From Polysemy to Semantic Change: Towards a typology of lexical semantic associations

From Polysemy to Semantic Change

Studies in Language Companion Series (SLCS) This series has been established as a companion series to the periodical Studies in Language.

Editors Werner Abraham University of Vienna

Michael Noonan

University of Wisconsin, Milwaukee

Editorial Board Joan Bybee

Christian Lehmann

Ulrike Claudi

Robert E. Longacre

Bernard Comrie

Brian MacWhinney

University of New Mexico University of Cologne Max Planck Institute, Leipzig University of California, Santa Barbara

William Croft

University of New Mexico

Östen Dahl

University of Stockholm

Gerrit J. Dimmendaal University of Cologne

Ekkehard König

Free University of Berlin

University of Erfurt

University of Texas, Arlington Carnegie-Mellon University

Marianne Mithun

University of California, Santa Barbara

Edith Moravcsik

University of Wisconsin, Milwaukee

Masayoshi Shibatani

Rice University and Kobe University

Russell S. Tomlin

University of Oregon

Volume 106 From Polysemy to Semantic Change. Towards a typology of lexical semantic associations Edited by Martine Vanhove

From Polysemy to Semantic Change Towards a typology of lexical semantic associations

Edited by

Martine Vanhove Llacan (Inalco, CNRS), Fédération TUL

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data From polysemy to semantic change : towards a typology of lexical semantic associations / edited by Martine Vanhove. p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 106) Includes bibliographical references and index. 1.  Semantics, Historical. 2.  Polysemy. 3.  Typology (Linguistics)  I. Vanhove, Martine. P325.5.H57F76

2008

401'.43--dc22 isbn 978 90 272 0573 5 (Hb; alk. paper)

2008031821

© 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Semantic associations: A foreword Martine Vanhove

vii

Part 1. State of the art Approaching lexical typology Maria Koptjevskaja-Tamm

3

Part 2. Theoretical and methodological issues Words and their meanings: Principles of variation and stabilization Stéphane Robert

55

The typology of semantic affinities Bernard Pottier

93

Cognitive onomasiology and lexical change: Around the eye Peter Koch

107

Mapping semantic spaces: A constructionist account of the “light verb” xordæn “eat” in Persian Neiloufar Family

139

Semantic maps and the typology of colexification: Intertwining polysemous networks across languages Alexandre François

163

A catalogue of semantic shifts: Towards a typology of semantic derivation Anna A. Zalizniak

217

Semantic associations and confluences in paradigmatic networks Bruno Gaume, Karine Duvignau & Martine Vanhove

233

Part 3. Case studies About “Eating” in a few Niger-Congo languages Emilio Bonvini

267

 From Polysemy to Semantic Change

Eating beyond certainties Christine Hénault

291

From semantic change to polysemy: The cases of “meat/animal” and “drink” Pascal Boyeldieu

303

Is a “friend” an “enemy”? Between “proximity” and “opposition” Sergueï Sakhno & Nicole Tersis

317

Semantic associations between sensory modalities, prehension and mental perceptions: A crosslinguistic perspective Martine Vanhove

341

Cats and bugs: Some remarks about semantic parallelisms Michel Masson

371

General index Index of languages Index of names

387 397 401

Semantic associations: A foreword Martine Vanhove

Llacan (Inalco, CNRS), Fédération TUL

The book whose outline is presented here is the outcome of a joint project, which began during the fall of 2002, at the Fédération Typologie et Universaux Linguistiques of the French National Centre for Scientific Research (CNRS, http://www.typologie. cnrs.fr), and of various collaborations with several European researchers. It gathered field linguists, semanticists, cognitivists, typologists, and one NLP specialist around different issues concerning the “Typology of semantic associations”, i.e., polysemy, heterosemy and semantic change at the lexical level. For a long time typologists have had doubts about the possibility of crosslinguistic studies of the lexicon since the relevant phenomena appear to be highly idiosyncratic and there seems to be no convincing solution to the problem of delimiting the field of inquiry. In the last few decades, however, there has been a growing interest in crosslinguistic research on the lexicon, from different perspectives and within different theoretical frameworks (see Koptjevskaja-Tamm et al. 2007). In the first part of the present volume, Maria Koptjevskaja-Tamm provides a very detailed state of the art article of the research in this domain. She shows how the very term of “lexical typology”, although seldom defined, is nonetheless often used as if it were self-explanatory, leading to a great diversity of approaches, and a lack of consensus about the domain of research itself. Her paper defines and illustrates various existing and possible approaches to lexical and semantic typology. She discusses some of the problems inherent in these approaches, and suggests reflections on possible directions for future research. Her review makes it clear that the papers presented in this volume are concerned by only one aspect of a vast issue, namely semantic shifts, be they synchronical or diachronical. Our joint project started from the well-known facts that the polysemy of lexical units is a universal phenomenon, and that within one linguistic phylum or one geographical area, the persistent occurrence and regular patterns of certain polysemies and semantic parallels are useful tools for establishing genetic affiliations. These observations go along a growing awareness that there exist also recurrent patterns of semantic associations, which transcend genetic or areal boundaries. This led us to a more general, typological survey about the possible existence of semantic universals that would go beyond the usual genetic or areal classifications. This in turn led to a

 From Polysemy to Semantic Change

discussion of several issues related to the fact that semantic associations are problematic from a comparative and crosslinguistic perspective: (i) the problem of the organization of the different meanings at the level of the individual linguistic system; (ii) the problem of how we have access to the relevant meaning, from a cognitive viewpoint; (iii) the problem of how to disambiguate meaning in context. In the second part of the book, within these general semantic and cognitive issues, Stéphane Robert examines the different reasons behind linguistic variation, the rule in natural languages. She proposes a multidimensional approach to the layering of the lexicon and its semantic organization, aiming at explicating the principles of variation and stabilization of the lexical network. She shows how the linguistic units trigger representations which are caught up in a complex network of relations, both language internal and external, semantic and formal, what she calls “the depth dimension of language”. This includes cultural categorization, referential paths, internal architecture (figure and ground), metaphor and metonymy, referential scales, application domains, scenarios and semantic universes, networks of formal and semantic relations between terms, connotations and social roles, associations between linguistic and extralinguistic representations, structured relations within a specific verbal or situational context, pragmatic inferences, (de)motivations, landmarks, attractors and “active zones”, prototypes, “semantic isotopics”, and anchor points. In addition, she also points to the two modes for forming meaning, construction or quotation, which both play a role in the construction of meaning and reference access. This overview, deeply entrenched in cognitive linguistics, provides a comprehensive background for a better understanding of the reasons behind semantic associations. Several other theoretical and methodological approaches are also proposed and discussed. They all aim at accounting for and studying the phenomenon of polysemy, heterosemy, and semantic change in the lexicon from a cognitive and typological perspective. Bernard Pottier offers a classification of the different mental processes at work in the domain of semantic associations (denominated as semantic “affinities”), which could be used as a tool for crosslinguistic comparison and typological classification in future research. Each process is named, described, visualized by use of dynamic graphs whenever possible, and exemplified in various languages and language phyla. Using the methodology of diachronic cognitive onomasiology, as developed in two projects at Tübingen University, the decolar, and the Lexical Change – Polygenesis – Cognitive Constants, Peter Koch discusses polygenetic semantic parallels in semantic change. He focuses on semantic parallels that are due to fundamental cognitive constants, leaving aside those that are triggered by genetic relationships, or by linguistic and cultural contacts. Through a two-dimensional grid, he identifies the cognitive and formal relations between a source and a target concept. He then shows, for the semantic domain of the eye (eyelash, eyebrow, eyelid, and eyeball),



Semantic associations: A foreword 

how cognitive constants can be revealed crosslinguistically, even in cases of complete diversity of formal devices. He provides a list of all the cognitive solutions chosen by the languages of the sample to create lexical innovations. His study explains how the degree of saliency of a concept interacts with semantic shift, producing or not patterns of demotivation and remotivation within the lexicon. Together with cultural and linguistic categorization, it also provides an explication for the different options chosen by the languages for lexical conceptualisation and gives insight to the ongoing debate on linguistic relativity. Within the framework of Construction Grammar, which accommodates idiosyncrasy, polysemy, compositionality, and productivity, Neiloufar Family has developed an analytical tool to map semantic regularities in semantic spaces. This is illustrated with the study of the semantic spaces of the Persian “light verb” xordæn “to eat”, which forms the basis of over 200 verbal notions. Her analysis implies that a light verb’s semantic space is populated by “notional islands” where groups of light verb constructions, expressing similar notions, combine the light verb with a restricted, but large, class of preverbs. Each island possesses linguistic and cognitive properties that allow intuitive disambiguation. The result of this analysis is a fresh insight into general linguistic issues, such as meaning construction, productivity, and compositionality. Furthermore, due to the compositional structure of the Persian verbal system, this study provides a basis for crosslinguistic investigations comparing and analysing processes of verb formation as they have evolved in different languages. It also allows for a crosslinguistic analysis of the status of these verbs in order to gain a better understanding of the semantic structures common to different languages. Alexandre François arrived at similar results within a crosslinguistic approach. He offers an adaptation to the lexicon of the methodological tool of semantic maps, now common among typologists for the comparison of grammatical devices. His aim is to define a standard of comparison, whereby the semantic network corresponding to a polysemous unit in one language (e.g., English right) may be compared with a similar semantic network in a different language (e.g., French droit). The method proposes to overcome the complexity of the polysemy networks and the pitfalls of idiosyncrasy, and to make crosslinguistic comparison possible, by breaking down the meaning into simpler, atomistic semantic units (or “senses”). This is illustrated with the study of the notion of breathe in a sample of 13 languages from 8 genetic phyla. The visual representation of the data shows how the semantic maps can provide heuristic tools also when trying to define semantic universals and typological tendencies across languages. Anna Zalizniak presents the ongoing project, at the Institute of Linguistics (Russian Academy of Sciences), of a database of semantic shifts. It comprises, in a unified, user-friendly format, a catalogue of parallel semantic shifts, both synchronic and diachronic, that took place in two or more words of different languages.



From Polysemy to Semantic Change

The inventory aims at revealing the most frequent, prominent and significant samples of semantic derivations, which occur independently in different languages and eras, and serve as a basis for semantic typology. The paper deals with the formal structure of the lexicographic entry of the catalogue of semantic shifts and discusses examples in several groups of the Indo-European phylum. The theoretical and methodological approach proposed by Bruno Gaume, Karine Duvignaud and Martine Vanhove is based on a new multidisciplinary approach using data processing resources, graph theory, and cognitive semantics. Recent work on the semantic networks of the French lexicon, based on dictionaries, and dictionaries of synonyms, as developed by Bruno Gaume and the research group DiLan (http://dilan. irit.fr/), meet typological and comparative issues. Dictionaries, once coded as graphs, show remarkable structural properties (technically called graphs of Hierarchical Small Worlds), which do not seem to be language dependent. These structural properties challenge compositional models of lexical semantics and support interactive linguistic and psycholinguistic models based on dynamic networks and the dynamic construction of meaning. The article explores a method for the automatic analysis of semantic groupings in order to distinguish, on typological and cognitive levels, which groupings are universal, and which are more limited geographically, genetically or culturally. The robustness of the method is tested on the polysemous verbs of sensory and cognitive perceptions. The third part of this volume is dedicated to empirical researches. The six case studies, covering five semantic domains, enlarge the limited set of semantic fields which have mainly be dealt with so far, i.e., colour terms, body parts and perception words, or which are closely intertwined with the evolution of grammatical morphemes, e.g., epistemic verbs. The original sample contained 45 languages for which we could rely on first hand data collected by the linguists of the research team, or on their mother tongues. For various reasons, it has not been possible to use them all systematically throughout the chapters, resulting in some discrepancies between the languages surveyed. Each case study takes into account, as much as possible, as outlined in Robert”s article, the role of cultural factors, co-textual and contextual environments, as well as universal cognitive findings, in establishing semantic associations. Emilio Bonvini’s article deals once more with the semantic notion of eat in a sample of languages from one phyla, the vast Niger-Congo family. His study is based on the lexical material and data of examples published in various dictionaries, and his own expertise in some of the languages. His aim is to identify the semantic specificity of each language in relation to universal semantic domains. He shows, through a synchronic and semasiological approach, how the lexical items meaning eat, a universal of human experiment, broaden out to a larger range of polysemous meanings (often comparable with the other findings about eat in other phyla). This is achieved in three steps: (i) the study of the notion of eat as lexical units, (ii) the study of eat in linguistic



Semantic associations: A foreword 

co-texts and its implications for the construction of orthonyms, (iii) the study of eat in the discursive context and its relations to the polysemous construction of meaning. Christine Hénault’s article about eat uses a different method. Focusing on a diachronic approach (with a blend of synchrony for the unscripted languages of the sample), she takes as a starting point the reconstructed meaning(s) in the protolanguages. She overcomes the risk of false etymologies by comparing the data to synchronic data within one and the same phylum and also by comparing them with a small sample of genetically and areally unrelated languages. It is noteworthy that the three articles dedicated in this volume to the concept of eat, whatever the method used, all point to a number of recurring semantic associations, and an ambivalence as regards to the control or non-controlled feature of the semantic extensions of eat, leading to cases of enantiosemy which seem to be polygenetic, but not a cognitive constant. Two sets of semantic relations are examined by Pascal Boyeldieu: those connected with the concepts of meat/animal and that of drink. His sample shows that the terms for “meat” and “animal” are identical (or similar) in a fair number of African languages, as well as in some languages of south-east Asia, a fact which strongly suggests a cultural motivation: the importance of game in (former) hunter societies, and certainly not a lexical universal. As for “drink”, it is commonly related to a number of main semantic fields: “smoke”, “drink alcohol immoderately”, “absorb” (non-living matter), “catch, get” (in a passive way), “be subjected to, suffer”. The latter three semantic fields point to the “passive, non-controlled” feature, which distinguishes drink from eat in the language sample under study. But the “control” feature of eat is far from being universal, as shown in the articles of Family, Bonvini and Hénault in this volume (as well as in Pardeshi 2006), and both “control” and non-controlled” features might even co-occur within one language. The joint study by Sergueï Sakhno and Nicole Tersis about the semantic fields of social relations, exemplified by friend and enemy, focuses mainly on Indo-European languages from a diachronic point of view, and on Inuktitut from a synchronic point of view. It shows that they are quite commonly cognates of words meaning “other”, “second”, “follow”, “nearness”, “opposite”, “matching” or “one’s own”. Some of them originally denoted a more specific kind of companionship, as in travel, business, lodging, bringing up. Their study also demonstrates that other semantic links, which seem at first sight quite remote such as “deceive” or “ghost”, are in fact connected with the semantic field of “friend” through the concept of “duality”. Their claim is that the notion of “duality” is to be integrated in a broader concept, that of “complementarity” which is particularly important in some cultures (“friend” as a part of a “whole”). They also show how the semantic networks of friend and enemy are deeply intertwined, and lead to several enantiosemic patterns that recur crosslinguistically. These semantic fields (with a few others discussed in this volume) also proved particularly interesting by putting to the fore the limits of a rigid approach of the

 From Polysemy to Semantic Change

lexicon: it is sometime necessary to compare lexical items in one language with a grammatical morpheme in another when one wants to understand and fully account for the semantic and cognitive networks in natural languages. The search for semantic universals also meets research on grammaticalization processes. After the works of Sweetser (1990), and Evans & Wilkins (2000), the issue of the semantic associations between the lexicon of sensory and cognitive perceptions is again investigated in Martine Vanhove’s article in a sample of 25 languages. It deals more specifically with the notions of “hear”/“listen”, “see”, “feel”, and their extension into the domains of internal perception and cognition: “heed”, “obey”, “learn”, “understand”, “know”, etc. They are also compared with some semantic extensions of “take”. It confirms for new phyla and areas (various African, Oceanic, Sino-Tibetan and Eskimo languages) the regularity of the semantic shifts already noted in other works, and the highly polysemous terminology and the consequences thereof for cognitive linguistics. Furthermore, the analysis of the data suggests a possible semantic and cognitive universal which groups, synchronically or diachronically at the lexical level, and independently of cultural factors, mental perceptions at large with the hearing sense, but far less systematically with sight, as an Indo-European biased cognitive approach could suggest. Michel Masson’s article provides a very deep insight into culturally based semantic parallels. He examines the recurrent associations between cats or monkeys and small insects in Indo-European languages, and concludes that they appear to stem from the cross-culturally recurrent conception of these living beings as connected to supernatural powers, giving rise to semantic shifts such as “scarecrow”, “frighten”, “devil” or “have the gift of witchcraft”. Our greatest hope is that this volume will fulfil its main aim, that is stimulating further crosslinguistic studies on the lexicon and semantics, a still under-developed domain of typological research.

Acknowledgements We are most grateful to Julienne Doko, who translated part of the chapters published in this book: that of Bonvini, Boyeldieu, Hénault, Pottier, Sakhno & Tersis, and to Margaret Dunham who was responsible for the translation of Gaume, Duvignaud & Vanhove, and of Robert.

References Evans, N. & Wilkins, D. 2000. In the mind’s ear: The semantic extensions of perception verbs in Australian languages. Language 76(3): 546–592.



Semantic associations: A foreword 

Koptjevskaja-Tamm, M., Vanhove, M. & Koch, P. 2007. Typological approaches to lexical semantics. Linguistic Typology 11(1): 159–185. Pardeshi, P. et al. 2006. Toward a geotypology of EAT-expressions in languages of Asia: Visualizing areal patterns through WALS. Gengo Kenky 130: 89–108. Sweetser, E. 1990. From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: CUP.

part i

State of the art

Approaching lexical typology Maria Koptjevskaja-Tamm University of Stockholm

The paper aims at situating the research direction presented in the volume within the larger domain of typological research in general. It gives a short summary of what is meant by typological research, discusses the relation between semantic and lexical typology and the general premises for lexical-typological research. The bulk of the paper is devoted to the three main lexical-typological research foci – what meanings can(not) be expressed by a single word, what different meanings can be expressed by one and the same lexeme or by words derivationally related to each other, and what cross-linguistic patterns there are in lexicon-grammar interaction. The paper ends up with the general discussion of the urgent methodological problems facing lexical typology as a field. Keywords:  lexical typology; lexicon-grammar interaction; linguistic categorization; motivation; semantic typology

1.  Introduction The aim of the present article is to situate the research direction presented in the volume within the larger domain of modern typological research in general. As witnessed by the title of the volume, “From polysemy to semantic change: Towards a typology of lexical semantic associations”, three key words are crucial here – typology, lexical and semantic. The paper starts by a short summary of what is meant by typological research in general, then discusses the relation between semantic and lexical typology and points out several different groups of questions asked within lexical-typological research, its different foci. Section 3 touches on the general premises for lexical-typological research – possible words, semantic generality vs. polysemy, and the meaning of “meaning”. Sections 4–6 are devoted to the three main lexical-typological research foci introduced in Section 2 – what meanings can(not) be expressed by a single word, what different meanings can be expressed by one and the same lexeme or by words derivationally related to each other, and what cross-linguistic patterns there are in lexicon-grammar interaction. Each of these chapters considers numerous examples of research and the various methodological and theoretical issues relevant for it. The whole paper ends up with the general discussion of the urgent methodological problems facing lexical typology as a field.



Maria Koptjevskaja-Tamm

2.  Typology, semantic and lexical typology The term “typology”, as is well known, has many different uses. What primarily matters for the present volume is typology understood as “the study of linguistic patterns that are found cross-linguistically, in particular, patterns that can be discovered solely by cross-linguistic comparison” (Croft 1990: 1). Typology can also refer to typological classification of languages into (structural) types on the basis of particular patterns for particular phenomena. Typological research is driven by the persuasion that the variation across attested (and, further, possible) human languages is severely restricted, and aims therefore at unveiling systematicity behind the whole huge complex of linguistic diversity. In pursuing their tasks, typologists raise – and often try to answer – important theoretical questions, such as the ones listed below. – – – – – – –

According to what parameters does a specific phenomenon vary across languages, in what patterns do these parameters (co-)occur? What generalisations can be made about attested vs. possible patterns? What is universal vs. language particular in a given phenomenon, what phenomena are frequent vs. rare? How are various linguistic phenomena distributed across the languages of the world? Which phenomena are genetically stable and which are subject to contact-induced change? How can the attested distribution of the different patterns across languages be explained? How can the attested cross-linguistic patterns/generalizations be explained?

The papers in the present volume do in fact focus on linguistic patterns that can be discovered only by cross-linguistic comparison – cross-linguistically recurrent patterns of polysemy, heterosemy and semantic change – and are therefore examples of typological research. The domain of research shared by the papers in the volume is, however, somewhat outside of the main interests of modern typological research, that has so far primarily focused on grammatical and, to a lesser degree, phonetic/phonological phenomena under the labels of “grammatical typology”, “syntactic typology”, “morphological typology”, “morphosyntactic typology” (or, quite often, just “typology”), “phonetic typology” and “phonological typology”. None of those would suit the direction of the volume. We are dealing here, first, with lexical and, second, with semantic phenomena – which are the primary objects of lexical vs. semantic typology. The terms “semantic typology” and “lexical typology” are often used as if there were self-explanatory, but are only rarely explicitly defined. According to Evans (forthc.), semantic typology is “the systematic cross-linguistic study of how languages express meaning by way of signs”. What can be meant by lexical typology is, however,



Approaching lexical typology

less clear, apart from the evident fact that it involves cross-linguistic research on the lexicon. Many linguists will probably agree with Lehrer’s (1992: 249) widely quoted definition that lexical typology is concerned with the “characteristic ways in which language […] packages semantic material into words” (cf. the excellent overviews in Koch 2001 and Brown 2001). Viewed as such, lexical typology can be considered a sub-branch of semantic typology concerned with the lexicon, as this is done in Evans (forthc.). Other definitions of lexical typology focus on “typologically relevant features in the grammatical structure of the lexicon” (Lehmann 1990: 163) or on typologically relevant vs. language-specific patterns of lexicon-grammar interaction (Behrens & Sasse 1997). I think that a reasonable way of defining what can be meant by “lexical typology” is to view it as the cross-linguistic and typological dimension of lexicology. The probably most updated overview of lexicology as a field is found in the two volumes (Cruse et al. eds. 2002, 2005), the title of which (“Lexicology. An international handbook on the nature and structure of words and vocabularies”) underlines the special orientation towards the two core areas which makes of lexicology an autonomous discipline, namely, the characterization of words and vocabularies, both as unitary wholes and as units displaying internal structure with respect both to form and content (Cruse et al. eds. 2002, 2005: viii–ix).

In the same vein as lexicology in general is not restricted to lexical semantics, lexical typology can include phenomena that are not of primary interest for semantic typology. Likewise, since lexicology is not completely opposed to either phonetics/phonology, morphology or syntax, cross-linguistic research on a number of word- and lexiconrelated phenomena is – or can be – carried out either from different angles and with different foci, or within approaches that integrate several perspectives, goals and methods. There are different kinds and groups of questions that can be addressed in typological research on words and vocabularies, or lexical typology, and that can therefore be considered as the different foci of lexical typology. Some of them (the most important ones, as I see them now) are listed below, but there are undoubtedly many others. The questions are often interrelated with each other, a point that will be stressed in the ensuing presentation. – What is a possible word, or what can be meant by a word? Possible vs. impossible words in different languages, different criteria for identifying words and interaction among them, universal vs. language-specific restrictions on possible, impossible, better and worse words. – What meanings can and cannot be expressed by a single word in different languages? Lexicalizations and lexicalization patterns, “universal” vs. language-specific lexicalizations, categorization within, or carving up of lexical fields/semantic





Maria Koptjevskaja-Tamm

domains by lexical items, the architecture of the lexical fields/semantic domains (e.g., basic words vs. derived words). – What different meanings can be expressed by one and the same lexeme, by lexemes within one and the same synchronic word family (words linked by derivational relations) or by lexemes historically derived from each other? Cross-linguistically recurrent patterns in the relations among the words and lexical items in the lexicon – a huge and heterogeneous category with many different subdivisions, a large part of which can be subsumed under the various aspects of motivation (for details see Koch 2001: 1156–1168 and Koch & Marzo 2007), e.g., semantic motivation (polysemy, semantic associations/semantic shifts) and morphological motivation (derivational patterns, including compounding). – What cross-linguistic patterns are there in lexicon-grammar interaction? The lexicon of a language is, of course, a dynamic and constantly changing complex structure where new words emerge, old words disappear or change in one or another way. Lexical-typological research has, thus, both synchronic and diachronic dimensions. Historically oriented lexical typology studies semantic change, grammaticalization and lexicalization processes (the latter understood as “a process by which new linguistic entities, be it simple or complex words or just new senses, become conventionalized on the level of the lexicon”, Blank 2001: 1603) as examples of diachronic processes showing cross-linguistically recurrent patterns. The lexicons of most languages show different layers of origin with many words coming from “outside” – as direct loans, loan translations, etc. A particularly interesting aspect of historical lexical typology is the search for cross-linguistically recurrent patterns in contact-induced lexicalization and lexical change, e.g., differences in borrowability among the different parts of the lexicon and the corresponding processes in the integration of new words, or patterns of lexical acculturation (i.e., how lexica adjust to new objects and concepts). Lexical-typological research can also be more local, e.g., restricted to a particular lexical field, a particular derivational process, a particular polysemy pattern, or more general, with the aim of uncovering patterns in the structuring of the lexicon that are supposed to have a bearing on many essential properties of the language. The latter includes various approaches to the issues of “basic” vs. non-basic vocabulary, or suggestions as to how characterize, compare and measure the lexical-typological profiles of different languages. In fact, some people prefer using the term “typological” (e.g., typological properties) for referring to what is considered as the more essential, central, or general properties of a language. In this understanding, a large portion of crosslinguistic research on words and vocabularies will not count as typological (Lehmann 1990) (this applies, among others, to what is called “local” lexical-typological research immediately above) – we will briefly discuss this position in Section 6.



Approaching lexical typology

In the ensuing sections I will give examples of lexical-typological research along these different lines, try to make generalizations on the state of art and, finally, point out some recurrent problems and directions for the future. Because of space restrictions, issues such as possible words or degree of borrowability will only be allotted a brief mention here.

3.  General premises: Words and meanings 3.1  Possible words As any introductory textbook in linguistics will tell us, “word” can denote different things. The huge issue of what can be meant by word, or what is a possible word across languages will only be touched upon in this paper – and mainly from the semantic point of view. Traditional morphological typology, with its focus on how much and what kind of morphology is allowed in words across languages (cf. isolating vs. polysynthetic languages, etc.), represents one way of comparing possible words cross-linguistically. The issue is, however, much more complex and requires a truly integrating approach, where morphological (and further grammatical), phonetic, phonological and semantic criteria are all relevant, as well as psycholinguistic considerations (holistic storage and processing), and sociolinguistic and pragmatic factors (e.g., degree of conventionalization). The important contributions include Aikhenvald & Dixon (2002) and the ongoing project “Word domains” within the research programme “Autotyp” (directed by Balthasar Bickel and Johanna Nichols, http://www.uni-leipzig.de/~autotyp/projects/ wd_dom/wd_dom.html) which both focus on words as phonological and grammatical domains;1 a useful discussion of compounds as words across languages is found in Wälchli (2005: 90–134). A new and promising way of investigating the question of “possible words” in a human language is laid out in Greville Corbett’s (2007, forthc.) “canonical approach” to inflection, that evaluates the various formal ways in which the word forms of one and the same lexeme can be related to each other. Since definitions of linguistic phenomena normally evoke several different criteria, the basic idea behind Corbett’s “canonical approach” in typology is to take definitions to their logical endpoint, where the different criteria converge and together produce the best, clearest, indisputable “canonical” instance of the phenomenon. Real “canonical” instances are rare or even unattested, but they serve as a point from which the actual phenomena can be evaluated in the theoretical space of possibilities. In the particular case of inflection we can,

.  Last visited on April 24th 2007.





Maria Koptjevskaja-Tamm

thus, evaluate agglutination vs. flection, various kinds of syncretism or suppletion, and, in general, various kinds of exceptions in inflectional morphology. Now, since much of the discussion in the present paper will deal with the link between words and their meanings, of primary concern here will be words as carriers of lexical meanings. Several interrelated distinctions are important for understanding the kinds of questions usually asked here and the controversies surrounding possible answers to them. These include the distinction between semantic generality and polysemy (Section 3.2.) and the various understandings of “meaning” – denotation vs. sense, and approximate vs. precise (Section 3.3.).

3.2  Semantic generality vs. polysemy, or when are meanings lexicalized? A classical issue in lexical semantics concerns the distinction between semantic generality and polysemy. Consider Table 1 for the English and Russian verbs designating motion in water (aquamotion). Table 1.  Aquamotion verbs in English and Russian

English Russian

Passive motion (‘float’)

Self-propelled motion of animate figure (‘swim’)

Motion of vessels and people aboard (‘sail’)

float

swim

sail plyt’/plavat’

English distinguishes among three different verbs – float, swim and sail. Float designates passive motion; swim – active, self-propelled motion of animate figure, and sail – active motion of vessels and people aboard. To simplify matters, we can say that we have three different meanings, or concepts here – ‘float’, ‘swim’ and ‘sail’ – each of which exists as the meaning of a particular lexeme in English, or is lexicalized. Russian has two aquamotion verbs plavat’/plyt, where the main difference is directionality of the designated motion – plavat’ is multidirectional, while plyt’ is unidirectional. Since each of these two verbs corresponds roughly to the same situations as the three English verbs together, we will for the sake of simplicity concentrate on just one verb, plyt’. The question is now whether the three meanings ‘float’, ‘swim’ and ‘sail’ are lexicalized in Russian. There are at least three theoretical and methodological possibilities here, e.g., semantic generality, polysemy and agnosticism. First, semantic generality: it could very well happen that plyt’ is semantically general and does not distinguish among “float”, “swim” and “sail” at all. In that case we could say that Russian does not lexicalize the differences among “float”, “swim” and “sail” in having just one and the same word (or one word couple) covering all the three meanings. The second possibility, polysemy, would mean that plyt’ does in fact distinguish at least among the three different meanings “float”, “swim” and “sail”. In that case we



Approaching lexical typology

could still say that each of these meanings is lexicalized in Russian – however, not as the meaning of its “own” particular lexeme, but rather as the meaning of a particular lexical unit. A lexical unit is, in turn, defined as the pairing of a single specifiable meaning/sense with a lexical form (Cruse 1986: 77–78), so that a polysemous word is a lexeme consisting of several lexical units. Fig. 1 visualizes the difference between potential semantic generality vs. polysemy in the case of plyt’. Each of the double-headed arrows represents a lexical unit (LU); the lexical form {plyt’} covers all the inflectional forms that together constitute the formal side of this lexeme.

a. Polysemy ‘float’ LU1

b. Semantic generality ‘swim’

LU2 plyt’

‘sail’

meaning

LU3

‘float / swim / sail’ LU

form

plyt’

Figure 1.  Representing polysemy (a) vs. semantic generality (b).

The third possibility is to leave aside the problem of semantic generality vs. polysemy and to remain agnostic about the correct semantic analysis of a particular word. This is the “default” interpretation of the data in Table 1. Under this view, what matters is the fact that Russian has only one lexeme (or, rather, a couple of directionalityrelated lexemes) corresponding to the three different English ones. There are various tests for distinguishing between semantic generality and polysemy, e.g., the distinct meanings within a lexeme having different syntactic properties. On the basis of this particular test it can be argued that each of the two Russian verbs plavat’/plyt’ distinguishes among several meanings, very much along the lines of the English system (cf. the analysis in Rakhilina 2006 and in Koptjevskaja-Tamm et al. forthc.), and that Fig. 1a gives a more faithful representation than Fig. 1b. Thus, when describing inanimate objects moving with water (≈ ‘float’), plyt’ has to combine with an overt indication of the path normally expressed by a combination of po “on” and the reference to the surface, as in (1a); no indication of the source or goal of the motion is allowed. Whenever plyt’ designates self-propelled motion of animate entities (≈ ‘swim’), it has to combine with an overt indication of either the source or goal of the motion, to the exclusion of the “surface” where it takes place, i.e., path (ex. 1b). Finally, when plyt’ refers to the motion of vessels (≈ ‘sail’), it can take either indications of the source/goal of the motion, or of the surface (path); in addition, both can be combined in one and the same sentence, as in ex. (1c).



 Maria Koptjevskaja-Tamm

(1) a.

Po rek-e ply-l-i želt-ye on river-dat.sg “plyt”-pret.pl yellow-nom.pl



list’-ja (*k bereg-u). leaf-nom.pl (*towards shore-dat.sg)



‘Yellow leaves were floating in the river (*to the shore).’

b. On ply-l k bereg-u (*po rek-e) he.nom “plyt”’-pret.m.sg towards shore-dat.sg (*on river-dat.sg) ‘He was swimming towards the shore (*in the river).’ c.

Korabl’ ply-l po Volg-e do Kazan-i ship.nom.sg “plyt’”-pret.m.sg on Volga-dat.sg to Kazan-gen.sg ‘The ship was sailing on the Volga river to the town of Kazan.’

Distinguishing between semantic generality and polysemy is, on the whole, a tricky business. Opinions on what polysemy amounts to and how to search for it differ considerably among different semantic theories and practices (such as dictionary entries), not to mention language users (see Riemer 2005 for a recent overview of the problem). In general, decisions on what should count as several meanings of one and the same lexeme vs. one more general meaning require sophisticated analyses and tests, difficult enough within one language, hard to carry out in several and impossible in many. The current practices of cross-linguistic lexical studies seem to take a fairly pragmatic stance on the choice between the three options represented by Table 1 and Fig. 1a-b, choosing the one that suits best the questions asked in a particular study (and also asking reasonable questions given the available data). Suppose that we are interested in the question of whether ‘swim’, ‘float’ and ‘sail’ belong to the universally lexicalized meaning. For this kind of question it hardly matters how exactly these meanings are expressed – as separate lexemes, as the different meanings of one and the same lexeme, or as words derivationally related to each other. So even if Russian does not have different words for ‘swim’, ‘float’ and ‘sail’, it still lexicalizes these meanings. On the other hand, the very fact that one and the same word plyt’ corresponds to the three different words in English is very interesting too: according to the common-sense interpretation, driven by considerations of iconicity, this is an indication of their semantic similarity. Further interpretations of this first impression may differ – whether we are talking about the cross-linguistic variation in carving up, or categorizing the corresponding conceptual domain, on the one hand, or about lexical semantic associations, on the other – but the fact of the semantic affinity remains and has to be accounted for. Consistent application of semantic tests may, in fact, lead to a huge proliferation and splitting of the meanings within a lexeme (cf. with some of the practices within the cognitive semantics). This richness is often of marginal interest for crosslinguistic and typological comparison, which is, by nature, reductionistic and tries to maintain a reasonable balance between language-specific details, on the one hand, and cross-linguistically applicable generalizations, abstractions and simplifications, on the



Approaching lexical typology

other hand. Here again, what is represented as one meaning vs. several meanings in cross-linguistic comparison tends to be governed by pragmatic considerations. To take an example, one of the meanings of the verb to sail (in combination with a human figure and a direct object) is “to control the movement of a boat or ship”, rather than just “to go in a boat or ship”. Even if this distinction is interesting per se, it can either be brought to the fore or neglected depending on what a particular study focuses on. The method of semantic maps that has been successfully used in cross-linguistic comparison of grammatical semantics is explicitly agnostic about the distinction between polysemy and semantic generality (cf. Haspelmath 2003: 231), the same position is advocated by François (this volume) for cross-linguistic studies of lexical associations and is (implicitly) taken in cross-linguistic studies of categorization of conceptual domains based on multidimensional scaling (Levinson & Mejra 2003; Majid et al. 2007; Wälchli 2006/2007; see also Section 3.3. in the present paper).

3.3  Th  e meaning of “meaning”: Denotation vs. sense, approximate vs. precise meaning definitions The meaning of “meaning” is a key issue in semantics, where opinions vary. For our purposes the main and generally recognized partition goes between denotation/extension vs. intension/(descriptive) meaning sense. The relations between the two are complicated, and the emphasis on the one or the other can have different implications for a cross-linguistic comparison. Consider the domain of temperature. The Swedish adjective ljummen (“lukewarm”) covers a narrow and well-defined range of temperatures, “neutral” temperatures, those corresponding to the temperature of the human skin and feeling neither warm nor cold. The Russian adjective teplyj, whose standard translation into English is warm, also denotes a relatively restricted range of temperatures, the greater portion of which is covered by ljum(men). From the denotational point of view, the two adjectives are, thus, fairly similar to each other, with the denotational range of teplyj slightly exceeding that of ljummen, as is schematically visualized in Fig. 2. Denotation / extension

teplyj

+37°C +35°C +33°C

(Descriptive) meaning / sense

ljummen

defined via direct reference to the human body

defined via neutrality of perception

Figure 2.  Extension/denotation vs. (descriptive) meaning/sense for the temperature adjectives ljummen (Swedish) and teplyj (Russian).



 Maria Koptjevskaja-Tamm

The descriptive meanings of the two adjectives, however, appear very different, as argued in (Koptjevskaja-Tamm & Rakhilina 2006: 259) and as also visualized in Fig. 2. First, teplyj does have a clear “warming” orientation: an entity qualified by its comparative form (bolee teplyj, teplee) has a HIGHER temperature than the one it is compared to. Ljummen often lacks this clear orientation. Thus, depending on pragmatic factors, an entity qualified by its comparative form (mera ljummen, ljummare) can have either a HIGHER (as in Hans öl är ljummare än min “His beer is more lukewarm than mine”) or a LOWER (as in Mitt te har svalnat, hans är ännu ljummare “My tea has cooled down, his is even more lukewarm”) temperature than the one it is compared to. In addition, teplyj is often used with body-part terms and has very positive connotations in metaphorical uses – e.g., teplye slova, čuvstva, otnošenija “warm (positive, friendly) words, feelings, relations”. On the basis of these observations we have chosen to define the meaning of teplyj via direct reference to the (human) body. Teplyj designates temperatures that correspond to or are not significantly higher than the temperature of the human body/skin or that maintain the temperature of the human body without too much effort on the part of the human being, and therefore cause an agreeable sensation of comfort and coziness. Ljummen, on the other hand, never qualifies body-part terms and lacks any positive connotations in metaphorical uses – e.g., ljumma känslor, reaktioner “weak, neutral feelings, reactions”. Thus, while teplyj covers temperatures that are “normal” with regard to the human body, that feel warm, but not exceedingly so, the meaning of ljummen lacks direct reference to the body, but evokes “neutrality” of perception. Although for many serious semanticists, lexicographers and lexicologists semantic analysis stands for coming to grips with descriptive meanings, or senses, the enterprise is far from obvious even in the researcher’s native tongue and gets easily insurmountable in other languages. As a consequence, much of cross-linguistic comparison is based on meanings defined as denotations, with various methods for eliciting, defining and evaluating expression-denotation couples (pictures, videoclips, Munsell colour chips, etc.). In other words, the question of “what meanings can be or cannot be expressed by single words in a language” often amounts to “what are possible/impossible denotational ranges of single words in a language”. There are various reasons for why such approaches are insufficient, including Quine’s (1960: 29) famous “gavagai”-problem (if a person whose language you don’t know says gavagai when a white rabbit appears in front of you both, how can you be sure about what (s)he really means?). Another big problem is that many meanings – or, rather, many conceptual domains – hardly lean themselves to being investigated via denotation-based techniques: for instance, how do you get at the meaning of “think” or “love”? (cf. also Evans & Sasse 2007). Other ways of dealing with the meaning in cross-linguistic lexical studies make use of “translational equivalents” found in dictionaries and word lists, questionnaires



Approaching lexical typology 

and parallel corpora. These are, again, problematic in various ways. In particular, dictionaries are a favourite object of ridicule in theoretical work on semantics and lexicography, for providing vague and circular definitions. We will come back to the issue of approximate vs. precise meaning definitions (and cross-linguistic identity of meanings, cf. Goddard 2001: 2–3), which is, in fact, crucial in cross-linguistic studies. Denotation-based techniques for data collection, questionnaires and parallel corpora effectively neglect the issue of semantic generality/polysemy, discussed in the previous section. They provide often a number of contexts, or “an etic grid” for capturing (logically) possible distinctions within a domain, with the results that the meaning of a word can easily become reduced to the set of its uses (an “etic definition”). The logical step from an etic definition to an emic one (i.e., finding out the commonalities behind the different uses and, ideally, arriving at a reasonable characterization of the descriptive meaning) goes hand in hand with deciding what constitutes one meaning, i.e., distinguishing between semantic generality and polysemy (cf. Evans forthc. for the discussion of etic vs. emic definitions in semantic typology). A central complication for cross-linguistic studies on the lexicon – and, further, in most cross-linguistic research where meaning is involved – is created by the problem of a consistent meta-language for representing meanings within and across languages. This, in turn, is related to the general enormous gap between theoretical semantics and theoretical lexicology, on the one hand, and actual lexicographic practices. The most serious candidate on the market is the Natural Semantic Metalanguage, originally advocated by Anna Wierzbicka. The proponents of the NSM take polysemy very seriously, strive for comparing descriptive meanings rather than denotational ranges, and aim at providing precise meaning definitions by means of reductive paraphrases based on a principled set of “universal semantic primitives” (see, for instance, Goddard & Wierzbicka (eds.) 1994; Goddard 2001; Wierzbicka 1990, 2007). The theory has both positive and negative sides (cf., e.g., the discussion in Krifka (ed.) 2003; Riemer 2002 and Evans forthc.), but on the whole it enjoys less popularity and attention in the typological enterprise than it deserves.

4.  What meanings can and cannot be expressed by a single word? What meanings can be or cannot be expressed by a single word in different languages, or what word meanings are universal, frequent, possible, impossible? Are there any universal (or at least statistically predominant) restrictions on the meanings that can or cannot be expressed by single words across languages, or are languages more or less free to choose here? This section will be devoted to research that focuses on the issues of cross-linguistically recurrent and typologically relevant aspects of lexicalizations and lexicalization patterns. Underlying it is the tension between two opposing

 Maria Koptjevskaja-Tamm

hypotheses on linguistic categorization, among others, on the concepts expressed by words. One hypothesis holds that categorization is universal, at least when it comes to basic, universal and daily situations, so that lexical meanings “originate in nonlinguistic cognition, and are shaped by perceptual and cognitive predispositions, environmental and biological constraints, and activities common to people everywhere” (Majid et al. 2007: 134). The other suggests that lexical meanings “do not reflect shared nonlingustic cognition directly, but are to some extent linguistic conventions that are free to vary – no doubt within limits – according to historical, cultural, and environmental circumstances” (Majid et al. 2007: 134, see also the references there). In this section we will look at studies dealing with categorization within, or carving up of lexical fields/semantic domains by lexical items.

4.1  C  ategorization within lexical fields and conceptual domains: A couple of examples to start with The basic idea underlying cross-linguistic research on categorization within lexical fields and conceptual domains (coherent segments of experience and knowledge about them) is that human experience is not delivered in nicely pre-packed units, categories and types, but has to be chunked, organized and categorized by human beings themselves. Categories correspond to experiences that are perceived to have features in common. When experiences are systematically encoded by one and the same linguistic label (e.g., by the same word) they are, most probably, perceived as being fairly similar to each other; that is they are taken to represent one and the same class, or to correspond to one and same concept or lexical meaning. A simple example of what can be meant by different ways of categorizing, or carving up a conceptual domain across languages is given in Table 2, which shows how the inventories of body-part terms in six languages differ in the extent to which they distinguish between hand vs. arm, foot vs. leg, and finger vs. toe by conventionalized, lexicalized expressions (“labels”). Table 2.  Hand vs. arm, foot vs. leg, finger vs. toe in English, Italian, Rumanian, Estonian, Japanese and Russian English

Italian

Rumanian

Estonian

Japanese

Russian

hand arm foot leg finger toe

mano braccio piede gamba dito

minaˇ brat, picior

käsi käsi(vars) jalg

te ude ashi

ruka

deget

sõrm varvas

yubi

palec

noga



Approaching lexical typology 

The table above follows the same practice of representing “lexicalization” in a fairly unsophisticated way as Table 1 in Section 3.2., without asking the question of whether ruka in Russian or yubi in Japanese are polysemous or semantically general. What matters here is simply how many different lexemes there are and how they partition the domain. A somewhat more complicated example, partly introduced in Section 3.2., is given in Table 3, which shows the verbs used for talking about water-related motion (aquamotion) in three languages – Swedish, Dutch and Russian. The table includes both motion of water itself (“flow” in English) and motion/location of other entities (other figures) with water as ground. Here, again, the Russian verbs plyt’/plavat’ are treated as one semantic unit, rather than two sets of different senses. Flyta in Swedish appears, however, at two different places – this does not per se imply any strong conviction that the case is much different from the Russian verb couple, but demonstrates problems with two-dimensional representations. Table 3.  A part of the aquamotion domain in Swedish, Dutch and Russian (based on Koptjevskaja-Tamm et al. forthc.)

Sw

Active motion of an animate Figure

Sailing boats

Motor-driven vessels

Rowing boats

simma

segla

(no specific aquamotion verbs)

Motion of vessels and people aboard

zeilen

Du zwemmen

Canoes

Motion out of control

Neutral motion/ Location

ro

paddla

driva

flyta

roeien

paddelen

varen

Ru plyt’/ plavat’ pod parusami

Passive motion; location on water

plyt’/plavat’ gresti

Motion of water

flyta, rinna

drijven

stromen

nestis’

teč, lit’sja

As these examples show, languages differ considerably as to how many different lexemes they have for talking about comparable domains and how exactly these words partition the domains. It is therefore reasonable to ask whether there is any systematicity underlying the obvious cross-linguistic variation. Whatever the answer is, it requires explanation. Only a handful of conceptual domains typically encoded by words (rather than by grammatical means) have been subject to systematic cross-linguistic research on their semantic categorization, primarily colour, body, kinship, perception, motion,

 Maria Koptjevskaja-Tamm

events of breaking and cutting, dimension (and posture not considered here).2 The list can be made slightly longer, if we include words and expressions with more grammatical meanings, such as indefinite pronouns (Haspelmath 1997), various quantifiers (cf. Bach et al. 1995; Auwera 2001; Gil 2001), interrogatives (Cysouw 2004), phrasal adverbials (Auwera 1998) and spatial adpositions (Levinson & Meira 2003) – these won’t be considered below.

4.2  Domain-categorization studies: Language coverage and focus The standard textbook example of underlying systematicity behind the striking crosslinguistic diversity, colour, remains, probably, the most widely researched on domain in lexical typology, in terms of the languages covered in systematic comparison by means of comparable and elaborated methodology, and the intensity, diversity and depth of theoretical discussions. Kay & Maffi’s (2005) chapter on colour in the World Atlas of Language Structures is based on 119 languages, the data coming primarily from Berlin, Kay & Merrifield’s World Color Survey in 1976–78, but there are many more languages the colour inventories of which have been subject to systematic research and which have figured in linguistic discussions (the relevant literature is too extensive for being listed here, cf. MacLaury 1997, 2001 and Payne 2006 and the references there, also http://www.icsi.berkeley.edu/wcs/for the World Color Survey Site). Kinship terminologies have for a long time been a favourite semantic field among anthropologists and anthropologically oriented linguists. Detailed and systematic descriptions of the domain are available for many hundreds of languages, and there is a long tradition of classifying the resulting systems into a small number of types. As a rule, such classifications do not consider the whole kinship systems, but concentrate only on their subparts. Probably the cross-linguistically most systematic study, Nerlove & Romney (1967), focuses on sibling terminologies in 245 languages. For the body domain, the most comprehensive – in terms of the number and representativity of the included languages – studies are the two chapters by Brown (2005a-b) in the World Atlas of Language Structures. One of them classifies 617 languages according to whether they use the same or different words for hand and arm, the other one asks the same question for finger and hand in 593 languages. Crosslinguistic studies on categorization of the whole body cover a handful of languages – the milestones here are Brown (1976), Andersen (1978) and, in particularly, Majid et al. (eds.) (2006), with detailed and systematic studies of ten languages. The modern lexical-typological research on motion verbs is for many people firmly associated with the tradition stemming from Talmy’s seminal chapter (1985), .  For the latest developments on the latter concept see Ameka & Levinson 2007, cf. also Newman 2002.



Approaching lexical typology 

which focuses on the components of a motion event that are systematically encoded within motion verbs in different languages (lexicalization patterns). It is not quite clear how many languages have been systematically studied from this point of view – my impression is that they are not so many. Wälchli’s on-going research on motion events in general covers more than 100 languages (Wälchli 2006; Wälchli & Zúñiga 2006). Since the motion domain is, in fact, very complex and heterogeneous and is normally encoded by many different linguistic means, it is reasonable to split it into smaller sub-domains for meaningful and doable cross-linguistic studies. Ricca (1993) focuses on the distinction between the deictic motion verb (“come”) and the non-deictic verb (“go”) in twenty European languages, whereas the papers in Maisak & Rakhilina (eds.) (2007) present detailed and systematic studies of aquamotion verbs across 40 genetically, structurally and areally diverse languages. The most systematic cross-linguistic study of perception verbs has been carried out by Viberg (1984, 2001) on the basis of fifty languages (and further confirmed by the Australian languages in Evans & Wilkins 2000), whereas the domain of cutting and breaking events (Majid & Bowerman eds. 2007) has been investigated in 28 genetically and structurally diverse languages (covering 13 languages families, four isolates and a creole language). Finally, Lang (2001) presents a cross-linguistic comparison of dimension terms (like “wide”, “long”, etc.) in a sample of forty languages coming from Europe and Asia.

4.3  Methodology Research on domain-categorization – as lexical typology in general – has on the whole made relatively little use of secondary sources, but often relies on primary data. Some of the studies on kinship terminology (e.g., Nerlove & Romney’s 1967; or Greenberg 1980) largely utilize the available earlier descriptions of particular languages, but this, in turn, depends on the long descriptive tradition of the domain starting with Morgan’s (1870) early typology of kinship systems. Viberg’s study (1984, 2001) of perception verbs involves a combination of secondary data sources (dictionaries, word lists and more detailed descriptions) and systematic translation of sentences in a questionnaire; Ricca’s (1993) study of deictic verbs is mainly based on a questionnaire. The majority of cross-linguistic colour studies have all involved elaborated and systematic techniques for data collection largely inspired by data collection in psychological and psycholinguistic research – Munsell colour chips, salience tests, number of connotations per colour term, frequency in texts etc. (cf. MacLaury 2001 for an overview and references). The major part of the data underlying Andersen’s (1978) and Brown’s (2005a-b) studies of body seem to come from available dictionaries and word lists. A new step in the study of body is undertaken in Majid et al. (eds.) (2006), where the papers on

 Maria Koptjevskaja-Tamm

ten different languages are all based on a consistent application of multiple methods for data collection, such as collecting replies to a detailed questionnaire and drawing outlines of the various body-part terms on a picture of a human body. Elicitation of verbal descriptions for visual stimuli underlies a large portion of cross-linguistic work on motion (the pictures in the Frog Story, initially coming from the cross-linguistic work on child-language acquisition by Berman, Slobin and their colleagues, Berman & Slobin 1994, or video-sequences depicting moving objects), on dimensional adjectives (Lang 2001) and, most recently, on the cuttingbreaking domain (videoclips showing various types of material separation, Majid & Bowerman (eds.) (2007). The papers on aquamotion in Maisak & Rakhilina (eds.) (2007) contain detailed descriptions of particular languages by language experts, who have conducted their own in-depth studies (e.g., using corpora and dictionary searches, work with informants, introspection, etc.), while at the same time following the common checklist, or guidelines. Finally, comparison of parallel translations of one and the same text into different languages is becoming a popular tool in typology. Viberg has been carrying out “smallscale” lexical-typological (or, rather, contrastive) studies of several verbs using parallel corpora in a few European languages (e.g., Viberg 2002, 2005, 2006). Parallel texts in more than 100 languages (the Gospel according to Mark) underlie Wälchli’s “large-scale” lexical-typological studies of motion verbs (Wälchli 2006; Wälchli & Zúñiga 2006).

4.4  Questions and generalizations Domain-categorization lexical typology, as typology otherwise, is driven by the curiosity to understand what is variable and universal in a particular linguistic phenomenon. Of central concern here are therefore questions such as according to what parameters a specific phenomenon can vary across languages, in what patterns these parameters (co-)occur and what generalisations can be made about attested vs. possible patterns. One possible result of cross-linguistic studies is, of course, a classification of the obtained data into patterns, or types (and, in some cases, the corresponding classification of the languages). While classifications of phenomena and languages are useful on their own, their value increases if they can be related to other phenomena – both linguistic and non-linguistic ones. In Ricca’s (1993) study of deictic verbs, the twenty languages are classified into three groups depending on the extent to which they make a systematic distinction between verbs showing centripetal (to the deictic centre, normally the speaker) vs. centrifugal (from the deictic centre) motion. Interestingly, the distribution of the types across the sample languages is dependent on a combination of genetic and areal factors and can therefore contribute to our general



Approaching lexical typology 

understanding of genetically stable vs. borrowable (or diffusible) phenomena with further implications for research in historical linguistics. Thus, the fully deictic languages are mainly found in Southwestern and Southern Europe (Portuguese, Spanish, Italian, Albanian, Modern Greek, with the two Finno-Ugric outliers Hungarian and Finnish), the non-deictic languages are Western and Eastern Slavic and Baltic, while the predominantly deictic ones are Germanic, French and the two Southern Slavic languages Serbo-Croatian and Slovenian. The next logical step to be taken from there would be to provide both a better coverage of the European language varieties and a broader nonEuropean sample. These will together contribute to a better understanding of the relative contribution of universal, genetic and areal factors in this domain (cf. Wilkins & Hill 1995 for similar considerations). Brown’s (2005a-b) body-part chapters in the World Atlas of Language Structures are further examples of classifications with interesting implications. First, they show different statistical asymmetries in the distribution of the types: while only 12% of the sample’s languages (72 languages of the 593 languages) use the same word for finger and hand, the corresponding share for the non-distinction of hand and arm is much higher – approximately 37% languages in the sample (228 languages of the 617 languages). Second, the chapters show that the skewings in the areal distribution of the languages in each category are not random, but correlate either with geography (languages without hand-arm distinction tend to occur more frequently near the equator) or with culture (languages without finger-hand distinction tend to be spoken by traditional hunter-gatherers or by groups having a mixed economy of cultivation and foraging). Brown suggests that the former can be dependent on the local clothing traditions (extensive clothing, sometimes also including gloves or mittens, greatly increases the distinctiveness of arm parts), whereas the latter can perhaps be explained by the greater use of finger rings among agricultural people which, in turn, makes fingers salient as distinct hand parts. In both cases, Brown’s explanation makes therefore appeal to the cultural practices. It should be remembered that the correlations are far from perfect: e.g., although Russian is being spoken in a much colder climate than Italian (and Russians are dependent on gloves and mittens to a much higher extent than Italians), the former uses one and the same word for hand and arm, while the latter has two. Some classifications might seem to play a more central, essential role in the linguistic system, they can thus be used as predictors for other linguistic phenomena and might have important implications outside of the language. Consider the impact of Talmy’s (1985, 1991) studies on the research on motion verbs. Talmy focuses on the different “conflations” of the motion meaning component with other meaning components, or different lexicalization patterns, where lexicalization is defined to be “involved where a particular meaning component is found to be in regular association with a particular morpheme” (Talmy 1985: 59). It is generally assumed that languages

 Maria Koptjevskaja-Tamm

tend to be consistent with respect to their lexicalization patterns in motion verbs (cf. the classification into path-conflating vs. manner-conflating vs. figure-conflating languages in Talmy 1985, its later modifications into satellite-framed vs. verb-framed languages in Talmy 1991 and even more radical reorganizations in Slobin & Hoiting 1994 and in Croft 2003: 222). The assumption is, thus, that lexicalization in this domain is not an issue of one particular meaning (or one particular combination of meaning components) being expressed via a word, but is more global – i.e., pertaining to the whole motion domain or, at least, to a major part of it. There is also ample research on possible connections between the lexicalization type of a certain language and, say, its discourse organization, child-language acquisition of the domain, even non-verbal communication such as gestures and “thinking for speaking”, i.e., mild cognitive effects of linguistic relativity (e.g., Slobin 2003; Slobin & Bowerman 2007; Kita & Özyürek 2003). In short, a language’s lexicalization pattern in the motion domain is often taken to belong to its typologically relevant properties. However, the role of the “Lexicalization Pattern” theory should not be exaggerated, since it is based on a limited number of languages, where, in addition, only a subset of motion verbs has been studied in detail. Wälchli’s (2006, 2006/7) on-going research on motion events based on parallel texts in more than 100 languages has already challenged some of the basic assumptions and predictions of the “Lexicalization Pattern” Theory. For instance, languages with a consistent (or predominantly consistent) “global” lexicalization-pattern type seem to be fairly unusual – the best examples are in fact found among the Eurasian and North American languages (exactly those that have figured prominently in Talmy’s seminal papers). Many more languages, however, turn out to show mixed behaviour. And in fact, even languages that are considered to fit into the Talmy-Slobin-Croft typology, often have verbs showing a deviant behaviour. For instance, the Path-conflating verbs ‘fall’ or ‘sink’ are present in many otherwise Manner-conflating languages like English and Swedish, while the otherwise Path-conflating language French has a general motion verb aller which does not conflate with any direction (Viberg 2006: 113). There are, therefore, asymmetries even among the verbs belonging to the same domain, which may be ordered rather than being random. As Viberg hypothesizes, Path-conflation might be most frequent with verbs for uncontrolled motion (down), least frequent with verbs for general direction (to/from) and intermediate with verbs for controllable motion (up/ down, in/out). These can be said to show a markedness hierarchy – the notion we will discuss shortly. Classifications become particularly interesting when they show various asymmetries – statistical preferences for certain combinations of parameters and the absence of attested though logically possible types. Consider Nerlove and Romney’s (1967) sibling terminologies. These are based on eight logical kin types as defined by three



Approaching lexical typology 

parameters (sex of ego, sex of sibling, relative age) – e.g., whether one and the same term is used for all siblings, whether there are two separate terms for ‘brother’ and ‘sister’, whether there are four different terms (‘younger brother’, ‘elder brother’, ‘younger sister’, ‘elder sister’), etc., which leads to 4140 logically possible types. However, only 10 of those are attested in more than one language in the 245-language sample! In a similar vein, even though, perhaps, less spectacularly, one of the five logically possible types is not attested in the kinship typologies focusing on the patterning of terms for parent / uncle /aunt (with the roots in Morgan 1870 and further modified by later research). Thus, languages can have the same term for ‘father’, ‘father’s brother’ and ‘mother’s brother’ (the “Hawaiian”, or “Generational” type), or the same term for ‘father’ and ‘father’s brother’, as opposed to ‘mother’s brother’ (the “Iroquois”, or “Bifurcate merging” type). However, no language has so far shown a system with one and the same term used for ‘father’ and ‘mother’s brother’, to the exclusion of ‘father’s brother’. The two kinship asymmetries cited above are examples of linguistic universals – generalizations on what is generally preferred/dispreferred (or even possible/ impossible) in human languages – the search for which has been a high priority on the typologists’ agenda for a long time. Linguistic universals within the lexical-typological research are often formulated in terms of lexicalization hierarchies/implications and lexical universals. The task is then both to unveil and to explain them. Continuing on the issue of the Morgan-inspired kinship terminologies – why is one of the five logically possible lexicalizations not attested? It has been suggested that the reason is cognitive: a definition of a term covering both ‘father’ and ‘mother’s brother’ would be cognitively more complex than the other four lexicalizations, since it will require disjunction (‘father’ or ‘mother’s brother’, cf. ‘male relative of one’s patriline’ for ‘father’ and ‘father’s brother’). It has also been suggested that the parent/aunt/ uncle terminology has the central role in predicting other properties in the kinship systems, such as, e.g., the sibling terminology (where the generalizations can be formulated as lexicalization implications); this, in turn, can be explained by the role of the corresponding relations in regulating marriage possibilities (for details cf. Evans 2001). Most lexicalization generalizations build on the unequal status of different words and other lexicalized expressions – either for encoding a particular conceptual domain, or in general. Some words are basic, while others are derived, some are less marked, whereas others are more marked – defined by various criteria stemming from Berlin & Kay’s (1969) classical study of colour terms, and the numerous publications by Greenberg (e.g., 1966), Brown (e.g., 1976) and by Berlin (e.g., 1992). One of Greenberg’s own examples nicely links to the discussion of kinship terminology in the preceding paragraph. Greenberg (1980) suggests that each of the different semantic

 Maria Koptjevskaja-Tamm

components used for analyzing kin terms defines its own markedness hierarchy, as shown in Table 4: Table 4.  Markedness relations among kin terms according to Greenberg (1980) Less marked

More marked

Example

lineal consanguineal less remote (measured in number of generations) senior (including both seniority within one generation and the distinction ascending-descending)

collateral affinal more remote (measured in number of generations) junior

‘father’ < ‘uncle’ ‘brother’ < ‘brother-in-law’ ‘father’ < ‘grandfather’ ‘older brother’ < ‘younger brother’

Markedness, according to Greenberg, would manifest itself in – zero expression vs. overt expression for certain categories cf. the consanguineal vs. the affinal relation in such pairs as brother vs. brother-in-law; – defectivation, i.e., the absence of a term in the marked category which would correspond to an existing one in the unmarked category cf. *cousin-in-law; – neutralization of certain distinctions in some categories (cf. neutralization of sex reference in cousin as against brother and sister), and – higher text frequencies for unmarked categories. When all the individual markedness hierarchies are combined the least marked of all the kin terms turn out to be parental terms, which is further corroborated by various properties singling out these terms across languages (cf. Dahl & KoptjevskajaTamm 2001). Since the discussions of basic vs. non-basic terms in the domain of colour are by now well known and widely quoted, the reader is referred to the relevant literature. Let us look at a couple of other examples of lexicalization generalizations pertaining to perception verbs and body-part terminologies. Viberg (1984, 2001) suggests that perception verbs across languages follow the sense-modality hierarchy sight >

hearing >

touch >

   smell taste

Not only does the hierarchy conform to the standard markedness criteria, but it also restricts patterns of intrafield polysemy, or semantic extensions of perception verbs within the perception domain: for instance, in Russian the verb slyšat’ “to hear” is often used in combination with the noun zapax “smell” for reference to smelling (a verb normally relating to a higher sense modality extends its uses to lower modalities,



Approaching lexical typology 

whereas the opposite does not seem to be attested) (cf. also Vanhove this volume; for a possible exception cf. Maslova 2004). There is a relatively long tradition of cross-linguistic markedness generalizations about the inventories of body-part terms, starting with Brown (1976) and Andersen (1978) (and further developed in Brown 2001; Wilkins 1996), e.g.,: – If both hand and foot are labelled, they are labelled differently (cf. English and Italian in Table 2); – If there is a distinct term for foot, then there will be a distinct term for hand (cf. English, Italian, which have both, vs. Japanese and Rumanian, which have a distinct term for hand, but not for foot, and Russian, that lacks either in Table 2). Body-parts differ, obviously, in how they relate to each other and to the whole body; the nose is, for instance, a part of the face, which, in turn, is a part of the head, which, in turn, is part of the body. Body-part terminology is therefore interesting for crosslinguistic generalizations on the partonomic levels of depth, or ethnolinguistic partonomy. Andersen (1978) suggests, for instance, that – there are never more than six levels of depth in the partonomy relating to body part terminology, and – there will be distinct terms for body, head, arm, eyes, nose and mouth The ten language descriptions in Majid et al. (2006) challenge a high portion of the earlier cross-linguistic research on categorization and linguistic/conceptual segmentation of the body. Thus, Lavukaleve, a Papuan language isolate spoken on the Russell islands within the central Solomon islands (Terrill 2006: 316) has one and the same word, tau, for both arm and leg, contradicting the claim that arm is always lexicalized by a distinct term. In addition, Lavukaleve has a distinct simple word, fe, for reference to foot, but nothing comparable for hand – contradicting therefore another of the claims above. Lavukaleve (as well as some of the other languages in the volume, e.g., Savosavo, another Papuan language in the Solomon islands, cf. Wegener 2006) shows also that linguistic evidence for partonomic relations between subset of body part terms can be difficult to achieve and that languages do not necessarily show any multi-level conceptualization in this domain, thus contradicting some of the universals proposed in Brown (1976) and Andersen (1978).

4.5  Explanations Linguistic typology in general seeks explanations for two different kinds of observed facts and generalizations – first, for the observed patterns in the linguistic phenomena themselves and, second, in their distribution across languages. A few examples of the

 Maria Koptjevskaja-Tamm

latter have been given above (e.g., in connection with deictic/non-deictic verbs and the hand/arm vs. finger/hand distinctions). Examples of the former are, for instance, the standard explanations for the possible colour-term systems as governed by the neurophysiology of vision (Berlin & Kay 1969; Kay & McDaniel 1978; Kay & Maffi 2000). Vision itself is also dominant among the sense modalities, as is well established within cognitive psychology and neurophysiology, which explains its highest position in the hierarchy for perception verbs (Viberg 2001: 1306–1307). Biology-rooted factors (e.g., perceptional discontinuity and other properties derived from perception), as well as functions have also been suggested as underlying segmentation of the body across languages (Brown 1976; Andersen 1978; also Enfield et al. 2006 for the discussion of the explanations and some counterexamples). The unmarked status of parental terms in the kin term systems follows, of course, from the unique biology-rooted status of parents (or at least of the biological mother) in any person’s life. All these examples, together with the earlier cognitive-based account for the cross-linguistically unattested lexicalization of ‘father + mother’s brother’ to the exclusion of ‘father’s brother’, suggest that lexical categories can be motivated – at least partly – by non-linguistic cognition and shaped by human perceptual and cognitive predispositions. There are also other possible explanations for cross-linguistic lexical preferences. We have already seen some examples of social and cultural practices as explanations for the cross-linguistically recurrent lexicalization patterns in connection with hand/ arm, finger/hand and parent/uncle/aunt kinship terminologies. Now, consider the above-mentioned cross-linguistic preference for conflation of path (down) and uncontrolled motion in verbs like fall or sink (Viberg 2006: 113). An obvious explanation here lies in the power of the omnipresent gravity in human environment: uncontrolled – and often unintended motion – under normal conditions in the default case means falling or sinking (in water). The long tradition of research on kinship has on the whole been an arena for hot disputes on the role of universal (biology-rooted) vs. social-construction explanations (see Foley 1997: 131–149 for an overview). Numerous examples of environmental, social and cultural explanations, combined with biology-rooted factors are found in the recent research on colour (e.g., Wierzbicka 1990; MacLaury 2001; Jameson 2005; Dedrick 2005; Paramei 2005 and the references there). Other possible explanations for cross-linguistically recurrent lexicalizations include the natural logic of events (rational and purposive connections among the components of the same event, e.g., Enfield 2007) and – last, but not least – innate concepts.

4.6  Universal vs. language-specific lexicalizations? Let’s come back to the initial questions asked in the beginning of Section 4. What meanings can be or cannot be expressed by a single word in different languages, or what word meanings are universal, frequent, possible, impossible? Are there any uni-



Approaching lexical typology 

versal (or at least statistically predominant) restrictions on the meanings that can or cannot be expressed by single words across languages, or are languages more or less free to choose here? It is in fact difficult to evaluate to what extent these questions have been approached and answered in the research up to now. Apart from the fact that very few concepts and conceptual domains have been studied at all, the different research traditions are hard to bring into line with each other, their results are often incommensurable and hardly mutually translatable. Take the hierarchy of perception verbs mentioned above. What does it actually mean for the universal-lexicalization enterprise? Does is imply that some languages will not lexicalize the meaning ‘to hear‘ and many more languages will not lexicalize the meanings ‘to taste’, ‘to touch’ and ‘to smell’? Conversely, if this is true (or at least partially true) – wouldn’t that imply that the word referring to hearing in a language which also has designated expressions for tasting, smelling and touching, will have a different meaning from the word that covers perception in all these modalities? Likewise, would the existence of the languages that merge ‘father’ and ‘father’s brother’, as well as ‘mother’ and ‘mother’s sister’ mean that biological parents do not correspond to universally lexicalized meanings? And how does this relate, in turn, to Greenberg’s markedness universals for kin-terms and the “parental prototype” suggested both there and in Dahl & Koptjevskaja-Tamm (2001)? As becomes clear, the answers to these questions are crucially dependent on what stance we take on the issues of polysemy and approximate vs. precise meaning identity discussed in Section 3. The overview in Goddard (2001; cf. also Brown 2001) which makes a serious attempt to take polysemy into consideration and aims at precise mean­ing definitions (in the NSM tradition), is probably the most updated one over suggested lexico-semantic universals; the important predecessors to the paper include the collective volume (Goddard & Wierzbicka 1994). As Goddard argues, “see” and “hear” seem to stand the proof of being universally lexicalized (at least as separate meanings within polysemous expressions). Some presumably basic and universal “concepts” seem to be doubtful as lexical universals or, at least, can only be viewed as approximate ones (e.g., “eat”, “give”). A few of the other surprises include the non-universal status of “water” and “sun” based on the fact that languages can have more than one word for each of those (cf. “hot water” vs. “non-hot water in Japanese, “sun low in the sky” vs. “hot sun overhead” in Nyawaygi, Australia). Emotions, contrary to many common assumptions, turn out to be highly culture- and language-specific. “Mother”, in its biological sense, has reasonable chances to survive as a lexical universal too, whereas “father’s” chances are considerably lower (since his role is much more subject to social factors). This, in fact, leads to interesting asymmetries in kinship terms, where “mother + mother’s sister” is supposed to be polysemous, whereas “father + father’s brother” is “allowed” to be semantically general – a fact not mentioned by Goddard (2001). All in all, according to Goddard, the best candidates for the universally lexicalized meanings

 Maria Koptjevskaja-Tamm

turn out to be overwhelmingly found within the set of semantic primitives suggested within the NSM (which now covers about sixty items). However, opinions vary as to what can count as universal lexicalizations and on the kinds of evidence for or against them (cf. Brown 2001), also among the individual researchers who have contributed to the papers in Goddard & Wierzbicka (1994). Consider ‘want’, which the NSM considers a semantic primitive, lexicalized in all languages. A recent study casting doubts on the suggested universality of ‘want’ is Khanina (forthc.), who shows that language after language in a sample of 73 genetically, areally and structurally diverse languages merge this meaning with other meanings (often modal and mental-emotive) in one and the same lexeme. In fact, exclusively desiderative expressions are predominantly found in the languages of Eurasia, excluding South Asia, and those of Northern America, while two thirds of the desiderative expressions in Khanina’s sample show other meanings as well. In many of these, a case can be made for polysemy; in others, however, there do not seem to be obvious morphosyntactic differences between desiderative and other uses. This does not exclude that further analysis will not provide arguments for polysemy even there. However, the fact itself that ‘want’ relatively seldom has a “lexeme of its own”, but tends rather to share the same lexeme with other meanings, is significant and has to be taken into consideration in discussions of universally lexicalized meanings. Khanina’s own conclusion is that the status of ‘want’ is subject to cross-linguistic and cross-cultural variation, as many other concepts: in some cases it is an indivisible and indefinable salient concept of the culture, whereas in others it is not salient, but is treated only as particular type of a more general situation. It is not always clear how precise semantic identity is established cross-linguistically and to what extent it is interesting and important to achieve that degree of exactness. Many scholars take cross-linguistic semantic comparability fairly easily, without always being aware of this. This is, for instance, characteristic of the research areas presented in the next two sections. In certain cases this is undoubtedly justified by the task and aims of the research whereby a higher degree of semantic precision might render the task undoable and create obstacles for interesting generalizations; in other cases, on the contrary, semantic vagueness cannot be justified and can even be shown to be detrimental for an effective cross-linguistic comparison.

5.  W  hat different meanings can be expressed by one and the same lexeme or by words derivationally related to each other? What different meanings can be expressed by one and the same lexeme, by lexemes within one and the same synchronic word family (words linked by derivational relations) or by lexemes historically derived from each other? These questions can be



Approaching lexical typology 

approached from different angles. Most of the other contributions to the volume start with an individual lexical item, or with several items belonging to one and the same individual lexical field, and ask what other lexical or grammatical meanings can be expressed by the same form(s) or by forms derived from it/them. In the sub-sections below we focus on semantic relations between particular lexical units, or particular meanings – i.e., particular instances of semantic motivation. But we can also compare whole classes or groups of words where one of the classes contains words derived from, or formed on the words in the other one, and ask about the semantic relations associated with a particular word formation device. The focus here is on the regularities in lexical motivation seen as an interaction of formal (morphological) and semantic motivation (cf. Koch & Marzo 2007). The next two subsections will give a short overview of the current crosslinguistic research along the two lines. The bulk of examples come from the body domain, but we will also briefly touch upon some studies on derivational morphemes and compounding.

5.1  F  ocusing on semantic motivation: Another look at the body and outside When foot and head are used with meanings different from their normal body-part meanings in the expressions the foot of the mountain and the head of the department, they present clear cases of polysemy. In other cases two meanings do not coexist within one and the same lexeme, but are related diachronically. The Italian testa “head” has developed from the Latin word testa “splinter”, but does not show the original meaning any longer. On the other hand, such meanings of the German word Haupt as “head (of the department, delegation etc.)” and “top (of a mountain)” have developed from its earlier body-part meaning “head”, for which Haupt in Modern German has been more or less replaced by Kopf. A useful cover term for all these cases is semantic shift, which refers to a pair of meanings A and B linked by some genetic relation, either synchronically or diachronically. Diachronic semantic shifts can sometimes lead to heterosemy (rather than to polysemy), which refers to “cases (within a single language) where two or more meanings or functions that are historically related, in the sense of deriving from the same ultimate source, are borne by reflexes of the common source element that belong in different morphosyntactic categories” (Lichtenberk 1991: 476). The pair head (like my head) and ahead (ahead of me) is an example of heterosemy. It is useful to distinguish between intrafield semantic shifts, that relate two meanings belonging to the same semantic domain, and interfield, or transfield semantic shifts, that relate meanings belonging to different semantic domains (all those quoted above). What might count as intrafield polysemy of body-part terms is a difficult question which brings to the fore the problem of distinguishing semantically general meanings,

 Maria Koptjevskaja-Tamm

on the one hand, from polysemy – cf. the discussion in sections 3.2. and 4. It has, for instance, been widely debated whether ruka “hand/arm” in Russian is semantically general or polysemous (cf. Wierzbicka 2007 for a recent hefty argumentation in favour of polysemy); it might very well turn out that a large portion (or even all) of the languages classified as neutralizing the hand /arm or the finger / hand distinctions in Brown (2005a,b) show in fact recurrent patterns of intrafield polysemy rather than semantic generality. I will not dwell on intrafield polysemy or on derivation of bodypart terms in this paper. There has been interesting cross-linguistic research on diachronic semantic shifts leading to the emergence of various body-part terms (both intra- and interfield shifts) – cf. Koch (this volume) and the references there. The best cross-linguistically studied cases of semantic associations pertaining to body involve body-part terms as grammaticalization sources for markers of spatial relations and reflexive-middle-reciprocal markers (in most cases leading to heterosemy). Various cross-linguistically interesting generalizations on grammaticalization body-parts → spatial relations and explanations for them are found in Svorou’s (1993) cross-linguistic study covering 55 languages of the world, and Heine’s (1989) and Bowden’s (1992) studies of the numerous African vs. Oceanic languages. Thus, frontregion relations are often expressed by markers coming from terms for eye, face, forehead, mouth, head and breast/chest, while those for back-region relations often emerge from body-part terms for back, buttocks, anus and loins (Svorou 1993: 71–72), e.g.,

(2) Examples of development body-parts → spatial relations (Svorou 1993: 71–72)

a. Halia (Austronesia, Oceanic, NW and Central Solomons)   i matana “in front of ” < i “in, at” + mata “eye” + -na (adv.suf) b. !Kung (Khoisan)   tsi’i “in front of ”

<

ts’i “mouth”

c. Navajo (Na-Dene)   bi-tsi “in front, at”

<

’atsii’ “head, hair”

d. Basque (Isolate)   gibelean “in back of ”

<

gibel “back” + -ean (loc)

e. Papago (Aztec-Tanoan)   -’a’ai “in back of ”

<

’a’at “anus”)

There are significant cross-linguistic differences here: terms referring to one and the same body-part can sometimes give rise to different developments; in addition, languages can also differ in their preferences for using particular body-part terms as sources for spatial markers. For instance, in Svorou’s sample, ‘head’ gives rise to markers of back-region in 12 cases and to markers of front-region in 2 cases, while ‘back’ gives



Approaching lexical typology 

rise to markers of back-region in 15 cases, of top-region in 3 cases and of bottomregion in 1 case. There are interesting explanations for these facts, based on universal preferences and on areal/genetic factors. Some of the differences in the semantic developments of “comparable” body-part terms have been interpreted as consequences of two different models, according to which anatomy is mapped into spatial relations – the anthropomorphic model (corresponding to the canonical upright position of a standing man with his/her back oriented backwards), and the zoomorphic model (corresponding to the canonical position of an animal standing on its four legs with its back oriented upwards) (Svorou 1993: 74–76; Heine 1997: 37–49). The anthropomorphic model is cross-linguistically preferred in being found more often across languages; in addition, even languages with predominantly zoomorphically modelled spatial concepts have at least some that emerge from the human body (Svorou 1993: 72–75; Heine 1997: 40). This is, of course, not particularly surprising given the general human predilection for anthropocentricity and embodiment in its different aspects. Zoomorphically modelled spatial concepts, on the contrary, seem to have a clearly areal and genetic distribution. There are also remarkable “areal differences in the relative weight given to the three major body regions” (i.e., head, trunk, and extremities, Heine 1997: 43. The proportions among the relevant semantic shifts in the Oceanic languages (Bowden 1992) more or less correspond to those in Svorou’s global sample in that the ‘head’ and the ‘trunk’ each provide about 49% of the sources. The African languages (Heine 1989), on the other hand, significantly differ in the proportions between the ‘head’ (38%) and the ‘trunk’ (60%) with the ‘belly’ as the more prominent source for spatial orientation in Africa than elsewhere in the world. The brief presentation above shows, thus, that the cross-linguistic studies on the grammaticalization path body-parts → spatial markers “live up” to such expectations of modern typological research as proposals of explanations for the cross-linguistic patterns and for the asymmetries in their distribution across a large sample of languages. Another relatively well-studied group of shifts with body-parts as source is involved in grammaticalisation of reflexive-reciprocal-middle markers. Schladt’s (2000) cross-linguistic study based on 150 languages shows that ‘body’ constitutes the absolutely most frequent source for reflexive markers, while ‘head’ is one of the other major ones. Since reflexive markers in turn tend to develop further into reciprocal and middle markers, ‘body’ and ‘head’ may also grammaticalize into those as well (cf. Heine & Kuteva 2002 for discussion and numerous examples). Although Schadt’s sample is geographically biased (and contains many more African languages than languages from the other parts of the world), it seems sufficient for manifesting interesting areal differences in the distribution of the grammaticalization sources. Thus, the reflexive markers in the African languages are derived much more frequently from ‘body’ and ‘body parts’ than elsewhere; in addition, within this category, ‘body’ is the much more preferred option

 Maria Koptjevskaja-Tamm

than ‘head’ in Africa than in Asia and Europe, whereas the languages of Northern America use exclusively ‘body’. There are other well-known grammaticalization paths from body-part terms attested across languages, but studied in a less systematic cross-linguistic way, e.g., the following ones: Body-parts → numerals, in particular, ‘hand’ → ‘five’ (e.g., lima “hand, five” in Samoan and all over Austronesian). As Heine & Kuteva (2002: 166) suggest, “[n]ouns for ‘hand’ probably provide the most widespread source for numerals for ‘five’ in the languages of the world”, but it is not quite clear what sample underlies this generalization. ‘Finger’ sometimes occurs in expressions for ‘six’, such as ‘a finger passes hand’, while ‘foot’ may occur in expressions like ‘two hands, one foot and one finger’ for ‘sixteen’ (Harald Hammarström p.c.); ‘hand’ → possession, for which Heine & Kuteva (2002: 167) give a few examples from African languages and suggest that this is an areally induced process. However, Estonian demonstrates a similar grammaticalization path (Ojutkangas 2000) – incidentally, several body-part terms in Estonian have also grammaticalized into spatial postpositions (related phenomena have a wider attestation in Finnic and even Finno-Ugric, Bernhard Wälchli p.c.). It is rather striking that none of the numerous cross-domain semantic extensions from body-part terms not leading to grammaticalization and attested all over the world, has been studied in a systematic cross-linguistic way, at least remotely comparable to the studies reported on above. A particularly interesting topic is the use of body-part terms in conventionalized descriptions of emotions and mental states, a phenomenon found all over the world, e.g., (3) Lao (Tai-Kadai) (Enfield 2002: 87) a. aj3-hòòn4 b. caj3-dii3 c. caj3-kaa4 heart-hot heart-good heart-daring ‘impatient, hot-headed’ ‘nice, good-hearted’ ‘daring, courageous’ (4)

Kuot (isolate, “Papuan”, New Ireland, Papua New Guinea) (Lindström 2002: 162) gigina-m [dal’p a] heavy-3pl stomach.nsg 3m.poss ‘he is worried/sad’ (lit. ‘his stomach is heavy’)



(5) Yélî Dnye (isolate, “Papuan”, Rossel island, Papua New Guinea) (Levinson 2006: 237) a nuu u tpile my throat his/its/her thing ‘A thing I really like it’ (lit. ‘My throat its thing’)

As the examples above demonstrate, languages can differ significantly in which body-parts can be seats for which emotions: the majority of emotional descriptions



Approaching lexical typology 

in Lao, Kuot and Yélî are based on ‘heart’ vs. ‘stomach’ vs. ‘throat’. Some areal tendencies have been suggested: thus, Lao seems to be representative for the languages in Southeast Asia (Matisoff 1986), which often use the word for ‘heart’ or ‘liver’ in most emotion descriptions. Enfield & Wierzbicka (2002) contain a number of enlightening papers on different languages, and a first important step towards a more systematic cross-linguistic comparison is provided by the volume on the role of heart in expressions of emotions (Sharifian et al. eds. forthc.). Large-scale crosslinguistic comparisons in this area are, however, still lacking – and are badly needed, in particular, given the prominent role of both body and emotions in the current cognitively-oriented semantic theories. The methodological and theoretical problems for such comparisons are, however, overwhelming. On the one hand, as we have seen, categorization of the body itself is subject to significant cross-linguistic variation. On the other hand, unveiling and describing the meaning of emotion expressions even in one’s native language is a much more difficult enterprise than in many other domains, emotions are to a high degree culture-specific, and systematic cross-linguistic comparison of emotion expressions has hardly begun (the very interesting papers in Harkins & Wierzbicka (2001) contain descriptions that are not directly comparable to each other). Some types of semantic associations involving body-part terms are probably restricted to certain linguistic areas or to particular linguistic families. Among the recurrent features of Australian lexical systems Evans (1992: 479) mentions examples of “synecdoche, by means of which animals or plants are named for their most salient body-part”. Thus, ‘tooth’ is extensively used in such names, leading to the polysemy of muyiny in Wadyiginy “dog, wild asparagus”, or to the different meanings within “the set of cognates of *waartu including Umbugarla waartu “mosquito”, Yolngu wart:u “dog”, Kayardild waardu “sandfly” and wardunda “mangrove rat””. The typical recurrent Australian metaphors include extensions of ‘eye’ to any point-like entities, including ‘star’, ‘well’, ‘small hole in ground’ and ‘bullet’, and the use of ‘ear’ as the seat of intelligence and apprehension, e.g., ‘ear-bad’ or ‘ear-without’ = ‘crazy’, etc. The cross-linguistic research on semantic associations and semantic shifts in domains other than body is even less systematic. The best-studied cases involve again heterosemy and grammaticalization from different sources (e.g., motion and posture verbs, “give”, “acquire”, etc.; cf. also the discussion of “want” in Section 3.6. – cf. Heine & Kuteva 2002 for the numerous examples and references). One of the notable few exceptions is the research on perception verbs developing cognitive meanings (Sweetser 1990; Evans & Wilkins 2001, cf. also Vanhove this volume). Brown (2001) reports on several other cross-linguistically recurrent connections (‘wood’ vs. ‘tree’, ‘seed’ vs. ‘fruit’, ‘wind’ vs. ‘air’), for which social and cultural factors have been suggested. Thus, speakers of languages where there is polysemy between ‘wood’ and ‘tree’ (two thirds of the languages in a big sample) usually live in small-scale, traditional

 Maria Koptjevskaja-Tamm

societies, while speakers of languages separating them usually live in large nation states (cf. also Section 4.3 for the potential body-part examples).

5.2  Formal motivation and its semantic correlates The preceding section has dealt with cross-linguistic studies on particular instances of semantic motivation. In this section we will briefly touch on the issue of the regularities in lexical motivation seen as an interaction of formal (morphological) and semantic motivation (cf. Koch & Marzo 2007). The issues we discuss here include the following: – what meanings correspond to “more basic” vs. “regularly derived” words?, – what formal (primarily morphological) devices are there in a language for forming words from other words, or lexical units from other lexical units – e.g., derivation, compounding?, and – what meaning relations can be expressed by these devices? Three examples of recent large-scale typological studies will give a flavour of how these questions can be approached and answered. The first of them, Nichols et al. (2004) approaches the issue of more basic vs. regularly derived words and how it interacts with the formal word-forming devices in a language. Given two sets of words such that the words in one set are semantically almost identical to those in the other apart from one and the same meaning component (or, put differently, the words in the two sets are related to each other by means of one and the same meaning relation), are there systematic formal relations between the words in the two sets and if yes, which of the words will be formally more basic vs. derived from the others? The two other studies, Juravsky (1996) and Wälchli (2005), focus on a particular word-forming device, derivation of diminutives and co-compounding, and ask the third of the questions stated above, namely, what meaning relations can be expressed by them. Although both studies concentrate on languages that have these devices, Wälchli (2005) is also interested in the question of absence vs. presence and, generally, of uneven distribution of cocompounding across languages, i.e., in the second of the above stated questions. The first example of large-scale cross-linguistic study that takes regularities of lexical motivation seriously is an important recent contribution by Nichols et al. (2004). It presents a typology of 80 languages based on their treatment of what the authors view as semantically basic and almost universal intransitive verbs such as ‘sit’, ‘fear’, ‘laugh’, ‘fall’ and their transitive counterparts (all in all 18 pairs). The list of verb pairs takes into consideration various parameters that are known or supposed to have an impact on derivational processes (animacy, agency, resistance to force), to be sufficiently common and easily found in lexical sources and/or translatable, and to show a wide spread in lexical semantics.



Approaching lexical typology 

The main question is whether the two sets of words are formally related to each other and if yes, how – i.e., which of the classes contains words derived from the words in the other one. It turns out that languages tend to be consistent in whether they treat intransitives as basic and transitives as derived by means of causative morphology (transitivizing languages), whether they derive intransitives by means of anti-causative morphology (detransitivizing languages), whether both intransitives and transitives are encoded by the same labile verb (neutral languages) or whether both intransitives and transitives have the same status (indeterminate languages), cf. Table 5. Table 5.  Examples of the types of lexical valence orientation illustrated by the derivational relations between the intransitive and transitive ‘hide’ (based on Nichols et al. 2004).

‘hide’ (go into hiding) ‘hide’ (put into hiding)

Chechen: transitivizing

Russian: detransitivizing

Thai: neutral

Nanai: indeterminate

dwa+lechq’ dwa+lechq’a-d-

prjatat’-sja prjatat’

àop àop

siridjaja-

The study is, thus, based on a principled sample of lexemes, and the resulting language types stem from their frequencies in the principled global sample of languages – which, as the authors themselves claim, defines this work as piece of lexical typology. There are also further various statistically significant generalizations on the “inner logics” of the types themselves, on their correlation to other linguistic phenomena (grammatical and lexical, e.g., alignment and voice alternations, complexity, aspect and Aktionsart) and on their distribution across the languages of the world – in the robust tradition of the standard modern large-scale typological research. Nichols et al. (2004) builds on a long tradition of cross-linguistic and typological studies on causatives, anti-causatives and, more generally, transitivity alternations (the cumulative results of which could be profitably used for constructing the list of verb pairs) – which does not reduce the groundbreaking character and the value of the study itself. Jurafsky (1996) and Wälchli (2005), each focusing on one particular wordforming (derivational vs. compounding) device across a large number of languages, are rather of a more exploratory nature, even if they both have theoretical and methodological predecessors (partly coming from the same tradition as Nichols et al. 2004; e.g., Kemmer 1993; Haspelmath 1987, 1993). Juravsky (1996) focuses on one particular derivational category – diminutives – identified via their ability to mean at least ‘small’. Wälchli (2005) studies co-compounds (dvandva compounds, pair words, copulative compounds) – compounds like ‘fathermother’ for parents, ‘milk-butter’ for ‘diary products’, etc., which are opposed to subordinate compounds, like ‘fingertip’ or ‘apple tree’, in which there is an asymmetric relation between the two parts (the head and its modifier). Each of these categories can be expressed by a variety of morphological devices – e.g., affixation, shifts in consonants,

 Maria Koptjevskaja-Tamm

vowels or lexical tones, and changes in noun-class or gender for diminutives. Each of them has also a number of different meanings and uses and, in addition, is not universal at all, even though both diminutives and co-compounds exist in many languages. It is instructive to compare how the two studies approach their object. Juravsky’s main interest lies in the amazing variety of semantic functions expressed by diminutives, in addition to the meaning ‘small’, e.g., ‘child/offspring’ (Tibetan dom “bear” vs. dom-bu “bear cub”), small-type (Ewe hɛ̄ “knife” vs. hɛ̄-vi “razor”), “imitation” (Hungarian csillag “star” vs. csillagocska “asterisk”), intensity/exactness (Latin parvus “small” vs. parvulus “very small”), approximation (Greek ksinos “sour” vs. ksinutsikos “sourish”) and individuation (Yiddish der zamd “sand” vs. dos zemdls “grain of sand, Juravsky 1996: 536). They are also known to easily acquire connotations (or uses) of affection, sympathy and endearment or, conversely, of contempt, and are extensively used for various pragmatic functions, such as politeness. The main question is then what other meanings and uses can be attributed to the diminutive marker and what is the rationale behind this. To account for this, Jurafsky proposes a structural polysemy model (inspired by Lakoff ’s radial-category notion) in which the different senses displayed by diminutives are modelled together with the metaphorical and inferential relations among them. The model has both synchronic and diachronic applications, where the latter cover, among others, possible lexical sources for the category itself (words semantically or pragmatically linked to children). Juravsky’s study leaves many questions. A major problem is the lack of “standard typological” systematicity in the account for the data and for the analysis, even though the study is based on extant grammars and work with consultants for more than 50 genetically, structurally and areally diverse languages. Thus, the different meanings and uses underlying the radial category are merely presented by illustrations from one or several languages, without any overview over their distribution. We do not know which of the functions are cross-linguistically more or less common, and what kinds of genetic, areal or other patterns there are in the distribution of diminutives and of their various meanings and uses. Grandi (2002) shows, for instance, that diminutives and augmentatives are an interesting areal phenomenon in the Mediterranean languages, with some properties distinguishing them from the other genetically related languages. The rationale behind Jurafsky’s semantic map is not quite clear either, as opposed to the logic of semantic maps common in typological research (Haspelmath 2003, François this volume). All this makes the use of “universal” in the title somewhat doubtful, but compared to the majority of studies anchored in cognitive semantics with its inclination to allegedly universal claims, Jurafsky does provide ample crosslinguistic data and opens opportunities for future research. Wälchli (2005) concentrates on several aspects of co-compounding. Cocompounds occur in several semantic types and functions, show a variety of formal patterns and considerable differences in frequencies of textual occurrences. Some



Approaching lexical typology 

languages lack co-compounds, even those that have productive noun-noun compounding (e.g., modern Germanic languages); another group use them sparsely (in just a few functions and relatively infrequently, e.g., Mari or Hindi); finally another group, like Vietnamese and Tibetan, are highly co-compounding – they show a very high frequency of co-compounds in texts and use them in many different functions. All these parameters of variation are subject to Wälchli’s investigation which involves several kinds of data, with the bulk of the data coming from texts – both original and parallel texts (the Universal Declaration of Human Rights and the Gospel according to Mark) in a large number of predominantly Eurasian languages. The data give rise to various generalizations on the patterning of co-compounds and their distribution across languages. Thus, for instance, it turns out that the meanings expressed by cocompounds in a language (its semantic profile) are intimately linked to their textual frequencies. On the other hand, the distribution of co-compounding across languages (including their semantic types and text frequencies) shows a macro-areal pattern of distribution, with a significant and steady decline from continental East and Southeast Asia westward in Eurasia. Wälchli’s study contains many interesting generalizations and explanations, a number of which are of primary methodological and theoretical relevance for future research, also on co-compounding (e.g., in non-Eurasian languages).

5.3  Lexical semantics in cross-linguistic research on motivation As shown by the discussion of universal lexicalization in Section 4.6., there are very few meanings that can easily translate among languages, in particular if precise semantic identity is required. This fact is normally not explicitly considered in cross-linguistic research on motivation that usually takes meanings for granted, self-evident, and easily identified across languages and in a particular language. Consider the grammaticalization paths hand → five, attested in various languages, including Samoan (Polynesian, Austronesian) and Turkana (Nilotic, Nilo-Saharan), and hand → possession, attested, among other languages, in Kono (Mande, Niger-Congo), Zande (Ubangian, Niger-Congo) (Heine & Kuteva 2002: 166–167) and Estonian (Finno-Ugric, Uralic). Since all these languages use the same lexeme for ‘hand’ and ‘arm’, how would we know that it is ‘hand’ that has been the grammaticalization source rather than, say, ‘arm (excluding hand)’ or ‘arm (including hand)’? A strict proof for the case would include, first, arguments in favour of polysemy ‘hand’/‘arm’ rather than semantic generality in all these languages and, second, evidence for the grammaticalized meanings being based on ‘hand’ to the exclusion of ‘arm’. I am not aware of any serious attempts to do anything along these lines and doubt that there have been any. The interpretation of these particular examples and the postulation of these particular semantic links are, most probably, founded on common sense and intuition rather than on strict

 Maria Koptjevskaja-Tamm

argumentation, and on parallels with other languages which clearly distinguish between ‘hand’ and ‘arm’. Likewise, almost none of the 18 verb pairs used in Nichols et al. (2004) and viewed by the authors as semantically basic and almost universals, belongs to Goddard’s (2001) list over lexico-semantic universals (with some, like ‘sit’ and ‘fear’, being explicitly excluded from it). The exact semantics and precise semantic identity of the verbs on the list is, however, not a point here: the 18 verb pairs have been chosen on pragmatic grounds, as representing certain combinations of general parameters, corresponding to frequently encoded situations and having approximate translational equivalents in many languages. Stricter requirements on semantic comparability would in fact create obstacles for achieving the principal objective of the study. Obtaining one-word expressions with the same semantics for 18 events (and, in addition, representing the various combinations of interesting parameters) in 80 languages is hardly conceivable, while choosing word combinations with the right semantics would most probably conceal the basic derivational relations. In other cases, the relatively low degree of semantic precision in the definitions is less justified and can be impeding for deeper insights and effective cross-linguistic comparison. Among the various grammaticalization paths building on motion verbs several are often defined as starting with ‘come’ and ‘go’ (cf. in Heine & Kuteva 2002 for examples). The English verbs “come” and “go” as semantic metalabels are not totally felicitous; among other things, they encode the deictic distinction between centripetal and centrifugal motion, absent from many languages of the world (see Section 4.4. for the discussion of Ricca 1993) and neutralize the distinction between motion on foot vs. in a vehicle (cf. also Goddard 2001: 28). Descriptions like come → continuous, or go → habitual are therefore too vague for understanding the underlying logic of the development – they do, however, fulfill functions as preliminary crude classifications and as guidelines for future research. The lack of consensus on the appropriate semantic meta-language and the form of meaning definitions creates obstacles for evaluating cross-linguistic connections even between studies of high semantic and lexicographic quality. Consider Enfield’s (2003) excellent book on the striking pattern of multi-functionality (polysemy and heterosemy) involving the verb ‘to acquire’ and shared by the languages of mainland Southeast Asia. The study suggests a fine-grained classification of the different meanings, illustrated by numerous relevant examples and provided with detailed semantic explications. Viberg (2002, 2006) also presents excellent studies on the semantically comparable verbs in European languages, primarily få in Swedish and get in English. In particular, få in Swedish shows an amazing diversity of uses, which has a clearly areal distribution (being replicated by its Norwegian cognate få and the etymologically unrelated verb saada in Finnish). Viberg (2006: 125) writes that “[e]ven if få has a relatively language-specific pattern of polysemy with respect to European languages in



Approaching lexical typology 

general, it is not without parallels in other parts of the world” and mentions Enfield’s study. He concludes that “there is no exact parallel between the meanings of få and Lao daj4 [“acquire”, MKT] but at a more general level the paths of extension are similar” (Viberg 2006: 126). I wish I could understand what this means. The very different ways of classifying phenomena and representing their meanings in Enfield’s and Viberg’s studies make it, in fact, very difficult to evaluate the degree of (dis)similarity between the two polysemy patterns (cf. Auwera et al. forthc. for a semantic map comparing parts of the relevant polysemy patterns – the North-European and Southeast Asian acquisitive modals, i.e., modals based on ‘get’/‘acquire’).

6.  W  hat cross-linguistic patterns are there in lexicon-grammar interaction? Lehmann (1990: 163) defines lexical typology as research which focus on “typologically relevant features in the grammatical structure of the lexicon”, rather than on “the semantics of individual lexical items, their configurations in lexical field or individual processes of word formation” (Lehmann 1990: 165), i.e., the issues that have been considered as definitely belonging to lexical typological in the preceding sections. Lehmann’s definition partly stems from a somewhat narrower understanding of typology than the one(s) suggested in Section 2. Typology, in this view, has to be based on essential properties that “vary regularly in the population under consideration” (Lehmann 1990: 165). As Lehmann explains this, the lexicon contains all that is completely idiosyncratic – and that does not therefore fit the premises for typological research – but also many regularities. It is a complex structure built upon categories and relations, with items falling into a number of lexical classes that are intimately connected to grammar. It is these typologically relevant features that are the primary object of the lexical typology. In a similar vein, Behrens & Sasse’s programmatic sketch (1997) promotes Lexical Typology (refers to a specific research framework), the aim of which is “to investigate cross-linguistically significant patterns of interaction between lexicon and grammar”, and mentions, among others, the following: Viewed in the context of comparative linguistic research, the concept of lexicogrammar leads to the assumption that we can expect, in different languages, quite divergent patterns of interaction between lexicon and grammar, and that these divergences are of great typological significance. It is therefore proposed that lexical semantics and its repercussions on grammar be assigned a central role in typological investigations. To this end, we will lay much emphasis on the discovery of principles of ambiguity and compositionality. These principles are presumably universal on a higher level of abstraction but typologically variable in their concrete individual manifestations. They therefore strongly influence the make-up of an individual language’s grammar and lexicon (Behrens & Sasse 1997: 1–2).

 Maria Koptjevskaja-Tamm

A number of various cross-linguistic studies can be attributed to lexical typology understood as a search for typologically relevant features in the grammatical structure of the lexicon, or as typologically significant correlations between lexicon and grammar. They vary in how and to what extent they fit into the typological research framework and tradition(s), and in how and to what extent they consider lexicon. On the whole, there is very little awareness that the relevant studies do focus on lexical phenomena. Some are restricted to lexicon-grammar interaction for a particular conceptual domain/lexical field or even for a particular lexical meaning, e.g., body-part terms in adnominal possession and in special syntactic constructions such as possessor ascension/external possession and body-part incorporation (Chappell & McGregor 1996; the literature is too extensive to be listed); kin terms in grammar (Dahl & Koptjevskaja-Tamm 2001); ‘give’ and argument linking (Haspelmath 2005a; Kittilä 2006), different classes of complement-taking verbs and the structure of complementation (Cristofaro 2003) ‘want’ and the structure of desiderative clauses (Haspelmath 2005b; Khanina 2005). Veselinova’s (2006) large-scale study of suppletion in verb paradigms is an excellent example of lexicon-grammar interaction: it shows that suppletion tends to be linked to verbs with particular lexical meanings (e.g., motion), with different meanings picked up by suppletion according to different grammatical categories (e.g., tense-aspect-mood, or imperative). But many other traditional grammatical phenomena can be viewed as lexical. First of all, consider the issue of word classes which has been subject to much debate and disagreements (with the relevant works being too numerous to be listed here). Word classes present an example par excellence of interaction – and significant correlation – between lexicon and grammar. The jump from individual language descriptions to large-scale cross-linguistic research tends, however, to reduce lexical information to very few representatives for each “presumptive” word class (like ‘big’ and ‘good’ for potential adjectives), not always systematically checked and/or completely comparable across the languages in the sample. Nonetheless, in a number of cross-linguistic works word-class behaviour is studied with more attention to lexical semantics and against the background of relatively fine-grained lexical distinctions. Thus, Dixon (1977) and later Dixon & Aikhenvald (2004) approach the problem of adjectives by breaking up the vague class of property concepts into a number of much more coherent semantic types (dimension, colour, physical properties, human propensities, etc.) which differ in their propensity to be lexicalized as “adjectives proper”, as verbs, as nouns or in still other ways. The suggested semantic classification and the labels for the semantic types are, in fact, less important than the much longer concrete lists of notions used more or less systematically across all the contributions in Dixon & Aikhenvald (2004). The specific language chapters provide various pieces of evidence that the notions included into one and the same semantic type vary in their propensity to be lexicalized as adjectives “proper” or in other ways, and a more fine-grained semantic classification might hopefully grasp the cross-linguistic systematicity here.



Approaching lexical typology 

A radically “lexicon-based” stance is taken in Pustet’s (2003) study of copulas in a global 131-language sample. Earlier work, primarily Stassen (1997), has suggested that copulas across languages show different inclination to combine with/to be required with different kinds of predicates along the hierarchy nominals > adjectivals > verbals. In other words, the first place where copulas occur in a language will be sentences like “Peter is a boy”, followed by “Peter is big”, with sentences like “Peter goes” having copulas rather infrequently. Pustet combines these generalizations with the various parameters suggested in earlier research as underlying the distinctions between verbs, adjectives and nouns (primarily in Croft 1991) and tests to what extent these are compatible. A part of her study is based on checking the behaviour of the items in large lexical samples (ranging from 530 to 850 items) as predicates in ten genetically, areally and structurally diverse languages. The lexical items fall into fourteen lexical classes based on various combinations of the three parameters of valence, transience and dynamicity which together define a three-dimensional semantic space and correlate with the presence vs. absence of copulas in a more principled and refined way than the earlier suggested hierarchy formulated in terms of word classes. It turns out that the majority of the lexical items in the sample represent just a small number of specific feature bundles – which, by and large, correspond to, or define the lexical prototypes of “entity”, or prototypical nominals (valence 0, –transient, –dynamic, e.g., ‘house’ and ‘old man’), “property”, or prototypical adjectives (valence 1, ±transient, –dynamic, e.g., ‘big’), and “event”, or prototypical intransitive and transitive verbs (valence 1 or 2, + transient, + dynamic, e.g., ‘to go’ and ‘to buy’). The items within these each of these classes tend to show uniform behaviour with respect to copularization in a particular language. Lexical items from the other feature combinations, e.g., ‘smart’, ‘hand’ and ‘son’ (valence 1, –transient, –dynamic), ‘to rain’ (valence 0, +transient, +dynamic), ‘to love’ and ‘to know’ (valence 2, ± transient, – dynamic), share the behaviour of the “semantically adjacent” major classes, but only to a certain extent – the exact cuts-off between copularizing and non-copularizing lexemes are quite language-specific, even though governed by universal considerations (defined by the position of an item in the semantic space). The fact that language-specific vocabularies include numerous minor (and normally small in size) classes, whose semantic profile does not coincide with those of prototypical nominals, adjectivals and verbals, is, thus, brought to the fore and taken seriously in Pustet’s study. One of the theoretically important implications is also an identification of zones that show cross-linguistically recurrent overlaps in their wordclass attribution, often even within one and the same language – e.g., terms denoting nationalities (‘French’), emotional states (‘to love’), bodily states (‘to be tired’). Pustet’s principled sample can be used for further work on word classes and on their interaction with various grammatical categories. And in general, future research on word classes will certainly benefit from taking lexical semantics seriously and working with more fine-grained lexical classes than what has often been done.

 Maria Koptjevskaja-Tamm

Cross-linguistic variation in categorization within major word classes also offers many opportunities for research on cross-linguistically significant patterns of interaction between lexicon and grammar. Verbs can be categorized in many different ways, for instance, depending on their argument structure, on how it is linked to a particular sentence structure and to what extent it can be subject to various valence alternations. The literature here is extensive and diverse, but is, in fact, very seldom explicitly linked to lexical phenomena, rather than being considered as primarily syntactic. As Nichols et al. (2004: 183) put it, “[i]f ergativity, for instance, were viewed as lexical, an ergative language would be one exhibiting ergativity as the default option or majority pattern in some broad category of verbs (e.g., all transitives, or some category of transitives), based on a standard sample of glosses. Stative-active alignment, viewed lexically, is lexically conditioned split intransitivity as Merlan (1985) presents it” (cf. Section 4.2. for the overview of Nichols et al.). A nice exception is Bossong’s (1998) study of experiencer constructions (like “I am cold”, “I am sorry”, “I see X”). It is built on the standard list of ten experience-denoting expressions across the European languages and shows significant genetic and areal differences in frequencies of the different syntactic patterns in which these expressions occur (among others, singling out the specific Standard Average European pattern). A huge research domain of primary relevance for verbs focuses on the categories of aktionsart, aspect and tense. Although it is generally acknowledged that all these categories are very sensitive to the differences in the semantics of different verb classes, the valid cross-linguistic generalizations here are still quite few. Most modern research on aktionsart has its roots in Vendler’s (1967) verb classes (states, activities, accomplishments, achievements), whereas “[a]n urgent desideratum is the investigation of the role of lexicon, in particular the subcategorization of situation types”, as Sasse puts it in his overview of the recent development in the theory of aspect (2002: 263). Tatevosov’s (2002) study is a very promising step in this direction. It is based on a principled list of 100 “predicative meanings” (normally expressed by verbs or verb-based expressions) coming from several cognitive domains (from being and possession, motion, physical processes and changes, to phasal and modal verbs) and covering the “basic” verbal lexicon; these are checked for all possible combinations with the verbal tense-aspect categories and their resulting meanings in four genetically unrelated languages – Bagwalal (Daghestanian, NE Caucasian), Mari (Finno-Ugric, Uralic), Tatar (Turkic) and Russian (Slavic, Indo-European). Already this comparison falsifies the common assumptions “that notions on which Vendlerian classes are based are logically universal, hence are not subject to crosslinguistic variation” and that verbs or verb phrases with “similar meanings” in different languages (i.e., translational equivalents) will belong to the same verbal class as their English equivalents (Tatevosov’s 2002: 322). “Actionality”, used by Tatevosov instead of “Aktionsart”, turns thus out to be a parameter based on a universal set of elementary semantic distinctions, but allowing



Approaching lexical typology 

for different settings in different languages. Different languages show therefore their own language specific subcategorizations of the verb lexicon that can only be discovered via empirical investigations rather than taken for granted. A particularly beautiful example of the latter comes from Botne’s (2003) cross-linguistic study of die and its correspondences in 18 languages. ‘Die’ is always quoted as the prototypical example of Vendler’s achievement verbs (telic, or bounded, and punctual) in that it refers to the acute point demarcating life and death. Botne shows, however, that languages can differ in their lexicalization of the different stages in the process leading to death, which, in turn, has important consequences for the aktionsart categorization of the corresponding verb in a particular language. Cross-linguistic variation in categorization within nouns also offers many interesting topics for research on lexicon-grammar interaction. For instance, subcategorization of nouns according to their interaction with the category of number, involves various fascinating and cross-linguistically still poorly understood issues such as count-mass distinction, collectives, singular vs. pluralia tantum, etc. (for some examples see Corbett (2000); Wierzbicka (1988); Rijkhoff ’s (2002) notion of “Seinsarten”, Koptjevskaja-Tamm & Wälchli’s (2001) areal-typological study of pluralia tantum in the Circum-Baltic area against a broader European context, Lucy (1992), Behrens (1995) and Koptjevskaja-Tamm (2004) on count-mass distinctions across languages). Table 6 gives a flavour of how differently count-mass categorization can work in relatively closely related languages (and, in addition, in cognates). Table 6.  Some examples of count-mass differences in Russian, German, Swedish and Italian Count (+) or mass (–) strawberry fruit hair furniture

Russian

German

Swedish

Italian

– klubnika + frukt + volos – mebel’

+ Erdbeer – Obst ± Haar + Möbel

+ jordgubbe ± frukt ­– hår + möbel

+ fragola + frutto + capello + mobile

Possible implications of such variation for the lexical semantics of the items under consideration are very rarely explicitly acknowledged and discussed in cross-linguistic studies on lexicon-grammar interaction. Consider Botne’s (2003) conclusions following his cross-linguistic study on the aktionsart categorization of the correspondences to die: This small, exploratory study has shown that … achievement verbs, though unified by the punctual, culminative nature of ther nucleus, may be conceptualized in

 Maria Koptjevskaja-Tamm

different languages as encoding durative preliminary (onset) or postliminary (coda) phases in addition to the punctual nucleus. Consequently, die verbs frequently have a complex temporal structure and do not simply encode a point of transition … [T]he same “concept” will not necessarily be encoded with the same phases in every language. Consequently, appropriate cross-linguistic comparison and analysis of these kinds of verbs will perforce require a close analysis of a particular verb in each language (Botne 2003: 276).

But if languages differ as to which of the phases leading to death they encode in their ‘die’-verbs, can we still view them as encoding “the same concept”? Likewise, the meaning of the German mass noun Obst is hardly identical to that of the Russian count noun frukt, even though they constitute translational equivalents to each other. Crosslinguistic identification of phenomena based on “approximate”, rather than “precise” semantic identity, can be justified when the primary focus of the cross-linguistic research is not on the lexical semantics per se (cf. with the discussion in Section 5.3.). However, it is also reasonable to take the next step and use the cross-linguistic variation in grammatical behaviour as evidence for the lexical-semantic differences. It is now widely acknowledged by various linguistic theories that a large portion of grammatical phenomena is rooted in the lexicon. Lexicon-grammar interaction will surely provide lots of challenges for the future lexical-typological research.

7.  Lexical typology: Past, present and future It is impossible to cover all the aspects of lexical-typological research in one paper. One important group of questions that have not been touched upon concerns crosslinguistically recurrent patterns in contact-induced lexicalization and lexical change: e.g., differences in borrowability among the different parts of the lexicon and the corresponding processes in the integration of new words, or patterns of lexical acculturation (i.e., how lexica adjust to new objects and concepts). The important contributions here include Brown (1999) and the on-going project on “Loanword Typology: toward the comparative study of lexical borrowability in the world’s languages” at Max Planck Institute for Evolutionary Anthropology (coordinated by Martin Haspelmath and Uri Tadmor, http://www.eva.mpg.de/lingua/files/lwt.html), which modify various traditional assumptions (like non-borrowability, or at least, “relative” non-borrowability of the items on Swadesh” “basic vocabulary” lists), come up with new cross-linguistic generalizations and suggest methodology for future research. Issues related to the “basic vocabulary” (in various understandings) and the “lexico-typological profile” of a language figure also in other connections in cross-linguistic research with lexical-typological ambitions (Viberg 2006; Koch & Marzo 2007; Kibrik 2003). There are also interesting questions on the interaction between lexicon and



Approaching lexical typology 

phonology, overall principles of taxonomic categorization or the organization of the lexicon, and surely many others. Possible theoretical implications of lexical typology are vast – for theoretical linguistics in general (cf. Koptjevskaja-Tamm et al. 2007 for some details) and for broader issues such as child-language acquisition (with Melissa Bowerman and Dan Slobin as the leading authorities) and “Linguistic Relativity”. In other words, there are numerous diverse and fascinating questions for the future lexical-typological research. Its basic and most urgent problems are primarily methodological. For instance, we need to: –

–

refine the existent methods of data collection and develop new ones, improve standards in cross-linguistic identification of studied phenomena and in their (semantic) analysis, achieve a reasonable consensus on the meta-language used for semantic explications and on the ways of representing meanings.

To start with the first issue, the methodology of data collection. Morphosyntactic typology has been largely dependent on secondary data sources, with reference grammars as the undoubtedly most often used data source, in many cases complemented by sporadic consultations with native speakers and/or language experts. Studies in morphosyntactic typology are typically a “one researcher’s job”: even when data collection involves filling in questionnaires and responding to other data elicitation stimuli, the people doing that part of job normally count as consultants, rather than co-authors (some of the exceptions being the tradition of the Leningrad/St.Peterburg Typological School, or the numerous collections edited by Aikhenvald and Dixon). For lexical typology, on the contrary, secondary sources are of marginal importance, in particular, if we take the three groups of questions that have been the main subject matter of this paper – categorization within conceptual domains, semantic and formal motivation, lexicon-grammar interaction. Relevant data are normally scattered across different kinds of secondary sources: a thesaurus might provide information on categorization within conceptual domains, while a “normal” dictionary may have something on the polysemy patterns and other formal-semantic relations within word families. Some information on lexicon-grammar interaction might be occasionally given in a reference grammar (which seldom lists all the words showing a particular grammatical behaviour), some might be appear in a dictionary. A desideratum would be to have a source that for every word in a language would give a precise meaning definition, show both its exact relations to other words and define its grammatical properties. There are a few attempts on the market towards this desideratum for the better described languages – e.g., “The interpretational-combinatorial dictionary” with roots in the Moscow School of Semantics (cf. Iordanskaja & Paperno 1996 for the excellent treatment of the Russian body-part terms in this tradition), the Berkeley

 Maria Koptjevskaja-Tamm

FrameNet project based on Frame Semantics (http://framenet.icsi.berkeley.edu/index. php?option=com_frontpage&Itemid=1)3 or the expanding enterprise of WordNet for several European languages (http://www.globalwordnet.org/).4 However, the lexicon for most languages of the world is – and will remain – relatively poorly described, at least for the purposes of consistent cross-linguistic research. Most lexical-typological research is therefore in need of constantly inventing, testing and elaborating its methods of data collection. Even more, the different methods are not easily transmittable among different research areas, even those that ask comparable questions. For instance, visual stimuli for eliciting words referring to cutting and breaking events can certainly serve as a model for research on some other conceptual domains involving dynamic situations with clearly visible actions and results (say, dressing / undressing, or putting). But already moving to domains based on other perceptual modalities is far from trivial – sounds, temperature, taste are still awaiting good data collection techniques and guidelines – whereas the emotional and mental world is even harder to cover with perceptual stimuli (cf., however, Pavlenko 2002 for a comparison of emotional descriptions in Russian and English narratives elicited through the same short film). The extent to which parallel texts can be used in lexical-typological research is, of course, extremely dependent on the object of study and on the genre of the parallel texts and is best suited for frequent phenomena. Thus, while motion verbs frequently occur in the New Testament, and generic statements in the Universal Declaration of the Human Rights, which are the easiest available texts in many languages, these sources will be of restricted value for the study of pain expressions, even though such examples do occasionally occur in the former. However, for a more limited number of languages other parallel texts have been successfully used in lexical-typological work (“Le petit prince” and “Harry Potter” belong to the favourites here). Finally, word lists, as we have seen, may well be used for some purposes (e.g., for checking the word-class categorization of “property” words or the aktionsart categorization of verbs, etc.), but are of marginal value when too little is known about the lexical meaning of phenomena under consideration or when the phenomena involve too many language-specific lexical idiosyncrasies. Consider pluralia tantum, e.g., nouns that only occur in the plural form, like scissors, and are very unevenly distributed across languages. In Koptjevskaja-Tamm & Wälchli (2001) we used two principled samples of lexemes that are encoded by pluralia-tantum nouns in Lithuanian vs. Russian for collecting comparable data across forty European languages. Since the distribution of pluralia tantum in a language is highly idiosyncratic, we hypothesized that

.  Last visited on April 26th 2007. .  Last visited on April 26th 2007.



Approaching lexical typology 

the degree of overlapping in the distribution of pluralia tantum across languages could be used as a measure for their contacts, in this case, the languages in the Northeastern part of Europe. While the two samples turned out to be useful for this particular end, the same fact (lexical idiosyncracies of pluralia tantum) causes difficulties for crosslinguistic studies of pluralia tantum in general. Although they do often occur in comparable domains (e.g., heterogeneous substances, like leftovers, diseases, like measles, festivities, like Weihnachten “Christmas” in German), they are very language-specific when it comes to the lexical meanings, which rules out the use of a consistent word list for cross-linguistic data collection. In my opinion, successful lexical-typological research should in most cases build on a collaborative work involving language experts (and, possibly, other specialists as well). The work on lexical universals within the NSM tradition (Goddard & Wierzbicka 1994), the different domain-categorization studies co-ordinated from the Max-Planck Institute in Nijmegen (Levinson & Meira 2003; Majid et al. eds. 2006; Majid & Bowerman eds. 2007), the project on aquamotion verbs directed by Moscow linguists (Maisak & Rakhilina 2007) are all examples of excellent semantic-typological research based on the methodology that had been elaborated, tested and improved by the group of language experts, who have further collected and analyzed the data in close collaboration with native speakers. The issue of data collection is, of course, intimately related to the issue of cross-linguistic identification of studied phenomena, which is a key concern for cross-linguistic and typological research in general. We have to be sure that we compare like with like, rather than apples with pears. However, another key concern for cross-linguistic and typological research is to find a reasonable level of abstraction, at which the richness of language-specific details can be reduced to manageable patterns. The two concerns interact in various ways; most importantly, what counts as “like and like” is often dependent on the research object and goal. In the course of this paper we have had several occasions to discuss the different levels of semantic precision appropriate for different types of lexical-typological research (e.g., domain-categorization vs. semantic motivation vs. lexicon-grammar interaction). It should be mentioned here that the grammatical typology on the whole hardly ever cares about precise semantics: the only prerequisite is that we can roughly identify linguistic phenomena across languages via certain conditions that they have to meet, e.g., via a certain function that has to be expressed by a construction. Thus, for instance, an possessive NP is recognized by its ability to refer to legal ownership (Peter’s bag), to kin relations (Peter’s son) or to relations between a person and his body-parts (Peter’s leg). The fact that the same construction in English can occasionally refer to temporal and local relations (yesterday’s magazine, London’s museums), whereas many other languages are much more restrictive in this respect is of marginal interest for the cross-linguistic identification of possessive NPs themselves. There are, of course, certain limits to the semantic

 Maria Koptjevskaja-Tamm

vagueness that can underlie systematic cross-linguistic identification of phenomena. I find it difficult to set up good methods for testing universality of some suggested metaphors cross-linguistically, like, for instance, anger is heat (Kövecses 1996). What can probably be done is to test some of its specific manifestations, e.g., whether the words for anger (and other emotions) can be described by temperature terms. Finally, in order to achieve a reasonable consensus on the meta-language used for semantic explications and on the ways of representing meanings is an urgent need – both in theoretical semantics, in semantic and lexical typology and in lexicography. Let’s hope that the contributions in this volume will be good points of departure for numerous future projects in lexical typology.

Acknowledgements This paper has a long prehistory. Parts of it have been presented in different versions at the Annual Meeting of the Finnish and Estonian Linguistic Society in Tallinn (May 2004), at the conference “Lexicon in linguistic theory” (Åbo/Turku, November 2004), at the workshop ”Semantic parallels in a cross-linguistic perspective” (Paris, December 2004), in lectures at the universities of Stockholm, Gothenburg, Lund, Pavia, Turin and Tübingen. I am grateful to the audiences for their questions and comments. I would also like to thank several friends and colleagues who have provided intellectual and humane support, inspiration and feedback in my work on the paper: Melissa Bowerman, Grev Corbett, Östen Dahl, Nick Enfield, Nick Evans, Giannoula Giannoulopoulou, Cliff Goddard, Olesya Khanina, Seppo Kittilä, Peter Koch, Christian Lehmann, Eva Lindström, Galina Paramei, Ekaterina Rakhilina, Edith Moravcsik, Carita Paradis, Frans Plank, Farzad Sharifian, Jae Jung Song, Martin Tamm, Martine Vanhove, Ljuba Veselinova, Bernhard Wälchli. All the faults remain mine. There is a certain overlapping between this paper and Koptjevskaja-Tamm et al. (2007). This concerns primarily parts of Section 6 and Section 7.

References Aikhenvald, A. & Dixon, R.M.W. (Eds). 2002. Word: A Cross-linguistic Typology. Cambridge: CUP. Ameka, F. & Levison, S. (Eds). 2007. Locative predicates. Linguistics, special issue 45(5–6). Andersen, E. 1978. Lexical universals of body-part terminology. In Universals of Human Language, J. Greenberg (Ed.), 335–368. Stanford CAC Stanford University Press. Auwera, J. van der. 1998. Phrasal adverbials in the languages of Europe. In Adverbial Constructions in the Languages of Europe, J. van der Auwera (Ed.), 25–145. Berlin: Mouton de Gruyter. Auwera, J. van der. 2001. On the lexical typology of modals, quantifiers, and connectives. In Perspectives on Semantics, Pragmatics, and Discourse: A Festschrift for Ferenc Kiefer, I. Kenesei & R.M. Harnish (Eds), 173–186. Amsterdam: John Benjamins.



Approaching lexical typology 

Auwera, J., van der, Kehayov, P. & Vittrant, A. Forthcoming. Modality’s semantic map revisited: Acquisitive modals. In Cross-linguistic Studies of Tense, Aspect, and Modality, L. Hogeweg, H. De Hoop & A. Malchukov (Eds). Amsterdam: John Benjamins. Bach, E., Jelinek, E., Kratzer, A. & Partee, B.H. (Eds). 1995. Quantification in Natural Languages. Dordrecht: Kluwer. Behrens, L. 1995. Categorizing between lexicon and grammar. The Mass/Count distinction in a cross-linguistic perspective. Lexicology 1(1): 1–112. Behrens, L. & Sasse, H.J. 1997. Lexical Typology: A Programmatic Sketch [Arbeitspapier Nr. 30, Neue Folge]. Cologne: University of Cologne, Department of linguistics. Berlin, B. 1992. Ethnobiological Classification. Princeton NJ: Princeton University Press. Berlin, B. & Kay, P. 1969. Basic Color Terms: Their Universality and Evolution. Berkeley CA: University of California Press. Berman, R.A. & Slobin, D.I., in collaboration with Aksu-Koç, A. 1994. Relating Events in Narrative: A Crosslinguistic Developmental Study. Hillsdale NJ: Lawrence Erlbaum Associates. Blank, A. 2001. Pathways of lexicalization. In Haspelmath et al. (Eds), 1596–1608. Bossong, G. 1998. Le marquage de l’expérient dans les langues d’Europe. In Actance et valence dans les langues de l’Europe, J. Feuillet (Ed.), 259–294. Berlin: Mouton de Gruyter. Botne, R. 2003. To die across languages. Linguistic Typology 7(2): 233–278. Bowden J. 1992. Behind the Preposition: the grammaticalisation of locatives in Oceanic languages. Canberra: Pacific Linguistics. Brown, C.H. 1976. General principles of human anatomical partonomy and speculations on the growth of partonomic nomenclature. American Ethnologist 3: 400–424. Brown, C.H. 1999. Lexical Acculturation in Native American Languages. Oxford: OUP. Brown, C.H. 2001. Lexical typology from an anthropological point of view. In Haspelmath et al. (Eds), 1178–1190. Brown, C.H. 2005a. Hand and arm. In Haspelmath et al. (Eds), 522–525. Brown, C.H. 2005b. Finger and hand. In Haspelmath et al. (Eds), 526–529. Chappell, H. & McGregor, W. (Eds). 1996. The Grammar of Inalienability. A Typological Perspective on Body Part Terms and the Part-whole Relation. Berlin: Mouton de Gruyter. Corbett, G.G. 2007. Canonical typology, suppletion and possible words. Language 83(1): 8–42. Corbett G.G. 2000. Number [Cambridge Textbooks in Linguistics]. Cambridge: CUP. Corbett G.G. Forthcoming. Higher order exceptionality in inflectional morphology. In Expecting the Unexpected: Exceptions in Grammar, H.J. Simon & H. Wiese (Eds). Berlin: Mouton de Gruyter. Cristofaro, S. 2003. Complementation. Oxford: OUP. Croft, W. 1990[2003]. Typology and Universals [Cambridge Textbooks in Linguistics], 1st-2nd Edn. Cambridge: CUP. Croft, W. 1991. Syntactic Categories and Grammatical Relations. Chicago IL: Chicago University Press. Cruse, A. 1986. Lexical Semantics. Cambridge: CUP. Cruse, A., Hundsnurscher, F., Job, M. & Lutzeier, P.R. (Eds). 2002[2005]. Lexicology. An International Handbook on the Nature and Structure of Words and Vocabularies, Vols. 1–2. Berlin: Walter de Gruyter. Cysouw, M. 2004. Interrogative words: An exercise in lexical typology. Ms, (http://email.eva. mpg.de/~cysouw/presentations.html) Dahl, Ö. & Koptjevskaja-Tamm, M. 2001. Kinship in grammar. In Dimensions of Possession, I. Baron, M. Herslund & F. Sørensen (Eds), 201–225. Amsterdam: John Benjamins.

 Maria Koptjevskaja-Tamm Dedrick, D. 2005. Color, color terms, categorization, cognition, culture: An afterthought. Journal of cognition and culture 5(3–4): 487–495. Dixon, R.M.W. 1977. Where have all the adjectives gone? Studies in Language 1: 19–80. Dixon, R.M.W. & Aikhenvald, A. (Eds). 2004. Adjective Classes. Oxford: OUP. Enfield, N.J. 2003. Linguistic epidemiology. London & New York: Routledge/Curzon. Enfield, N.J. 2002. Semantic analysis of body parts in emotion terminology: Avoiding the exoticisms of “obstinate monosemy” and “online extension”. Pragmatics and Cognition 10(1–2): 81–102. Enfield, N.J. 2007. Lao separation verbs and the logic of linguistic event categorization. In Majid & Bowerman (Eds), 287–286. Enfield, N.J., Majid, A. & van Staden, M. 2006. Cross-linguistic categorisation of the body: Introduction. In Majid et al. (Eds), 134–147. Enfield, N.J. & Wierzbicka, A. (Eds). 2002. The body in description of emotion. Pragmatics and cognition 10(1–2) (special issue). Evans, N. 2001. Kinship terminology. In International Encyclopedia of Social and Behavioral Sciences, N.J. Smelser & P. Baltes (Eds), 8105–8111. Amsterdam: Elsevier Science. Evans, N. 1992. Multiple semiotic systems, hyperpolysemy, and the reconstruction of semantic change in Australian languages. In Diachrony within Synchrony, G. Kellerman & M. Morrissey, (Eds), 475–508. Bern: Peter Lang. Evans, N. Forthcoming. Semantic typology. In The Oxford Handbook of Typology, J.J. Sung (Ed.), Oxford: OUP. Evans, N. & Sasse, H.J. 2007. Searching for meaning in the Library of Babel: Field semantics and problems of digital archiving. Archives and Social Studies: A Journal of Interdisciplinary Research 1. (http://socialstudies.cartagena.es/index.php?option=com_content&task=view &id=53&Itemid=42) Evans, N. & Wilkins, D.P. 2000. In the mind’s ear: The semantic extensions of perception verbs in Australian languages. Language 76: 546–592. Foley, W. 1997. Anthropological Linguistics. An Introduction. Malden MA: Blackwell. Gil, D. 2001. Quantifiers. In Haspelmath et al. (Eds), Vol. 2, 1275–1294. Goddard, C. 2001. Lexico-semantic universals: A critical overview. Linguistic Typology 5: 1–65. Goddard, C. & Wierzbicka, A. (Eds). 1994. Semantic and Lexical Universals. Amsterdam: John Benjamins. Grandi, N. 2002. Morfologie in contatto. Le costruzioni valutative nelle lingue del Mediterraneo. Milano: F. Angeli. Greenberg, J.H. 1966. Language Universals [Janua Linguarum, Series Minor 59]. The Hague: Mouton. Greenberg, J.H. 1980. Universals of kinship terminology. In On Linguistic Anthropology: Essays in Honor of Harry Hoijer, J. Maquet (Ed.), 9–32. Malibu CA: Udena Publications. Harkins, J. & Wierzbicka, A. (Eds). 2001. Emotions in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Haspelmath, M. 1987. Transitivity Alternations of the Anticausative Type [Arbeitspapier Nr. 5, Neue Folge]. Cologne: University of Cologne. Haspelmath, M. 1993. Inchoative/causative verb alternations. In Causatives and Transitivity, B. Comrie & M. Polinsky (Eds), 87–111. Amsterdam: John Benjamins. Haspelmath, M. 1997. Indefinite Pronouns. Oxford: OUP. Haspelmath, M. 2003. The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison. In The New Psychology of Language, M. Tomasello (Ed.), Vol. 2, 211–242. Mahwah NJ: Lawrence Erlbaum Associates.



Approaching lexical typology 

Haspelmath, M. 2005a. Ditransitive constructions: The verb ‘give’. In Haspelmath et al., 426–429. Haspelmath, M. 2005b. ‘Want’ complement clauses. In Haspelmath et al., 502–505. Haspelmath, M., Dryer, M, Gil, D. & Comrie, B. 2005. The World Atlas of Language Structures (WALS). Oxford: OUP. Haspelmath, M., König, E., Oesterreicher, W. & Raible, W. (Eds). 2001. Language Typology and Language Universals, Vol.1–2. Berlin: Walter de Gruyter. Haspelmath, M. & Tadmor, U. (Eds) In preparation. Lexical Borrowability. 2 Vols. Berlin: Mouton de Gruyter. Heine, B. 1989. Adpositions in African languages. Linguistique Africaine 2: 77–127. Heine, B. 1997. Cognitive Foundations of Grammar. Oxford: OUP. Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Iordanskaja, L. & Paperno, S. 1996. A Russian-English Collocational Dictionary of the Human Body. Columbus OH: Slavica Publishers. (Also http://LexiconBrige.Com/for the electronic edition) Jameson, K. 2005. Culture and cognition: What is universal about the representation of color experience? Journal of Cognition and Culture 5(3–4): 294–347. Juravsky, D. 1996. Universal tendencies in the semantics of the diminutive. Language 72(3): 533–578. Kay, P. & Maffi, L. 2000. Color appearance and the emergence and evolution of basic color naming. Journal of linguistic anthropology 1: 12–25. Kay, P. & Maffi, L. 2005. Colour terms. In Haspelmath et al. (Eds), 534–545. Kay, P. & McDaniel, C. 1978. The linguistic significance of the meanings of basic color terms. Language 54: 610–46. Kemmer, S. 1993. The Middle Voice. Amsterdam: John Benjamins. Khanina, O. 2005. Jazykovoe oformlenie situacii želanija: opyt tipologičeskogo issledovanija [Linguistic Encoding of ‘Wanting’: A Typological Approach]. Ph.D. dissertation, Moscow State University. Khanina, O. Forthcoming. How universal is wanting? Studies in Language. Kibrik, A.A. 2003. Lexical semantics as a key to typological peculiarities. Paper at the Workshop on Lexical/Semantic Typology at the ALT conference in Cagliari, September 2003. Kita, S. & Özyürek, A. 2003. What does cross-linguistic variation in semantic coordination of speech and gesture reveal? Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language 48: 16–32. Kittilä, S. 2006. Anomaly of the verb ‘give’ explained by its high (formal and semantic) transitivity. Linguistics 44(3): 569–609. Koch, P. 2001. Lexical typology from a cognitive and linguistic point of view. In Haspelmath et al. (Eds), Vol. 2, 1142–1178. Koch, P. & Marzo, D. 2007. A two-dimensional approach to the study of motivation and lexical typology and its first application to French high-frequency vocabulary. Studies in Language 31(2): 259–291. Koptjevskaja-Tamm, M. 2004. Mass and collection. In Morphology: A Handbook on Inflection and Word Formation, G. Booij, C. Lehmann & J. Mugdan (Eds), Vol. 2, 1016–1031. Berlin: Walter de Gruyter. Koptjevskaja-Tamm, M., Divjak, D. & Rakhilina, E. Forthcoming. Aquamotion verbs in Slavic and Germanic: A case study in lexical typology. In Multiple Perspectives on Slavic Verbs of Motion, V. Dragina & R. Perelmutter (Eds). Amsterdam: John Benjamins. Koptjevskaja-Tamm, M. & Rakhilina, E. 2006. “Some like it hot”: On semantics of temperature adjectives in Russian and Swedish. STUF (Sprachtypologie und Universalienforschung), a

 Maria Koptjevskaja-Tamm special issue on Lexicon in a Typological and Contrastive Perspective, G. Giannoulopoulou & T. Leuschner (Eds) 59(2): 253–269. Koptjevskaja-Tamm, M., Vanhove, M. & Koch, P. 2007. Typological approaches to lexical semantics. Linguistic Typology 11(1): 159–186. Koptjevskaja-Tamm, M. & Wälchli, B. 2001. The Circum-Baltic languages: An areal-typological approach. In Circum-Baltic Languages: Their Typology and Contacts, Ö. Dahl & M. Koptjevskaja-Tamm (Eds), Vol. 2, 615–750. Amsterdam: John Benjamins. Krifka, M. (Ed.). 2003. Natural Semantic Metalanguage. A special issue of Theoretical Linguistics 29(3). Berlin: Mouton de Gruyter. Kövecses, Z. 1996. Anger: Its language, conceptualization, and physiology in the light of crosscultural evidence. In Language and the Cognitive Construal of the World, J. Taylor & R. MacLaury (Eds), 181–196. Berlin: Mouton de Gruyter. Lang, E. 2001. Spatial dimension terms. In Haspelmath et al. (Eds), 1251–1275. Lehmann, C. 1990. Towards lexical typology. In Studies in Typology and Diachrony. Papers Presented to Joseph H. Greenberg on his 75th Birthday, W. Croft, K. Denning & S. Kemmer (Eds), 161–185. Amsterdam: John Benjamins. Lehrer, A. 1992. A theory of vocabulary structure: Retrospectives and prospectives. In Thirty Years of Linguistic Evolution. Studies in Honour of René Dirvén on the Occasion of his Sixtieth Birthday, M. Pütz (Ed.), 243–256. Amsterdam: John Benjamins. Levinson, S. 2006. Parts of the body in Yélî Dnye, the Papuan language of Rossel Island. In Parts of the Body: Cross-linguistic Categorization, N. Enfield, A. Majid & M. van Staden (Eds), special issue of Language Sciences 28: 221–240. Levinson, S. & Meira, S. 2003. “Natural concepts” in the spatial topological domain – adpositional meanings in cross-linguistic perspective: An exercise in semantic typology. Language 79(3): 485–516. Lichtenberk, F. 1991. Semantic change and heterosemy in grammaticalization. Language 67(3): 475–509. Lindström, E. 2002. The body in expressions of emotion: Kuot. In Enfield, & Wierzbicka (Eds), 159–184. Lucy, J.A. 1992. Grammatical Categories and Cognition: A Case Study of the Linguistic Relativity Hypothesis. Cambridge: CUP. MacLaury, R.E. 1997. Color and Cognition in Mesoamerica: Constructing Categories as Vantages. Austin TX: University of Texas Press. MacLaury, R.E. 2001. Color terms. In Haspelmath et al. (Eds), 1227–1251. Maisak, T. & Rakhilina, E. (Eds). 2007. Glagoly dviženija v vode: leksičeskaja tipologija (Aquamotion Verbs: Lexical Typology). Moskva: Indrik. Majid, A., Enfield, N.J., & van Staden, M. (Eds). 2006. Parts of the body: Cross-linguistic categorization. Language Sciences 28(2–3): 137–360 (special issue). Majid, A. & Bowerman, M. (Eds). 2007. Cutting and breaking events: A cross-linguistic perspective. Cognitive Linguistics 18(2): 153–177 (special issue). Majid, A., Bowerman, M., van Staden, M. & Boster, J.S. (Eds). 2007. The semantic categories of ‘cutting and breaking’ events across languages. Cognitive Linguistics 18(2): 133–152. Maslova, E. 2004. A universal constraint on the sensory lexicon, or when hear can mean ‘see’? In Tipologičeskie obosnovanija v grammatike: k 70-letiju professora Xrakovskogo V.S. (Typological Foundations in Grammar: For Professor Xrakovskij on the Occasion of his 70th Birthday), A.P. Volodin (Ed.), 300–312. Moscow: Znak. (Also http://www.stanford. edu/~emaslova/Publications/PubFrame.html)



Approaching lexical typology 

Matisoff, J. 1986. Hearts and minds in Southeast Asian languages and English: An essay in the comparative lexical semantics of psycho-collocations. Cahiers de linguistique Asie Orientale 15(1): 5–57. Merlan, F. 1985. Split intransitivity: Functional oppositions in intransitive inflection. In Grammar Inside and Outside the Clause: Some Approaches to the Theory from the Field, J. Nichols & A. Woodbury (Eds), 324–362. Cambridge: CUP. Morgan, L.H. 1870. Systems of Consanguinity and Affinity. (Reprinted in 1997, Lincoln NB: University of Nebraska Press). Nerlove, S. & Romney, A.K. 1967. Sibling terminology and cross-sex behavior. American Anthropologist 74: 1249–1253. Newman J. 2002. The Linguistics of Sitting, Standing and Lying. Amsterdam: John Benjamins. Nichols, J., Peterson, D.A. & Barnes, J. 2004. Transitivizing and detransitivizing languages. Linguistic Typology 8: 149–211. Ojutkangas, K. 2000. Grammatical possessive constructions in Finnic: Käsi ‘hand’ in Estonian and Finnish. In Facing Finnic. Some Challenges to Historical and Contact Linguistics [Castrenianumin toimitteita 59], J. Laakso (Ed.), 137–155. Helsinki: Finno-Ugrian Society – Dept. of Finno-Ugrian Studies of the University of Helsinki. Pavlenko, G. 2002. Emotions and the body in Russian and English. In Enfield & Wierzbicka (Eds), 207–241. Paramei, G. 2005. Singing the Russian blues: An argument for culturally basic color terms. Cross-cultural research 39(1): 10–34. Payne, D. 2006. Color terms. In Encyclopedia of Languages and Linguistics, 601–605. Oxford: Elsevier. Pustet, R. 2003. Copulas. Universals in the Categorization of the Lexicon. Oxford: OUP. Quine, W.V. 1960. Word and Object. Cambridge MA: The MIT Press. Rakhilina, E. 2006. Glagoly plavanija v russkom jazyke (Aquamotion verbs in Russian). In Maisak & Rakhilina (Eds), 267–285. Riemer, N. 2005. The Semantics of Polysemy. Berlin: Mouton de Gruyter. Ricca, D. 1993. I verbi deittici di movimento in Europa: Una ricerca interlinguistica. Firenze: La Nuova Italia Editrice. Rijkhoff, J. 2002. The Noun Phrase [Oxford Studies in Typology and Linguistic Theory] Oxford: OUP. Sasse, H.J. 2002. Recent activity in the theory of aspect: Accomplishments, achievements, or just non-progressive state? Linguistic Typology 6(2): 199–271. Schladt M. 2000. The typology and grammaticalization of reflexives. In Reflexives: Forms and Functions, Z. Frajzyngier (Ed.), 103–124. Amsterdam: John Benjamins. Sharifian, F., Dirven, R., Yu, N. & Neiemier, S. (Eds). Forthcoming. Culture, Body, and Language: Conceptualizations of Internal Body Organs Across Cultures and Languages. Berlin: Mouton De Gruyter. Slobin, D.I. 2003. Language and thought online: Cognitive consequences of linguistic relativity. In Language in Mind: Advances in the Study of Language and Thought, D. Gentner & S. Goldin-Meadow (Eds), 157–191. Cambridge MA: The MIT Press. Slobin, D.I. & Bowerman, M. 2007. Interfaces between linguistic typology and child language research. Linguistic Typology 11: 213–226. Slobin, D.I & Hoiting, N. 1994. Reference to movement in spoken and signed languages: Typological considerations. BLS 20: 487–505. Stassen, L. 1997. Intransitive Predication. Oxford: OUP.

 Maria Koptjevskaja-Tamm Svorou, S. 1993. The Grammar of Space. Amsterdam: John Benjamins. Sweetser, E. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: CUP. Talmy, L. 1991. Path to realization – Via aspect and result. A typology of event conflation. BLS 17: 480–519. Talmy, L. 1985. Lexicalization patterns. In Language Typology and Synchrohic Description, T. Shopen (Ed.), Vol. 3, 57–149. Cambridge: CUP. Tatevosov, S. 2002. The parameter of actionality. Linguistic Typology 6(3): 317–401. Terrill, A. 2006. Body-part terms in Lavukaleve, a Papua language of the Solomon Islands. In Majid et al. (Eds), 304–322. Vendler, Z. 1967. Linguistics in Philosophy. Ithaca NY: Cornell University Press. Veselinova, L. 2006. Suppletion in Verb Paradigms. Amsterdam: John Benjamins. Viberg, Å. 1984. The verbs of perception: a typological study. Linguistics 21: 123–162. Viberg, Å. 2001. Verbs of perception. In Haspelmath et al. (Eds), Vol. 2, 1294–1309. Viberg, Å. 2002. Polysemy and disambiguation cues across languages. The case of Swedish få and English get. In Lexis in Contrast, B. Altenberg & S. Granger (Eds), 119–150. Amsterdam: John Benjamins. Viberg, Å. 2005. The lexical typological profile of Swedish mental verbs. Languages in contrast 5(1): 121–157. Viberg, Å. 2006. Towards a lexical profile of the Swedish verb lexicon. Sprachtypologie und Universalienforschung (STUF) 59(1): 103–129. (special issue on The Typological Profile of Swedish, Å. Viberg (Ed.)). Wälchli, B. 2005. Co-compounds and Natural Coordination. Oxford: OUP. Wälchli, B. 2006. Lexicalization patterns in motion events revisited. Ms. (http://ling.unikonstanz.de/pages/home/a20_11/waelchli/waelchli-lexpatt.pdf) Wälchli, B. 2006/2007. Constructing semantic maps from parallel text data. Ms. (http://ling.unikonstanz.de/pages/home/a20_11/waelchli/waelchli-semmaps.pdf) Wälchli, B. & Zúñiga, F. 2006. The feature of systematic source-goal distinction and a typology of motion events in the clause. Sprachtypologie und Universalienforschung (STUF) 59(3): 284–303. (special issue on The Lexicon: Typological and Contrastive Perspectives, T. Leuschner & G. Giannoulopoulou (Eds)). Wegener, C. 2006. Savosavo body part terminology. In Majid et al. (Eds), 344–359. Wierzbicka, A. 1988. The Semantics of Grammar. Amsterdam: John Benjamins. Wierzbicka, A. 1990. The meaning of color terms: Semantics, culture, and cognition. Cognitive linguistics 1: 99–150. Wierzbicka, A. 2007. Bodies and their parts: An NSM approach to semantic typology. Language Sciences 29: 14–65. Wilkins, D.P. & Hill, D. 1995. When GO means COME: Questioning the basicness of basic motion verbs. Cognitive Linguistics 6(2–3): 209–259.

part ii

Theoretical and methodological issues

Words and their meanings Principles of variation and stabilization Stéphane Robert

Llacan (Inalco, CNRS), Fédération TUL This chapter, entrenched in cognitive linguistics, proposes a multidimensional approach to the layering of the lexicon and its semantic organization, explicating the principles of variation and stabilization of lexical networks. Semantic variation is considered as inherent to language structure and driven by common universal cognitive mechanisms which are accounted for by a dynamic conception of meaning construal. Intra-linguistic plasticity of meaning echoes inter-linguistic variation. The discourse level is the seat of meaning construal mechanisms which contribute to the general polysemy of lexical units and to the stabilization of their meaning within a particular utterance. Units appear to be the seat of most variations, within and across languages, because meaning is construed in extremely varied ways according to common mechanisms. Keywords:  comprehension; discourse; meaning; polysemy

1.  Introduction* For both structural and cognitive reasons, natural languages are characterized by their plasticity, by the ease with which the representations borne by the units composing them are subject to change. Polysemy and polyreference are the general rule among languages. A single unit can thus have several different meanings and point to several different referents. In English for example the word greens can refer to village commons, leafy vegetables or members of a political party. Inversely, different units can refer to the same thing, such as roe and caviar, or hepatitis and jaundice. One could even state that local synonymy (limited to a certain context) is what makes it possible to paraphrase a term or phrase using another. Thus reflect can be paraphrased by either “think” or “throw back light”. The ability to build equivalences is in fact a fundamental

*Our deep thanks go to Margaret Dunham for her precious help in translating and accommodating this paper to English.

 Stéphane Robert

property of language: equivalences between terms (synonymy) or between phrases (paraphrasing), but also between languages (translation). There is no one-to-one relation between form and meaning, either within a language or across languages. From this view point, variation within languages (polysemy, synonymy), echoes variation from language to language and raises the question of how it is possible to say “the same thing” differently. Whereas this plasticity in meaning ensures both the referential power of a language and a form of optimization for the system, it also entails another of language’s defining characteristics – ambiguity and its communicative corollary: misunderstandings. That communication remains nonetheless possible is because the factors of variation in language are submitted to processes of regulation and meaning stabilization. I will begin by attempting to highlight a certain number of variation factors at the level of the isolated units, then I will try to show that in language activity, virtual units undergo certain operations whereby they are incorporated into utterances, and to highlight this different operations of the sentence level, which permit a certain stabilization in meaning but also occasion communicative failures. We will take this opportunity to also question the causes of these language characteristics and possible consequences from a cognitive viewpoint. For language is the seat of tensions between opposing forces which can all be functionally justified.

2.  Language malleability and variation at the unit level There is no one-to-one correspondence between form and meaning in language: a form almost always has several meanings which vary according to context, and several forms can refer to a same item. This plasticity constitutes one of language’s fundamental principals. It is made manifest in different ways but follows consistent procedures. Units appear to be the seat of the most variation, within and across languages, not only because meaning is construed in extremely varied ways (categorizing and segmenting the world, selecting properties and reference pathways, cf. 2.1.), and because linguistic units are subjected to regular meaning changes (cf. 2.2.), but also because words contain what I have called a “depth dimension” (see below 2.3.), also extremely variable. Units thus show variable specificities depending on the language and on the culture, which most probably plays an important role in cognition’s access to reference. We will limit ourselves here to lexical units, but grammatical units also undergo regulated variation.

2.1  Different means of reference accessing (on synonymy) Linguistic reference is always mediated. Firstly because words are not things, they are substitutes for the reality they designate (independently of the nature of the reality),



Words and their meanings: Principles of variation and stabilization 

or more precisely they are the representatives of representations (Culioli 1990:  22). However, this reality does not constitute a pre-segmented, stable, given, for which words would be but the labels. Indeed, reality is presented to perception as a continuum, whereas language is composed of discrete units. Therefore it must segment the perceived or conceived reality, in order to build the referential values of its units, and this segmentation varies from language to language. Although traces of iconicity in language exist (i.e., resemblances between form and meaning), generally speaking, the relation between a form and its referent is arbitrary, which also contributes to inter-linguistic variation. This arbitrary character is moreover what makes languages so powerful: if words necessarily resembled the objects they designated, languages, which make sparing use of phonetic means, would be extremely limited. Thus the signifier (the form), is variable, and applies to meanings which vary from language to language.

2.1.1  Variable categorization, segmentation and construals To illustrate the variable segmentation carried out by languages, I will take examples from two domains which could a priori appear as the most constrained by physical and perceptive data, and thus the most stable: body parts and spatial reference. Despite the fact that the data is shared, the body is “segmented” into different referential units depending on the language. The word leg in English designates, following the referential scale,1 either the whole of the lower member, or the part below the knee, whereas in Wolof, tànk, in its wider sense, refers to the part below the knee, and in its narrower sense, to the foot. Thus the segmentation differs between the two languages. Some languages contain terms which refer to body parts that do not exist in other languages, so the body cannot be considered a specific language unit. Contrary to French and English, Ibo (a Kwa language spoken in Nigeria) and Langi (a Bantu language spoken in Tanzania) do not lexically differentiate arm from hand. Moreover, language can view body parts in relation to each other or in relation to outside elements in various ways. Mandarin Chinese establishes a link between the terms leg, thigh and foot, as there is a common term for the three: tuǐ. But Chinese can also specify whether the “leg” is a “small tuǐ ”, xiǎo tuǐ, or to refer to the thigh as “big tuǐ ”, dà tuǐ. Contrary to English, French establishes a link between the “fingers of the foot” doigts de pied, and the “fingers of the hand” doigts de la main (on body parts, see Maria Koptjevskaja-Tamm’s article, in this volume and Brown 2005a and 2005b). Sanskrit uses the same term, pradeśinī, to refer to the forefinger or the corresponding toe. Thus our body, which is the same for everyone physically speaking, is not conceived of in the same way by all languages. Furthermore, these differences in segmentation affect the grammar.

.  On scale of predication, see Langacker (1991b: 283).

 Stéphane Robert

Certain African languages for example classify objects according to whether they occur in pairs (hands, eyes, feet…), and these objects are grouped into one category (“class”), with a specific grammatical agreement. In many languages, the fact that body parts are inalienable possessions triggers specific syntactic constructions. Thus in French one says je lave ma voiture (lit. “I am washing my car”) but je me lave les mains (lit. “I am washing me/myself the hands”). Spanish includes clothing in constructions for inalienable objects, contrary to French. Concerning spatial orientation, languages show three major reference systems: an absolute reference system, like the cardinal points; an anthropomorphic reference system such as right and left which are defined with respect to the observer, and a relative (or intrinsic) reference system which takes one object as a reference point for locating another (“on the roof ”, “near the house”…). Languages generally use all three systems but apply them along varying scales. Thus in French, to situate a building, one tends to use the relative/intrinsic referential system (“the post office is on the corner” or “beside the town hall”) whereas in the US one tends more to use absolute references (“it is north of the campus”), which French usually reserves for a larger scale, to locate one city in reference to another (“Amiens is north of Paris”). One could be tempted to think that these reference systems are universal, but such is not the case. Some languages, such as Malagasy and most Austronesian languages, use only one system, namely absolute reference, independently of scale. In these languages, one never says “the book is on your right”, but “the book is to the north (or south) of the table” (cf. Ozanne-Rivierre 1999). Lastly, the cardinal point system is also variable: some American Indian languages have not four cardinal points but six, as they also include the zenith and the nadir as spatial references. Thus, even though the physical properties of the world allow one to make certain predictions as to linguistic categories, one sees that these are not absolutes, because in language, everything is constructed, and therefore variable. Let us add one last example of the variable categorization in languages. Even a tool which may seem as fundamental as “yes/no” is not universal: certain languages (such as French and German) have a third term (si in French, doch in German), which serves to contradict a negative sentence; others, such as Latin and Chinese, have no words for “yes” or “no”. However, let us make it immediately clear that the fact that a concept has no corresponding linguistic category in a given language does not imply that its speakers cannot conceive of it or perceive it. Berlin and Kay (1969) show that speakers’ color perception does not depend on the (very variable) number of color names in their language. Just as languages which have no word for “yes” or “no” still have means for signifying agreement or contradiction, but using other processes, for example by repeating just the verb with or without negation (“eat”/“not eat”), or yet by using the verb “to be” (“that be yours?” answer: “be”). Let us mention in passing that in this way Chinese has several negation possibilities: the notional negation marker bú and the



Words and their meanings 

event negation marker méi (see also the Greek mè and ouk). Thus these languages use different linguistic categories for expressing these shared notions. Languages therefore show equivalency relations, although construals and reference constructions are extremely variable. Firstly because of the previously mentioned segmentation and linguistic categorization, but also because of a second fundamental mechanism. To gain access to a same referent, languages construct variable reference pathways.

2.1.2  Property selections and referential paths The meaning of a word is not limited to its referential value, i.e., the referent it designates. Languages usually choose one of the referent’s properties to designate it, for example a physical or functional characteristic. So, to come back to body parts, the index in French, or German, is the finger used for pointing (Zeigefinger), whereas in Greek it is the one which is used to lick (likhanós). In both cases, a different functional property is selected to designate the same referent. Access to reference therefore follows a different path in each language, a variable “referential path” (Corbin & Temple 1994).2 These examples show that the referential path chosen by a given language is both motivated (here linked to the functional properties of the referent) and therefore non-random, but since only one property is chosen, the choice is also arbitrary, or at least not strictly deterministic. Thus English designates a “used car” not by the fact that it is something one buys under favorable financial conditions, as in French (une voiture d’occasion), but rather by the fact that it was previously owned (or previously owned by only one other person as in a second-hand car). Therefore the property retained varies from language to language and probably refers to the trait considered the most salient for a given culture at a given point in time. But this does not imply that the meaning of the term is reduced to this one property: the referential path is just one means of reference accessing. The variability of referential paths across languages, as well as inside a given language, is due to a more general property of language, as claimed by cognitive linguistics, namely its ability to “construe” a particular situation in different ways (Langacker 1991a). The construction of different construals and variable referential paths to designate a referent explains the existence of synonyms within languages, such as voiture and automobile (“car”) in French. In the case of voiture (from the Latin vehere) the trait retained is that of being useful for “transporting” people or objects, whereas the trait retained for automobile is that of “being able to move on its own”. Just as record player and turn-table refer to the same thing, but after having followed different referential paths – the first term referring to the function and the second to the instrument’s

.  See also Langacker (1991b: 284) on compositional path.

 Stéphane Robert

mechanical apparatus (cf. Corbin & Temple 1994:  10 on électrophone and tournedisque in French). The fact that only one of a referent’s diverse properties is retained also explains the polyreference of certain terms. This is because very different objects can present a common property and thus be designated by a same term referring to that property. That greens can refer at once to expanses of grass, members of a political party and vegetables is due to the fact that English has chosen to designate these referents by a common property (the color green) which is considered salient and typical for each of the referents. In the same way, in French, the expression un bleu (“a blue one”) can designate a beginner, a new recruit, a work suit, a cheese, or a bruise, all of which have, in different ways, the common property of “being blue” (Corbin & Temple, ibid.). This economy in designation contributes to the referential power of words and makes it productive. One can easily imagine the language using the same term green to designate new referents presenting the same salient characteristic, as indeed it already does in compound nouns such as green-card. This process of constructing reference by selecting properties considered common to different referents is therefore the source of internal meaning variation phenomena. But the semantics of a term cannot be reduced to its referential value, it also encompasses dimensions other than the referential path, dimensions which are part of its meaning and also constitute variation factors.

2.1.3  The internal architecture of meaning and the referential background The manner in which the referent is designated also brings a complex semantic architecture into play. Designating an element generally entails the construction of a referential “background”. Thus the term hypotenuse usually designates the longest side of a right-triangle, which is opposite the right angle. The term refers to the side, designates it, but this designation only makes sense within the global representation implied by the right-triangle in the background (Langacker 1991a and b); the word tip refers to the extremity of an entity, but the meaning of the term takes into account the presence of the entity in the background (ibid). In the same way, concerning body-part terms, “essential to the characterization of expressions like head, arm, and leg is the position of the profiled entity relative to the body as a whole” (Langacker 1991b: 283). The same is true for the term uncle for example, which refers to a particular element within family relationships. The meaning of uncle encompasses both the designated element and the structure of parental relations that it is part of. It should be noted that the categorization of these parental relations varies from language to language: some languages, such as Wolof, distinguish between the maternal uncle and the other uncles and aunts; others, such as German and English, have a category which groups brothers and sisters together, independently of their gender (Geschwister “siblings”). The point which we find important here is that the meaning of a term is part of a hierarchical architecture, a sort of landscape which includes both a background,



Words and their meanings 

“ground” (in Talmy 1978), “fond” (in Vandeloise 1986) or “base” (in Langacker 1991b), and a salient sub-structure within the background, the “figure”, “cible” (“goal”) or “profile” (ibid). The “figure” represents the designated element and the “ground” the background into which the figure is inserted. The base and profile constitute two components of meaning, they do not have the same status but are linked in forming a term’s meaning. According to Langacker (ibid), the construal of a term’s meaning is an operation through which one profiles a sub-structure upon a base. The profiled element constitutes the referential value, it is part of the meaning, along with the base. Therefore there is an architecture of meaning, marked by a grounded structure. To gain access to a same referential value, languages may carry out profilings on different grounds.

2.2  Meaning’s malleability (polysemy and meaning shifts) Depending on the context, the meaning of a term varies. This variation is regulated by different mechanisms. There is always interaction between the terms present in the utterance (and between their respective properties). The association between one term and another, or even between a term and a given context, contributes in effect to the specification of its referential value. Thus a setting will not refer to the same thing depending on whether one is talking about a play or a ring. A tender steak is definitely not the same thing as a tender man. In both cases, the term’s application domain is different, which not only produces additional specifications but also “works” on the meaning of setting or tender which therefore are subject to deformation. We will come back to the modes of interaction a term has with its usage context (section 3). However, through these different values, the term presents a certain stability of meaning, manifested by the fact that the language considers it a single unit. Between a square foot and a square person there is both a shift in the adjective’s meaning, and semantic properties which are kept. The question then becomes to arrive at a description of the term’s unity, the nature of the relations between its different meanings as well as the mechanisms which produce the regulated variation.

2.2.1  Metaphor There are two well known major mechanisms which pilot these meaning shifts: metonymy and metaphor. These are not simply elements of rhetoric, but fundamental linguistic mechanisms which regulate the variation in the meaning of units.3 Metaphor is the transfer of properties from one domain to another to create a new referential value: some of a term’s semantic properties are selected (abstracted) and

.  I am speaking from the internal viewpoint of the meaning of units and not on the discourse level; the rhetoric of discourse distinguishes numerous figures of style for which an abundant literature exists, and which goes well beyond the scope of my paper here.

 Stéphane Robert

applied to another domain to designate a new entity in virtue of the properties considered shared by the two referents. For example, between a merchandise train and a train of thought, the word train does not have the same meaning, but the two meanings are linked together by a common semantic schema. Of the notion “train”, what is retained is the organization in successive units with identical function, linked to each other to form a complex unit. The shared properties are very abstract. They constitute a semantic schema that is present throughout all of the uses and which founds the semantic unity of the term. This is what Langacker (1991b) calls the “image schema”, Michaelis (1996) a “semantic super-structure” and Culioli a “schematic form” (Culioli 1990: 115–135). This schematic form can be applied to different domains that it will inform. In the case of train, for example, it is applied both to a vehicle (an element in space) and to a series of thoughts (elements in time): by switching application domains, the term switches referential values. The schematic form (or image-schema) is thus defined as a form which generates other forms, a sort of meaning-producing matrix. Similarly, the adjective square presents in its different uses an identical schema where an object takes on the shape of a square, where all sides are of equal length. When applied not to an object but to a person, having a certain shape but not being a geometrical shape, the meaning of the term shifts to the mental properties of the individual, conceived of as encompassing certain angles, certain boundaries, a certain rigidity. In general, one speaks of “metaphor” when the shift takes between one particular use (generally a concrete one), considered the primary meaning, and another (generally more abstract), through a process of selecting properties which are transferred from the primary domain to the other, which is probably the case for the two meanings of the adjective square, or yet for the temporal meaning of the verb to go, probably derived from its meaning as a verb of movement. However, it is not always possible to reconstruct the history of a word, nor to say exactly what the primary meaning was from which a schematic form was abstracted and then applied to another domain. It is probable that in certain cases the terms represent an abstract semantic schema from the start which, during a same period in the history of the language, is applied to different domains: there is no shift from a primary meaning to a metaphorical meaning, but from the beginning the word functions in various domains. This is the case of the word nú for example in Gbaya (Central African Republic) which designates the active part of an element, and can therefore refer to the tip of a pin, the edge of a field, the opening of a basket, embers of a fire, and language, conceived of as humans’ activity par excellence (Roulon-Doko 2003). Furthermore, metaphors can be dynamic (creative and perceived as transfers, as in the wings of desire) or fixed and lexicalized (waiting in the wings of a stage). In the variation mechanism we are attempting to describe, the different meanings of a term are linked together through a common semantic schema (schematic



Words and their meanings 

form or image-schema) which represents a set of shared abstract properties. The schematic form which serves as the foundation of the semantic unity of the term (train for example) is never bare, but is always instantiated in a particular domain and with a usage context which gives it its specific meaning (“train” in merchandise train or train of thought):

Metaphorical process: abstraction/ instanciation in a domain usage 1 meaning 1

semantic invariant: schematic form (does not appear)

usage 2 meaning 2

usage 3 meaning 3

Figure 1.  Schematic form and polysemous network (vertical relation).

Thus we would say that the semantic invariant (schematic form) represents the “signification” of the term and that its different usage values constitute its various “meanings”. Linguists present different models for organizing these meanings among themselves and in relation to the schematic form (see Kleiber 1999; Lakoff 1987). According to Langacker (1991b), the different meanings themselves are organized in a radial manner, with a more or less high degree of schematicity. Moreover, one of the meanings is often considered prototypical, i.e., it often appears as the best representative of the term’s values (for example the meaning “means of transportation” for the word train). One is thus faced with what I would call a “vertical” type of relation between the term’s different meanings. The relation in effect passes through a common relation to a schematic form which transcends all the meanings but never appears directly: to explain the shift from one meaning to another, one must go back to the schematic form which is at the base of the term’s semantic unity. From one usage to another, one does not find all of a term’s semantic properties,4 the properties specific to each use are linked to the term’s variable application domain (cf. 2.3. below) and to its particular properties (see the two meanings for pit in 4.1. below). Thus one sees that they are an important variation factor for a word’s meaning.

.  In certain cases the organization of the different meanings is more complex and combines metaphorical (vertical) relations and metonymical (horizontal) relations. See below 2.2.3.

 Stéphane Robert

Whether one calls it a metaphor or a schematic abstraction, the linguistic mechanism described here stems from a much more general and fundamental cognitive mechanism, that of analogy. Analogy rests on a homology between sometimes very different domains and on the perception of (abstract) properties seen as shared. From one meaning to the next, one finds both a common schema linked to these shared properties, and semantic properties specific to each usage, linked to the application domain. In metaphorical transfer, as shown by Lakoff & Johnson (1980), one transports a “form” but also inferences linked to the properties of the form.5 This is an important point in the case of scientific vocabularies which, far from escaping the metaphorical process, on the contrary have frequent recourse to it, notably because it makes it possible to take something known as a basis for describing and naming something unknown. The “milky way”, “electric current”, the “earth’s crust”, the “hammer” in the middle ear, “noise” in information theory are coded metaphors whose inferences are probably conscious and controlled because they are part of precise scientific models which strongly constrain their referential values and limit the transfer of inferences. Moreover one notes that it is the knowledge of the theoretical background (and therefore the term’s application domain) that stabilizes the meaning effects of these metaphors. When the context is unknown, as in pedagogical situations, inference transfers are probably very powerful and may lead to an important gap between the conventional meaning and the meaning construed by the public who very normally proceed by analogy. Thus, for example, the term black hole also rests on metaphor. It designates “cosmic objects so massive that they attract light rays, bend them in on themselves, prevent them from escaping, whence their absence of color, their “blackness” which makes them invisible”6 (Allègre 1995: 282, translation by Margaret Dunham). The astronomical metaphor rests on several shared properties between “holes” and these cosmic objects: both are containers, into which one falls, which are difficult to get out of and which trap you, furthermore they are black. But a “hole” supposes an emptiness which a priori risks being transferred (though inference) to the cosmic objects. Whereas for the latter, it is not their emptiness but rather their considerable mass which attracts

.  “Metaphor is a cross-domain mapping with preserved inferences”. Besides, for Lakoff, metaphors do not reside in words but in systems, as he showed in particular for mathematics (Lakoff 1993; Lakoff & Núñez 2000). .  “Des objets cosmiques si massifs qu’ils attirent les rayons lumineux, les courbent sur eux-mêmes, les empêchent de sortir, d’où leur absence de couleur, leur “noirceur” qui les rend invisibles”.



Words and their meanings 

objects and prevents them from escaping. The idea of mass and its physical effects are not part of the habitual meaning of “hole” where, on the contrary, emptiness plays the role of container-trap. To block this inference transfer, it is necessary to first set up the theoretical background of physics. It is not certain that the theoretician who created the neologism by metaphor to designate a new scientific object was aware of all the inferences transported by the metaphor. These can be very powerful and do actually play a structuring role in the scientific domain. For example, the computational theory of the mind which is prevalent in cognitive sciences rests on an initial metaphor, that of the mind seen as a computer (Bruner 1992). This metaphor has generated an entire theoretical apparatus (the brain’s “hardware” and “software”, “computation”, cognitive “pre-wiring”, “input”, “output”…). However the analogical process was erased: in the initial approach, it was a question of simulating mental processes using computers, it then became a case of describing them using computers, then it was a question of describing them using the computer as a model (metaphor), lastly, in a third stage, some began considering the brain as being a computer, a thinking machine (whence identification between the two domains, disappearance of the analogy). This founding metaphor whose heuristic process was erased, had considerable and often unwitting consequences, linked to the transfer of inferences. Thus, notably, because of the computer model, human thought has been seen as an autonomous system based on the manipulation of formal symbols which could be described in terms of logic and algebra, and everything that did not belong to the rational domain (emotions, perception) was removed from its workings. Because of the transfer of computer properties to thought, another shift took place, surreptitious but crucial, from the notion of signification to that of information (Bruner 1992). The problem of meaning in cognitive science has thus unconsciously been reduced to the domain of information processing. Information theory deals with the modalities of the transfer of information but not with those of constructing information, which was thereby removed from the field of cognitive science. Signification was then treated as a stable product (information to be transmitted), already a given in the input and thus not submitted to construction. The initial metaphor here had considerable impact on the definition of the object to be described and the model produced. When metaphorical denominations are new, their scientific impact is therefore not always quantifiable. Thus physicists trying to explain nuclear forces using properties associated with nuclear particles (one of which was even baptized gluon, meaning “that which sticks”), attributed qualities to them which represent inferences based on metaphorical transfers (quarks have “colors”, “flavors”, “charm”) for which the corresponding physical properties are not very clear (Allègre 1995: 230). It is therefore a question of thinking of a domain in terms of another by virtue of analogy and shared properties. But what the impact of the transfer of inferences in the construction of a model in particle physics will be, is difficult to say at the start.

 Stéphane Robert

So metaphors, in both the scientific domain and in general, are based on a fundamental cognitive mechanism which makes it possible to think of one domain in terms of another, through analogy. This process surely has heuristic and/or pedagogical virtues, and a certain cognitive efficiency. From a linguistic point of view, it allows a remarkable systemic economy and adaptability: a single unit gives access to several referents, an old term can be adapted to new realities or new concepts (on this last point, see the detailed studies carried out by Vidalenc 1997). But words are not concepts, they are “representation triggers” which present specific structural and functional properties and carry, along with their referential values, a whole fabric of structured relations (see 2.3. below on depth). Whence the “danger” which menaces language communication, that of the surreptitious import of representations and properties through inference. This danger is partly controlled, generally speaking, by the specification of a term’s meaning within the utterance (section 4), and in science through linking the term to a model which is most often explicit and constrained (through definitions, explicit descriptions of the properties and insertion in a specific model). The model constructs the value of the term in the background, and constitutes the term’s application domain. It is when the application domain is entirely specified that the term becomes a technical one, linked to a true scientific concept. It is therefore, in science as in ordinary language, first and foremost the articulation within a specific context which stabilizes the ambiguities in the meanings of a term.

2.2.2  Metonymy The second well known major mechanism for regulating meaning is metonymy. Traditionally speaking, metonymy is described as a shift in the referential value based on a relation of contiguity: the meaning of a term is transferred from one referent to another, by virtue of the contiguity relations between the two referents. Thus through metonymy, the blue helmets refer to the soldiers of the U.N. instead of referring to the helmets themselves; in to have a glass, the term glass can either designate the object or, through metonymical shift, its contents. These meaning shifts are based on the widespread mechanism of metonymy. Let us note that the term contiguity here is used in a very abstract sense; it can refer to relations of a variable nature such as the container for the contents (a glass), a part for the whole7 (a roof for a house), but also a cause for an effect (I like Schubert = I like Schubert’s works), the place of origin for the product (a Bourgogne), a place for the institution which resides there (the decisions of the White House), a body part for the moral properties associated with it

.  In this case it is called a synecdoche, but at this level of analysis, the distinction does not seem important as they both make use of the same mechanism.



Words and their meanings 

(have guts) … One can also consider as metonymical transfer the use of a brand name (or of an element of a category) to designate any element of that category, as in the case of fridge (Frigidaire in French) for refrigerator. Metonymy can take place through syntagmatic reduction: it is possible that the use of Schubert to refer to his works is based on the reduction of the phrase I like (the works of ) Schubert, the same for a (wine from) Bourgogne. Certain syntagmatic reductions are historically attested: thus the French term foie (“liver”) comes from the Latin expression iecur ficatum, a culinary term which originally designated the “liver (of a duck), iecur, fattened on figs, ficatum” of which only the beginning remained, ficatum (“enfigged”) > foie “liver” (Traugott & Hopper 1993: 81). Through metonymy, the term ended up designating not only this particular type of liver, but any liver. From the viewpoint of linguistic processes, the foie is therefore a variant of the refrigerator! The contiguity which links these different referential values is therefore always conceptual but is sometimes also accompanied by contiguity between the syntactic constituents. Beside these well known cases, Traugott & Hopper (1993: 80–93) mention a particularly interesting type of metonymy, where two meanings of a term are linked by a relation of inference. They give the example of the Germanic hwile (> wile) “time” which is the origin of the English while and the German weil (“because”). The adverbial phrase “at the time when” which uses this term (along with a distal demonstrative in the accusative and an invariable subordinator equivalent to “that” which were later morphologically reduced), first expressed the simultaneity of two events then, through inference, a causal link between the two events. Thus in Old English, in the sentence corresponding to “that disaster lasted the nineteen winters while (wile) Stephen was king”, the subordinate took on the meaning of “because Stephen was king”. From concomitant links one infers a link of causality. It is this value that was lexicalized in the German weil, which comes from the same hwile with a temporal meaning (as in Weile “moment, time”, verweilen “stay”) but which lost its temporal origins and no longer has any meaning but the causal one of “because”. In Tswana (Bantu), the verb “to get up (in the morning)” also functions as an auxiliary meaning “do (something) the next day”. D. Creissels (2001) analyzes the emergence of this second meaning through a process of semanticization (or lexicalization) of pragmatic inferences, linked to the fact that humans tend to make the alternation sleep/wake coincide with the alternation night/day. So if a human says “when I get up” one can, in the absence of contrary indications, infer that the person is referring to “tomorrow morning” because the prototypical getting up is the getting up which follows the night’s sleep. From the meaning “to get up doing something” one passed, through the lexicalization of the pragmatic inference, to the meaning “do something the next morning”. The semantic shift was accompanied by a syntactic reanalysis (auxiliarization process), but also belongs to the domain of metonymy from a semantic view point: the contiguity link is not simply a physical contiguity between the two referents but a contiguity of events.

 Stéphane Robert

Which is why I will follow here the more general definition given by Kövecses & Radden (1998: 39), following Lakoff: “Metonymy is a cognitive process in which one conceptual entity, the vehicle, provides mental access to another conceptual entity, the target, within the same domain”. The first important point in this definition of metonymy is that this linguistic mechanism is defined as a cognitive process and that it is described in terms of “access” to a conceptual entity. As for metaphors, it is not just a question of relations between words and things, but a question of the relations between the conceptual representations carried by words, as is shown by the diverse contiguity relations described here (part/whole, cause/effect…). Words are representation triggers and metonymy is a cognitive process which makes it possible for one word to trigger access to a new representation. It is noteworthy that in the case of metonymy, there is a dissymmetry: one of the representations is the vector for the other, it is the entrance point through which the target is accessed; therefore it functions as a salient feature of the second representation8 and the contiguity link between the two representations constitutes the referential path which gives access to the second one. The second important point is that, contrary to metaphors which are based on the transfer of properties from one domain to another, metonymy operates within the same domain: it allows the transfer of referential values within a single semantic domain. Which is why I propose to describe the metonymical links between a term’s two meanings as relations of a “horizontal” type; one must note however that, there too, the relation between the two meanings is mediated by an abstraction process that here is not based on analogy (as for metaphors) but rather on a link between properties of a single referent: metonymical process: conceptualization of contiguity

The blue helmets meaning 1 = part

The blue helmets meaning 2 = U.N. soldiers

Figure 2.  The metonymical link (horizontal relation).

.  For a more detailed analysis of the different types of metonymy as well as the cognitive processes at work, see the article by Kövecses & Radden (1998).



Words and their meanings 

Metonymical shifts can happen repeatedly in the history of a word. Thus the French term bureau initially designated a piece of rough cloth (bure) placed on the table where one worked. Then, through metonymy, it designated the table itself (“desk”), before, through a second metonymy, coming to designate the room where the table is found (“office”). It is probably undergoing further metonymy in designating the activities carried out in the room, as in des horaires de bureau “office hours”. Similarly, the term pen, from the Late Latin penna “feather”, first served to designate a feather object for writing (“quill”), then the pointed metal object which replaced the feather, then the stylistic qualities of those using the instrument (a witty pen). One can see through this example that the referential value of a term can survive its demotivation (loss of the link between the object “feather” and the value “writer”). We will see in the following section, with the example of fox (and also in 3.1. for the example of souris), that metaphor and metonymy can also be combined.

2.2.3  Combining metaphor and metonymy Interestingly, metaphor and metonymy can combine in the polysemous network. For instance a fox can refer to the wild animal, but also to its fur (metonymy), a coat made of its fur (second metonymy) as well as to an attractive woman (metaphor). As mentioned by Balbachan (2006), Lipka (1990a and b) “identifies two typical processes where metaphors and metonymy take place, showing a general schema as a lexical rule for semantic shift or transfer: radial shift and chaining shift”. To illustrate these two types of networks, he gives the example of two polysemic words: English head which shows a radial shift and English volume which manifests a chaining shift, as illustrated in the following figures (3 and 4).

S5

ym on

y

S3

et m metonymy m et ap ho r

S4

metaphor

S1

or

et ap h

r ho ap et m metaphor m

S2

metaphor

S9

S7

S8

S6

sememe

Meaning

‘head’

S1

upper part of human body

S2

seat of intellect

S3

life (cf. it cost him his head)

S4

image of head on one side of coin

S5 S6

knobbed end of nail foam on top of liquor

S7

top of page

S8

fully developed part of boil

S9

end of table occupied by host

Figure 3.  Radial shift (from Balbachan 2006).

 Stéphane Robert metonymy S1

metonymy S2

metaphor S3

S4

sememe

Meaning

‘volume’

S1 S2

roll of parchment (disappeared) book tome

S3

size, bulk of a book

S4

size, bulk of other things

Figure 4.  Chaining shift (from Balbachan 2006).

2.2.4  Active zones and contextual interactions Let us further mention an important factor in the semantic variation of terms: interaction with the context. Plasticity in terms is also largely conditioned by their interaction with the verbal and situational contexts, which produce a veritable “work” on the meaning of lexical units, defining landmarks, attractors and “active zones”, producing coercion, semantic shifts or semantic layerings. These processes will be presented in section 3. because they contribute to the stabilization of the word’s meaning in language use. 2.3  The depth dimension of language 2.3.1  Semantic universes: Frames and scenarios So far, we have described semantic structures and mechanisms allowing meaning shifts, but the meaning of linguistic units is not limited to these meaning matrices. Linguistic units, in effect, are linked to semantic universes, to representational backdrops which contribute to the value of a term’s meanings and which themselves can be highly structured. The terms buy and sell for example, designate a particular action between two participants which implies a history of variable but codified mercantile relations depending on the language and culture, which Fillmore has called “frames” or “scenes” (Fillmore 1977, 1982).9 However, these extra-linguistic factors have an impact, either direct or indirect, on the semantics of the terms and on their use; the notion of “frames” is intended to capture useful chunks of encyclopedic knowledge relevant to the usage of linguistic units (Goldberg 1995: 26). Thus the term bachelor is often defined as a man who is not married. But this definition is not sufficient for rendering either its values or its usage; the term implies a precise cultural background

.  See Martin (2001) for an elaboration on the notion of “frame” and its role in polysemy.



Words and their meanings 

which explains why one would not easily say that the Pope or a hermit is a bachelor. The term weekend of course profiles a certain part of the seven-day cycle, but a full understanding of its meaning implies to know a larger semantic (and cultural) frame by which Saturday and Sunday are non-working days therefore associated with leisure, sport, camping … (Fillmore 1982). Furthermore, the notion of frame often explains the difference between two synonyms: for instance the words roe and caviar refer to the same entity but are associated with different frames, anatomical or gastronomical (Langacker 1987: 164–65). In Ibo (a Kwa language of Nigeria), one thus finds eighteen terms for “to buy” depending on the nature and conditioning of the object, but also on the circumstances of the sale, the particular gesture associated with it, the quantity or fractioning of the object, whether the seller is obliged to travel, if the person one asks to buy will pay or not, etc. … (Chukwe 1997). The different customs in the background directly intervene in the semantics of the verb as they are categorized in the language. And one sees that the scenarios underlying the signification of a meaning are culture-dependant. But these scenarios can also be indirectly linked to the term’s meaning. Thus in English, white is associated with marriage because of particular customs, namely the marriage ceremony and the color of the bride’s dress. The presence of this scenario in the background has the effect of generating connotations associated with this color; it induces diverse “resonances”: it is a positive color, it evokes purity, virginity, the intact nature of an entity, its innocence (white as snow for “innocent”). Again, the associated scenarios and the connotations stemming from them vary from culture to culture: in China, red is the color of weddings and white that of mourning; white therefore will not have the same connotations as in English, and will certainly not evoke virginity. These connotations linked to background scenarios are indeed part of the term’s meaning, and play an important role in a term’s stylistic effects and meaning variations within utterances. We have seen that linguistic units often function through the selection of one of the referent’s properties to designate it, which leads to a property being used to designate several different referents (cf. greens or bleu). Linguistic units thereby constitute access paths to a complex representation fabric or network. Through its different values, a single term refers to different scenarios: that greens can refer to vegetables reflects the fact that the tops, the leaves, of vegetables are of that color; that greens can refer to members of a political party is due to the fact that they tend to use green banners; in British and American history, greens were pieces of land reserved for common use in each village, first for grazing purposes, then for recreational uses. Thus one sees that these scenarios are historically and culturally grounded. In certain cases, when a scenario no longer has historical validity, it becomes demotivated, and can even disappear. This is the case for the French bleu, which, among many other meanings, referred to young army recruits who usually showed up wearing their blue work clothes. With the end of obligatory military service, this term may eventually fall into disuse.

 Stéphane Robert

However, history has given us new oppositions, as with the greens (ecologists) and the reds (communists), which are probably metonymic designations (“who bear green banners”, “who bear red banners”). There are also new background scenarios which lead to the emergence of new referential values, such as un blanc which formerly in French referred to a royalist soldier (whose uniform was white), but now belongs to a different paradigm linked to completely new reference values referring to wine.

2.3.2  Connotations As we saw above, languages create network relations within the semantics of words (metonymic or metaphorical relations between meanings, relations between a schematic form and its different instantiations, relations between different referential values, different scenarios or semantic universes), but they also associate various connotations with a term’s meaning. As we saw with the example of white, a term’s connotations vary from culture to culture, and also according to its different uses. These connotations explain certain synonymic variations: car, automobile and jalopy are synonyms but are distinguished by their different registers and connotations, as is also true for jaundice and hepatitis. These connotations serve to signal a social role (which can be momentary) played by the speaker, or the speaker’s belonging to a specific social group. Similarly, using the expression father to designate a priest signifies that the speaker is a practicing catholic, contrary to using priest or clergyman. In the same way, using the heat for “policeman” signals belonging to a certain age group and general ideology. In fact, choosing a term for its connotations allows speakers both to situate themselves intersubjectively (as regards the group) and to express one’s position, one’s judgment on what is being talked about. On the discourse level, connotative choices permit argumentative strategies based at least partly on identification phenomena, largely exploited by publicists (see Honeste 1997; Grunig 1990). Certain connotative values can be more generally associated with words, so that they carry uncontrolled resonances in a given culture. Such is the case for the term North for example. For many French people, the term calls up thoughts of cold, grey, wet weather, and for Parisians, the daily grind, which are all negative values; moreover it is opposed to diverse positive representations of the South: sun, joy, feeling good, vacation, because of seasonal habits which are quite specifically French. Therefore avoidance strategies are developed for example by the departments and institutions in the North of France. The Artois University Press has thus been prettily renamed “Septentrion Press”. The term septentrion is a synonym for North, but has neither the same distribution (usage contexts) nor the same semantic resonances; it is an old term, associated with a poetic and literary universe which calls up all sorts of other associations. These two terms then have the same referential value but not at all the same meaning. And if the department Côtes du Nord successfully changed its name to Côtes d’Armor, it was both to avoid the negative associations with the North and to



Words and their meanings 

endow itself with a more fitting denomination, both geographically and culturally: the Côtes d’Armor are in the north of Brittany rather than in the north of France, moreover they have a specific history which the term Armor positively evokes. This new name, in effect, is wound up in a very different network of associations and connotations: it not only brings to mind the Celtic legends, but also the formal echoes between Armor and Amor.

2.3.3  The depth dimension of language: Fabric of networks To the fabric of diverse semantic relations which units enter into (Armor and Celtic legends), one may also add a network of formal relations, either etymological or not (Armor and amor), between units. These relations vary greatly from language to language, and probably even from one individual to another because they are built on both social and individual experiences, and each one generates diverse association representations. For the present writer, the term uncle of course calls up the domain and structure of kinship relations, but also the book Uncle Tom’s Cabin, and her own uncle who had a house in the Alps and hence the memory of winter sports, and so forth. The cognitive reality of these formal relations between non related terms is also visible both in slips of the tongue and in puns. I will not go into the details of the diverse morphological relations that are set up in the paradigms here (cf. Robert 2003). These morphological relations (etymological or not) thereby produce echo phenomena between the terms of a language (amor – Armor): the formal relationship induces semantic relationships between the different notions, connotations or values associated with each of the terms. Depending on the language, words resonate in an extremely variable way, depending both on the physical and cultural contexts and on the rest of the language’s lexicon. The linguistic units trigger representations which are caught up in a complex network of relations, at once language internal and external, semantic and formal. This web of relations and associations that links linguistic to extra-linguistic matter, constitutes what I call “the depth dimension of language” (“l’épaisseur du langage”, Robert 1999, 2003). Depth constitutes a third dimension in language, as opposed to the syntagmatic dimension (relations between the utterance’s terms) and the paradigmatic dimension (relations between the terms that may potentially occupy the same spot in the utterance); it is what makes the meaning “subjective and open-ended” (Lichtenberk 1991).10 This depth dimension constitutes the semantics of a term, and in a way represents the extremely variable harmonics that the semantic-structure-as-fundamental-frequency

.  “A term may have a primary meaning, but its total meaning subsumes not only this primary meaning, central designation, but also all the other more or less peripheral aspects of the situations in which the term is used.” (Lichtenberk 1991: 480).

 Stéphane Robert

gives rise to. The depth of language is a complex area where linguistics associates both with linguistic and extra-linguistic matter and which plays an important role in the construal of an utterance’s meaning (Robert 1999). Table 1 below, which is certainly far from being exhaustive, lists the different components of the meaning of linguistic units that we have highlighted here. They gather together variation factors that are at once internal (within languages) and external (from language to language): Table 1.  Processes involved in the words’ meaning –  world segmentation and categorization –  referential paths –  profiling: internal architecture (figure and ground) –  plasticity and meaning shifts (metaphor, metonymy) –  referential scales –  variable application domains (instantiation) –  depth: –  scenarios and semantic universes –  networks of formal and semantic relations between terms –  connotations –  associations between linguistic and extralinguistic representations

3.  The problem of transparency and referential accessibility 3.1  Motivation Thus far, we have examined the different mechanisms for constructing the meaning of terms, such as property selection and referential paths, transfer processes through metaphor or metonymy. In these different cases, meaning is construed through a referential process which is indirect but also transparent and motivated. This motivation most probably plays a role in how these terms are stored in the memory (by linking different meanings together or by linking a meaning with the physical and cultural properties of the referent) as well as in the cognitive accessibility of the referent. However, this referential transparency varies within a language as well as from language to language. Within languages, the semantics of terms is not always motivated and the modalities of reference accessing may be opaque for different reasons (see below). Furthermore, referential efficiency among terms may vary within languages, as well as from language to language, for the “same” term. Thus languages present varied strategies for reference accessing which are partially linked to their morphological, and therefore formal, properties.



Words and their meanings 

The specific problems raised by technical term translation and terminological creation are particularly interesting on this head, as they bring to light the necessity for efficient designations, whether it be a question of procedures to follow in case of emergency, translating traffic regulations, instruction manuals, or even school books and teaching materials. It is most probably necessary to introduce as much motivation in the designations as possible. This entails either transparency in the referential paths, or retaining the most salient properties within the culture to designate the referent. Thus to indicate the blinking cursor which shows its position on a computer screen, French used, in succession, first curseur then souris (“mouse”). The term curseur has fallen into disuse because visibly connected to an era when screens had a different presentation, and rested on a metaphor linked to a slide rule and to the movement of a mobile element along the ruler. The term souris (“mouse”) is based on a metaphor then on a metonymical extension. It began by designating the element which serves to transmit the hand’s movements to the screen: the metaphor was based on the shape of this element (small and oval) and on the (rapid) movements it made possible; then, through metonymy, the term souris (“mouse”) came to signify not the element moved by hand, but the element it affected on the screen. The salient properties that made this metaphorical shift of the term souris (“mouse”) possible were the size and movements of the object. While trying to create terminology in Banda (Central African Republic), as described by M. Diki-Kidiri, for the same element it was the term for firefly that was retained; the salient properties which seemed the most suggestive being size and luminosity on a dark background (as were the screens at that time). These privileged paths are certainly linked to the cultural world. In fact, the cognitive efficiency of metaphors is often based on the existence of a world of wider cultural references which are not always transposable from one language to another. As J.L. Vidalenc showed (1997: 143), the metaphorical expression scientific frontier, used in American scientific presentations, refers more to Westerns and to the American “frontier” culture associated with them than to the sources usually called upon in scientific communications. Such an expression would certainly not have the same meaning for a French public. Certain general metaphors do not exist in all languages. In English, up and down are associated with turning a machine on and off, as in the expression to shut down the computer. This analogy between movement and turning something on or off is not the same in French, where downward movement is rather associated with something falling, and probably breaking. This association is so strong that it prevented me for a long time from using the entry shut down in the scroll menu on my computer, for fear of breaking something. … In other words, we are faced with an apparent paradox: it is probably by taking what is most typical culturally in a language that one is able to construct the best

 Stéphane Robert

“equivalences” between languages and not by taking universal invariants conceived of as having a minimum of common contents. It would certainly be interesting to carry out experiments on the possible existence, in different cultures, of privileged access paths to reference: spatial trajectories, functional property selection (cf. index above) or tactile properties (linked for example to manipulability which probably plays a role for certain classifiers in Chinese), visual properties (it is because of their long and thin shape that the terms “fish”, “stick”, “street” and “necklace” in Mandarin have the same classifier tiáo). The olfactive properties of referents seem to be more rarely selected as salient. Corbin & Temple (1994) note however the French term fenouillette which designates a variety of apple whose smell evokes that of fennel (French fenouil). But the degree of transparency in terms also depends on language specific morphological factors. German presents a remarkable degree of transparency in compound words, transparency which is linked to the clarity of its compositional rules and the flexibility of its particles (Pérennec 1997). Furthermore, specialist vocabularies in German, much more than in French, with English somewhere in the middle, make widespread use of so-called folk roots. One can compare the German Unterhaut (lit. “under-skin”) to its English equivalent derm, or Einbaumboot (lit. “one-tree-boat”) to its English equivalent monoxyl canoe (also known as a dugout in everyday speech). The German terms thus show a remarkable referential transparency as compared to English. Although the semantic interpretation still necessitates recourse to encyclopedic knowledge, the mode for accessing the referent is transparent. Concerning this last example, one notes that the precision of the reference path or its explicit character do not necessarily imply that the reference is accessed more quickly. The English monoxyl is a constructed term which is explicit in its referential path (“made of a single piece of wood”) but it makes use of (Greek) roots which are opaque for most speakers. Similarly, most chemistry terms, such as cupritetramine and desoxyribonucleic are analytical terms, explicit and free of ambiguity, but opaque for non specialists. Because referential accessibility implies not only an explicit (analytic) reference path but also knowledge of the theoretical background, i.e., the term’s application domain. Depending on the speaker’s knowledge, the referential path may be transparent even though the reference is opaque. Inversely, a vague term may be referentially efficient, because of its usage conditions.

3.2  Opacity and accessibility All terms in a language are not always “constructed words” or “defined descriptions” which furnish the speakers with (always partial) descriptions of the referent. In effect, there are, within languages, different strategies for accessing references, especially through analytical processes (as with the preceding examples) or “direct” processes.



Words and their meanings 

Of course referenciation is always mediated as it is transmitted via units which refer to representations constructed by the language, but access to the reference can be carried out either through constructions (analytic processes), or through encoded units as such, which are unanalyzable (as with proper names for example, or more generally, mono-elementary units such as table or glass), which form meaning blocks. Let us quote the famous distinction proposed by Frege between the different denominations for a single planet: Venus, which is a proper name and constitutes a mono-elementary unit, the evening star, or the morning star (to which one could add the shepherd’s star), which constitute definite descriptions, i.e., analytical ones. The same is true for the Castafiore and the Milanese Nightingale. In strategies of reference construction, the referential path may be opaque, either because the coded unit is not analyzable (table) or because demotivation has taken place, and the compound meanings have been lost. This is the case for example for turkey, a term which originally designated a fowl from Turkey, but with the fowl becoming widespread, the metonymical path was lost. The American states of Louisiana, Virginia, Georgia and the Carolinas all bear testimony to the monarchs ruling at the time of their conquest. The name Alsatia (“Alsace”), literally “other seat”, designated a foreign settlement, referring to the Germans who had settled west of the Rhine. The fact that this area was considered a sort of enclave led to the term being used derogatively in London to refer to the White Friars precinct which had become a sanctuary for debtors and law breakers, and thus an asylum for criminals. This complicated path followed by the semantic shift is totally opaque nowadays (Shipley 1984: 344). The case of grève in French is another nice example of demotivation. This name has two meanings: (1) it designates a “beach strand or river bank” and (2) it refers to one of France’s national specialties, namely “strikes” (to be on strike). Originally, these two meanings were linked by a double metonymy: the “place de Grève” (lit. The Strand place) was the name of a place in Paris, close to the Seine’s bank; at a certain time in history, the workers would meet in this place to protest against their working conditions. Hence, the phrase “être en (place) de grève” took on the meaning “to stop working and go to this place for protesting”. Later, this expression became autonomous (as in une grève importante “an important strike”) and the link with the particular geographical place was lost: the metonymic shift was demotivated and the two meanings appeared to belong to two homophonic terms, corresponding to what Lichtenberk (1991) calls a case of “heterosemy”.11 Today, the referential path of grève is opaque. But its meaning is not.

.  “In heterosemy, the semantic (as well as the formal) properties of the elements are too different to form a single conceptual category. Rather, the category has only a historical basis: what unites its members is their common ultimate source” (Lichtenberk 1991: 480).

 Stéphane Robert

In fact, opacity of the referential path does not necessarily imply opacity of the term’s referential value, nor its inaccessibility, just as the path’s transparency does not guarantee transparency of reference. In effect, most acronyms such as LASER, AIDS or DNA represent opaque referential paths for most speakers. However, their referents remain accessible (at least to a certain degree), especially as these objects are part of a familiar universe; the term then functions as a sort of coded unit within the language. Similarly, the referential value of a term generally survives its demotivation, as is the case for example for the plumber (from the French plomb “lead”) which still designates the same category of workers even though they no longer repair lead piping. Another interesting example is also given by Lee (2001: 10) which is the case of the English bug (1. “insect”, 2. “fault in a computer program”). This term was first used when a problem with one of the early computers was found to be due to the presence of a dead insect in its innards and therefore used in its original sense concerning a problem with the computer. However, this situation involved the activation of a new frame (computer programming), which was the source of new semantics for the term that came to refer to any fault in a computer program, even when unrelated to the presence of an insect in the machine. The rate of the (formal) evolution in words does not necessarily follow that of their referents. This discrepancy does not hinder speakers because the relationship between form and meaning is fundamentally arbitrary and coded, even if occasionally motivated. What is crucial is that the term have meaning for the speakers, namely that it permit access to a common representation; if the relationship between the linguistic form and the representation attached to it is most often arbitrary as concerns the system of the language, from the speaker’s viewpoint, it no longer is from the moment the representation is acquired: a table is a table. It is therefore most probably when it is a question of gaining access to a new representation, as in the case of terminology creation, that motivation and transparency in the referential path are the most important. But path transparency and referential accessibility do not necessarily go together. This discrepancy between path opacity and referential transparency can be explained, in my view, by a more general linguistic mechanism. I think that on the discourse level, namely when units are used in an utterance, there are two modes for forming meaning: by quotation or by construction. In fact, from a structural point of view, discourse makes use of different types of units: either simple units (table), constructed units (be they derived: dancer, compounded: pillowcase or phrases: head of hair). These different structures probably give rise to different modes of constructing meaning and reference access, in the production or comprehension of the utterance: on one hand certain junctures are formed at the time of speaking (construction formation mode); on the other hand, certain structures function as fully fledged units (“coded units”), stored in the memory as wholes and used more or less as such in the sentence (quotation formation mode). These two utterance modes are most



Words and their meanings 

probably both necessary for speech. The first ensures the creativity and plasticity necessary for language, the second ensures economy in individuals’ efforts and interpersonal comprehension. However, it seems to me that these two meaning production modes do not necessarily follow the linguistic structure of the units being used: some complex units may, from the point of view of production and reference access, function as simple units, produced through citation and not construction. The expression head of hair probably usually functions as a simple unit and the speaker (and listener) probably do not construct its meaning by analytically following the term’s referential path. Just as in toothbrush one does not necessarily hear tooth, and in an instrument’s mouthpiece one does not necessarily activate the term mouth. This is why transparency of reference paths does not necessarily go hand in hand with the accessibility of the referent: it all depends on the reference construction mode during discourse. These two construction modes also apply to structures larger than the word, and even entire sentences. Proverbs (April showers bring May flowers) and certain set expressions (hard as Job, to smoke like a chimney, to keep a stiff upper lip) generally belong, on one count or another, to the quotation mode: when speakers use them, they do not usually build them up from their individual components, but quote them as fully formed units. However, the latent referential path of set expressions can be reactivated. This is often what happens in puns or advertisements which frequently consist in bringing to the surface opaque referential paths. The varying activation of component meanings then depends on the specific dynamics of the sentence.

4.  Construing meaning in discourse: Stabilization mechanisms Linguistic units present ambivalences and potential semantic overloads due to their polysemy and their representational depth, the complex fabric of relations they enter into. However, in language activity, units never appear on their own, but always in a verbal and situational context, inserted in utterances where all of their values are not present. Following the tenets of cognitive semantics (Langacker 1987 and 1991; Talmy 2000), we consider that “instead of thinking in terms of words as expressing “concepts”, we should think of them as tools that cause listeners to activate certain areas of their knowledge base, with different areas activated to different degrees in different contexts of use” (Lee 2001:  10). Being used in discourse, the context “acts on” the meaning of the units and constrains their interpretation. More generally, discourse, through different relating mechanisms, makes it possible to progressively build the reference frame and “verbal scene” (Victorri 1997) which will specify both the meaning of each unit and that of the sentence. Thus reference is always construed contextually through a dynamic process, for which we will mention a few of the mechanisms here.

 Stéphane Robert

These meaning construal mechanisms in discourse contribute both to the general polysemy of the term and to the stabilization of its meaning within a particular utterance.

4.1  Application domains, meaning attractors and semantic isotopics 4.1.1  Application domains In their different uses, words are always invested, instantiated in “application domains” which define their semantic incidence and contribute to creating their referential value and contextual meaning. Incidence domains are important for terms because they contribute both to the variation (plasticity) of their potential meanings and to their stabilization in the utterance. A change in the application domain and semantic universe of a term can produce a meaning shift and a radical change in its referential value. As mentioned by Sinclair (1998: 7), the meaning of white when followed by wine (as in white wine), refers to a different color range (from almost colorless to yellow, light orange or light green), than when it is not so followed. Furthermore, depending on the context, a term can be linked to a different semantic universe, while keeping its profile. This is the case for example of the word pit, which refers both to a hole or cavity in the ground and a certain seating area in a theatre or auditorium. In both cases, it represents an element of the same general shape as well as certain shared functional properties (it is usually hollowed out with the intention of containing people or objects, and is usually below ground or below the level of surrounding people) but its application to different semantic universes entails a completely different referential value in either case, as well as wholly divergent associated properties (negative as in the saying “it’s the pits!” and positive as being some of the best seats for seeing the scene in the theatre). The different semantic universes the terms are linked to are therefore the source of the variation in the units’ meanings, but also play the inverse functional role in disambiguating the construal of a term’s meaning in discourse. How are these application domains and semantic universes specified? By the verbal context (the relationships between the sentence constituents and the relationship between the sentence and what precedes it) and by the situational context (extra-linguistic factors pertaining to the discourse situation): these together construe different reference points which steer the term’s meaning. 4.1.2  P  rimitive meaning attractors (prototypes, personal attractors and discourse situations) The terms we use are caught up in the representation depth that we mentioned earlier where extremely diverse relational networks are woven and which vary according to the cultures and individuals, as they are bearers of an individual’s experiences, both material and psychological. However, this representation depth affecting words is crossed by different reference-concentrating areas, landmarks or anchoring points, which serve as “interpretative attractors” or “meaning attractors”, i.e., elements which



Words and their meanings 

attract/steer a term’s interpretation in a particular direction. The prototype is one such element. Individuals also have their own meaning attractors: out of context, a linguist will tend to interpret the word instrumental in its grammatical meaning (that of a morpheme serving to indicate that the complement corresponds to the instrument of the process) whereas for musicians, the first thing to spring to mind will be their violins or pianos. The discourse situation also functions as a factor specifying a term’s application domain and as a meaning attractor: depending on whether one is at a concert, in a bakery or at a Chinese restaurant, the French term baguette will be connected to the semantic domain construed by one’s location and will refer either to a conductor’s baton, to a loaf of bread or to chopsticks. The discourse situation therefore functions as the default “meaning attractor”: it calls up a reference domain that the terms used will naturally be connected to. The reference domain acts as the backdrop or ground against which the figure defined by the term’s signification will be profiled, the figure and ground together constituting the contextual meaning of the unit. The pragmatic context (i.e., the situation where the utterance is produced) can also lead to a variety of meanings on the grammatical level which overthrow the meaning of the whole sentence: in the French sentence je vous coupe la tête (lit. “I am going to cut you the head”), depending on the situation, the personal pronoun vous (“you, for you”) has two different possible values (benefactive or applicative), so that the sentence as a whole takes on a completely different meaning: if you are at a fishmonger’s, it would mean “I’m going to cut the (fish’s) head for you” (benefactive), whereas if you are under threat from a mad man, it would mean “I’m going to cut off your head” (applicative). A term’s meaning is construed through interpretative mechanisms which are conditioned by different factors. Communication is only possible because the reference points of the verbal context take precedence over the rest. But interference between the different “meaning attractors” is always possible, as is shown for example by misunderstandings and puns (see e.g., Arnaud 1997).

4.1.3  Contextual meaning attractors: Semantic isotopics The most important reference points for communication are those which are created by the verbal context, e.g., by the creation of relationships between a given term and the rest of the utterance, and between the utterance and those preceding it. The relationships between the terms of a sentence is notably governed by a fundamental mechanism of “semantic isotopic”12 which consists in linking a term’s meaning to the semantic universe of the preceding term to create interpretative continuity in the line

.  The concept is from Greimas (1966: 96). It was further elaborated by diverse linguists. For a detailed analysis of the different types of isotopics, see Rastier (1987: 87–141) for example.

 Stéphane Robert

of thought. Thus, in the absence of particular contextual indications, in the sentence the pilot pulled back on the stick to fly higher, the terms pilot and fly lead to interpreting stick as an “airplane control handler” and not as a “tree branch”. Through this isotopic process, concatenation draws a guiding thread through the depth dimension of language, which orientates the meaning of a term towards an interpretation congruent with the semantic field established by what precedes it. It thereby contributes to removing the potential ambiguities inherent to linguistic units due to their polysemy. I consider that the same principle of semantic isotopic comes into play in the disambiguation process evoked by Paprotté (1998: 248) concerning the two meanings of port in English: “safe harbor” or “red wine”. Thus two different isotopics are created in the following two examples (Fig. 5): The violent hurricane did not damage the ships which where in the port isotopic1:

sea world

→

sea world

→

harbor

Deceived by the identical color, the host took a bottle of Barolo instead of one of port isotopic2:

color

→

beverage

→

wine

Figure 5.  Two different isotopics for port.

This disambiguation process which draws a guiding thread through the depth dimension of language can be schematized as in Figure 6:

language depth

isotopic

utterance Figure 6.  Semantic isotopic and language depth.

Setting up contextual relations creates interpretative landmarks and semantic fields which, apart from specific psychological situations which lead to the interference of personal attractors (preoccupation, fatigue, obsession), prevail over the other meaning landmarks and attractors.

4.2  Contextual linkage and multiple landmarks The meaning of a word in context is the result of a multifactor process. In effect, all of the factors, contextual, lexical and grammatical, constantly intervene in the progressive



Words and their meanings 

construal of an utterance’s meaning and in the specification of the values of its terms. When it appears in a sentence, a unit is linked, concomitantly, to elements at different levels: in relation to the verbal context and preceding situation, in relation to the other lexical elements, in relation to the syntactic structures. Everything is linked in language and the relational mechanisms produce meaning through constant interaction between the elements involved. Putting words into sentences thereby activates one or another of its latent meanings and produces a contextual linkage (it clears a pathway through the forest of meanings). In the following sections we will first present the different linguistic components interacting at the utterance level in order to specify the meaning of a word (4.2.1., 4.2.2. and 4.2.3.), we will then mention some of the main mechanisms characterizing these interactions (4.2.4.).

4.2.1  Lexical interactions Linking a term to various elements (the context, linguistic units and structures) does not constitute a simple filter among a unit’s possible values (as in the example cited for pit), it produces a veritable working over of the term’s meaning, which is construed by interactions. Thus in a floury hand and a floury pear, the adjective always refers to the fact that the object in question at a certain point in time presents certain qualities linked to flour (which is its meaning), but its meaning varies considerably as it designates in one case an object covered with flour and in the other, the texture of flour (Corbin & Temple 1994). However, this specification of the value floury is not foreseeable outside of the connection of the adjective to the particular nouns it determines. Similarly, it is the specific values of steak and man that will inform the variable values of the adjective tender in a tender steak or a tender man, while at the same time tender will specify the steak or man in question. We also saw that the shift in the meaning of square in a square person or a square foot is brought about by the interaction between the nominal referent’s properties and those of the determinant. Moreover, in some cases the precise meaning of a word or phrase is determined more by the verbal environment than by the parameters of the lexical entry, as in the case of white in white wine (cf. supra 4.1.1.), which is what Sinclair (1998: 6) calls a “semantic reversal”. From a linguistic point of view, these semantic shifts can be at least partly predicted by a corpus-based analysis of the word’s collocations (cf. Sinclair 1998; Deignan 2006). 4.2.2  The framing role of the verbal context Thus the simple linking together of two notions produces an effect on their semantic values, due to their respective properties. In the examples we have looked at, the term which triggered the variation and meaning specification of the adjective was identifiable and located in the immediate vicinity since it involved a noun determined by the adjective. But there is not always a one-to-one correspondence between the elements which interact and it is not always one unit which acts upon another unit. In effect, a

 Stéphane Robert

preceding utterance (no longer simply a preceding unit) can orient the value of a following term or utterance: the meaning of setting in I’m going to change the setting will vary according to whether it follows the sentence your ring looks very old-fashioned or this scenery doesn’t seem quite right for Shakespeare. Similarly, the verbal context can largely constrain the value of a unit or a whole phrase. Thus the whole meaning of he laid the table will vary according to whether one is talking about a child or a woodworker. Even if the terms “child” or “woodworker” were not explicitly mentioned in the context, the context nonetheless functions as a thematic landmark.

4.2.3  Interactions between syntax and semantics Grammatical factors also affect terms’ values. To mention only a few examples, in French word order plays a role in specifying the meanings of units, as can be seen in the opposition between un homme grand “a tall man” (physical value of the adjective) and un grand homme “a great man” (appreciative value of the adjective), the place of the adjective thus plays a semantic role in French which constrains its behavior (shown by the fact that *la verte herbe lit. “the grass green” is impossible, the adjective can only follow the noun as in l’herbe verte whereas English shows the exact reverse) and produces meaning shifts: because of the adjective’s location, in un bel imbécile (“a great fool”), bel does not designate a physical quality but serves as an intensifier for imbécile (as in English, where to obtain a positive reading for great, it would have to follow the noun: this fool is great). Similarly, the plural can also produce semantic linkage. Thus in English the word term has numerous possible meanings: it can designate a “word” (as in the expression a technical term) or an “end” (as in to put a term to one’s life), it can also refer to an expected end, a qualitative meaning (to be born at term) (cf. Robert 1999). The simple use of the plural, terms, produces semantic effects as it implies fragmentation which renders the word countable, and thus leads to its taking on the meaning “relations” as in to be on good terms with someone, or “conditions”: the terms of the contract. As terms referring to a quality are not fragmentable (cf. *whitenesses), the plural thus eliminates the qualitative interpretation of term. In general, lexical and grammatical factors interact and mutually condition each other. For a construction 〈verb + to + complement〉, the nature of the introductory verb constrains the choice and the grammatical category of the complement: take to sub-selects an activity, whereas go to sub-selects a place. The semantics of the verb thus limits the choice of complement by creating both syntactic constraints (for take to to be able to select an entity as complement, the preposition could not directly follow the verb, as that position would be occupied by the beneficiary; He took John to the zoo vs. He took to swimming in the morning) as well as “semantic isotopics”. However, as these examples show, the lexical semantics (the value of the introductory verb) also specifies the semantic value of the syntactic construction (value of the complement introduced by to). These interactions between semantics and grammar are also visible



Words and their meanings 

in the syntax of metaphors, as shown by Deignan (2006) through analysis of a large corpus. For instance in Spanish (Balbachan 2006), the metaphorical expression matar el tiempo (lit. “killing time”) implies both a selectional constraint violation and a syntactic anomaly (the absence of the preposition “a”). In French, depending on whether they are used metaphorically or not, the following movement verbs have different syntactic constructions, with different prepositions: one says courir vers la maison “run towards the house” but à la victoire “(run) to victory”, nager en piscine “swim in a pool” but dans le bonheur. lit. “(swim) in(to) happiness”. As shown by Yaguello (1998: 98–106), figurative expressions have their own syntax: although in French one can say elle a l’oreille fine (lit. “she has a fine ear”, meaning “she hears well”) or elle a le coeur gros (lit. “she has a big heart”, meaning “she is sad”), the constructions son oreille est fine (lit. “her ear is thin”) or son coeur est gros (lit. “her heart is big”) are impossible with a figurative reading, whereas they are acceptable if the terms are taken literally: son oreille est fine (“her ear is thin”) which is constructed, and interpreted, in the same way as ses yeux sont bleus (“her eyes are blue”). What is at play here is that in the figurative sense, it is not the body part which presents the predicated property, but rather the quality associated with it, hearing for ear or feelings for the heart. However, body parts, as we mentioned above, belong to the category of inalienable possessions having specific syntactic properties which are not found in the metaphorical or metonymical uses of body parts in French (cf. 2.1.1. je me suis lavé les mains/j’ai lavé ma voiture). Similarly, whereas by metonymy one may say in French il a la gâchette facile (lit. “he has an easy trigger”, meaning “he is trigger-happy”), one cannot really say *sa gâchette est facile (lit. “his trigger is easy”). Thus the figure leaves a trace in the syntactic constraints. More generally, the corpus-based analysis of Hunston & Francis (2000) and Deignan (2006) have shown an interesting point for the disambiguation of polysemy: the different meanings of polysemous words have a tendency to be realized in distinctive grammatical patterns. Let us note that the interaction between syntax and semantics can happen retroactively. In French, pied-de-biche rouillé (lit. “foot-of-doe rusted” meaning “rusty crowbar”), the (postponed) adjective retroactively converts the preceding expression from a genitive construction into a compound noun referring to a tool. Thus the factors that determine the meaning of a term vary in their nature. They can be either linguistic or pragmatic, and generally belong to an incidence domain which is also variable: their scope can cover a word, a group of words, or a whole sentence. The diversity of a term’s specifying factors (context, units, grammatical constructions, sentences) and their variable scope (incidence on the following unit or on the sentence as a whole) present a difficulty when one tries to model the processes of construing meaning in discourse. However, the different factors that specify a term’s meaning in discourse follow regular processes which are based upon a general mechanism that Culioli calls “repérage” (anchoring) (Culioli 1982). This anchoring is most

 Stéphane Robert

probably a fundamental cognitive mechanism, also at work in construing the figure and ground, topic and focus in language, as well as in visual perception. The anchoring process sets up a relationship between two terms through which one of the terms is taken as an anchor point for localizing (in its abstract sense) the other term. Thus a term is located in reference to another term which serves as its reference point and this relative localization of one term in reference to another produces new determinations. These “terms” can be of varying natures and dimensions: notions (through the different elements in the lexicon, such as a name in relation with an adjective for example), temporal reference points (a moment in time), or subjective ones (a subject) but also topic and focus, propositions, sentences, or even a word’s prototypical meanings (in relation to its contextual meaning). These reference relations between an utterance’s terms produce contextual links, activate meaning attractors, create semantic isotopics and specify word meanings.

4.2.4  Some semantic mechanisms at the utterance level One can characterize some of the different semantic mechanisms operating at the utterance level and producing semantic variations. The following list is, of course, not exhaustive. Profiling active zones As shown by Langacker (1991a: 189–201), different semantic components of a word can be activated, depending on the context. For instance, in the following two sentences, different parts of the window are activated:

(1) He cleaned the window



(2) He opened the window

Because of the semantics of the verb, (1) refers more specifically to the glass of the window, whereas (2) draws more attention to the frame of the window. Therefore, two different zones of the word’s meaning are profiled in the different sentences, for which two different synonyms could be used. Constructions and coercion The grammatical context can at times cause the language-user to reinterpret all or parts of the semantic features of a lexeme that appears in it. This phenomenon has been referred to by computational and generative linguists as “coercion” and was mainly studied for aspectual shifts (Pustejovsky & Bouillon 1996; De Swart 1998). Consider a sentence like (3), taken from DeVelle (2003):

(3) The tourist photographed the sunset until nightfall.



Words and their meanings 

The verb to photograph normally refers to a punctual event, as well as the singular object (the sunset); however, the adverbial until implies duration. The conflict between the two different aspectual specifications causes the verb to be reinterpreted as referring to an iterative process. This repetitive effect is absent both from verbs referring to a durative activity such as in The tourist watched the sunset until nightfall and in the other uses (i.e., without the adverbial until) of the verb to photograph. The aspectual value of to photograph has been coerced by the durative adverbial. In cognitive semantics, this phenomenon is considered an effect of a more general principle: (grammatical) constructions have meanings distinct from those of words and these meanings interact with the meaning of the words (see Goldberg 1995).

(4) As they have waved us along the raised causeway and into the rocky cleft …

In this sentence, the particular interpretation of the predicate as “to signal permission to move to a place by waving” is produced by the so-called “caused-motion” construction applied to the verb to wave. Michaelis (2003) considers that there is a general override principle stating that “if lexical and structural meaning conflict, the semantic specifications of the lexical element conform to those of the grammatical structure with which that lexical item is combined”. This principle is illustrated by the interpretation of a sentence like They have good soups there. The nominal construction which licences the combination of a noun and a plural suffix –s requires that its nominal head denote a count entity. While soup, as a liquid, is prototypically viewed as a mass, the noun soup, when combined with the plural construction as here, receives the individual construal associated with count entities, and is thereby seen as denoting a portion or type (Michaelis 2003: 172). Semantic shift More generally, Talmy indicates that ‘when the specifications of two forms in a sentence are in conflict, one kind of reconciliation is for the specification of one of the forms to change so as to come into accord with the other form. This change of accommodation is termed a shift.’  (Talmy 2000: 324)

Talmy (2000: 324–336) distinguishes different types of shifts and also various other processes for resolving semantic conflicts (blends, juxtapositions, schema juggling). I would like to mention just one example of semantic shift which enabled me to represent the connections at work between the linear axis of the sentence and the depth dimension of language (Robert 2003).

 Stéphane Robert

When Balzac describes Eugenie Grandet as a poor rich heiress (une riche et pauvre héritière), the reader reinterprets the two contradictory adjectives either by displacing the contradiction temporally (she is potentially rich as a future heiress but is currently poor), or by giving poor a subjective reading (“unhappy”) instead of an objective one (“who isn’t rich”), i.e., by displacing the adjective’s meaning onto the modal plane. The reader thus carries out a change in the reference point which shifts the meaning of the adjective from one plane onto another. This reference change makes it possible for the meaning to follow another path in the depth of language. This semantic process can be schematized as in Figure 7:

language depth

utterance shifting Figure 7.  Shift in language depth.

4.3  Semantic layering Because of the polysemy of words and the “depth dimension” of language, in some utterances it is possible to activate several meanings of the same word. This sort of semantic layering is the mainspring of rhetoric, puns, and also of advertising, as Grunig (1990) has shown, from whom the following examples are taken. One example is the advertisement for a brand of pochettes d’emballage (wrapping bags): Ces sacs qui nous emballent, literally “Those bags that wrap us up”, which can also figuratively mean “Those bags which delight us”. The advertisement thus plays on the two meanings of the French verb emballer, which has a literal meaning, that of wrapping something up, and the figurative one of delighting someone, similar to the slightly different English figurative meaning of wrap as in They’re completely wrapped up in each other. A possible English rendering of the advertisement would be “Wrap yourself up in these bags”. A second example is an advertisement for an oven which runs Mettez-lui une grosse tarte, which means “Put a big pie in it” but also “Give it a big slap”. These phenomena of semantic layering can cover several terms: this entails several isotopics being constructed within a single sentence. The following advertising slogan, which actually pertains to a type of car, thus plays on a double isotopic (see Fig. 8): Quand je vois du trafic, je sors mon automatique, which can be almost directly translated into English by “When I see traffic, I take out my automatic”.



Words and their meanings  Quand je vois du trafic, je sors mon automatique contrebande (“contraband”) circulation (“traffic”)

revolver automobile

Figure 8.  Semantic layering (activation of two isotopics).

Note that the two meanings are not actually activated at once: it takes time for the (French) reader to realize that the intended meaning (car) is not the first one that came to mind (gun) (as the word trafic is not usually applied to driving conditions, contrary to English). In advertisements, the illustration often triggers the activation of the second meaning as is the case in this example. The second isotopic is certainly the least probable as the meaning of trafic for “too many cars” remains marginal in French, just as the nominalization of the adjective automatique (again, contrary to English), but this isotopic is activated by the illustration that accompanies the advertisement: the association between the universes of the two isotopics is probably not without psychological effects. Thus the two paths activated in the depth of language interact (Fig. 9): language depth

isotopic 1 isotopic 2

utterance Figure 9.  Activation of two paths in the language depth.

Another kind of semantic layering is produced by replacing a word in a set expression such as a proverb, the title of a movie or a famous song. One example is the advertisement for a cigarette brand called Kool: Some like it Kool which is a play on the title of the film Some like it hot. Another example is the advertisement for “Dim” hosiery: en avril ne te découvre pas d’un Dim, based on an alliterative French proverb warning against the sudden return of cold weather in springtime, en avril ne te découvre pas d’un fil “in April, don’t remove a stitch (of clothing)”. The insertion of a single term (Kool or Dim) in the utterance activates two utterances, the actual slogan and the backgrounded proverb, thereby creating layers of meaning with semantic interaction between the two utterances. Thus we can see that the end of the utterance is the privileged location for what I call “semantic bombs” whose effects are not additional as they induce phenomena of meaning restructuring, resonance, diffusion and layering: on the different non linear meaning factors, one may consult Robert (1999 and 2003).

 Stéphane Robert

5.  Conclusion Because of the absence of one-to-one relations between forms and meanings in language, linguistic units are by nature polysemous; furthermore they are caught up in a fabric of various associations (the language depth) and serve as representation triggers; lastly, linguistic units are semantically deformable: when they are inserted in an utterance, the verbal and situational contexts act upon their meaning. This plasticity of meaning in words makes for a functional optimality of linguistic systems by conferring upon them remarkable referential power and adaptability. It probably also plays a role of cognitive optimization through memory storing economy. This deformability of linguistic units comes nonetheless with an important drawback for communication as it generates ambiguities, sources of misunderstandings. It is then through the progressive construal of meaning over the whole utterance that the meanings of terms are stabilized, through relation processes which constantly intervene during discourse. But this meaning stabilization makes use of a construction dynamic and interpretative adjustments whose results are never guaranteed. Which shows that language is the seat of opposing forces which confer a particular power upon it, and where the speaker is at once the driver and the passenger.

References Allègre, C. 1995. La défaite de Platon ou la science du XXe siècle. Paris: Fayard. Arnaud, P.J.L. 1997. Les Ratés de la dénomination individuelle: Typologie des lapsus par substitution de mots. In C. Boisson & P. Thoiron (Eds), 307–332. Balbachan, F. 2006. Killing time: Metaphors and their implications in lexicon and grammar. metaphorik.de 10 (http://www.metaphorik.de/10/index.htm). Berlin, B. & Paul, K. 1969. Basic Color Terms: Their Universality and Evolution. Berkeley CA: University of California Press. Boisson, C. & Thoiron, P. (Eds). 1997. Autour de la dénomination [Travaux du C.R.T.T]. Lyon: Presses Universitaires de Lyon. Brown, C. 2005a. Hand and arm. In Haspelmath et al. (Eds), 522–525. Brown, C. 2005b. Finger and hand. In Haspelmath et al. (Eds), 526–529. Bruner, J. 1992. Acts of Meaning. Harvard MA: Harvard University Press. Chukwu, U. 1997. Les verbes ibo pour [acheter]. In C. Boisson & P. Thoiron (Eds), 71–106. Corbin, D. & Temple, M. 1994. Le monde des mots et des sens construits: catégories sémantiques, catégories référentielles. Cahiers de lexicologie 65(2): 5–28. Creissels, D. 2001. Auxiliarisation et expression de significations aspecto-temporelles en tswana, conference at the Rivaldi seminar, Paris, France, 14 juin 2001. Culioli, A. 1982. Rôle des représentations métalinguistiques en syntaxe [Collection ERA 642]. Paris: Université Paris 7. Culioli, A. 1990. Pour une linguistique de l’énonciation. Opérations et représentations, Tome I.  Paris: Ophrys.



Words and their meanings 

Deignan, A. 2006. The grammar of linguistic metaphors. In Corpus-Based Approaches to Metaphor and Metonymy, A. Stefanowitsch & S.T. Gries (Eds), 106–122. Berlin: Mouton de Gruyter. De Swart, H. 1998. Aspect shift and coercion. Natural Language and Linguistic Theory 16: 347–385. DeVelle, S. 2003. On-line effects of semantic coercion: Simple versus enriched compositional processing. Paper presented at the 14th Australian Language and Speech Conference. The University of Queensland. Fillmore, C. 1977. Scenes-and-Frames Semantics. In Linguistics Structures Processing, A. Zampolli (Ed.), 55–81. Amsterdam: North Holland. Fillmore, C. 1982. Frame semantics. In Linguistics in the Morning Calm. Linguistic Society of Korea (Ed.), 111–38. Seoul: Hanshin. Goldberg, A. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Chicago IL: The University of Chicago Press. Greimas, A.J. 1966. Sémantique structurale. Paris: Larousse. Grunig, B.N. 1990. Les mots de la publicité. Paris: Presses du CNRS. Haspelmath, M., Dryer, M., Gil, D. & Comrie, B. 2005. The World Atlas of Language Structures (WALS). Oxford: OUP. Honeste, M.L. 1997. De la dénomination aux stratégies argumentatives. In C.  Boisson & P. Thoiron (Eds), 279–306. Hunston, S. & Francis, G. 2000. Pattern Grammar: A Corpus-driven Approach to the Lexical Grammar of English. Amsterdam: John Benjamins. Kleiber, G. 1999. La sémantique du prototype. Catégories et sens lexical. Paris: Presses Universitaires de France. Kövecses, Z. & Radden, G. 1998. Metonymy: Developing a cognitive linguistic view. Cognitive Linguistics 9(1): 37–77. Lakoff, G. 1987. Women, Fire and Dangerous Things: What Categories Reveal About the Mind. Chicago IL: University of Chicago Press. Lakoff, G. 1993. The contemporary theory of metaphor. In Metaphor and Thought, A. Ortony (Ed.), 202–251. Cambridge: CUP. Lakoff, G. & Johnson, M. 1980. Metaphors We Live By. Chicago IL: The University of Chicago Press. Lakoff, G. & Núñez, R.E. 2000. Where Mathematics Comes From: How the Embodied Mind Brings Mathematics into Being. New York NY: Basic Books. Langacker, R.W. 1987 and 1991a. Foundations of Cognitive Grammar, 2 Vols. Stanford CA: Stanford University Press. Langacker, R.W. 1991b. Cognitive grammar. In Linguistic Theory and Grammatical Description, F.G. Droste & J.E. Joseph (Eds), 275–306. Amsterdam: John Benjamins. Lee, D. 2001 [2004]. Cognitive Linguistics. An Introduction. Oxford: OUP. Lichtenberk, F. 1991. Semantic change and heterosemy in grammaticalization. Language 67(3): 475–509. Lipka, L. 1990a. An Outline of English Lexicology. Tübingen: Niemeyer. Lipka, L. 1990b. Metaphor and metonymy as productive processes on the level of the lexicon. In Proceedings of the XIVth International Congress of Linguistics, W. Bahner, J. Schildt & D. Viehweger (Eds), 1207–1210. Berlin: Akademie-Verlag. Martin, W. 2001. A frame-based approach to polysemy. In Polysemy in Cognitive Linguistics [CILT 177], H. Cuyckens & B. Zawada (Eds), 57–82. Amsterdam: John Benjamins.

 Stéphane Robert Michaelis, L. 2003. Word meaning, sentence meaning, and syntactic meaning. In Cognitive Approaches to Lexical Semantics, H.  Cuykens, R.  Dirven & J.R.  Taylor (Eds), 163–210. Berlin: Mouton de Guyter. Ozanne-Rivierre, F. 1999. Spatial orientation in some Austronesian languages. In Language Diversity and Cognitive Representations, C. Fuchs & S. Robert (Eds), 73–84. Amsterdam: John Benjamins. Paprotté, W. 1998. Word sense disambiguisation. In Contrastive Lexical Semantics, E. Weigand (Ed.), 243–262. Amsterdam: John Benjamins. Pérennec, M. 1997. Le mot complexe en allemand. In C. Boisson & P. Toiron (Eds), 205–220. Pustejovsky, J. & Bouillon, P. 1996. Aspectual coercion and logical polysemy. In Lexical Semantics. The Problem of Polysemy, J.  Pustejovsky & B.  Boguraev (Eds), 133–162. Oxford: Clarendon Press. Rastier, F. 1987. Sémantique interprétative. Paris: Presses Universitaires de France. Robert, S. 1999. Cognitive invariants and linguistic variability: From units to utterance. In Language Diversity and Cognitive Representations, C.  Fuchs & S.  Robert (Eds), 21–35. Amsterdam: John Benjamins. Robert, S. 2003. L’épaisseur du langage et la. linéarité de l’énoncé: Vers un modèle énonciatif de production. In Parcours énonciatifs et parcours interprétatifs – Théories et applications, A. Ouattara (Ed.), 255–274. Paris: Ophrys. Roulon-Doko, P. 2003. Les parties du corps et l’expression de l’espace. In Perspectives synchroniques sur la grammaticalisation. Polysémie, transcatégorialité et echelles syntaxiques, S. Robert (Ed.), 69–84. Leuven: Peeters. Shipley, J.T. 1984. The Origins of English Words. Baltimore MD: Johns Hopkins University Press. Sinclair, J. 1998. The lexical item. In Constrative Lexical Semantics, E.  Weigand (Ed.), 1–24. Amsterdam: John Benjamins. Talmy, L. 1978. Figure and ground in complex sentences. In Universals of Human Languages, Vol. 4, J. Greenberg, C. Ferguson & E. Moravcsik (Eds), 625–649. Stanford CA: Stanford University Press. Talmy, L. 2000. Toward a Cognitive Semantics (I & II). Cambridge MA: The MIT Press. Traugott, E.C. & Hopper, P. 1993. Grammaticalization. Cambridge: CUP. Vandeloise, C. 1986. L’Espace en français. Paris: Le Seuil. Victorri, B. 1997. La polysémie: Un artefact de la linguistique? Revue de sémantique et pragmatique 2(2): 41–62. Vidalenc, J.L. 1997. Quelques remarques sur la métaphore comme outil de dénomination dans un corpus d’histoire des sciences. In C. Boisson, & P. Thoiron (Eds), 133–156. Yaguello, M. 1998. Petits faits de langue. Paris: Le Seuil.

The typology of semantic affinities Bernard Pottier

Université Paris Sorbonne, Institut de France, Fédération TUL Following previous research by the author, this chapter shows how semantic proximity (or proxemy) may result from meaning divergence in one lexical item (polysemy) or from meaning convergence of different lexical items (parasemy). Polysemous variations are explained by the interplay of several parameters: domains of instantiation, synaesthesic fields, and universal visualized cognitive schemas. In parasemy, a referential domain is conceived and uttered through several optional solutions, i.e., polysemiosis, within one and the same language as well as crosslinguistically (e.g., correspondences between proverbs in spite of distinct cultural environments). The paper shows how taxemic paradigms are expressed by roots, stems, classifiers, determiners, a compromise between the necessary multiplicity of lexical items and memory limitations. Keywords:  cognitive schema; conceptualization; mental imagery; noeme; parasemy; polysemiosis; polysemy; proxemy; taxeme

1.  Introduction The (relatively) recent interest of American linguists such as Fillmore & Kay (1996), Lakoff & Johnson (1980), Langacker (1990) or Talmy (2000) for cognitive research, as well as the European tradition initiated by Gustave Guillaume as early as 1920, which propose an indepth semantic reflection about grammatical mechanisms, and, later on, lexical mechanisms, put to the fore the prevalence of the mental level within the enunciative process. The studies of individual languages provide a very rich data of phenomena which, although specific to certain areas, disclose, nevertheless, general tendencies at a certain level of abstraction. Such is the case of semantic affinities between grammatical and lexical elements all along their historical evolution as well as in their synchronical interrelationships. In line with previous works (Pottier 1955, 1992, 1999, 2000), the present study offers a catalogue of the different mental processes at work in the domain of semantic affinities that can be a useful tool for crosslinguistic comparison in future research. Each process is named, described, visualized by use of dynamic graphs whenever possible, and exemplified.

 Bernard Pottier

“Semantic affinities” may result from a divergence (i.e., when a sign has several values, that is polysemy) or from a convergence (i.e., several signs share several common values, that is parasemy). This general phenomenon is called proxemy (or semantic proximity, cf. Gaume et al., this volume.), and can be represented and illustrated as follows (Fig. 1):

Si 1 Si

**

Sds POLYSEMY (of one sign)

Si 2 Si 3 PARASEMY (of several signs)

PROXEMY Signified relations top: of the stairs, of the milk, of a bottle

to spoil ~ to damage ~ to ruin

Figure 1.  Representation of proxemy.

2.  Semantic divergence or polysemy 2.1  Polysemy and homonymy Let us consider the utterance: Iˉputˉtheˉfileˉwithˉtheˉplane

The co-occurrence of two polysemous terms leads to a reciprocal selection of likelihood: file: “tool”ˉratherˉthanˉ“folder” plane:ˉ“tool”ˉratherˉthanˉ“tree”ˉorˉ“aircraft”

The semic affinities between these two terms, which lead to a likely interpretation, are called semic harmony or isosemy. It is interesting to note that the semic density plays a part in the phenomenon of polysemy in which the terms that are less dense and those that are very dense are the



The typology of semantic affinities 

most subject to variation. The first ones, because their intension is weak, have a great extension and the second ones which are full of imagery, are easily metaphorically transposed (Fig. 2): —

+

to go, to work, to put, to run

consult, perforate

my hat has gone to put an end colour that runs in the wash the lift isn’t working

burst, feed

her heart was ready to burst to feed the mind

Figure 2.  Representation of semic density.

Polysemy is called homonymy when the signifieds have no connected semantic relations (Fig. 3):

Si

file

kind of tool folder

Sd 1 Sd 2 plane

kind of tool tree

Figure 3.  Representation of homonymy.

When comparing the lexicon of several languages, one notices that words share a number of “semantic molecules” which correspond to the intersection of two semantic domains (cp. in François, this volume, where the term colexification is proposed, and the example of English straight and French droit).

2.2  The role of cultural habits Cultural habits are situated between the Universal and the Individual. The Universal refers to the mental imagery linked to nature’s phenomena, the human body, the “whole/part” relations, everyday objects, fundamental actions. The Individual depends on each person’s own history (even the existential traumas). The Cultural characterizes groups, and concerns colours, forms, beliefs, good and evil, life cycles. The prototypical representations, such as the essential forms of the moon or of a bottle, evoke cultural orthomorphy.

 Bernard Pottier

So do the connotative semes included in the signified of certain signs: “Friday the thirteenth” (ill-fated), “to have a yellow streak” (bad), “touch wood” or “knock on wood!” (good). They are latent elements of the virtueme of any sign: “to have two left feet” (bad), “touch wood!” (good).

2.3  The parameter of the domains of instantiation A morpheme (be it a lexeme or a grammeme) can alter according to the domains of instantiation it applies to. These variations are characterized by their use on a regular basis, e.g.,: Existential Spatial Temporal Notional Modal

theˉvaccineˉtook Johnˉtakesˉaˉbook itˉtookˉmeˉtwoˉhours toˉtakeˉaˉdecision IˉtookˉyouˉforˉanˉEnglishman

Iˉsawˉhimˉinˉperson inˉJapan inˉtheˉnight expertˉinˉeconomics inˉmyˉopinion

2.4  The parameter of synaesthesic fields This parameter concerns the numerous lexemes (often belonging to the adjectivisation category) whose values depend on the sensorial modality which is highlighted: Sight Smell Taste Hearing Touch

douxˉauˉregardˉ(fr.) unˉdouxˉparfum uneˉpommeˉdouce uneˉdouceˉmélodie laˉpeauˉdouce

(soft) toˉsmellˉsweet sweetˉcream sweetˉtoˉhear (soft)

2.5  Mental schemas Mental schemas, as cognitive studies have repeatedly shown, are an essential element of the semantic path which leads to polysemy. As I proposed in former studies,1 looking for more abstract representations than the highly iconic ones of the American cognitive tradition, four types of mental schemas (Fig. 4) can be distinguished, even though it is not always easy to oppose them. They are referred to as dominants, and they are useful for the understanding of the mental phenomena.

.  Many of the concepts already proposed in Pottier (1955) meet current research in cognitive linguistics: “schèmes représentatifs” is parallel to mental schemas, “mouvement” to trajector, “limite” to landmark, “point de visée” to viewpoint. For event conceptual schemas, see Pottier (1992).



The typology of semantic affinities  VISUEME

IDEEME

NOEME

SCHEMA

entity

abstraction

morphodynamic

event

bridge

middle

separate

modify

+ –

Figure 4.  Mental schemas.

2.5.1  The visueme A visueme is a mental representation which selects, among the semic constellation of an entity or its characterizations (properties, activities, location) a semantic feature meant to be metaphorical and considered to be salient. Cultural prototype is a particular instance among others, e.g.: Theˉhouseˉdoor theˉmiddleˉofˉtheˉcentury Toˉshutˉtheˉdoorˉonˉanyˉdiscussion I’mˉinˉtheˉmiddleˉofˉreading

In several writing systems, the pictogram of “water” (three undulating lines in parallel), makes us understand why a same sign can refer to “ water”, “tears”, “hair”, “grass”.

2.5.2  The ideeme The ideeme, as a notional concept, is a mental representation based on a typical abstraction that evokes properties, activities, general relations and can be represented as in Figure 5. ‘pend-’ pendant la séance (Fr.) la solución queda pendiente (Sp.) the pending authorization o nosso caso continua pendente (Ptg.) ‘from’ In several languages a same sign can mean ‘trunk’, ‘beginning’, ‘source’, ‘because’

‘duality’ In several languages polysemies are mentioned: ‘two’, ‘opposed’, ‘rival’, ‘companion’ can be expressed by a same sign (cf. Sakhno & Tersis, this vol.). Fr. ‘contre l’amiral’ (as opposed to), ‘contre-amiral’ (‘next to’ in military rank) nú

‘active part’ nú (Gbaya Roulon p.c.)

Figure 5.  Representation of ideemes.

man’s mouth edge of knife selvage of woman’s wrap point of needle

 Bernard Pottier

2.5.3  The noeme The noeme is a component of a noemia, or a kinetic and often a dynamic three phase mental schema (the trimorph) of the morphology of an event. It is useful for numerous instantiations, whether lexemic or grammemic, such as in the following examples (Fig. 6): 1

2

3

1 to enter

2 3 to stay to leave

Figure 6.  Representation of noemes.

Below in Figure 7 are examples of noemes and their metaphorizations: frôler (Fr.) ‘to skim’

S: frôler un mur (to brush against a wall) N-M: frôler le ridicule (to border ridicule) T: frôler les 10 secondes au 100 mètres (to run a 100 metres in under 10 seconds)

to catch

S: to catch something (‘grasp’) N-M: to catch the wind (‘benefit’) M: I caught a few words (‘understand’)

+

Figure 7.  Examples of noemes and their metaphorizations.

2.5.4  The analytic schema The analytic schema (AS) is a mental representation of a conceived event, from the most simple to the most complex, which combines noemes. The following are some graphic representations together with the corresponding examples (Figure 8): A

A+

>>>> a

b (A turns red) a to become b

====== B A to cut B

A

A

A to wish ([A] to leave) A+

A A to drink a beer A to drink in somebody’s words

A to eat a cake A to eat one’s words

Figure 8.  Examples of analytic schemas.

The following example from Latin: pugnas bibit aure vulgu (bibit = “drinks”) “the crowd is all ears to the battle narratives” is also an illustration of the left part of the above graphic representation.



The typology of semantic affinities 

The difference between the above mentioned two types of activities can be characterized as follows: ReceptiveˉpositionˉofˉA

ActiveˉpositionˉofˉA+

The Chinese word miè: “turn off, eliminate, be stifled, erase the traces, annihilate …” reminds Portuguese apagar which means “erase the blackboard” as well as “turn off the light” or “eliminate a memory” (Fig. 9): A+

>> /existence/ → /non-existence/ Figure 9.  The analytic schema of Chinese miè and Portugese apaga.

3.  Semantic convergence or Parasemy 3.1  From referentiality to signs: Polysemiosis When the enunciator uses a referent ℜ (entity or event), it undergoes a process of conceptualization (Co). Conceptualizations may be numerous and therefore may lead to choosing, in an utterance or a chain of utterances, several signs, successively, for a same referent (coreferentiality). We deal here with polysemiosis. Polysemiosis may concern the lexicon, leading to cases of polynomies as in the following example: “It’sˉaˉsortˉofˉbrochure,ˉofˉleafletˉaboutˉtheˉcityˉmainˉmonuments” or it may concern different syntactic constructions, named polysyntaxies, for equivalent conceptualizations as in: “-ˉLook!ˉItˉstoppedˉraining. -ˉFinally! -ˉYes,ˉitˉisˉnotˉrainingˉanymore”

All co-referential signs in a particular discourse form some sort of a paradigm that corresponds to the parasemiosis phenomenon (Fig. 10).

3.2  Semiotization Semiotization corresponds to the need to give a name to conceptualized referents. “Immediate”, spontaneous naming, for instance when we call dog an animal accompanying a blind person to guide him, shows a relation of orthonymy.

 Bernard Pottier Coa

Sign a

Cob

Sign b

ℜ

PARASEMY

POLYSEMIOSIS Figure 10.  Representation of polysemiosis.

Some parasemies, i.e., when several signs have several common values, are every day rhetorical figures (including tropes), such as metonymy (London rejects the treaty), metaphor (the elder daughter of the Church stands for France, land of the free for America), hyperonymy and hyponymy (this dog is a mammal whose breed is the spaniel), or peronymy (the baker’s oldest son). For the receiver, in a given environment (situation, context, co-text), if a signified appears to be “immediate”, it is referred to as orthosemy (London for the city where the Tower Bridge is located). Any other intention causes a metasemy (London for the U.K.). Parasemy is clearly explained by the polynomy of a noeme or of a conceptual schema: to spring up, to emit, elocution, to talk ‘impassable’ opaque, waterproof, watertight, airtight (“not allowing x, y, z to pass”)

to come close, to near, to skim to skim through a book to come very close to the 5 meter mark (sports) the country is nearing disaster; my book is nearing completion “il était difficile de passer de l’un à l’autre sans friser une chute qui pouvait entraîner dans l’abîme commun” (G.Sand, in Frantext) ([N.d.T.] “it was difficult to go from one to the other without being on the verge of falling into the common abyss”.) Figure 11.  Polynomy of a noeme.

The same phenomenon also occurs with grammatical elements: Descartes

→ x (x) accordingˉto Descartes; selonˉDescartes,ˉd’après Descartes, suivantˉDescartes,ˉàˉlaˉsuiteˉdeˉDescartesˉ(Fr.)



The typology of semantic affinities 

In the context of dictionaries, parasynonymic series can be found, either in dictionaries of synonyms, or in monolingual dictionaries that have cross references (i.e., to get damaged, to go bad, to spoil, to ruin), or in bilingual dictionaries where there is an attempt to grasp the semantic field of the word to be translated by using several terms in the target language. Signs whose signifieds have in common one conceptual element show a common parasemy as in e.g., to grow, to increase, to raise. This common parasemy could be written as follows: /goes towards the +/. In certain constructions, there is a possibility to alternate between lexemes: anxietyˉgrows/raises/increases,

while in others, the alternation is not accepted or hardly: Peterˉgrowsˉ(*increases) Theˉpricesˉareˉincreasingˉ(*areˉgrowing)

In French, the term voire functions as a semantic reagent of signified affinity and organizes terms on a progressive axis, similarly to “even” in the English translations: ilˉcomprend l’admiration séculaires

voire voire voire

approuveˉ(heˉunderstands,ˉheˉevenˉapprovesˉofˉit) l’éblouissementˉ(admirationˉevenˉbedazzlement) millénairesˉ(centennialˉevenˉmillennial)

This use is comparable to that of “pour ne pas dire” (“if not to say”) in the following examples: “avarié, pour ne pas dire pourri” (spoilt, if not to say rotten), “passionné, pour ne pas dire exalté” (passionate if not to say elated). A similar case is found in English with “indeed”: I’m astounded, indeed disgusted I feel that he is right, indeed I know he is He was happy, indeed delighted, to hear the news.

The parasemic path from parasemy (also known as parasynonymy) to synonymy can be represented as follows, where Signified 1 is included in Signified 2. Si 1

Si 2

⊃

Sd 1

Sd 2

Parasemy ‘to spoil ~ to damage’

Figure 12.  Representation of parasynonymy.

Synonymy French = franco-(phone)

 Bernard Pottier

3.3  The fields homology: Co-hyponymy and co-semy When a concept is applied to different domains, the selected terms are in a co-hyponymic relationship with the concept because they share, at least, the nuclear content of this concept. Thus,//to bring something back into its initial (good) state//which corresponds to a minimal analytic schema, can be formulated by the metaterm (ad-hoc hyperonym) //REPAIR//, and be realized, in the language, by co-semic lexemes such as: toˉcure toˉmend toˉrestore toˉresole ______ _______ ________ _______ MED. SEWING BUILD. SHOE

toˉbrightenˉup _______ COLOR

In the same way, the concept of//FILIATION//(an element stemming from a whole keeps some of the properties of the whole) may be formulated in different languages such as: “telˉpère,ˉtelˉfils”ˉ(Fr.) “deˉtalˉpalo,ˉtalˉastilla”ˉ(Sp.) “suchˉasˉtheˉtreeˉis,ˉsuchˉisˉtheˉfruit”

The /a little while ago/ concept is found in various adjectival constructions such as: Freshˉpaint,ˉfreshˉnews,ˉfreshˉtroops Newˉborn,ˉtheˉnewˉrich Inˉrecentˉyears,ˉhisˉmostˉrecentˉbook

Furthermore, the fields’ homology can be very abstract. It can be based on a synaesthesic backgroup, more or less clear depending on the culture, as is the case in the Taoist philosophy which contrasts two complementary principles: YANGˉ(+):ˉtheˉsun,ˉhard,ˉdry,ˉstable,ˉawaken,ˉman,ˉreality,ˉwhite,ˉdayˉ… YINˉ(–):ˉtheˉmoon,ˉsoft,ˉhumid,ˉvariable,ˉasleep,ˉwoman,ˉdream,ˉblack,ˉnightˉ…

3.4  Paradigmatic parasemy: The taxeme The taxeme is an element of sense recurring over a large paradigm whose forms are related by a minimum of semantic affinities. Crosslinguistically and within one language, the taxeme can be expressed by different forms. a. The taxeme can be represented by a root or a stem, what I call the morphosemy phenomenon. Such is the case of the consonantal roots in Semitic, such as the root KTB in Arabic: KaTaBa: “heˉwrote” KiTāB: “book”

KāTiB: “secretary,ˉwriter” ma-KTūB: “written”

and of the stem in many Indo-European languages: read,ˉreader,ˉreading,ˉreadable,ˉreadership



The typology of semantic affinities 

b. The taxeme can also be a word, which acquires a generic value when used in compound forms: open-:ˉopen-door,ˉopen-handed,ˉopen-hearted,ˉopen-necked,ˉopen-plan

In Chinese, the compound words clearly reveal the intersections of the signifieds. e.g., duì is part of the composition of a whole series of forms which have in common the concept of “duality”: duìchènˉ“symetric” duìdĕngˉ“equivalent” duìfāngˉ“adversary” duìhuàˉ“toˉenterˉintoˉdialogue” duìmiànˉ“across” duìshŏuˉ“rival” duìliánˉ“distich” duìzhàoˉ“toˉcompare”

c. The taxeme can also be a classificatory element, i.e., classifier, determiner, which applies to a whole series of lexemes. Chinese, numerous African and Amerindian languages are well known examples. A similar phenomenon occurs in European languages when the introducer of series of entities is selected as in the following examples: aˉbundleˉof:ˉfiles,ˉletters,ˉtickets,ˉsticks,ˉclothes,ˉnewspapers aˉwadˉof:ˉdocuments,ˉbanknotes,ˉtobacco aˉsliceˉof:ˉbread,ˉsausage,ˉmeat,ˉcake,ˉcredit,ˉluck,ˉaction,ˉfilm,ˉplay,ˉbook

In the individual and discursive realizations, a great variation can appear. In French, for instance, to indicate a very small quantity of concrete or abstract entities, the word zeste (zest) is attested in Frantext (http://www.frantext.fr/) in combination with the following words: citron,ˉfierté,ˉraison,ˉpassé,ˉdétritus,ˉcouperoseˉ(lemon,ˉproud,ˉreason,ˉpast,ˉlitter, blotches)

and soupçon (hint) with: rouille, réserve, détachement, désir, pleurésie, sourire, vitriol (rust, reserve, detachment, desire, pleurisy, smile, vitriol)

Both are found, together with larme (tear), in the following excerpt from Françoise Chandernagor’s novel L’Allée du Roi (1981; p. 293): “Ninon disait trouver elle-même dans ses “caprices”: un soupçon de désir, un zeste de plaisir et, au pis, une larme de repentir”.2

.  [N.d.T.] “Ninon said that even she could find in her own “whims”: a hint of desire, a zest of pleasure and, at the worst, a drop of repentance”.

 Bernard Pottier

A suffix can determine a specific characteristic of the lexeme: Sp.:

-adaˉ(cuttingˉobjects):ˉcuchillada,ˉestocada,ˉlanzada -azoˉ(bluntˉobjects):ˉcodazo,ˉmartillazo,ˉlatigazo

The capital letter at the beginning of proper names can be considered to be a class mark as well: Rose, Martin (a rose, a martin) and, without any possible opposition, Manchester or Cyprus. d. The above-mentioned Arabic root in (a) suggests a morphemic motivation of the invariable element KTB for a whole semantic field. Conversely, this semantic field is expressed in English by words without any formal link such as write, book, secretary, office, library. Such a phenomenon may be called lexemic dissemination which can be considered as an implicit taxeme, well perceived by the enunciator. French offers a good example of dissemination for the taxeme which alludes to /the sense of sight/. /∢/: v oir, regarder, miroir, lunette, télescope, œil, ophtalmologiste, loupe, observer, optique, aveugle, spectateur, scruter … (to see, to look, mirror, glasses, telescope, eye, ophthalmologist, magnifying glass, to observe, optical, blind, spectator, scrutinize …)

This is not mere speculation: we can notice that in the texts, there is a cooccurrence of many of these terms, as in the following example from Frantext: “Le médecin des yeux eût rougi de s’appeler œilliste (…); déjà la qualification d’oculiste, insuffisamment barbare, humilie ses prétentions: il est ophtalmologue”3  (Remy de Gourmont). “Ce regard jeté à la loupe sur mes moindres détours de pensée, cette scrutation (P. Bourget, Le Disciple 1889; p. 85). continue de mon être le plus caché …”4

4.  Conclusion Even though reflection and thought are built on theoretical reasoning, it is only when languages are embodied in texts that the detailed study of examples throws light on and justifies the initial hypotheses.

.  [N.d.T.] “The eye doctor would have been ashamed of being called eyeist (…); the designation oculist, not barbarian enough, already is a humiliation for his pretensions: he is ophthalmologist.” .  [N.d.T.] lit. “My thoughts being looked through a magnifying glass in the smallest detail, this continuous scrutinizing of my deepest inner self …”



The typology of semantic affinities 

The operations of semantic choice concern both the enunciator and the interpreting receiver, and these are constructed through mental schemas, several of which can be visualized on graphs. These schemas are approximations that can be compared with the ideograms of several ancient or modern writing systems, the catastrophe schemas in mathematics (Thom 1974), or the symbolic forms of various cultures. The specificity of each individual language does not contradict the supposedly universal thought mechanisms. Semantic typology is ground in the to-ings and froings between these two extreme poles.

References Base textuelle Frantext, (http://www.frantext.fr/) Fillmore, C.J. & Kay, P. 1996. Construction Grammar. Ms, University of California at Berkeley, Department of Linguistics. Guillaume, G. 1919. Le problème de l’article et sa solution dans la langue française. Paris: Hachette. Lakoff, G. & Johnson, M. 1980 [2003]. Metaphors We Live By. Chicago IL: University of Chicago Press. Langacker, R.W. 1999. Grammar and Conceptualization. Berlin: Mouton de Gruyter. Pottier, B. 1955. Systématique des éléments de relation. Paris, thèse. (Also 1962; Paris : Klincksieck) Pottier, B. 1992. Sémantique générale. Paris: PUF. Pottier, B. 1995. Le cognitif et le linguistique [Acta Romanica Basiliensia, ARBA 3], Linguistique et modèles cognitifs: 175–199. Pottier, B. 1999 [1997]. Mental activities and linguistic structures. In Proceedings of the 16th International Congress of Linguists, Paris 20–25 July 1997, B.  Caron (Ed.), paper No. 7. Oxford: Pergamon (Elsevier Science, CDRom). Pottier, B. 2000. Représentations mentales et catégorisations linguistiques. Louvain: Peeters. Talmy, L. 2000. Toward a Cognitive Semantics. Cambridge MA: The MIT Press. Thom, R. 1974. Modèles mathématiques de la morphogenèse. Paris: Union Générale d’Editions.

Cognitive onomasiology and lexical change Around the eye* Peter Koch

University of Tübingen Using the methodology of diachronic cognitive onomasiology, as developed in two projects at Tübingen University, the study discusses polygenetic semantic parallels in semantic change, focussing on those that are due to fundamental cognitive constants. The cognitive and formal relations between a source and a target concept are identified through a two-dimensional grid. The approach is exemplified for the semantic domain of eye (eyelash, eyebrow, eyelid, and eyeball). The study provides a list of all the cognitive solutions to create lexical innovations chosen in the language sample. Together with cultural and linguistic categorization, it also explains the different options chosen by the languages for lexical conceptualisation and gives insight to the ongoing debate on linguistic relativity. Keywords:  body parts; cognition; contiguity; frame; metaphor; metonymy; onomasiology; polygenesis; semantic change; semantic parallels; typology

1.  Theoretical and methodological preliminaries 1.1  The search for semantic parallels Cognitive semantics has not only given a fresh impetus to synchronic, but also to diachronic linguistics, in so far as cognitive approaches to the description of metaphor, metonymy, subjectification, etc. shed new light on well-known problems of semantic change (cf. Blank & Koch 1999). As Lakoff & Johnson (1980) and Traugott (cf. recently Traugott & Dasher 2002) have shown mainly for vocabulary, and as Heine & Kuteva (2002) have shown mainly for grammar, but in part also for vocabulary, certain paths of change are of particular interest, especially those which are followed again and again when it comes to semantic change in language, and which therefore seem to point

*I would like to thank Martine Vanhove for her helpful suggestions as well as Sam Featherston for the stylistic revision of this paper.

 Peter Koch

to constant cognitive factors. Presumably these factors intervene in linguistic change like an invisible hand (in the sense of Keller 1994) and time and again produce similar results in a polygenetic fashion. If this were true, it would be much easier to predict lexical change (cf. Koch 1997; 2000: 75–81, 89–92; 2001a: 8–17, 25–31; 2003: 154–162, 2005a; Blank 2003). My paper is intended to show how diachronic cognitive onomasiology proceeds in this domain and what insights it offers. From both a theoretical and a methodological point of view, it is first of all necessary to discuss several questions: – – – – –

What would the results have to look like if the term “polygenesis” is to be applied legitimately (see immediately below and 1.2.)? Will our approach be semasiological or onomasiological (1.2.)? On the basis of what kinds of data can our hypotheses be checked (1.3)? What is lexical change (1.4.)? What kind of lexicological model are we to adopt (1.5.) to account for a realistic conception of lexical change (1.4.)?

When we began to study the designations of parts of the human body, we first established a project limited to Romance languages (decolar)1 and intended to analyse 14 languages or varieties in total in order to document the cognitive types present in the Romance area as completely as possible. Our second project, LexiTypeDia,2 is based on a worldwide sample of languages, and here, within the domain head, we describe designations of body parts with a different aim in mind: We check for semantic parallels in languages all over the world. According to our hypothesis, semantic parallels between languages may be due to fundamental cognitive constants (a). However, it is obvious that semantic parallels may also be triggered either by genetic kinship of languages (b) or by linguistic and cultural contact (c). Especially in case (c), we suppose semantic parallels to show a significant areal distribution. Since it is rather trivial to find semantic parallels in cognates belonging to genetically related languages (b), and since it is quite natural to find them in languages in contact (c), we have to radicalize our starting hypothesis: Semantic parallels arouse

.  The DECOLAR sample comprises the following Romance languages/language varieties/ language states: Catalan, Engadinian, Old French, Modern French, Friulian, Galician, Italian, Ladin, Occitan, Portuguese, Romanian, Sardinian (Campidanian), Sardinian (Logudorian), Spanish. For DECOLAR cf. Blank et al. (2000); Gévaudan et al. (2003). .  The data of LexiTypeDia will be presented and interpreted in Steinberg (in prep.). For LexiTypeDia cf. Koch & Steinkrüger (2001); Koch (2003); Mihatsch & Steinberg (2004) (especially Koch 2004a; Mihatsch & Dvořák 2004).



Cognitive onomasiology and lexical change 

our interest only in so far as there is a chance of these having been triggered by cognitive constants and of being independent of linguistic kinship as well as of language contact. So we have to search for semantic parallels that are likely to be polygenetic in nature. In order to find this kind of results, we have to stick to a rigorous methodology, as described in 1.2.–1.5.

1.2  Onomasiology Our starting point has to be an onomasiological one. Onomasiology is like a sieve filtering out everything that corresponds to a pre-established criterion – the fact of designating a given concept – without our being able to manipulate the results. So we have to accept everything that is filtered out, whether it confirms our hypothesis or not. Onomasiology has a second advantage: It enables us to discover material that is interesting independently of any etymological relationship. In this way, we can postulate a potentially polygenetic evolution even within one and the same language family: (1) a. b. c.

LateˉLat.ˉciliumˉeyelashˉ