions, activations and propagations. These operations suppose "potential differences" which are valuations.

Cognitive semantics and diachronic semantics 123

2.3. Evolutive paths: endoxal and paradoxal We distinguish two kinds of differential thresholds, qualitative thresholds and acceptability thresholds. They divide as we have seen the taxeme into two types of zones (doxal and paradoxal). Whether the evolution crosses or not the acceptability thresholds, it will be called doxal or paradoxal. The paradoxal evolution is obviously complex, because it supposes the referral of two contradictory dynamics (which correspond to two opposed doxas): to cross an acceptability threshold in a dynamic A, the litigious content must be below the acceptability threshold in a dynamic Β (see Chamfort: "a little boy asked his mother for jam: 'Give me too much', he said to her" [1968: 341]). One can further distinguish two kinds of doxal evolution, whether it crosses or not a qualitative threshold. Among the evolutions which cross a threshold, we shall finally distinguish between neutralizing paths and valuating paths. We can then deal with a third type of evolution, which the law of panchronic valuation does not account for: the extension from the neutral term. This endoxal evolution starts from the middle or neutral terms with respect to evaluation (which are often the most frequent terms), to designate the whole taxeme. In this way, rue 'street' can express in French all types of traffic lanes, avenues, walks (in which streets are included).23 It can be considered that the neutral term has the largest number of contexts (simply because it contains less specific attributes: it is therefore compatible even isotopically with a large number of sememes). Thus, the sememe which in the taxeme has the widest and most stable basin (coinciding with the minimum) may designate the whole taxeme.

124 François Rastier

1 extension

avenue

boulevard

street 1

street 2 Figure 6. Extension from the neutral term

The same endoxal evolution can take place several times in history. For instance passer in Latin ('sparrow') becomes pájaro ('bird') in Spanish, passereau in French (OFr. passe(re) -> passerei > passereau 'passerine', which designates all kinds of small birds); in a renewed way, piaf, which in the first half of this century meant 'sparrow' has extended its meaning to 'bird' in familiar French. Whereas the law of panchronic valuation expressed the relationships between extense and intense zones, this form of evolution expresses the relationships inside the extense zone: being established inside a same zone, they characterise themselves by the conservation of evaluation.

2.4. Force and form The description of taxemes as sets of static relations has indeed to its credit simplicity and economy: these relations can be characterized as logical relations, and this brought quite a success to descriptions through semantic features (actually used by defenders as well as by opponents of structural semantics). But the description of forms has no relevance if one cannot account for their evolution. Cognitive semantics acknowledges abstract

Cognitive semantics and diachronic semantics 125

forces, especially of course in its analyses of actancy when, following Talmy, it enacts a kind of transcendental mechanics which owes a lot to Aristotelian physics (see the concept of impetus in Talmy 1988). Forces and semantic forms must nevertheless be articulated. The morphodynamic description may account for forms in terms of forces. To say the least, these two aspects, form and force, are complementary: a force is felt and measured by the distortions it entails; a stabilized form results from an always momentary balance of forces. The effect of forces can be understood in two ways: the movement of critical points; the concomitant distortion of the "normal" sections of the form. The sections that exceed a threshold gain a potential of disturbance. From this point of view, the edges of the evaluative zones are privileged spots of disturbance. Beyond a certain level, peripheral disturbances modify the main basin of the taxeme.

2.5. The conditions of form evolution In morphodynamics, regular points and singular points are distinguished. As a form is better recognized by its singular points rather than its regular points, some of the relationships which are characterized, by analogy with perception, as relationships form/content can be described or reformulated as relationships between the regular sections of the form and its singular sections. For instance, at the textual level, we have described isotopies as products of the Gestaltist law of good continuation: they are as such regular portions of textual forms, and appear then as semantic backgrounds. On the contrary, allotopies are singular points and certain tropes introduce qualitative discontinuities by a disruption of isotopy.24 At the lexical level, the isotopie contexts of a sememe keep a regularity, whereas allotopic contexts create singularities which may sustainably disturb the basin of attraction of the sememe. This point must be qualified according to the typology of sememes: sparse with low associated values, sememes are compatible with a large number

126 François Rastier

of contexts. For instance the one of grammatical morphemes, which, as a general rule, does not present any generic feature of domain, can show a great deal of isotopie contexts - and this is the reason why the same grammatical morphemes are found in all discourses: hence the diachronic evolution of grammatical morphemes is slower than the one of lexemes.

2.6. Value and values Let's admit that forms (here semantic forms) are inhibited movements. As a normed set of evaluations, a doxa is precisely composed of prescriptions and inhibitions which ensure a synchronic or diachronic stability to semantic configurations. We retrieve here Barthes's intuition that the lexicon is a frozen doxa (1984: 129) but we pluralize it: frozen doxas. For instance in French there is no taxeme of levels of temperature: the sememes expressing levels of temperature are organized in different taxemes according to contexts. The same goes for sizes: 5 ft 10 is considered normal for a man, tall for a woman, and giant for a child. This size is hence respectively below and beyond the qualitative threshold, then beyond the acceptability threshold. Whereas cognitive semantics, after having reinvented the Kantian oversimplicity, is looking for the descriptive categories around transcendental aesthetics as a prior frame to any perception, we have used the term aesthetics in a more restricted meaning. Therefore the historic project of diachronic semantics leads us rather to look amongst social evaluations to find the forces which shape and distort the lexicon. Thus only the structural principles of semantic organization pertain to language; relevant categories pertain to the specific language considered; but the particular organization of its configurations pertains to evaluative norms which are subject to variations. The changes in evaluation resulting from a crossing of a qualitative threshold or mainly an acceptability threshold introduce modifications that can be both fast and long lasting. η ς

Cognitive semantics and diachronic semantics 127

We clearly understand that everything depends upon the position of the evaluative thresholds (whether they are qualitative or acceptability thresholds). According to the position of these thresholds, the structure of the taxeme changes; and this position does not depend on language but on doxa. Therefore, every stabilized semic formula results from an undefined series of contextual relations, i.e. from an interpretative tradition which varies with discourses and social practices. One cannot talk about doxas, as systems of valuation, without mentioning the social values which they reify. Without trying to play with words, one can wonder if the value (relevant linguistic difference) is not ultimately founded upon the values, in the social sense of the term, including the ethic and aesthetic judgements.26 The law of distribution and the absence of perfect synonyms enable us to outline a positive answer and to second it with examples: an icterus is more posh than a jaundice, ethylism (especially if it is social) than alcoholism, etc. In this case, interpreting an occurrence is not or is not anymore a simple operation of relating it to a type, but of positioning it in a taxeme, and of locating it in one of the evaluative zones of this taxeme.

3.

Implementation: the denominations of FACE in French

For our proposition of illustration, a study of the joint evolution of the words face and visage (figure) in French, we rely on Vaugelas (1647), F. Brunot (1905) and especially Renson (1962).27

3.1. The different steps of this evolution According to Renson (1962:1, 227) face is first attested in the 12th century with the meaning of 'visage' 'face'. In fact, it means essentially the surface of the face and in particular the cheeks, as it appears in contexts like cent foiz li baise de randon / Les ueuz la face et le menton 'one hundred times he kisses at random / eyes, face and

128 François Rastier

chin' (Roman de Thèbes, 6378). At the time, the plural faces with the meaning of 'cheeks' is attested.28 This acceptation justifies the abundance of colour adjectives (vernatile, palie, etc.). Yet, from the 13th to the 15th century, the number of contexts of face grows substantially, face meaning then, apart from the physical aspect of the figure, its expression (feature /moral/): for instance in Greban (doulce et tant prudente, benigne) or in Molinet (ayreuse et furibonde, humble etfortaccointable; cf. Renson 1962:1, 229). This evolution is general and also noticed for visage. It is evidence of the psychologization of literature (which in fact constitutes the best part of our documents). But, as soon as the 12th century, as has been stressed by J. Trenel, a new class of contexts appears with the use of face in the translations of the Bible (with the meaning of presence: la face de Dieu [which translates the Hebrew plural ρ 'nim Elohim] and also of surface: la face de la terre 'the face of the earth'). The face of God is without any precise features (see Renson 1962:1, 233). This new acceptation, translated literally from Hebrew, shows perhaps a relation with the spiritualization of the acceptation which describes the human figure: thus we find in Arnoul Greban adjectives like dampnee, digne and saintissime, or in Molinet angélique. On the opposite, visage is more rarely found in religious contexts, and in the assumption of a reciprocal reinforcement of neighbouring semes, it can be said then that face shows the afferent feature /religious/, brought to the forefront in certain contexts. It remains that in the 16th century face presents two acceptations, for God and for the human beings, the first being ameliorative and the second neutral, liable to physical as well as moral uses, and sharing with visage many common contexts. The configuration is then as follows: 'face 2' (/human/) and visage share a common general basin of attraction and are only separated by a low pass, whereas 'face l ' (/divine/) lies beyond.

Cognitive semantics and diachronic semantics

129

qualitative ¡threshold /physical/

M * fece 1 /divines'

/moral/

* fece 2 /human/ Figure 7. FT. face, end 15th c.

Yet, at the end of the century appears the expression face du grand Turc, which designates a totally different body part; thus Sieur Tabourot des Accords wrote : son mari qui était tout nud sur le lict, avait la face du grand Turc tournée de ce côté là 'her husband who was stark naked on the bed had the face of the Grand Turk turned that way' (Escraignes dijonnaises, 42 v°). Must we recall the establishment of diplomatic relations by François the First with the "Sublime Porte" and suppose that the wars between Christendom and Islam could well make of this expression the ludicrous and discreetly blasphematory reverse of the divine face? Whatever the case may be, the word face itself, according to Ferdinand Brunot and in spite of Vendryes's reluctance to admit that this could be the reason of it, finds itself marked with pej oration. In 1627, Mlle de Gournay feared that the "new critics" might decide suddenly to refuse to "escrire face, ... généralement refusée du nouveau jargon, parce qu'on dit la face du Grand Turc" [write face ... generally rejected from the new lingo, because one says the face of the Grand Turk] (L Ombre, 1627: 958, cit. in Livet, 1895: 303). Twenty years later, Vaugelas already emphasized that one "would not any longer dare say face for visage except for certain expressions" (1647: Préface, EX). Hence its sudden rarefaction according to Renson's counting: from 23% of the

130 François Rastier

designations of figure in the 16th century, it drops to 3% in the 17th 29

century. Thanks to the afference of the feature /pejorative/, the acceptation 'face 3' is now separated from visage by an acceptability threshold. Now visage covers the doxal zone, surrounded by two antipathetic and paradoxal acceptations of face, one religious and the other one infamous. This configuration is unstable, not by its form, but most certainly because 'face 1' and 'face 3' share the same signifier. I acceptability threshold

acceptabily I threshold

qualitative threshold

face 1 /divine/

face 2 /human/ Figure 8. Έτ./ace, end 16th c.

The reduction of the basin offace benefits then to figure, which, as soon as the middle of the 16th century, by a specialization of the meaning of 'exterior shape', was used for 'shape of the human face', and later, in the 17th century, was also used for the facial expression or look (Fr. mine, 1662), and came to replace visage and face in the current usage (see Brunot: 1905, s.v.).

Cognitive semantics and diachronic semantics 131

acceptability threshold 1

1

ι

•

•

I

acceptability threshold

qualitative threshold face 1 /divine/

visage

figure

face 3 /obscene/

Figure 9. Taxeme of FT. face (end 17th)

In today's French, figure has become a neutral and generic term, liable to the most diverse uses; face keeps its feature /pejorative/, as shown by the insulting phrase face de.... Finally, visage, reserved to the written language or to an elevated spoken style, has gained a feature /ameliorative/ (widely used in advertisements for cosmetic products). Visage keeps the feature /human/ but acquires a new acceptation (although already very common in Montaigne) as it is suitable for the aspect of very diverse objects (socialisme à visage humain, 1968), which anticipates perhaps, as formerly for 'face 1', its disappearance from the class of the denominations for human face.

132 François Rastier

Vi

face 1

visage

figure

face 3

Figure 10. 20th c.

3.2. Discussion 3.2.1. About the unity of the taxeme It can be objected that face in face du grand Turc belongs to another taxeme, the taxeme of the designations of the posterior. Yes, but 'face 2' remains a pejorative designation for visage: "Caesar believed that long and lean figures (visages) were real faces (faces) of conspirators" (Voltaire, letter to d'Argentai, February 1 Ith 1764).30 Furthermore, it could be objected that, strictly speaking, 'face 1 ' belongs to the religious domain. Fine, but the human visage is also designated there by face: the expression face-à-face (face to face) comes thus from religious texts describing the encounter of man and God. Moreover, expressions like face d'abbé (literally "face of a reverend") 'red and illuminated face' (Cotgrave, 1611; Oudin, c. 1640), face de carême 'pale and pallid face' (Panckoucke, 1749), are already attested in the 17th century in Bois-Robert (Epìtres) or Racine (Les Plaideurs). These two antithetic acceptations of face persist in religious and moral contexts. Littré places side by side face de réprouvé (literally 'face of outcast') 'sinister and frightening physio-

Cognitive semantics and diachronic semantics 133

gnomy', and avoir une face de prédestiné (literally 'having the face of a predestined') 'having a full, rosy and serene face'.31 To account for these opposed acceptations of face, the 17th century lexicographers used the concept of style. Richelet (1680), the first to relate the condemnation offace, said: "This word is still used in solemn and majestic poetry but not in gallant, lively poetry". Furetière (1690) claims on the contrary that "the word face is only used for figure in mockery for a face that is too thick or too large" ("le mot face pour visage ne se dit plus guère en ce sens qu'en raillerie d'un visage qui est trop gros ou trop large"). The Dictionnaire de l'Académie, in its second edition, will suggest a synthesis by distinguishing the "serious style when talking of God" and the "familiar style: une face réjouie, enluminée ('a joyful, illuminated face')". The theoretic problem raised by the notion of style should not be underestimated; in fact, we admit that a taxeme belongs to one and only one semantic domain. In so far as taxemes are classes of sememes which constitute a basis for term selection within a given practice, and where domains reflect different practices, the sememes of a given taxeme pertain to the same domain. In today's French, 'face 1' remains in religious texts, since, in spite of the aggiornamenti, the diachrony of religious discourses does not follow the same evolution as the other discourses grouped under the enigmatic name of langue générale [general language].32 This diachronic autonomy of a discourse is not surprising: in medicine, face still means 'visage', or more precisely 'the front part of the head'. More generally, it can be pointed out that the different discourses and the different social practices that they reflect are moving around differentiated diachronies: the evolution of a language, and in particular of a lexicon, obeys to very various types of temporality. The faces d'abbé and the faces de prédestinés, which looked very much alike by their florid complexion, have disappeared from today's French: no ameliorative contexts are found where face would mean 'visage'. Thus, in an expression like face d'ange 'angel's face', the pejorative feature of 'face' is propagated to 'angel' and not the ameliorative feature of 'angel' which contradicts the one of 'face': see e.g. Télérama about the trumpeter Chet Baker who died

134 François Rastier

of an overdose: "La mort donna des ailes à une face d'ange" [death gave wings to an angel's face]. The disappearance of the ameliorative acceptation o f f a c e , which appeared in the 12th century with the translations of the Bible, is perhaps an indication of a secularization of society, where the terms taken from the Scriptures are not any longer a pledge of what the Academy called a "serious style". It indicates also that /pejorative/ has become an inherent feature of 'face 2'.

3.2.2.From syntagms to phrases We must point out here a deficiency of the lexicographical approach: traditionally linked to an ontology of the concept and the isolated word, it deals only with simple lexical items and neglects complex lexical items. Yet, in the religious discourse, naturally prone to formulas through its links with the ritual, complex lexical items like face de Dieu are particularly important. In fact, for textual linguistics (and I believe that lexicology should be founded on such linguistics), the minimal semantic unit is the syntagm. Words are units deprived of their preferential contexts, decontextualized artefacts both of ontology and lexicography. In this perspective, simple lexical items are defined as strongly integrated syntagms (fact actually confirmed by linguistic evolution: each word is a fragment of a myth). Yet, the diachronic status of syntagms and complex lexical items are different: • A non integrated syntagm like face du grand Turc can be an initiator: semantic innovation can create a local disruption, which, amplified by other factors, will lead to a revision of the taxeme (creation or deletion of an attractor, or simple distortion of the global basin). • Integrated to a complex lexical item, like face-à-face, a simple lexical item does not retain the properties of its original taxeme: its sememe is redefined in a new taxeme and is not revised by the disruptions of the original taxeme. Thus, as Vaugelas pointed out:

Cognitive semantics and diachronic semantics 135

"Pour les personnes, on dit encore regarder en face, reprocher en face, soustenir en face, résister en face mais toujours sans l'article là" [for people, we still say regarder en face, reprocher en face, soustenir en face, résister en face but always without the definite article la] (1647: 60). And indeed, face-à-face has survived without any ameliorative feature (although it was drawn from translations of the Bible), nor subsequent pejorative feature; just like visà-vis, although vis owes probably its obsolescence to some form of ostracism if not some taboo (see below 3.2.3.).

3.2.3. The two networks of content and expression If the norms of the doxa may account for the evolution of the signifieds, they also participate in the evaluation of the signifiers, and the signs find themselves caught into two networks: (i) an onomasiological network which accounts for the evolution of the signs according to the evaluative variations of the signifieds; (ii) a semasiological network which takes into account the evaluative variations of the signifiers. The evolution of a sign obeys to these two networks of constraints. Yet, and this is worth being noticed, the same evaluative forces are at work in the onomasiological and semasiological networks: in the first case, we have seen that the proscription concerning the face of God has extended to the human face. For the second, let us consider the case of a prohibition concerning the signifier and let's take the example of vis which belonged in Old French to the designations of the visage. The dropping of the last consonant has transformed it into a homophone of vit ('prick'). The homophony with vit was obviously not lost on facetious minds.33 The opposition between noble and evil parts of the body was sufficient to inhibit vis in favour of visage?4 Thus the same kind of prohibition may contribute to inhibit a signifier (vis) or a signified ('face' 2). And in both cases, the controversial sign is eliminated from the taxeme.

136 François Rastier

4.

Conclusion

4.1. The historic contingence and the ideological conditions Let us imagine that the hilarious Sieur Tabouret, being or not the inventor of it, had never published this face du grand Turc, then the designations of visage might perhaps have been changed. Without returning to Plekhanov's conclusions on the role of the individual in history, it is enough to extend to languages the well known fact in the dialectology of small communities that a lexical innovation can always be attributed to one person. In the written languages, it is not rare that literature innovates and that its innovations are taken up in the oral language (see in Chinese the four character expressions). Nevertheless, for an innovation to be taken up, it must use salient semantic categories: it is useless to recall, at the end of the 16th century, the disturbance caused by the wars of religion, the expansion of scepticism and philosophical dissoluteness. During the following century, the censorship of the social, pious or academic good taste, contributed, as we know, to normalize the language and to put an end to semantic ambiguities: hence face was sacrificed on the altar of decency.

4.2. The anthropological foundation of diachronic evolutions We have seen that the evolution of taxemes can be described as a succession of distortions. Among the forces that are instrumental in it, ameliorations and pejorations play an antithetic role. The more a content is highly valued, the narrower its basin and the more it can be distorted. The zones neighbouring the acceptability thresholds are unstable, the pej oration of a very valorized term (as was the case for face) is enough to disturb a complete arrangement. Thus, the dogmatic value that was attached to face enabled to transform a new context into a blasphemy (as if faith was closer to blasphemy than indifference).

Cognitive semantics and diachronic semantics 137

More generally, as doxal contradictions correspond to social contradictions, a conflict between groups can express itself by a conflict between doxas, and a paradiastole is always possible (example: your heroes are murderers). We can nevertheless retain antithetic evolutions. Valuation is expressed through: an activation, the broadening of the attraction basin (and therefore of the contexts), the acquisition of a typical value, the use in an elevated style. On the opposite, tabooand prohibition are expressed by: the active inhibition of the singularity and of the related points, the narrowing of the attraction basin, the quantitative rarefaction of form, and often a euphemistical shift and the use of a sloppy style.35 One wonders then, in the continuation of what we said about value and values, if the founding of semantic oppositions does not lie in the doxal opposition between the valorized and the devalorized, and, ultimately, between the prohibited and the prescribed. If this is the case, semantics, be it cognitive or not, can only have an anthropological foundation, articulated upon ethnology and history.

Notes 1. 2.

3.

4. 5.

Cf. e.g. 1971: 164. Unless you believe, like Rosch following Berlin and Kay, that prototypes are founded by nature. Surprisingly, for these authors, nature is achronic (and yet, for instance, the mammoth is not any longer a prototype of big game animal). One can indeed wonder if her insistence on fuzzy membership does not come from an insufficient definition of the categories (which do not correspond to any lexical classes listed by linguists). In fact, nature is only a naturalization, i.e. a naturalization of doxa or the ideology of common sense. One knows that semasiology takes the signifier as invariant and considers the problem of polysemy as fundamental, whereas onomasiology starts from the signified and considers synonymy as prevalent. From the thesis that synonyms exist in any language, advanced by Abbé Girard (after Prodicos) and pursued by the synonymists of the 18th century, was born today's differential semantics (see n. 9).

138 François Rastier

6.

7.

8.

9. 10. 11.

12.

13. 14.

15.

16.

This ontological conception of meaning can be found in Guillaume with the theory of the potential signified, as in the theory of prototypes. The main discordance between Guillaume's approach and the cognitive approach lies in the fact that the potential signified is an abstract form, whereas for cognitivist semanticists (furthermore as opposed to Rosch), it is a better specimen. One knows that the transcendental theories of meaning, like Guillaume's or Culioli's which was born from it and, lastly, cognitive semantics, describe a time that is not the time of history, whereas diachronic semantics is a historical discipline. The concept of value explains as well the law of distribution, which Bréal, developing in a historical perspective the synonymists' findings defines as follows: "We call distribution the intentional order as a result of which words which should be synonymous and which were such in fact, have nevertheless taken different meanings and cannot any longer be used one for the other" (1897: 22). He concludes: "The history of language is a series of distributions" (1897: 22). With which the semic analysis bears incidentally some relation. See Bréal about "our fathers from the Condillac School" (1897: 277). And for us the reference is an effect, not a starting point (see Rastier 1991, eh. VII, 1994, ch. II). Anxious to found in nature the linguistic evolution, Sweetser makes her own the legend, long ago cleared up by Meillet and Benveniste, that the concrete meaning comes first in comparison to the abstract meanings. Metaphor and metonymy, emblems of the restricted rhetoric since Jakobson has strangely enough coupled them, are neither symmetric nor converse; and their choice is furthermore problematic: why exclude the other tropes? To the contrary and complementarity, as we will see (Fig. 1), bifteck came to mean 'food' by extension from the valorized form. This law has perhaps anthropological foundations; at least according to Louis Dumont, the hierarchy subordinates the includer to the included, the extense term to the intense term, the unmarked to the marked (see 1992: 1124). We consider language not as the reflection or the transposition of perceptive forms, but as an object of perception (hence our proposals for a theory of semantic perception): the link between language and perception lies here. We leave open the question of the incidence of semantic perception on the other forms of perception. The diachronic structural semantics and the cognitive semantics, in so far as they take into account general conditions of the capture of meaning and find probably a common foundation in Husserl's phenomenology (especially the concept of qualitative discontinuity presented in the third Recherches

Cognitive semantics and diachronic semantics 139

17. 18. 19.

20.

21.

22.

23. 24. 25.

logiques 1901), will perhaps give us a way to articulate the cognitive (as transcendental) and the cultural by taking into account cultural and historical conditions of cognition as languages force and articulate it. Due to lack of space, we refer you to his excellent presentation, 1997: 167210. And not the whole series of contexts associated to one expression (as Victorri and Fuchs [1996] did for instance in a semasiological perspective). Representational theories of meaning do not succeed in accounting for qualitative inequalities. Yet, distinguishing evaluative zones inside the taxeme gives us a means to break with the representational theory of meaning, since no metric enables us to distinguish the big from the huge and the cold from the icy. See Rastier 1996. Among the levels of acidity of a wine (weak, soft, cool, vivid, nervous, sour, green), the same level, measured in pH, and expressible in abstracto by nervous, will be categorised as vivid (non pejorative) for a young wine, and by sour (pejorative) for a wine that aged. In traditional semic analysis, for instance, the introduction of a sememe in a taxeme modifies the semic composition, i.e. the relational structure of the other sememes. The problem raised by the sememes which would mean one thing and its opposite (as the famous addad of the Arabic tradition) is in our opinion only raised in cases of antiphrasis or syllepsis, and therefore in context. Thus the Annuaire des rues de Paris (Directory of Parisian streets) contains not only streets but also boulevards, avenues, courts, etc. In this way, we reformulate, without linking it to any zero level, the problem of the deviation which preoccupies stylistics. The human universe is not made of knowledge on the one hand and emotions on the other hand. This omnipresent distinction, up to nowadays cognitive sciences, reiterates without any foundation the archaic separation of heart and reason. Undoubtedly, the neutrality of information is only a modern artefact which agrees with the persistent prejudice that language is a simple ideographic instrument for the use of rational thought. Let's agree that the human universe is made of social and individual appreciations, which are the object of fundamental aesthetics. Fundamental aesthetics pertains to linguistics when it takes as object the linguistic material itself. At the morphological level, all languages contain appraising morphemes (see for instance the affix -acci- in Italian). At the immediately superior level, the lexicon of languages swarms with evaluations, and acceptability thresholds structure the elementary lexical classes. All the more for phraseological units, very numerous in any text, which reflect and propagate a social doxa. At the level of the sentence, it can be considered that any predication is an evaluation. At the textual level at last, narrative analysis for instance

140 François Rastier

26.

27. 28.

29.

30.

31.

has underlined many times the importance of modalities qualified as thymic. To put it shortly, fundamental aesthetics defines the semiotic substratum on which arts of language are built, and remains well short of philosophical aesthetics. It has nothing in common with any aesthetic, poetic or stylistic function. This question may take Saussure as an authority when he defines two indissoluble aspects of value: the internal value, which draws its principle from differential semantics, and the external value, for which he gives the example of the coin. This counter value remains metaphorical, but the articulation of the two types of value raises a problem (see Piotrowski 1997): it is in our opinion the correlation between linguistic valuations and social values (of which the economical exchange values are only an isolated case, exemplary because normed). My thanks to Evelyne Bourion for having passed to me the elements of this document. Example: D'andeus ses oiz ses faces moille 'Both his eyes wet his face' (Benoit, Ducs de Normandie, 5114; six similar examples are found in this piece) to be compared with Plure de ses oils, si li moille sa face 'His eyes weep and soak his face' (Chanson de Guillaume, 478). One example: Corneille still uses it in the sense of visage in Médée (1635), but uses it only from then on in religious contexts. Vaugelas may find that "usage put it out of use" (1647: Préface, IX) for ridiculous, extravagant and very insulting reasons. He recognizes that "qu'en même temps que je condamne la raison pour laquelle on nous a osté ce mot dans cette signification, je ne laisse pas de m'en abstenir" [whereas I condemn the reason why this word has been taken away from us with this meaning, I can only abstain from using it]. The first edition of the Dictionnaire de l'Académie (1694), already behind usage, simply defines face by visage; but, if the fifth edition (1798) confirms this, it adds that "Dans le sérieux, il ne se dit en ce sens qu'en parlant de Dieu" [seriously, it is only said with this meaning when referring to God]. Even if the comparison with the posterior remains possible, in Zola for instance where the eaters in L'Assommoir "avaient des faces pareilles à des derrières" [had faces like bottoms] (ch. VII). We make sure to distinguish the metaphor from its expression: it would have been preferable to associate the word visage to a comparison of the face and the bottom. Renson quotes l'autre visage in Voiture and in various authors luckily forgotten visage sans nez 'face without nose', gros visage 'big face', visage à rendre un lavement 'face to bring up a lavement' (1962: I, 210). But these occurrences remain isolated and were not adopted by usage. The opposition of these two expressions can already be found in the Dictionnaire de l'Académie (1835). Victor Hugo unites brilliantly the two ac-

Cognitive semantics and diachronic semantics 141

32.

33.

34.

35.

ceptations: "Le voyage qu'ils [mes parents morts] font est profond et sans bornes, / On le fait à pas lents, parmi des faces mornes, / Et nous le ferons tous" (Hugo, Feuilles d'automne, 6). In this context, 'face 2' appears in the religious domain (a funeral). The evaluative norms belonging to this kind of discourse have saved face from infamy. Using an understatement, the Jesuit authors of the Dictionnaire de Trévoux (1721) notes on expressions containing the word face that "these phrases {face-à-face, etc.) are imitated from the Scriptures ... This makes the use of certain terms more acceptable". The discourse of Law is probably not subject to taboo. At least, the taxeme of designations of the face does not know the same diachrony in the religious discourse than in others. One of Molinet's ballads, by and large obscene in view of its rhymes obligingly itemised in the manuscript for absent-minded readers: "Madame, j'ai sentu les façons / Du feu d'amour, puisque je vis / Les yeux plus aspres que faucons / De vostre gent et plaisant vis" (cf. Dupire 1936: 866-867). The competition of visage, suggested by Renson, is all the more unconvincing since visage is derived from vis; the monosyllabic character of vis has also been put forward, but we can find monosyllabic words which are quite durable. Like face, vis finds itself banished because visage is traditionally opposed to parts of the body considered as evil. This theme, as we know, has been developed by Freud, who takes up Schopenhauer's polarization ("the head and the genitals are, so to speak, the opposed poles of the individual", excerpt from Le monde, in Insultes, Monaco, Le Rocher, 1988, p. 29). Euphemization has certainly an anthropological generality: for instance the name of the living body is extended to the dead body (the word cadavre 'corpse' is not used during a funeral). The use of a neutral term like corps 'body' (ex. levée du corps 'funeral' is not limited to an extension: we thus avoid the pejoration of the dead body).

References Barthes, Roland 1984 Le bruissement de la langue. (Essais critiques 4.) Paris: Seuil. Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. (Beihefte zur Zeitschrift für Romanische Philologie 285.) Tübingen: Niemeyer. Berlin, Brent and Paul Kay 1969 Basic Color Terms. Berkeley: University of California Press.

142 François Rastier

Bréal, Michel 1897 Essai de sémantique. Paris: Hachette [new edition Brionie: Montfort, 1983]. Brunot, Ferdinand 1905-1937 Histoire de la langue française des origines à 1900, 12 Volumes. Paris: Armand Colin. Chamfort, Sébastien-Roch Nicolas 1968 Maximes et pensées, caractères et anecdotes. Edited by Jean Dagen. Paris: Flammarion. Clédat, Louis 1895 Les lois de la dérivation des sens appliquées au français. Revue de Philologie 9: 49-55. Coseriu, Eugenio 1964 Pour une sémantique diachronique structurale. Travaux de linguistique et de littérature II/l : 139-186. 1968 Les structures lexématiques. In: W. Theodor Elwert (ed.), Probleme der Semantik, 3-16. (Beihefte zur Zeitschrift für französische Sprache und Literatur, N.F. 1.) Wiesbaden: Steiner. Darmesteter, Emile 1877 De la création actuelle de mots nouveaux dans la langue française et des lois qui la régissent. Paris: Vieweg. 1887 La vie des mots étudiée dans leurs significations. Paris: Delagrave [new edition Paris: Champ Libre, 1979]. Dumont, René 1992 Anthropologie, totalité et hiérarchie. In: Anthropologie et philosophie. Paris: Centre Pompidou, 11-24. Dumarsais, César C. [1730] 1988 Traité des tropes. Edited by Françoise Douay-Soublin. Paris: Flammarion. Dupire, N. (ed.) 1936 Les Faictz et dictz de Jean Molinet, 3 Volumes. Paris: Picard. Hjelmslev, Louis 1971 Essais linguistiques. Paris: Presses Universitaires de France. Husserl, Edmund [1901] 1963 Recherches logiques. Volume 3. Eléments d'une élucidation phénoménologique de la connaissance. Paris: Presses Universitaires de France. Katz, Jerrold J. and Jerry A. Fodor 1963 The structure of a semantic theory. Language 38: 170-210. Koch, Peter 1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46.

Cognitive semantics and diachronic semantics 143

LakofF, George 1987 Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. Chicago/London: University of Chicago Press. Langacker, Ronald 1986 An introduction to cognitive grammar. Cognitive Science X/l : 1-40. Livet, Charles-Louis 1895-1897 Lexique de la langue de Molière comparée à celle des écrivains de son temps. 3 Volumes. Paris: Imprimerie Nationale. Martin, Robert 2 1992 Pour une logique du sens. Paris: Presses Universitaires de France. Nyckees, Vincent 1997 Pour une archéologie du sens figuré. Langue française 113: 49-65. Petitot, Jean 1983 Les catastrophes de la parole. Paris: Maloine. 1985 Morphogenèse du sens. Paris: Presses Universitaires de France. 1996 Les modèles morphodynamiques en perception visuelle. Visio I: 6573. Piotrowski, David 1997 Dynamiques et structures en langue. Paris: Editions du CNRS. Pottier, Bernard 1974 Linguistique générale. Théorie et description. (Initiation à la linguistique B, 3.) Paris: Klincksieck. Rastier, François 1987 Sémantique interprétative. Paris: Presses Universitaires de France. 1991 Sémantique et recherches cognitives. Paris: Presses Universitaires de France. 1994 Sémantique pour l'analyse. With collaboration of Marc Cavazza and Anne Abeillé. Paris: Masson. 1996 Chamfort: le sens du paradoxe. In: Ronald Landheer and Paul J. Smith (eds.), Le paradoxe en linguistique et en littérature, 119-143. Genève: Droz. Reisig, Karl 1839 Vorlesungen über lateinische Sprachwissenschaft. Leipzig: Lehnhold. Renson, Jean 1962 Les dénominations du visage en français et dans les autres langues romanes. Paris: Les Belles Lettres. Rosch, Eleanor 1978 Principles of categorization. In: Eleanor Rosch and Barbara Lloyd (eds.), Categorization and Cognition, 27-48. Hillsdale: Erlbaum.

144 François Rastier

Sweetser, Eve 1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press. Tabouret des Accords, Jehan 1922 Escraignes dijonnaises (after Cioranesco XVIth). Dijon. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12: 49-100. Vaugelas, Claude Faivre de [1647] 1934 Remarques sur la langue française. Edited by Jeanne Streicher. Paris: Droz. Victorri, Bernard and Catherine Fuchs 1996 La polysémie. Paris: Hermès.

Section II Descriptive categories

Losing control: grammaticization, subjectification, and transparency Ronald W. Langacker

My purpose is to explore a common type of semantic change involving attenuation in the degree of control exerted by an agentive subject. Attenuation figures crucially in many cases of grammaticization. When carried to extremes, it results in subjectification (Langacker 1990b) and in the transparency of highly grammaticized forms (Langacker 1995a).

1.

Background

As background, I must first introduce a few basic notions of cognitive grammar (Langacker 1987a, 1990a, 1991). Its central claim is that grammar is symbolic in nature, forming a continuum with lexicon and being fully describable as assemblies of symbolic structures (form-meaning pairings). The meaningfulness of grammar becomes evident given an independently motivated conceptualist semantics which properly accommodates our multifaceted capacity to conceive and portray the same situation in alternate ways. An expression's meaning is a function of both the content it evokes and the particular construal it imposes on that content. One aspect of construal is the level of specificity (conversely, schematicity) at which a situation is characterized, e.g. in describing someone as a person, a scholar, a linguist, or a Romance linguist specializing in diachronic semantics. Among the numerous other aspects of construal are scope, subjectivity, and various kinds of prominence. An expression's overall scope is the full range of conceptual content it evokes as the basis for its meaning. Within this, there is a lim-

148 Ronald W. Langacker

ited range - called the immediate scope - describable as the general locus of attention (the "onstage region"). For example, the overall scope of elbow is the conception of the human body, and its immediate scope is the conception of an arm, as seen in Figure 1(a). Heavy lines indicate profiling, one kind of prominence. An expression's profile is the specific focus of attention within its immediate scope. It can also be described as the entity the expression designates, i.e. its referent within the conception it evokes. The profiled entity can either be a thing or a relationship (assuming very general definitions of these terms). As shown in Figure 1(b), for instance, across profiles a relationship such that one participant successively occupies all the points along a spatial path traversing the other. (We see in this example that there need be no distinction between immediate scope and overall scope, just as there need be none between the nucleus of a syllable and the syllable as a whole.) An expression's profile determines its grammatical class: a noun profiles a thing, whereas verbs, adjectives, adverbs, prepositions, participles, and infinitives profile relationships of various kinds. (a)

elbow

overall

e

(b)

across overall scope

/ /

A\ Figure 1. Scope and profiling

The conception of a relationship presupposes that of its participants, which are accorded varying degrees offocal prominence. It is useful to think of this metaphorically in terms of primary and secondary spotlights which can be directed at various elements of the

Losing control: grammaticization, subjectification, and transparency 149

scene onstage. The primary focal participant is termed the trajector (tr). When there is a secondary focal participant, it is called a landmark (lm). In a construction (a larger symbolic assembly), an expression which specifies a relational element's trajector is thereby identified as its subject, and one which specifies its landmark is an object (if it profiles a thing) or a non-subject complement. As a fully general characterization, therefore, subject and object status are claimed to reside in the focal prominence directed at onstage elements, rather than being defined in terms of any particular semantic role. As primary focal participant, the trajector is the initial point of access - the "starting point" - in conceptualizing the profiled relationship (MacWhinney 1977; Gernsbacher and Hargreaves 1992; Chafe 1994; Tomlin 1995; Langacker 1997,in press). I must next explain what is meant by subjectivity and subjectification (Langacker 1985, 1990b, 1995b). As I myself use these terms, the crucial factor is vantage point, and in particular the relative positions of the subject and object of conception (which are of course to be distinguished from the grammatical subject and object). The subject of conception is an implicit locus of consciousness ("perspective point") which apprehends the object and - if it is only the subject is not itself apprehended. The subject's activity determines the nature of the conceptual experience but does not per se figure in its "content". To the extent that an entity functions as the subject or object of conception, it is said to be subjectively or objectively construed. It is construed with maximal subjectivity when it remains offstage and implicit, inhering in the very process of conception without being its target. It is construed with maximal objectivity when it is put onstage as an explicit focus of attention. Thus profiling and trajector/landmark status imply a highly objective construal. To forestall possible confusion, I should note that Traugott also speaks of subjectivity and subjectification (cf. also: Traugott, this volume). For her, the latter is "a very general tendency toward greater pragmaticization of meaning" (1986: 540). It includes at least three overlapping kinds of changes: (i) the shift from externally based descriptions to internally grounded assessments, e.g. boor 'farmer' > 'crude person'; (ii) extension to textual and metalinguistic

150 Ronald W. Langacker

uses, as when observe develops the meaning 'state'; and (iii) increased involvement of speaker judgment, as in while 'during' > 'although' (1989: 34-35). In Traugott's inclusive sense, subjectification may just affect the domain in which a property or relationship is manifested. For me, however, it is a matter of perspective. Moreover, only a particular entity (not an expression overall) is said to be subjectively or objectively construed. Whether boor means 'farmer' or 'crude person', for example, its profiled referent is the onstage focus of attention, hence objectively construed, whereas the speaker remains an implicit locus of judgment and is thus construed subjectively. This is not to deny that the speaker's attitude becomes more evident in the later meaning, the primary content being evaluative, but the speaker carries out this evaluation without becoming the focused target of conception. There is of course no point in trying to decide between Traugott's version of subjectivity and my own, since both will figure in an overall account of grammaticization. The issue is purely terminological. The speaker (S) and hearer (H) function as conceptualizes, i.e. subjects of conception, with respect to the meanings of linguistic expressions. They are normally offstage and subjectively construed. With a simple noun like dog, for instance, they have only a conceptualizing role - indicated by the dashed arrows in Figure 2(a) - and thus remain external to its scope. A deictic element such as an article or demonstrative makes a specification concerning the speaker's and hearer's apprehension of the profiled entity, thus bringing them within the expression's overall scope. Nonetheless, being referred to only tacitly, they are offstage and subjectively construed, as shown for the dog in Figure 2(b). It is however possible for the speaker and hearer to go onstage and even be profiled as a focused object of conception, notably in the case of pronouns like I, you, and we, as seen in 2(c).

Losing control: grammaticization, subjectification, and transparency 151

a) dog overall scope immediate scope

b) the dog overall scope immediate scope I

Λ

1

(c) we overall scope"

t

Figure 2. Positions of the conceptualizes in relation to scope

2.

Subjectification

In previous works (1990b, 1991), I characterized subjectification in terms of replacement: some relationship within the objective situation under description is replaced by a comparable but subjectively construed relationship inherent in the very process of conception. However, as suggested by Verhagen (1995) and Harder (1996: 352), I have come to believe that this subjective component is there all along, being immanent in the objective conception, and simply remains behind when the latter fades away. This revised notion of subjectification is represented abstractly in Figure 3.

152 Ronald W. Langacker

Initial Configuration overall scope

Attenuation overall scope

Subjectification overall scope immediate scope tr Im

>

>

Q»P

Figure 3. Attenuation in degree of subject control

é

In the initial configuration, the solid arrow in bold depicts an objectively construed, profiled relationship. The direction of the arrow indicates some kind of objective asymmetry which motivates the choice of trajector; the trajector is in some sense more active - typically as an agent, experiencer, or mover. The solid arrow labeled Τ stands for processing time. The import of the diagram is that the conceptualizer, C, does some kind of mental scanning from the trajector to the landmark in conceiving of the profiled relationship (i.e. the trajector functions as a "starting point" or initial point of access in conceptualizing it). The dashed arrows represent the conceptualizer's mental activity, carried out through processing time. In the second diagram, the bold arrow representing the profiled relationship is given in dashed lines to indicate attenuation, involving both the objectively conceived relationship and the trajector's role in it. Although the conceptualizer continues to carry out the same or a comparable mental scanning as in the initial configuration, the objective situation now offers less motivation for it. We will be investigating the progressively more tenuous objective basis for both the relationship and for the trajector's selection as primary focal participant. Finally, the last diagram indicates the full disappearance of any objective basis for the conceptualizer's mental scanning. A relationship is still established between the trajectorand landmark, but the

Losing control: grammaíicization, subjectifìcation, and transparency 153

basis for it now resides exclusively in the conceptualized activity. This relationship is subjectively construed because it inheres in the process of conceptualization itself, rather than being an onstage object of conception. To the extent that the same mental operations figured in the initial configuration, we can say that the subjective relationship was immanent in the objective one. Observe that subjectifìcation per se need not have any effect on the choice of focal participants (trajector/landmark alignment). It merely removes any objective basis for selecting the trajector as initial point of access. A first example is the semantic extension relating the two senses of across exemplified in (1): (1)

a. The child hurried across the street. b. There is a mailbox right across the street.

In (l)a, the trajector of across (instantiated by the child) has that status by virtue of being a mover, successively occupying all of the points along a spatial path traversing the static landmark. This is shown in the first diagram of Figure 4. In (l)b, on the other hand, the trajector of across - the mailbox - is static, as seen in the second diagram; it occupies only a single position vis-à-vis the landmark (equivalent to the final position in (l)a). This illustrates attenuation with respect to both the objective relationship and the basis for the choice of trajector. The profiled relationship is less inclusive and less dynamic in this second sense of across, and the trajector does not stand out as primary focal participant by virtue of moving. In conceiving of the trajector following the objective spatial path in (l)a, the conceptualizer necessarily scans mentally along the same path as an inherent aspect of tracking the subject's motion. Although objective motion is lacking in the derived sense of across, we see from the second diagram in Figure 4 that the conceptualizer nonetheless follows the same path subjectively in locating the trajector with respect to a reference point (R). Once the objective motion is stripped away, the subjective mental scanning becomes more apparent. Observe that this fading away of the objective motion does not inherently involve any change in trajector/landmark alignment; only

154 Ronald W. Langacker

the extent of the trajector's role is modified. Instead of successively occupying all of the locations constituting a spatial path, it occupies only the final location. across J overall scope

across |

overall scope

immediate scope

immediate scope Im

J

—

^

tr

tr

[«'ΚΊΊΙΙ'»] ΗΙϋ&βΛιβ· 1

\ \ . i / /

τ

W

>

1 \\

'/

>

è Figure 4. An example of subjectification

Although I consider this to be a case of subjectification, I have left a dashed arrow to show that an objective basis for the conceptualizer's scanning along the spatial path has not disappeared entirely; usually some vestige of it remains. It might appear that the transition connecting the two senses of across would have to be discrete: the trajector either traverses all the locations constituting the spatial path, or occupies only the final location, but there is no intermediate sense in which, say, it traverses only a portion of the path. But the extent of the motion is only one dimension of possible variation other dimensions involve the status of the motion as well as who does the moving. Differences along these parameters provide a series of cases that are intermediate between that of profiled, objective motion by the trajector, and at the other extreme, that of purely subjective motion by the conceptualizer imposed with no objective basis. Such a series is given in (2).

Losing control: grammaticization, subjectification, and transparency 155

(2)

a. The child hurried across the busy street, [profiled objective movement by trajector] b. The child is safely across the street, [static location resulting from unprofiled, past, actual movement of trajector] c. You need to mail a letter? There's a mailbox just across the street, [static location as goal of unprofiled, potential, future movement of addressee] d. A number of shops are conveniently located just across the street, [static location as goal of potential movement by a generalized or generic individual] e. Last night there was a fire across the street, [static location, no physical movement necessarily envisaged at all]

Only in (2)a is movement along the entire spatial path put in profile. This constitutes an important and discrete difference: since the other sentences profile only a single, static configuration, they are imperfective (or "stative") and take the verb be. There nevertheless remains a vestige of objectively construed motion. It undergoes a change in status from profiled to unprofiled, and from actual to potential to generic. The mover also changes, from being an objectively construed participant (the trajector) to being subjectively construed (the unmentioned addressee), and from a specific mover to a generalized or unspecified one. It is only in the last example, (2)e, that the conception of physical movement may be entirely absent, leaving only subjective motion by the conceptualizer, who mentally traces along the path in order to specify the trajector's location. Thus I do not envisage attenuation and eventual full subjectification as occurring in a single step. It is more likely a gradual evolutionary process involving small steps along a number of possible parameters. This will usually result in the coexistence of alternative values at a given diachronic stage, with gradual shifts in preference being responsible for changes that in retrospect appear to be discrete (cf. Heine 1992). Attenuation can be observed with respect to at least four parameters (the grouping is somewhat arbitrary). We have already noted change in status: from actual to potential, or from specific to generic.

156 Ronald W. Langacker

A second parameter is change in focus, i.e. the extent to which particular elements stand out as focus of attention, notably in terms of profiling. We saw that actual motion by the trajector is profiled in (2)a, but unprofiled in (2)b, which designates only the final locative configuration resulting from that motion. Full elimination, exemplified by the absence of any objective movement in (2)e, might be thought of as the extreme case of defocusing. A third kind of attenuation is a shift in domain, e.g. from a physical interaction to a social or experiential one, as in the evolution of modals (Sweetser 1982, 1990). The final parameter is change in the locus of activity or potency. This is illustrated in (2) by the change in mover: from a focused onstage participant (the trajector) to an offstage one (the addressee), or from a specific mover to a non-specific, generalized one.

3.

Grammaticization

Attenuation figures in many cases of grammaticization. Let us start with the common evolution of a verb meaning 'go' into a marker of futurity, as with the English be going to construction. We can first observe that a sentence like (3) is ambiguous. It may indicate actual movement through space by the subject, in order to initiate an action at the endpoint of the spatial path. It can also indicate the futurity of the infinitival event, with no implication of spatial motion; in this case I posit subjective movement through time by the conceptualizer. The two senses of be going to are sketched in Figure 5. (3)

Sam is going to mail the letter, [physical, objective movement through space by the subject OR subjective movement through time by the conceptualizer]

Losing control: grammaticization, subjectification, and transparency 157

be going to' overall scope

be going to overall scope

immediate scope

immediate scope space

Im

Im tr

71ÚL \ 1 I

I I \ Ml I W i l l Will

i r -w-

è

> fonftp-» V i\ i ι ιI * ι ι / / \ \ I ι I Λ W i l l

\ÏÏ77~ -W—

è

Figure 5. Subjectification as part of grammaticization

In the initial sense, the trajector follows a spatial path, at the end of which he intends to initiate some activity, which constitutes a relational landmark. (This landmark is specified by the infinitival complement, and the trajector by the subject noun phrase, at higher levels of grammatical organization.) The movement of course takes place through time (the arrow labeled t). In the future sense of be going to, the conceptualizer traces a mental path along the temporal axis and situates the infinitival event downstream in the flow of time relative to some reference point. As we saw before with across, there is no change in trajector. The trajector does however have a diminished role in the profiled relationship: since it no longer moves through space, its activity is confined to whatever it does in the landmark event. While this may seem peculiar, it is actually quite common (see Langacker 1995a). Moreover, it is unproblematic in cognitive grammar because the trajector is characterized as primary focal participant, not in terms of any particular semantic role.

158 Ronald W. Langacker

This clearly qualifies as subjectification as defined above. In following the subject's movement through space, which unfolds through time, the conceptualizer implicitly traces a mental path through time - the same path which stands alone as the profiled temporal configuration in the future sense of be going to. The conceptualizer's subjective motion through time is immanent in the conception of the subject's objective motion through space, and remains behind when the latter fades away. Once more, the heavy dashed arrow in the second diagram of Figure 5 indicates that the relationship is not purely subjective; some vestiges remain of the original objective basis for the conceptualizer's mental path. We can discern a number of intermediate stages: (4)

a. Sam was going to mail the letter - but he never reached the post office. b. Sam was going to mail the letter - but he never got around to it. c. If Sam isn't careful he's going to fall off that ladder. d. Something bad is going to happen -1 just know it. e. It's going to be summer before we know it.

In the physical motion sense of (4)a, the subject does not just move but also has the intention to carry out the infinitival event at the end of the spatial path. This intention perseveres in many nonmotion uses, as seen in (4)b. The trajector thus continues to be a locus of activity tending toward realization of the infinitival process. The activity is however attenuated by virtue of having lost its physical component; only its mental aspect remains. We can regard this as either a change in focus (an extreme case, resulting in the full absence of physical motion) or else a change in domain (from physical to mental/experiential). Intention is a sort of potency directed toward realizing the envisaged event. One prevalent kind of attenuation involves progressive diffusion in the locus of potency. We see this in examples like (4)c, where the future event is conceived as being accidental; the subject does not act with the specific intent of bringing it about. He is none-

Losing control: grammaticization, subjectification, and transparency 159

theless still plausibly attributed a certain amount of responsibility for the falling. Given the overall circumstances in which the subject finds himself, his other actions and/or his inattentiveness create the potential and even the likelihood of the event being realized. In short, the locus of potency is no longer concentrated and specifically identified with the subject, but is much more diffuse. It resides in a combination of factors merely associated with the subject (e.g. degree of attentiveness) and the external circumstances in which the subject's activity takes place. Naturally, the specifics vary greatly from one example to the next. In (4)d, the subject is no longer even a locus of responsibility; in fact, the subject does not exist at the reference time (R) and is incapable of responsibility. It is rather just some aspect of the present circumstances that induces in the conceptualizer an expectation of the future event. The source of this expectation may be quite diffuse, a matter of speaker assessment that cannot be tied to any specific objective elements. Finally, (4)e approximates the limiting case of a purely temporal relationship. At the extreme, the profiled relationship resides solely in the conceptualizer temporally locating the situation by mentally scanning forward through time from the reference point. I note in passing that this account is not based on metaphor. Contrary to my own previous discussion of be going to (1986), as well as Sweetser's (1988), there is no transfer from the spatial to the temporal domain, but merely the retention of a temporal relationship that was there all along. Moreover, it is specifically not claimed that the subject of be going to is metaphorically construed as moving along a temporal path (analogous to a spatial path) - only the conceptualizer is claimed to move along a temporal path, mentally and subjectively.

4.

Transparency

Towards the end of this grammaticization process - in cases like (4)d-e where the subject no longer has any role in bringing about the infinitival event - we have the situation I call transparency: anything

160 Ronald W. Langacker

eligible to be the subject of the infinitival complement is also eligible to be the subject of the entire expression, as seen by the socalled "dummy" or "expletive" subjects in (5). This is the kind of situation that was dealt with in transformational grammar by positing a rule of Subject-to-Subject Raising. (5)

a. There is going to be another storm tonight. b. Tabs are going to be kept on all the dissidents.

I have argued elsewhere (1995a) that raising rules are superfluous in a cognitive grammar description. What we find in this construction, and in others to be examined, is progressive attenuation and diffusion in the locus of control, as seen in (4). Such attenuation does not per se have any effect on the choice of trajector, but only on the extent of the objective motivation for this choice. Beyond a certain point, therefore, the trajector of the profiled relationship no longer has any role in effecting that relationship, so it need not have any particular properties (as it does when it is still construed as a mover or as having intention or responsibility). What the trajector actually does, in such extreme cases, is limited to whatever is implied by virtue of its status as trajector of the process temporally downstream from the reference point. Hence all restrictions on its choice are determined by the infinitival complement. I reiterate that in cognitive grammar it is unproblematic for the trajector of a relationship to have no direct role in it - trajector status is characterized in terms of primary focal prominence, and the "spotlight" of focal prominence can in principle be directed wherever desired. I should note that progressive diffusion in the locus of potency is nothing out of the ordinary. We can observe it in the normal range of variation permitted with virtually any agentive verb in regard to the specific role of the subject. Consider the examples in (6), where the nature of the subject's involvement is described in brackets: (6)

a. Edward frightened the hikers by jumping out of the bushes and shouting at them, [source of volitional physical action]

Losing control: grammaticization, subjectiflcation, and transparency 161

b. Edward frightened the other hunters by accidentally firing his rifle, [source of non-volitional physical action] c. Edward frightened the priest by believing in satan. [locus of mental attitude] d. Edward frightened the children by being so ugly, [locus of property reacted to by others] e. Edward frightened his parents by not being among the children getting off the bus. [mere association with circumstance reacted to by others] Edward has a fully agentive role in (6)a, but in (6)b his agentivity is attenuated through absence of volitionality. In (6)c Edward's participation shifts from the physical to the mental domain, and changes from a temporally bounded action to a steady-state attitude. Also, the person being frightened now has a greater proportion of the responsibility for the fright being induced; holding a belief is not frightening per se, but only if someone finds its content objectionable. In (6)d-e the subject's responsibility diminishes still further, while the object's increases. In (6)d, Edward is merely the passive and unwilling locus of a static property. And in (6)e, even that is lacking - it is only due to someone else's expectations that Edward is even associated with the fright-inducing circumstance. This is not yet a case of full transparency. One cannot, for example, say (7): (7)

*Tabs frightened civil libertarians by being kept on all the dissidents.

Thus even in (6)e there is some tenuous respect in which Edward is still held responsible for inducing the fright. But it may be quite tenuous indeed. This illustrates the great flexibility we usually have in construing agentivity or responsibility; the individual put in focus as trajector or subject may in fact be only metonymically related to the actual locus of potency (Langacker 1995a). The full transparency characteristic of highly grammaticized forms (like be going to, as

162 Ronald W. Langacker

seen in (5)) merely carries the diffusion of responsibility to its ultimate conclusion. The verb have is an especially interesting example but can be mentioned here only in passing (cf. Brugman 1988). As a main verb taking NP complements, have shows a broad range of uses with varying degrees of attenuation in subject control: (8)

a. Be careful - he has a knife! [source of immediate physical control] b. I have an electric saw (but I seldom use it), [source of potential physical control] c. They have a good income from investments, [locus of experience, abstract control] d. They have three children, [locus of social interaction, generalized responsibility] e. He has terrible migraine headaches, [passive locus of experience] f. We have some vast open areas in the United States, [locational reference point, diffuse locus of potential experience]

Yet some vestige of subject involvement always remains. Even in examples like (8)f (analyzed in Langacker 1993), where a generalized subject functions mainly as a locative reference point, there is still a vague notion of possible experiential consequences. Hence the main verb have is non-transparent. The auxiliary verb have in the perfect construction is fully transparent, however: (9)

a. There may have been a serious breach of security. b. Tabs should have been kept on those dissidents all along.

This construction represents a considerably more advanced stage of grammaticization and no longer implies any necessary subject involvement. (For detailed analysis, see Langacker 1990b, 1991: 211225; Carey 1994.)

Losing control: grammaticization, subjectification, and transparency 163

5.

Modals

The English modals are likewise highly grammaticized and for the most part transparent, as seen in (9). They have evolved from main verbs with meanings like 'want', 'know how to', 'have the strength to', etc. Observe that such verbs have two crucial properties both reflected in grammaticized modals: they are force-dynamic (Sweetser 1982, 1990; Talmy 1988), and the action serving as target of the force vector, i.e. the event expressed by the verb's complement, remains potential rather than being actual. The profiled relationship involves some kind of effectiveness or potency tending toward realization of the type of action expressed by the complement, but no actual instantiation of that action is implied. I cannot go into the details of either the synchronic analysis of the modals or their diachronic evolution (see Langacker 1990b, 1991: 269-281). Here I will simply note that their development illustrates the attenuation of subject control, in that the locus of potency is no longer identified with the subject. Consider first the root or deontic interpretations of modals. As noted by Talmy and Sweetser, root modals generally convey forcedynamic relationships in the domain of social interaction. This shift from physical to social force constitutes attenuation in regard to domain. Moreover, the source of potency is no longer identified with the subject, but is implicit and subjectively construed. It may be the speaker but need not be, as seen in (10). It is not necessarily any specific individual, but may instead be some nebulous, generalized authority. In other words, the source of potency is highly diffuse. (10) a. You may not see that movie -1 won't allow it! b. You must go home right away -your wife insists! c. Passengers should arrive at the airport two hours before their flight. d. You must not covet your neighbor's wife. Nor is the subject necessarily the target of the potency, which is also diffuse. Although the modal force may be directed at a specific

164 Ronald W. Langacker

individual - be it the subject, the addressee, or some third party - we see from (11) that this is not always the case. The force is simply directed toward realization of the target event, to be apprehended by anyone who might be in a position to respond to it. Because the subject is usually not the source of potency, and need not be its target, the root modals exhibit transparency. (11) a. The next patient can come in now. b. This fence must be painted by tomorrow - you had better get busy. c. This fence must be painted by tomorrow - you had better tell Harry. d. There may not be any alcohol served at the party. e. Tabs must be kept on those dissidents! What about epistemic modals, as in (12)? (12) a. They should be able to find what they need. b. Tabs will probably be kept on all the dissidents. c. There may be some rain tonight. Epistemic modals are maximally diffuse in regard to the source and target of potency, hence transparent. I have described their potency as inhering in the evolutionary momentum of reality itself, as assessed by the speaker/conceptualizer: given how reality has been evolving up through the present, what is the likelihood of it continuing to evolve in such a way as to "reach" the target process? This is basically equivalent to saying that the conceptualizer carries out a mental extrapolation of ongoing reality, projecting into the future, and senses the degree of force impelling this mental extrapolation in the envisaged direction, or the degree of resistance encountered in projecting it through to the target. In other words, the force dynamics are inherent in the conceptualizer's mental activity, hence subjectively construed in a strong sense.

Losing control: grammaticization, subjectification, and transparency 165

6.

Be-Auxiliaries

Let us next consider the frequent path of grammaticization whereby verbs of motion or posture evolve into auxiliary verbs roughly glossed as 'be'. A case in point is Spanish estar, whose etymological value was 'stand'. Like English be, it takes various types of relational expressions as complements, including adjectives, prepositional phrases, and active participles: (13) a. Está enfermo. 'He is ill.' b. Está en la cocina. 'He is in the kitchen.' c. Está trabajando. 'He is working.' What exactly is the meaning of an auxiliary verb like Spanish estar or English be? Elsewhere (1987a, 1987b) I have argued that the characteristic property of a verb or a finite clause - the property that distinguishes them from non-verbal relations like adjectives, prepositions, and participles - is a subjective factor, namely whether the conceptualizer follows the profiled relationship's temporal evolution sequentially or construes it in summary or holistic fashion. A be-type auxiliary verb embodies this subjective factor, but does not specify any particular relationship for its profile, i.e. the objectively construed relationship it follows through time is highly schematic. Betype verbs can thus be used in constructions like those in (13), in which they combine with a ήοη-verbal element to form the complex head of a finite clause. When the schematic but sequentially viewed relationship profiled by the auxiliary is equated with the specific relationship profiled by the non-verbal complement, it lends to the complement the sequential viewing required for the head of a finite clause. The historical evolution leading from a motion or posture verb to a ¿>e-type auxiliary clearly involves attenuation. Let us consider it in regard to a participial construction, as in (13)c. The initial configuration is shown in the first diagram of Figure 6. In this construction, the onstage profiled relationship is that of the trajector maintaining a certain posture. Since 'stand' is a verb, the relationship is scanned

166 Ronald W. Langacker

sequentially by the conceptualizer (note the arrow labeled T, for processing time). The import of using 'stand' in this participial construction is that maintaining the posture accompanies and perhaps even renders possible another activity, carried out at the same time by the same individual. This is shown diagrammatically by the dotted correspondence line equating the two trajectors, as well as the double arrow, which indicates that maintaining the posture enables the participial activity to occur. The resulting expressions are roughly comparable to (14)a. Expressed by the participle, the accompanying activity is offstage and unprofiled - (14)a profiles the standing, not the looking - and since sequential viewing requires a high degree of focus, the offstage activity is viewed holistically, in summary fashion. (This holistic construal is one semantic contribution of the participial inflection; another - ignored here - is the "internal perspective" imposed by a progressive construction [Langacker 1995b].) (14) a. b. c. d. e.

He stood there looking over the fence. The clock stood ticking on the table. The cup was leaking. It was raining. Tabs were being kept on all the dissidents.

The second diagram in Figure 6 represents attenuation with respect to both the nature of the profiled process and its role in effecting the accompanying activity. In (14)b, for instance, stand is attenuated due to being predicated of an inanimate subject. Whereas a person exerts muscular control in order to maintain a vertical orientation, a clock stands passively, merely by virtue of having a certain shape. The effective relationship between stand and the participial activity is also greatly attenuated; presumably the clock will tick in any orientation.

Losing control: grammaticization, subjedification, and transparency 167

stand' + PARTICIPLE

stand + PARTICIPLE overall scope

α "•¿Λ

immediate scope

>

>

wmz \ '. 1/

be + PARTICIPLE be + PARTICIPLE overall scope immediate scope

Figure 6. Evolution of a be-type auxiliary verb

Carried to its extreme, attenuation of subject control erases even the orientational specifications of a verb like stand. Nothing then remains of any specificity - all that is left onstage is the highly sehe-

168 Ronald W. Langacker

matic notion of the trajector being involved in some wholly unspecified relationship followed sequentially through time. This is the value ascribed to a be-type verb, as described above. Moreover, since the profiled process no longer has any specific properties that would distinguish it from the activity expressed by the participle, these come to be equated, as indicated by the additional correspondence line in the third diagram of Figure 6. In other words, be profiles a schematic process followed sequentially through time, and the participle describes the same process in more specific terms. Hence the two relationships collapse into one, as seen in the final diagram. The be-type verb contributes the sequential viewing required for the head of a finite clause, resulting in a progressive construction, as in (13)c or (14)c. Because the profiled relationship is wholly schematic, it imposes no restrictions on its trajector. And since the profiled relationship is identified with the one expressed by the participle, the participial subject is ipso facto the subject of the entire construction. This is a situation of full transparency, as seen in (14)d-e. It is also a case of full overlap between the profiled relationship and that of its complement: owing to attenuation, nothing is left onstage that is not subsumed by the complement. This is a common outcome of grammaticization, which in general leads to greater overlap between the grammaticized element and the structure it combines with (Langacker 1992, 1995c). Needless to say, this is additionally a case of subjectification, since the sequential viewing which remains as the only essential contribution of be was immanent in the original value of stand.

7.

Get-Passives

Finally, consider the grammaticization of get in combination with a passive participial complement. In (15), we find a series of expressions representing an evolutionary path not unlike the one leading to the progressive be. In each example, get is a finite verb, hence it profiles a process viewed sequentially, whereas the passive particip-

Losing control: grammaticization, subjectification, and transparency 169

ial complement is non-finite and viewed holistically. The evolution of get involves progressive attenuation in both the nature of the profiled relationship and the degree of control exercised by its subject. The result is that get comes close to being just a passive auxiliary, like be, serving only to provide the sequential viewing required for the head of a finite clause. It has not gone quite that far, however. (15) a. b. c. d.

Sue got (herself) appointed to the governing board. Ralph got fired again. All my books got stolen. Another bank got robbed last night.

In all the examples, the main clause subject - the trajector of the finite verb get - is also the trajector of the complement. And since the complement is based on a passive participle, its trajector is the same element which functions as landmark (or patient) of the verb stem from which the participle derives. Our interest, though, lies in the trajector's role in the main clause relationship. In (15)a, the trajector and subject (Sue) is construed as a volitional agent who manages to bring about the participial event, and who also, secondarily, is an experiencer who enjoys the benefits of its occurrence. This volitional construal is almost necessary with the reflexive herself, but it is at least possible in the simpler construction without the reflexive. That is our starting point, sketched in Figure 7(a).

170 Ronald W. Langacker

Figure 7. Attenuation in the get-passive

More typical are examples like (15)b, where the subject is not a volitional agent with respect to the participial complement. The subject may still be responsible in some way - Ralph may have been fired because of what he did or did not do, or simply because his boss recognized certain undesirable qualities in him - but not necessarily. He is still however an experiencer in regard to the participial event, as shown by the dashed arrow in Figure 7(b). But under any of these construals, the subject's primary role is that of passive undergoer of the complement event. Its external, specifically main clause role has been attenuated and now ranges from some indeterminate kind of responsibility to merely that of an experiencer. If the main clause subject in (15)b is the locus of experience, this cannot be true in (15)c, where the subject is inanimate. We have further attenuation in the degree of subject control, for the experiencer is no longer the subject per se, but rather an individual associated with the subject, the possessor in this example. Moreover, even further diffusion is possible, as the locus of the experience need not be overtly specified or clearly delimited. Thus (15)d does not imply any particular experiencer, which may be construed as the owners or employees of the bank, or perhaps just lawful members of society who

Losing control: grammaticization, subjectification, and transparency 171

feel menaced by the rising level of crime. Hence in Figure 7(c) the experiencer is not identified with the trajector, nor is it put in focus as a specified, profiled participant. The passive be in English merely adds sequential viewing to the participial complement. The passive get may come close to this, but so long as any vestigial notion of experience remains, the relationship it profiles is not merely a schematic representation of the participial relationship. Moreover, ge/-passives approximate but do not quite achieve full transparency: (16) a. b. c. d.

?A lot of headway got made last night. ??Tabs got kept on all the dissidents. ? *It got claimed that there are wombats on Venus. *There got claimed to be wombats on Venus.

(Observe that all of these sentences are well formed if get is replaced with be.) Geí-passives are non-transparent in that the subject must in some way be implicated in the experiential relationship, not necessarily as the locus of experience, but maybe just by providing a link to the implicit experiencer. Headway is easily associated experientially with those who make it, but with tabs the matter is less clear (keep tabs on is less analyzable than make headway). Although I believe that it and there are also meaningful, their value is more abstract and does not pertain to the objective situation under description (Langacker 1995a: 55). Thus they are not readily implicated in an experiential relationship as examples like (16)c-d would imply. Further examples of attenuation in subject control are readily found. A case in point is the contrast between the main verb do, implying some degree of subject responsibility, and the auxiliary verb do, which is fully transparent. Another is the relationship between the so-called "equi" and "raising" senses of verbs like promise, threaten, and many others (cf. Ruwet 1991: 56-81; Traugott 1993; Farrell 1995; Yerhagen 1995).

172 Ronald W. Langacker

(17) a. Felix promised to get rid of his trombone. b. There promises to be a bright future for all of us. In the former, as in (17)a, the subject has to be capable of volition, planning, and communicative interaction. By contrast, (17)b merely expresses the speaker's assessment based on an overall appreciation of the current situation, hence there are no restrictions on the subject. The overall pattern documented here suggests that the so-called "raising" constructions are best seen as the limiting case of "equi" constructions, representing the extreme situation where attenuation of subject control (or its counterpart in the object-raising constructions) results in transparency (cf. Newman 1981, 1982; Langacker 1995a).

8.

Conclusion

Let me conclude by noting some limitations of the work described here. In articulating some general patterns observable in the evolution of numerous grammatical constructions, I have naturally been unable to examine any one construction in full depth and detail. In particular, I have not attempted to determine how many distinct conventional values have to be attributed to a particular element or construction at a given historical stage to account for the variety of contextual interpretations it displays. I should also emphasize that this work has not been based on serious historical investigation. In presenting series of examples representing progressive degrees of attenuation, I have not intended to suggest that they necessarily correspond to the actual order of diachronic development, which clearly has to be established in its own terms. On the positive side, I have offered a unified account of diverse phenomena usually considered in isolation from one another. Attenuation in subject control has been shown to be a pervasive, multifaceted phenomenon that plays a major role in certain kinds of grammaticization, with important consequences for synchronic analysis and description.

Losing control: grammaticization, subjectification, and transparency 173

References Brugman, Claudia 1988 The syntax and semantics of HAVE and its complements. Ph.D. dissertation, Department of Linguistics, University of California, Berkeley. Carey, Kathleen 1994 Pragmatics, subjectivity and the grammaticalization of the English perfect. Ph.D. dissertation, Department of Linguistics, University of California, San Diego. Chafe, Wallace 1994 Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago / London: University of Chicago Press. Farrell, Patrick 1995 Lexical binding. Linguistics 33: 939-980. Gernsbacher, Morton Ann and David Hargreaves 1992 The privilege of primacy: Experimental data and cognitive explanations. In: Doris L. Payne (ed.), Pragmatics of Word Order Flexibility, 83-116. (Typological Studies in Language 22.) Amsterdam / Philadelphia: Benjamins. Harder, Peter 1996 Functional Semantics: A Theory of Meaning, Structure and Tense in English. (Trends in Linguistics. Studies and Monographs 87.) Berlin / New York: Mouton de Gruyter. Heine, Bernd 1992 Grammaticalization chains. Studies in Language 16: 335-368. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In: John Haiman (ed.), Iconicity in Syntax, 109-150. (Typological Studies in Language 6.) Amsterdam / Philadelphia: Benjamins. 1986 Abstract motion. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 12: 455-471. 1987a Foundations of Cognitive Grammar. Volume 1, Theoretical Prerequisites. Stanford: Stanford University Press. 1987b Nouns and verbs. Language 63: 53-94. 1990a Concept, Image, and Symbol: The Cognitive Basis of Grammar. (Cognitive Linguistics Research 1.) Berlin / New York: Mouton de Gruyter. 1990b Subjectification. Cognitive Linguistics 1: 5-38. 1991 Foundations of Cognitive Grammar. Volume 2, Descriptive Application. Stanford: Stanford University Press.

174 Ronald W. Langacker

1992

Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81: 287-309. 1993 Reference-point constructions. Cognitive Linguistics 4: 1-38. 1995a Raising and transparency. Language 71: 1-62. 1995b Viewing in cognition and grammar. In: Philip W. Davis (ed.), Alternative Linguistics: Descriptive and Theoretical Modes, 153-212. (Current Issues in Linguistic Theory 102.) Amsterdam / Philadelphia: Benjamins. 1995c A note on the Spanish personal "a". In: Peggy Hashemipour, Ricardo Maldonado, and Margaret van Naerssen (eds.), Studies in Language Learning and Spanish Linguistics in Honor of Tracy D. Terrell, 431-441. New York: McGraw-Hill. 1997 A dynamic account of grammatical function. In: Joan Bybee, John Haiman and Sandra A. Thompson (eds.), Essays on Language Function and Language Type Dedicated to T. Givón, 249-273. Amsterdam / Philadelphia: Benjamins. in press Dynamic conceptualization in grammatical structure. MacWhinney, Brian 1977 Starting points. Language 53: 152-168. Newman, John 1981 The semantics of raising constructions. Ph.D. dissertation, Department of Linguistics, University of California, San Diego. 1982 Predicate adjuncts. Australian Journal of Linguistics 2: 153-166. Ruwet, Nicolas 1991 Syntax and Human Experience. Chicago / London: University of Chicago Press. [Ed. and trans, by John Goldsmith.] Sweetser, Eve E. 1982 Root and epistemic modals: Causality in two worlds. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 8: 484-507. 1988 Grammaticalization and semantic bleaching. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14: 389-405. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12: 49-100. Tomlin, Russell S. 1995 Focal attention, voice, and word order. In: Pamela Downing and Michael Noonan (eds.), Word Order in Discourse, 517-554. (Typological Studies in Language 30.) Amsterdam / Philadelphia: Benjamins.

Losing control: grammaticization, subjectification, and transparency 175

Traugott, Elizabeth 1986 From polysemy to internal semantic reconstruction. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 12: 539-550. 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31-55. 1993 The conflict promises/threatens to escalate into war. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 19: 348-358. this volume The rhetoric of counter-expectation in semantic change: A study in subjectification. Verhagen, Arie 1995 Subjectification, syntax, and communication. In: Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectivisation: Linguistic Perspectives, 103-128. Cambridge: Cambridge University Press.

The rhetoric of counter-expectation in semantic change: a study in subjectifícation Elizabeth Closs Traugott

1.

Introduction

My aim in this paper is to exemplify how strategic language use together with cognitive processes that constrain choices and understanding of inferences leads to semasiological change over time.1 In particular, I will discuss how "rhetorical uses" of a "linguistic device" (Kay 1990) such as in fact give rise to changes of a highly regular kind. The groundwork for this study was laid in syntactic-semantic investigations of the development of discourse markers like anyway, besides, indeed, in fact (Traugott 1995a, in press). In these studies I identified a structural path of change from: i)

Verb Adverb (VAdv); this is clause-internal, oblique, often a manner adverb and it typically occurs clause-finally, to ii) Sentence (or I[nflectional] P[hrase] Adverb (IPAdv); this has scope over the clause and occurs either after Complementizer] or adjacent to the tensed verb, to iii) Discourse Marker (DM); this occurs outside of Comp and has scope over the following complex structure. In short: (1)

VAdv > IPAdv > DM A constructed example to clarify the distinctions is:

(2)

a. There is no basis in fact. (VAdv). b. If in fact (IPAdv) there is no basis in fact (VAdv) / If there is in fact (IPAdv) no basis in fact. (VAdv).

178 Elizabeth Closs Traugott

c. In fact (DM) if in fact (IPAdv) there is no basis in fact. (VAdv). As (2) shows, old and new uses can (and typically do) coexist synchronically in a relationship known as "layering" (Hopper 1991). My main purpose was to stress that: a) Semantic change is regular not only in well-known domains such as space > time, deontic > epistemic, but also in other domains like manner or spatial adverbial > discourse marker. Evidence for the latter is provided by Hanson (1987) on modal adverbs like probably, possibly, Powell (1992a) on stance adverbs like actually, generally, loosely, precisely, Brinton (1990, 1996) on a variety of pragmatic markers including Old English hwcet 'what!, listen up!', anon, and bifel, gelamp 'happen' constructions, and Jucker (1997) on well. b) Recruitment to the new syntactic position2 was not possible until a prior semantic change had occurred permitting use of the form in that position (see also Bybee, Perkins and Pagliuca 1994 on evidence that semantic leads syntactic change). c) The changes should be considered cases of grammaticalization (see also Onodera 1995, Brinton 1996) although they do not fulfill the usual criterion of increased bonding or syntactic scope reduction (see Lehmann 1995 [1982]). From a cognitive semantic point of view, what is striking about the adverbs is that both contemporary and historical examples, especially in their IPAdv use, exhibit characteristics of the rhetorical strategy often called "counter-expectation". Furthermore, they are typically used in contexts where counter-expectation is redundantly marked, e.g. modals, or adversatives like although. When a speaker expresses counter-expectation, he or she expresses beliefs or points of view contrary to his or her own or the interlocutor's expectations regarding the states of affairs under discussion (see König 1986, 1991; Heine, Claudi and Hünnemeyer 1991; Fraser and MalamudMakowski 1996; Schwenter 1997; also, from a somewhat different perspective, Anscombre and Ducrot 1989).3 This can provisionally be summarized as:4

The rhetoric of counter-expectation in semantic change 179

(3)

U2 is not expected given Ui.

Counter-expectation is a matter of point of view and is therefore an example of the subjectivity of language. As has often been pointed out (see Benveniste 1971 [1958]; Lyons 1982; Stubbs 1986; Langacker 1990; papers in Stein and Wright 1995) we cannot say or write anything without expressing some aspect of point of view. Everything is the product of a speaker (the "speaking subject" or "sujet d'énonciation") and "involves the expression of self and the representation of a speaker's ... perspective or point of view in discourse - what has been called a speaker's imprint" (Finegan 1995: 1). However, some utterances are more subjective than others in that they specify the subjectivity involved. Compare: (4)

a. Bill spoke frankly. b. Frankly. Bill spoke.

In (4a) subjectivity is minimal, but not absent: past tense is anchored in the speaker, as is choice of unmarked topic structure, etc. By contrast frankly in (4b) explicitly draws attention to speaker's subjective point of view or stance to what is being said. While subjectivity is pervasive in language use, only a subset of elements in language serve to express it explicitly, and often only in certain constructions (e.g. frankly clause-initially in (4b) but not clause-finally as in (4a)). These elements come to do so through the process of subjedification which I characterize as follows: If the meaning of a lexical item or construction is grounded in the sociophysical world of reference, it is likely that over time speakers will develop polysemies that are grounded in the speaker's world, whether reasoning, belief, or metatextual attitude to the discourse. Subjectification, then is the semasiological development of meanings associated with a form such that it comes to mark subjectivity explicitly.5 I return to this point in section 5, where I also discuss the relationship of this view of subjectification to Langacker's.

180 Elizabeth Closs Traugott

2.

The distinction between IPAdvs and DMs

There is currently far from common agreement about how to distinguish adverbs, despite extensive work on syntactic and semantic distinctions by, among others, Greenbaum (1969), Jackendoff (1972), Allerton and Cruttendon (1974), Ernst (1984), Quirk et al. (1985), and McCawley (1988). Most contemporary analyses of adverbs recognize a multiplicity of syntactic positions for adverbs in English, but make a major distinction only between what I am here calling VAdvs (adverbs of direction, manner, etc. that are oblique arguments of the verb), and one other. This other class of adverbs is usually characterized as "modal", "evidential", "stance" (e.g. Biber and Finegan 1988), or as "pragmatic" (e.g. Fraser 1996). Distinctions in meaning within either class are treated as homonymous (Fraser) or unitary and monosemous, with variations on a core meaning dependent on context (Sperber and Wilson 1995). Here I distinguish IPAdv and DM on grounds not only of the distributional differences sketched in Section 1 but also of pragmatic differences to be outlined below. Furthermore, I regard the distinctions as polysemous; polysemy may be pragmatic (see Horn 1985; Sweetser 1990) as well as semantic. Methodologically and theoretically it is not possible to investigate semantic change on the assumption of homonymy, because if the forms are homonymous they are by definition not related, and therefore no historical connection between uses can be postulated. It is also impossible to investigate them on the assumption of monosemy since the meanings that license new contextual uses cannot be accounted for sufficiently explicitly. DMs have been a major topic in pragmatics and discourse analysis since the publication of Schiffrin's groundbreaking (1987) book entitled Discourse Markers, in which she analyzed a variety of pragmatic markers such as y 'know, I mean, but, oh, then. More recently there has been a tendency to bifurcate the domain into: a) those markers that signal relationships between clauses; Schiffrin (1990) terms them "discourse deictics", while Fraser (1988,1996) calls them "discourse markers",6

The rhetoric of counter-expectation in semantic change 181

b) other pragmatic markers that serve primarily interpersonal functions, such as hedging, or turn-taking cues, e.g. well, y 'know. DMs are here taken to be adverbs that: "signal a comment specifying the type of sequential discourse relationship that holds between the current utterance ... and the prior discourse" (Fraser 1988: 2122). They are "metatextual"7 in that their function is to comment on the text being constructed; they are also "procedural" in the sense that they are primarily pragmatic and non-truth-conditional. Being deictic, they "point to" the speaker's metatextual attitudes, and also guide hearers to those attitudes. D

3.

A synchronic account of in fact

3.1. An earlier analysis Fraser cites in fact in several of his writings on DMs, but classifies it in different ways. For example, Fraser (1988) lists in fact under two kinds of DM function (topic resumption and elaboration) while Fraser (1996) lists it only once (elaboration). Fraser and MalamudMakowski (1996: 871) list it under contrastive DMs and say that it: "signals that U2 is to be interpreted as a denial of the explicit proposition asserted with Ui, with the added condition that the assertion of Ui is not the direct responsibility of the speaker of U2". Among their examples of in fact are: (5)

a. I thought the rock I found was granite. In fact, the entire quarry was quartz. b. A. The paint is purple. B. In fact, it's mauve. (Fraser and Malamud-Makowski 1996: 871-872)

Being clause-initial in (5), and without a complementizer, contrastive in fact occurs in the same surface position as elaborative in fact, as can be seen by constructing (5b) as a monologue:

182 Elizabeth Closs Traugott

(5)

c. The paint is purple. In fact it's mauve.

Here in fact would probably be construed not as a denial of the preceding utterance, but as the speaker's improvement on the preceding utterance. In other words, whereas in fact in (5b) is probably an IPAdv (B implies, No it is not purple, it is mauve, and denies any assumption that purple and mauve are similar), in (5c) it is probably a DM (the speaker adjusts the color term on a scale of precision and implies that mauve is a more precise kind of purple). I say "probably" because the examples in (5) are constructed. The cognitive task of interpreting non-redundant utterances is a misleading one for the researcher who seeks to account for how human beings communicate with each other. Most naturally occurring data provide not only verbal context but redundant verbal context which typically disambiguates the uses in question. Spoken data also provide distinct intonational cues (see work by Allerton and Cruttendon 1974; Ferrara 1997). I now turn to naturally occurring written data. The problem for a historical linguist is that until the present century written texts are all we have, and appeals to non-verbal context must be speculative at best. So in comparing the historical with contemporary data, it is best to use contemporary written data.

3.2. In fact as VAdv, IPAdv, and DM Contemporary data evidence at least three syntactically and semantically distinct uses of in fact. First, there is the VAdv, which is clause-final and, in (6), contrasts with in feeling (other contrasts in the data include in law, and sensationalism). (6)

Humanity, comfortably engaged elsewhere in the business of living, is absent in fact but everywhere present in feeling. (United Airlines Magazine, 5/1997)

The rhetoric of counter-expectation in semantic change 183

A second type, the IPAdv, is illustrated by (7), where in fact follows the finite verb and contrasts with e.g. however. (7)

It's almost as if he were avoiding the evening's moment of truth. The moment is, in fact, about 45 minutes, the main body of the play, during which Mr. Quinton ... holds the stage and acts up a primeval cataclysm all by himself. (New York Times, 7/29/1996, review oí Phaedra)

In Ui the frame It is as i f , the scalar almost, and the semantically negative avoiding all serve to warn that Ui is problematic, dissonant, and contrary to the speaker's view. This dissonance is further highlighted by the reorientation of "moment" from something normally thought to be very short to a lengthy 45 minutes. Examples like (7) show that Ui may indeed be the responsibility of the speaker - speakers set up false scenarios and then show that the assumptions manifest in them are wrong, inappropriate to the occasion, etc. We may characterize this as follows: (8)

IPAdv in fact signals that U2 is to be interpreted as a denial of one or more manifest assumptions in Ui.

The third (DM) use of in fact typically occurs clause-initially; it may also occur after a finite verb. Unlike IPAdv in fact, it is used in a semantic set with in other words, I mean, but not however. A typical example is: (9)

Polling isn't The World's Oldest Profession, although around election time it might seem like it. In fact, once upon a time, way back in the first third of this benighted century, modern polling wasn't yet even a gleam in the eye of a small-town Iowa kid named George Gallup. Today... (United Press International, 10/8/1990)

184 Elizabeth Closs Traugott

As (9) shows, the main function of DM in fact is elaboration. Yet it does also signal some level of contrast. As in (9), it is often used after a negative statement. It serves to confirm the inference from that negative statement (which itself counters a prior expectation), and to provide an elaboration which is a clarification or reformulation of the negative statement. Here the elaboration is presented as a justification of the denial in the preceding sentence. The DM use of in fact can be characterized as: (10) DM in fact signals that U2 is an elaboration of Ui, and a more precise formulation than Ui.

4.

A brief history of in fact

In fact has its historical beginnings in the early sixteenth century. In the earliest uses we find fact as a lexical noun (Stage 0, the initial stage before change has set in). My first example of it is in a prepositional phrase is: (11) Stage 0: Full Lexical Ν A sort of naughty persons..., Have practised dangerously against your state, Dealing with witches and with conjurors, Whom we have apprehended in the fact. Raising up wicked spirits from under ground. (cl592 Shakespeare, 2 Henry VI, Il.i. 163-170) Fact begins to appear in the Helsinki Corpus by the late 17thC in a bare prepositional phrase as the object of in (Stage I, the initial construction which later split into VAdv and IPAdv). Here it is favored in coordinate constructions, either as a member of a taxonomically contrasted set as in (12a) or in an explicitly contrastive context like (12b):

The rhetoric of counter-expectation in semantic change 185

(12) Stage I: VAdv a. But it is evident in fact and experience that there is no such universal Judge, appointed by God over the whole World, to decide all Cases of temporal Right. (1671 Tillotson, Sermons, p.II.ii.445 [HC]) b. This company, therefore, have always enjoyed an exclusive trade in fact, though they may have no right to it in law. (1776 Smith, Wealth of Nations, Bk 5, p.701) In these kinds of contexts, in fact came to be endowed with evidential meanings ('in practice/reality/actuality') that assess something as physically accessible, empirically attested, real and true. (12a) cites fact (action, performance) and experience (as opposed to speculation) as sources of evidence, (12b) cites it as the pragmatic and experiential locus of daily transactions, as opposed to abstract legal right. Such uses, as well as independent developments in the philosophical and scientific construals of 'fact', are likely to have contributed to the recruitment of in fact to the class of adverbs that are potentially epistemic modals, specifically those that signal epistemic certainty. By the mid-18thC in fact is used as an IPAdv in clause-initial post-complement position. Often it occurs in the already contrastive environment of but as in (13a) or in highly rhetorical contexts, as in (13b) which is a rhetorical question, followed by a denial: (13) Stage Π: IPAdv a. Every particle eludes the grasp by a new fraction, like quicksilver, when we endeavour to seize it. But as in fact here must be something which terminates the idea of every finite quantity... (1739-40 Hume, Treatise Human Nature, Bk. 1, p. 44 ) b. If this be their true interest, have they in fact pursued it? Has it not, on the contrary, invariably been found, that... ? (1787 Federalist Papers, p. 31)

186 Elizabeth Closs Traugott

As these examples indicate, the IPAdv use points to the truth of ρ despite prior contrary expectations regarding it. The first example in my data of the third, DM, stage, occurs in a novel by Jane Austen, in dialogue. In the new use, in fact functions at the discourse level to express the speaker's attitude to the appropriateness of the discourse itself. In this sense it serves a metatextual function and is entirely procedural. In (14a) in fact introduces justification of what has just been said, in other words, self-corrective elaboration which is contrastive not in terms of truth but of appropriateness of expression. Similarly in (14b) in fact signals that what follows strengthens and specifies what precedes (lead inevitably to): (14) Stage ΠΙ: DM a. "I was wrong, " he continued, "in talking of its being broke to you. I should not have used the expression. In fact, it does not concern you - it concerns only myself (1815 Austen, Emma, Vol. 3, Chap. 10, p. 393) b. Thus in various ways ethical questions lead inevitably to psychological discussions; in fact, we may say that all important ethical notions are also psychological. (1874 Sidgwick, History of Ethics, Chap. 1, p. 4) Note that in both instances the context explicitly refers to language use (expression, discussion). The data suggest that the epistemic meaning of in fact originally arose in contexts of lexical contrast, specifically as an alternate in lexical fields concerning human action as it pertains to judgment, whether divine or human, but in any case where action is taken as evidence. In these contexts in fact is used to refer to physical acts as contrasted with thought, belief, etc. Once understood as epistemic, it could be recruited to the class of epistemic IPAdvs. Initially, this recruitment takes place in contexts that render in fact redundant, such as contrastive but in (13a), or in the question and conditional in (13b). Here it eventually takes on the force of contrast and can be understood as adversative in less explicitly adversative contexts, and eventually even out of context.

The rhetoric of counter-expectation in semantic change 187

5.

Implications for historical semantics

Cognitive linguists have tended to favor a metaphorical view of semantic change (see Sweetser 1990; Heine, Claudi, and Hünnemeyer 1991), though recently there has been growing interest in conceptual métonymie change, also known as change via context-induced inferencing (see Traugott and König 1991). If we consider the history of in fact, a metaphorical interpretation is not ruled out at the earliest stage. On the metaphorical view, physical deeds come to be understood as mapped onto anything that is observable. On the other hand, a métonymie inference from doings is that they are observable because they are actions (not thoughts, opinions, etc.). Good metonymies are often metaphorical as well, so there is no difficulty in looking at the change either way. However, when we consider later changes involving the use of in fact to signal counter-expectation at the propositional level and then metatextual rejection of earlier formulations, etc., an account from metaphor becomes less plausible. An interpretation based on processes of inferencing from the context, i.e., from the linear progression of utterances and the redundancies in them, is required, together with processes of subjectification that lead speakers to semanticize (conventionalize) those inferences that saliently express their attitude.9 Among semantic changes, subjectification has been of special interest. Although already mentioned in Bréal's work (1964 [1900]: Chapter 25), subjectivity did not come to be of major topic of research until Benveniste raised the question whether a language could still function and be called a language unless it were "marked so deeply by the expression of subjectivity" (1971: 225), and distinguished between the syntactic subject/"sujet d'énoncé" and the "speaking subject"/"sujet d'énonciation" (see also Lyons 1982, Langacker 1985, this volume). A historical view of subjectification has been developed by e.g. Traugott (1982,10 1989, 1995b), Langacker (1990), papers in Stein and Wright (1995). Although Langacker's and my own views on the ubiquity of subjectivity coincide, we are using the term "subjectification" to refer to considerably different phenomena, and from different perspectives

188 Elizabeth Closs Traugott

on theory and methodology. The differences are therefore not "purely terminological" (Langacker, this volume). For Langacker subjectification is primarily a synchronic phenomenon characterizing perspectivization of those constructions in which the subject of conception is "offstage" and implicit, and in which speaker and hearer are "normally offstage", although "immanent in the objective conception". He also conceives subjectification as a diachronic phenomenon, at least with respect to those verbs that involve "raising", but the exact relationship between the synchronic perspectivization and the diachronic change remains unclear. For me, subjectification is a diachronic phenomenon only. It may result synchronically in layerings of less or more subjective meanings for the same lexical item or construction, but there is no necessary isomorphism between the diachronic and synchronic variation. For Langacker, the data in question are constructed, typically non-redundant, and considered out of context. For me, the data are linguistically highly contextualized and always naturally occurring (even if written). For Langacker, subjectification is the result of attenuation, most especially of the "objectively profiled subject". However, in my view, subjectification is not limited to constructions involving the syntactic subject, as is demonstrated by the development of DM out of IPAdv uses of adverbs like in fact, or of scalar particles like even out of a manner adverb meaning evenly (Traugott 1995a). Most importantly, subjectification is not characterized by attenuation, as this same example illustrates (even originates in the meaning 'level, uniform'; see also degree modifier kinda, sorta, which derive from (a) kind of, (a) sort of (Tabor 1993)). Far from attenuation, it involves "pragmatic strengthening" and enriching of the form-meaning pair in question with the speaker's perspective. Subjectification is very wide-spread, indeed the most pervasive tendency in semantic change. Its effects can be detected in the development of lexical items that express events and their participants (e.g. the development of pejorative and ameliorative changes such as are cited in the literature for boor, churl, knight, etc.). More interestingly for cognitive semantics, its effects are also found in the development of evidential and epistemic domains, as for instance in

The rhetoric of counter-expectation in semantic change 189

the development of highly subjective meanings of epistemic modals like must. Other prime examples are the development of performative uses of locutionary verbs such as promise, recognize, insist (ultimately derived from spatial and mental terms), and especially of the IPAdv and DM meanings of adverbs like in fact discussed here. Thus: >

VAdv Least subjective

IPAdv

>

DM •

Most subjective

Figure 1. Adverbials: Path of subjectification

While the first stage (VAdv > IPAdv) might conceivably be construed as a case of "loss of control" by the subject of the event-structure that VAdv modifies, the second (IPAdv > DM) cannot. In both the IPAdv and DM cases in fact is grounded in the speaker's perspective. The scope of that perspective is the clause in the case of the IPAdv, of the relationship between successive discourse units in the case of the DM. Similarly, subjectification effects in morphological domains (see e.g. Company 1993) cannot be identified with loss of control. Subjectification involves speakers recruiting forms with appropriate meanings to externalize their subjective point of view. This is an activity that draws on cognitive principles but takes place in the context of communication and rhetorical strategizing. As Stubbs (1986) has pointed out: "whenever speakers (or writers) say anything, they encode their point of view towards it: whether they think it is a reasonable thing to say, or might be found to be obvious, irrelevant, impolite, or whatever" (Stubbs 1986: 1). Stubbs's key word here is "encode". It has been suggested (Diewald 1993, Keller 1995) that subjectification defined as preemption of old meanings to encode and externalize speaker subjectivity is really objedification. From the point of view developed here, which focuses on speakers' use of old structures with new functions, such a suggestion confuses objectivity with making explicit and manifest features of the discourse situation.

190 Elizabeth Closs Traugott

The particular semantic change undergone by in fact discussed in this paper is an example of the development of a class of elements that Talmy has called "force-dynamic logic-gaters [that] limn out the rhetorical framework, ... direct the illocutionary flow, ... specify the logical tissue" (Talmy 1988: 88-89). If in fact is indeed a "logicgater", it might seem to be objective. However, as I have shown, it is highly subjective, expressing speaker's alternative points of view to some prior discourse in the case of the IPAdv, and to the discourse being negotiated in the case of the DM. In neither use does it add anything to the truth-value of what is said, or express a logical relationship. However, because it has the appearance of a logical and objective marker, it can be said to "locate the claim ... in the communally available realm of intersubjectivity" and therefore is partially intersubjective (Powell 1992b: 349), although its prime function is subjective. Does objectification then occur? Indeed. In semantic change it is the highly conscious and deliberately interventive sort of change that comes about when ordinary words are preempted for technical or legal purposes, e.g. linguists' use of "competence", "relevance", "quantity", or the redefinition of the word "harassment" following on the U.S. Civil Rights Act of 1964. But these are cases of "external", "non-natural" change, not of "internal", "natural" change, which is the central concern of historical cognitive semantics.

Notes 1.

2.

Thanks to Lisbeth Lipari, whose work on fact in 1994 provided the original impetus for work on this construction, and especially to Scott Schwenter, for numerous comments, bibliographical suggestions, and inspiration. The main historical data base is The Helsinki Corpus of the English Language (HC); (see e.g. Rissanen et al. 1993). Stanford Academic Text Services made access to this and a variety of other computerized corpora possible. The position is "new" with respect to the construction not the grammar. Both the IPAdv and a position related to DM (left-most "Topic") have been available from Indo-European times (see Kiparsky 1995).

The rhetoric of counter-expectation in semantic change 191

3.

4. 5.

6.

7.

8.

9. 10.

Another way to conceptualize this is as a "polyphony" (see Nalke 1992, working in the tradition of Anscombre and Ducrot 1989). For example, in connection with negation, N0lke says not "induces (sic) a polyphonic structure into the sentence consisting of two (incompatible) viewpoints" (Nelke 1992: 190; italics original). U = utterance. Ui and U 2 are not necessarily adjacent. As a semasiological change, it concerns the histories of individual lexical items. It has nothing to do with choice of styles, which may be more "objective" or "subjective" at different times (Adamson 1995); nor with the choices made by certain groups for use in certain registers, in renewing the lexicon or expanding it for new political, academic, legal, medical or other purposes (Marchello-Nizia, in press). The claim is only that if a lexical item is available in the system, over time it is likely to acquire more subjective meanings. The term "discourse marker" has also sometimes been applied to elements of a conjunctive rather than adverbial type, such as and, but. The latter are, however, more commonly called "discourse connectives", cf. Blakemore (1987) on and, but, so, etc. Pons Borderiá (1995) discusses the various terminological distinctions. As Dancygier (1992) points out, this term is preferable to "metalinguistic" for expressions of attitude to the text being constructed; this allows "metalinguistic" to be used for corrections, and what Horn (1995) has called "metalinguistic negation". The term was introduced into relevance-theory by Blakemore (1987) to characterize pragmatic markers that "guide the inferential phase of comprehension" (Sperber and Wilson 1995: 12-13). Despite adopting this term, I do not endorse the view that procédurals are primarily addressee-oriented. Traugott (in press) presents a model of this process. The term "subjectification" is not used in Traugott (1982), but the "expressive" tendency of language identified there is roughly equivalent to subjectification as defined here.

References Adamson, Sylvia 1995 From empathetic deixis to empathetic narrative: Stylisation and (de-) subjectivisation as processes of language change. In: Stein and Wright (eds.), 195-224. Allerton, D. J. and Alan Cruttenden 1974 English sentence adverbials: Their syntax and their intonation in British English. Lingua 34: 1-30.

192 Elizabeth Closs Traugott

Anscombre, Jean-Claude and Oswald Ducrot 1989 Argumentativity and informativity. In: Michael Meyer (ed.), From Metaphysics to Rhetoric, 71-87. (Synthese Libraiy 202.) Dordrecht: Kluwer. Benveniste, Emile [1958] 1971 De la subjectivité dans le langage. In: id., Problèmes de linguistique générale, 258-85. Paris: Gallimard, (trans, by Mary Elizabeth Meek as Problems in General Linguistics, 224-246. Coral Gables, FL: University of Miami Press). Biber, Douglas and Edward Finegan 1988 Adverbial stance types in English. Discourse Processes 11: 1-34. Blakemore, Diane 1987 Semantic Constraints on Relevance. Oxford: Blackwell. Bréal, Michel [1900] 1964 Semantics; Studies in the Science of Meaning. New York: Dover Publications, (trans, by Mrs. Henry Cust). Brinton, Laurel J. 1990 The development of discourse markers in English. In: Jacek Fisiak (ed.), Historical Linguistics and Philology, 45-71. (Trends in Linguistics. Studies and Monographs 46.) Berlin / New York: Mouton de Gruyter. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Function. (Topics in English Linguistics 19.) Berlin / New York: Mouton de Gruyter. Bybee, Joan, Revere Perkins and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Company, Concepción 1993 Old forms for new concepts: The recategorization of possessive duplications in Mexican Spanish. In: Henning Andersen (ed.), Historical Linguistics 1993, 77-92. Amsterdam: Benjamins. Dancygier, Barbara 1992 Two metatextual operators: Negation and conditionality in English and Polish. Proceedings of the 18th Meeting of the Berkeley Linguistics Society: 61-75. Diewald, Gabriele 1993 Zur Grammatikalisierung der Modalverben im Deutschen. Zeitschrift für Sprachwissenschaft 12: 218-234. Ernst, Thomas Boyden 1984 Toward an Integrated Theory of Adverb Position in English. Bloomington: Indiana University Linguistics Club.

The rhetoric of counter-expectation in semantic change 193

Ferrara, Kathleen 1997 Form and function of the discourse marker anyway: Implications for discourse analysis. Linguistics 35: 343-348. Finegan, Edward 1995 Subjectivity and subjectivisation: An introduction. In: Stein and Wright (eds.), 1-15. Fraser, Bruce 1988 Types of English discourse markers. Acta Linguistica Hungarica 38: 19-33. 1996 Pragmatic markers. Pragmatics 6: 167-190. Fraser, Bruce and M. Malamud-Makowski 1996 English and Spanish contrastive discourse markers. Language Sciences 18: 863-881. Greenbaum, Sidney 1969 Studies in English Adverbial Usage. London: Longman. Hanson, Kristin 1987 On subjectivity and the history of epistemic expressions in English. Papers from the 23rd Annual Regional Meeting of the Chicago Linguistic Society: 133-147. Heine, Bernd, Ulrike Claudi and Friederike HUnnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hopper, Paul J. 1991 On some principles of grammaticization. In: Traugott and Heine (eds.), Vol. I, 17-35. Horn, Laurence R. 1985 Metalinguistic negation and pragmatic ambiguity. Language 61: 121-174. Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. (Studies in Linguistics series 2.) Cambridge, MA: MIT Press. Jucker, Andreas H. (ed.) 1997 The discourse marker well in the history of English. English Language and Linguistics 1:91-110. Kay, Paul 1990 Even. Linguistics and Philosophy 13:59-111. Keller, Rudi 1995 The epistemic weil. In: Stein and Wright (eds.), 16-30. Kiparsky, Paul 1995 Indo-European origins of Germanic syntax. In: Adrian Battye and Ian Roberts (eds.), Clause Structure and Language Change, 140169. Oxford: Oxford University Press.

194 Elizabeth Closs Traugott

König, Ekkehard 1986 Conditionals, concessive conditionals and concessives: Areas of contrast, overlap and neutralization. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly and Charles A. Ferguson (eds.), On Conditionals, 229-246. Cambridge: Cambridge University Press. 1991 The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In: John Haiman (ed.), Iconicity in Syntax. (Typological Studies in Language 6.) Amsterdam / Philadelphia: Benjamins. 1990 Subjectifîcation. Cognitive Linguistics 1: 5-38. this volume Losing control: grammaticization, subjectifîcation, and transparency. Lehmann, Christian [1982] 1995 Thoughts on Grammaticalization. Munich: Lincom Europa. (First published as Thoughts on Grammaticalization. A Programmatic Sketch. Vol. I. Arbeiten des Kölner Universalien-Projekts 48. Köln: Universität zu Köln. Institut ftlr Sprachwissenschaft.) Lyons, John 1982 Deixis and subjectivity. Loquor, ergo sum? In: Robert J. Jarvella and Wolfgang Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics, 101-124. New York: John Wiley. Marchello-Nizia, Christiane in press Language evolution and semantic representations: From "subjective" to "objective" in French. In: C. Fuchs and S. Roberts (eds.), Language, Diversity, and Cognitive Representations. (Human Cognitive Processing.) Amsterdam: Benjamins. McCawley, James D. 1988 The Syntactic Phenomena of English. Chicago: University of Chicago Press, Vol. 2. Nelke, Henning 1992 Semantic constraints on argumentation: From polyphonic microstructure to argumentative macro-structure. In: Frans H. van Eemeren (ed.), Argumentation Illuminated, 189-200. Amsterdam: SICSAT. Onodera, Noriko Okada 1995 Diachronic analysis of Japanese discourse markers. In: Andreas Jucker (ed.), Historical Pragmatics, 393-437. (Pragmatics and Beyond 35.) Amsterdam / Philadelphia: Benjamins.

The rhetoric of counter-expectation in semantic change 195

Pons Borderiá, Salvador 1995 A prototype approach to the concept of connective. MS, Stanford University. Powell, Mava Jo 1992a The systematic development of correlated interpersonal and metalinguistic uses in stance adverbs. Cognitive Linguistics 3: 75-110. 1992b Folk theories of meaning and principles of conventionality: Encoding literal attitude via stance adverb. In: Adrienne Lehrer and Eva Feder Kittay (eds.), Frames, Fields, and Contrasts: New Essays in Semantic and Lexical Organization, 333-353. Hillsdale, NJ: Lawrence Erlbaum Assocs. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rissanen, Matti, Merja Kytö, and Minna Palander-Collin (eds.) 1993 Early English in the Computer Age: Explorations through the Helsinki Corpus. (Topics in English Linguistics 11.) Berlin / New York: Mouton de Gruyter. Schiffrin, Deborah 1987 Discourse Markers. (Studies in Interactional Sociolinguistics 5.) Cambridge: Cambridge University Press. 1990 Between text and context: Deixis, anaphora, and the meaning of then. Text 10: 245-70. Schwenter, Scott A. 1997 The pragmatics of conditional marking: Implicative, scalarity and exclusivity. Ph.D. dissertation, Stanford University. Sperber, Dan and Deirdre Wilson 2 1995 Relevance: Communication and Cognition. Oxford: Blackwell. Stein, Dieter and Susan Wright (eds.) 1995 Subjectivity and Subjectivisation in Language. Cambridge: Cambridge University Press. Stubbs, Michael 1986 Ά matter of prolonged field work': Notes toward a modal grammar of English. Applied Linguistics 7: 1-25. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. Tabor, Whitney 1993 The gradual development of degree modifier sort of and kind of: A corpus proximity model. Papers from the 29th Regional Meeting of the Chicago Linguistic Society: 451-465.

196 Elizabeth Closs Traugott

Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12: 49-100. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Directions for Historical Linguistics: A Symposium, 245-271. Austin: University of Texas Press. 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 57: 33-65. 1995a The role of discourse markers in a theory of grammaticalization. Paper presented at the 12th International Conference on Historical Linguistics, Manchester. 1995b Subjectification in grammaticalization. In: Stein and Wright (eds.), 31-54. in press Constructions in grammaticalization. In: Richard Janda and Brian Joseph (eds.), A Handbook of Historical Linguistics. Oxford: Blackwell. Traugott, Elizabeth Closs and Bernd Heine (eds.) 1991 Approaches to Grammaticalization. 2 volumes. (Typological studies in language 19.) Amsterdam / Philadelphia: Benjamins. Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In: Traugott and Heine (eds.), Vol. I, 189-218.

Synecdoche as a cognitive and communicative strategy Brigitte Neriich and David D. Clarke The urge to classify is a fundamental human instinct; like the predisposition to sin, it accompanies us into the world at birth and stays with us to the end. (A. Tindell Hopwood)

1.

Defining synecdoche

Until quite recently, research in rhetoric, literary theory and cognitive semantics has been characterised by the dominance given to metaphor. Metonymy has seldom been subjected to the same type of scrutiny. However, this situation seems to be changing rapidly as metonymy is being discovered as a cornerstone of human cognition and ordinary language use (as demonstrated by the recent Workshop on "Conceptual Metonymy" held at Hamburg University, June 23 to 24,1996). However, "confusion piles upon obscurity when we consider the treatment given to synecdoche" (Bredin 1984: 45). Not only is there no clear definition of synecdoche, there is also neither agreement on the various types of synecdoche, nor on synecdoche's relation to metonymy, or on synecdoche's function in actual discourse. Bernard Meyer summarises the situation well when he writes: "La catégorie de la synecdoque apparaît donc comme une classe rhétorique d'extension flottante, une nébuleuse de figures variant autour d'un noyau stable" [The category of synecdoche appears therefore as a rhetorical class with a vague extension, a nebula of figures which vary around a stable core] (Meyer 1993: 85) - and we shall see that even this core has been cracked open.

198 Brigitte Neriich and David D. Clarke

Brigitte Neriich (in press) has provided an overview of the historical shifts and changes in the status of synecdoche (for metonymy, read Bonhomme 1987). We shall provide a short summary of the main stages here, but it has to be stressed that this overview is not at all exhaustive: In Antiquity, namely in the work of Aristotle, synecdoche was still studied as part of metaphor, or rather, the three main figures of speech, metaphor, metonymy and synecdoche, were not yet strictly differentiated. 'Metaphor' in a wider sense

i

metaphor

Figure 1. Antiquity

In classical rhetoric (epistomised perhaps by Pierre Fontanier, [1821/1827] 1968) scholars distinguished between three, sometimes four, main tropes or master tropes: metaphor, metonymy, synecdoche and irony (cf. also Burke 1969). Synecdoche itself was seen as subsuming various subtypes. The kernel consisted for a long time of two subtypes of synecdoches: the part for whole one and the genus for species one (and vice versa), with the part-whole subtype being the most stable part of this typology. Classical examples for the part-whole and genus-species subtypes of synecdoche are: sail for ship and bread for food.

Synecdoche as a cognitive and communicative strategy 199

figures of speech

metaphor

metonymy

synecdoche

part/ whole

genus/species Figure 2. Classical rhetoric

In the process of reducing all tropes to two, especially under the influence of Roman Jakobson (1956, [1956]1983), that is to metaphor (based on similarity) and to metonymy (based on contiguity), metonymy was seen as including the part-whole type of synecdoche, and the genus-species type of synecdoche was largely forgotten. figures of speech

metaphor

metonymy [including part/whole] Figure 3. Jakobson

Synecdoche was rediscovered by the Groupe μ and declared the master-trope by Tzvetan Todorov (1970) in the 1970s, an enterprise that was however fraught with difficulties, as a new terminology only served to hide the older conception of synecdoche in which it was dressed up. Todorov for example gave as an example for the type of

200 Brigitte Neriich and David D. Clarke

synecdoche of the part for the whole the one of voile for bateau, which is, as we shall see, a metonymy, a point already made by Michel Le Guern (1973: 29-38), Nobuo Sato (1979), and Adolphe Nysenholc (1981). The Groupe μ were aware of the distinction between the part-whole type of synecdoche (exploiting what is nowadays called partonomies or meronymies; cf. Tversky 1990; Cruse 1986), and the genus-species type of synecdoche (exploiting taxonomies), but ended up confusing them again. What the Groupe μ calls particularising synecdoche, that is the figure of the part for the whole, should, as we shall see, be regarded as part of metonymy. In short, whereas synecdoche was absorbed by metonymy in the case of Jakobson, synecdoche absorbed part of metonymy in the case of the Groupe μ. The discussion in France and else-where surrounding the Groupe μ'β general rhetoric would deserve a separate article. Only quite recently (but going back to some pre-Jakobsonian writers such as Nyrop 1913, Esnault 19251 and others) the ancient kernel of synecdoche, it's "noyau stable", has been broken up in a different way. One constituent (the part-whole subtype) has been attributed to metonymy, whereas the genus-species subtype of the kernel has been preserved to define synecdoche as a third member in a triplet of essential tropes, namely metaphor, metonymy and synecdoche. This last option has been most forcefully defended by Kenichi Seto in his paper entitled "Metonymy and the Cognitive Triangle" (Seto, in press a), as well as in his contribution to the Hamburg workshop entitled "On Distinguishing Metonymy from Synecdoche" (Seto, in press b). Armin Burkhardt summarises this view well when he writes: Der alte Streit um die Zuordnung der Teil/Ganzes-Beziehung kann begraben werden, sobald einsichtig wird, daß er sich nur einer Äquivokation verdankt - kann doch pars sowohl 'Bestandteil' als auch 'Teilmenge' bedeuten. Nur im letzteren Falle darf die pars/totum-Beziehung als eine quantitative und damit synekdochische betrachtet werden, während die Beziehung des materiellen Bestandteils zum Ganzen in die Reihe metonymischer Relationen gehört. (Burkhardt ms., 1995: 2-3) [The old quarrel about the whole-part relation can be forgotten as soon as we become aware of the fact that it is only based on an equivocation, for

Synecdoche as a cognitive and communicative strategy 201

pars can mean 'component' as well as 'subset'. But only in the latter case should we regard the pars/totum-relation as a quantitative and therefore synecdochical one, whereas the relation of a material component to the whole belongs to the set of métonymie relations.]

To summarise, one can therefore say that metonymy is based on qualitative, synecdoche on quantitative relations, that is on set-inclusion. Metonymy is based on our world-knowledge about space and time, cause and effect, part and whole, whereas synecdoche is based on our taxonomic or categorical knowledge. Metonymy exploits our knowledge of how the world is, synecdoche of how it is ordered in our mind (Nerlich, in press). This view is to some extent opposed to two other more modern approaches to metaphor, metonymy and synecdoche, one written in the French tradition of "general rhetoric", the other against it. In their 1977 article "Metaphor, Metonymy and Synecdoche (Re)visited", Peter Schofer and Donald Rice argued that: Synecdoche is characterized by a semantic or referential relationship of inclusion made possible by the fact that one of the signifieds is also a semantic feature of the other signified. Whereas we eliminated all but causal relationships from metonymy, synecdoche is expanded here to include not only the part for the whole but also the container for the contained. (Schofer & Rice 1977: 141)

However, this opinion can be regarded as a rather isolated deviation from the view that part-whole synecdoches are metonymies, a conception which was by then becoming generally accepted even in French speaking countries. In his 1984 article on metonymy, Hugh Bredin distinguished between structural relations and extrinsic relations and claimed that synecdochic relations are structural, and métonymie relations are extrinsic ones (Bredin 1984: 54): The theory I want to propound is that synecdochic relations are structural, and metonymical relations are extrinsic - relations, in the one case, between particulars and their parts, and in the other case between particulars and other particulars. ... It is of great importance, though, to note that what

202 Brigitte Neriich and David D. Clarke

is taken to be the particular in any given case is dependent on the context, and not necessarily upon some inherent nature in things. In other words, the concept particular, as it is being used here, has a strongly epistemic character - as perhaps it always has. For example, "wheel" is a synecdoche for an automobile; but if a racing driver is given the nickname "Wheel", this is a metonymy. In one case the particular is an automobile, and wheels are part of it, structurally related to the automobile as part to whole. In the other case, wheels are a particular, and are extrinsically related to the driver, who is another particular. We must also note, that, in the synecdochic relation of genus to species, the concept of the genus is taken as a particular, even though it is a concept. One reason is that, when we begin to analyze a concept into its constituents, its having those constituents is the having of certain properties: thus, mortal has the property of including the extension of the concept man; it may be said to instantiate the having of that property as a particular instantiates a universal. (Bredin 1984: 54) Relations

structural

synecdoche

simple (metonymy)

dependent

similarity (metaphor)

others

Figure 4. Bredin

However, we would argue with Burkhardt and Seto that kind-of relations or taxonomical relations or category relations are one thing, and part-of relations or partonomical relations or real word relations are quite another (cf. Seto, in press b). Kind-of relations are exploited in synecdoche, part-of relations are exploited (amongst

Synecdoche as a cognitive and communicative strategy

203

others) in metonymy. Both types of relations are relations between constituents or components, but in the case of metonomy these relations link occurrences in the "real world", whereas in the case of synecdoche the relations that are exploited are those which pertain between 'components' of concepts or categories. The crucial difference therefore between the classical and the modern conception of synecdoche ist that part-whole relations as relations in the real world are seen to be the basis for metonymies and are therefore put on the same level as relations between cause and effect, container and contained and so on, whereas genus-species relations as relations based on set-inclusion are considered to be the basis for synecdoches. figures of speech

metaphor

metonymy

including part/whole

synecdoche

genus/species

Figure 5. Esnault, Burkhardt, Seto, etc.

2.

Defining the communicational function of synecdoche

After having looked at synecdoche from a semasiological point of view, we would now like to turn to a more onomasiological perspective. In the following we would like to argue that metaphor, metonymy, and synecdoche, thus defined and delimited, can be regarded as universal cognitive and communicative strategies. We would also like to argue that all three "figures of speech" can be studied on three levels of analysis: on the level of synchrony (or panchrony), on the level of diachrony, and on the level of developmental linguistics (see figure 6). On the synchronic level, synecdoche can be studied as

204 Brigitte Neriich and David D. Clarke

part of the systematic analysis of semantic relations, that is of taxonomy or hyponymy ("the lexical relation corresponding to the inclusion of one class in another", Cruse 1986: 88), as well as on the conceptual level and on the level of discourse or actual language use, to which we shall come shortly (cf. figure 7); on the diachronic level synecdoche can be studied as bringing about particularisation and generalisation or restriction and extension of meaning; on the developmental level synecdoche can be studied as underlying small children's overextensions (cf. Neriich, Todd & Clarke, in press). metaphor

metonymy

synecdoche

synchrony

diachrony

development

hyponymy

particularisation

overextension

Figure 6. The synchronic-diachronic-developmental interface

synecdoche

communicational

structural

conceptual (based on categories)

(based on semantic relations)

^

(based on social relations) Figure 7. The levels of synchronic analysis

Synecdoche as a cognitive and communicative strategy 205

Now, whereas the function of metaphor in discourse is fairly well understood, and the function of metonymy is gradually being unravelled, the function of synecdoche in actual discourse is not so easy to pin down. There is a general consensus that synecdoche is the least figurative of the three figures of speech. As Klinkenberg and Seto (amongst others) have pointed out, synecdoche seems to belong to the normal functioning of ordinary language (Klinkenberg 1983: 291). Seto writes: In advanced countries a TV means a color TV and a fridge means an electric refrigerator. That this is possible is, I suppose, due to the mechanism of the genus-for-species synecdoche. This kind of synecdoche may be related to the so-called maxim of quantity. If this is possible, it seems reasonable to suppose that the genus-for-species synecdoche is not a special figure of speech, but a very general semantic mechanism which governs a wide range of language use .... (Seto, in press a: 7).

Heinrich Lausberg had already pointed out in his Elemente der literarischen Rhetorik that the genus-for-species type of synecdoche is especially frequent "um im Kontext die Wiederholung eines die Art bezeichnenden Wortes zu vermeiden (yariatio...), wobei der bestimmte Artikel als rückverweisendes Bekanntschafts-Signal ... auftritt" [so as to avoid in a certain context the repetition of a word that designates a species (yariatio ...), with the definite article serving as a cataphoric signal of familiarity] (Lausberg 1971: §194). Synecdoche would therefore be one of the mechanisms used to establish coherence in texts (Harweg 1979: 186-187), and could or should be studied by text-linguists. An example, randomly chosen from a Sunday colour-supplement, demonstrates this mechanism nicely. In an article on North Korea, we can read: By the Arch of Triumph tiny girls in white ballet tutus went cartwheeling across lay-bys, bigger girls practised gravity-defying gymnastics routines; never had I seen youngsters so supple and fit. Here, 10,000 kids assembled to march with silver spears; there, 20,000 headed for the 120,000-seat Kim II Sung sports stadium to rehearse a series of dazzling visual tricks with

206 Brigitte Neriich and David D. Clarke

multi-coloured shields. When, finally, we tore ourselves away, the gymnasts kept waving until we had turned the corner and were out of sight. (The Observer, Life magazine, 7 January 1996, p. 10).

The writer can introduce "variation" by going up and down in hyponymic or synecdochic space or by going sideways in choosing from an array of (register specific) synonyms. Having lost its most figurative part, the part-whole synecdoche, to metonymy, synecdoche thus seems to have lost at one and the same time its status as a figure of speech and become a general semantic or stylistic mechanism. The figurative nature of the genus-for-species synecdoche can certainly be revived, as demonstrated in the famous Raymond Queneau-example quoted by the Groupe μ (1970: 103),2 but still, our much reduced synecdoche is certainly the least figurative of our three master-tropes: metaphor, metonymy and synecdoche (Nerlich, in press). However, all is not lost, as apart from its function of variatio in ordinary language, synecdoche seems to have a more specific function in one class of extraordinary language use or in one specific functional style or register. A style or register can be defined in the following way: Stilarten, die sich aus den Funktionen der Sprache (Verständigung, Mitteilung, Wirkung) als gebrauchsfertige (rekurrente) Verwendungsweisen ergeben, z.B. der wissenschaftliche und der umgangssprachliche Stil, der Verkehrs- und Geschäftsstil, der Stil öffentlicher Verlautbarungen, der publizistische und der künstlerisch-literarische Stil. (Lewandowski 1979, I: 217) [Styles which emerge from the various functions of language (understanding, communication, effect) as readily usable (recurrent) ways of language use, as for example the style of scientific discourse and ordinary language, the style of social interaction and commerce, the style of official announcements, the style of publishing and the artistic and literaiy style.]

It seems that synecdoche is particularly common in any type of "official" register, such as "police-speak" or the style of public announcements - and here we might have a hint as to one of the most

Synecdoche as a cognitive and communicative strategy 207

common communicative functions of synecdoche. Exploiting the fact that the genus-species type of synecdoche is always true (whereas in the case of metaphor the truth-function of a word is undermined), and that it is easier and safer to use a generic term rather than a specific one, one can hear utterances such as "I apprehended a person/vehicle on the premises". Synecdoche is used here "as an instrument to avoid introducing too many and especially too specific referents" (Dirven 1993: 22). Not using the generic term can indeed have legal consequences. These uses of synecdoche in police-speak, in aircraft-speak, busspeak etc., in officialese and legalese in general, may strike us as somewhat odd, but they are generally not perceived as "figures of speech". We are dealing here rather with stylistic or register variations, where register is the product of functional variation correlated with contextual variation, as researchers in the Hallidayan linguistic tradition are apt to say (cf. Halliday 1978: 23-31). When we look back at the history of the concept of register itself, we find that John Rupert Firth first introduced this notion under the heading of 'restricted languages' and "looked to 'science, technology, politics, commerce', 'industry', 'sport', mathematics', and 'meteorology', or to 'a particular form or genre', or to a 'type of word associated with a single author or a type of speech function with its appropriate style' or 'tempo' (1968: 106, 98, 112, 118-120)" (Beaugrande 1993: 8). According to Firth, linguistics can regard each "'person' 'as being in command of a constellation of restricted languages, satellite languages' ..., but these are 'governed' by 'the general language of the community' (1968: 207-208)" (Beaugrande 1993: 8). Spinning out Firth's thoughts, one could say that the remoter the satellite or register orbiting round the common core language, the more effective its use in communication, the more impact its use has on the hearer. This remoteness can be based on the prevalent use of metaphors as in poetic texts, or on the prevalent use of synecdoches as in official texts. That is to say, metaphors deviate from ordinary language by jumping or crossing conceptual boundaries, and synecdoches deviate from ordinary language by moving

208 Brigitte Neriich and David D. Clarke

above (or below) basic level concepts (Rosch 1973). Both break "ordinary" expectations. Firth also claimed that language is important in creating persons and said that: The meaning of person in the sense of a man or woman represented in fictitious dialogue, or as a character in a play, is relevant if we take a sociological view of the personae or part we are called upon to play in the routine of life. Every social person is a bundle of personae, a bundle of parts, each having its lines ... the continuity of the person, the development of personality, are paralleled by the continuity and development of language in a variety of forms. (Firth 1950, cit. Matthiessen 1993: 24)

Now, to be able to live life to the full in the social community, people must be able to use and understand a variety of registers or styles. So as to communicate effectively, we must be able to adopt the right role or persona correlated with the right register, we must be able to use a special language for a special purpose, and we must be able to show solidarity with a group, or to express dominance by the use of the appropriate register. How an official persona is created through the use of an official language is nicely illustrated in the following passage. In his novel A Suitable Boy, Vikram Seth describes a bus journey in India in the 1950s in what one can only call a clapped out old bus (but things are very similar today). The driver took particular pride in the signs (in standard Hindi) displayed in the bus. Above his seat for instance, it said Officer Seat and Don't talk to the driver when the vehicle is in motion. Above the door it said: Only disembark when the bus has come to a halt. Along one wall of the bus, the following message was painted in a murderous scarlet: Do not travel when drunk, intoxicated or with a loaded gun. But it said nothing about goats, and there were several in the bus. (Seth 1993: 700-701, our emphasis)

The pride in the bus and the pride in being a bus driver, is derived from the special language used in it or on it, not from the bus's state of repair.

Synecdoche as a cognitive and communicative strategy 209

We were reminded of this passage when sitting on the Nottingham to Heathrow bus and asked by the steward to "put the towels in the receptacle provided when using the toilet". When putting the paper towel in the bin later on we looked at it with different eyes. There is an element of euphemism inherent in this use of synecdoche, but it is based on a different mechanism than the one where we replace bin men with waste disposal operatives for example. In Figure 8 we have tried to correlate three types of stylistic devices with three types of registers or styles. Metaphor is predominantly used in the poetic register, but it can also be used in ordinary language. Metonymy is predominantly used in ordinary language but it can also be used in poetic texts. Synecdoche can ,be used in ordinary language, mostly so as to introduce variation and coherence into texts, but it also characterises official styles. It is therefore the least "figurative" of our figures of speech. Stilmittel - stylistic devices

I

metaphor

metonymy

synecdoche

poetic

ordinary

official [variatio]

Stilform - style, register Figure 8. The main communicational functions/registers of metaphor/metonymy/synecdoche-use

210 Brigitte Neriich and David D. Clarke

In conclusion one can say that the genus-for-species synecdoche plays a vital part in language and life. On the conceptual level it reflects and exploits the order in our categories, on the linguistic or structural level it exploits semantic relations, and on the communicational level it brings order into texts and into social relations.

Notes 1.

2.

At the tum from the nineteenth to the twentieth century, French historical semanticists or those writing in the French language, had started to include part-whole relationships under the heading of metonymy. Examples are Nyrop's (1913) work and, more specifically, Esnault's (1925) work, where he points out that : "Le sémantisme du mot métaphore, c'est Transfer; celui de métonymie Changement (de dénomination); celui de synecdoque Annexion" [The semantics of the word metaphor is Transfer; that of metonymy Change; that of synecdoche Annexation] (1925: 29). "Synecdoque et métonymie ont, en effet, ceci de commun qu'elles respectent le 'cosmos', l'ordre constant des phénomènes naturels, ceci de différent et d'opposé, que la synecdoque considère les êtres par leur classification, leur 'extension', la métonymie par leur activité, leur 'compréhension'" [Synecdoche and metonymy have, in fact, that feature in common that they respect the 'cosmos', the constant order of natural phenomena, they differ and are in fact opposed to each other in so far as synecdoche considers phenomena by their classification, their 'extension', whereas metonymy considers them from the point of view of their activity, their 'intension'] (1925: 35). "Il reprit son chemin et songeusement quant à la tête, d'un pas net quant aux pieds, il termina sans bavures son itinéraire. Des radis l'attendaient, et le chat qui miaula espérant des sardines, et Amélie qui craignait une combustion trop accentuée du fricot. Le maître de la maison grignote les végétaux, caresse l'animal et répond à l'être humain qui lui demande comment sont les nouvelles aujourd'hui: — Pas fameuses." [He continued on his way and, dreamily as to his head, with a clear step, as to his feet, he finished his itinerary without any hitches. The radishes were waiting for him, and the cat mewed in hope of sardines, and Amélie, who feared a marked combustion of the stew. The master of the house nibbled the vegetables, stroked the animal and answered the human being, who asks him how his day has been: — Not brilliant.]

Synecdoche as a cognitive and communicative strategy 211

References Beaugrande, Robert de 1993 'Register' in Discourse Studies: A concept in search of a theory, in: Mohsen Ghadessy (ed.), 7-25. Bonhomme, Marc 1987 Linguistique de la métonymie (Sciences pour la communication 16.). Frankfurt/M. etc.: Peter Lang. Bredin, Hugh 1984 Metonymy. Poetics Today 5/1: 45-58. Burke, Kenneth 1969 A Grammar of Motives. Berkeley / Los Angeles: University of California Press. Burkhardt, Armin 1995, ms. Zwischen Poesie und Ökonomie. Die Metonymie als semantisches Prinzip. To appear in: Euphorien 1996. Cruse, David A. 1986 Lexical Semantics. Cambridge: Cambridge University Press. Dirven, René 1993 Metonymy and metaphor. Different mental strategies of conceptualization. Leuvense Bijdragen 82: 1-28. Esnault, Gaston 1925 L'imagination populaire: Métaphores occidentales. Paris: Presses Universitaires de France. Firth, John Rupert [1950] 1957 Personality and language in society. In: Papers in Linguistics. 19341951, 177-189. London: Oxford University Press. 1968 Selected Papers ofJ.R. Firth 1952-1959. Edited by Frank R. Palmer. London / Harlow: Longmans, Green and Co. Fontanier, Pierre [1821; 1827] Les Figures du discours. Paris: Flammarion. 1968 Ghadessy, Mohsen (ed.) 1993 Register Analysis: Theory and Practice. London: Pinter Publishers. Groupe μ (Jacques Dubois, Francis Edeline, Jean-Marie Klinkenberg, Philippe Minguet) 1970 Rhétorique générale. Paris: Larousse. Halliday, M.A.K. 1978 Language as a Social Semiotic: The Social Interpretation of Language and Meaning. London: Arnold.

212 Brigitte Neriich and David D. Clarke

Harweg, Roland 2 1979 Pronomina und Textkonstitution. (Beihefte zur Poetica 22.) München: Fink. Hopwood, A. T. 1959 The development of Pre-Linnaean taxonomy. Proceedings of the Linnaean Society of London 170: 230-234. Jakobson, Roman 1956 The metaphoric and métonymie poles. In: Roman Jakobson and Morris Halle (eds.), Fundamentals of Language, 76-82. Gravenhage: Mouton & Co. [1956] 1983 Two aspects of language and two types of aphasie disturbances. In: id., (eds.), Language in Literature. Edited by Krystyna Pomorska and Stephen Rudi, 95-120. Cambridge/Mass. / London: The Belknap Press of Harvard University Press. Klinkenberg, Jean-Marie 1983 Problèmes de la synecdoque. Du sémantique à l'encyclopédique. Le Français Moderne 51: 289-299. Lausberg, Heinrich 4 1971 Elemente der literarischen Rhetorik. Eine Einführung für Studierende der klassischen, romanischen, englischen und deutschen Philologie. München: Hueber. Le Guern, Michel 1973 Sémantique de la métaphore et de la métonymie. Paris: Larousse. Lewandowski, Theodor, (ed.) 3 1979 Linguistisches Wörterbuch. Volume 1. Heidelberg: Quelle & Meyer. Matthiessen, Christian 1993 Register in the round: Diversity in a unified theory of register analysis, in: Mohsen Ghadessy (ed.), 221-292. Meyer, Bernard 1993 Synecdoques: Etude d'une figure de rhétorique. Volume 1. Paris: L'Harmattan. Nerlich, Brigitte in press Synecdoche: A trope, a whole trope and nothing but a trope? In: Neal R. Norrick and Armin Burkhardt (eds.), Tropical Truth. Nerlich, Brigitte, Zazie Todd and David D. Clarke in press 'Mummy, I like being a sandwich'. Metonymy in language acquisition. In: Klaus-Uwe Panther and Günter Radden (eds.). Nyrop, Kristoffer 1913 Grammaire historique de la langue française. Volume 4. Sémantique. Copenhagen: Gyldendalske Boghandel Nordisk Forlag.

Synecdoche as a cognitive and communicative strategy 213

Nysenholc, Adolphe 1981 Métonymie, synecdoque, métaphore: Analyse du corpus chaplinien et théorie. Semiotica 34: 311-341. Panther, Klaus-Uwe and Günter Radden (eds.) in press Metonymy in thougt and language. Proceedings of the Hamburg workshop on 'ConceptualMetonymy'. Amsterdam: Benjamins. Rosch, Eleanor 1973 On the internal structure of perceptual and semantic categories. In: Terrence E. Moore (ed.), Cognitive Development and the Acquisition of Language, 111-144. New York: Academic Press. Sato, Nobuo 1979 Synecdoque, un trope suspect. Revue d'Esthétique 1-2: 116-127. Schofer, Peter and Donald Rice 1977 Metaphor, metonymy, and synecdoche revis(it)ed. Semiotica 21: 121-149. Seth, Vikram 1993 A Suitable Boy. London: Phoenix. Seto, Ken-ichi, 1995 On the cognitive triangle: The relation between metaphor, metonymy and synecdoche, ms. in press a Metonymy and the cognitive triangle. in press b On distinguishing metonymy from synecdoche. In: Klaus-Uwe Panther and Günter Radden (eds.). Todorov, Tzvetan 1970 Synecdoques. Communications 16: 25-35. Tversky, Barbara 1990 Where partonomies and taxonomies meet. In: Savas L. Tsohatzidis (ed.), Meanings and Prototypes. Studies in Linguistic Categorization, 334- 344. London: Routledge.

Laws of thought, knowledge and lexical change Beatrice Warren There are universal laws of thoughts which are reflected in the laws of change of meaning ... even if the science of meaning ... has not yet made much advance towards discovering them. (Jespersen 1925: 212)

The concern of the present contribution is the interpretative mechanisms which make it possible for words to take on meanings that are not memorized. The first section of the paper will deal with topics such as meaning and reference, meaning and categorization, criterial and non-criterial features of meaning, meaning and context. This will lead up to a presentation of a model of the process of lexical change. Along with current cognitive approaches, it will be suggested that certain types of mental operations on certain types of knowledge are involved in lexical change. The nature of these operations and these types of knowledge will be described in the section that follows. A distinction is made between three main types of lexical change, viz. novel hyponymic senses, non-literal senses and appended senses.

1.

Preliminaries

1.1. Meaning and reference The advent of cognitive approaches to linguistic theory has brought along certain metaphors for the cognitive activities involved in interpreting figurative uses of words, notably mental space, domain, source, target, mappings, blends, projections. By means of these

216 Beatrice Warren

terms we may describe in some detail how the mind goes about creating a new meaning, in particular figurative meanings. Consider, for instance, the interpretation of the compound land yacht} We can posit that we have two mental spaces representing our concept for 'yacht' (input 1) and for its counterpart on land (input 2). By means of cross-space mappings which will connect water with land, skipper with driver, course with road, yacht with car, tycoon with owner, a new mental space emerges, giving us the meaning 'impressive, luxurious car'. Admittedly, these terms are illuminating. They do provide "a handle on concepts" as Lakoff and Johnson (1980) have taught us good metaphors do. However, the theory of blending is not a sufficient explanation for the interpretation of metaphors. Consider another metaphorical compound, i.e. asphalt beach, which - in Swedish translation happens to be the title of a book. Presumably, all of us have mental spaces representing asphalt and beach; all of us are capable of agile cognitive activity. Yet, I doubt that one can work out the intended interpretation. However, as soon as we are informed that it is an expression employed by New Yorkers for rooftops used for sunbathing, we will no doubt be able to produce a rationale for the compound by means of conceptual blending. By this I wish to illustrate the importance of retrieving a referent in interpreting words. Meaning cannot be created solely by extracting meanings from morphemes, but must involve matching a linguistic unit with a referent (or, as I will presently argue, a class of referents). Lexical meaning, it has been said, is that which takes us from a combination of phonemes to a non-linguistic entity, be it actual or hypothesized. Lexical meaning without a referent is unthinkable.

1.2. Meaning and categorization It was just claimed that if there is no referent, there is no meaning. The reverse is, however, not true. The traditional view is that words

Laws of thought, knowledge and lexical change 217

may have referents without having meanings. Proper names, which have unique reference, are a case in point. For there to be lexical meaning, there must be a set of referents. This is true provided I may use the term referent not only of the entities that nouns pick out, but also of the non-linguistic phenomena that verbs and adjectives denote. Man is a classifying animal, Jespersen said. However, not only man but probably most other animals are completely dependent on being able to form categories to survive. Without categories there would be chaos. The knowledge we may have acquired about one entity would not be transferable to others of the same kind, precisely because we would not see it as a member of a set. It is therefore natural that we have common names for the members of sets and so we can look upon words with lexical meanings as class labels, as do Eleanor Rosch (1978) and her followers, for instance. This approach throws considerable light on the nature of word meaning. Linguists assume that word meaning is inferred. Innate principles of categorization are probably fundamental in this inference. The characteristics that make an entity or phenomenon a member of a category become features of meaning. This approach also explains why it seems so natural to think of word meaning as composed of features (or components), as so many semanticists have done and still do. Moreover, postulating that features of meaning derive from our perception of class-distinctive features of referents, we can see how the generalization and abstracting that we believe is part and parcel of determining word meaning can be done.

1.3. Criterial and non-criterial features of meaning Giving up the idea that there is a connection between our ability to form categories and our ability to form meanings for words would indeed be difficult. However, I do not think that all features of meaning can be of this kind. There are features of meaning that do not specify the kind of entity or phenomenon the speaker wishes to

218 Beatrice Warren

denote, but his or her attitude to it. We must distinguish between features of meaning which are criterial and fix reference and which we can connect with our ability to form categories and features of meaning which do not fix reference and for this very reason express speaker evaluation and which in my view cannot have truth values. Words containing these non-criterial evaluative-attributive features of meaning are of fundamental importance, i.e. words such as good, bad, beauty, folly, wisdom, justice, right, wrong, difficult, easy, danger, safety, etc., etc. Without words containing such features, freedom of speech would be of little value, because we would hardly have verbal means of expressing an opinion. Bolinger expresses this point in the following manner: "... language is more than logic and meaning is more than truth and to exclude other values is to insist that language is nothing but a transmission belt for factual knowledge" (1975: 160). The distinction between defining and evaluative-attributive features of meaning that I have just described is of course reminiscent of the time-honoured distinction between denotation and connotation. I hope, however, that I have made the distinction more precise in that I have pointed out that evaluative features are definitely communicated not vague associations forming a background noise and above all in that I have pointed out that features express speaker evaluation because they do not fix reference. That is to say, it is up to the speaker what entities or phenomena words containing such features apply to. The question left unanswered is what is the source of these features? I have elsewhere suggested (Warren 1992: 19) that the mood of the whole situation in which a word occurs, i.e. more precisely the character of the speaker, the manner in which and the reason for which (s)he uses a word may all contribute to giving it its expressive force. It is claimed that if paralanguage expresses the speaker's dislike but his actual words express approval, we trust the former source of information rather than the latter. For survival it is often no use knowing what kind of thing we are faced with, if we do not also know whether this thing is bad or good or indifferent. Evaluations must be as important as forming categories.

Laws of thought, knowledge and lexical change 219

1.4. Meaning and context In the last two decades or so, the assumption that words have stable meanings which we retrieve ready-made as we come across them in an utterance has been challenged by a number of linguists, probably independently of each other. For instance, Bransford and McCarrell (1974) suggest that words are cues or instructions to create meaning. Fillmore thinks of sentences as "blueprints" of which the interpreter constructs interpretations (1985), Sperber and Wilson (1986) wish to replace the traditional code model with an inferential model of interpretation. The Norwegian philosopher-linguist Rommetveit (1988) insists that words do not have literal, basic, invariant meanings, but meaning potentials. I myself have suggested (Warren 1992) that we must distinguish between the meaning of a word out of context and its meaning in context. The former type of meaning I refer to as dictionary meaning. It is the meaning that the lexicographer would be interested in. The latter type of meaning I refer to as contextual meaning, which is the value we give a word in context. Further, I have suggested that whenever we come across a word in context, we contextualize its meaning and reference. This contextualization is a matter-of-course process in interpretation. It involves a readiness to adapt the dictionary meaning if necessary and if at all possible so that it fits the context in question. The question is what modifications are possible. I will presently attempt to address this issue, but first I wish to point out that the contextualization process is normally followed by a decontextualization process. This process may involve new meanings being generalized further or, if they are deemed to be of nonce value, scrapped altogether, which is probably frequently the case. However, sometimes they may be considered useful for some reason and be memorized. This memorized new meaning may spread to a sufficiently large group of people to become a new dictionary meaning. In other words, a distinction is made between a new contextual meaning, which is part of parole and the work of a moment, and a new dictionary meaning, which is part of langue and the creation of which is gradual. The distinction between

220 Beatrice Warren

contextual and dictionary meaning is hardly controversial. Nevertheless it is necessary to make it explicit. This approach makes it possible to agree that the traditional view of clearcut, precise senses should be abandoned and at the same time maintain that dictionary meanings are stable, although sometimes rather shapeless. It is when words occur in a context that their meanings become flexible and may assume a specific shape. This approach also makes a radical monosemic position à la Ruhl (1989) superfluous, i.e. the view that words have single abstract senses from which contextual senses are derived. The assumption that the creation of contextual senses is compulsory does not preclude the possibility that some words have several memorized senses. Needless to say, the flexibility of word meaning in context serves a purpose: it enables great precision in communication. The points of this section can be summed up as follows: i. ii.

iii.

The source of lexical meaning is our perception of the character of the referent. In principle words with lexical meaning will not have unique reference. Words with purely referential meaning can therefore simply be considered to be category labels. To form a category normally involves perceiving at least one common trait, disregarding dissimilar ones. Hence it is natural to think of meaning as componential. We must distinguish between features of meaning which fix reference. That is to say, those about which there is consensus among the members of a language community as to what kind of entities and phenomena they denote (criterial features) and features of meaning concerning which the speaker decides what entities or phenomena they are applicable to (evaluativeattributive). Criterial features derive from man's need and ability to form categories; evaluative-attributive features derive from man's need and ability to form judgements.

Laws of thought, knowledge and lexical change 221

iv.

We must distinguish between out-of-context meanings of words which are static and meanings of words in context which are flexible. The fact that word meaning is modifiable allows precision with least effort.

2.

A model of interpretation of words in context

As illustrated in Figure 1, knowledge of the dictionary meaning and/ or the context - both linguistic and non-linguistic - and knowledge of worldly "things" will suggest a plausible referent (or some plausible referents). The dictionary meaning (provided we have one) will then be contextualized to confirm that it does fit or can be made to fit the character of the favoured referent or referents, a process which will finally yield some chosen contextual meaning and referenti). 2

Figure 1. A model of interpretation of words in context

If we apply this model to the example asphalt beach above, we see that we did not arrive at any good idea of its reference and meaning because we did not have sufficient context to access our knowl-

222 Beatrice Warren

edge that people sometimes sunbathe on rooftops, which, had we had, predictably would have led to the metaphorization of beach. As already pointed out, it is normally the contextualization process that gives rise to new meanings. The question is then: in what ways can dictionary meanings be modified to yield new meanings. Assuming that certain natural strategies representing universal cognitive processes will be appealed to, we can predict that interpreters will of course assume that the speaker has used a particular word because (s)he believes that it will enable interpreters to retrieve the intended referent(s). Therefore there must be some connection between dictionary and contextual meaning and reference and this connection must be deemed so self-evident that - given the particular context - it can be worked out. It is to the description of these types of connections that we now turn.

3.

Creating hyponyms

Lyons makes use of the term denotatum, which he defines as "the class of objects, properties, etc. to which a word correctly applies"3 (1977: 207). The members of the denotatum of a word are mental entities, whereas the referents of a word are normally actual entities connected to a particular context. Applying this distinction to my approach, we can say that the dictionary meaning of a word is that which takes us to its denotatum, the contextual meaning of a word is that which takes us to its referent or referents. Whereas the denotatum of a word encompasses all possible members in the set, a word in context often refers to one or some members. If context suggests that the entities referred to must have some property or properties over and above the features specified by the dictionary meaning of the word in question, there will be a natural addition of features of meaning creating a subcategory. If such a subcategory is of lasting interest, such contextual meaning may become conventionalized. Indeed, examples of conventionalized specialized meanings are not hard to find. Missile is not used to denote

Laws of thought, knowledge and lexical change 223

any thrown object but often about ballistic robots. Drug often suggests narcotics and when we say of a woman that she is liberated, we do not simply mean that she is freed but that she is freed from traditional ideas concerning woman's role in society. In literature on semantic change, it is often pointed out that narrowing of meaning is a common type of lexical sense shift and that it is paralleled by widening of meaning. Whereas, needless to say, I naturally accept that narrowing is natural in interpretation, I am not so convinced that widening of meaning is a possible interpretative strategy. Consider the following text: In engineering it is rare to find iron used in its pure form. Generally the metal is alloyed with carbon and other elements to form wrought iron, steels and cast iron.

The definite article preceding metal signals to the reader that something which has already been mentioned is referred to again. Going back in the text, the only possible coreferent of the metal is the referent of iron, but in order for metal to be co-referential with iron, it must be specialized to mean "that kind of metal that is iron". That is to say, we have an unconventional contextual meaning and referent for metal. If we reverse the order of iron and metal in this text so that it runs as follows: In engineering it is rare to find metal used in its pure form. Generally the iron is alloyed with carbon and other elements to form wrought iron, steels and cast iron.

we see that we have difficulties in making iron co-referential with metal. This suggests that generalization of meaning is not a natural interpretative mechanism. In other words, it seems that we can constrain the reference of - e.g. - fruit in a context so that it refers exclusively to, say, apples (and not oranges, pears or bananas), but we cannot generalize the reference of apple to include oranges and pears and bananas, etc.

224 Beatrice Warren

I have just suggested that one of the conditions for making a word acquire some non-conventional reference and meaning is that there is such a well-established connection between the non-conventional reference and/or meaning and its conventional denotatum and meaning that the encoder can trust his audience to access it. The connection between a superordinate term and one of its hyponym is of this kind. However, as I try to illustrate below, this link can be used in one direction only. that which is a kind offruit/metal-> viz. in this context "apple", "iron" •that which apple/iron is a kind of -> viz. in this context "finit", "metal"

This may seem strange at first sight, but perhaps the explanation is simply the following: if you are told to look for a credit card in a purse which is in a handbag, it would be very strange to construe this as an instruction to look for the card anywhere in the handbag but not in the purse. Our assumption must be that it is somewhere in the purse. Perhaps I should emphasize that I do not deny that generalization of meanings occur. They demonstrably do. What I doubt is that generalization is an interpretative mechanism. In other words, generalization must be the result of some interpretative mechanism and not a mechanism itself.

4.

Creating non-literal senses

The rule that I have just suggested, i.e. that the entity or phenomenon referred to should be within the denotatum of the word in question, is a rule which we all know can be violated and indeed often is. Again, provided there is some decipherable connection between the conventional and non-conventional meaning and reference, we have then what is recognised as a non-literal use of a word. According to tradition, as is illustrated in Table 1, this connection may be (i) one or more reminiscent properties, in which case we

Laws of thought, knowledge and lexical change 225

have a metaphor, (ii) a polarized property, in which case we have reversals, (iii) one or more properties which need to be down- or upgraded, in which case we have hyperbole and litotes, respectively (I have put hyperbole and litotes within parenthesis for reasons I will clarify later on), or (iv) a relation of contiguity, in which case we have metonymy. We see here reflected the two basic ways in which associations may be formed: either X and Y have common or reminiscent properties (as in metaphors, reversals, hyperbole and litotes) or X and Y form a pair by having co-occurred in time and/or space (as in metonymy). I will clarify also this point later on. Table 1. Survey of non-literal senses Connection

Sense extension

•

reminiscent property

•

• •

polarized property (property intensity of which need to be up- or downgraded) relation of concomitance

• •

metaphor (memory of a computer) reversal (a fine friend) (hyperbole, litotes)

•

metonymy (hunchback)

•

As we have seen, the difference between metaphor and metonymy is traditionally that metaphor involves relations of resemblance, whereas metonymy involves relations of concomitance. I have elsewhere emphasized that the interpretation of metaphors involves retrieval of properties, whereas the interpretation of metonyms involves retrieval of relations and that this is the important difference (Warren, in press a and b). Metonymy, I have also suggested, is basically an abbreviation device. Consider some well-known examples of metonymy: (1) (2) (3)

She heard the piano from next door. The kettle is boiling. That's Churchill, (pointing to a sculpture)

226 Beatrice Warren

Paraphrasing the interpretation of these, we get: (1) (2) (3)

a. She heard (the sound produced) by the piano. a. The (water in the) kettle is boiling. a. That's (a sculpture representing) Churchill.4

We see here that what is left out is that which is so closely connected with what is mentioned that it goes without saying in a context which requires its retrieval to make sense. What is left of the noun phrase has greater information value relative to that which is left out. Let us now compare (4) and (5) below: (4) (5)

Arthur is Hamlet in this play. {Hamlet = metonym) Arthur is a real Hamlet. (Hamlet = metaphor)

In (4) we have - according to my definition - a metonym, in (5) a metaphor. The interpretation of Hamlet in (4) can be accounted for by supplementing this noun phrase as in (5): (4)

a. Arthur is the actor representing Hamlet in this play.

It is not possible to account for the interpretation of Hamlet in (5) in similar fashion. In order to interpret Hamlet in (5), we need to retrieve at least one property shared by Arthur and Hamlet, for instance that Arthur is a brooding young man. As we see from this particular example, as well as from example (3), a métonymie relation can be iconic. The relation between the sculpture of Churchill and the person Churchill could surely be said to be one of resemblance. According to the traditional definition, Churchill in (3) would consequently be a metaphor and so would Hamlet both in (4) and in (5). Since, according to my analysis, we employ different strategies working out interpretations for Hamlet in (4) and (5), it seems reasonable that we separate them also terminologically.5

Laws of thought, knowledge and lexical change 227

To strengthen my argument I ask the reader to consider one more example, viz.: (6)

Arthur has his father's eyes. the interpretation of which spells out as:

(6)

a. Arthur has (eyes which resemble) his father's eyes.

Since it is possible to account for the interpretation of the noun phrase his father's eyes simply by expanding it without having to retrieve shared features, I do not hesitate to think of it as a metonym in spite of the resemblance relation. So far I have contrasted my approach to the metaphor-metonymy distinction with the traditional definition. Let me also contrast it with the favoured current approach which is that metaphor involves mapping across domains, whereas metonymy involves mapping within a domain (Lakoff and Johnson (1980) and Lakoff and Turner (1989), further developed by Croft (1993)). I have some problems with this: one is that I fail to see much difference between claiming that metaphor involves mappings across domains and the traditional view that metaphor involves seeing similarity in dissimilarity. And if we are to see, for instance, ecstasy and a particular drug as belonging to the same domain, is that not because experience has supplied an associative link between them? And is that then radically different from the traditional claim that metonymy is based on contiguity? Another difficulty is the looseness of the term domain. How do we determine whether entities belong to the same domain or not? Personally I would have imagined that all handbags belong to the same domain, but if I were to hear somebody say of a handbag which this person found elegant that it is a real Guzzi, although it manifestly is not, I would take a real Guzzi to be a metaphor and be willling to form two domains: one with Guzzi handbags and one comprising all other handbags. The point I am trying to make is that since domains do not seem to be static and invariant constructions of

228 Beatrice Warren

our experience but adaptable to context (on this see Ungerer and Schmid (1996)), surely it is the interpretation we favour that induces the formation of domains and not the domains that basically induce interpretations. I concede, however, that the creation of domains enables access to features intended to be communicated in this case "elegant". Finally, I fail to see that viewing metaphor and metonymy in or across domains can account for important differences between these two different figures of speech. Viewing metonymy as basically an abbreviation device, on the other hand, we can explain why, for instance, metonyms comparatively often lack expressive force (as the examples cited here demonstrate) and comparatively rarely supply names for unnamed entities. Metaphors are different in this respect: they either are used to name (at least on word level) or to have expressive force or both functions simultaneously. (For other differences between metaphor and metonymy which remain unaccounted for by the domain approach, see Warren (in press a).) I conclude this subsection by giving the promised explanation of why hyperbole and litotes are within parenthesis in Table 1. As already pointed out, we have hyperbole if, in fitting the conventional meaning of the linguistic unit to our favoured referent, we find that a match is only possible if the value of some feature of meaning is decreased in strength. Similarly, we have litotes if we find that a match demands that the value of some feature of meaning is increased in strength. However, the encoder need not find any such down- or upgrading necessary. In other words, there may be disagreement between encoder and decoder as to whether adjustment of the conventional sense is called for. For this reason, it is somewhat problematic to look upon hyperbole and litotes as invariably non-literal uses of words.

Laws of thought, knowledge and lexical change 229

5.

Creating appendant meanings

Finally I would like you to consider the following three examples: (7) (8) (9)

It won V happen while I still breathe, [live] - Why didn 't you pick the tomatoes I asked you for? - They were still green, [unripe] He ought to know. He is a professor, [expert]

The first example is offered by Halliday (1985: 319-320) as an example of metonymy. If breathe in (i) is a metonym, then green in (ii) and professor in (iii) ought to be, too. Yet none of these examples fit my criteria for metonymy. It does not seem possible to think of them as abbreviated noun phrases, nor as non-literal. I concede, however, that just as so often is the case in metonymy, we are induced to go from conventional to non-conventional sense and reference because of concomitance relations. The difference is, however, that in the case that I term metonymy, the concomitance relation is between entities, whereas in the examples above, the concomitance relation is between propositions, which translates into an antecedent-consequent relation. That is to say in these cases: • if one breathes, then one lives • if tomatoes are green, then they are probably unripe • if somebody is a professor, then (s)he is probably an expert of some kind I therefore refer to senses arising from antecedent-consequent relations as implications. Implications do not replace the senses from which they derive. Consequently green, breathe, professor in the examples above can be said to have two senses: the conventional explicit sense and the appended implicit sense. Provided the implication is not absolute, this appended sense is cancellable as we see in (10):

230 Beatrice Warren

(10) The tomatoes are green [unripe], but strangely enough ripe anyway. This is not possible in the case of my version of metonymy: (11) *The [water in the] kettle is boiling, but not the water. Whereas implications will not cause violations of truth conditions, métonymie senses will, since that which the predication concerns is left out: it is not the kettle, but the water in it that is boiling. Implied senses can, however, be conventionalized and no longer be dependent on a particular standard sense to be elicited. It is possible to hang up on somebody by putting down the receiver; it is possible to go to the toilet behind a bush; it is possible to go to bed in a tent without a bed. Sense changes of this kind have not infrequently been described in the literature. In English the most famous example is possibly that of how today's meaning of beads developed from gebet, which used to mean 'prayer', into meaning 'balls of a necklace': if one told one's prayers, one would be likely to tell (i.e. count) the balls of the rosary (Stern (1965)). My main reason for insisting on keeping implications terminologically apart from the non-literal métonymie senses is that here we have a case in which conventional and novel senses may peacefully co-exist. The fact that we do not have to choose either a conventional or a novel sense makes a gradual switch-over from a conventional to a novel sense possible. One consequence of this is that there may be remnants of the conventional sense affecting the novel sense for a long time. It is often pointed out that the grammaticalization process is gradual and that there tend to be constraints in the new grammatical item connected with the original lexical sense (Hopper and Traugott 1993). Implication can account for these features of grammaticalization. Consider, for instance, be going to as a future tense auxiliary, the meaning of which is said to be 'future culmination of present intention or cause' (Leech 1977: 54).

Laws of thought, knowledge and lexical change 231

6.

Summing up

This account of lexical change is not complete in that I have concentrated on changes dependent on context, these being by far the most frequent ones. That is to say, I have not considered changes induced by changes in our perception of the class-distinctive features of the members of the sets. This is a process referred to as substitution by Stern (1951) and linguistic conservatism by Ullmann (1957) - conservatism because we keep the name although the definition has changed: pens, for instance, are no longer parts of wings of birds. Nor have disseminating processes such as (notably) analogy been considered. Analogy may cause a novel sense to spread from one word to others with which this word is related by means of, for instance, synonymy, hyponymy, antonymy or derivation. There are also other processes by means of which new meanings are created that have been ignored. However, the main processes are the ones that have been brought up. The account is basically cognitive in spirit in that it is accepted that in making words assume meanings, perception and categoryforming principles, context and world knowledge (be it frames, scenes, schémas, cognitive categories or domains) are essential ingredients. I have, however, found weaknesses both in traditional and cognitive approaches. That is to say: i.

ii.

iii.

The cognitive linguists' rejection of objectivism (on this, see Lakoff (1988)), is not sufficient to account for the distinction between criterial and non-criterial features of meaning, a distinction which is of fundamental importance in semantics. Although I accept that widening of meaning may be a result of interpretative strategies, there are reasons to doubt that it is an interpretative mechanism per se. I have not accepted that the crucial difference between metaphor and metonymy is that the former is based on iconic relations, whereas the latter is based on contiguity relations, or

232 Beatrice Warren

iv.

that the former involves mapping across domains, whereas the latter involves mappings within a domain, I have not accepted a wide definition of metonymy but argued that the term metonymy should be restricted to non-literal referential senses. I have pointed out that the literal types of metonymy, which I would term implications, are different from the non-literal ones, mainly in that these senses are based on antecedent-consequent relations between propositions and occur originally as appended senses.

Notes 1.

2.

3.

4.

5.

I owe the example to Gilles Fauconnier, who brought it up at a lecture given in Stockholm in June 1996. For an account of his theories, see Fauconnier (1985). I have presented retrieval of some plausible referent and the contextualization process as being sequential. These two processes may, however, be more or less simultaneous. Note that this definition can only apply to purely referential words, since words with attributive-evaluative features have no correct application, this being the very feature that enables words to convey attitudes. The manner of paraphrasing I have suggested is more precisely: She heard that which the piano produced... That which is in the kettle... That's something which represents Churchill... This mode of paraphrasing demonstrates better how knowledge of context and interpretative strategies collaborate to produce a plausible referent. That which represents Churchill could, e.g., refer to a photo, a painting, an actor, etc. depending on context. Nunberg (1996) and Jackendoff (1997) consider similar examples as cases of metonymy, although they prefer the term meaning/reference transfer.

References Bolinger, Dwight 1975 Aspects of Language. New York etc.: Harcourt, Brace and World.

Laws of thought, knowledge and lexical change 233

Bransford, John and Nancy Mc Carrell 1974 A sketch of a cognitive approach to comprehension: some thoughts about understanding what it means to comprehend. In: Walter Weimer and David Palermo (eds.), Cognition and the Symbolic Processes, 189-229. Hillsdale, New Jersey: Lawrence Erlbaum Associations. Croft, William 1993 The role of domains in the interpretation of metaphors and metonymies. Cognitive Linguistics 4: 335-370. Fauconnier, Gilles 1985 Mental Spaces. Cambridge, Mass.: The MIT Press. Fillmore, Charles 1985 Frames and the semantics of understanding. Quaderni di Semantica 6: 222-254. Halliday, Μ. Α. Κ. 1985 An Introduction to Functional Grammar. London etc.: Edward Arnold. Hopper, Paul and Elizabeth Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Jackendoff, Ray 1997 The Architecture of the Language Faculty. (Linguistic Inquiry Monographs 28.) Cambridge, Mass.: The MIT Press. Jespersen, Otto 1925 Menneskehed, nasjon og individ i sproget. Oslo: Aschehoug. Lakoff, George 1988 Cognitive semantics. In: Umberto Eco, Marco Santambrogio and Patrizia Violi (eds.), Meaning and Mental Representations, 119-54. Bloomington / Indianapolis: Indiana University Press. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Lakoff, George and Mark Turner 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: The University of Chicago Press. Leech, Geoffrey 1976 Meaning and the English Verb. London: Longman. Lyons, John 1977 Semantics. 2 volumes. London / New York: Cambridge University Press. Nunberg, Geoffrey 1996 Transfer of meaning. In: James Pustejovsky and Branimir Boguraev (eds.), Lexical Semantics. The Problem of Polysemy, 109-133. Oxford: Clarendon Press.

234 Beatrice Warren

Rommetveit, Ragnar 1988 On literacy and the myth of literal meaning. In: Roger Säljö (ed.), The Written World, 13-40. (Springer Series in Language and Communication 23.) Berlin / Heidelberg: Springer-Verlag. Rosch, Eleanor 1978 Principles of Categorizations. In: Eleanor Rosch and Barbara B. Lloyd (eds.), Cognition and Categorization. 27-48, Hillsdale, New Jersey: Lawrence Erlbaum. Ruhl, Charles 1989 On Monosemy: A Study in Linguistic Semantics. Albany: State University of New York Press. Sperber, Dan and Deidre Wilson 1986 Relevance: Communication and Cognition. Oxford: Basil Blackwell. Stern, Gustav [1931] 1965 Meaning and Change of Meaning. Bloomington: Indiana University Press. Ullmann, Stephen [1951] 1957 Principles of Semantics. Oxford: Basil Blackwell. Ungerer, Friedrich and Hans-Jörg Schmid 1996 An Introduction to Cognitive Linguistics. London / New York: Longman. Warren, Beatrice 1992 Sense Developments. Stockholm: Almqvist & Wiksell. in press a What is metonymy? In: The proceedings of The International Conference on Historical Linguistics, August, 1995. in press b No more ham sandwiches, please. In: Günter Radden and KlausUwe Panther (eds.), Metonymy in Language and Thought. Amsterdam / Philadelpia: Benjamins.

Section III Case studies

Intensifies as targets and sources of semantic change Ekkehard König and Peter Siemund

1.

How regular is semantic change?

In contrast to the established view on sound change and morphological change, which seem to be governed by clear restrictions and principles, the prevailing view on semantic change has been for a long time and probably still is that "each word has its own history" (J. Grimm, J. Gilliéron). This pessimistic view has been seriously called into question by some recent studies, which have shown that it is possible to formulate some pervasive generalizations about semantic change. First of all, it has been shown that all semantic changes are instances of a very limited set of possible processes, such as metaphor, metonymy, ellipsis, narrowing, broadening, etc. (Traugott 1985; König/Traugott 1988; Blank 1997; Koch 1997). Secondly, it is possible to formulate at least some generalizations about the source domains and target domains involved in such changes, in particular, with regard to metaphorical change. The structuring of temporal domains in terms of spatial notions (SPACE TIME) or, more generally, the pervasive change from concrete to abstract are cases in point. Thirdly, various attempts have been made to formulate some generalizations about possible directions of semantic change. E. Traugott's hypothesis that meanings grounded in the socio-physical world of reference develop into meanings grounded in the speaker's world and further into markers of metatextual attitude to the discourse is one of the best known examples (Traugott 1989,1995). That semantic change is far from erratic or random and is in fact general and regular to a certain extent is particularly obvious if the focus of investigation is on minor lexical classes (function words)

238 Ekkehard König and Peter Siemund

rather than on the major word classes (Ν, V, Adj, P). The semantic development of such function words does not only reveal some general tendencies of semantic change within a particular language, but also allows cross-linguistic generalizations to be made about typical source and target domains and possible directions of change.

2.

The meaning of intensifiers

Since this paper is centrally concerned with general processes of semantic change involving intensifiers either as target or as source, our analysis has to begin with a brief analysis of their meaning (cf. Edmondson/Plank 1978; König/Siemund 1996a, b). Intensifiers are expressions like the following: (1)

Germ, selbst·, Russ. sam; It. stesso; Engl, x-self, Fr. x-même, etc.

There is no general agreement as to how such expressions should be categorized. We find at least a dozen different labels in grammar handbooks or specific studies. The reason for choosing the term intensifiers is not that it is particularly illuminating, but simply that it avoids misleading connotations, in contrast to terms like emphatic reflexives. The morphological properties of such intensifiers differ considerably across languages: Intensifiers may be invariant (German, Mandarin, Japanese) and they may inflect for some or all of the following features: person, gender, number and case (Turkic, Slavic, Romance, English). In the latter case they manifest adjectival behaviour and typically exhibit agreement with a nominal co-constituent. Among their syntactic properties there is one that seems to be the most reliable criterion for identifying them across languages: In nearly all of the fifty languages investigated so far we found that intensifiers are used as adjuncts to noun phrases, i.e. they combine with a noun phrase to form another noun phrase:

Intensifiers as targets and sources of semantic change 239

(2)

a. I would like to talk to the Pope himself. b. The work of Picasso himself is what I came to see. c. The work of Picasso itself...

Furthermore, there seem to be semantic criteria which clearly allow their identification across languages. Intensifiers seem to have at least two uses: an adnominal one and an adverbial one. In addition to functioning as adjuncts to noun phrases (cf. (2)), intensifiers can also be used as part of a verb phrase: (3)

a. The President wrote that speech himself. b. I had a car like that once myself.

As is shown by examples like (3a-b), many languages further differentiate between two such adverbial uses: (a) an exclusive use, roughly paraphrasable by 'alone', 'without help' and (b) an inclusive use, paraphrasable by 'too'. Since the relevant contrast is not found in all languages, however, it will play no role in the subsequent discussion. To describe the meaning of adnominal intensifiers simply as expressing 'emphasis', 'intensification' or 'unexpectedness' is clearly not very illuminating. Examples like (2) show that such expressions evoke alternatives to the denotation of the expression they combine with. Since this property is generally associated with focusing (identificational focus), it suggests that these expressions should be regarded as focus-sensitive adjectives or particles. Perhaps the evoking of alternatives is then a property of the focusing intensifier are associated with, and the co-constituent they combine with can be analyzed as their focus. This is exactly the view that will be taken here. What then is the real contribution that an adnominal intensifiers makes to the meaning of a sentence? To answer this question for examples like (2a), we need to ask what the alternatives evoked in such cases are. A plausible answer certainly is that they are people associated with the Pope in some way: his Cardinals, his secretary, his collaborators or clergy. An assertion of (2a) is thus tantamount to asserting that the speaker does not want to talk to any of those people

240 Ekkehard König and Peter Siemund

from the environment of the Pope. Generalizing from such examples, it is suggested in König (1991) that adnominal intensifiers associate a periphery of alternatives (Y) with a center (X), constituted by the denotation of their focus. Building on this idea, Baker (1995) has distinguished several types of such relations between a center (characterized as 'discourse-prominent' by Baker) and a periphery, which can be illustrated by examples like the following: (4)

a. Bill Clinton himself will sign the document. b. The passengers got away injured. The driver himself died on the scene of the accident. c. Mary's husband looks after the children. Mary herself has a regular job. d. He was not particularly tall, a little taller than Jemima herselfperhaps, but his shoulders in the tweed suit were broad, giving an air of authority, and he himself if not exactly heavy, was certainly a substantial man. [Antonia Fraser, A Splash of Red, 1981: 88]

A center can be established on the basis of hierarchies in the real world (cf. (2a), (4a)). Intensifiers in their adnominal use can thus always be combined with expressions for people of high rank. In fact, it seems that these are the examples that we think of first whenever we are asked to produce examples with adnominal intensifiers. In our modern egalitarian societies central roles are more typically temporary, rather than permanent. A driver is central, in contrast to the passengers, in all matters of driving and safety, a guide is central during the time of an excursion, etc. (4b). Centrality may also be the result of taking a certain person as point of departure for the identification of others, as in (4c). Note that the intensifier in this example could not be combined with the expression Mary's husband. Finally, a person may be used as the center of observation or perspective in a narrative and become central in that sense. The final example (4d) is a case in point. Following Baker (1995), the conditions for using adnominal intensifiers can thus be summarized as follows:

Intensifier s as targets and sources ofsemantic change 241

(5)

conditions for the use of adnominal x-self (Baker 1995) a. X has a higher rank than Y on a real-world hierarchy b. X is more important than Y in a specific situation c. Y is identified relative to X (kinship terms, part-whole, etc.) d. X is the subject of consciousness, center of observation, etc. (logophoricity)

In German there are expressions often used in captions which nicely illustrate the points made above: (6)

Maria selbdritt Mary self-three 'Mary together with two persons'

In contrast to the adnominal use, the adverbial (exclusive) use of intensifiers will not play a prominent role in the subsequent discussion. We will therefore only say that sentences with exclusively used adverbial intensifiers make an assertion which can be roughly represented by the relevant sentences without intensifiers, but with an information structure in which the agent subject is focused against the rest of the sentence as background. Moreover, there is a presupposition to the effect that it is the denotation of the subject which profits or suffers most as a result of the relevant action: (7)

a. The President wrote the speech himself. b. [[the President^ wrote the speech] (assertion) c. The President is primarily affected, (presupposition)

The expressions listed in (1) are the most prominent representatives of intensifiers in the languages mentioned, i.e. they are the ones most frequently used, exhibiting the fewest restrictions and the greatest versatility in their use. The list given in (1) should not be taken to suggest, however, that there is only one single intensifier in each language. The following examples from German show that languages typically have more than one intensifier, even if their number is very limited in each case:

242 Ekkehard König and Peter Siemund

(8)

(höchst)selbst, (höchstpersönlich, eigen, leibhaftig, in Person, von selbst, von sich aus, an sich, (von) allein

3.

Processes of semantic change with intensifiers as targets and sources

In this section only a brief overview of all changes involving intensifiers as sources or targets will be given. Each of the changes listed will then be discussed in detail in a subsequent section. Intensifiers frequently derive from expressions for body parts. Leibhaftig in German is the clearest example we can find in that language. Intensifiers may further develop into reflexive anaphors. The development of reflexive pronouns in English as a result of a fusion between personal pronouns and the intensifier self (him + self himself) as well as similar developments in Afrikaans are clear examples. (9)

BODY PARTS

INTENSIFIERS

REFLEXIVE ANAPHORS

Intensifiers may also, so it seems, adopt the meaning of their focus and be used in the sense of 'master', 'boss', etc., i.e. they may come to express roles of high rank, normally expressed by a typical focus. It is a very similar development that gives rise to the use of intensifiers as polite or honorific pronouns, a usage that is found in Turkish and Japanese, for example. Given such developments, it also is not surprising that intensifiers should have a euphemistic use, as is found in the Celtic areas of the British Isles. (10)

INTENSIFIERS

SOCIAL ROLES

HONORIFIC PRONOUNS

EUPHEMISMS

A very different development is indicated by the polysemy found in many languages for intensifiers. In addition to the adnominal use discussed above, many languages use the same expressions prenominally as scalar focus particles expressing unlikely instantiations

Intensifìers as targets and sources of semantic change 243

of values in certain contexts. In other words the relevant meanings correspond to Engl. even. (11)

INTENSIFÌERS

SCALAR ADDITIVE FOCUS PARTICLES

In German selbst exhibits both uses, in contrast to selber, which is only used as an intensifier: (12) a. Der Papst selbst/selber wird uns besuchen. 'The Pope himself will come to visit us.' b. Selbst der Papst wäre hier ratlos. 'Even the Pope would not know what to do.' Especially the Romance languages show that intensifiers may be involved in the development of demonstrative pronouns (Sp. ese). From these demonstratives, definite articles and personal pronouns may develop (Sardinian, Catalan). (13)

INTENSIFIERS

DEMONSTRATIVE PRONOUNS

DEFINITE ARTI-

CLES / PERSONAL PRONOUNS

Finally, we also find that intensifiers frequently develop into expressions indicating type identity (Sp. mismo).

4.

From expressions for body parts to intensifiers

A wide variety of languages for which the relevant connection is still visible synchronically show that intensifiers often develop from expressions for body parts ('body', 'soul', 'head', 'eye', 'bone', 'person' (< persona 'mask'). The following examples illustrate this historical and synchronic connection (cf. Moravcsik 1972): (14) Arab, ayn 'eye'; Arab, nafs 'soul'; Amharic ras- 'head'; Georgian tviton, tavi 'head'; Germ, leibhaftig·, Hausa ni dakaina Ί with my head'; Hebrew etsem 'bone'; Hung, maga 'seed'; Jap.

244 Ekkehard König and Peter Siemund

zisin 'body'; Okinawan duna 'body'; Rum. insumí 'person'; etc. Given such convincing cross-linguistic evidence, Grimm's (1967) hypothesis that Germ, selb(st) derives from si-liba 'his body' is not all that implausible, even if it is not folly convincing. For most intensifiers in European languages the etymology is unclear, except for those cases where the source is provided by Lat. persona (< 'mask'), such as personally, in person, etc. To reconstruct the path leading from expressions for body parts to intensifiers is no easy matter, given that we only have the beginning point and endpoint, but no information on stages in between. It is quite plausible to assume that the relevant change started in contexts like (2a) and (4a) and thus was essentially based on hierarchical structures in social groups. Persons of high rank could get things done by proxy. A formulation like "The duke did that" was quite appropriate for such scenarios. Only their presence ("in person") was remarkable and noteworthy. The addition of such a proto-intensifier to a noun had the effect of eliminating the entourage of the highranking person as a possible referent and thus had the overall result describable as semantic narrowing. The next step could have been a métonymie change. The intensifier picked up the feature of characterizing a referent as central from its context and thus came to be associated with imposing a structure on a set of persons including the referent of its co-constituent and contextually given alternatives in terms of center vs. entourage and later center vs. periphery. If intensifiers were originally only used in combination with the names of persons of high rank, the other three uses listed in (5) can be assumed to be later developments. They would be, in fact, another instance of a change towards further subjectivization in the sense of Traugott (1989). These assumptions are supported by a substantial body of evidence: i.

In the texts of older periods of European languages, intensifiers are primarily used with names of high-ranking persons. In Old English, for example, self-was primarily used with nouns

Intensifiers as targets and sources ofsemantic change 245

ii.

iii.

5.

like Crist, Haeland, God, Drihten, deofol, cyning, etc. (Farr 1905: 19). In many languages at least some intensifiers can only be combined with a human focus. Examples are Turk, kendi, Jap. zisin, Mand. ziji, Bengali nije, It. in persona. In other words, here we find a situation which is closer to the original selezionai restrictions. In other languages intensifiers can be assumed to have extended their territory to foci of all types. Some languages seem to have developed new expressions which recreate, as it were, the original selectional restriction, i.e. that only combine with expressions for persons of exalted rank: höchstpersönlich, höchstselbst in German, selveste in Norwegian, sahst in Arabic, in persona in Italian, etc. That these new expressions are more complex than the original intensifiers, whose combinatorial potential had been extended, is to be expected. This fact finds a parallel in the frequent phenomenon that new, more complex local prepositions are created once the older ones are primarily used in a temporal sense.

From intensifiers to reflexive anaphors

A wide variety of languages use the same expression both as intensifier and as a marker of co-reference in a local domain, i.e. as reflexive pronoun. In Europe this phenomenon is relatively rare, but it can be found in the Finno-Ugric and Celtic languages and in English. Outside of Europe the double use of the same expression both as intensifier and as reflexive pronoun seems to be the majority pattern. It is found inter alia in Turkic languages, in Semitic languages, in many Caucasian languages, in Indie languages, in Persian, Mandarin and many others. For most of these languages it does not seem to be possible to identify a direction of change. Clear evidence for such a direction is, however, provided by West Germanic languages such as English or Afrikaans. In Old English there were no reflexive pronouns and personal pronouns did double duty as markers of disjoint reference and as markers of co-reference in a local domain. After

246 Ekkehard König and Peter Siemund

modest beginnings in Old English and even in Middle English the intensifier self was more and more used in combination with object pronouns to indicate co-reference unambiguously: (15) a. Hannibal ... bine selfrie mid atre acwealde [King Alfred's Orosius 4 11.110.2; Bately, 1980] 'Hannibal killed himself with poison.' b. Judas se arleasa pe urne Hcelend belœwde for pam lyöran sceatte pe he lufode unrihtlice aheng hine selfhe [Admonitio ad filium spiritualem 1 9.25, Norman 1848] 'Judas the disgraceful who betrayed our Lord for that wicked money that he loved unrighteously hanged himself.' These combinations of personal pronouns and intensifiers were later fused into one word and developed into reflexive anaphors in a process that is still not completely understood. At the same time, the original monosyllabic intensifier self was replaced by such compounded forms, so that intensifiers and reflexive anaphors are identical in form, even if not in distribution, in Modern English. To this piece of positive evidence for a direction of change from intensifier to reflexive anaphor we can add the general argument that reflexive anaphors are more strongly grammaticalized than intensifiers and can also for this reason be assumed to be the target of a change connecting these two classes of expressions. What we have not considered so far is the question of how a fusion of personal pronouns and intensifiers can lead to reflexive pronouns, i.e. result in expressions that mark co-reference in a local domain. Recall that in Old English a sentence of the form (16) could either express co-reference or disjoint reference of the third person pronoun: (16) he acwealde hine 'he killed him/himself Now it is quite plausible to assume that an interpretation of disjoint reference was the preferred or unmarked option. Most activities

Intensifiers as targets and sources of semantic change 247

denoted by verbs in a language are preferentially other-directed, i.e. the standard assumption is that these activities are outward directed, away from the agent. The opposite assumption seems to be the unmarked choice only for a very small subset of verbs: verbs of grooming, of defending, preparing seem to be self-directed in the standard case. So, whenever a simple personal pronoun appeared in a sentence like (16), the standard interpretation was one of disjoint reference. In order to indicate the marked option unambiguously, the intensifier could be used in Old and Middle English and was standardly used from Early Modern English onwards. The relevant meaning of the intensifier seems to have been the one discussed for the adnominal use above (cf. König/Siemund 1996c; see Keenan 1996 for a different account). When an intensifier was added to the pronoun, the referent was characterized as center and opposed to a periphery. In the absence of any contextual information of the type listed in (5) the center was clearly the agent, the referent of the subject, as opposed to other persons towards the relevant activity could be directed. For hine in (16) to be interpreted as disjoint from the subject after the addition of self, alternatives to such a referent would have to be given in the context. So far our assumption has been that the path from body parts to reflexive anaphors necessarily involves intensifiers as an intermediate stage. There is, however, clear evidence that such an intermediate step is not necessary. Expressions for body parts may lead directly to reflexive markers. In Basque, for example, the reflexive marker burua 'head' is not used as an intensifier and similar facts have been reported from other languages. Moreover, it is a well-known fact that inherently reflexive verbs or reflexive uses of verbs may develop from constructions with objects denoting part-whole relations as a result of métonymie change. The following English example from the end of the 18th century, which clearly has a co-referential, reflexive interpretation, is an interesting case in point: (17) Women who have lost their husbands' affection, are justly reproved for neglecting their persons, and not taking the same

248 Ekkehard König and Peter Siemund

pains to keep, as to gain a heart... [Mary Wollstonecraft, The Wrongs of Woman, 1798]

6.

From intensifiers to prominent roles

Intensifiers in Balto-Slavic languages exhibit an interesting type of polysemy. The expressions for intensifiers are also used in the sense of 'husband', 'lord/lady of the house', i.e. they may denote important, prominent social roles. (18) Lith. pàts 'self, 'husband', patì 'wife'; Latv. pats 'self, 'Lord of the house' Similar usages are also found in Latin, Classical Greek and Russian, even if this does not seem to be a matter of polysemy. (19) a. Russ. sam skazal 'the master/the husband said' b. Lat. ipse/ipsissimus dixit 'the master said' c. ClassGreek autòs 'the master', 'the lord of the house', 'the teacher' That such usage is not strictly a matter of polysemy, i.e. of la langue, is pointed out by Benveniste in the following quotation: L'emploi de ipse pour le maitre de maison est un simple fait de 'parole', il n'a jamais atteint le niveau de la 'langue'. (Benveniste 1966: 302)

A similar usage can also be found in Hiberno-English as the following examples show: (20) a. b. c. d.

It's himself is going to speak. Herself isn't too good again. How is himself? Herself isn't here right now. (i.e. the person salient in a specific context)

Intensifiers as targets and sources of semantic change 249

The relevant changes in Balto-Slavic as well as the incipient changes or extended uses are clearly based on ellipsis. The intensifier itself expresses the central role of a contextually given focus. In Irish English himself ox herself can simply refer to a person who is salient in a given context. The conditions for such changes or extensions in usage are nicely summarized by Benveniste in the following quotation: Pour qu'un adjective signifiant 'soi-même' s'amplifie jusqu'au sens de 'maitre', une condition est necessaire: un cercle fermé de personnes, subordonné à un personnage central qui assume la personalité, l'identité complète du groupe au point de la résumer en lui-même, à lui seul, il l'incarne. (Benveniste 1969: 91)

The relevant condition is a hierarchically structured social group with a prominent personality representing this group. The use of intensifiers in contexts where personal pronouns are normally used, particularly in subject position, is a related phenomenon. Due to the basic meaning of adnominal intensifiers, such usage is polite and honorific. In Turkish kendi can be used for all persons, speaker, hearer and other. In Japanese zibun is only used for speaker or hearer, and, to give a third example, nerrorek in Basque is an archaic form which could be used as a honorific pronoun to address persons of high rank such as priests. Finally, there is the use of intensifiers as euphemisms. On the Isle of Man, for example, sentences like (21) are apparently used to refer to small people: (21) Themselves were not out on the streets today.

250 Ekkehard König and Peter Siemund

7.

Intensifiera and scalar focus particles

In a wide variety of languages intensifiers, or related expressions, are also used in the sense of English 'even', i.e. as scalar focus particles, which characterize their focus as an extreme, or maximally unlikely value for the open sentence which constitutes their scope. In a sentence like the following George is characterized as a highly unlikely candidate for the predication 'x will go to that party': (22) Even George will go to that party. A further aspect of meaning that distinguishes scalar focus particles from intensifiers is the 'additive' implication that alternatives to the value given will make the sentence true, i.e. other people than George will go to the party. This use of intensifiers or related expressions as scalar additive focus particles can be found in languages like the following: (23) Germ, selbst (but not selber)·, Norw. selv; Dutch zelfs\ Fr. même As is shown by this list, the relevant polysemy is a wide-spread phenomenon among the languages of Europe. What the exact path of the relevant extension or change is, is completely unclear. The only relevant observation that can be made here is that there are contexts in which intensifiers and additive scalar particles are interchangeable. In the relevant contexts the denotation of the focus must be characterized by the context as an extreme case for a predication that is also made true by other values. Note that in the case of intensifiers these aspects of meaning come from the context. All the intensifier does is establish the usual relation between center and periphery: (24) a. In such basic issues the Pope himself would not know what to do. b. (This picture is very valuable) The frame itself would cost a fortune.

Intensifies as targets and sources of semantic change 251

Such sentences clearly allow a paraphrase with even. (25) a. Even the Pope would not know what to do. b. Even the frame would cost a fortune. Note that the additive implication is due to the conditional mood of these sentences. The characterization as an extreme value for an open sentence is based on contextual knowledge. What the intensifier shares with scalar focus particles is the evoking of alternatives, and compared to the plausible alternatives, the focus of the intensifier in sentences like (24) is clearly an extreme, maximally unlikely value. Sentences like (24) can thus be assumed to provide the transition point for the relevant change, which could again be characterized as an abductive one. Different compositional processes can be assumed to lead to the same overall result. There is one more detail in the change from an intensifier to an additive scalar particle that needs to be mentioned. Intensifiers in their adnominal use do not have scope over a clause, but take their scope purely within the NP with which they combine. Evidence for this claim is provided by the fact that adnominal intensifiers can never take scope over other scope-bearing elements. Moreover, the rest of the sentence other than the NP which they follow never plays any role in spelling out the contribution to the meaning of a sentence. All an adnominal intensifier contributes to a sentence is the evoking of alternatives and their characterization as periphery to the referent of the focus, characterized as center. Scalar additive focus particles, by contrast, do usually take scope over the clause that contains them, which is clearly demonstrated by the fact that they have the additive implication in sentences which do not normally suggest that alternatives to the focus referent will make the relevant predication true: (26) a. The President himself will address the meeting. b. Even the President will address the meeting.

252 Ekkehard König and Peter Siemund

8.

From intensifiers to expressions of defîniteness

As pointed out in (13) above, intensifiers may develop into demonstrative pronouns and further into definite articles. Intensifiers may also develop into expressions indicating type identity, personal pronouns and other anaphoric expressions. The following observations will mainly be based on the Romance languages (cf. Ranoth 1990; Selig 1992). That the Latin intensifiers ipse developed into a demonstrative pronoun is well-documented in the modern Romance languages, cf. Cat. eix, Pg. esse, Sp. ese. In view of the fact that Lat. ipse is the result of compounding the Latin demonstrative pronoun is with the intensifier -pse, the subsequent development of ipse into a demonstrative pronoun is certainly not entirely unexpected. It is also a well-known fact that demonstrative pronouns can develop into definite articles. This can be observed in German, where the articles der, die, das were originally used as demonstratives, but also within the group of Romance languages itself. Here the source of most modern definite articles can be found in Lat. ille, cf. Fr. le, la; It. il, la; Pg. o, a; Sp. el, la. The underlying use of ille that allowed for this development is that as a determiner, not as a pronoun. We are here confronted with a clear process of semantic bleaching. Demonstratives are deictic expressions of defîniteness, and when these expressions lose their deictic component and only retain the ability to pick out a unique referent in a certain discourse domain or world, they develop into definite articles. Hand in hand with this development has gone the emergence of an obligatory definite/indefinite marking on nomináis. Sardinian and those dialects of Catalan spoken on the Balearic Islands did not use ille as source for the development of the definite article, but derived it from ipse instead, cf. Sard, su, sa; Cat. es, sa. Interestingly enough, the relevant dialects of Catalan have another set of definite articles, viz. those derived from ille (el, la). Their use, however, is mainly restricted to adverbials (a l 'hora de la mort) and proper names (I'Havana, el Papa). There are even some minimal pairs. Thus, s'església refers to a particular church as a building

Intensifies as targets and sources of semantic change 253

whereas l'Església denotes the (Catholic) Church as an institution. Also, es sertyor means 'gentleman' or 'owner', but el Senyor is reserved for 'the Lord' (cf. Hualde 1992: 281; Seguí i Trobat 1993: 35). To base the definite article on ipse appears to have been the original strategy in the linguistic domain where Catalan is spoken today. Those derived from ille were superimposed during the 12th and 13th century. The original situation is preserved in a number of place-names: Sant Joan Despi, Collserola, Sant Esteve Sesrovires (Seguí i Trobat 1993: 33). Another possible development of demonstrative pronouns is that of 3rd person personal pronouns. Although most of the modern Romance languages used Lat. ille as the source of this development (Fr. il, elle, Pg. ele, eia; Sp. él, ella), Sardinian derived them from Lat. ipse, cf. isse, issa (Campidanese issu). As in the case of definite articles, this development can also be assumed to be due to semantic bleaching. Note, however, that the relevant source are demonstratives in their use as pronouns, not as determiners. Again, the deictic component is lost and what remains is the referential function. It appears noteworthy that Sard, isse may also be used as a respectful form of address and that it is by and large restricted to animate referents (cf. section 6). Apart from Lat. ipse, Lat. ille too left its imprint on the Sardinian pronominal system. The clitic pronouns found there are clearly derived from this demonstrative and are in fact preferred in direct/indirect object positions (cf. Jones 1993: 201): (27) a. ?Appo datu su dinari a issos. b. Lis appo datu su dinari. Ί gave them the money.' This does not mean that isse could not be found in object positions. However, if it is used there, it always implies some element of contrast: (28) Appo vistu a issos, ma no 'a tie. Ί saw them, but not you.'

254 Ekkehard König and Peter Siemund

Another possible path of development that intensifiers can take is the one to expressions of type identity (Pg. mesmo, Sp. mismo). In the French expression même, we even find formal identity between Lat. ipse and idem. In the course of this development, the Lat. enclitic pronominal intensifier -met became prefixed to ipsu (mismo < me-ísmo < medipsimu < -met ipsimu). What seems to have been grammaticalized here is the discourse-anaphoric or discourse-deictic function of ipse. This expression could regularly be used to pick up information which had been established in the previous discourse. Anaphoricity can be analyzed as the expression of identity, but this is at first restricted to cases of token identity. When this expression of identity is extended to also cover cases of type identity, we arrive at the meaning of Sp. mismo in contexts like the following: (29) Tengo el mismo coche que mi hermano. 'I've got the same car as my brother.' In Sardinian we also find anaphoric uses of the pronominal forms derived from Lat. ipse. Still, this use of Sard, isse appears to be restricted to those positions where it is not a co-argument to its antecedent. In co-argument positions the reflexive clitic si is used invariably. In contrast to the reflexive clitic, isse in itself is not confined to the co-referential interpretation (cf. Jones 1993: 241): (30) Juanne¡ credei ki Gavinij I'at comporatu pro isse¡/j/k. 'John thinks that Gavin bought it for him/himself.' The co-referential interpretation can be forced by adding an intensifier (e tottu (lit. 'and all') or mattessi 'same') to the anaphoric pronoun. This strategy to restrict the binding domain of an anaphoric expression is strikingly parallel to what we find in Old English, Afrikaans or Frisian (cf. section 5). (31) Juanne\ credet ki Gavinij I'at comporatu pro isse e tottu/isse mattessij. 'John thinks that Gavin bought it for himself.'

Intensifies as targets and sources of semantic change 255

9.

Conclusion

However tentative some of the observations and assumptions are that were made above, the picture that emerges provides further evidence for the view that semantic change is not as random and unpredictable as it is often assumed to be. The development of minor lexical categories, in particular, exhibits a great deal of similarity across languages and seems to follow one of a limited number of possible paths. The processes of semantic change exhibited by the development of function words are partly those found in the development of major lexical categories, but there is also a certain element of generality, predictability and unidirectionality not found in the development of nouns, verbs or adjectives.

References Baker, Carl Leroy 1995 Contrast, discourse prominence, and intensification, with special reference to locally free reflexives in British English. Language 71: 63-101. Bately, Janet (ed.) 1980 The Old English Orosius. Oxford: Oxford University Press. Benveniste, Emile 1966 Problèmes de linguistique générale. Paris: Gallimard. 1969 Le vocabulaire des institutions indo-européennes, Volume 1. Paris: Editions de Minuit. Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Niemeyer. Edmondson, Jerold and Frans Plank 1978 Great expectations: An intensive self analysis. Linguistics and Philosophy 2:373-413. Farr, James 1905 Intensives and Reflexives in Anglo-Saxon and Early Middle English. Baltimore: J.H. Fürst. Grimm, Jacob 1967 Deutsche Grammatik. Hildesheim: Georg Olms Verlagsbuchhandlung (reprint).

256 Ekkehard König and Peter Siemund

Hualde, José Ignacio 1992 Catalan. London: Routledge. Jones, Michael Allen 1993 Sardinian Syntax. London: Routledge. Keenan, Edward L. 1996 Creating Anaphors. An Historical Study of the English Reflexive Pronouns. UCLA, ms. Koch, Peter 1997 Ein Blick auf die unsichtbare Hand: Kognitive Universalien und historische romanische Lexikologie. In: Thomas Stehl (ed.), Unsichtbare Hand und Sprechwahl. Typologie und Prozesse des Sprachwandels in der Romania. Tübingen: Narr. König, Ekkehard 1991 The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. König, Ekkehard and Elizabeth Traugott 1988 Pragmatic strengthening and semantic change: The conventionalizing of conversational implicature. In: Werner Hüllen and Rainer Schulze (eds.), Understanding the Lexicon. Tübingen: Niemeyer. König, Ekkehard and Peter Siemund 1996a iSWto-Reflektionen. In: Gisela Harras (ed.), Wenn die Semantik arbeitet. Festschrift für Klaus Baumgärtner, 277-302. Tübingen: Niemeyer. 1996b Emphatische Reflexiva und Fokusstruktur. In: Inger Rosengren (ed.), Sprache und Pragmatik AO, 1-42. Lund: Lunds Universitet. 1996c On the development of reflexive pronouns in English: A case study in grammaticalization. In: Uwe Böker and Hans Sauer (eds.), Anglistentag 1996 Dresden, Proceedings, 95-108. Trier: Wissenschaftlicher Verlag. Moravcsik, Edith 1972 Some cross-linguistic generalizations about intensifier constructions. Papers from the Regional Meetings of the Chicago Linguistic Society 8: 271-277. Norman, H.W. 1848 The Anglo-Saxon Version of the Hexameron of St. Basil ... and the Anglo-Saxon Remains of St. Basil's Admonitio ad filium spiritualem. London. Romoth, Susanne 1990 Die Identitätspronomina in der Romania. Genève: Droz.

Intensifìers as targets and sources ofsemantic change 257

Seguí i Trobat, Gabriel 1993 Iniciado a la morfosintaxi catalana. Palma: Aina Moll. Selig, Maria 1992 Die Entwicklung der Nominaldeterminanten im Spätlatein. Romanischer Sprachwandel und lateinische Schriftlichkeit. Tübingen: Narr. Traugott, Elizabeth Closs 1985 On regularity in semantic change. Journal of Literary Semantics 14: 155-173. 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 57: 33-65. 1995 Subjectification and grammaticalization. In: Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectmsation in Language, 31-54. Cambridge: Cambridge University Press.

Cognitive ease and lexical borrowing: the recategorization of body parts in Romance1 Thomas Krefeld

0.

The Romance challenge

True problems will never disappear - science loves them too much. Romance Linguists, for instance, have been fascinated by the parallelism in the evolution of different Romance languages ever since the birth of their science; this parallelism has so much more appeal, because it is essentially of a typological or structural order, without always implicating the etymological identity of the linguistic material. Whereas diachronic morphosyntax has accepted this descriptive (and explanatory) challenge, semantics, specially lexical semantics, have for the most part remained reluctant. The cognitive orientation could though stir up this state of things.

1.

The parts of the human body and the Torso-Extremities-Model

From a cognitive point of view the categorization of those domains that do not depend, at least in an indirect way, on the form of socialcultural life, merits special curiosity. As a field of application, let us take the designations of the parts of the human body2 excluding, of course, the anatomical nomenclature, which would not be representative of the pre-scientific and naïve lexical categorization of the "natural" spoken language. Our languages conceive the human body as a whole, more precisely as an organized group of high complexity; this group consists of parts, some of which are conceived as natural sub-groups, which themselves show a higher or lesser degree of complexity.3

260 Thomas Krefeld

French, for example, distinguishes the following groups (corps) and sub-groups (parties): tête 'head', cou 'neck', tronc 'torso', extrémités (membres)·, further, certain joints linking the sub-groups themselves (hanche 'hip') 4 and other joints linking certain parts of the sub-groups (as genou 'knee', cheville 'ankle', coude 'elbow', poignet 'wrist'). I will provisionally call this type of categorization the "Torso-Extremities-Model".

-hanche

Figure 1. The Torso-Extremities-Model (French version)

From an onomasiologie point of view, the sub-categorization of the constitutional parts (or sub-groups) depends on the following criteria: • • •

the respective vertical position (superior vs. inferior); the direction of view (front, rear, lateral); the exposition to view in an immobile position (exterior vs. interior).

Cognitive ease and lexical borrowing 261

The respective vertical position is fundamental. We find it again in the sub-categorization of all parts (all except the neck). The direction of view is important for the sub-categorization of the head, the neck and the torso, but is quite insignificant with regard to the extremities (in French, only jarret 'hollow of the knee' and talon 'heel'). The least important, but at the same time the most specific, is the exposition to view, which concerns the neck (gorge 'throat') and, only in a very marginal way, the extremities (aisselle 'armpit'). It is therefore interesting to see that the French type of sub-categorization, reflects to a certain extent, from joint to joint so to speak, the articulation of the skeleton. For example, the lexical categorization of the arm is analogue to that of the leg, according to the analogy of their physical constitution. Compare: •

sub-group bras; comprising of coude 'elbow', poignet 'wrist', main 'hand', doigt 'finger'; • sub-group jambe; comprising of genou 'knee', cheville 'ankle', pied 'foot', orteils (doigts) 'toe'. Noteworthy are the synonymous names of the toe (orteil/doigt): the lexicalization of a specific designation (orteil) shows that the "anatomic principle" cannot be completely neglected. The existence of the less specific term (doigt), however, shows the dominance of the "organic principle". The difference of their organic functions is obviously more important than their anatomic analogy. The leg is defined as a sub-group just on account to its main function, which is assuring the movements by involving the functions of all of its parts. For the arm, though, this is not the case, because the foot (pied) and the hand (main) do not at all have analogous organic functions. Due to its functional importance, the hand must be considered a supplementary sub-group, which is not the case with the foot. This is also the reason why the word for the foot frequently refers to the entire leg (Rum. picior, Germ. dial. Fuß 'foot', 'leg'), whereas the name of the hand very rarely seems to refer to the entire arm.

262 Thomas Krefeld

Note that the relative autonomy of the hand and the resulting functional differences between the arm and the leg are, according to the envisaged language, more or less evident. At first glance, one might object that the lexical categorization of French (as well as of Romance and already of Latin) would correspond to the skeletal symmetry ("anatomic principle") because the articulations of the hand and the foot have the same names: Lat. digitus means 'finger' as well as 'toe', and so do Fr. doigt, It. dita, Rum. deget etc., which can be, if necessary, specified as Fr. doigt du pied, It. dita del piede, Rum. deget de la picior etc. However, one must take into account the fact that the designations of the hand were transposed to that of the foot, and not vice versa. Above all, we ascertain the existence of individual designations for each of the different fingers (in French: pouce 'thumb', index 'forefinger', médius 'middle finger' [not really popular], annulaire 'ring finger', auriculaire 'little finger'), the inside part of the hand (paume 'palm') and the joints between the fingers and the hand (nœuds 'knuckles'). Some of these designations (those of the most salient fingers) have been transposed to the articulations of the foot, despite their having completely different organic functions: • • • • •

2.

Lat. pollex 'thumb' andpollexpedis; Romansh polisch and polisch dil per, Fr. pouce and pouce du pied; It. pollice and pollice del piede (as well as alluce); It. mignolo 'little finger' and mignolo del piede.

The etymological point of view: the non-Latin element

Looking at the fundamental importance of the body for human orientation in the world, we are not surprised to note a strong Latin-Romance continuity in this lexical domain; it suffices to cite the quasiPan-Romance designations of the EAR (type Lat. auricula), the NOSE (type Lat. nasus), the EYE (type Lat. oculus), the LIP (types Lat. labi-

Cognitive ease and lexical borrowing 263

urn, labrum), the TEETH (type Lat. dens), the TONGUE (type Lat. lingua), the HAIR (type Lat. capillus), the SKIN (type Lat. pellis), the STOMACH (type Lat. stomachus), the RIB (type Lat. costa), the HAND (type Lat. manus), the FINGER (type Lat. digitus) etc. However this essentially basic lexicon contains a pile of lexemes of non-Latin origin, as e.g. in French: joue 'cheek' < pre-Lat. *gabota (FEW 4: 9-10); nuque 'neck' < Lat. (medical) nuca < Arab. nukha·, échine 'vertebral column' < Frankish *skina; flanc < Frankish *hlanka; hanche < Frankish *hanka; bras < Greek-Lat. brac(c)hium\ jarret < Gallic *garra\ jambe < Greek-Lat. campa. It has to be emphasized that French, which often plays a soloist role, this time is in harmony with the other Romance languages. The absence of Latin elements is not particularly specific to French; the situation, e.g., of Romansh (whose lexicon is often archaic) and Italian are both analogous, even despite the fact that Italian belongs to what Dámaso Alonso (1978) called the "Romania continua". Even Rumanian, the other great soloist among the Romance Languages, behaves in an identical manner. If the etyma of the loan words integrated in French, do not appear in the three other languages, we find other terms of equally non-Latin origin or, at least, Latin elements that are more or less isolated and regional and which can be found nowhere else. If, inversely, French distinguishes itself by a geographically isolated Latin lexeme, the other languages, in contrast, present borrowings. The following list shows a sample of non-Latin elements among the French names for body parts and their Italian, Rumanian and Romansh5 equivalents: 1.

2.

Fr. joue < pre-Lat. *gabota vs. Romansh gauta < Gallic gaita; It guancia < Longobardic *wankja; Rum. obraz < Slavic obrazu. NECK: Fr. nuque (It. nuca) < Arab, nukha vs. Romansh totona/ tatona < ? (without doubt pre-Latin); Rum. ceafa < perhaps Turk. (cf. Alban, qafé). CHEEK:

264 Thomas Krefeld

3.

3. 4.

5. 6.

7.

8.

Fr. échine < Frankish *skina vs. Romansh spina dorsola < Lat. spina·, It. schiena < Long. *skena, Rum. spina < Lat. spina. FLANK: Fr. flanc (It fianco) < Frankish *hlanka vs. Romansh fomau (Surselv.) / fama (Engd.) derived from fom 'faim'. HIP: Fr. hanche < Germanie *hanka vs. Romansh calun (Surselv.)/ gialun, chalun (Engd.) < pre-Lat. *calon-/*galon- (cf. HR I: 144); It. anca < Longobardic or Frankish *hanka; Rum. çold < Transsilv.Germ. scholder. ARM: Fr. bras, Romansh bratsch; It. il braccio·, Rum. braf < Greek-Lat. brac(c)hium. NUCKLE: nœud < Lat. nodus vs. Romansh dartugf < Lat. *artuculum < articulum (cf. HR I: 242); It. nocca < Longobardic biohha; Rum. încheietura degetelor. HOLLOW OF THE KNEE: Fr. jarret (It. garretto, besides cavità del ginocchio) < Gallic garra·, also Romansh garlet (Surselv.)/ giarlet (Engd.) < Gallic *garrulu de *garra{' vs. Rum. scobitura genunchiului. LEG: Fr. jambe·, Romansh comba (Sursei\.)/chomma (Engd.); It. gamba < Greek-Lat. kampe vs. Rum. picior < Lat. petiolus. VERTEBRAL COLUMN:

Given the heterogeneity of the origins - there are words of substratum languages (Gallic, pre-Latin), of superstratum languages (different Germanic languages, Slavic) and of adstratum languages (Arabic)8 - and considering the long duration of development, we must ask ourselves why Latin and the Romance languages adopted so many loan words to express this very fundamental conceptual domain.

3.

The motivation for the borrowings

The sociolinguistic situations, that is to say the situations of linguistic contact to which this handful of words is submitted, are so divergent that it would be absurd to imagine one sole historical explanation. This doesn't exclude that certain borrowings nevertheless

Cognitive ease and lexical borrowing 265

can be explained from their cultural context, i.e. by the cultures of the speech communities in contact. Arabic nuca would, without any doubt, not have been adopted by Medieval Latin without the knowledge of the highly developed Arabic medicine. But this case is an exception. In general, the habitual opposition of "necessity borrowing" and "luxury borrowing"9 has actually no explanatory power in this lexical domain, insofar as the necessity is defined in the traditional, referent-based onomasiological way. Commonly the borrowings that designate new things, institutions etc., i.e., realities that were unknown before the adopting speech community got in contact with the "giving" culture and its language, are called "necessary". In contrast with this, the integration of a borrowing without being onomasiologically necessary is considered to be a mere result of social prestige. It goes without saying that each "foreign" language gets prestigious when the culture it represents is accepted as exemplary in certain domains. These are the "luxury borrowings" that double or even replace already existing words without bringing new senses ("signifiés") as, for example, the English loanword star in French besides the original vedette. Thus, it is evident that the "newness" doesn't play a role in the domain of the human body and its parts. Neither is the prestige of the contact language important, as long as parts of the body that are tabooed by the borrowing society are not concerned: the latter is often the case with the genital organs. In this case, the use of a loan word that enjoys the prestige of its original language facilitates the violation of the taboo as, for example, the Latin names for the genital organs. We are therefore allowed to search the motivation for borrowing in the very categorization of the body parts, or more precisely, in the differences of categorization that distinguish Latin, on the one hand, and French (or Romance) on the other. And in fact, the loan words cited above participated in a veritable process of recategorization, which favoured the organic principle that predominates, as we saw, in French. This predominance actually doesn't seem to be an isolated particularity of French, it is rather a constant that is able to explain

266 Thomas Krefeld

the restructuring of the Romance vocabulary in the domain of the body parts. When we reconstruct the corresponding lexical field in Latin, we immediately meet with a somewhat contradictory categorization:

1, 2, 3, 4 Lat. membra 3,4 Lat. artus, Fr. membres

a. ARM and LEG missing in Latin

b. EXTREMITIES and BODY PARTS in Latin

Figure 2. The idiosyncratic categorization of body parts in Latin

We have, at one end, the rather subtle anatomic opposition between two collective terms: •

Lat. membra 'constituent parts of the body, extremities and head and torso' (GII: 863-864);

Cognitive ease and lexical borrowing 267

•

Lat. artüs 'articulated extremities' (G I: 603; cf. the latinism It. arto).

At the other end, there apparently weren't any specific terms for the sub-groups of the superior and inferior extremities; that is to say, there is neither a word for the entire arm, nor for the entire leg. It is obvious that the Torso-Extremities-Model is only superficially rooted in Classical Latin (by the existence of artüs), but in the course of history it grew stronger and stronger, operating on two levels: 1. The opposition of membra vs. artüs is neutralized in the way that membra takes the signification of artüs 'articulated extremities'. The older sense of membra 'constituent body parts including extremities and torso' looses its lexical representation; it is "delexicalized". 2. The two lexical "gaps", concerning the designations of the arm and the leg, were filled in, in different eras, but in a similar way.

3.1. The ARM With regard to Fr. bras, we observe that Greek βραχίων 'arm in general' (Gemoll: 163) was borrowed by Latin at quite an early moment. If this had not been the case, this originally Greek element wouldn't have got its Pan-Romance character (G I: 859; REW: 1256). I would like to emphasize the fact that this borrowing doesn't just fill in a lexical gap10, the loan word brac(c)hium is part of a rather complicated process of re-categorization. The Latin lexicography shows that the Greek import brac(c)hium designates the arm in general, and the forearm in particular. In Romance, we find something quite curious, because the same term not only stands for the arm in general, but also for the upper arm in particular. Why is this so? Classical Latin uses several names for designating, among others, the upper arm:

268 Thomas Krefeld

•

armus 'upper part of the upper arm and the shoulder, upper arm' (G I: 582); • (h)umerus 'bone of the upper arm; the section of elbow right to the collarbone and the shoulder blade, i.e. upper arm and shoulder' (G I: 3294); • lacertus 'muscles; muscles of the upper arm; the strong upper arm, with many muscles' (G II: 527). We cannot, a priori, exclude the possibility that the loan word brac(c)hium would have completed this paradigm. But looking at the very weak presence of the three cited Latin words in the Romance languages, at least as designations of the arm (or upper arm),11 we should rather ascertain that brac(c)hium absorbed their function, or more exactly, a part of their function: the loan word only replaces them to the extent as they refer to the arm. It is true that the chronology can't be reconstructed in a reliable way and, of course, brac(c)hium didn't make the three terms disappear immediately; nevertheless, we can suspect, going by what we have said, that the three Latin words might have disappeared, because they didn't refer to the organic sub-group arm, nor to a constituent part of it, but to anatomic sections that don't correspond at all to the Torso-ExtremitiesModel, because the three obsolete Latin terms designate at the same time an upper part of the torso (the shoulder) and a part of the (upper) arm. The most illustrative example in this respect is the case of (humerus that was best maintained: this word doesn't refer anywhere in Romance to the shoulder and the arm (except the anatomy technique term It. omero) but only to the shoulder. This semantic restriction manifests very clearly the adaptation to another type of categorization. All those Romance languages that didn't manage to adapt the sense of (h)umerus, have abandoned it completely. An evident result of this process is the missing of a term for the upper part of the torso. The new "gap" is compensated for by extending the referential range

Cognitive ease and lexical borrowing 269

of a contiguous anatomic term, i.e. Latin spatula (> Fr. épaule, It. spalla-, cf. G II: 2746). This Latin word referred to the shoulder blade in a very precise way, and therefore to a part of the skeleton linked to the shoulder.12 The example makes evident that it doesn't suffice to note a métonymie change, that bases itself on a relation of contiguity (shoulder blade and shoulder), but that we also have to explain why other, equally contiguous terms have not been metonymically extended. Remember that Lat. lacertus, which already meant 'upper arm' and 'shoulder' would have been an "ideal" candidate for metonymy. It did not succeed - I should say: it could not succeed because the successful concurrent spatula had the advantage of being in harmony with the categorization of the Torso-ExtremitiesModel. Latin spatula designates a part of the torso and only of the torso, whereas lacertus designates a part as well of the arm as of the torso.

3.2. The LEG In the domain of this sub-group, we have to deal with a situation somewhat analogous to that of the sub-group arm. Classical Latin presents an opposition between: • •

femur 'thigh, femur (thighbone)' (G I: 2715) and crus 'lower part of the leg' (G I: 1774-1775).

From the Romance perspective, there are two things to remember: firstly, neither the one nor the other term survived in Romance13 and, secondly, all the Romance languages have a word to designate the entire leg, or the sub-group leg, as we prefer to call it. We therefore observe the same type of re-categorization. But unlike the re-categorization (of the designations) of the arm, the Romance languages show different solutions. Iberoromance chose a "satellite" word, i.e. a word that in Classical Latin already designated a part of the leg and which was therefore in métonymie

270 Thomas Krefeld

relation with the two lost words: Lat. perna 'hip' 14 and primarily 'ham of the posterior thigh' (G II: 1618-1619), which gave Cat./Pg. perna, Span, pierna. Rumanian is characterized by a Latin diminutive petiolus 'small foot, small leg (of a lamb etc.)' (G II: 1670) based on pes, which gave picior. Let us remember that this Rumanian word means 'leg' and 'foot'. Sardinian, Corsican and certain Mediterranean Italian dialects adopted the feminine form anca of the Latin adjective ancus 'bent' (REW: 446; G I: 422). However, in Central Romance, it is still a borrowing dating back already to the Latin period, Greek καμπή 'curve, inflection' (Gemoll: 406; G I: 947) that gave Fr. jambe, It. gamba, Romansh comba/ chamma etc. The FEW, s.v. camba, states that this word was a technical term of veterinary medicine designating the 'ankle joint of a horse, hock' (FEW II: 119). But one must say that the FEW's commentary is not apt to instruct us about the process of borrowing because it doesn't offer any outline of the semantic background against which it unfolds. Without any doubt, Wartburg is right to discuss the case of camb in the context of other Romance words that took the meaning of 'leg'. He resumes that the words that substituted crus are all taken from the coarse names of the animal body.15 Thus, his reconstruction focuses only on the "vulgar" aspect. In his Einführung in Problematik und Methodik der Sprachwissenschaft (1962), he continues with comments on the same example, saying that "this apparition of slangy (in German, burschikos), familiar and expressive words, in general characterizes the late empire" (1962: 117). In my opinion, this quite stereotypical argumentation is, despite its explanatory power, not very helpful in this case. Firstly, it makes an abstraction of the Rumanian and Sardinian solutions, further, it doesn't take into account that Classical Latin perna also signifies 'hip' and, maybe, 'leg' and, at last, it doesn't consider the fact that the Latino-Greek camba was far from being exclusively a term of the veterinary medicine. The word had the very general sense 'bent', already attested in Latin in Plautus (G I: 947) and thus is a synonym

Cognitive ease and lexical borrowing 271

of the adjective ancus, which lead, as we have seen, to other Romance designations of the leg (Sard., Cors, anca 'leg') and which doesn't seem to be vulgar. Elsewhere, a technical veterinary term by itself has no rude connotation. It could, eventually though, receive this connotation but on the condition that it leaves the technical domain. On the level of the same variety, vulgarity (or expressivity) and technicality are mutually exclusive. Finally, one must not forget that a name for the entire leg was missing in Classical Latin, just as it was for the arm, and that a lexeme signifying 'bent' would actually lend itself ideally, to "mending" the lexical network.

3.3. The HIP I have now to treat the most common of the loan words in this domain, which is a Germanic designation of the hip. Walther von Wartburg discussed the problem in his already cited Einführung (1962: 117-118); his explanation can be summarized as follows: the Latin designations of the inferior and superior parts of the leg {crus and femur) have disappeared; in the case of femur, it resulted from the homonymie gênante with fimus 'animal waste, manure', caused by the fact that fimus had changed its declension under the influence of its synonym stercus (fimus > femus, -oris). Thus, the formal changing of a word designating 'animal waste, manure' would have made the designation of the thigh obsolete. It would result a lexical gap, provisionally filled by the métonymie extension of the designation of the contiguous part: the Latin coxa 'hip' would have taken the additional meaning of the thigh (> Fr. cuisse, It. coscia), designating the segment of the leg from the hip right to the knee. Unlike the disappearance of femus, the semantic change of coxa wouldn't end with an empty case, but only with a situation of "semantic distress" (in German, Notlage), released finally by the integration of the Germanic *hanka 'hip' (> Fr. hanche, It. anca), that was "heard from time to time from the mouths of Germanic soldiers and colo-

272 Thomas Krefeld

nists" (118). Two aspects of this rather mechanical explanation give rise to the following critical remarks: 1.

2.

Lat. femus, -oris 'animal waste, manure' plays a marginal role, mainly because it has survived only in Old French, Occitan and Catalan (REW: 3311). It isn't very "elegant" to suppose that a word of a quite reduced spatial distribution would have set off a more or less Pan-Romance process: Lat. coxa didn't conserve anywhere (except Engd. cossa; cf. HR II: 640) the meaning of 'hip' and its supposed Germanic substitute *hanka is again spread out over a vaster territory (Italy, the Iberian Peninsula). It is not very convincing to assume the métonymie extension of coxa and to pretend at the same time that this metonymy is immediately replaced because it brought "confusions" (Unklarheiten) with it. One would at least have to specify of what nature these confusions could be; surely they are qualitatively of a type different from those much less probable confusions that provoked the homonymy between *femus, -oris 'animal waste, manure' and femur, -oris 'thigh': In this last case, all metonymy appears excluded; there is hardly a linguist who will be able to imagine a context in which a risk of confusion could happen.

In fact, the metonymy between hip and thigh is so trivial that it isn't worth searching for a justification and, in terms of the Romance results, it probably already existed in the Latin of the Roman Empire. Still, it remains curios that coxa has completely lost the broad sense 'hip' and 'thigh' (corresponding to the sense of Lat. perna which underwent a similar modification). The best witness to the obligatory nature of this semantic restriction is the situation of Romansh, where reflexes of Lat. coxa are attested in a closed geographic area. The semantics are now strictly alternative and the word means either 'thigh' (queissa in Surselvan) or 'hip' (cossa in Engadine dialects), but has never both meanings (HR II: 640).

Cognitive ease and lexical borrowing 273

It goes without saying that this type of alternative semantic reduction is not due to an unavoidable tendency towards more lexical precision: Rum. picior 'leg' and 'foot' has shown us that Romance lexicalization, in this domain as in others, also worked in the opposite direction. The Rumanian word (of Latin origin) integrates two senses in métonymie relation, although in Classical Latin, two different forms (crus vs. pes) already existed. The case of coxa (i.e. the alternative semantic reduction and the parallel integration of Germanic *hanka) is unsuitable for illustrating the traditional rule: phonic accidents may provoke semantic confusions. We prefer to see a confirmation of re-categorization as sketched out above. The existence of a lexical category such as Lat. coxa that designates the section of the human body beginning at the hip and reaching to the knee doesn't fit into the Torso-ExtremitiesModel. According to this categorization, the hip is a part of the subgroup of the torso, whereas the thigh is a constitutive part of the subgroup of the leg.

4.

Conclusion: recategorization and cognitive ease

From an anatomical point of view, the Latin way of categorizing the human body (see above fig. 2a and 2b) might be called ingenious; certain terms refer to very complex organic compounds as the UPPER ARM-SHOULDER- or THIGH-HIP-categories. This categorization, nevertheless, suffers from a fundamental disadvantage: it does not correspond with the perceptive saliency of the arms and the legs in visual appearance of the human body. Arms and legs are without any doubt cognitively privileged. It suffices to recall children's drawings (fig. 3a and 3b) which in their first stages (see fig. 3a) even omit the torso.

274 Thomas Krefeld

a. Very early stage

TT b. Latter stage

Figure 3. Perceptive saliency of the arms and the legs in children's drawings

Based on this, the presence of many non-Latin lexemes of completely heterogeneous origins is no longer surprising; it has to do with lexical categories that have appeared in Romance languages completely by chance, more precisely, by the contingent existence of a more or less important bilingual minority. These lexical categories succeeded in being borrowed by the unilingual majority because their concepts are imposed by perceptive experience. However, the re-categorization evidently wasn't introduced all at once by means of a specific loan situation (or of one specific language of contact); it occurred thanks to the contribution of many languages, step by step along history's path. The cognitive perspective couldn't justify the appearance of a bilingual population, but it enables us to explain the integration of borrowings, i.e., its acceptation by a community of non-bilingual native speakers. In other words, these borrowings are neither luxury nor necessity borrowings but moreover borrowings of cognitive ease.

Notes 1. 2. 3.

I am grateful to Max Krefeld for drawing the figures 1, 2a, 2b. For rich lexical material cf. Zauner 1894. For further philosophical details see Husserl 1992: 227-300.

Cognitive ease and lexical borrowing 275

4. 5. 6.

7.

8.

9.

10.

11.

12. 13. 14. 15.

Curiously enough, there is no word for the SHOULDER JOINT, Fr. aisselle meaning only 'armpit'. Romansh is the name for the Swiss varieties of Rhaeto-Romance. The etymologies are due to Gamillscheg, HR, Tiktin, REW and DELI. This word doesn't have the meaning 'joint' which is lexicalized by lisiira (Engd.; cf. HR I: 438; < ?) and by giugadira (Surselv.; cf. HR I: 368) deverbal noun, of giugar 'to play'. The type Fr. y'a/rei/Romansh g(i)arletlIt. garretto has many meanings; it still designates the Achilles tendon (Romansh), the heel of the horse (Italian) where the jointure "bends the behind leg" at the mammiferous ungulates. (Robert: 1042) The "substrata" just as the "superstrata" are at the time of the borrowing evidently adstrata of Latin, resp. of Romance; the terminology isn't particularly happy since the senses of the two terms oscillate between the relative chronology (before vs. after the romanization) and the sociopolitical status (dominating vs. dominated speech communities). The terms of Luxuslehnwort 'luxury borrowing' and Bediirfhislehnwort 'necessity borrowing' have been coined by the Swiss dialectologist Ernst Tappolet; cf. Tagliavini 1998: 214. This explanation goes perhaps for the case of the loan words designating the cheek (as Fr. la joue), that doesn't seem to have had a proper name in Latin: the meaning of Lat. gena wasn't simply 'check', but rather 'part of the face between the forehead, the temple and the chin, including the eyelid and the eye-socket' (Georges I: 2913) Cf Lat. armus > Rum. (dialectal) arm 'cuisse des animaux', ORum. '(jointure) of the hip'; cf. Tiktin I: 214-215; REW: 4822; the relatively well established Lat. (h)umerus survives, but exclusively in the signification 'shoulder' (Rum. umar, It. omero, Sp./Pg. hombro·, cf. REW: 4232). The fact that there exists in Latin a lexicalized metaphoric use (the word also designates a type of spoon) is without importance in our context. The REW doesn't give any reflex. Unfortunately, G: 1618, is not very precise in its definition. "Therefore some coarse words came into use for 'leg', taken from animal bodies" (FEW II: 119); cf. also Wartburg (1962: 117).

276 Thomas Krefeld

References Alonso, Damaso 1978 Die Ausgliederung der westromanischen Sprachen. In: Reinhold Kontzi (ed.), Zur Entstehung der romanischen Sprachen, 163-186. Darmstadt: Wissenschaftliche Buchgesellschaft. Bernardi, Rut 1994 Handwörterbuch des Rätoromanischen: Wortschatz aller Schriftsprachen, einschliesslich Rumänisch Grischun, mit Angaben zur Verbreitung und Herkunft. 3 Volumes. Zürich: Offizin. (= HR) Cortelazzo, Manlio and Zolli, Paolo 1979 Dizionario etimologico della lingua italiana. 5 Volumes. Bologna: Zanichelli. (= DELI) Gamillscheg, Ernst 1934 Romania Germanica. Sprach- und Siedlungsgeschichte der Germanen aitf dem Boden des alten Römerreichs. 3 Volumes Berlin / Leipzig: de Gruyter. Gemoll, Wilhelm 9 1965 Griechisch-deutsches Schul- und Handwörterbuch. München / Wien: Freytag. Georges, Karl E. 8 1976 Ausführliches Lateinisch-deutsches Handwörterbuch. 2 Volumes. Hannover: Hahn [1913]. (= G) Husserl, Edmund 1992 Logische Untersuchungen. Hamburg: Meiner (= Gesammelte Schriften 3.). Meyer-Lübke, Wilhelm 3 193 5 Romanisches etymologisches Wörterbuch. Heidelberg: Winter. (= REW) Robert, Paul 1984 Le petit Robert. Dictionnaire alphabétique et analogique de la langue française. Paris: Le Robert. Tagliavini, Carlo 2 1998 Einführung in die romanische Philologie, Tübingen / Basel: Francke [1973]. Tiktin, Hariton 2 1986 Rumänisch-deutsches Wörterbuch. 2., überarbeitete und ergänzte Auflage von Paul Miron. Wiesbaden: Harrassowitz.

Cognitive ease and lexical borrowing 277

Wartburg, Walther von 2 1962 Einführung in Problematik und Methodik der Sprachwissenschaft. Tübingen: Niemeyer. Wartburg, Walther von 1928ss. Französisches etymologisches Wörterbuch. Leipzig / Tübingen / Basel: Zbinden. (= FEW) Zauner, Adolf 1894 Die romanischen Namen der Körperteile. Romanische Forschungen 14: 339-530.

Cognitive aspects of semantic change and polysemy: the semantic space HAVE/BE Peter Koch

1.

The semantic space

HAVE/BE

At least since Lyons's famous (1967) article, linguists have been interested in the interlacements of possessive, existential, and locative predicates from a cross-linguistic perspective. Localist and/or generative approaches, accounting for the relevant interlacements in terms of a unitary underlying structure,1 have been challenged by typological and/or cognitive approaches which, while acknowledging the lexical diversity of natural languages, nevertheless make cautious semantic generalizations, where cross-linguistic insights enable us to do so (cf. Bickerton 1981; Wilson 1983; Hengeveld 1992; Heine 1997; Feuillet 1998). The point I want to stress here is the importance of the diachronic evidence from semantic change and - indirectly - from polysemy,2 which allows us to formulate cognitively valid generalizations even in lexical semantics. In the area HAVE/BE, as I term it for the sake of convenience, Bickerton (1981: 244-246) proposes a kind of "semantic space" like Figure 1 in order to account for the joint lexicalization of different semantic categories on the bases of adjacency: OWNERSHIP

LOCATION

POSSESSION

EXISTENCE

Figure 1. The semantic space HAVE/BE (after Bickerton 1981: 245)

280 Peter Koch

In the meantime, this fourfould schema has turned out to be neither sufficient nor fully adequate for providing all the adjacencies we need (cf., e.g., Wilson 1983). Furthermore, the theoretical basis of 'adjacency' itself is not quite clear. Is it a relation of adjacent storage in our mind (whatever that may mean)? Is it a relation between categories of the same conceptual taxonomy (so that from a diachronic perspective we could really speak of "extension" from one category to the other)? Is it a relation of elements of the same conceptual frame (so that we could speak of conceptual "contiguity": cf. Note 8)? Above all, does it always represent the same kinds of relationship (so that OWNERSHIP : POSSESSION = OWNERSHIP : LOCATION etc.)? In Koch 1993, I proposed a refined onomasiological grid (Figure 2) for the semantic space HAVE/BE. First of all, I distinguish four conceptual categories, which, albeit in varying combinations and constellations and with differences in terminology, regularly occur in publications on central conceptual distinctions in verbal and predicate semantics (cf. Lyons 1967; Pottier 1974: 109-116; Clark 1978; Bickerton 1981: 245; Koch 1981: 260-276; Hagège 1982: 46; Wilson 1983; Hengeveld 1992: 73-126; Heine 1997; Feuillet 1998: 670-673):3 • with only one participant x: 1. ASCRIPTION, as in example (1); 2. EXISTENCE; as in example (2); • each with two participants χ and y: 3. possession, as in example (3); χ = the possessed; y = the possessor; 4. location, as in example (4): χ = the located; y = the locality. (1) (2) (3) (4)

Engl. John% is ill. Engl. THERE ARE many unhappy people*. Engl. Johny HAS a bookX. Engl. The bookx WAS on the tabley.

ASCRIPTION EXISTENCE POSSESSION LOCATION

Cognitive aspects of semantic change and polysemy 281

I would like to stress that the (partly displaced) "adjacency" of these four categories in Figure 2 is only a matter of graphic representation. The actual semantic relationships between the four domains (or parts of them) have to be specified on the conceptual level, as we will see in sections 2-5. On the basis of cross-linguistically observable lexical facts, there are two further conceptual distinctions to be made, which cut across the four main categories represented in Figure 2:

Figure 2. The semantic space HAVE/BE (according to Koch 1993)

First, there is the distinction between the (inherent) thematicity and the rhematicity of participant x. Thus, for LOCATION we can distinguish, with respect to participant x, THEMATIC LOCATION from RHEMATIC LOCATION. (4)

a. Engl. The bookx WAS on the tabley.

THEMATIC LOCATION

(x = thematic)

282 Peter Koch

b.

Engl.

THERE WAS

a bookx on the tabley.

RHEMATIC LOCATION

(oc = rhematic) But as illustrated by Figure 2, this distinction applies to POSSES4 SION and to EXISTENCE as well. At first glance, the connection between thematicity/rhematicity and certain predicative concepts instead of sentences seems a bit misleading, because we are used to assigning degrees of thematicity or rhematicity rather to elements of a sentence (e.g. subject, direct object etc.). Nevertheless, inherent degrees of thematicity or rhematicity attributed to participant slots are essential for differentiating between semantically similar verbs and for relating them to each other (cf. Koch 1981: 93-95, 341-342, 352-356,1991: 297-301; Oesterreicher 1991). This is conclusively demonstrated, for instance, by Sp. estar (THEMATIC LOCATION) as opposed to haber (RHEMATIC LOCATION):

(5)

a. Sp. El librox ESTABA sobre la mesay. 'The book was on the table. ' b. Sp. HABÍA un librox sobre la mesay. 'There was a book on the table. '

THEMATIC LOCATION

(x = thematic) RHEMATIC LOCATION

(x = rhematic)

In the final analysis, the synchronic situations in Spanish (5a/b) and in English (4a/b) are not so different. In my view, English has nowadays two distinct verbal expressions is, i.e., to be (THEMATIC LOCATION) and there is!there are (RHEMATIC LOCATION), even if the second has developped diachronically from the first (cf. also section 5). Obviously, the inherent thematicity-rhematicity reliefs of verbs are only valid for unmarked usage. Marked departures from the rule are always possible (cf. [5c]), but this does not impair the inherent thematicity-rhematicity relief of the verb in question: (5)

c. Sp. Librosx, no HABÍA sobre la mesay. 'Books were not on the table.'

(χ

Φ

rhematic !)

Cognitive aspects of semantic change and polysemy 283

The second distinction I want to introduce concerns the boundedness vs. the unboundedness of the predication, i.e. its limitation in a spatial or temporal respect:5 (2)

a. Engl. THERE ARE many lions* in Africa. BOUNDED EXISTENCE

b. Engl.

THERE ARE many

unhappy people*. UNBOUNDED EXISTENCE

(3) (6)

a. Engl. Johny has a bookx in his hand. BOUNDED POSSESSION b. Engl. Johny has a bookx. UNBOUNDED POSSESSION a. Sp. Albertox esta pálido. 'Albert has a pale complexion. ' BOUNDED ASCRIPTION b. Sp. Albertox es pálido. 'Albert is (being) pale.' UNBOUNDED ASCRIPTION

The conceptual distinctions represented in Figure 2 provide us with a cognitive map6 of the semantic space HAVE/BE, that allows us to retrace characteristic paths of semantic change recurring polygenetically - that is, independently of each other - in different languages of the world and in different epochs of linguistic history. I will allude here to several recurrent patterns and paths (sections 2 and 3) and then later focus on some of them (sections 4 and 5).

2.

From LOCATION via BOUNDED POSSESSION to

POSSESSION

We can take as our first example a path that is extremely wellknown to language typologists. It leads from RHEMATIC LOCATION to BOUNDED POSSESSION and then to POSSESSION in general (cf., e.g., Lyons 1967: 394; Clark 1978: 114; Hagège 1982: 48; Heine 1997: 50-53; for the examples cf. Freeze 1992: 576-577):7 (7)

a. Russ. Ha crojiôy EblJlA Knnrax. 'There was a book on the table.'

RHEMATIC LOCATION

284 Peter Koch

b. Russ. yMeHXy BblJIA KHBT3.\. Ί had a book.' a. Finn. Pöydälläy ON kynäx. (8) 'There is a pencil on the table.' b. Finn. Liisalldy ON miesx. 'Lisa has a husband.' (9) a. Hindi Kamree-mëëy aadmiix HAI. 'There is a man in the room.' b. Hindi Larkee-keey paas kuttaax HAI. 'The boy has a dog.' (10) a. Tagalog MAY gerax sa Ewropay. 'There is a war in Europe. ' RHEM. b. Tagalog MAYrelosx a naanaiy. 'Mom has a watch.'

POSSESSION RHEMATIC LOCATION POSSESSION RHEMATIC LOCATION POSSESSION LOCATION/EXISTENCE POSSESSION

A relationship of BOUNDED POSSESSION (JC, y) is prototypically linked, i.e., is contiguous, to one of LOCATION (.χ, y) within the same frame (what I have with me, is, e.g., in my hands, in my pocket etc.). So, the first step of change must be a metonymicaP one. Via a figure-ground effect we slip from RHEMATic LOCATION to RHEMATIC 9 BOUNDED POSSESSION (p. 291, Figure 3: arrow O). The second step, 10 then, is an extension of meaning from RHEMATIC BOUNDED POSSESSION to RHEMATIC POSSESSION in general (arrow Θ). This last step of extension of meaning, necessary for reaching the target concept in the examples (7b) - (10b), is part of a more general phenomenon of extension from BOUNDED POSSESSION to POSSESSION in general. To name just one further, well attested type of examples: a very special, but surely prototypical kind of BOUNDED POSSESSION is HOLDING SOMETHING IN ONE'S HAND. There are numerous cases of extension of meaning from this prototype to the whole category of POSSESSION (cf. Buck 1988: s.v. 11.11 HAVE; Koch 1991: 291-292; 1995: 30-31):11 (11) (IE. *kap- 'to grasp, to seize' >) Goth, haban 'to hold' > 'to have' (cf. Engl, have; Germ, haben)

Cognitive aspects of semantic change and polysemy

285

(12) (IE. *ghabh- 'to grasp, to seize' >) Lat. habere 'to hold' > 'to have' (13) Lat. tenere 'to hold' > Sp. tener, Sard, tènnere, Southern It. tenere 'to have' (14) Fr. tenir 'to hold' > Guad.Creole (ti)ni 'to have' (15) ChSl.y'efz 'to take' in relation to im-é-ti 'to have'

3.

POSSESSION, EXISTENCE a n d LOCATION

A very widespread change consists of transforming verbs of POSSESSION into verbs of EXISTENCE and further on into verbs of (RHEMATIC) LOCATION. We find the most famous example of this change in the evolution of Latin into the Romance languages. Take, for instance, the meaning 'to have' = RHEMATIC POSSESSION of Lat. habere (cf. also [12]): (16) a. Lat. Marcusy librum* HABET. 'Marcus has a book. '

RHEMATIC POSSESSION

In preclassical and postclassical Latin, we have several attestations of lat. habere as an impersonal expression of RHEMATIC BOUNDED EXISTENCE, as for example in (16b), and of RHEMATIC LOCATION, as in (16c) (cf. also ThLL: s.v. habere, III C 2, p. 2461, 78 2462, 5): (16) b. OLat. Animae pauxillumx in mey HABET. (Naevius, before 200B.C.,cit.ManoliuManea 1985: 111) 'There is very little life in me.' RHEMATIC BOUNDED EXISTENCE

c. VulgLat. Misi tibe... chiloma. Entroy HABET collyramx I ... (2nd Century A.D., cit. Cugusi 1981: 724) Ί sent you a box [...]. There is a loaf of bread in it [...].' RHEMATIC LOCATION

286 Peter Koch

It is well known that many Romance languages have continued to maintain these senses of habere. Sp. haber (which, by the way, totally abandoned the original possessive sense) may serve as an example: (17) a. Sp. En elpuebloy HABÍA muchas casasx.

'In the village, there were many houses.' RHEMATIC BOUNDED EXISTENCE b. = (5 b )

RHEMATIC LOCATION

Another instance is Fr. il y a (examples (18a) and (18b)), which has a somewhat more complicated history (cf. Kawaguchi 1991: 134-177): (18) a. Fr. Dans le village, IL Y AVAIT beaucoup de maisons. 'In the village, there were many houses.' RHEMATIC BOUNDED EXISTENCE

b. Fr. IL Y AVAIT un livrex sur la tabley. 'There was a book on the table.'

RHEMATIC LOCATION

In Romance languages yet another sense developed, based on the Latin innovation, namely RHEMATIC UNBOUNDED EXISTENCE: (17) c. Sp. HAY mucha gente* infeliz.

'There are many unhappy people.' RHEMATIC UNBOUNDED EXISTENCE

(18) c. Fr. iL Y A beaucoup de gensx malheureux.

'There are many unhappy people.' RHEMATIC UNBOUNDED EXISTENCE

There is overwhelming evidence that all these semantic changes are not idiosyncratic facts of Latin-Romance lexical history. We may compare other completely independent instances of analogous changes in different languages (cf. Clark 1978; Bickerton 1981: 66-

Cognitive aspects of semantic change and polysemy

287

67; Buchholz 1989; Hengeveld 1992: 159-160; Heine 1997: 95,137138): (19) a. Mod.Gr. (coll.) Ο Σπύροςy ΕΧΕΙ èva γλήγορο αυτοκίνητοχ. ' Spiros has a fast car. ' RHEMATIC POSSESSION b. Mod.Gr. (coll.) Σ'αυτό τό σπίτι^ ΕΧΕΙ τρεχούμενο νερόχ. (lit. 'in this the house it-has running water') 'In this house there is running water.' RHEMATIC BOUNDED EXISTENCE

c. Mod.Gr.(colL) Στην βαλίτζαΊ ΕΧΕΙ παντελόνια καί παπούτσια*. (lit. 'in the suitcase it-has trousers and shoes') 'There are trousers and shoes in the suitcase.' RHEMATIC LOCATION

(20) a. Pg. Joäoy tinha urna máquina de escreverx. ' Joäo had a typewriter. ' RHEMATIC POSSESSION b. Braz.Pg. Um dia näo tinha mais ratosx. 'One day, there were no more mice.' RHEMATIC BOUNDED EXISTENCE

(21) a. Germ. Martiny hat eine Schreibmaschine* 'Martin has a typewriter. ' RHEMATIC POSSESSION b. Southern Germ. Es hat kein Brotx mehr. 'There is no more bread.' RHEMATIC BOUNDED EXISTENCE (ex. 22-26 cf. Bickerton 1981: 66-67) (22) Guy.Creole Dem get\ wan umanx\ we geti gyal-piknixi. (23) Haw.Creole Get\ wan wahinix\ shi geh wan datax% (24) Hait.Creole Gê\ you fâmx\ ki gèi you pitit-fixi. (25) Papiamentu Tim u muhex\ cu tim un yiu-muhexi. (26) Sào Tomé Creole Te\ ua mwalax\ ku tei ua mina-mosaX2· 'There is a woman who has a daughter.' Verb,: RHEMATIC EXISTENCE Verb2: RHEMATIC (INALIENABLE) POSSESSION

288 Peter Koch

(27) Guad.Creole Ni\

sùx

i ki

nii, ni ι sax\ kipa

nij.

(Poullet et al. 1984: s.v. ni) 'There are those who have got something and those who haven't got anything.' M',: RHEMA TIC UNBOUNDED EXISTENCE ili2: RHEMATIC POSSESSION

(28) a. Bulg. UMAMm EFLHA Jimaaa^ rojiaMa. ' I have a large meadow. ' RHEMATIC POSSESSION b. Bulg. IIpc3 JISRROTOY HMA ΜΗΟΓΟ NJIOMOEEX. (Buchholz 1989: 333) 'In the summertime there are many fruits.' RHEM. BOUNDED EXISTENCE

c. Bulg.

CaMO OrmmoBa^

HflMAWE

raM.

(Buchholz 1989: 334) ' Only Ognj anov was not there. ' RHEMATIC LOCATION (29) a. Nubi áánayfií ma yaláx tinin. (Heine 1997: 137) Ί have two children. ' RHEMATIC POSSESSION b. Nubi ákiljú náa? (Heine 1997: 137) 'Is there food?' RHEMATIC BOUNDED EXISTENCE I think that by taking a cognitive perspective we can try to retrace and understand the different conceptual steps involved in these semantic changes.

4.

From bounded possession to bounded existence

Even if we take into account the rather well attested Latin material in ex. (16b) and (16c), several questions arise. Besides other problems that I do not consider insurmountable,12 it is not entirely clear which was the first sense to develop from an initially possessive sense of habere: RHEMATIC EXISTENCE or RHEMATIC LOCATION? Our oldest attestation of the impersonal Latin verb habere (16b) rather expresses RHEMATIC BOUNDED EXISTENCE. The second oldest attestation (16c) clearly expresses RHEMATIC LOCATION.

Cognitive aspects of semantic change and polysemy

289

Let's suppose that the passage from RHEMATIC POSSESSION to was actually the first step of semantic change (Figure 3: arrow Θ). An argument for this hypothesis could be that verbs like have are the simplest and most typical verbs which serve to introduce entities as existing in the universe of discourse (e.g. Once upon a time, there was a king who had three daughters. They all lived in an old castle in the forest....). In this context, note the Modern Greek personal variant (19d) of (19b): RHEMATic BOUNDED EXISTENCE

(19) d. ModGr. Αυτό τό σπίτι^ΕΧΕΙ τρεχούμενο lit. 'This house has running water.' BOUNDED

νερόν EXISTENCE

In Latin, we can hypothesize an analogous sentence like (16e) as opposed to the "traditional" version (16d)(cf. Bassols de Climent 1948; García Hernández 1992: 167): (16) d. Lat. Domi est multum vinumx. 'In the house there is much wine.' RHEMATIC BOUNDED EXISTENCE

e. Lat. Domus habet multum vinumx. lit. 'The house has much wine.' RHEMATIC BOUNDED EXISTENCE

How could we explain that in (16e) an expression for POSSESSION like habere directly replaced an expression of EXISTENCE like esse in (16d)? Perhaps it would be easier to relate (16e) to (16f): (16) f. Lat. Dominus habet multum vinum^. ' The master has much wine. '

RHEMATIC POSSESSION

The step from (16f) to (16e) can easily be accounted for as a case of metonymy. In a given frame, comprising MASTER, HOUSE and WINE, we observe a figure-ground effect from MASTER-WINE to HOUSE-WINE (cf. Note 8). But at this point (16e) could have been

290 Peter Koch

reinterpreted as a metaphor13 replacing (16d), i.e. as a metaphor that conceptualizes BOUNDED EXISTENCE in terms of BOUNDED POSSESSION. When a given object exists in a given PLACE, the PLACE can be regarded as a CONTAINER that is similar to the POSSESSOR of the object. Metaphor always involves a conceptual leap, but interpreting PLACES as POSSESSORS is a very bold leap, and syntactically, expressions of PLACES, such as participant^ in (19d) and in (16e) are non-prototypical subjects of transitive verbs. The same problem can be observed in well-known English examples like (31) with respect to (30): (30) Engl. Johtiy HAS moneyx in his pocket. RHEMATIC BOUNDED POSSESSION

(31) Engl. The tree(y) HAS a nestx in ity. RHEMATIC BOUNDED EXISTENCE

The metaphorical and syntactic "shock" provoked by sentences like (19d) or (16e) can be attenuated by also indicating, within the same clause, the fact of LOCATIVE BOUNDEDNESS of the EXISTENCE of χ. The locative phrase may be co-referential with the subject (31). But the locative phrase also may be introduced by switching from a personal construction such as (19d) or (16e) (where the subject participant^ still has the form of a POSSESSOR) to an impersonal construction like (19b) (where the subject participant y has simply "evaporated"14 in favour of a new participant y that again expresses the LOCATIVE BOUNDEDNESS of the EXISTENCE of χ). In Latin, we arrive at a construction like (16g) that virtually can have arisen from a blend of types (16d) and (16e):15 (16) g. Lat. Domi habet multum vinum%. 'In the house there is much wine.' RHEMATIC BOUNDED EXISTENCE

Cognitive aspects of semantic change and polysemy

291

According to solutions such a (19b) or (16g), the PLACEPOSSESSOR metaphor frequently and polygenetically provides for impersonal expressions of EXISTENCE (Figure 3: arrow Θ, corresponding to ex. [19b], [20b], [21b], [22]-[26], [27], [28b], and [29b]). It remains to be seen whether or not the reinterpretation of a metonymy as a metaphor like in (16e), followed by a blend with a construction like (16d), is a factor in analogous processes in other languages, too, or if a direct - and very bold - metaphor from RHEMATic POSSESSION to BOUNDED EXISTENCE is the ordinary way.

Figure 3. Recurrent paths of semantic change in the space HAVE/BE

5.

From EXISTENCE to LOCATION and vice versa

Many languages lexicalize RHEMATIC LOCATION as opposed to MATIC LOCATION. French is one of these languages:

THE-

292 Peter Koch

(18) b. (see above) c. Fr. Le livre* ÉTAIT sur la tabley. 'The book was on the table.'

RHEMATIC LOCATION THEMATIC LOCATION

On the other hand, it is striking how many of these languages have identical expressions for RHEMATIC EXISTENCE and for RHEMATIC LOCATION (cf., e.g., [2] and [4b], [17a] and [17b], [18a] and [18b], [19b] and [19c]). We could doubt if it is possible at all to distinguish these two categories (cf. Feuillet 1998: 706-707). In my opinion, the close proximity of the concepts of RHEMATIC LOCATION and RHEMATIC EXISTENCE, which is supported by psychological, philosophical, and linguistic observations (cf. also Lyons 1967: 390391; Bolinger 1977: 99; Holenstein 1980: 32; Bogacki 1988: 24-25), can, and must, be interpreted in terms of contiguity. As we have already seen, the BOUNDEDNESS of EXISTENCE prototypically, though not always, is due to a LOCATIVE limitation. In such a prototypical frame of BOUNDEDNESS of EXISTENCE, there necessarily is a strong contiguity between EXISTENCE and LOCATION. Thus, if you have at your disposal an expression for BOUNDED EXISTENCE like Lat. habere in (16b/g), it easily slips via a figureground effect into an expression of LOCATION (16C). This is a metonymical effect (Figure 3: arrow ©) (cf. Note 8). But in order to grasp this effect, we have to distinguish between two figure-ground constellations: figure = RHEMATIC LOCAIION/ground = RHEMATIC EXISTENCE on the one hand, and figure = RHEMATIC EXISTENCE /ground = RHEMATIC LOCATION on the other. This twofold perspectivization is confirmed by several further linguistic observations (cf. Koch 1993: 181-183 for the first and second points): First, there are languages such as Spanish, that use largely the same lexical unit for RHEMATIC LOCATION and RHEMATIC EXISTENCE (Sp. haber), but that in some specific cases use the same lexical unit for THEMATIC and RHEMATIC LOCATION (Sp. estar), completely excluding the latter verb, however, from the realm of RHEMATIC EXISTENCE:

Cognitive aspects of semantic change and polysemy

(3)

293

c. Sp. Delante del edificioy ESTABAN el cura y dos monaguillosχ. (Koch 1993: 182) 'In front of the building, the priest and two choirboys were standing.' RHEMATIC LOCATION

In such a case, the distinction between RHEMATIC LOCATION and RHEMATIC EXISTENCE is of indéniable importance. Secondly, there seem to be languages, like German, that in a very clear-cut way, i.e., by using different lexical units, mark the distinction between RHEMATIC (and THEMATIC!) LOCATION (Germ, sein/liegen etc.: [33a] and [33b]) and EXISTENCE (Germ, es gibt: [33c]): (32) a. Germ. Auf dem Tischy WAR/LAG ein Buchx. 'There was a book on the table.' RHEMATIC LOCATION b. Germ. Das Buchx WAR/LAG auf dem Tischy. 'The book was on the table. ' THEMATIC LOCATION c. Germ. In Afrikay GIBT ES Löwen*. 'There are lions in Africa.' RHEM. BOUNDED EXISTENCE In such a case, the distinction between RHEMATIC LOCATION and is of crucial importance. Thirdly, the contiguity between EXISTENCE and LOCATION seems to be so strong that we not only observe metonymies from RHEMATIC EXISTENCE to RHEMATIC LOCATION (as described above), but also metonymies in the opposite direction, as, e.g., in English and Italian. How does this come about? First of all, these two languages developed particular expressions for RHEMATIC LOCATION by integrating into the verb for THEMATIC LOCATION an originally deictic element, that is co-referential with the thematic locative participant y (Figure 3: arrow 0). Compare for Italian esserci (33b) in relation to essere (33a): RHEMATIC EXISTENCE

(33) a. It. Il librox ERA sul tavoloy. 'The book was on the table.'

THEMATIC LOCATION

294 Peter Koch

b. It.

un libro* sul tavoloy. 'There was a book on the table.'

C'ERA

RHEMATIC LOCATION

(Cf. for English: there is (4b) in relation to to be (4a)). Compared with the originally neutral verb of LOCATION ((4a), (33 a)), the new verbal expression ((4b), (33b)), integrating Engl. there!It. ci by a phraseological device, designates a restricted type of LOCATION, namely RHEMATIC LOCATION, whereas the verb Engl, to befit, essere restricts its meaning to THEMATIC LOCATION. Now, on diachronic grounds it is obviously the novel RHEMATIC LOCATIVE expression that by way of metonymy also became an expression of BOUNDED EXISTENCE in these languages (Figure 3: arrow ©) and not vice versa: (33) c. It. In Africay CI SONO molti leoni*. 'There are many lions in Africa.'

BOUNDED EXISTENCE

(Cf. for English: there is/are (2a)). If we did not distinguish in principle RHEMATIC EXISTENCE and RHEMATIC LOCATION, we would not be able to adequately retrace the opposite diachronic paths in this semantic space: English, Italian

RHEMATIC BOUNDED POSSESSION

t

RHEMATIC BOUNDED EXISTENCE

or'

e.g. French, Greek

LOCATION

Figure 4. From RHEM. BOUNDED EXISTENCE to RHEM. LOCATION and vice versa

Cognitive aspects of semantic change and polysemy

295

Let us now consider BOUNDED EXISTENCE. This is a special, but prototypical case of EXISTENCE in general. Often speakers, relying heavily on the context, do not even make the boundedness of EXISTENCE explicit (cf. Hengeveld 1992: 97). Thus, (20b)/(21b) could, according to the context, mean that 'there were/is no more mice/bread in the house, in the village, in the country, on earth etc.' Put the other way around, any EXISTENCE is bounded in a certain sense. Even UNBOUNDED EXISTENCE is BOUNDED by the universe of discourse (cf. Bolinger 1977: 99). So it is not surprising, that by extension of meaning, expressions for BOUNDED EXISTENCE - whatever their origin may be (cf. Figure 4) - are extended to EXISTENCE in general (Figure 3: arrow Θ): (19) e. ModGr. ΕΧΕΙ πολλούς δυστυχισμένους ανθρώπουςχ. (cf. [19b]) (lit. 'it-has many unhappy people') 'There are many unhappy people.' RHEMATIC EXISTENCE

(28) d. Bulg. HMA vyffHOBamxopax.

(Buchholz 1989: 332) (cf. [28b])

'There are strange people.' RHEMATIC EXISTENCE (32) d. Germ. Es GIBT viele unglückliche Menschen*. (cf. [32c]) 'There are many unhappy people.' RHEMATIC EXISTENCE (33) ά. li. C'È molta gente* infelice. (cf. [33c]) 'There are many unhappy people.' RHEMATIC EXISTENCE cf. also: (2b) in relation to (2a) for English; (17c) in relation to (17a) for Spanish; (18c) in relation to (18a) for French; (27): for Guadeloupe Creole (ra,).

296 Peter Koch

6.

Final observations

Lexical change within the semantic space HAVE/BE is not arbitrary. We have demonstrated this partly by referring to ample cross-linguistic evidences elaborated by others (esp. Heine 1997) and partly by carrying out a detailed analysis of apparently recurrent processes of change in this area, giving us insight into the nature of "semantic space". It would certainly be interesting to collect more cross-linguistic data that would corroborate the recurrent nature of the metaphors, metonymies, and extensions of meaning interrelating the concepts of POSSESSION, EXISTENCE and LOCATION. In a rather conventional way, we followed grosso modo the diachronic succession of changes, asking ourselves "in which direction" a given semantic class of lexical units moved. This may have been a convenient mode of presentation, but we have to realize that it is totally artificial in relation to what really happens in language change. Speakers do not intend to change anything in language (cf. Coseriu 1958), so lexical items are not "directed" anywhere. On the contrary, speakers use lexical units to communicate in the most efficient and successful manner possible that which they want to express. If speakers affect semantic change (and they do), they affect it not by providing existing words with novel meanings (semasiological perspective), but instead indirectly by expressing things through other and/or new words (onomasiological perspective). So if there really are regularities and "invisible-hand processes" (cf. Keller 1994) in semantic change and if they are, at least to a large extent, by-products of the speakers' activities, then we have to read the results of our investigation "backwards" in order to give them methodological significance (cf. Koch 1997 and in press a). Very tentatively, then, we can make the following statements based on our material:16 A new expression for EXISTENCE in general is created with more than random probability by extension of meaning from an expression for BOUNDED EXISTENCE (Figure 3: arrow ©).

Cognitive aspects of semantic change and polysemy

297

A new expression for (RHEMATIC) BOUNDED EXISTENCE is created with more than random probability either by metonymy from an expression for RHEMATIC LOCATION (arrow ©) or by metonymy, reinterpreted as metaphor, from an expression for RHEMATIC POSSESSION (arrow Θ). A new expression for RHEMATIC LOCATION is created with more than random probability either by metonymy from an expression for BOUNDED EXISTENCE (arrow O ) or by a phraseological device involving restriction from an expression for LOCATION (arrow ©). A new expression for RHEMATIC POSSESSION is created with more than random probability by extension of meaning from an expression for RHEMATIC BOUNDED POSSESSION (arrow Θ). A new expression for RHEMATIC BOUNDED POSSESSION is created with more than random probability by metonymy from an expression for RHEMATIC LOCATION (arrow O). It would perhaps be possible to demonstrate that these paths of semantic change (that must not be used at all) reappear polygenetically in the history of different languages and should be explained by certain cognitive universale underlying our conceptualization of POSSESSION, EXISTENCE a n d LOCATION.

In any case, the activation or non-activation of these types of semantic change and the resulting interlinguistic differences in the articulation of the semantic space HAVE/BE constitute features highly relevant to language comparaison. These features contribute to what I would like to call 'lexical typology' (cf. Koch, in press b). Notes 1.

2.

Cf. Lyons (1967); Langacker (1968); Clark (1978); Foley/Van Valin (1984: 47-53); Kawaguchi (1991); Freeze (1992); Kayne (1993: 4-7). See the critical survey in Heine (1997: 214-222). At least since the seminal work of Bréal (1921: 143-144) we know that with respect to semantic change polysemy is only the other (synchronic) side of the coin. Consequently, the old and the new meaning of a lexical, or grammatical, unit can coexist for a while, but at some moment in

298 Peter Koch

3.

4.

5.

diachrony the old meaning (or sometimes the new meaning) can be abandoned (cf. Koch 1991: 293, 1994a: 203-209; Blank 1997: 114-130). Even though Cognitive Semantics and Grammaticalization Theory have "rediscovered" these fundamental facts (Wilkins 1996: 267-270; Heiiie 1997: 82-83), many linguists still seem perplexed when faced with phenomena of this kind. I do not take into consideration here a possible, and probably necessary, additional category EQUATION (e.g. Athens IS the capital of Greece), because it will not be relevant for the processes I want to analyze. - I am also aware of the internal, possibly prototypical, structure of a large category of POSSESSION, including PHYSICAL POSSESSION, INANALIENABLE POSSESSION, etc. (cf. Taylor 1995: 202-203; Heine 1997: 33-41; cf. also Koch 1981: 314-317, 359). For the present purpose, I can exclude all kinds of INANALIENABLE POSSESSION, the only relevant sub-category being what I call BOUNDED POSSESSION (as exemplified by ex. [3a] and [30] and as represented in Figure 2). - 1 do not go into details concerning the verbal or non-verbal character of the predication in cases where the categories of the semantic space HAVE/BE are expressed by copulas or by zero (cf. Hengeveld 1992: 26-30; on the other hand Feuillet 1991, 1998: 664). For the category of ASCRIPTION, inherent thematicity of participant χ seems to be generalized (cf. example [1]), which does not exclude marked uses of ASCRIPTION predicates as in Engl. Stupid he isn't. - Until now, the thematicity-rhematicity conditions have not been accounted for systematically in studies concerning the semantic space HAVE/BE. However they are often indirectly involved in the description of single categories or parts of this area. Thus, corresponding to THEMATIC vs. RHEMATIC POSSESSION we find: "prédicat d'appartenance" vs. „prédicat de possession" (Benveniste 1960: 196197); "possess2" vs. "possess," (Clark 1978); "ownership" vs. "possession" (Bickerton 1981: 245); "non-presentative possessive" vs. "presentative possessive" (Hengeveld 1992: 125-126); '"belong'-constructions" vs. '"have'-constructions" (Heine 1997: 29-33). Corresponding to THEMATIC vs. RHEMATIC EXISTENCE we find: "modèle Cogito, ergo sum" vs. "modèle il y a + GN" (Feuillet 1998: 730-707; cf. also Koch 1993: 180-181). Corresponding to THEMATIC vs. RHEMATIC LOCATION, we find "nonpresentative locative" vs. "presentative locative" (Hengeveld 1992: 125126, with a clear-cut distinction from "existential"), or "locative/situative" vs. "existential" (Clark 1978; Hagège 1982: 46, 49; Freeze 1992: 553). Note that in the latter, very current, terminology the distinction between RHEMATIC LOCATION and EXISTENCE proper - as in (2) -fades away unduly (cf. section 5 for more details). For the category of LOCATION, inherent boundedness of the predication is logically necessary (cf. ex. [4]). - Until now, the inherent boundedness of

Cognitive aspects of semantic change and polysemy 299

6.

7. 8.

the predication has not been accounted for systematically in studies concerning the semantic space HAVE/BE. However they are often indirectly involved in the description of single categories or parts of this area. Thus, corresponding to BOUNDED VS. UNBOUNDED EXISTENCE we find: "presentative locative" vs. "existential" (Hengeveld 1992: 97, 125Í). Roughly corresponding to BOUNDED vs. UNBOUNDED POSSESSION we find: "physical possession" vs. other types of "possession" (Heine 1997: 34; cf. above n. 3). Corresponding to BOUNDED VS. UNBOUNDED ASCRIPTION, we find: "énoncé [à localisation temporelle et modale] et situé par rapport au locuteur" vs. "énoncé hors de toute localisation temporelle et modale et hors de la subjectivité du locuteur" (Benveniste 1960: 160, cf. also 167); "attribut" vs. "essence" (Hagège 1982: 48-49); "qualification incidentielle" vs. "qualification essentielle" (cf. Feuillet 1998: 725; cf. 711-725). For the principle of semantic or cognitive maps, cf. Anderson (1982); Bybee (1985: 195-196); Croft et al. (1987); Haspelmath (1997: 59-62). Cf. however the critical remarks above concerning the interpretability of "adjacencies". In Heine's 100 language sample, this designation type for POSSESSION covers 20.9% (cf. Heine 1997: 75). As for this understanding of metonymy and contiguity in relation to prototypicality and figure-ground effects, cf. Koch (1995: 40-41, 1999); Blank (1997: 235-243). (What Traugott and König [e.g. 1991] call "pragmatic strengthening" or "conventionalization of a conversational implicature" is a kind of metonymy.) Note that the notion of prototypicality that we have to apply here is an onomasiological one (cf. Koch 1996). In this sense, contiguity is an external relation between distinct categories. So, it would not be legitimate to denominate a metonymical process, like the one described, as a "metonymical extension" of the category RHEMATIC LOCATION.

9.

10.

11.

Just as I use "RHEMATIC LOCATION" for "LOCATION with a rhematic participant x", so also "RHEMATIC BOUNDED POSSESSION" for "BOUNDED POSSESSION with a rhematic participant x" etc. As for this understanding of (true) extension to a whole category, a process that is virtually, but not necessarily, in relation to prototypicality, cf. Koch (1995: 30-31); Blank (1997: 200-206). According to Heine (1997: 75), the so called '"action'-type" for the designation of POSSESSION covers 13.6% of the 100-language sample. The term "action", however is misleading in this context, because the immediate conceptual basis of these designations for POSSESSION is HOLDING IN ONE'S HAND. (The designation for HOLDING IN ONE'S HAND may in turn go back to a word designating the action of GRASPING/TAKING; but for some languages, as far as we know, this is not at all certain: cf. L. tenere 'hold'

300 Peter Koch

12.

13.

14. 15.

16.

as antecedent in ex. [13]-[14]. Anyway, it is not the concept of ACTION as such that motivates new designations for POSSESSION, but the - in my view prototypical - case of BOUNDED POSSESSION). First of all, Greek influence as been invoked (cf. Svennung 1932: 84; Luque Moreno 1978: 138-140; García Hernández 1992: 166-167). If this should be correct, we would have to count the Greek and the Romance attestations of the POSSESSION EXISTENCE pattern as only one occurrence. But this would not detract from the broad polygenetic evidence for this pattern. Secondly, a different meaning of L. habere has been considered the source of the existential meaning, namely 'to be in a state' (with possible Greek parallels in this case as well). This explanation cannot be excluded a priori on lexicological grounds, so that it would be possible to interpret the syntactically ambiguous noun phrase animae pauxillumx in (16b) not as the object of an impersonal verb habet, but as the subject of a personal form habet (cf. Luque Moreno 1978: 138-146; García Hernández 1992: 170; Manoliu Manea 1985: 111; indeed, habet is also attested with unambiguous subject participants x: cf. ThLL: s.v. habeo, III C 2, p. 2461, 65-77). But even on this assumption, one has to postulate that the impersonal type domiy habet multum vinumx=D0, lit. 'in the house it-has much wine', is the result of a contamination between the personal type domiy habet multum vinumx=s, lit. 'in the house is much wine', and the personal type domusrS habet multum vinum,^, lit. 'the house has much wine' (Luque Moreno 1978: 145). Thus, it seems impossible to totally disregard the role of the latter, metonymical/metaphorical type on a clearly possessive basis that is much closer to the etymological meaning of habere (cf. ex. [12]) and fits perfectly into the polygenetic pattern illustrated by ex. (19) ff. For the understanding of metaphor cf. Lakoff/Johnson (1980); Lieber (1992: 14-82); Koch (1994a, 1995: 39-40); Blank (1997: 157-171). Since metaphor always involves a conceptual leap, the (presumed) similarity is an external relation between distinct categories. So, it would not be legitimate to denominate a metaphorical process, like the one described, as a "metaphorical extension" of the category POSSESSION. - As for reinterpreting metonymies as metaphors, cf. Taylor (1995: 138); Koch (1997: 234). For the "evaporation" of highly thematic subject participants in the genesis of impersonal verbal expressions cf. Koch (1994b). Interestingly, our example (16c), although it goes one step further (and expresses RHEMATIC LOCATION), is ambigous with respect to types (16e)/(31) and (16g), since the participant y is anaphorical and therefore implicit. It would be more accurate to begin each statement conditionally: "If a new expression X is created...".

Cognitive aspects of semantic change and polysemy 301

References Anderson, Lloyd 1982 The 'Perfect' as a universal and as a language particular category. In: Paul Hopper (ed.), Tense and Aspect: Between Semantics and Pragmatics, 227-264. (Typological Studies in Language 1.) Amsterdam / Philadelphia: Benjamins. Benveniste, Emile 1960 "Etre" et "avoir" dans leurs fonctions linguistiques. In: Emile Benveniste, Problèmes de linguistique générale. Volume 1, 187-207. Paris: Gallimard. Bickerton, Derek 1981 Roots of Language. Ann Arbor: Karoma. Blank, Andreas 1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. (Beihefte zur Zeitschrift für Romanische Philologie, 285.) Tübingen: Niemeyer. Bogacki, Christophe 1988 Les verbes à argument incorporé. Langages 89: 7-26. Bolinger, Dwight 1977 Meaning and Form. (English Language Series 11.) London / New York: Longman. Bassols de Climent, Mariano 1948 Origen de la construcción impersonal del verbo habere. Revista de estudios clásicos 3: 215-225. Bréal, Michel 5 1921 Essai de sémantique (Science des significations). Paris: Hachette. Buchholz, Oda 1989 Zu Konstruktionen mit impersönlich gebrauchtem 'haben' in den Balkansprachen. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42: 329-338. Buck, Carl D. 3 1988 A Dictionary of Selected Synonyms in the Principal Indo-European Languages. Chicago / London: University of Chicago Press. Bybee, Joan L. 1985 Morphology: A Study of the Relation Between Meaning and Form. (Typological Studies in Language 9.) Amsterdam / Philadelphia: Benjamins.

302 Peter Koch

Clark, Eve V. 1978 Locationals: existential, locative, and possessive constructions. In: Joseph H. Greenberg (ed.), Universals of Human Language. Volume 4. Syntax, 85-126. Stanford: Stanford University Press. Coseriu, Eugenio 1958 Sincronía, diacronia e historia. El problema del cambio lingüístico. Montevideo: Universidad de Montevideo. Croft, William, Hava Bat-Zeev Shyldkrot and Suzanne Kemmer 1987 Diachronic semantic processes in the middle voice. In: Anna Giacalone Ramat, Onofrio Carruba and Giuliano Bernini (eds.), Papers from the 7th International Conference on Historical Linguistics, 179-192. (Current Issues in Linguistic Theory 48.) Amsterdam / Philadelphia: Benjamins. Cugusi, Paolo 1981 Gli ostraka latini dello Wâdi Fawâkhir. Per la storia del latino. In: Letterature comparate. Problemi e metodi. Studi in onore di Ettore Paratore, II, 719-753. Bologna: Pàtron. Feuillet, Jack 1991 Phrases nominales et phrases à verbe "être". EUROTYP Working Papers n° 1. 1998 Typologie de "être" et phrases essives. In: Jack Feuillet (ed.), Actance et Valence dans les Langues de l'Europe, 663-767. (Empirical Approaches to Language Typology. EUROTYP 20-2.) Berlin / New York. Foley, William and Robert D. Van Valin jr. 1984 Functional Syntax and Universal Grammar. (Cambridge Studies in Linguistics 38.) Cambridge: Cambridge University Press. Freeze, Ray 1992 Existentials and other locatives. Language 68: 553-595. García Hernández, Benjamin 1992 Nuevos verbos impersonales en latin tardío e influencia griega. In: Maria Iliescu and Werner Marxgut (eds.), Latin vulgaire - latin tardif III, 159-172. Tübingen: Niemeyer. Hagège, Claude 1982 La structure des langues. (Que sais-je? 2006.) Paris: Presses Universitaires de France. Haspelmath, Martin 1997 Indefinite Pronouns. Oxford: Clarendon. Heine, Bernd 1997 Possession. Cognitive Sources, Forces and Grammaticalization. (Cambridge Studies in Linguistics 83.) Cambridge: Cambridge University Press.

Cognitive aspects of semantic change and polysemy 303

Hengeveld, Kees 1992 Non-verbal Predication. Theory, Typology, Diachrony. (Functional Grammar Series 15.) Berlin/New York: Mouton de Gruyter. Holenstein, Elmar 1980 Von der Hintergehbarkeit der Sprache, (suhrkamp taschenbuch Wissenschaft 316) Frankfurt/M.: Suhrkamp. Kawaguchi, Junji 1991 AVOIR et les problèmes de la localisation en français. Tokyo: France Tosho. Kayne, Richard S. Towards a modulary theory of auxiliary selection. Studia 1993 Linguistica 47: 3-31 Keller, Rudi 2 Sprachwandel. Von der unsichtbaren Hand in der Sprache. (Uni1994 Taschenbücher 1567.) Tübingen / Basel: Francke. Koch, Peter 1981 Verb - Valenz - Verfügung. Zur Satzsemantik und Valenz französischer Verben am Beispiel der Verftigungsverben. (Reihe Siegen 32.) Heidelberg: Winter. 1991 Semantische Valenz, Polysemie und Bedeutungswandel bei romanischen Verben. In: Koch and Krefeld (eds.), 279-306. HABEN und SEIN im romanisch-deutschen und im innerromanischen 1993 Sprachvergleich. In: Giovanni Rovere and Gerd Wotjak (eds.), Studien zum romanisch-deutschen Sprachvergleich, 177-189. (Linguistische Arbeiten 297.) Tübingen: Niemeyer. Gedanken zur Metapher - und zu ihrer Alltäglichkeit. In: Annette 1994a Sabban and Christian Schmitt (eds.), Sprachlicher Alltag. Linguistik - Rhetorik - Literaturwissenschaft. Festschrift für Wolf-Dieter Stempel, 201-225. Tübingen: Niemeyer. 1994b Dépersonnalisation (et repersonnalisation). L'information grammaticale 62: 9-11. 1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: eine kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46. La sémantique du prototype: sémasiologie et onomasiologie? Zeit1996 schrift für französische Sprache und Literatur 106: 223-240. La diacronia quale campo empirico della semantica cognitiva. In: 1997 Marco Carapezza, Daniele Gambarara and Franco Lo Piparo (eds.), Linguaggio e cognizione, 225-246. (Società di Linguistica Italiana 37.) Roma: Bulzoni. 1999 Frame and contiguity. On the cognitive bases of metonymy and certain types of word formation. In: Klaus-Uwe Panther and Günter

304 Peter Koch

Radden (eds.), Metonymy in Language and Thought, 139-167 (Human Cognitive Processing) Amsterdam / Philadelphia: Benjamins. in press a TREE and FRUIT: A cognitive-onomasiological approach. Studi Italiani di Linguistica Teorica ed Applicata. in press b Lexical typology. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Typology and Universals. An International Handbook. Berlin / New York: de Gruyter. Koch, Peter and Thomas Krefeld (eds.) 1991 Connexiones Romanicae. Dependenz und Valenz in romanischen Sprachen. (Linguistische Arbeiten 268.) Tübingen: Niemeyer. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Langacker, Ronald W. 1968 Observations on French possessives. Language 44: 51-75. Liebert, Wolf-Andreas 1992 Metaphernbereiche der deutschen Alltagssprache. Kognitive Linguistik und die Perspektiven einer Kognitiven Lexikographie. (Europäische Hochschulschriften I, 1355.) Frankfurt/M.: Lang. Luque Moreno, Jesús 1978 En torno al sintagma "haber impersonal + sustantivo" y sus orígenes latinos. Revista española de lingüistica 8: 125-147. Lyons, John 1967 A note on possessive, existential and locative sentences. Foundations of Language 3: 390-396. Manoliu-Manea, Maria 1985 Tipologiá e historia. Elementos de sintaxis comparada románica. (Biblioteca románica hispánica II, 337.) Madrid: Gredos. Oesterreicher, Wulf 1991 Verbvalenz und Informationsstruktur. In: Koch and Krefeld (eds.), 349-384. Pottier, Bernard 2 1985 Linguistique générale. Théorie et description. (Initiation à la linguistique Β, 3.) Paris: Klincksieck. Poullet, Hector, Sylviane Telchid and Danièle Montbrand 1984 Dictionnaire des expressions du créole guadeloupéen. Fort-deFrance: Hatier. Raible, Wolfgang 1981 Von der Allgegenwart des Gegensinns (und einiger anderer Relationen). Zeitschrift für Romanische Philologie 97: 1-40.

Cognitive aspects of semantic change and polysemy 305

Svennung, Josef 1932 Wortstudien zu den spätlateinischen Oribasiusrezensionen. Uppsala: Almqvist & Wikseil. Taylor, John R. 2 1995 Linguistic Categorization. Prototypes in Linguistic Theory. Oxford: Clarendon. ThLL = Thesaurus linguae Latinae. 8 vol. Leipzig: Teubner, 1936-66. Traugott, Elizabeth C. and Ekkehard König 1991 The semantics-pragmatics of grammaticalization. In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, I, 189-218. (Typological Studies in Language 19.) Amsterdam / Philadelphia: Benjamins. Wilkins, David P. 1996 Natural tendencies of semantic change and the search for cognates. In: Mark Durie and Malcolm Ross (eds.), The Comparative Method Reviewed, 264-304. Oxford: Oxford University Press. Wilson, Bob 1983 An examination of cross-linguistic constraints on the lexicalization of predications of ownership, possession, location and existence. University of Hawaii Working Papers in Linguistics 15/2: 1-18.

List of contributors

Andreas Blank Philipps-Universität Marburg Institut fur Romanische Philologie Marburg, GERMANY David D. Clarke University of Nottingham Department of Psychology Nottingham, UNITED KINGDOM Dirk Geeraerts Katholieke Universiteit Leuven Departement Linguistiek Leuven, BELGIUM Peter Koch Eberhard-Karls-Universität Tübingen Romanisches Seminar Tübingen, GERMANY Ekkehard König Freie Universität Berlin Institut fur Englische Philologie Berlin, GERMANY Thomas Krefeld Ludwig-Maximilians-Universität München Institut für Romanische Philologie München, GERMANY Ronald W. Langacker University of California, San Diego Department of Linguistics La Jolla, California, USA

308

Conributors

Helmut Lüdtke Christian-Albrechts-Universität Kiel Romanisches Seminar Kiel, GERMANY Brigitte Neriich University of Nottingham Department of Psychology Nottingham, UNITED KINGDOM François Rastier Université Paris IV, Sorbonne Centre de Linguistique Française Paris, FRANCE Peter Siemund Freie Universität Berlin Institut fur Englische Philologie Berlin, GERMANY JohnR. Taylor University of Otago School of Languages, Linguistics Dunedin, NEW ZEALAND Elizabeth Closs Traugott Stanford University Department of Linguistics Stanford, California, USA Beatrice Warren Lund University Department of English Lund, SWEDEN

Index

acceptability threshold 120, 122, 124, 127,128, 131, 137, 140 adverbial 178 adverb 178, 179, 180, 181, 186, 189, 190 agent 152, 169, 170 agentive 148, 161 agentive subject 148 agentivity 161, 162 appended senses 216,230,233 attenuated 158, 161, 166, 167, 170 attenuation 3, 148, 152, 153, 156, 159, 160, 162, 163, 166, 168, 169, 170, 171, 172,189,291 attractor 120, 121, 122, 123, 135 auxiliarization 58

basin 120, 121, 122, 123, 124, 126, 129, 131, 135, 137, 138 blend 216,291,292 body parts 7, 30, 242, 243, 244, 247, 259,266,267

categorization 6, 216, 217, 218, 260,262,266,269,273,274 cognition 198 cognitive ease 259, 273,275 cognitive semantics 198 collusion 55, 58 concept 19, 20, 22, 23, 27, 39, 40, 41,42 conceptual blending 217 contiguity 6, 7, 8, 52, 74, 79, 200, 225, 228, 232, 269, 280, 293, 294

contrary to the speaker's view 184 contrast 21, 23, 24, 26, 27, 28, 30, 31,32, 33,35, 37,41,42 control 147, 148, 152, 160, 162, 163, 166, 168, 169, 170, 171, 172 counter-expectation 177, 179, 188 creativity 12,50,51 cross-over 52, 54, 55, 56, 57 cross-space 216

definite article 244,253,254 designation 20, 23, 24, 25, 26, 27, 30, 38, 39, 41 discourse 3, 33, 58, 64, 77, 81, 112, 128, 134, 135, 142, 180, 181, 187, 190, 191, 198, 204, 205, 207,238,253, 255,290,296 discourse marker 178, 180, 181, 182, 183, 184, 187, 189, 190, 191 domain (conceptual) 6, 7, 11, 39, 40, 41, 42, 72, 81, 83, 111, 112, 117, 118, 127, 133, 134, 142, 150, 156, 159, 161, 163, 178, 181, 216, 228, 229, 232, 233, 238, 246, 247, 253, 254, 255, 260, 263, 265, 266, 270, 271, 273,281 dynamic systems 118 dysphemism 82

efficiency 3, 8, 63, 64, 65, 66, 72, 73, 78, 100, 101, 103, 104, 105, 106

310 Index

euphemism 65, 82,209,243,250 explanation (of semantic change) 92, 98, 100, 101, 103, 104, 106 expressivity 3, 8, 63, 65, 66, 70, 73, 78, 83, 106 extension of meaning 3, 7, 74, 82, 76, 96, 153, 204, 223, 232, 285, 296,297,298

feature 32, 34, 37,41 figure-ground effect 285, 290,293 figures of speech 7, 198, 204, 205, 208,210,229 focus 238, 240, 241, 243, 244, 246, 250,251,252,257 form categories 217,218,221 frame (conceptual) 6, 7, 8, 39, 72, 75,76, 77, 118,127,232 functional motivation (of semantic change) 98, 99, 101, 102, 103, 104,105,106

generalization 218,224,225 genus-for-species 199, 200, 201, 204,206,207,210 grammaticalization 7, 9, 49, 57, 147, 148, 151, 156, 157, 160, 163, 165, 168, 169, 172, 179, 231 grammaticization see grammaticalization grammaticized 148, 162, 163, 168 group (of concepts) 260, 261, 262, 269, 270,273

hearer-oriented strategy 3, 8, 64, 65, 103, 104, 105, 106 homonymie clash 53, 54, 69, 70, 101

homonymy 65, 69, 70, 101, 181, 272 honorific pronouns 243

Inflectional Phrase Adverb 178, 179, 180, 181, 182, 183, 184, 185,186,187, 189, 190, 191 instance (of a schema) 20, 21, 25, 41 intensifier 4, 10, 237,238,239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 250, 251, 252, 253, 255,257 interpretative mechanism 216, 224, 225,232 interpretative strategy 223, 232,233

language change 8, 10, 38, 50, 52, 61, 62, 63, 64, 65, 66, 72, 297 lexical change 2, 8, 10, 50, 104, 106,215,216, 232 linguistic sign 19, 22, 23, 26, 34, 39,42 luxury borrowing 265, 375

meaning (criterial and non-criterial features of) 216, 218, 221, 232 metaphor 2, 3, 4, 7, 8, 9, 10, 51, 52, 62, 63, 64, 65, 70, 71, 72, 73, 82, 83, 84, 99, 114, 115, 117, 141, 159, 188, 198, 199, 200, 201, 202, 204, 205, 207, 208, 210, 216, 217, 225, 226, 227, 228, 229, 232, 234, 290, 291, 292, 298 metaphorical 38, 115, 141, 149, 159, 188, 238,291 metaphorization 222

Index 311

metaphor-metonymy distinction 228 metonym see metonymy métonymie 59, 80, 115, 188, 201, 202, 227, 231, 245, 248, 269, 270,272,273,285,293 metonymy 2, 3, 4, 7, 8, 9, 10, 31, 51, 52, 64, 65, 72, 75, 80, 82, 83, 99, 114, 115, 117, 139, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 210, 225, 226, 227, 228, 229, 230, 231, 232, 233, 235, 269, 272, 273, 290, 292, 295,297,298 motivation for borrowing 266

Natural Lexicology 98, 106 Natural Phonology 92, 98, 102, 103, 104, 105 necessity borrowing 265, 275 neutralisation 30,31

objectification 190, 191 onomasiological 10, 11, 28, 67, 112, 113, 120, 136, 204, 265, 280,297 onomasiology 9, 112, 138

paragon 110,115,116,123 part-of-relation 203 part-whole-relation 123, 199, 200, 201, 202,203,206,242,248 perceptive experience 275 perceptive saliency 274 performance 50, 51 polysemous 30, 33, 52, 113, 114, 119, 138, 181 polysemy 2, 30, 31, 32, 33, 34, 38, 51, 52, 55, 65, 68, 70, 75, 80,

180, 181, 243, 249, 251, 279, 280,298 possession 11, 76, 280, 281, 283, 284, 285, 286, 288, 289, 290, 291,292,297, 298 pragmatic strengthening 3, 8, 189 profile 40, 41, 148, 149, 155, 165, 166, 168,169, 171 profiled 40, 149, 150, 151, 152, 153, 154, 155, 156, 158, 159, 160, 163, 165, 166, 168, 169, 171 profiling 6, 8, 20, 39, 41, 148, 149, 150, 156 prototype 2, 3, 6, 10, 18, 20, 26, 27, 28, 30, 31, 32, 33, 64, 72, 76, 77, 110, 111, 113, 123, 138,285 prototype theory 91, 93, 94, 95, 97, 98, 100 prototypical 6, 7, 10, 21, 32, 34, 77, 78,81, 111,285,291,293,295 prototypical change 76 prototypicality, prototype effects 91,92, 93,98, 99, 101, 105, 106

qualitative threshold 121, 122, 123, 124, 127

raising 160, 171, 172, 174 recategorization 259, 266, 268, 270, 273, 275 reflexive anaphor 243,246,247 restriction of meaning 76, 81, 269, 273 rhetoric 7, 114, 139, 177, 190, 191, 198, 199,200, 201,202

semantic change 2, 3, 5, 6, 7, 11, 37, 49, 61, 62, 63, 66, 67, 68, 69,

312 Index

71, 72, 75, 77, 78, 79, 80, 81, 83, 84, 92, 93, 95, 96, 97, 99, 100, 101, 106, 107, 108, 148, 177, 179, 181, 188, 189, 191, 223, 237, 238, 243, 256, 272, 257, 279, 280, 284, 290, 292, 297, 298 semantic extension see extension of meaning semantic primitive 29 semantic space 4, 297, 280, 282, 284,295,297, 298 semantic restriction see restriction of meaning semasiological 67, 91, 94, 98, 112, 113, 114, 118, 136, 140, 204, 297 semasiology 9, 111, 138 sentence meaning vs. utterance meaning 21 signified 19, 21,22,, 41 signifier 19, 21, 22, 25,41 similarity 21, 56, 62, 79, 200, 228 speaker-oriented strategy 2, 64, 65, 78, 103, 104, 105, 106 strategies working out interpretations 227 sub-categorization 261 sub-groups (conceptual) 260, 261, 267 subject 148, 149, 150, 151, 152, 154, 156, 157, 158, 159, 160, 161, 162, 163, 164, 166, 168, 169, 170, 171, 172 subjectification 3, 6, 8, 9, 10, 147, 148, 150, 152, 153, 154, 156, 157,, 158, 168, 173, 177, 180, 188, 189,190 subjectivity 8, 179, 180, 188, 190 synecdoche 117

Synecdoche 3, 7, 10, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206,207,208, 209,210

taxeme 110, 111, 112, 115, 118, 120, 122, 123, 124, 126, 127, 128, 133, 134, 136, 137, 140, 142 torso-extremities-model 260, 267,269,273 trajector 149, 150, 152, 153, 155, 156, 157, 158, 160, 166, 168, 169, 171 transparency 101, 103, 104, 147, 148, 160, 161, 162, 168, 171, 172, 174 transparent 162, 163, 164, 171

117, 125, 135, 261, 154, 162, 105, 164,

unprofiled 155, 156, 166 utterance meaning see sentence meaning 21

valuation 110, 115, 116, 124, 125, 128, 138 value 109, 110, 113, 126, 127, 128, 137, 138,139, 141 Verb Adverb 178, 183, 185, 186, 190

whole-part-relation see part-wholerelation widening of meaning see extension of meaning