(Cognitive Linguistic Research) Andreas Blank, Peter Koch - Historical Semantics and Cognition (1999, Mouton de Gruyter) PDF

Cognitive Linguistics Research
W
DE
G
MOUTON
DE GRUYTER
W
DE
G
MOUTON
DE GRUYTER
Historical Semantics
and Cognition
Andreas Blank
Peter Koch
(Editors)
W
DE
G
MOUTON
DE GRUYTER
Historical Semantics and Cognition
1749 ~
!.f
~
Qj
~
~
~ 1999
13
Editors
Rene Dirven
Ronald W. Langacker
John R. Taylor
Mouton de Gruyter
Berlin . New York
Historical Semantics
and Cognition
Edited by
Andreas Blank
Peter Koch
Mouton de Gruyter
Berlin' New York 1999
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Waiter de Gruyter GmbH & Co. KG, Berlin
@ Printed on acid-free paper

which falls within
the guidelines of the ANSI
to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data
Historical semantics and cognition I edited by Andreas Blank,

Peter Koch
p. cm. - (Cognitive linguistics research; 13)
Includes bibliographical references and index.
ISBN 3-11-016614-3 (cloth; alk. paper)
I. Semantics, Historical - Psychological aspects. 2.
Cognition. I. Blank, Andreas. H. Koch, Peter, 1951- .
Ill. Series.
P325.5.H57H48 1999
401' .43-dc21 99-32695
CIP
Die Deutsche Bibliothek - Cataloging-in-Publication Data
Historical semantics and cognition I ed. by Andreas Blank ; Pe-

ter Koch. - Berlin; New York: Mouton de Gruyter, 1999
(Cognitive linguistics research; 13)
ISBN 3-11-016614-3
© Copyright 1999 by WaIter de Gruyter GmbH & Co. KG, 0-10785 Berlin
All rights reserved, including those of translation into foreign languages. No part of this book
may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopy, recording, or any information storage and retrieval system, without
permission in writing from the publisher.
Printing: Wemer Hildebrand, Berlin
Binding: Liideritz & Bauer, Berlin
Printed in Gennany
Preface
The papers collected in this volwne evolved from a symposiwn that

was held September 19-21, 1996, at the "Clubhaus" of the Freie
Universitat Berlin. The symposium was organized with the double
intention of providing a forum in which synchronically and dia-
chronically oriented scholars would have to exchange their ideas and
where American and European cognitive linguists would be con-
fronted with representatives of different directions in European
structural semantics. While the confrontation indeed happened as
planned, the expected synergetic effects were perhaps not as inten-
sive as we had hoped. However, we are convinced that some of the
discussions we had will bring long-term results, thanks to the op-
ponents' modified perception of each other generated by this en-
counter.
We would like to express our gratitude to the "Au6enamt" of the
Freie Universitat Berlin for all its various forms of support, and es-
pecially to the Volkswagen-Foundation, without whose grant this
symposium would not have been possible.
All the work, the preparations including the program and the
schedule of meetings, the duplication and distribution of hand-outs
and papers, as well as the organizing of coffee-breaks, restaurants,
accomodations and transfer from airports to hotels, could not· have
been done without a devoted team of co-workers. We take this op-
portunity to thank once again Mary Copple, Genevieve Gueug, Paul
Gevaudan, Richard Waltereit and especially Sigrid Kretschmann
whose experience and readiness were an enormous support and con-
tributed to the success of the symposium.
Ideas of how the proceedings could best be published were dis-
cussed during the Berlin symposiwn itself. Due to changes in both
our academic affilations, some time went by until it was decided that
a greater part of the papers read at the Clubhaus should be published
in a volume rounded off with two articles that fit the volume's the-
vi Preface
matic framework better than the papers originally presented in Ber-

lin. A lot of work by Cinzia Cazzaro, Mary Copple, Angela Dom,
Cristina Fossaluzza, Keith Myrick, Eberhard Matt and Alexandra
Twardy went into elaborating the decisive version of the book.
Finally, we would like to express our gratitude to all of the contri-
butors for their comprehensive cooperation, to the editors of the
Cognitive Linguistics Research series and to Anke Beck of Mouton
de Gruyter.
Marburg / Tubingen Andreas Blank / Peter Koch

Contents
Introduction: historical semantics and cognition 1

Andreas Blank and Peter Koch
Section I: Theories and Models
Cognitive semantics and structural semantics 17

John R. Taylor
Diachronic semantics: 49
towards a unified theory of language change?
Helmut Ludtke
Why do new meanings occur? A cognitive typology 61

of the motivations for lexical semantic change
Andreas Blank
Diachronic prototype semantics. A digest 91

Dirk Geeraerts
Cognitive semantics and diachronic semantics: 109

the values and evolution of classes
Franfois Rastier
Section ll: Descriptive categories
Losing control: 147

grammaticization, subjectification, and transparency
Ronald W Langacker
viii Contents
The rhetoric of counter-expectation in semantic change: 177

a study in subjectification
Elizabeth Closs Traugott
Synecdoche as a cognitive and communicative strategy 197

Brigitte Nerlich and David D. Clarke
Laws of thought, knowledge and lexical change 215

Beatrice Warren
Section DI: Case studies
Intensifiers as targets and sources of semantic change 237

Ekkehard Konig and Peter Siemund
Cognitive ease and lexical borrowing: 259

the recategorization of body parts in Romance
Thomas Krefeld
Cognitive aspects of semantic change and polysemy: 279

the semantic space HA VE/BE
Peter Koch
List of contributors 307
Index 309
Introduction:
historical semantics and cognition
Andreas Blank and Peter Koch
1. General survey
Cognitive linguistics has had considerable influence on the develop-

ment of theories and methods of description in semantics (cf. Lakoff
1987; Langacker 1987/90; Taylor [1989] 1995; Kleiber 1990; Unge-
rerlSchmid 1996). Nowadays, even manuals of historical linguistics
refer to issues in cognitive research relevant to problems of diachro-
ny.) Indeed, some of the favourite subjects of cognitive semantics
(metaphor, metonymy, polysemy etc.) deal precisely with the syn-
chrony/diachrony-interface. In our opinion, investigation of dia-
chronic problems can, in turn, sharpen our view for fundamental se-
mantic processes and should therefore be able to advance theorizing
in cognitive linguistics. In this sense, historical semantics is an ideal
testing ground for semantic models and theories, as cognition and
our basic human conceptual system are highly involved in lexical
and grammatical change. The authors of this volume approach the
synchrony/diachrony-interface from various theoretical points of
views and apply or develop different conceptions of cognitive lin-
guistics.
1.1. The first group of articles deals with fundamental theoretical

issues in synchronic and especially diachronic linguistic description.
John Taylor discusses the foundations and basic issues of cogni-
tive semantics in contrast with European structural semantics, as it is
paradigmatically represented by the work of Eugenio Coseriu. The
central point of this controversy is the question of whether it is use-
ful and efficient to distinguish encyclopedic semantic structures from
internal, language-specific semantic structures.
2 Andreas Blank and Peter Koch
On the ground of his more general model of linguistic change (cf.

LUdtke 1980; 1986), Helmut Liidtke studies a number of cases of
semantic change in lexicon and grammar, in order to demonstrate the
cognitive linkage of the different levels of language on which change
can occur.
Andreas Blank discusses traditional classifications of the moti-
vations for lexical semantic change and develops a comprehensive
typology of these motivations on the basis of recent issues in cog-
nitive as well as in modem diachronic linguistics.
Dirk Geeraerts focuses on two major topics in his "diachronic
prototype semantics" (cf. Geeraerts 1997): i) the mapping of dia-
chronic semantic processes for several aspects of the protypical
structure of categories (e.g. typicality, family resemblance, blurred
edges, importance of encyclopedical knowledge), and ii) the typolo-
gy of motives for lexical change based on speaker-oriented or hear-
er-oriented strategies aimed at increasing either communicative effi-
ciency or expressivity.
Fran~ois Rastier reflects upon the epistemological status of the
definition of a prototype as the "best" representative of a category -
especially with regard to the valorization of the prototype by the
speakers. He interprets certain types of semantic change as a dis-
placement of "evaluative thresholds" dependent upon social values
and practices.
1.2. The second group of contributions develops categories for the

linguistic decription of diachronic processes.
By analyzing examples taken from different word classes, Ron-
aid Langacker describes several semantic processes whose com-
mon denominator is the gradual change from physical movement to
a merely virtual movement in the speaker's mind (e.g. Engl. The
mailbox is across the street; I'm going to sing). The resulting attenu-
ation of the semantic aspect [control] in the meaning of linguistic
entities is what Langacker calls "subjectification".
The same term is defined in quite a different way by Elizabeth
Traugott in her study of the semantic development of Engl. in fact:
Introduction 3
"subjectification" in her understanding is the rise of a new sense

from pragmatic inferences in typical discourses ("pragmatic
strengthening"). In contrast with the older sense, the new one fo-
cuses on the subject of a discourse because either subjective valua-
tions are emphasized or because the new sense has acquired a prag-
matic function at the speech-act level itself.
Brigitte Nerlich and David Clarke elaborate a number of criteria
to distinguish the traditional, but usually not well defined trope "syn-
ecdoque" from "metonymy" and "metaphor". They further explore
the cognitive background of synecdoque, as they have defined it, as
well as its rhetoric, pragmatic and semantic potential in synchrony
and diachrony.
Beatrice Warren introduces a model for the contextualization of
word-meanings based on semantic and encyclopedic knowledge. On
the ground of this model, she develops three major types of semantic
innovation called "novel hyponymic senses", "non-literal senses"
and "appended senses".
1.3. In the third group, theoretical options and categories related to

cognitive approaches are applied to describe selected diachronic phe-
nomena.
Ekkehard Konig and Peter Siemund explore the main cognitive
strategies for conceptualizing and verbalizing "intensifiers" in a
great number of languages as well as the semantic development of
intensifiers into genuine reflexive pronouns.
Analyzing the changes in conceptualization of the human body
and the limbs of the body from Latin to Romance, Thomas Krefeld
retraces the passage from the Latin model with "overlapping" deno-
minations to a clear-cut torso-extremities-model in the Romance lan-
guages. The latter seems more natural from a point of view of
Gestalt theory.
Starting from basic conceptual distinctions in the "semantic
space" HA VE/BE, Peter Koch detects typical paths of change in this
area. Certain patterns of metonymy, metaphor and semantic exten-
sion seem to occur polygenetically in different languages and thus
reveal modes of how we can conceptualize fundamental relations

like POSSESSION, EXISTENCE, and LOCATION.
In each of the contributions to this volume, fundamental topics of

cognitive linguistics (cf. section 2) are in some way connected to
recent issues in diachronic linguistics or pragmatics (cf. section 3).
2. Cognitive models and approaches
2.1. European structural semantics has pleaded for a strict theoretical

separation of encyclopedic knowledge from language-specific se-
mantic features and has determined the latter to be the only object of
linguistic semantics. In contrast to this, cognitive linguistics has
strongly emphasized the importance of encyclopedic knowledge for
semantics. Indeed, certain phenomena that are relevant to linguistic
theory and description cannot be explained on the level of intralin-
guistic regularities as, e.g., the "associative anaphor" in (1), which
the hearer can only interpret against the background of his world
knowledge:
(l) We arrived at the village. Unfortunately, the church was

closed.
The papers brought together in this volume show that it is neces-

sary to partially or even entirely anchor diachronic studies in ency-
clopedic knowledge. While some authors do not discuss this pro-
blem explicitely, others claim that semantic knowledge is exclusive-
ly extralinguistic (4 Langacker and esp. 4 Taylor).2 4 Geeraerts
clearly gives priority to the encyclopedic knowledge, but neverthe-
less recognizes the relevance of intralinguistic semantic facts. ~
Blank emphasizes the overall importance of encyclopedic knowl-
edge for semantic change, but also accounts for changes induced by
intralinguistic constellations. On the one hand, ~ Krefeld highlights
diverging segmentations of the human body in different languages
Introduction 5
(Latin vs. Romance), on the other hand, his analysis is rooted in fun-
damental anthropological and gestaltist categories. While observing
very accurately intralinguistic semantic factors, ~ Rastier neverthe-
less stresses the importance of social values for semantic change.
In our opinion, linguists should not renounce completely the dis-
tinction between encyclopedic aspects of meaning and intralinguistic
semantic features. It is true that intralinguistic features are not sub-
stantially different from encyclopedic information, but they have ac-
quired a categorially different status, insofar as they reflect semantic
oppositions that in some languages are expressed by a simple
lexeme, while other languages either have recourse only to a com-
plex word or a paraphrase or even simply cannot realize them at all. 3
Divergent semantic structures of this kind must be interpreted as
emanating from cognitive constellations, because the diversity of
pragmatic and social relevance and the resulting differences in the
profiling of a concept determine the linguistic strategies used by the
speakers of one language. Thus, distinguishing intralinguistic from
encyclopedic knowledge opens a new field of research to cognitive
semantics, esp. with regard to cross-linguistic (and to "cross-cultur-
al") studies.
2.2. The verbalizing of extralinguistic entities is always related to the

problem of categorization. According to the framework of cognitive
linguistics, categories have a prototypical internal structure and their
external hierarchical relations show a different cognitive profiling
(superordinatelbasic/subordinate level). It now appears that dia-
chronic semantic processes often involve questions of categorization
and of prototipicality (cf. Geeraerts 1997; Koch 1995, 1996; Blank
1997). For example, the phenomenon of semantic change can be un-
derstood as the immediate corollary of the blurred boundaries of pro-
totypically organized categories (~ Geeraerts). From a different per-
spective, prototypical conceptual constellations are viewed as neces-
sary - but not sufficient - conditions for certain types of semantic
changes (~ Blank). Indeed, as demonstrated by ~ Koch,
metonymies and metaphors operate on a prototypical view of source
and/or target domains. While most work in cognitive linguistics

takes prototypicality for granted, ~ Rastier raises the question of
how new prototypes in language emerge.
2.3. Information relevant to meaning organizes not only in categories

- be they prototypically structured or not -, but also in conceptual
networks, i.e., frames, scenarios, domains etc. This is another
important point for historical semantics, because semantic change
can derive from altered perspectivization, profiling or highlighting of
concepts or conceptual aspects inside these cognitive networks.
These processes play an important role in ~ Langacker's work, esp.
with regard to his conception of subjectification (cf. section 2.4.) as
well as in the interpretation of changes based on contiguity in ~
Blank, ~ Traugott, ~ KoniglSiemund and ~ Koch. In contrast to
frame and scenario, the notion of "domain" is rather blurred, as it is
used indifferently to describe structures based on contiguity and
taxonomic relations (cf. the terminological distinctions made in
Taylor 1995: 83-87). This terminological inaccurateness can even
lead to explicit rejection of the term "domain" for the description of
semantic change (~ Warren).
Many studies in cognitive linguistics have emphasized the role of
the human body as a fundamental reference point of cognition. In a
diachronic perspective, this frame has a double function. First, if we
take the body as a target domain, we can find examples for innova-
tive denominations of body parts, which are of great cognitive inter-
est, and we can even find evidence for a change of the conceptual-
ization of the body itself (~ Krefeld). Secondly, the body also
serves as a source domain for diachronic processes and for gramma-
ticalization, e.g., the creation of intensifying adverbs out of co-refer-
ential pronouns, which themselves derive from words for body parts
(~ KoniglSiemund).
2.4. Specific properties of prototypically organized categories and

particular conceptual structures build the cognitive background of
semantic change. When it comes to a concrete semantic transfer,
Introduction 7
speakers intentionally or accidently perceive or reinterpret a given

concept in relation to another concept. The question is which types
of associative links can relate the source to the target domain and
how the resulting linguistic processes of semantic transfer can be de-
scribed systematically.
This observation leads us directly to figures of speech like meta-
phor, metonymy, synecdoque or ellipsis and some others whose his-
tory goes back to antique rhetoric. On the basis of the work done in
cognitive linguistics, these tropes are now considered as notions of
theoretical linguistics, instead of tools of practical rhetoric, and have
to be submitted to systematic analysis and definition. The contribu-
tions of ~ NerlichlClarke, ~ Warren and, partially, of ~ Liidtke
are going in this direction; ~ Blank shows some typical correlations
between certain types of associations and the motivations for seman-
tic change.
A highly interesting aspect in a large number of papers is the
great, if not to say, outstanding relevance of conceptual contiguity
("metonymy"). It is fundamental for the studies of ~ Konig/Sie-
mund, and - together with metaphor and semantic extension - it also
plays a central role in ~ Koch. Seen from this perspective, even
"subjectification" (according to both ~ Langacker's and ~ Trau-
gott's understanding) can be completely reduced to the profiling of
concepts against a background that is constituted by the respective
frames or contexts.
3. Recent issues in diachronic linguistics
In the last two decades, diachronic linguistics have been strongly in-
fluenced by pragmatics, a tendency that has also marked the present
volume.
First of all, we note that linguists have "rediscovered" the impor-
tance of the speaking subject, but the hearer's role has also been
reconsidered. Thus, language as a means of self-presentation and ex-
pression of subjectivity (~ Traugott) is coming into view. Speaker-
and hearer-oriented linguistic strategies (~ Geeraerts) and, in a

more general way, the importance of expressivity and efficiency in
lexical change are emphasized (~ Liidtke; ~ Geeraerts; ~ Blank;
~ Nerlich/Clarke). Finally, one encounters the phenomenon of the
"valorization" of words and concepts (~ Rastier) and the process of
"pragmatic strengthening" (~ Traugott; cf. also Konig in several
other publications).
The greatest progress in diachronic linguistic theory during the
last years has been the conception of language change as an "invi-
sible-hand process" (cf. detailedly Keller 1994). The theory of the
"invisible hand" provides us with an explanation for language
change that combines a framework taken from pragmatics (e.g., the
speaker- or hearer-oriented strategies as mentionned above) with
cognitive regularities of linguistic innovation (see section 2.), which
are corroborated by the interpretation of empirical data (~ Liidtke;
~ KoniglSiemund; ~ Koch). In this context, it has been discovered
that many diachronic processes are unidirectional and therefore nor-
mally are not reversible.
4. Grammaticalization
During the last two decades, grammaticalization has been a major

line of study in diachronic linguistics. In as much as grammaticaliza-
tion constitutes both a formal and a semantic process, linguists have
inevitably resorted to concepts such as "semantic bleaching", subjec-
tification, metaphor and metonymy (cf. Heine, Claudi and Hiinne-
meyer 1991; Hopper and Traugott 1993). Therefore, exploring the
"grammaticalization channels" and "scales" (Lehmann 1995: 25)
opens a broad field of study to cognitive linguistics. Conceptual
mechanisms involved in grammaticalization are explored by ~
Langacker (who uses the term "grammaticization"), ~ Traugott and
~ KoniglSiemund.
Insofar as grammaticalization is typically unidirectional (cf. ~
Ltldtke, ~ Traugott, ~ Konig/Siemund), it serves a good example
Introduction 9
for invisible-hand processes. Once a word or a syntagmatic construc-

tion is conventionalized as a grammatical rule there seems to be no
way back to the lexicon.
5. Two perspectives in semantic investigation:

semasiology and onomasiology
Traditional synchronic and diachronic semantics distinguish between

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

sents some of his examples in a way that allows a semasiological as
well as an onomasiological reading; e.g., figure 2 in his contribution
shows the change of meaning of Lat. manducare 'to chew' > Fr.
manger 'to eat' as well as the change of the expression for the
concept EAT from Lat. edere to Fr. manger.
Change of meaning, which is principally a semasiological pro-
cess, is approached from an onomasiological perspective in Blank's
study of the speakers' motivations for inventing new expressions for
concepts.
Rastier combines both perspectives in analyzing, dealing with the
semantic evolution of Fr. face on the one hand, and with the history
of the expressions for FACE in French on the other.
Konig/Siemund first focus on the concept of INTENSIFIERS which
is thus onomasiologically defined. Then, the particular semantic de-
velopment of the corresponding expressions in their sample of lan-
guages is subject to a double semasiological study, retrospectively as
"targets of semantic change" and prospectively as "sources of se-
mantic change".
Insofar as it investigates Latin and Romance (changes of) expres-
sions for parts of the HUMAN BODY, Krefeld' s study is onomasiolo-
gically oriented. But by discovering changes in the segmentation of
the conceptual frame itself, it is essentially dependent on semasiolo-
gical insights.
In Koch's article, POSSESSION, EXISTENCE, LOCATION, ASCRIPTION,
and their subdivisions constitute onomasiologically defined target
concepts. A retrospective (semasiological) view leads, then, to the
source concepts that serve as cognitive reference points for express-
ing the target concepts.
The semasiological approach not only gives us access to the his-
tory of particular linguistic phenomena, but, more importantly, it
also focuses our understanding of the cognitive basis and interpreta-
tion of diachronic processes. The onomasiological approach shows
the continuous change in the way we express concepts and concep-
tual domains while at the same time sharpening our view for recur-
rent types of expression and for their motivations.
Introduction 11
Combining the onomasiological approach with a well-founded

semasiological typology of diachronic semantic processes will en-
able us to understand, in a sort of "panchronic" perspective, the basic
cognitive patterns of how man conceives the world. We can hope to
identify the source-concepts that serve as typical reference points for
verbalizing a given target-concept. We can hope to describe accura-
tely the semantic path from source- to target-concept. 5 A diachronic
approach applied to a large language sample should help us to "neu-
tralize" historical idiosyncrasies and to make fundamental cognitive
patterns transparent.
According to this view, "Historical Semantics and Cognition"
does not constitute a gratuitous side track of cognitive linguistics,
but rather proves to be a central field of activity for what we could
call "anthropological linguistics" or "linguistic anthropology",
exploring the limits that the specific structure of human perception
imposes upon linguistic creativity. The contributions to this volume
lay some fundamental groundwork towards this promising project.
Notes
I. Cf., e.g., Trask (1996); Posner (1997); Campbell (1998); Fritz (1998); still
no reference to cognitive approaches is found in Hock (1991).
2. Consider also typical statements in Haiman (1980); Langacker (1987: 63);
Croft (1993: 336).
3. For further discussion of this topic cf. Ludi (1985, 91-94); Koch (1998:
118-120) and Blank (in press, section 11).
4. Onomasiological case studies are found in Geeraerts/GrondelaerslBakema
(1994: 117-153).
5. Two research projects at the university of TUbingen are attempting to make
this twofold program a reality. They are studying the lexical and semantic
evolution of the words for parts of the body (and for some related concep-
tual domains), in the Romance languages (project DECOLAR = Dictionnaire
etymologique et cognitif des langues romanes) and in a representative
sample of other languages of the world (project Lexical change - polygene-
sis - cognitive constants as part of the interdisciplinary Research Center
441 "Linguistic Data Structures"). Their goal is to discover the typical stra-
tegies for verbalizing these concepts, and, moreover, to establish empirical

evidence for their polygenetic origin, their areal distribution, and possibly
their idiosyncratic nature. We hope to learn which concepts have relatively
stable expressions and which are submitted to continuous change. Cf.
BlanklKoch, in press; BlanklKoch/Gevaudan, in press.
References
Blank, Andreas
1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der ro-
manischen Sprachen. (Beihefte zur Zeitschrift ftlr Romanische Phi-
lologie 285.) Ttlbingen: Niemeyer.
in press Neuere Entwicklungen der lexikalischen Semantik. In: Gtlnter Hol-
tus, Michael Metzeltin and Christian Schmitt (eds.), Lexikon der Ro-
manistischen Linguistik, Volume I (art. 34b).
Blank, Andreas and Peter Koch
in press Onomasiologie et etymologie cognitive: l'exemple de la T~TE. In:
Mario Vilela (ed.), Atas do 1 Encontro de Linguistica Cognitiva,
Porto 29 e 30 de Mayo 1998.
Blank, Andreas, Peter Koch and Paul Gevaudan
in press Onomasiologie, semasiologie et l'etymologie des langues romanes:
esquisse d'un projet. In: Actes du XXlIe Congres International de
Linguistique et Philologie Romanes.
Campbell, Lyle
1998 Historical Linguistics. An Introduction. Edinburgh: Edinburgh Uni-
versity Press.
Croft, William
1993 The role of domains in the interpretation of metaphors and metony-
mies. Cognitive Linguistics 4: 335-370.
Fritz, Gerd
1998 Historische Semantik. Stuttgart / Weimar: Metzler.
Geeraerts. Dirk
1997 Diachronic Prototype Semantics. Oxford: Clarendon Press.
Geeraerts, Dirk, Stefan Grondelaers and Peter Bakema
1994 The Structure of Lexical Variation. Meaning, Naming, and Context.
(Cognitive Linguistics Research 5.) Berlin / New York: Mouton de
Gruyter.
Introduction 13
Haiman, John
1980 Dictionaries and encyclopedias. Lingua 50: 329-357.
Heine, Bemd, Ulrike Claudi and Friederike HOnnemeyer
1991 Grammaticalization. A Conceptual Framework. Chicago, London:
University of Chicago Press.
Hock, Hans Henrich
21991 Principles of Historical Linguistics. Berlin 1 New York: Mouton de
Gruyter.
Hopper, Paul J. and Elizabeth C. Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
Keller, Rudi
21994 Sprachwandel. Von der unsichtbaren Hand in der Sprache. TObin-
gen: Francke.
Kleiber, Georges
1990 La semantique du prototype. Paris: Presses Universitaires de France.
Koch, Peter
1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine
kritische Bestandsaufhahme. Romanistisches Jahrbuch 46: 27-46.
1996 La semantique du prototype: semasiologie ou onomasiologie? Zeit-
schriftfur Jranzosische Sprache und Literatur 106: 223-240.
1998 Saussures mouton und Hjemslevs trQ!: zwei Schulbeispiele zwischen
Semstruktur und Polysemie. In: Edeltraud Wemer, Ricarda Liver,
Yvonne Stork and Martina Nicklaus (eds.), et multum et multa.
Festschrift fUr Peter Wunderli zum 60. Geburtstag, 113-136. TO-
bingen: Narr.
Lakoff, George
1987 Women, Fire and Dangerous Things. What Categories Reveal about
the Mind. Chicago etc.: University of Chicago Press.
Langacker, Ronald W.
1987/91 Foundations of Cognitive Grammar. 2 Volumes. Stanford: Stanford
University Press.
Lehmann, Christian
1995 Thoughts on grammaticalization. (LINCOM studies in theoretical
linguistics 1.) MUnchen: Lincom Europa.
LOdi, Georges
1985 Zur Zerlegbarkeit von Wortbedeutungen. In: Christoph Schwarze
and Dieter Wunderlich (eds.), Handbuch der Lexikologie, 64-102.
K6nigsteinlTs.: Athenaeum.
LUdtke, Helmut
1980 Sprachwandel als universales Phlinomen. In: Helmut LUdtke (ed.),
Kommunikationstheoretische Grundlagen des Sprachwandels, 1-19.
Berlin / New York: Mouton de Gruyter.
1986 Esquisse d'une theorie du changement langagier. La Linguistique
22: 3-46.
Posner, Rebecca
1997 Linguistic Change in French. Oxford: Clarendon.
Taylor, John
21995 Linguistic Categorization. Oxford: Clarendon.
Trask, Robert L.
1996 Historical Linguistics. London: Amold.
Ungerer, Friedrich and Hans-J6rg Schmid
1996 An Introduction to Cognitive Linguistics. London / New York:
Longman.
Section I
Theories and Models

Cognitive Semantics and Structural Semantics
John R. Taylor
This contribution is not specifically about historical semantics. My

focus, rather, is some basic issues in semantic theory, especially as
these arise from a confrontation of Structural Semantics and Cog-
nitive Semantics.
By "Structural Semantics" } refer to the well-established conti-
nental European tradition, represented above all by Coseriu (e.g.
1977). Structuralist approaches are also evident in Lyons (1968) and
Cruse (1986). Structuralism has profoundly influenced historical lin-
guistic studies, especially as these have pertained to the familiar
European languages (e.g. Coseriu 1974).
By "Cognitive Semantics"} refer to the study of semantics within
the framework of "Cognitive Grammar", as developed above all by
Langacker (1987, 1991). The work of Lakoff (1987), Talmy (1988),
and many others, is broadly compatible with Langacker's approach.
Studies of grammaticalisation, e.g. Heine (1993, 1997), can also be
assimilated to the Cognitive Grammar programme. For lexical histo-
rical semantics within the Cognitive Grammar framework, especially
important is the work of Geeraerts (e.g. 1985, 1997).
Coseriu (1990) sees a profound gulf between Structural Seman-
tics and the Cognitive Grammar approach to semantics. In recent
years, however, others (e.g. Koch 1995, 1996) have been keen to in-
corporate insights of Cognitive Grammar into historical linguistic
studies, without, however, wanting to give up some of the basic as-
sumptions of Structuralism. In this connection, it should be noted
that the aspect of Cognitive Semantics that has been most commonly
seized upon, has been the idea of categorisation by prototype. Al-
though "cognitive semantics" appears in the title of Coseriu (1990),
the only aspect of Cognitive Semantics that is dealt with in any depth
in the article is categorisation by prototype. This emphasis on proto-
18 John R. Tay/or
types is unfortunate, in that there is much more to Cognitive Gram-

mar than categorisation by prototype!
I have several aims in this chapter. One is to dispel what appear to
be some current misunderstandings about Cognitive Grammar. I also
question some of the assumptions underlying Structural Semantics,
arguing that some of the postulated distinctions may be unnecessary,
and, to the extent that these distinctions do have validity, they can be
incorporated unproblematically into the Cognitive Grammar model. I
also tentatively point to some aspects of the Cognitive Semantics ap-
proach that I believe are likely to be especially relevant to historical
semantic investigations.
I begin by observing that although Cognitive Grammar and Struc-
turalism have developed independently, with little mutual interac-
tion, the two approaches can be seen to have a common origin. Both,
namely, have developed, albeit in'different ways, some basic insights
of Saussure.
1. The Saussurian heritage
Not the least of the achievements of Ferdinand de Saussure was to

have established Linguistics as an autonomous academic discipline.
Contrary to the naive view of the matter, it is not, according to Saus-
sure (1916: 23), the existence ofa certain subject matter (in our case,
language, and languages) that justifies and circumscribes a dis-
cipline, but rather a "point of view", a distinctive way of treating the
subject matter. Scholars with all manner of interests have had things
to say about language. Yet if it is to achieve disciplinary autonomy,
Linguistics cannot simply be the grand sum of whatever literature
students, anthropologists, sociologists, psychologists, etc. say about
language. Paradoxically, the very centrality of language to human
existence sharpens the need to define the proper object of Linguis-
tics. As is well known, Saussure's proposal was that the distinctively
"linguistic" study of language had to treat language as a semiotic
system. The proper object of linguistic inquiry is therefore the "lin-
Cognitive Semantics and Structural Semantics 19
guistic sign", the symbolic association of a signifier (an "acoustic

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

instance to a prototype.
A special case of the "is-a" relation obtains between the use of an
item on a specific occasion and the item as stored in a speaker's
memory.2 When the word tree is uttered with reference to a specific
tree, the semantic pole of the utterance (the specific tree referred to,
or, to be more precise, the speaker's conceptualisation of the tree) is
an instance of the more abstractly characterised tree-concept associ-
ated in the speaker's mind with the stored lexical item. In parallel
manner, the pronunciation of the word on a specific occasion is an
instance of the more abstractly characterised phonological represen-
tation stored in the speaker's mind. Note that the properties of the in-
stance may not fully match the more abstract schema. (Suppose that
the word tree is applied, not to a "prototypical" tree, but to a date
palm, or that the word is pronounced in a non-standard way.) In such
cases, the instances may still count as instances of the symbolic unit
[TREE] in virtue of the "is-like" relation; the usage would count as
"marginal", but would still be attracted to the stored unit in virtue of
its similarity to it.
A related point, is that there is no need to make a principled dis-
tinctiol!. J~~.twe~,n . "Ji.p.guIstic·· tneaning;;aIt~~~~~I!c.Y~J()il.~dic -knowl-
edge"! Consequently, Cognitive Grammar does not draw a distinc-
tiO-n'in principle between "sentence meaning" and "utterance mean-
ing". Traditionally, sentence meaning is the meaning that a sentence
has in virtue of the "linguistic meaning" of its parts, whereas utter-
ance meaning is the meaning that an utterance acquires in a particu-
lar communicative context. Both kinds of meaning properly belong
in the semantic representation associated with the symbolic unit, and
both need to be characterised relevant to appropriate background
knowledge. Naturally enough, "utterance meaning" may need to be
characterised against a much richer array of background assump-
tions, which appeal to specific aspects of the speech situation. But
this fact is fully consistent with the view that utterance meaning
stands in an "is-a" relation to sentence meaning. Sentence mearung,
to the extent that the notion is vanao'al all~- is- schematic for the range
of utterance meanings that an expression may have on specific oc-

casions of its use.
I have emphasised the Saussurian roots of Cognitive Grammar, in
order to better contextualise a comparison with Structural Semantics.
Structural Semantics has also drawn its inspiration from Saussure,
albeit with an emphasis on other aspects of Saussure's thought.
Saussure, as we all know, asserted that the link between the signifier
and the signified is arbitrary (Saussure 1916: 100-102). There are, to
be sure, slightly different (though not incompatible) ways of under-
standing "the arbitrariness of the sign". In the first place, the sign is
arbitrary in the sense that there is no inherent association between
sound and meaning. On this view, arbitrariness contrasts with moti-
vation. A sign may be motivated, to the extent that the language user
can perceive some reason why the signifier should have the meaning
that it does (and vice versa). Internally complex signs are usually
motivated, to varying degrees. On a slightly different understanding
of arbitrariness (and it is this understanding that is especially em-
phasised in Cognitive Grammar), the linguistic sign is arbitrary in
the sense that it is the product of conventionalisation. Speakers act in
the belief that the signifier-signified relation (at least for established
units, cf. footnote 3) is shared by other members of a speech com-
munity. And, as Saussure (1916: 104) pointed out, a speaker is un-
able to single-handedly modify the established and shared conven-
tions.
Saussure (1916: 155-157) goes further, and maintains that it is not
only the signifier-signified relation that is arbitrary, the signifiers and
the signifieds in any given language are themselves arbitrary, in the
sense that there is no intrinsic reason why just these meanings should
receive symbolic expression, nor why just these phonological forms
should serve as signifiers. Saussure emphasised that the signs that
make up a language do not constitute a nomenclature, i.e. they are
not labels for an independently given list of concepts. It is the lan-
guage itself that structures cognition, thereby creating the concepts
through the very process of symbolising them. Likewise, there is no-
thing intrinsic to a sound that renders it suitable to function as a lin-
guistic signifier. Sounds have the status of speech sounds only in
22 John R. Tay/or
virtue of the structuring of sound by the semiotic system that is a

language.
If pursued, Saussure' s views on the arbitrariness of concepts and
sound patterns must inevitably lead to a position of radical rela-
tivism, of a form that perhaps not even Whorf would have wanted to
endorse!4 Research on semantic and phonological universals has, of
course, revealed rather severe constraints on the concepts and sound
patterns that may come together in symbolic association, while still
allowing, within the boundaries set by these constraints, consider-
able cross-language diversity,s But perhaps the most challenging
component of Saussure's thought, and one that has fired the imagi-
nation of generations of his readers, lies in his thesis that units on the
phonological and semantic levels have a status within the language
only by virtue of the relations which they contract with other units
on the same level (phonological or semantic). Saussure (1916: 158-
160) introduced the term "value" to refer to this aspect of the lin-
guistic sign. A "concept" receives its "value", not in virtue of any in-
trinsic semantic content, but in virtue of the relations (syntagmatic
and paradigmatic) which it contracts with other symbolised con-
cepts. Likewise, sound units have a value in the language system in
virtue of the sound units they contrast with, and combine with. Para-
doxically, the value of a linguistic unit is determined, not by its in-
trinsic content, but by what it is not, by "ce qui existe en dehors de
lui" [what exists outside of itself] (1916: 160).
In proposing the notion of "value", Saussure is not denying the
obvious fact that signs do have a positive content, alongside their
contrastively defined value. Structuralist Semantics captures this dis-
tinction by the terms "signification" and "designation" (or, in Ger-
man, "Bedeutung" and "Bezeichnung,,).6 The "signification" of a
sign is the concept understood contrastively i.e. in terms of its rela-
tions to neighbouring concepts; the "designation" is the concept un-
derstood in its positive aspects, Le. in terms of its potential to refer
to actual states of affairs in the world. To give a simple example: the
morphophonemic category [SINGULAR] has the same designation,
both in a language which has a simple two-way contrast between
singular and plural, and in a language which has a three-way contrast
between singular, dual, and plural. Yet the signification ("value") of

[SINGULAR] is different in the two cases. In the one, [SINGULAR] en-
ters into a simple binary contrast with [PLURAL], in the other, it en-
ters into a ternary contrast with [DUAL] and [PLURAL]'
We can make a further distinction, between "signification" and
"designation" on the one hand, and "reference" on the other. Here,
we need to return once again to Saussure. Saussure (1916: 98) made
it very clear that for him, the linguistic sign was a mental entity. The
linguistic sign [TREE] did not associate a particular tree growing in
the yard with a specific utterance [tri:]. The semantic content of the
sign was a "concept" in the mind of the language user. Saussure like-
wise insisted that the acoustic image was a mental representation,
distinct from any physical manifestation, and thus neutral with re-
spect to pronunciation and perception. But in any particular act of
speech, there is, obviously, a specific articulatory/acoustic/perceptual
event, probably (though not necessarily) associated with reference to
a particular entity in the real world. Thus, it is easily possible for two
signs to be used with the same reference, but having different desig-
nations (and eo ipso, different significations). To extend on the earli-
er example: The category [PLURAL], in the two kinds of languages
mentioned above, has the same reference when used of a group of
three entities. But the designation (and signification) of [PLURAL] is
different in the language which only has the two-way contrast be-
tween singular and plural, from the designation that it has in a lan-
guage which knows a three-way contrast between singular, dual, and
plural.
Saussure represented the sign as a simple bipartite entity (Fig. 1).
More elaborate schemas have been proposed; the "semiotic penta-
gon" in Fig. 2 has been adapted from Koch (1996), who attributes it
to Raible (1983). As mentioned, the sign for Saussure was a purely
mental entity. In order to link the two elements of the sign to entities
outside the mind (i.e. to an acoustic-phonetic event on the one hand,
and, on the other, to a referent in the world), the bipartite sign is ex-
tended to include a "name" and a "referent". A fifth element is intro-
duced, in order to capture the distinction between signification and
designation.
24 John R. Tay/or
Figure 1. The Saussurean sign
, ,
,,
(linguistic sign)
signified
, ,,
, ,,
signifier designation
,
,,
~--- ----------------
(act of speech)
name referent
Figure 2. The "semiotic pentagon" (after Raible 1983 I Koch 1996)
I have already hinted at what might be the Cognitive Grammar

approach to the issues touched on above. The distinction between an
acoustic-phonetic event and a signifier (Saussure's "acoustic im-
age") is an Ilia-a" relation, i.e. the relation between a fully specified
instance and a more abstractly characterised schema. The same goes

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

Structural Semantics concerns the special status accorded in the
latter to "signification" in contrast to "designation" and "reference".
What is at issue here, essentially, is whether it is justified to postu-
26 John R. Tay/or
late a level of purely "linguistic" meaning, in contradistinction to a

level of encyclopedic knowledge, between "Wla semantica lingiiis-
tica" [a linguistic semantics] and "Wla semantica de las cosas" [a se-
mantics of things] (Coseriu 1990: 281).7
For Coseriu (1990: 267), Cognitive Semantics commits "el error
mas grave y mas elemental que pueda cometerse en semantica" [the
most serious and most error that it is possible to commit in seman-
tics]; this is the capital error of confusing linguistically structured
meaning with experientially derived knowledge about the states of
affairs that linguistic expressions refer to. The confusion manifests
itself, according to COSeJ1U, in the treatment of prototype effects. On
the Structuralist view, prototype effects lie outside the language sys-
tem proper; they have to do with difficulties a person may encoWlter
in properly applying a word to a state of affairs, i.e. they are matters
of designation and reference, not of signification. That it might be
difficult to determine, at a given time and place, whether it is
"night", or "day", in no way entails that the significations of the
words night and day are "fuzzy" or indeterminate, or structured
around a prototype. On the contrary, Coseriu argues, prototype ef-
fects arise precisely because the linguistic meanings of night and day
are absolutely clear-cut; were this not the case, we could have no
confidence in asserting that a certain state of affairs constitutes a
good example, or a less good example, of the application of the word
(Coseriu 1990: 258). Likewise, penguins and ostriches can only be
recognised as "marginal" examples of the bird category if the cate-
gory is already clearly defined. In order for a bird to be a "less good"
example of the category, it must already have been categorised as a
bird (1990: 279). And the clear-cut concept of what a bird is, is a
matter of signification, not of designation, or of associating exem-
plars with a prototype. Whether birds, as entities in the world,
constitute a clear-cut category or not, is a matter of biology, not of
linguistics.
Coseriu (1990: 268) observes that Cognitive Semanticists have
selectively focussed on just those words (such as to lie, and names of
natural kinds) whose real-world applications tend to give rise to pro-
totype effects, whilst ignoring linguistic-semantic contrasts which
are clear-cut, not only on the level of signification, but also on the le-
vels of designation and reference. He mentions the example of mo-
tion verbs. Spanish venir and ir contrast with respect to 'motion to
the place of the 1st person' vs. 'motion to the place of the 2ndl3rd
person'. In Italian and Catalan, the contrast between venire/andare,
venir/anar, is drawn differently, between 'motion to the place of the
1st/2nd person' vs. 'motion to the place of the 3rd person'. (Hence,
Catalans, speaking Spanish over the telephone, will tend to make the
error of saying • Maiiana vengo a verte 'Tomorrow I come to see
you', instead of the correct Maiiana voy a verte 'Tomorrow I go to
see you'.) Moreover, the notion of "structure", and of a "structured
lexicon", suggests that a semantic contrast might serve to differentia-
te more than one word pair. And indeed, the above mentioned con-
trast in Spanish shows up with verbs of carrying: traer vs. llevar. (In
Italian, though, the contrast is not made: portare serves for both
senses.) It is difficult to imagine, Coseriu remarks, what "proto-
types" could be associated with these clear-cut meanings, and what
deviations therefrom could look like. There is a further point. This is
that the distinctions in question are language-specific, and therefore
cannot plausibly arise from any natural categorisation of non-linguis-
tic reality. By focussing on the referential possibilities of lexical
items, and on the naming of real-world (and therefore universally ac-
cessible) categories, Cognitive Semantics has ignored the structured,
language-specific relations that exist between significations. In brief,
Cognitive Semantics falls into the trap that Saussure warned us
about, of viewing a language's lexicon as a nomenclature, a list of
names for pre-existing categories.
With respect to its allegedly onomasiological orientation, Coseriu
brings in a third player, in addition to Structural Semantics and Cog-
nitive Semantics, namely the theory of word definition by necessary
and sufficient conditions. Although Fillmore (1975) presented proto-
type semantics in opposition to "check-list" theories of meaning, Co-
seriu groups both together as examples of onomasiologically orient-
ed approaches, and both stand in contrast to Structuralism, which
looks in the first place at relations of contrast between linguistic
units, not at states of affairs in the world. Thus, for Coseriu (1990:
28 John R. Tay/or
245), the Katz and Fodor (1963) analysis of bachelor, which defines
the word as a conjunction of the features [HUMAN], [ADULT], [MALE],
[NEVER MARRIED), suffers from the same fault as prototype theories,
in that it defines the word in terms of the conjunction of (real-world)
features of its potential referents (Le. in terms of the word's
designation), rather than in terms of the word's linguistic value.
In evaluating Coseriu's critique, let us first consider the content of
the proposed contrasts at the level of significations. Let us accept
that day and night stand in a simple two-way contrast. The contrast
has to do, presumably, with the presence vs. absence of sunlight (as-
suming an open-air environment). Note that the contrast appeals in-
trinsically to a real-world phenomenon, one that can only be appre-
hended empirically, through experience of the world. Coseriu, gen-
erally, is quite happy to give natural language glosses (in French,
German, Spanish, or whatever) to the content of distinctive semantic
features. Now, Jackendoff (1990: 33) has remarked that one cannot
create a semantic feature simply by taking any old expression and
putting a pair of square brackets around it. Behind Jackendoff"s quip
is the idea that if linguistic meanings are to be distinct from encyclo-
pedic knowledge(the features that go into the linguistic defini,tions
m~~t . ~e:()~1oJQg!c;~lh'. distinct from attrlliuteSOf"the real W6f1.d.,kor if
there is no such distinctron"beiWeeii'Hiigulstic~'semantic features, and
attributes of extra-linguistic reality, the methodological basis of the
distinction becomes vacuous.
And indeed, a common strategy of many two-level theorists (see
footnote 7) is to propose that semantic features have the special
status of semantic primitives, presumably innate to human cognition,
and that are mdepeiiderit-'of expenence.·1~off. foreX-ample,
postulates a set of "conceptual constituents\-o'f the kind [THING),
[PLACE], [GO], [STAY], [MOVE], [CAUSE], etc., which are combined in
accordance with "conceptual well-fonnedness rules". These generate
the general architecture of all possible concepts, whose substance is
filled in by information derived from acquaintance with the world.
Such an approach will tend to emphasise the universality of se-
mantic structures, at least at a certain level of abstraction. CqSe"riu,
on the other hand, makes no pretence that distinctive semantkiea-
tures ("semes") might be, or might be built up out of, universal se-
mantic primitives. 8 Distinctive semantic features have to be deter-
mined case by case, according to fhe structUral relations'obtaining in
a giverrtanguage;-ancrareasslmple or-complexas thedata requires.
FurthelllIOre;-stgn1fica:tlonsare-ii.ot··~COuiIf·up""'otif'orfeamr~s-;·1tisthe
features that emerge from the contrasts, not vice versa (Coseriu
1977: 17). Coseriu (1990: 261) cites with approval Pottier's (1964)
well-known analysis of seating objects in French, which lists such
real-world features as "avec pieds" [with feet], "avec bras" [with
arms], "avec dossier" [with back]. Note here that the very notions of
a "(chair)-leg", "(chair)-arm", and "(chair)-back7already presuppose
(encyclopedic) familiarity with the domain of furniture, and with the
conventional practice of naming parts of furniture metaphorically in
terms of animal (or human) body partr9 It would indeed be "patently
ridiculous" (Jackendoff 1990: 33) to propose "avec dossier" as a uni-
versal semantic feature. But it is also difficult to imagine what the
"linguistic" meanings of chaise, fauteuil, tabouret, etc., could be, if
not knowledge of what these kinds of objects actually are, and how
they are to be differentiated one from the other. to
Since, for Structuralist Semantics, the distinction between the lin-
guistic and encyclopedic levels does not reside in the content of the
distinctive features, we need to ask whether there are other charac-
teristics of significations,' which· rendertltis levef"ofdescfiption onto-
logicf:l11YJfi~:tiJlct'fr6m-designatiori andieference.-Two aspects ap-
pear to be relevan!i.orCQ$eriu. The first I have already mentioned.
This is thatslgmfications (within a given semantic field) are clearly
contrastive, and betray no "fuzziness" or prototype effects. A corol-
lary of contrastiveness, which I shall address in the next section, is
that significations are taken to be unitary entities, i.e. betray no poly-
semy. The second aspect is the possibility of neutralisation (Coseriu
1977: 17-18). The notion is familiar from phonology. In certain en-
vironments, the contrast between two otherwise contrastive pho-
nemes is suspended. A well-known example concerns the neutralisa-
tion of the voicing contrast in word-final obstruents in German and
Russian. Coseriu (1990: 260) views neutralisation as a specifically
linguistic (not a conceptual) phenomenon; consequently, the possibi-
30 John R. Tay/or
lity of neutralisation can serve as a diagnostic, as it were, of a

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

more, there is no inherent conflict between prototype categorisation
and semantic contrast. On the contrary, Rosch (1978) argued that
"basic level" categories achieve salience largely because their proto-
types maximise the distinctiveness of the categories (cf. Taylor
1995: 50-51). Neither is it fair to charge Cognitive Semantics with
undue concern with real-world, and hence "universal" (Coseriu
1990: 252) categories, for which a language merely supplies a list of
names. From its very inception, Cognitive Grammar has emphasised
the role of "construal" in semantics; linguistic expressions do not
refer "directly" to states of affairs in the world, but to speakers'
conceptualisations of these states of affairs (Langacker 1987: ch. 3).
Furthermore, it is fully accepted that different languages may make
available to their speakers different sets of "conventionalised"
modes of construal.
Secondly, it is not always the case that words contrast so clearly
as in the examples that structuralists like to quote. This is most obl .
vious in the case of (near) synonyms. Crose (1986: 266) characte-
rises (near) synonyms as items which have "a low degree of implicit
contrastiveness". Thus, in declaring that a building is "high", one is
not implicitly denying that it is ''tall'' (and vice versa). Although high
and tall do not share exactly the same meaning, the difference can
hardly be stated in terms of the presence vs. absence of some distinc-
tive semantic feature. I I Crose (1986: 285) also drew attention to
what he called "plesionyms" - sets of words that are only weakly
contrastive, and which stand mid-way, so to speak, between (near)
synonyms on the one hand, and fully contrastive word sets on the
other. Take Crose's examples fog, mist, haze. Whereas other words
for meteorological phenomena, such as rain, snow, hail, arguably do
form a clearly contrastive set, this is certainly not the case withfog,
mist, haze. Precisely because the words are only weakly contrastive,
the boundaries of their meanings are not clearly defined - either con-
ceptually, or referentially. Even so, I still have a fairly clear concep-
tion of what a prototypical fog etc. is like. Consequently, if I attempt
to apply one of these words to a specific state of affairs, I can do no
other than appeal to a conception of a prototypical fog~ etc., and as-
32 John R. Tay/or
sess how well the actual situation confonns to the prototype(s), and
on this basis, decide which of the three words might be most appro-
priate. But if this is the case, there is no reason to suppose that a
similar process does not apply when I use the words snow and hail,
day and night. The only difference is, that in the latter case, the pro-
totypes are clearly distinct, and characterisable in tenns of the pre-
sence vs. absence of some easily statable attribute, whereas the
prototypes ofJog, mist, haze are not.
3. The question of polysemy
For Structural Semantics, as for other "two-level" approaches, it is

axiomatic that the linguistic meanings ("significations") of words are
unitary entities, i.e. that at the level of significations, polysemy does
not exist. It is not denied that a word may be used in a variety of
senses. 12 But these senses only arise when uniquely specified values
get filled out with semantic content, either as a matter of conven-
tional usage, or in a specific discourse context. In this connection,
we may refer to Coseriu's well-known distinction between "system",
"nonn", and "discourse". The "system" is specified at the level of
the language-detennined significations, the "nonn" comprises estab-
lished elaborations of significations, while "discourse" pertains to
specific readings that emerge within a text. (The phonological ana-
logy should be obvious. Phonemic contrasts pertain to the "system",
while the "nonn" comprises established allophonic realisations.) Co-
seriu's position is that Cognitive Semanticists are inclined to find
polysemy everywhere, because of their fixation on "nonn" (and even
"discourse"), and their neglect of the "system".
As with prototypes, there is unfortunately some misunderstanding
of what the Cognitive Grammar position is with regard to polysemy.
It is certainly true that Langacker (1988: 50) has asserted that the
nonnal, expected state of affairs in lexical semantics is that a word
(especially a word in frequent use) will be polysemous, i.e. will have
a range of established senses. Thus, the semantic pole of a symbolic
unit may need to be represented as a network of units, linked by rela-
tions of schematicity and resemblance to a prototype. Although the

matter has not been as extensively studied in Cognitive Grammar,
the phonological pole, to the extent that a unit may receive a variety
of pronunciations, may also need to be represented as a network of
possibilities. 13
A number of "case studies" of individual lexical items pursued
within the Cognitive Grammar tradition, have portrayed these as
sometimes highly polysemous, and this fact may well have contri-
buted to the perception that Cognitive Grammar encourages the pro-
liferation of polysemy. 14 On the other hand, whether or not an item is
to be regarded as polysemous is an empirical question, to be deter-
mined case by case. There is certainly nothing in the Cognitive
Grammar framework which excludes the possibility that a linguistic
unit may have a constant, invariant value. And some analyses have
indeed emphasised the unitary value of some linguistic signs. This is
especially the case with respect to grammatical categories. Thus,
Langacker (1987) argued that the lexical categories [NOUN] and
[vERBl can be associated with a single, highly abstract (schematic)
value. IS
With regard to many lexical items, however, polysemy is surely a
brute fact, which simply cannot be argued away.16 Consider Fill-
more's (1982) well-known analysis of climb. Fillmore, it will be re-
called, postulated a prototypical sense, which involves the features
"clambering (with the limbs)" and "ascending". Both are present in
climb a tree. But in climb down a tree, the feature "ascending" is de-
feated. Coseriu (1990: 256-257), addressing Fillmore's analysis,
observes that the very possibility of "climbing down" a tree demon-
strates that Fillmore's analysis was incorrect; the proper characteri-
sation should be, not "ascend", but "(move) in a vertical or inclined
plane" (sobre un piano vertical 0 inclinado). Concerning the feature
"clambering", given that monkeys, snails, and even plants can climb,
the proper characterisation should be "keeping hold with the extre-
mities" (agarrandose con las extremidades). The fact that, in the ab-
sence of specifications to the contrary, "climbing" is taken to be in
an upward direction, is a default interpretation, associated with the
"norm", not with the "system".17
34 John R. Taylor
)j Unfortunately, this proposal fails to cover some further uses of

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

has a range of different readings, and we can state them quite preci-
sely, and point to their similarities (as Fillmore did), irrespective of
whether there is (or whether we recognise that there is) a unitary
meaning. However, to the extent "that a speaker is creatively extend-
ing the usage range of a word, it may well be true that the speaker
does need to recognise some commonality between an accepted
usage A and the new usage situation B.19 (Still, the pertinent mean-
ing invariant common to A and B need not coincide with the invari-
ant which justifies the extension of the word from A to another situa-
tion C, which is tantamount to claiming that there will be no invari-
ant that unifies all three readings.) In Cognitive Grammar terms,
these commonalities would be captured by means of low-level sche-
mas that cover the relevant cases. But with respect to a range of al-
ready established (and conventionalised) uses, nothing excludes the
possibility that these uses are simply learned, on the basis of input
data. Indeed, without some such assumption, it would be difficult to
explain how different readings of a word can drift so far apart over
time. A speaker of modem English probably no longer perceives any
relationship at all between type 'kind' and type 'printer's character'
(cf. Geeraerts 1985), or between buff 'dull pale-brown' and buff
•amateur enthusiast' .
Not only is the structuralist level of signification not strictly ne-
cessary in order to guarantee a person's adequate use of a language,
it is difficult to imagine how significations, as understood by Cose-
riu, could be learned in the first place. Recall that significations do
not emerge from usage events: "le releve des procedes employes
dans la production des phrases ne pourrait jamais amener cl la delimi-
tation du signifie" [listing the procedures employed in the production
of sentences could never lead to the delimitation of significations]
(1977: 12). A little further down, we read that from "des acceptions
ou des variantes isolees", "on ne peut pas, en principe, deduire d'une
f~on immediate le signifie" [from isolated readings or variants, it is
not possible, in principle to directly deduce the signification]. Signi-
fications, in fact, appear to inhabit an idealist world, distinct from
the world in which and of which language is used: "el mundo de 10s
36 John R. Tay/or
significados es un mundo ordenado; no es el mundo ca6tico y con-

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

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

pole of the linguistic sign, many semanticists have been reluctant to
appeal to concepts at all. Concepts, by defInition, are private, mental
entities; a person can have no access to another person's concepts -
except, of course, through the medium of language. But if language
is defIned as a means for symbolising concepts, there is no methodo-
logy for independently establishing the nature of another person's
concepts. Appeal to concepts, therefore, could be circular (cf. Lyons
1977: 113). Interestingly, Lyons (1968: 443) favoured a structuralist
approach to semantics precisely because it frees the linguist from the
need to refer to "concepts". The meaning of a word becomes nothing
other than the set of relations that the word contracts with other
lexical items. 22
I do not think that "concepts" need be such mysterious entities as
Lyons and others make out (cf Taylor, in press a). A common view
amongst psychologists is that a concept is a P,!inciple of categorisa-
tion (Komatsu 1992). To "have" a concept, is to have tIle means to
categorise entities as examples of thatconcept. Put CrUdely:i~ have
the concept TREE, is to have the ability to recognise a tree when one
sees one. Understood as schemas for categorisation, concepts are by
no means restricted to n"Omlnal entities. Oneof Langacker's major
achievements is to have proposed a theoretical apparatus for the elu-
cidation of the conceptual structure, not only of various relational
units, such as verbs, prepositions, and adjectives, but also of "func-
tional" morphemes such as the articles and case categories.
What goes into a concept? Coseriu (1990: 261) - rightly - criti-
cises the view that concepts might be "imagenes de las clases" -
mental representations ("pictures") of categories. A crucial notion of
Cognitive Grammar (Langacker 1987: 183-185) is that the meaning
of an expression involves the~'profIling" (or designation) of an enti-
ty, against backiCoW}d assumptIOns. (These latter are referred to
variously as domains, frames, idealised cognitive models, etc.) The
(by now) classic example is the word hypotenuse (Fillmore 1985).
The word designates a straight line, no more, no less. A hypotenuse,
however, is categorised (recognised) as such, in virtue of the fact

that the straight line functions as part of a non-designated (non-pro-
filed) entity, namely, a right-angled triangle. It is not the straight line
as such, nor the right-angled triangle, that constitutes the concept
HYPOTENUSE, but the profiling of the straight line against the notion
of a triangle.
The example of hypotenuse is relatively simple, in that the con-
cept presupposes a fixed, and easily circumscribed domail!.2.f.~()_~I
e~~~~~ 1\:1ost concepts need to be characterised against €iultiple do-
mains, ~f varying centrality, which may be selectively activated on
partic~ occasions of their use. (Thus, read a book, print a book,
drop a book, etc., construe BOOK slightly differently in each case,
and highlight different background domains, even though the pro-
filed entity arguably remains the same.) Concepts, therefore, turn out
not to be fixed entities, but rather "emerge" in the act of conceptuali-
sation. By the same token, complex expressions are rarely fully com-
positional, in the sense that their meaning can be computed from the
fixed meanings of their component parts. Combining concepts is not
just a matter of combining the profiles, it also involves the integra-
tion of background knowledge.
Coseriu accuses Cognitive Linguists of exaggerated concern with
"objective" categories; there are classes of things out there in the
world, which the words of a language pick out. This view, I think,
seriously misunderstands the Cognitive Grammar programme. Ob-
jectively speaking, I dare say there is much in common between
writing (with a pen) on a piece of paper and drawing (with a pencil)
on a piece of paper. Both involve a person holding a slim instrument
and making marks on a surface. Why do we not categorise the two
kinds of events in the same way? The Structuralist view would be
that it is the language system itself that presents us with the struc-
tured opposition between write and draw. But there is surely more to
it than this. Writing and drawing are understood against broader con-
stellations of knowledge. Murphy and Medin (1985), not mappropri-
ate1y,speatcnf-"ttleories", which serve to give coherence to catego-
ries?3 Writing is understood against a theory of written linguistic
communication, drawing against a theory of visual representation. It
40 John R. Tay/or
is in virtue of the background theory, that writing (with a pen) and

writing (with a wora processor)'::::·
two very different kinds of activ-
ity, objectively speaking - are nevertheless both categorised as in-
stances of "writing".
Changes in worE.._I!!eanin~ are likely to have as much to do with
changes in background assumptions, i.e. domain-based knowledge
configurations, as with designation ("profiling"). Indeed, the aspect
of Cognitive Semantics that promises the greatest scope for insight-
ful studies of meaning change, could well be the importance attribut-
ed to background assumptions. The development of Latin scribere
from 'make marks on a surface' to 'write', and the development of
legere from 'pick out' to 'read', are not just a matter of "restriction"
or "specialisation" of meaning (nor of the addition, or subtraction, of
semantic features). In each case, the profiled activity remains much
the same. What has changed are the background assumptions (the
"theories") against which this activity is profiled.
5. Conclusion
Let us return to Saussure's original insight that a "concept" needs to

be characterised both positively (in terms of its actual content), and
negatively (in terms of what it is not). Structuralist Semantics chose
to separate out these two aspects, proposing a level of designation
(the positive content of the signified), and a level of signification
(the signified in contrast to other signifieds). This, I think, was an er-
ror. Just as the Saussurian sign resided in the integration of the signi-
fied and the signifier, so too the signified resides in the integration of
designation and signification. Cognitive Grammar achieves this inte-
gration by means of the notion of profile and base. The profile is the
concept in its positive aspects, i.e. the entity (or category) actually
referred to. The base comprises background knowledge that is not
specifically designated. But without the base, there can be no profile,
and the base, without profiling, lacks structure.
The major achievement of Structuralist Semantics is to have em-
phasised the semantic relations between lexical items. Some of the
earlier studies of lexical items within the Cognitive Semantics tradi-

tion (e.g. Brugman 1981; Coleman and Kay 1981) probably did tend
to study words in isolation from other lexical items with which they
stand in contrast. But it would certainly not be fair to say that Cog-
nitive Semanticists have in general been insensitive to matters per-
taining to lexical fields, and to implicit contrasts between lexical
items. These implicit contrasts belong in the domain-based knowl-
edge against which an entity is profiled. The background knowledge
against which a concept is profiled may comprise not just "encyclo-
pedic" knowledge pertaining to a conceptual domain, but equally,
"linguistic" knowledge pertaining to the paradigmatic and syntagma-
tic relations that the linguistic unit contracts with other linguistic
units.
Notes
1. Cf. Langacker (1987: 11): "Language is symbolic in nature. It makes

available to the speaker - for either personal or communicative use - an
open-ended set of linguistic signs or expressions, each of which associates
a semantic representation of some kind with a phonological representation"
[bold in original]. Lakoff's (1987: 473) characterisation of constructions as
"pairings of form and meaning" can also be taken as an endorsement of the
symbolic nature of language.
2. Whether or not an item is stored in memory is a function of "entrenchment"
(in turn a function of frequency of successful use; Langacker 1987: 59).
Obviously, entrenchment is a matter of degree. There is therefore no clean
cut-off point between "stored" units and ad hoc constructed units.
3. This is not to deny that certain facets may be more intrinsic to an expres-
sion's meaning, and relevant to just about all its uses; nevertheless, even
highly central aspects can sometimes be defeated, and outweighed by other,
more circumstantial aspects. See Langacker (1987: 158-161).
4. As a matter of fact, Saussure (in the representation of his thought that has
come down to us) appears to shy away from the full implications of his
theory. Thus, he observes (1916: 160) that if, of the three "synonyms" re-
douter, craintire, and avoir peur, redouter did not exist, its meaning would
be shared out amongst its competitors. Saussure, therefore, appears to pre-
suppose the existence of a conceptual content, which is independent of lan-
guage, and which has to be lexicalised, some way or other.
42 John R. Taylor
5. This was the principal theoretical import of Berlin and Kay's (1969) work
on colour categories. Essentially, Berlin and Kay demonstrated for a se-
mantic domain (colour) the same kinds of universal constraints that Jakob-
son (1968) had claimed for phonology.
6. Tenninology, however, is far from unifonn.
7. In maintaining this distinction, Structuralist Semantics aligns itself with a
number of other ''two-level'' approaches to semantics (e.g. Searle 1980,
Bierwisch 1981, Kirsner 1993, Wunderlich 1993). Although these ap-
proaches may differ in their details (especially concerning the manner in
which "linguistic" meanings are represented and get projected onto ency-
clopedic meanings), a common theme is the assumption that linguistic
meanings are unitary, clearly-defmed entities, which lack the rich detail de-
rived from experience of the world. For discussion, see Taylor (1994; 1995:
ch. 14).
8. For Coseriu, "universals" are to be found, if anywhere, in extralinguistic
reality and its categorisation, not in significations. Thus the label "semanti-
ca 'universal'" is applied to both prototype theories and theories of neces-
sary and sufficient conditions (Coseriu 1990: 252). Coseriu (1977: 10-11)
even charges generative grammar with an exclusively onomasiological (and
therefore universalist) perspective: "la grammaire generative part de la
realite extra-linguistique designee, ou bien d'une pensee prelinguistique
'universelle' (c'est-a-dire non encore structuree par telle or telle langue), et
passe pour ainsi dire a travers et par-dessus les langues pour aboutir a la
parole." [generative grammar starts from designated extra-linguistic reality,
or from 'universal' prelinguistic thought (i.e., from thought which is not yet
structured by a particular language), and by-passes, so to speak, the lan-
guage system, in order to arrive at the utterance.]
9. The proverbial linguist from Mars, on learning that afauteuil is an object
for sitting on, which has arms, legs, and a back, could be excused for sup-
posing that a man giving a piggy-back to his young son, is a Jouteu;l. The
point of this flippant example, of course, is that word meanings are not the
"minimalist" (cf. Coseriu 1990: 263) constructs envisaged by Structuralist
Semantics, but are likely to be extremely rich in detail and background
(encyclopedic) assumptions. E.g. knowledge of seating objects pertains not
only to the parts of which they are composed, but also to how humans typi-
cally interact with these objects.
10. Concerning Pottier's analysis, Coseriu and Geckeler (1981: 42) do indeed
raise the question whether we are here dealing with "an analysis of linguis-
tic content" or "a description of a series of ... objects, which is to say, of a
part of extralinguistic reality". The authors maintain that although Pottier
begins his analysis by considering the objects as such, and the real-world
features that distinguish them, he proceeds to eliminate the linguistically ir-

relevant features, thereby arriving at the (linguistically) "pertinent features".
Still, it is legitimate to ask what these "pertinent features" are supposed to
be, if not the necessary and sufficient features of check-list theories.
11. For discussion of high and tall, see Dirven and Taylor (1988) and Taylor
(in press b).
12. Cf. Coseriu (1977: 10): "Poser I'existence des unites fonctionnelles ne sig-
nifie nullement qu'on n'admette dans chaque cas qu'une seule 'significa-
tion' (= acception), mais au contraire qu'on s'efforce justement de definir
les limites, donnees par la langue, a I'interieur desquelles une infinite d'ac-
ceptions peuvent se presenter." [To postulate the existence of functional
units by no means entails that we allow, in each case, only one 'significa-
tion', or reading; rather, we attempt to circumscribe the limits, set by the
language system, within which an infinity of readings are possible.] (Note
that in this passage, "signification" appears to be used in the sense of "de_
signation", while "unite fonctionnelle" corresponds to my "signification".)
13. For some observations, see Taylor (1995: 223ss.).
14. Particularly influential has been Brugman's (1981) analysis of over, subse-
quently elaborated by Lakoff(1987).
15. In Taylor (1996), I argued, within the Cognitive Grammar framework, for a
unitary, schematic account of the possessive morpheme in English, and
against the adequacy of prototype accounts.
16. Similarly, for many grammatical categories, it would be fruitless to search
for a unitary phonological representation. In this connection, it is notewor-
thy that while Jakobson (1936) insisted on the methodological necessity to
assign a constant, albeit highly abstract, semantic value to each of the Rus-
sian cases (otherwise, he argued, the linguistic sign would fracture into nu-
merous fonn-meaning relationships), he was quite unperturbed by the ab-
sence of a unique phonological representation for each of the cases. If the
absence of a unique representation can be tolerated with regard to signi-
fiers, one wonders why polysemy should be outlawed with signifieds?
17. Coseriu notes that the phenomenon is not unknown in other languages, cf.
Gennan steigen. Or consider the English verb grow. His debts grow day by
day would be understood to mean that his debts get bigger (i.e. that they
grow "upwards"). (The example is mine, not Coseriu's.) But it is equally
possible to defeat the default interpretation: His debts are growing smaller
day by day. Coseriu would probably argue, therefore, that "upward motion"
is not intrinsic to the semantics of grow - the word "really" means 'change
in the vertical extent of an entity'. However, still other uses, e.g. The sound
of the music grew less as the band marched away (LDCE) suggest an even
more schematic sense, i.e. 'become', 'change in state'. But now, the seman-
44 John R. Taylor
tic content of grow has become so impoverished, that it is scarcely possible

to differentiate the word from other change-of-state verbs, such as become.
18. It is noteworthy that in addressing the linguistic value of climb, Coseriu
proceeds by abstracting what is common to a range of specific uses. Curi-
ously, he fails to implement what is surely the central idea of a structuralist
semantics, namely, the possibility of semantic contrasts between climb and
other items in the same lexical field.
19. But note that metonymic extensions are not based on similarity at all, but on
contiguity (within a conceptual domain). The development of English bead
had nothing at all to do with the "similarity", at any level of abstraction, be-
tween a prayer and a spherical object on a string.
20. Cf. Coseriu (1977: 17): "Les unites fonctionnelles correspondent d'une fa-
~on immediate a des intuitions globales unitaires." [Functional units corres-
pond immediately to global unitary intuitions.]
21. No doubt, the use of mouse to refer to the computer gadget, was once a "ca-
tegorizaci6n de emergencia". Now, however, mouse is the standard term.
(What else is one to call the thing?) Do we therefore say that the "value" of
mouse as a name for the small mammal has changed? Surely not. Mouse
has simply acquired an additional meaning, and the two meanings (which
are related in a fairly transparent way) happily coexist.
22. Structuralism is not the only conceivable alternative to a conceptualist se-
mantics. On a behaviourist semantics, knowledge of a word resides in fol-
lowing the rules for using the word correctly. This is the essence of Witt-
genstein's (1978: §43) aphorism that "the meaning of a word is its use in
the language".
23. The notion of "theories" can help explain the conundrum touched on at the
beginning of this section, viz., by what right can we base a theory of lan-
guage on such irredeemably, subjective entities as "concepts"? One answer
is, that each of us attributes to other people a mental life (replete with
"concepts") which is very much of a kind with our own, precisely on the
basis of a ''theory'' that other human beings function in much the same way
as we do. Cf. Fodor (1980).
References
Berlin, Brent and I\ay, Paul

1969 Basic Color Terms: Their Universality and Evolution. Berkeley:
University of California Press.
Bierwisch, Manfred
1981 Basic issues in the development of word meaning. In: Werner
Deutsch (ed.), The Child's Construction of Language. London:
Academic Press, 341-387.
Brugman, Claudia
1981 The Story of Over. MA Thesis, University of California, Berkeley.
Coleman, Linda and Kay, Paul
1981 Prototype semantics: The English word lie. Language 57: 26-44.
Coseriu, Eugenio
1974 Diachronie, Synchronie und Geschichte: Das Problem des Sprach-
wandels. (Intemationale Bibliothek fUr allgemeine Linguistik 3.)
MUnchen: Fink.
1977 L'etude fonctionnelle du vocabulaire: Precis de lexematique. Ca-
hiers de Lexicologie 29: 5-23.
1990 Semantica estructural y semantica "cognitiva". In: Manuel Alvar et
a1. (eds.), Profesor Francisco Marsa: Jornadas de Filologia, 239-
282. (CoHeci6 homenatges 4.) Barcelona: Publicacions Universitat
de Barcelona.
Coseriu, Eugenio and Geckeler, Horst
1981 Trends in Structural Semantics. (TObinger Beitrage zur Linguistik
158.) TObingen: Narr.
Cruse, D. Alan
1986 Lexical Semantics. Cambridge: Cambridge University Press.
Dirven, Rene and Taylor, John
1988 The conceptualization of vertical space in English: The case of
"tall". In: Brygida Rudzka-Ostyn (ed.), Topics in Cognitive
Linguistics, 207-229. (Current Issues in Linguistic Theory 50.) Am-
sterdam: Benjamins.
Fauconnier, Gilles
1985 Mental Spaces: Aspects of Meaning Construction in Natural Lan-
guage. Cambridge, MA: MIT Press.
Fillmore, Charles
1975 An alternative to checklist theories of meaning. Proceedings of the
Berkeley Linguistic Society I: 123-131.
1982 Towards a descriptive framework for spatial deixis. In: Robert J.
larvella and Wolfgang Klein (eds.), Speech, Place, & Action in
Deixis and Related Topics. Chichester: John WHey, 31-59.
1985 Semantic fields and semantic frames. Quaderni di semantica 6: 222-
254.
46 John R. Tay/or
Fodor, Jerry
1980 The present status of the innateness controversy. In: id., Repre-
sentations: Philosophical Essays on the Foundations of Cognitive
Science, 257-316. (Harvester Studies in Cognitive Science 13.)
Cambridge, MA: MIT Press.
Geeraerts, Dirk
1985 Cognitive restrictions on the structure of semantic change. In: Jacek
Fisiak (ed.), Historical Semantics - Historical Word Formation,
297-323. (Trends in Linguistics. Studies and Monographs 29.)
Berlin: Mouton de Gruyter.
Heine, Bemd
1993 Auxiliaries: Cognitive Forces and Grammaticalization. New York:
Oxford University Press.
1997 Possession: Cognitive Sources, Forces, and Grammaticalization.
(Cambridge Studies in Linguistics 83.) Cambridge: Cambridge
University Press.
Jackendoff, Ray
1983 Semantics and Cognition. (Current Studies in Linguistics Series 8.)
1990 Semantic Structures. (Current Studies in Linguistics Series 18.)
Jakobson, Roman
1936 Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der russi-
schen Kasus. In: id., Selected Writings, 2, 23-71. The Hague:
Mouton.
1968 Child Language, Aphasia, and Phonological Universals. The
Hague: Mouton. (Tr. by Allan R. Keiler of German version of
1941).
Katz, Jerrold and Fodor, Jerry
1963 The structure of a semantic theory. Language 39: 170-210.
Kirsner, Robert
1993 From form to message in two theories. In: Richard A. Geiger and
Brygida Rudzka-Ostyn (eds.), Conceptualizations and Mental Pro-
cessing in Language, 81-114. (Cognitive Linguistics Research 3.)
Koch, Peter
1995 Der 5eitrag der Prototypentheorie zur Historischen Semantik: Eine
kritische Bestandsaufitahrne. Romanistisches Jahrbuch 46: 27-46.
1996 La s6mantique du prototype: S6masiologie ou onomasiologie? Zeit-
schriftfiJr jranz(Jsische Sprache und Literatur 106: 223-240.
Komatsu, Lloyd
1992 Recent views of conceptual structure. Psychological Bulletin 112:
500-526.
Lakoff, George
1987 Women, Fire, and Dangerous Things. What Categories Reveal
about the Mind. Chicago: Chicago University Press.
1987 Foundations o/Cognitive Grammar. Vol l. Stanford: Stanford Uni-
versity Press.
1988 A view of linguistic semantics. In: Rudzka-Ostyn 1988, 49-90.
1991 Foundations o/Cognitive Grammar. Vo12. Stanford: Stanford Uni-
versity Press.
Lyons, John
1968 Introduction to Theoretical Linguistics. Cambridge: Cambridge Uni-
versity Press.
1977 Semantics. 2 Vols. Cambridge: Cambridge University Press.
Murphy, Gregory and Medin, Douglas
1985 The role of theories in conceptual coherence. Psychological Review
92: 289-316.
Pottier, Bemard
1964 Vers une semantique modeme. Traveaux de Iinguistique et de litte-
rature 2: 107-137.
Raible, Wolfgang
1983 Zur Einleitung. In: Helmut Stimm and Wolfgang Raible (eds.), Zur
Semantik des Franzosischen, 1-24. (Beihefie zur Zeitschrifi fUr fran-
zOsische Sprache und Literatur, N.F. 9.) Wiesbaden.
Rosch, Eleanor
1978 Principles of categorization. In: Eleanor Rosch and Barbara B.
Lloyd (eds.), Cognition and Categorization. Hillsdale: Erlbaum, 27-
48.
Rudzka-Ostyn, (ed.)
1988 Topics in Cognitive Linguistics. (Current Issues in Linguistic Theory
50.) Amsterdam: Benjamins.
Saussure, Ferdinand
1916 Cours de Iinguistique generale. Paris: Payot.
Searle, John
1980 The background of meaning. In: John R. Searle, Ferenc Kiefer and
Manfred Bierwisch (eds.), Speech Act Theory and PragmatiCS, 221-
232. (Synthese Language Library 10.) Dordrecht: Reidel.
48 John R. Tay/or
Talmy, Leonard
1988 The relation of grammar to cognition. In: Rudzka-Ostyn 1988, 165-
205.
Taylor, John R.
1994 The two-level approach to meaning. Linguistische Berichte 149: 3-
26.
1995 Linguistic Categorization: Prototypes in Linguistic Theory. 2nd
edition. Oxford: Clarendon Press.
1996 Possessives in English: An Exploration in Cognitive Grammar. Ox-
ford: Clarendon Press.
in press a Concepts and domains.
in press b Synonymy.
Wittgenstein, Ludwig
1978 Philosophical Investigations. Tr. by Gertrude E. M. Anscombe. Ox-
ford: Blackwell.
Wunderlich, Dieter
1993 On German um: Semantics and conceptual aspects. Linguistics 31:
111-133.
Diachronic semantics:
towards a unified theory of language change?
Helmut Ludtke
1. Introduction
In linguistic description, phonology, morphology, syntax and vocab-

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

bound to arise as regards semantics: will it be possible to fit changes
of meaning convincingly into patterns - let alone laws - similar to
those found in the other fields?
2. Cognitive diachrony
Ever since Hermann Paul (1966: 32 [§16]) it has been stated that the
main cause ("Ursache") of language change is normal human speech
activity (or performance); this is a basic tenet of empirical lin-
guistics.
When treating phenomena of language change it is advisable to
distinguish between an observation level relating to performance,
and a description level relating to language systems (competence);
while on the former we state what is happening or what is going on,
on the latter we declare either what is the case now or what has hap-
pened in the past.
The whole chain of events that leads from a given state of some
language system to another-may be conveniently divided into three
stages: OUTSET ~ INTERMEDIA TE ~ OUTCOME. The outset of change
is human creativity in single speech acts; the intermediate is imita-
tion and repetition ("following suit"), the outcome, difference in
state. The latter is described as a property while the former two are
observed (if possible) as phenomena.
The above tripartite model:
CREATIVITY ~ IMITATION ~ DIFFERENCE
or, in other terms:
INNOVATION ~ DIFFUSION ~ RESULT
is applicable especially to performance phenomena that require
full consciousness on the part of both speaker and hearer, such as
lexical borrowing and coinage of new words or locutions. In these
cases, the outset may be a single act performed by one speaker. But
it is equally possible, and has occurred several times in my own ex-
perience, that a number of persons at different places within a short
period produce the same innovation without knowing about each
Diachronic semantics: towards a unified theory? 51
other. Even then we may speak of independent single events con-

stituting the first stage in what may result in a change, viz. lexical
enrichment. The second stage in the process will be a period when
speakerlhearers recognise a word or locution as being commonly
heard but still remember it coming into use during their lifetimes;
hence they consider it as a loan or a neologism. For the next genera-
tion of speakerlhearers, who learn the item in question during child-
hood, it becomes part and parcel of ordinary language behaviour:
this is the third stage.
Another type of cognitive development in a language is the se-
mantic expansion of existing lexical items through metaphor and/or
metonymy. As regards the former, we may distinguish between first
using a given word with a metaphoric meaning as an act of creativity
(stage 1), a period of imitation (stage 2) characterized by the speak-
erlhearers' awareness of using a (well known) metaphor, and a state
of affairs in which the hierarchical relationship between the basic
and the extended meaning of the word has become blurred; if this
has come about we have an example of polysemy. Such a cognitive
shift occurred with Engl. law and its counterparts in other European
languages: although - historically speaking - the facts are obvious
(legislation came first, "laws" of nature were discovered later), to
many people, especially intellectuals, the formerly metaphoric mean-
ing of the word appears to be basic; therefore, law has become poly-
semous. Incidentally, there are linguists who question the appropri-
ateness of my speaking of universal "laws" of language change 1 on
the ground that these "laws" are susceptible of being transgressed.
They obviously no longer regard the term law of nature as being
metaphoric.
While leaving the eternal dispute as to where polysemy ends and
homophony (or homography) begins to lexicographers, I would like
to mention one very extreme case, viz. Fr. voler 'to fly' and 'to
steal', with its latter meaning arising in the 16th century (and thereby
ousting the rival verb rober, which had had that meaning earlier). At
the beginning of etymological research linguists were not sure
whether voler 'fly' and voler 'steal' were one and the same word.
The missing link was to be found in the medieval practice of falcon-
52 Helmut Ludtke
ry (ars venandi cum avibus): voler un oiseau was used with refer-
ence to birds of prey, and hence the verb could be used as a euphe-
mistic metaphor implying swiftness, for 'theft'.
3. Metonymy
Metonymy seems to be based on some sort of "nearness": from ma-

terial contiguity (e.g. Lat. bucca 'cheek' > Span., Pg., Cat. boca, It.
bocca, Fr. bouche 'mouth,)2 via pars pro toto (e.g. Lat. visus 'faculty
of seeing' > It. viso, OFr. vis 'face') to the use of a word in a new
context as, for example, German bi/tig 'cheap' (discussed by Rudi
Keller, forthcoming), which formerly meant something between
'just' and 'adequate'. The semantic shift undergone by this adjective
results from the change in nouns it can qualify: from bi/tiger Preis
'just (or adequate) price' towards bi//ige Ware 'cheap merchandise'.
The tertium comparationis of this shift is some sort of metalinguistic
concept like BENEFITING THE CUSTOMER and, of course, reference to
the market situation.
Moreover, there is a special phenomenon which might be quali-
fied as "semantic encroachment". It may be generally described with
a term familiar from biogenetics: as a sort of cross-over (or recombi-
nation) between two signifiants (A; B) and their corresponding sig-
nifies (a; ~). The following example involves two (or rather three)
forms of the Latin verb esse 'to be' - es, eris (est) - with their re-
spective meanings:
SIGNIFIANT SIGNIFIE
A eris 'future' ex
B es 'present' 13
Figure 1: Latin> Spanisch 'thou art'
On the way to Spanish, eris 'thou wilt be' became eres 'thou art',
while es disappeared as a result of homonymic clash with (Latin) est
Diachronic semantics: towards a unified theory? S3
'is' > (Spanish) es. One may wonder how a grammatical item could
possibly, as it were, "exchange" its former future meaning (a) for
'present' (~). How can a meaning become detached from a form it
innately belongs to, in order to cross over to another form? If we rule
out purposeful language engineering, we have to look for an invisi-
ble-hand explanation, i.e. for some speaker/hearers' behaviour that
will account for the above example (cf. Keller 1997: 422-427).
We can imagine a scenario where es « es and < est) had both
meanings, '(thou) art' and '(he/she/it) is'. In some speech situations
this would be tolerable if the overall context eliminated ambiguity;
in other cases es would be misunderstood.
It has often been observed in European languages that verb forms
of the present are used with reference to future events (in connection
with an adverb); there seems to be a general rule allowing the substi-
tution of unmarked, or less marked forms for marked, or more
marked ones. So there would be nothing strange about Latin es being
used instead of eris; but what about the contrary? It would violate
the rule. Only in a particular situation when es (2nd person) would
clash with es(t) (3fd person), e.g. in slurred speech, would speakers,
realizing that they had been misunderstood, possibly resort to full
forms on repeating, such as *este (3 rd person), with a paragogic vow-
el (according to Latin sandhi rules reconstructible from Romance
dialect material). In order to emphasize that not the third but the sec-
ond person was intended they might choose the form eris, reasoning
that the hearers would certainly understand the present instead of the
future meaning, given the particular situation. 3 In a second stage of
the process ("imitation"), speakers with such experience in their
background would be tempted to use eris > eres prophylactically,
i.e. in order to ward off the possibility of misunderstanding es as an
allegro form of est. When additionally the old synthetic future tense
forms came to be more and more often replaced by periphrasis with
the auxiliary verb habere (esse-habes instead of eris > eres) the ab-
normality of eris being used with the present meaning dwindled, so
that eres eventually acquired the meaning "(thou) art".
Another example of semantic encroachment may be adduced.
Latin manducare originally meant 'to chew' whereas its Romance
54 Helmut Liidtk,e
descendants, i.e. Fr. manger, It. mangiare, Rum. minca, have acquir-
ed the meaning 'to eat'. The semantic "nearness" is as obvious here
as it was in the previous example. The "invisible-hand" explanation
is also similar although this is an instance of lexical and not gram-
matical cross-over. There was homonymic clash between some very
frequent fonns of edere (the classical word for 'to eat') and those of
esse 'to be', and the table illustrating the situation is analogous:
SIGNIFIANT SIGNIFIE
A manducare 'chew' a
B edere 'eat' J3
Figure 2. Late Latin 'to eat'
In the history of Latin> Romance the case is a little more compli-

cated since two different words came to be used instead of edere:
manducare 'to chew' and comedere 'to eat up'. The latter survived
in the Iberian Peninsula (> Span., Pg. comer). The Emperor Augus-
tus is reported to have used manducare with the new meaning, while
in Bible Latin both words occur along with classical edere (Garcia
de la Fuente 1994: 85-88 and 126-162). As time went by, many
words that had rather been somewhat colloquial variants (like cabal-
Ius 'horse' besides equus) came to be considered as vulgar (in the
sense of 'not used in educated writing'). This happened also to man-
ducare meaning 'to eat' so that it was a shock for educated Romans
to read Jesus' words at the Holy Communion: "Accipite et mandu-
cate! Hoc est corpus meum" (Matthew 26,26 in the Vetus Latina, cf.
Hilicher 1938. In the Vulgate manducate was corrected into com-
edite).
4. Randomness or directionality?
Both the manducare and the eris case have in common that, as an
outcome of the whole process, the more frequent signijie wins the
Diachronic semantics: towards a unified theory? SS
(in a pseudo-Darwinian sense) fitter signifiant over to its side: A

goes with ~ leaving a. to fmd some new partner, i.e. in the eris case
to the periphrastic future tense form, in the case of manducare to a
loan word masticare borrowed from the Greek f.la.crnx,a.m 'to gnash
the teeth'. During the intermediate stage 2 there must have been a
period of polysemy; but since the hearer would be more easily in-
clined to interpret the word in question giving it the more frequent
meaning, this recombination eventually gained the upper hand.
The general speaker-to-hearer strategy (at stage 1) that brings
about the cross-over may be formulated as follows: if a frequently
occurring meaning is attached to a precarious form (i.e. one liable to
misunderstanding) don't worry! Use a makeshift, choosing another,
semantically close, form and trust the hearer! He will surely be eager
to make good sense of what you are saying. Once people have been
successful with this maxim, both as speakers and as hearers, the pro-
cess is in stage 2, when "collusion" becomes normal: communis er-
ror facit ius. The cross-over is accomplished when a new form is
established for the less frequent meaning a..
There are further instances of semantic encroachment and cross-
over. In contemporary French, the equivalents of Engl. day, evening,
year are patterned couples:
jour so;r an
journee soiree annee
Figure 3. Time units in French
The upper row stands for 'time' in a strictly chronological per-

spective, while the derived words in -ee refer to the respective peri-
ods as 'being lived through' (le temps vecu). This is the general rule.
It is violated, however, in speech practice since an exception is quite
regularly made for annee, because its partner an is "precarious".
This means that speakers shrink from using it in certain contexts and
replace it by annee even though, strictly speaking, the general se-
mantic rule would require an. The expressions from day to day, from
evening to evening, from year to year have as their French equi-
56 Helmut Ladtlce
valents de jour en jour, de soir en soir, d'annee en annee, and not, as

one might expect, "'d'an en an (pronounced /danana!). The reason,
cacophony, is obvious.
In a similar way, several hundred years ago, there happened an-
other cross-over with French terms referring to time:
B hui aujourd'hui A
13 'today' 'nowadays' a
Figure 4. French adverbs of time
The meaning 13 occurs, of course, more often than a. The form B

seems to have been precarious for its shortness or maybe for its pho-
netic similarity to oui 'yes'. Anyway, a cross-over ~ ~ A took place
(~ ModFr. aujourd'hui 'today'), a coming to be expressed by de
nosjours.
Classical Latin had two words meaning 'to go': ire for normal
and vadere for rapid, violent, or purposeful movement; in addition
there was ambulare 'to walk'. In colloquial language, since the fIrst
century A.D., ire and vadere were put together in a single paradigm:
vado - vadis - vadit - imus - itis - vadunt, and the slight difference
in meaning disappeared. In some Gallic Latin texts from the begin-
ning of the sixth century onward, we fInd the then shortest forms
imus - itis replaced by the corresponding forms of ambulare: vado -
vadis - vadit - ambulamus - ambulatis - vadunt, which develops
into Fr. vais - vas - va - allons - allez - vont. Obviously, 'to go' oc-
curs more often than 'to walk', and both meanings are similar. We
fInd a cross-over according to the model already mentioned above:
B imus / itis ambulamus / ambulatis A

> allom / allez
13 '~e/you go' 'we/you walk' a
Figure 5. Suppletion in Late Latin

The outcome again was p ~ A, Le. nous allons / vous allez 'we/ you
go'. The meaning 'to walk' was then expressed by denominal
cheminer « camminare) or by marcher. borrowed from Germanic.
In all these instances we find the same process of recombination
of the stronger (Le. more frequent) meaning with a phonetically
stronger form.
s. Grammaticalization
This is an often used (or should I say "abused"?) term whose mean-
ing is far from clear. It will be applied here to a change consisting of
lexical items turning into grammatical ones. Since the reverse ap-
pears to be extremely rare, gramrnaticalization may be regarded as
another "one-way road", i.e. a directional, not random shift, compa-
rable therefore with semantic cross-over.
A good cross-linguistic example is the formation of future tense
forms from verbs meaning 'to go', a process that has happened in
French, Spanish, Portuguese, and English (cf. je vais chanter / voy a
cantar / vou cantar / I am going to sing). It should be stressed that
the shift occurred independently in these four languages. In three of
them, the go-future coexists along with an older future tense form
(Fr. je chanterai / Sp. cantare / Engl. I'll sing ); only in colloquial
Portuguese has vou cantar practically supplanted cantarei and he i-de
cantar.
How is the phenomenon to be accounted for? Necessity can safely
be ruled out: there was no semantic "empty slot" to be filled. If it is
reasonable at all to speak of a chain shift, it is a push-chain, not a
drag-chain. The question is therefore: what may have induced speak-
erlhearers to confuse LOCOMOTION with ENSUING ACTION OR STATE
(since two verbs with these types of meaning are involved in the
process)?
Purposeful locomotion in the present has something in common
with volition and obligation in that it implies another possible, or
rather probable, event in the future. Therefore, verbs denoting these
states of the mind (or of the body in the case of 'to go') are candi-
58 He/mut Lad/Ice
dates for the auxiliarization of future tense. If it happens that the

speaker's and the hearer's centres of interest do not coincide - the
former expressing his present locomotion, the latter paying more at-
tention to the ensuing event - it is a question of chance whether the
hearer's interpretation will come to the fore, outweighing the speak-
er's intention. The next step, and the decisive one (the real innova-
tion!), will be to use the locomotion verb even when no bodily
movement is referred to. In doing so more or less regularly, speaker/
hearers enter into the "collusion" phase (stage 2 of the process) when
ambiguity is tacitly accepted by both participants in discourse. The
speaker vs. hearer dialectics leads us over to considering a possible
interplay between the learner and the competent adult. Some such
interplay must necessarily be assumed if the change from lexical to
grammatical item is to be accomplished.
6. Quantitative vs. qualitative change
The most difficult question that remains to be answered is whether

apparently directional changes in the field of cognition can be func-
tionally related to such quantitative diachronic universals as irre-
versible phonic shrinkage, lexical and/or syntactic accretion, and fu-
sion (or merger) of contiguous items in the spoken chain. These can
be handled resorting to notions of information theory (as shown in
Liidtke 1980a: 10-15). But since information theory categorically ex-
cludes semantics and cognition, the question remains open as to how
the findings that suggest directionality in the latter fields can be
accommodated to those stated in purely quantitative terms.
Notes
1. LUdtke 1980a, 14; 1980b: 205-216; 1986: 14-31; 1988: 1634; 1996: 531-
532.
2. More examples of metonymic extension and deplacement, from the Latin >
Romance domain, are to be found in Wright 1994: 68-105.
3. Cf. the description of a similar process by Heine 1997: 77.
References
Garcfa de la Fuente, Olegario

1994 Latin biblico y latin cristiano, Madrid: Ed. CEES. Second revised
and augmented edition of the Introduccion al latin biblico y cris-
tiano. Madrid: Ediciones Clasicas 1990.
Heine, Bemd
1997 Possession. Cognitive Sources, Forces, and Grammaticalization.
Cambridge: Cambridge University Press.
JUlicher, Adolf (ed.)
1938 Itala, Volume 1. Berlin: de Gruyter.
Keller, Rudi
1997 Kodewandel. In: Roland Posner, Klaus Robering and Thomas A.
Sebeok (eds.), Semiotik / Semiotics. Ein Handbuch zu den zeichen-
theoretischen Grundlagen von Natur und Kultur / A Handbook on
the Sign-Theoretic Foundations of Nature and Culture, Volume 1,
414-430. (HandbUcher zur Sprach- und Kommunikationswissen-
schaft 13.) Berlin / New York: de Gruyter.
LUdtke, Helrnut (ed.)
1980a Sprachwandel als universales Phllnomen. In: id. (ed.), Kommunika-
tionstheoretische Grundlagen des Sprachwandels, 1-19. Berlin /
New York: de Gruyter.
1980b Auf dem Weg zu einer Theorie des Sprachwandels. In: id. (ed.),
Kommunikationstheoretische Grundlagen des Sprachwandels, 182-
252. Berlin / New York: de Gruyter.
1986 Esquisse d'une theorie du changement langagier. La linguistique
2211: 3-46.
1988 Grammatischer Wandel. In: Ulrich Ammon, Norbert Dittmar and
Klaus J. Mattheier (eds.), Sociolinguistics / Soziolinguistilc. An In-
ternational Handbook of the Science of Language and Society / Ein
internationales Handbuch zur Wissenschaft von Sprache und Ge-
sellschaft, Volume 2, 1632-1642. (HandbUcher zur Sprach- und
Kommunikationswissenschaft 3.) Berlin / New York: de Gruyter.
60 Helmut Ladtke
1996 Changement linguistique. In: Hans Goebl, Peter H. Nelde, Zden~k

Swy and Wolfgang WOlck (eds.), Kontaktlinguistik / Contact Lin-
guistics / Linguistique de contact. Ein internationales Handbuch
zeitgenossischer Forschung / An International Handbook of Con-
temporary Research / Manuel international de recherches contem-
poraines, Volume I, 526-540. (Handbtlcher zur Sprach- und
Kommunikationswissenschaft 12.) Berlin / New York: de Gruyter.
1997 Sprachwandeltheorie: Erkenntnisse und Probleme. In: Maren Huber-
ty and Claudia Perlick (eds.), Studia Historica Romanica. In honor-
em Johannes Klare, 61-83. (Abhandlungen zur Sprache und
Literatur 90.) Bonn: Romanistischer Verlag.
in press Teorla del cambio linguistico.
Paul, Hermann
51920 Prinzipien der Sprachgeschichte. Halle: Niemeyer. Reprint Ttlbin-
gen: Niemeyer '1966.
Wright, Roger
1994 Early Ibero-Romance. (Estudios lingtlfsticos 5.) Newark: Juan de la
Cuesta.
Why do new meanings occur?
A cognitive typology of the motivations
for lexical semantic change
Andreas Blank
1. Language change and the speaker's mind
1.1. Three levels ofmotivation
One of the central topics of linguistics is the problem of language

change. It is especially intriguing to structuralists: why should well-
balanced linguistic systems change and thus put the stability of the
system at risk?
The challenge to structural linguistics - and linguistics in general
- presented by language change was formulated and discussed some
forty years ago by Eugenio Coseriu (1958). He states that language
is, rather than a product (ergon), a process (energeia) that exists vir-
tually in the speaker's mind as a mere potentiality (dynamis) and
fmds reality only in concrete utterances. While communicating, we
reify what is in our mind and thereby reinvent language every time
we speak. Speaking - and writing, of course - is the only way to in-
troduce innovations that might be adopted by other speakers and
thus become new language rules (Coseriu 1958: 44-46). It is the
dynamic view of language as something that is reified every time
someone speaks that bridges the "aparente aporia del cambio lin-
guistico" [the apparent aporia of linguistic change] (1958: 7). !
I
The differenciation between innovation and adoption by other

speakers leads us to a first typology of motivations: there is, on one
side, the motivation a speaker has while he is innovating and, on the
other side, the motivations other speakers feel to adopt this innova-
tion, e.g. the prestige of the innovating speaker, or the adequacy or
62 Andreas Blank
persuasive character of the innovation itself (cf. Blank 1997: 374-

375). Adoption thus shows a sociolinguistic ia.cet, which we cannot
investigate here (cf. e.g. Labov 1972), a pragmatic and a cognitive
side: when speakers decide to accept an innovation because it is
convincing to any extent, this is a pragmatic decision mainly based
on the innovation's good cognitive perfonnance. If, for example, we
find that to trash is a good expression for 'erasing files or e-mails
from the computer', speakers decide to use it for that purpose be-
cause it is "convincing"; the persuasive power of this metaphor lies
in the clear-shaped similarity between the concept THROWING WASTE
INTO THE GARBARGE CAN and TO ERASE DATA. As a consequence of
adoption, the semantic innovation becomes lexicalized (cf. Blank, in
press, section 2).
On the cognitive level, lexicalization "confirms" that the innovat-
ing speaker has made a good choice. Ifwe want to explore the cogni-
tive aspect of semantic change, we may therefore safely concentrate
on the motivations for semantic innovation and examine lexicalized
material. As to the reasons why speakers innovate, Coseriu (1958:
37-38) differenciates between three levels of causes or motivations: 1
• the gen&;l~otivation for language change, which, according to

Coseriu, is the "finalidad expresiva (y comunicativa) de los hab-
lantes" [the speakers' expressive and communicative purpose]
(Coseriu19~8: 116),
• the gene(~lconditions for language change, and
• the specific motivation for a concrete innovation.
~ ....... .-----
On the level of specific motivations, we have to reveal the con-

crete conditions behind an innovation: when the American inventor
of the 'self-propelled submarine explosive' baptized it torpedo after
the Latin name for the electric ray, the very reason why he did it was
his need for a name for the new weapon. He probably called it tor-
pedo, because he thought that this was a pretty convincing meta-
phor. 2 The specific motivation was, therefore, the need for a new
name in a concrete situation, and the general motiva~on for the inno-
Why do new meanings occur? 63
vator's choice was the wish to give it a somewhat suggestive and

hence successful name. Only by combining the individual and the
general aspects can we provide satisfying explanations for semantic
mnovadOHS. 1 Typologies of the motivations for semantic change do
not always distinguish clearly between these ~vels, but mainly
focus on the specific motivations in order to fiM.regularities. By this
process of abstraction, we can empirically group individual motiva-
tions into ''types of motivations", that correspond to a certain extent
to Coseriu's second level, insofar as they constitute the sufficient
conditions for change in the part of language that interests us. 4 In
section 2 and 3 we shatl matriIy concentrate on these typical motiva-
tions for semantic innovations.
1.2. The ge~otivation for language change,

expres~~'~~d efficiency
The general motivation for language change is widely accepted and
was summed up as early as 1949 in a rather unpretentiously and
slightly behaviouristic manner by George K. Zipf: "Man talks in or-
der to get something." (1949: 19). According to Zipf, the main moti-
vation for speaking is to achieve success. This hypothesis implies
that language change is a mere side-effect of the speakers' pragmatic
goals. In Rudi Keller's adaption of Adam Smith to linguistics (Keller
1994, esp. ch. 4) natural languages are defined a "phenomenon of the
third kind", a product of social interaction, created through a series
of "invisible-hand processes" by speakers whose main purpose was
to "get something". Thus, language change is not inherent in lan-
guage itself, although it occurs at any time in any language, nor are
there certain tendencies (e.g. "simplification") inherent in language.
Rather, language change is a consequence of inherent characteristics
of man's mind and human social interaction.
The pragmatic view of language implies that speakers want to
communicate successfully and that they produce innovations any
time they judge it to be the most successful strategy. Innovations can
be produced with regard to the speaker him- or herself or with regard
64 Andreas Blank
to the interlocutor (cf. also Geeraerts 1997: 116-119). 5 Speaker-ori-

ented strategies aim at reducing linguistic effort by shortening
words, integrating "orphaned words" into the lexicon (cf. section
3.5.2.), making use of analogies etc.; in short: they aim to render
speaking more efficient. In certain communicative situations speak-
er-oriented efficiency can also be increased by using the word for the
prototype instead of the category (and vice versa) or by creating a
metonymy or a metaphor, because the underlying concepts are more
salient and thus allow easier access to the corresponding words (cf.
section 3 for details).
Such strategies enable speakers to maximize the communicative
relevance of their discourse (cf. SperberlWilson 1986, esp. 157-159).
Enhancing relevance is already a hearer-oriented strategy to the ex-
tent that a rarticularly relevant communication seems to be more
convincing. Hearer-oriented strategies generally aim at assuring the
correct understanding of what the speaker wants to express and at
influencing the hearer in favour of the speaker. Typical lexical stra-
tegies are to avoid ambiguity ("disturbing" homonymy or polysemy)
or to use more explicit complex words, to create expressive or eu-
phemistic metaphors, metonymies etc.
While speaker-oriented strategies are more or less used to in-
crease communicative efficiency, hearer-oriented strategies seem,
furthermore, to be oriented towards the communicative principle of
expressivity. Dirk Geeraerts (1983, 1997: 104-106) sees expressivity
at work, when speakers verbalize newly introduced or differently
perceived concepts or give a new stylistic use to an already existing
word (esp. by creating euphemisms). While in Geeraerts (1983) the
distinction between "expressive factors" and "efficiency principles"
is rather strict, both conceptions are later defined as being "comple-
mentary sides of the same coin" (1997: 108): expressivity playing
the role of the force creatrice behind language and linguistic innova-
tion and efficiency being used for optimizing what has been created
by means of expressivity, following the well-known principle of
least effort (cf. also Geeraerts, this volume).
Geeraerts uses the term "expressivity" in a rather wide sense (cf.
Geeraerts 1997: 92), and, in this wide sense, we can surely admit
that expressivity and efficienc can be treated on the Same level as

comp emen conce ·ons. Expressivity, however, is usually un-
erstood as an aspect of certain speech acts by which speakers ex-
press real or faked emotions (cf. Bally 1965: 75; Fiehler 1990: 222-
223; Mair 1992: 32-61). I therefore consider Geeraerts twofold typo-
logy of the general motivations (Q!.~~e_change problematic,
especiallyslnce1ilSwide use of the two terms may Ieaa iiito" anlm-
passe, as new creations can be efficient without being strictly expres-
sive at all:
(1) OFr. panser 'to care for a wounded person' > 'to bandage a
person's wounds'
(2) Engl. glass 'a hard, transparent substance' > 'drinking contai-
ner made out of glass' > 'glassful'
In my opinion, efficie~.9:~9:. ..~:c~ressi:Y.~t.L~~.P-ot !!~.~~~.~~.:~mpa

rable or.c~IE-.pl~~!!.~_c:>,!!.~th~ ,J,~~J~Y,,!i, If we bear in nuncrThat
spea:Kei-'S'oon't want to change their language, and that their principal
goal is to communicate as successfully as possible and to reach this
communicative goal with minimal linguistic effort, then communica-
tive success is efficient in a rather abstract sense. I would like to call
this the efficiency 0/ communication: it is, in other words, the general
purpose of communication and the general motivation behind lan-
guage change.
It is true, of course, that avoiding paraphrases or complex words
and creating metaphors, metonymies or ellipses instead is efficient
too, but, in this case, the efficiency operates on a more concrete
level, and it is indeed a linguistiC-strategy speakers adopt more or
less Willingly. On the same level, expressivity is also a strategy that
speakers can adopt for optimizing their communicative success
when they want to impress their interlocutors, treat him or her gent-
ly, manifest emotions, show things under a different light etc.; in
short, when they want to come out on top. Both strategies then con-
tribute in their specific manner to the general efficiency of communi-
cation and of language change.'
66 Andreas Blank
To conclude this section, we can say that efficient communication

is to maximize success by either reducing or increasing linguistic
8
effort. Thus, a study of the motivations for lexical semantic change
must be concerned with both developments, gather individual moti-
vations for particular changes into type(mid try t.? iden!!frt!l:~.~Qgni
tive basis for these types. This will be tne main Issue of section 3.
'Section 2 will look back into the history of semantics and discuss
some of the problems raised by traditional typologies of the motiva-
tions for lexical semantic change.
2. Traditional approaches to the motivations for semantic

change
2.1. Stephen Ullmann's typology
The essentials of the 19th and early 20th centuries' reflections on

lexical semantic change are summarized in Stephen Ullmann's two
books Principles 0/ Semantics (1957) and Semantics (1962). In the
latter, Ullmann distinguishes three aspects of semantic change: its
"causes", "nature" and "consequences" (cf. the critique in Geeraerts
1983, 1997: 85-92; Blank 1997: 34-44). This has been for decades
the most popular and important theory in this domain. As to the cau-
ses o/semantic change, Ullmann gives a list of six types (1962: 197-
210), which integrates other, less comprehensive typologies, as e.g.
those of Meillet ([1905] 1965) and Sperber ([1923] 1965).
A closer look at Ullmann's typology in the light of what has been
said in section 1 is rather deceiving: Ullmann's list lacks both a cog-
nitive and an .em..E!~!£alJ~ac~~~ and is merely an eclectic collec-
tlono(ffioilvations, necessary con<!!!!.~" ~d acces~_ elements.
9
From his six types only two are relatively unprobiematic, i.e. the
"social causes" and the "psychological causes". The first type con-
cerns words that are often used by a group of speakers in a ~cted
a
conteXt and, as result, becomesemantlcaIly restriCted to the actual
sensethey have in this context:
(3) Lat. cubare 'to lie' > Fr. couver 'to breed'
Vice versa, the meaning of a word can become generalized when

this word is used outside its usual context:
(4) Engl. lure 'decoy' > 'anything that attracts'
The theory of the psychological grounding of semantic change

was introduced by Hans Sperber (1965). According to him, emotion-
ally marked concepts can serve as an onomasiological "center of at-
traction" for other words to verbalize the "attractive" concepts and,
vice versa, serve as a cognitive basis, as a semasiological "center of
expansion" for verbalizing other concepts (cf. also Blank 1998a: 15-
17). Concepts within emotionally marked fields (cf. section 3.7.) are
frequently akin to new verbalization, be it in an expressive, "exagge-
rating" way, as in ex. (5), or in the manner of euphemistic "under-
statement", as in ex. (6):
(5) MHGerm. sere 'wounded, sore' > 'very'

(6) VulgLat. male habitus 'in a bad state' > Fr. malade, It. malato
'ill'
While the relevance of these two types is convincing, this is not

the case for the remaining four. According to Ullmann, a semantic
change can be due to newly developed objects or ideas (need for a
new name, cf. ex. 7), to technical, scientific, political or sociocultural
developments which influence our conception of things, people,
ideas etc. (historical causes, cf. ex. 8), as well as to changes that
have already occurred in one language and that are subsequently co-
pied in another (foreign influence, cf. ex. 9). Finally, change can
happen when two words are habitually collocated in speech and the
sense of one word is transferred to the other (linguistic causes, cf.
ex. 10):
(7) Engl. torpedo 'electric ray' > 'self-propelled submarine ex-

Plosive' I.'
, / \ 01 , _ ' "
68 AmJreas Blanlc
(8) Fr. plume 'goose feather for writing' > 'pen with metal nib'
(9) OGr. aggelos 'messenger' > 'angel', copying the polysemy of
Hebr. ml 'k 'messenger', 'angel'
(10) Fr. pas 'step' > 'not' (~ ne ... pas, lit. 'not (a) step')
First of all, I believe that Ullmann's historical causes and need

for a new name are facets of one and the same type, insofar as in
both cases there are new concepts - the submarine bomb, the mo-
dem pen - that need to be expressed. 10 The same holds true for for-
eign influence, where it is not the adstrat situation itself that is the
cause for semantic change, but the need to verbalize a new concept.
Semantic loan is only a rather smart device adopted for that purpose,
because it imitates a polysemy that seems to function well in another
language and which thus promises to be successful.
As to Ullmann's example for linguistic causes, it must be ob-
served that in fact the sense of the whol~ colJocatiQn. is transferred to
the simple word ~d not only the sense of the part that is omitted.
Lexical ellipsis, therefore, must be defined in a different way than it
is interpreted by Ullmann (cf. Blank 1997: 288-292). In our context
it is important that the collocation of French ne and pas is not the
motivati?.!!.f~~_!!!.~_~~~~ti~_.cllange ..wi1j.9Jtpas h~_~~_t?rgo.!1Jh..hut a
necessary con~itj9n __that.Jl:mkes . a._change possible. If it was the
motivati()n, all collocations (and complex lexemes in general) would
necessarily become elliptic, which is clearly not the case.
2.2. Other causes ofsemantic change
We can thus reduce Ullmann's six types to three. Furhtermore, ~e

find that his typology lacks traditional causes such as "irony" or
what Nyrop called "conneXlOn entre les choses" [relation between
the things] (1913: 80). To my knowledge, Nyrop was the first to
state that concepts in our mind are interconnected and that one
- 11
concept can evoke those concepts related to it. ~ptual
relations seem....m induce semantic change, as happened in the
following example: 12
(11) Lat.focus 'fireplace' > 'fire'
Another traditional cause for semantic change that does not

appear on Ullmann' s list is homonymic clash:
(12) Lat. vicarius 'village mayor' > Gasc. bigey 'rooster'

(~ Lat. gal/us 'rooster' > Gasc. ·gat; LateLat. cattus 'cat' >
Gascongat)
The traditional interpretation, which goes back to Gillieron and

Roques (1912), sees the semantic change of vicarius motivated by
the homonymy of Latin gal/us and gattus in Gascon, caused by the
parallel sound change of intervocalic -1/- > -t- and -tt- > -t-. In a
medieval rural society this homonymy could indeed have given rise
to confusion. The question is whether the metaphor for vicar ius was
only created
___._.~when-·theliomoii'-iC--cr~'·li'-"alread--occure(i-·
__...______ ._. _-____ .l}'!J:! .•.. ,.•..,.Mh. ,ad,,, ..... __ . _.~, _. .~_ In
agreement with Wartburg (1971), I would rather say that the already
existing expressive "Trabantenworter" [satellite words] vicariuslbi-
gey were used more and more as the normal names for the 'rooster'
when the homonymy arose, while the original word gat was falling
into disuse. Thus, the homonymy of gat is not the m<?tivation fQ!Jhe
metaphor bigey but tlie cause of its ~~_..!!~~U~~ ..!19Q11.~IYiQffi..for
'rooster' ID Gascoii~an-conchide that the avoidance of misun-
derstandings due to homonymy (as well as to polysemy) is situated
on the level of motivations for adoptions of innovations, while the
creation of the metaphor of Gascon bigey 'rooster' was due to ex-
pressivity, viz. "psychological causes" (for details cf. Blank: 1997:
354-359).
70 Andl'eas Blank
3. A new typology of the motivations for lexical semantic

change
3.1. An empirical approach based on cognitive foundations
Traditional historical semantics have not succeeded in conceiving a

consistent theory of the motivations for semantic change, but rather
they provide a mix of mechanisms, such as metaphor, necessary (but
not sufficient) conditions, such as collocations, and actual causes. 13
Furthermore, rarely any consideration is given to the different levels
of motivations introduced in section 1.1. Considering the eclecticism
of a typology like Ullmann' s, which until the 1990s was reputed to
represent the "state-of-the art" in historical semantics, one should fa-
vour an empirical approach to this matter in order to reduce the risk
of omittin es that are less striking than others. Such a typol;gy
should concentrate maJ. y on the moiIyatlons for semantic innova-
tion. [If It is self-evident that recent developments in pragmatics and
;emantics (esp. cognitive semantics) should be integrated whenever
they support classifying concrete examples and further theorizing in
this domain of historical semantics.
A corpus of more than 600 particular semantic changes taken es-
sentially from Romance languages and completed by examples from
German and English as well as from some other languages was
compiled and classified by Blank (1997: 497-533). The items were
mainly collected in view of a classification of the mechanisms for
lexical semantic change and not with regard to its motivations. This
assures a certain randomness and objectivity of the corpus with re-
spect to the motivations for semantic innovation.
For each exam e I tried to find the most plausible reason for its
c~[Qn, and then Isimilar cases were groupe together. I have to ad-
mit that some examples simply refuse classification, as the specific
conditions and motivations for their creation have not yet been
establishedY The great majority, however, fits into six main cate-
gories, some of which show subdivisions, which I would like to call
linguistic (section 3.6.) or cognitive constellations (sections 3,5.,
3.7.) favouring semantic change. The six main types of motivations
are established empirically, but show significant correlations to theo-

retical frameworks of cognitive semantics, viz. prototype theory,
basic-level theory and frames-and-scenes semantics. Furthermore, all
six types are consistent with the general motivation for language
change, insofar as their common denominator is to enhance commu-
nicative efficiency.
3.2. New concept (need for a new name)
Our first type has already been discussed in section 2. New concepts
arise when we change the world around us or our way of conceiving
it (ex. 13 and 14), but also when we leave our "habitat" and enter a
new one (ex. 15 and 16). New concepts can of course be verbalized
by paraphrase, but it is more efficient, and in most cases more perva-
sive, to express them by semantic change. The common background
of all four examples is the confrontation with a new referent or con-
cept:
(13) Lat. pecunia 'cattle (used as a currency)' > 'money'

(14) Engl. mouse 'small rodent' > 'small, hand-guided electronic
device for executing commands in computer programs'
(15) EurSp. lean 'lion' > AmerSp. 'puma'
(16) Fr. lhard 'lizard' > Creole (Reunion) 'gecko'
3.3. Abstract concepts, distant and usually invisible referents
The second type ofmotivations concerns conceptual domains whose

referents are either abstract or usually distant or hard to see and thus
rather difficult for us to seize intellectually. Abstraction explains the
usual metaphorical verbalization of e.g. TIME, UNDERSTANDING,
SENSE-PERCEPTION or EMOTIONS (cf. Sweetser 1990: 32-37; Blank
1998b: 19-23). Verbalizing abstract concepts by metonymy occurs
more rarely:
72 Andreas Blank
(17) Lat.luna 'moon' > Rum.lunii 'month'
It can happen that, with time, metaphors (or metonymies) lose

their concrete sense and become opaque, as happened in ex. (18):
(18) It. capire 'to understand' « Lat. capere 'to catch')

In this case, the cognitive perception of the mental process by
means of physical grasping is no longer possible. Here indeed, Ital-
ian speakers were quick to create and adopt a new metaphor that re-
news the old conceptual link between UNDERSTANDING and GRASP-
ING:
(19) It. afferare 'to grip, to grasp' > 'to understand'

"",., ..-
Concretely ~le but distant concepts and things that are diffi-
cult to see can be brought "closer" into our view by metaphor, as in
ex. (20) and (21):
(20) Fr. gorge 'throat' > ~e, canyon'

(21) Lat. pupilla 'little girl or doll' > 'pupil (of the eye)'
This type of motivation once more illustrates the difficulty of ap-

plying the labels "expressivity" and "efficiency" on the level of lexi-
cal semantics: verbalizing an abstract or somehow difficult notion by
metaphor certainly is an expressive issue that makes communication
more convincing, but it is also - and mainly through its expressivity
- a very efficient tool of lexical enrichment.
3.4. Sociocultural change
Changes in our c,?nception of the world can also lead to the transfor-
mation of an already existing complex conceptual system by the loss
of one or more concepts, by shifting concepts or by introducing new
ones. For example, a change in the legal system made the Latin dis-
tinction between relatives on one's mother's side (avunculuslmater-

tera) and relatives on one's father's side (patruuslamita) obsolete.
This led, among other things, to the extensions of meaning in ex.
(22) and (23), which are a part of a greater restructuration in the
semantics of Latin kinship terms (cf. Coseriu 1978: 136-138):
(22) Lat. avunculus 'uncle on one's mother's side' > Fr. onc/e,
Rum. unchiu 'uncle'
(23) Lat. amita 'aunt on one's father's side' > OFr. ante, ModFr.
tante, Occ. tanto, Engd. amda, Rum. miitu~ii 'aunt'
Fascinating examples which demonstrate the multiple reorganisa-

tion of a lexical structure due to sociocultural change are the French
denominations for the different meals of the day: until the 16th cen-
tury people used to have the main meal (disner/desgeuner, both from
VulgLat. *disjejunare 'to defast') in the middle of the morning and a
second, lighter meal (souper) in the afternoon. Changes in lifestyle
of the noble society and in the (imitating) urban bourgeoisie shifted
the main meal of the day to noon. The long gap between getting up
and eating made it necessary to introduce a breakfast in the morning
and led to the semantic differenciation of dejeuner 'breakfast' and
diner 'main meal'; souper now meant 'evening meal'. In the 19th
century, it appeared more suitable to the members of urban profes-
sional societies to have the main meal in the evening, which made
diner to acquire the sense of 'dinner' and dejeuner 'lunch'. For
'breakfast', a new complex lexeme was created (petit dejeuner) and
souper now serves to designate a late-evening meal. The binary sys-
tem of the Middle Ages thus developed through a ternary to a four-
partite structure shifting the words along the temporal contiguity of
meals (cf. DHLF, ss.vv.):
74 Andreas Blank
Figure J. The diachrony of French words for MEALS
3.5. Close conceptual or factual relation
When we speak, it can happen that we use a word in a sense that is

different from its usual one. Normally, our interlocutor understands
what we mean because the context may help and because the word
we have chosen usually refers to a concept that is somehow closely
linked to the concept we have made it refer to in this concrete speech
act. Close links between concepts make name transfers possible and,
when they are considered tQ be eftic.i~n.t,ih~y mjght be£S2wekxical-
--
ized~-'and-the word that has undergone semantic change becomes
polysemous. According to the nature of this conceptual relation,
three types of cognitive constellations can be distinguished:
1. Frame relation: A strong and habitual relation between two con-

cepts within a frame makes speakers express them by using only one
word: the frame relation is "highlighted" (cf. Croft 1993: 348; Koch,
in press b). This was the type of motivation Nyrop called "connexion
entre les choses" (cf. section 2.2.). The examples of this type are
mainly metonymies (24), but frame relations may also lead to auto-
c~se change, as in ex. (25):
(24) Lat. testimonium 'testimony, witness' > Fr. temoin 'witnesser'

(25) It. noleggiare 'to lend' > 'to borr:ow'
Completely different frames led Latin plicare 'to fold' to take op-
posite semantic directions in Thero-Romance and Rumanian, produ-
cing a kind of "interlinguistic antonymy":
(26) Lat. plicare 'to fold' > Rum. a pleca 'to leave'
(27) Lat. plicare 'to fold' > Sp. llegar, Pg. chegar 'to arrive'
The reason for this curious development is probably the

following: in the shepherd society of Rumania folding the tents was
associated with leaving, while in the marine society of Spain folding
the sails was associated with arrival. 16
2. Prototypical change: In contrast with Geeraerts (1983), I would

like to restrict "prototypical change" to the following cases:
In the first case, a word is constantly used to refer to the prototype
of the usually designated category (cf. also Rastier, in this volume).
This gives rise to restrictions of meaning, as in ex. (28) and (29),
which rely on the fact that in France WHEAT is the prototypical CERE-
AL, and in patriarchal societies the prototypical HUMAN BEING is the
MAN:
(28) Lat. frumentum 'cereal' > Fr. froment 'wheat' > 'special type
of wheat', Fr. ble 'cereal' > 'wheat'
(29) Lat. homo 'human being' > VulgLat. 'male human being'
The opposite happens when the word which usually designates

the prototype of a category is extended to refer to the whole catego-
ry, as in the following examples:
(30) VulgLat. hostis 'opposed, hostile [!] army' > Olt. oste, Otr.
ost 'army' (> MEngl. (h)ost)
(31) Lat. tenere 'to hold' > Sp. tener, Sard. teniri 'to have'
Here we can say that TO HOLD (IN THE HANDS) is the clearest and
most typical instance of POSSESSION mirroring an anthropologically
motivated salience (cf. Koch 1991: 31, this volume), while the OP-
76 Andreas Blank
POSED ARMY fonns a kind of "moral prot0!.yp~" _because of its

elementaryJb.teat. Note that ftlePrototypical structure of a category
if-6ruyanecessary condition. The sufficient condition that triggers
this type of change seems to be the speakers' strong emotional or
factual fixation on the prototype. In the first case, they then use the
word usually linked to the whole category in a specific context
where only prototypical instances can be referred to (ex. 28 and 29),
e.g. in a public discourse when the orator addresses the whole com-
munity, but indeed speaks to an entirely male assembly. In the
second case, the word for the prototype is expressively taken to refer,
in fact, to another member of the category or the category itself (ex.
31). Both "abuses" will probably lead to a higher frequency of use of
the semantically extended or restricted word which makes their new
senses even more attractive.
The question is why in one case we have semantic restriction and
in another semantic extension. In my opinion, this depends essen-
tially on what's on the left of the arrow: if this is a word whose refer-
ential class is prototypically structurable (like CEREALS or HUMAN
BEINGS). then we can only have restriction to the prototype; if. on the
other hand, the referential class of the word we examine shows no
prototypical structure, but is itself a prototypical instance, then it can
only be extended to the whole category. We must keep in mind that
in both cases the fixation on the prototype triggers the semantic
change. 17
Sometimes we find extensions or restrictions of meaning which
are difficult to be explained by fixation on a prototype, as ex. (32)
and (33), where at first sight it is hard to imagine that one of the two
referential classes involved could be a prototypical instance of the
other:
(32) MEngl. hound 'dog' > 'dog trained to pursue game'

(33) VulgLat. adripare 'to get on shore' > 'to arrive'
In these cases; it helps to see if the transfer could have happened

inside a specific frame, and, in fact, the typical dog in the frame
"HUNTING" is the HOUND, and in the frame "SEAFARING" the proto-

typical arrival is GETTING TO SHORE.
3. Blurred concepts: So far, we have discussed cases where a name

transfer is considered to be efficient. In some cases, however, speak-
ers make transfers without being aware of it, because their knowl-
edge about the limits of these concepts and the respective categories
is momentarily or permanently1>lurreq. Confusions of this kind hap-
pen every day to almost everybody. If, however, the confusion be-
tween two concepts is a widespread matter of fact in a speech com-
munity, this can lead to a co-hyponymous transfer. A conceptual
field where changes of thiSkIiiOliave'happened several times in the
history of Romance languages is constituted by RATS, MICE and simi-
lar small animals (cf. for details Blank 1998c):
(34) LateLat. talpus ~ > It. topo, Sard. topi 'mouse'

(35) Lat. sorex 's~e' > It. sorcio, Fr. souris, Rum. ~oarece
'mouse'
(36) ? *ratt- 'rat' > Pg. rato, Fr. rat, It. (dial.) rat, rat 'mouse'
3.6. Complexity and irregularity in the lexicon
A fundamental speaker-oriented strategy is to communicate at the

"lowest possible costs". Consequently, speakers reduce iq(!,g,ularities
or superfluous complexity.it!Jhe Jexj"Oii~lll.QifQftiii·iiine wi.tpout
being- aware of ,it. rITcgUlarity may, in fact, be a consequence of
"expressive factors" and the efficiency principle pushing speakers to
linguistic optimization (cf. Geeraerts 1997: 108), but for the follow-
ing situations this is not necessarily the background. In any case, like
in other instances of semantic change, "good" innovations are subse-
quently adopted by the community. Four different lexical constella-
tions can be distinguished:
1. Lexical complexity: The more frequently a word is used, the more

speakers tend to reduce its signifiant ("Zipfs law"; cf. Zipf 1945:
78 Andreas Blank
142-144). If the word in question is a compound or a syntagmatic

construction (e.g. mother-in-law, matter offact), the reduction may
concern one part of the complex lexeme. This process is usually
called "lexical ellipsis". Semantically, a simple lexeme also receives
the meaning of a complex word of which it is formally a part, so that
"absorption" or "incorporation" (of the meaning of the complex lex-
erne) would be more appropriate terms (cf. Koch 1991: 287; Blank
1997: 291):
(37) It. portatile 'portable' > 'notebook-computer' (~ computer

portatile)
(38) Lat. separare 'to separate' > Fr. sevrer 'to wean' (~ MFr.
sevrer de la mamele) ~
The causal relation between lexical complexity, high frequency of

use and the semantic process of ellipsis is extraordinarily strong, but
it is not the exclusive motivation for ellipsis (e.g. to drink 'to drink
alcohol' is probably due to decency; cf. section 3.7.).
2. "Orphaned word": Here we have, in a certain manner, the oppo-

site scenario to the one presented in the last paragraph: a lexically
isolated word, which is the unique member of its derivational class
and rather restricted in use, is interpreted by speakers as belonging to
another derivational class and, therefore, becomes formally and se-
mantically integrated into this class. This process of lexical integra-
tion is usually called "popular etymology" or "reinterpretation" (cf.
Blank 1993, 1997: 303-308). Popular etymology can be restricted to
the phonetic or graphetic level (MEngl. soverein > sovereign, under
the influence of MEngl. reign). Where it occurs with semantic chan-
ge, the formal similarity is backed up by a conceptual relation
(usually contiguity), as in the following examples:
(39) Fr. forain 'non-resident' > 'belonging to the fair' (~ Frrl!:!:.e

'fair, market')
(40) OGr. nelcromanteia 'art of divination through communication
with the dead, necromancy' > ClassLat. necromantia >
LateLat. nigromantia 'black art', 'magic' (~ Lat. niger

'black')
While ex. (40) shows the fate of a loan word, which by definition
is lexically isolated, French forain (from VulgLat. foranus) was
previously "orphaned" in the French lexicon, because it didn't
participate in the sound change of its lexical root (Lat. foris 'outside'
> Fr. hors). Through lexical reinterpretation it became "adopted" by
the family offoire (from Lat.feria) in which it synchronically serves
as an adjective ifoire - forain following the scheme of gloire -
glorieux).
3. "Lexical gap": A lexical gap, in my understanding, is caused by

an asymmetric lexical structure. Such an asymmetry existed in Clas-
sical Latin with eques, meaning 'cavalryman' and, metonymically,
'knight' as a member of the social class that formed the cavalry in
the Roman army. One of the co-hyponyms of eques in the military'
sense was pedes 'infantryman'. The lexical disproportion between
the polysemic eques and the monosemic pedes incited speakers to
create an analogous metonymy for pedes which then meant also
'plebeian', although at least two other words for this social class
(plebs, plebeius) already existed.
Analogous semantic change induced by an already current poly-
semy is esp. frequent in slang and other linguistic substandards,
where the phenomenon is known under the name of "synonymic
derivation" (cf. Klein 1997: 35-36, citing earlier studies from Meillet
and Kroesch): synonyms or co-hyponyms of a word having a slang
sense develop the same or a similar slang sense. Synonymy or co-
hyponymy make the innovations easier to understand for the hearer
and thus reduce communicative nsks, whileguaranteeing at the
same time the specific cryptic character of slang. An example from
Meillet (cited in Klein 1997: 36):
(41) Fr. polir 'to polish', 'to steal' Q fourbir 'to polish' > 'to steal',
nettoyer 'to clean' > 'to steal', etc.
80 Andreas Blank
4. Untypical meaning/untypical argument structure: The last lexical

constellation that Can motivate semantic change concerns wc;;ds
whose meaning is somewhat untypical for the word class they
belong to. There is a general tendency to give more prototypical sen-
ses to such words. For example, nO~~_.~.~r.es.siog results or..places
are mo~eprotQWjc.al_.thao."tla.miii.i..aaiolli.t. (cf. HopperlThompson
1985: 152-154); the latter thus show a tendency to develop resultant,
static or local meanings:
(42) Lat. prehensio 'act of seizing so.' > Fr. prison 'captivity' >
'prison'
Prototypical transitive verbs place the most active argument in the

subject position (cf. Hopper/Thompson 1980: 252). This offers an
explanation for auto-converse changes, as in ex. (42), where the
CAUSATIVE argument shifts into the subject position, or in ex. (43),
in which an agentive argument structure has developed:
(43) Pg. aborrecer 'to hate sth.lso.' > 'to vex so. (= to cause
hatred)'
(44) Lat.lucere 'to be visible' > Cat. lIucar 'to see'
3.7. Emotionally marked concepts
Some conceptual domains are emotionally marked, as e.g. EATING

and DRINKING, SE'X,-OEATH, FEAR:wANGER, BEAUTY, HOPE, GREAT
QUANTITY/INTENSITY, THE FUTURE, ORIENTATION in TIME, SPACE and
DISCOURSE (cf. for details KochlOesterreicher 1996: 79-87). A
certain number of these domains are additionally marked with taboo.
What's tabooed in a speech community is partially culture-specific,
but certainly contains a supracultural, if not universal core.
When it comes to a tabooed topic, successful communication
does not totally but largely depends on the context of the speech act.
One can address a taboo directly, e.g. by saying ''the princess died in
a car accident", but often this is not the recommended strategy, as
speaking of DEATH is principally a "face-threatening act" (cf. BrownJ

Levinson 1978: 70-72) and one risks jeopardizing communicative
success by offending the interlocutor's feelings. To avoid this "loss
of face" speakers can adopt a euphemistic strateg)' by saying e.g.
"the princess passed away, expired, etc." It is important to see that a
semantic euphemism does not avoid the problematic topic, but al-
lows a different approach to it: the cognitive ratification of the se-
mantic relation between ~A5i~!NJt,~AfirigDjlN,ra:essentJ.aj .fm
the euphemIsm· to- wo~k~· That is to say that a good semantic euphe-
iiiism 1t~ttr11eVemiig and explicit at the same time, if we want it to
be communicatively efficient. If that is the case, it can become
lexicalized.
Euphemisms are often created via metonymy (ex. 45) and meta-
phor (ex. 46), but also through semantic restriction (ex. 47), ellipsis
(cf. section 3.6., no. 1) or ironic antiphrasis (ex. 48 and 49):
(45) Lat. injirmus 'weak' > 'ill (euphemistic)' > OFr. enferm, Sp.
enfermo 'ill', Lat. languidus 'feeble' > Rum. linced 'ill
(euph.)' > 'ill'
(46) Fr. tuer 'to extinguish (fire)' > 'to kill'
(47) It. casino 'little house' > 'brothel'
(48) OF. oste 'guest' > '~e (euph.)'
(49) Sard. masetu 'gentle, good-natured' > 'irascible'
A completely different way of treating tabooed concepts is to ver-

balize them by dysp'~~mism, i.e. by an offensive, ~plicite'!y. ~ce
threatenin~~~~!egy.MhAl~19.2l:.26), e.g. by saying
"the·pn~c.~~.~Jg~*~.9..tlle...hucket". Dysphemisms are esp. vivacious
among· 'members of social groups where expressive and drastic
speech is considered to be "tough" or signifies "coolness". Under
these specific conditions, dysphemisms are used with the same in-
tention as euphemisms: they guarantee and enhance communicative
success. Dysphemisms are usually located on diastratjcaIJy.or dia-
phasically low marked level.§.Q{.SIX=ec b, but can be cited by "normal"
speakers as a special expressive effect.
82 Andreas Blank
Expressive verbalization does also occur with emotionally mark-

ed but not tabooed conceptual domains, like MONEY, WORK, PLANS
AND GOALS, QUANTITY etc. Here, instead of being cautious and
discreet, we must exaggerate and say things in a drastic or
hyperbolic manner, if we want to convince our interlocutor and
make him or her share our point of view (Stempel 1983: 89-90).
Thus, expreSSi~.. llota moli.!!!!i.9.!!J!!r _ t i c cb""!!,,, but.
communicative.. o!:.ategy (cf. section 1.2.). The typical mechanisms
1tte agatfi°-inetapll and metonymy (ex. 50 and 51), but extension of
meaning (ex. 52), co-hyponymous transfer (ex. 53) or contrast-based
semantic change (ex. 54) is also found:
(50) VulgLat. "'tripalium 'torture' > MFr. travail 'work (expr.)' >
ModFr. 'work'
(51) Fr. bordel, It. bordello 'brothel' > 'disorder, brawl'
(52) Lat. caballus 'bad horse, jade' > VulgLat. 'horse (expr.)' >
Cat. cavall, Engd. kaval, Fr. cheval, It. cavallo, Pg. cavallo,
Sard. kat!4u, Rwn. cal, Sp. caballo 'horse'
(53) Sp., Pg. aborrecer 'to vex so.' > 'to bore so. (expr.) > 'to bore
so.'
(54) Engl. bad 'not good' > Engl. (slang) 'good, excellent'
Euphemistic and expressive words are subject to a general ten-

dency: their veiling or drastic-hyperbolic power weakens the more
frequently they are used. As most of the examples above show, the
expressivity or the euphemistic character totally wears away and new
euphemisms or expressive words have to be created.
4. Conclusion
The aim of this paper was to give a typology of the motivations for
semantic .imw..YAtion that integrates what we have learned aoout lan-
gu~giand its ~ction since the times of Ullmann. I am convinced
that;by drawing upon the background of pragmatic and cognitivist
models a deeper insight into the reasons why speakers change their
lexicon has been gained/
The corpus analysis confirmed the general assumptions made in
section 1 and the tripartite typology )If motivations elaborated by
Coseriu. There is, on one side, a pa,rticular, situativ . 'on for
I!P individlUlI to risk a semantic inn tion; on the other side, inno-
vation is gdlerally motivated by the wish to cOI1.lm.unicate as effi-
ciently as p~ible, i~. to influence the interlocutor in the desired
manneralldto do this at the least possible expense. Finally, one
finds that the e!c!l!.a.LmQ.ti.Yil1Wns c~~ gro'!Eed in!9...slx.wes of
general motivatlOns or sufficient conditions for semantic innovation,
as presenko'iii-'section 3. These conditions derive either from our
perception of the world and our way of structuring our concepts or
from the structure and form of one language's lexicon. The severat
subtypes are to be understood as more specific frameworks that trig-
ger specific mechanisms of semantic change.
Notes
1. Instead of "causes". I prefer the tenn "motivation". because it emphasizes

that semantic innovations (like any linguistic innovation) are speaker-based;
they occur because speakers have a motive that makes them innovate.
2. I emphasize that the metaphor itself is not. on any level. the cause of the
change. but only the mechanism or "procedure" the speaker uses (cf. Blank
1997. ch. 5). As Geeraerts puts it. metaphors. metonymies etc. "indicate the
associative mechanisms that defme and delimit the set of possible (or
plausible) semantic changes" (1997: 103).
3. Coseriu uses the following comparison to underline this necessity: "Dar
s610 una explicaci6n generica de un cambio hist6ricamente detenninado es
como decir que una casa se ha incendiado 'porque el fuego quema la made-
ra·. lo)ldiil es cierto desde el punto de vista generico .... pero no nos dice
nada acerca de la causa historica (particular) del incendio." (1958. 104-
106) [To give a general explanation ofa historically occured change is like
saying that a house caught fIre 'because wood burns'. which is correct from
a general point of view. but doesn't tell us anything about the historical
(particular) reason for the fire].
84 Andreas BlanJc
4. The existence of an individual motivation and a sufficient condition does

not necessarily lead to a semantic innovation. Speakers can also take an-
other pathway of verbalization, as e.g. word formation or borrowing.
5. It is self-evident that the interlocutor doesn't need to be a concrete person,
but can. also be a fictive hearer/reader imagined by the author.
6. This central hypothesis of SperberlWilson (1986) is not unanimously ac-
cepted; cf. e.g. Horn (1989, 194-195); Keller (1995, 211-213).
7. The efficiency of language change has not necessarily to be of panchronic
validity; it suffices that a group of speakers is convinced that the innovation
serves their communicative goals to make it appear efficient. A lexicalized
innovation can turn out later to be inefficient, as e.g. in the case ofhomony-
mic clash as a consequence of - probably efficient - phonetic change or in
the case of "polysemic clash" as a consequence of-certainly efficient - se-
mantic change. This happened e.g. to the polysemous OSp. pregon 'mes-
sage', 'messenger': the latter sense was later expressed by the derivation
pregonero. Cf. Koch (in press a: section 6.3).
8. Maybe the distinction between expressivity and efficiency is also misled
because when linguists say "language change" they often mean "sound
change" or, at the most, ''morphQlogical change". It is obvious that by in-
vestigating the diachrony of the lexicon one gets different results.
9. For a detailed discussion cf. Blank (1997: 347-354).
10. The difference is that in the case of torpedo we have an entirely new object
and, consequently, a new word, while in the case of pen there already
existed a word to designate a former stage of development. Obviously, the
word is maintained by what could be called "linguistic conservatism" (Ull-
mann 1957: 211-212), but, to begin with, writing instruments with a metal
nib must have been classified as being similar to traditional pens. To that
extent, the metal-nibbed pen was once a new object in search of a name.
11. Cr. Nyrop (1913,80): "Quand on ditj'attends le courrier, on pense moins
a l'homme qui apporte les lettres, qu'a ces lettres elles-memes, et courrier
prend ainsi le sens de 'lettres': tout le courrier est pour vous." [When one
says: I'm waitingfor the 'courrier' (lit. courier), rather than thinking of the
man who brings the letters, one thinks of these letters themselves, and thus
courrier acquires the meaning 'letters': all the 'courrier' isfor you.]
12. Nyrop's "connexion entre les choses" is the type of motivation that often
stands behind the tendencies of semantic change detected by Elizabeth
Traugott (cf. e.g. 1989, 1990, this volume), which usually run under the
label of "subjectification".
13. This critique includes the theories of semanticists not mentioned in section
2, as e.g. de la Grasserie (1908), Stem (1931), Gamillscheg (1951), Kron-
asser (1952), Duchaeek (1967) and even the otherwise very rich and exem-
plary study by Nyrop (1913).
14. The reader may notice that this objective is in slight contradiction to the
heading of the present section: if we keep the title "motivations for seman-
tic change" this is mainly due to tradition. But as we have already noticed
in section 1.1., linguists usually only touch upon successful innovations, i.e.
those that have led to semantic change.
15. One example: Lat. collocare 'to dispose' was restricted to Fr. coucher, It.
coricare 'to lay down' and to Sp., Pg. colgar 'to hang'. Both ways of re-
striction seem to be likewise fortuitous.
16. Cf. Rohlfs (1971: 138); Tagliavini (1973: 123-124). A different interpreta-
tion is given by Klein (1997: 239-240).
17. For other diachronic aspects of prototypes cf. Koch (1995, 39-41).
References
Allan, Keith and Kate Burridge

1991 Euphemism and Dysphemism. Language Used as a Shield and a
Weapon. New York etc.: Oxford University Press.
Bally, Charles
4 1965 Linguistique generale et linguistique fran~aise. Bern: Francke.
Blank, Andreas
1993 Das verwaiste Wort. Zum Bedeutungswandel dUTch Volksetymo-
logie. In: Foltys, Christian and Thomas Kotschi (eds.), Berliner
Romanistische Arbeiten. Fur Horst Ochse, 43-61. Berlin: Freie Un i-
versitllt.
1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der
romanischen Sprachen. TObingen: Niemeyer.
1998a Der Kopf in der Romania und anderswo - Ein metaphorisches (und
metonymisches) Expansions- und Attraktionszentrum. In: GiI, AI-
berto and Christian Schmitt (eds.), Kognitive und kommuni/cative
Dimensionen der Metaphorik in den romanischen Sprachen, 11-32.
BOM: Romanistischer Verlag.
1998b Les adjectifs temporels du type long/court dans les langues romanes:
un cas de 'metaphoricite etroite'. In: Dupuy-Engelbardt, Hiltraud
and Marie-JeaMe Montibus (eels.), L 'organisation lexicale et
cognitive des dimensions spatiale et temporelle. Actes d'EUROSEM
96, 15-37. Reims: Presses Universitaires.
86 Andreas Blank
1998c Topo et al. - Onomasiologie, Semasiologie und Kognition am Bei-

spiel der Bezeichnungen von MAUS, RAITE und MAULWURF in der
Italoromania. Zeitschriftfor romanische Philologie 114: 505-531.
in press Pathways of lexicalization. In: Haspelmath, Martin, Ekkehard
KOnig, Wulf Oesterreicher and Wolfgang Raible (eds.), Language
Typology and Language Universals. An International Handbook.
Berlin / New York: De Gruyter.
Brown, Penelope and Stephen Levinson
1978 Universals in language use: politeness phenomena. In: Goody,
Esther N. (ed.), Questions and Politeness. Strategies in Social
Interaction, 56-289, 295-310. Cambridge: Cambridge University
Press.
Coseriu, Eugenio
1958 Sincronia, diacronia e historia. El problem a del cambio lingiiistico.
Montevideo: Universidad de la Republica.
1978 FUr eine strukturelle diachrone Semantik. In: Geckeler, Horst (ed.),
Strukturelle Bedeutungslehre, 90-163. Darmstadt: Wissenschaftliche
Buchgesellschaft.
Croft, William
mn..F
1992 Dictionnaire historique de la langue franfaise. Paris: Robert.
Ducha~ek, Otto
1967 Precis de semantique fram;aise. Bmo: University J.E. Purkyn~.
Fiehler, ReUlhard
1990 Kommunikation und Emotion. Theoretische und empirische Unter-
suchungen zur Rolle von Emotionen in der verbalen Interaktion.
Berlin / New York: De Gruyter.
Gamillscheg, Emst
1951 Franzosische Bedeutungslehre. TUbingen: Niemeyer.
Geeraerts, Dirk
1983 Reclassifying semantic change. Quaderni di semantica 4: 217-240.
1997 Diachronic Prototype Semantics. A Contribution to Historical Lexi-
colog)!. Oxford: Clarendon.
Gilli~ron, Jules and Mario Roques
1912 Etudes de geographie linguistique d'apres I 'Atlas Linguistique de la
Franc~. Paris: Champion.
Grasserie, Raoul de la
1908 Essai d'une semantique integrale. 2 volumes. Paris: Leroux.
Hopper, Paul J. and Sandra A. Thompson

1980 Transitivity in grammar and discourse. Language 56: 251-299.
1985 The iconicity of the universal categories 'noun' and 'verbs'. In: Hai-
man, John (ed.), Iconicity in Syntax, 151-183. Amsterdam: Benja-
mins.
Horn, Laurence R.
1989 A Natural History of Negation. Chicago: University of Chicago
Press.
Keller, Rudi
21994 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tubin-
gen: Francke.
1995 Zeichentheorie. Zu einer Theorie semiotischen Wissens. TUbingen:
Francke.
Klein, Franz-Josef
1997 Bedeutungswandel und SprachendijJerenzierung. Die Entstehung
der romanischen Sprachen aus wortsemantischer Sicht. TUbingen:
Niemeyer.
Koch, Peter
1991 Semantische Valenz, Polysemie und Bedeutungswandel bei romani-
schen Verben. In: Koch, Peter and Thomas Krefeld (eels.), Conne-
xiones Romanicae. Dependenz und Valenz in romanischen
Sprachen, 279-306. TUbingen: Niemeyer.
1995 Der Beitrag der Prototypentheorie zur Historische Semantik: Eine
kritische Bestanelsaufnahme. Romanistisches Jahrbuch 46: 27-46.
in press a Ein Slick auf die unsichtbare Hand: Kognitive Universalien und
historische romanische Lexikologie. In: Stehl, Thomas (ed.), Un-
sichtbare Hand und Sprecherwahl. Typologie und Prozesse des
Sprachwandels in der Romania. TUbingen: Narr.
in press b Frame and contiguity. On the cognitive basis of metonymy and
certain types of word formation. In: Radden, GUnter and Klaus-Uwe
Panther (eds.), The Metonymy in Thought and Language. Amster-
dam: Benjamins.
Koch, Peter and Wulf Oesterreicher
1996 Sprachwande1 und expressive MUndlichkeit. Zeitschrift fur Litera-
turwissenschaft und Linguistik 26: 64-96.
Kronasser, Heinz
1952 Handbuch der Semasiologie. Heidelberg: Winter.
Labov, William
1972 Sociolinguistic Patterns. Philadelphia: University of Pennsylvania
Press.
88 Andreas Blank
Mair, WaIter N.
1992 Expressivitlit und Sprachwandel. Studien zur Rolle der Subjektivitlit
in der Entwicklung der romanischen Sprachen. Frankfurt: Lang.
Meillet, Antoine
1965 Linguistique historique et Iinguistique generale. Paris: Champion.
Nyrop, Kristoffer
1913 Grammaire historique de la langue fran~aise, Volume 4. Copen-
hague: Gyldendal.
Rohlfs, Gerhard
1971 Romanische Sprachgeographie. MUnchen: Beck.
Sperber, Dan and Deirdre Wilson
1986 Relevance. Communication and Cognition. Oxford: Blackwell.
Sperber, Hans
21965 Einftihrung in die Bedeutungslehre. Bono: Schroeder.
Stempel, Wolf-Dieter
1983 'lch vergesse alles'. Bemerkungen zur Hyperbolik in der Alltags-
sprache. In: M. Faust (ed.), Allgemeine Sprachwissenschaft. Sprach-
typologie und Textlinguistik. Festschrift fur Peter Hartmann, 87-98.
TObingen: Niemeyer.
Stem, Gustaf
1931 Meaning and Change of Meaning. Bloomington, Ind.: Indiana Uni-
versity Press.
Sweets er, Eve E.
1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects
of Semantic Structure. Cambridge: Cambridge University Press.
Tagliavini, Carlo
1973 Einftihrung in die romanische Philologie. MUnchen: Beck.
Traugott, Elizabeth C.
1989 On the rise of epistemic meaning in English: an example of subjecti-
fication. Language 65: 31-55.
1990 From less to more situated in language: the unidirectionality of se-
mantic change. In: Adamson, Silvia, Vivien Law,Nigel Vincent and
Susan Wright (eds.), Papers from the Fifth International Conference
on English Historical Linguistics, 496-517. Amsterdam: Benjamins.
Ullmann, Stephen
21957 Principles ofSemantics. Oxford: Blackwell.
1962 Semantics. An Introduction to the Science of Meaning. Oxford:
Blackwell.
Wartburg, Walther v.
J 1971 Einjahrung in die Problematik und Methodik der Sprachwissen-
schajt. TObingen: Niemeyer.
Zipf, George K.
1945 The repetition of words, time-perspective, and semantic balance.
The Journal a/General Psychology 32: 127-148.
1949 Human Behaviour and the Principle 0/ Least Effort. Cambridge,
Ma.: Addison-Wesley.
Diachronic prototype semantics. A digest
Dirk Geeraerts
The purpose of this contribution is to present the major points of the

argumentation contained in my book Diachronic Prototype Seman-
tics (1997). The book is a systematic exploration of the way in which
the notion of prototypicality can be made to bear fruit in historical
semantic research; it draws on the various articles on the subject that
I have written over the last fifteen years or so, and sketches a con-
ceptual framework in which the different case studies that I have
dealt with in the individual papers, fit together into an encompassing
view of semasiological change, considered from the point of view of
prototype theory. The present resume (which could be seen as an up-
dated version of the overview that was given in Geeraerts 1992) will
not be concerned with examples and case studies, but will boil down
the framework presented in the monograph to its essentials. For
those who are familiar with the book, the present paper can function
as a summary; for other readers, it might possibly serve as an appeti-
zer.
1. Prototypicality from a descriptive point of view
The crucial distinction to be kept in mind with regard to the use of

prototypicality in diachronic semantics is the distinction between a
descriptive and an explanatory approach. Prototypicality is basically
a descriptive concept, but it fits perfectly well into an explanatory
model that links up with the "Natural" models of the lexicon repre-
sented by Natural Phonology and Natural Morphology. While the
present paragraph will be concerned with the use of prototypicality
as a descriptive concept, the following one will deal with the expla-
natory aspects.
92 Dirk Geeraerts
If the prototypical view is accepted as an adequate model for the

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

ture, prototype theory highlights the fact that changes in the referen-
tial range of one specific word meaning may take the form of modu-
lations on the core cases within that referential range. Changes in the
extension of a single sense of a lexical item are likely to take the
form of an expansion of the prototypical centre of that extension. If
the referents that may be found in the range of application of a parti-
cular lexical meaning do not have equal status, the more salient
members will probably be more stable (diachronically speaking)
than the less salient ones. Changes will then take the form of modu-
lations on the central cases: if a particular meaning starts off as a
name for referents exhibiting the features ABCDE, the subsequent ex-
pansion of the category will consist of variations on that type of re-
ferent. The further the expansion extends, the less features the peri-
pheral cases will have in common with the prototypical centre. A
fIrst layer of extensions, for instance, might consist of referents exhi-
biting features ABCD, BCDE, or ACDE. A further growth of the peri-
pheral area could then involve feature sets ABC, BCD, CDE, or ACD (to
name just a few).
In the monograph 1997, this hypothesis is supported by a case
study involving the close inspection of the development of a recent
Dutch neologism, viz. the clothing term legging. (For a separate pub-
lication of the case study, see Geeraerts, in press.)
1.2. The development ofradial sets
By stressing the intensional non-equality of lexical-semantic struc-

ture, prototype theory highlights the clustered set structure of
changes of word meaning. This hypothesis shifts the attention from
the extensional structure of an individual meaning of a lexical cate-
gory, to the intensional structure of the lexical item as a whole, that
is, to the overall confIguration of the various readings of the word.
The hypothesis suggests that the structure of semasiological change
mirrors the synchronic semantic structure of lexical categories, given
that the latter involves family resemblances, radial sets, and the dis-
tinction between central and peripheral readings. Semasiological
change, then, involves the change of prototypically clustered con-
cepts. This general statement can be broken down into two more
94 Dirk Geeraerts
specific ones. First, the structure of semasiological change as a

whole is one of overlapping and interlocking readings; specifically, a
novel use may have its starting-point in several existing meanings at
the same time. Second, there are differences in structural weight
among the readings of an item; specifically, there are peripheral
meanings that do not survive for very long next to more important
meanings that subsist through time.
In the monograph 1997, this hypothesis is corroborated by means
of an extended case study involving the Dutch item vergrijpen (ori-
ginally published as Geeraerts 1983a).
1.3. Semantic polygenesis
By stressing the extensional non-discreteness of lexical-semantic

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

spring into existence at any moment in the history of a word, they
are at the same time an illustration of the synchronically flexible
character of word meanings: exactly what belongs to a category at
one particular moment is not necessarily clear.
1.4. Semantic change from subsets
By stressing the intensional non-discreteness of lexical-semantic

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

even if ABC does not constitute a separate meaning of the category
ABCDE in any psychologically or intuitively important interpretation,
it is a subset that has to be taken into account in the (disjunctive,
clustered) definition of the category. Hence, it is impossible to main-
tain the view that only the senses of a category (and not those exten-
sional subsets that do not constitute separate meanings, or those de-
scriptive features that are supposed to have "encyclopedic" rather
than "semantic" status) are structurally important.
From a diachronic point of view, this means that semantic
changes may take their starting-point on the extensional level just as
well as on the intensional level, or in the domain of encyclopedic in-
formation just as well as in the realm of semantic information. Even
where a classical definition is possible, extensional subsets or inten-
sional features with an "encyclopedic" rather than a "semantic"
status may play a crucial role in processes of semantic change. The
examples discussed in the monograph (previously published in Geer-
aerts 1994) show precisely this: the semantic extensions through
which new meanings arise may take their starting-point in exten-
sional subsets that do not correspond with senses in the structuralist
sense, even in those cases where the categories in question might be
classically defined.
To round off the overview, it should be stressed that the aspects

of semantic change enumerated here are not necessarily new to dia-
chronic semantics. What is indubitably new, however, is the fact that
these more or less known aspects of change can now be incorporated
into a global model of lexical-semantic structure. That is to say, from
a descriptive point of view the importance of prototype theory pro-
bably resides less in the novelty of its observations, taken separately,
than in the fact that it brings them together in an overall model of the
structure of lexical meaning.
2. Prototypicality from an explanatory point of view
In the cluster of approaches exploring the concept of naturalness in

linguistic theory, some areas of linguistics are better represented than
others. Best known are probably Natural Phonology (Stampe 1979,
1987; Donegan and Stampe 1979) and Natural Morphology (Mayer-
thaler 1981; Wurzel 1983; Dressier 1985), while Natural Syntax
(Haiman 1980, 1985) and Natural Text Linguistics (DressIer 1989)
are somewhat more recent additions. But what about Natural Lexico-
logy? Elaborating the line of thought started in Geeraerts (1983 b), I
will argue that prototype theory is part and parcel of a "Natural" the-
ory of the lexicon.
2.1. The functional motivation for prototypicality
Given a prototypical conception of semasiological structure, the first

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

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

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

semantic change, because it relates them to general epistemological
beliefs about the working of the human conceptual system, beliefs it
shares with other cognitive theories. And at the same time, of course,
the overall conception of a prototypical organization of conceptual
categories can itself be explained on functional grounds.
2.2. Prototypicality in an explanatory framework
As a next major step in the argumentation, we should recognize that

the type of efficiency achieved by the prototype-based conception of
lexical-semantic structure is part of broader range of efficiency phe-
nomena in lexicology. Without trying to be exhaustive, two additio-
nal examples of efficiency may be mentioned.
First, let us consider homonymic clashes. Gillieron's famous ex-
ample involves the collision of Latin caltus ('cat') and gallus
('cock') into Gascon gat (Gillieron and Roques 1912). The tension is
resolved by replacing gat ('cock') by bigey, a local equivalent of vi-
caire ('curate'), or by azan, the local equivalent of faisan (,pheas-
ant'), or by the cognates of Latin pullus. The moral of the story is
usually taken to be that homonymic ambiguities set off therapeutic
diachronic changes towards their resolution. The rationale behind the
avoidance of homonymy might be called a principle of formal effi-
ciency, more particularly a "one form, one meaning" principle: for-
mally disambiguated languages are functionally superior, because
they avoid communicative misunderstandings.
Second, popular etymology instantiates a tendency (at least in
some cases) towards formal, morphological transparency. In Dutch,
for instance, the loan-word hamac 'hammock' is changed into hang-
mat 'hanging carpet'. The semantic transparency of the latter expres-
sion (which is composed of the verbal stem hang- 'to hang' and the
noun mat 'carpet, mat') is communicatively efficient; those who are
not familiar with the foreign word may grasp (or at least get an idea
of) what is referred to.
As a further step, let us now compare the approach developed so

far with the functional principles at work in the theory of Natural
Phonology. Natural Phonology assumes that research into phonolo-
gical phenomena has to take into account whatever is known about
the physiological structure of our auditory and articulatory organs:
the physiological possibilities of and restrictions on those organs de-
termine what is "natural" in phonetics and phonology. The fact, for
instance, that sounds may be subject to a process of assimilatory
voicing in a voiced environment is a natural phonological process,
because our articulatory organs would generally encounter difficul-
ties producing a sequence of a voiced, an unvoiced, and then again a
voiced segment. Natural phonological processes do not have to be
learned; they follow automatically from the physiological difficulties
that occur in the production and perception of speech. Conversely,
learning a language implies learning which natural processes have to
be suppressed according to the phonological system of a particular
language. Word-final devoicing, for instance, is a natural process,
but children learning French or English have to learn not to give in
to the natural tendency.
Natural Phonology distinguishes between two major types ofpho-
no logical processes: lenition and fortition. (There is a third type re-
lating to suprasegmental, prosodic phenomena, but the present dis-
cussion will be restricted to processes involving single segments.)
Fortition occurs when a sound segment is pronounced in a more out-
spoken manner, in a way, that is, in which it can be more clearly
distinguished from its surroundings. Dissimilation, diphthongization,
and epenthesis are frequently occurring examples of fortition pro-
cesses. Lenition, on the other hand, occurs when the contrast be-
tween a segment and its surroundings is weakened, as in the case of
assimilation, monophthongization, shortening, and deletion. Forti-
tion and lenition processes have a tendency to occur preferentially in
specific environments. Fortition is typical for "strong" positions,
such as vowels in stressed syllables, or word-initial consonants. Le-
nition favours segments in "weak" positions, such as word-final
segments or in unstressed syllables. In addition, there is a stylistic
difference between both types: fortition is more likely to occur in
102 Dirk Geeraerts
slow, careful, and formal speech, whereas lenition occurs more

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

PHONOLOGICAL FORM lenition processes
CONCERNING THE prototype-based • isomorphism
LEXICAL RELAnON flexibility and stability • iconicity
BETWEEN FORM AND
MEANING
• transparency and
motivation
Figure 1. A classification of efficiency principles
We can see, in other words, that the prototype-theoretical ap-

proach to lexical-semantic structure could easily fonn the basis for a
"Natural" theory of the lexicon, provided that the existence of other
types of lexicological efficiency is taken into account. It should be
noted that the distinction between the basic forms of the efficiency
principle (speaker-based optimization of production versus hearer-
based optimization of perception) features in various forms in the
recent literature on linguistic change. In Langacker (1977) and Kem-
mer (1992), for instance, it appears in the form of a distinction be-
tween a Principle of Least Effort, and a Principle of Maximal Dis-
tinctiveness. Lewandowska-Tomasczcyk's classification of causes of
change (1985) makes clear that a term such as "Principle of Least
Effort" - although traditionally receiving a speaker-oriented interpre-
tation - may also be interpreted from the perspective of the hearer:
maximal coding distinctiveness on behalf of the speaker favours mi-
nimal decoding effort on the part of the hearer. In addition, note that
the distinction is far from new: one of the oldest fonnulations is von
der Gabelentz's distinction between Bequemlichkeitstrieb and Deut-
lichkeitstrieb (1891).
The comparison with Natural Phonology suggests a specific pro-
blem for a functional explanatory theory of lexical change, and this
recognition should help to avoid any exaggerated optimism. In
general, it has to be recognized that the functional approach sug-
104 Dirk Geeraerts
gested here is far from answering all questions that arise in the con-
text of historical semantics. To illustrate, let us go back to our dis-
cussion of popular etymology. It was suggested that a form such as
Dutch hangmat (from hamac) may be explained by an economical
tendency to have morphologically transparent word forms. Some
cases of popular etymology are less clear, however. For instance,
while Dutch cichorei 'chicory' is sometimes transformed into sui-
kerij 'sugary', semantic transparency is far from achieved: chicory
and sugar have nothing in common, chicory does not even taste
sweet. Perhaps we might say that the functional principle at work
here is a tendency to exploit the morphological possibilities of the
lexicon (that is, to maximize the number of morphologically com-
plex words at the expense of newly introduced base forms). This ten-
dency in itself would then be an illustration of a more fundamental
tendency towards an economical lexical organization (keeping the
number of lexical base forms down is efficient because it diminishes
the memory load of the system).
However, even apart from the fact that the semantic opacity of
suikerij increases rather than diminishes the strain on lexical mem-
ory (the language user has to remember that suikerij has nothing to
do with sugar), the operation of the economic principle with regard
to the number of lexical forms is unsuccessful, since cichorei actual-
ly continues to exist next to suikerij: the transparency principle crea-
tes a situation that is in conflict with the isomorphic principle. In
short, the operation of the functional principles does not guarantee
success: some changes seem to miss their probable goal, or at least
yield results that are incompatible with other instantiations of the ef-
ficiency factor. Specifically, given that Natural Phonology accepts
that hearer-oriented and speaker-oriented processes may be in con-
flict, the question arises how tensions between the hearer-oriented
principle of isomorphism, and the speaker-oriented principle of pro-
totypicality are resolved.
3. Conclusion
To round off the discussion, it may be useful to specify how the pre-
sent analysis of diachronic prototypicality from an explanatory point
of view, relates to my earlier attempts to define a classification of
functional causes of semantic change, as in Geeraerts (1983b). The
present classification is a direct continuation of the older one, to the
extent that it makes crucial use of the distinction between proto-
typicality on the one hand, and factors like isomorphism and iconi-
city on the other. However, I earlier connected this distinction with
the difference between expressivity and efficiency, respectively, as
fundamental factors shaping lexical changes. In the present frame-
work, the two types of factors (prototypicality on the one hand, iso-
morphism etc. on the other) are rather seen as different forms of the
efficiency principle - a speaker-oriented one, and a hearer-oriented
one, respectively.
This does not mean that expressivity no longer has a role to play
in the classification of causes of lexical change. Expressivity (in-
tended here in the sense of 'the need to express something verbally')
remains the basic motivating force behind any form of lexical
change, either because a new concept has to be put into words, or be-
cause an existing concept receives a different form of linguistic ex-
pression. The different (and competing) types of efficiency, on the
other hand, suggest different formal ways in which this communica-
tive, expressive intention may be realized. In this sense, the present
framework makes explicit the asymmetrical relationship between ex-
pressivity and efficiency that was inherent in my original proposals
(see Blank 1997: 362): expressivity is always a primary cause of
change, whereas efficiency involves the choice of the linguistic
means realizing the expressive intention.
At the same time, by highlighting the distinction between speak-
er-related and hearer-related types of efficiency, the explanatory ap-
proach described here as part of a summary of my book Diachronic
Prototype Semantics (1997), opens the way towards the development
of a theory of "Natural Lexicology".
106 Dirk Geeraerts
References
Blank, Andreas
1997 Prinzipien des lexikalischen Bedeutungswandel am Beispiel der ro-
manischen Sprachen. Ttlbingen: Max Niemeyer Verlag.
Donegan, Patrica and David Stampe
1979 The study of natural phonology. In: Daniel A. Dinnsen (ed.), Cur-
rent Approaches to Phonological Theory, 127-173. Bloomington:
Indiana University Press.
DressIer, Wolfgang U.
1985 Morphonology. Ann Arbor: Karoma Press.
1989 Semiotische Parameter einer textlinguistischen Natiirlichkeitstheo-
rie. Vienna: Verlag der Osterreichischen Akademie der Wissen-
schaften.
Gabelentz, Georg von der
1891 Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherigen
Ergebnisse. Leipzig: Tauchnitz.
Geeraerts, Dirk
1983a Prototype theory and diachronic semantics. A case study. Indoger-
manische Forschungen 88: 1-32.
1983b Reclassifying semantic change. Quaderni di Semantica 4: 217-240.
1985a Paradigm and Paradox. Explorations into a Paradigmatic Theory
of Meaning and its Epistemological Background. Leuven: Univer-
sitaire Pers.
1985b Semantische polygenese: een bijzondere vorm van historische bete-
kenisvariatie. Forum der Letteren 26: 120-130.
1989 Prospects and problems of prototype theory. Linguistics 27: 587-
612.
1993 Cognitive semantics and the history of philosophical epistemology.
In: Richard A. Geiger and Brygida Rudzka-Ostyn (eds.J, Concep-
tualizations and Mental Processing in Language, 53-80. (Cognitive
Linguistics Research 3.) Berlin / New York: Mouton de Gruyter.
1992 Prototypical effects in diachronic semantics: A round-up. In: Gtlnter
Kellermann and Michael D. Morrissey (eds.), Diachrony within Syn-
chrony: Language History and Cognition, 183-203. (Duisburger Ar-
beiten zur Sprach- und Kulturwissenschaft 14.) Frankfurt etc.: Laog.
1994 Subsets in semantic change: generalizing inductive generalization.
In: Keith Carlon and Emma Vorlat (eds.), Perspectives on English,
128-139. (Orbis: Supplementa 2.) Leuven: Peeters.
in press Neologism at short range. In: Rainer Schulze (ed.), Making Mean-
ingful Choices. TUbingen: Gunter Narr Verlag.
Geeraerts, Dirk, Stefan Grondelaers and Peter Bakema

1994 The Structure 0/Lexical Variation. Meaning. Naming. and Context.
Gruyter.
Gillieron, Jules and Mario Roques
1912 Etudes de geographie linguistique. Paris: Champion.
Haiman, John
1980 The iconicity of grammar: isomorphism and motivation. Language
56: 515-540.
1985 Natural Syntax. (Cambridge Studies in Linguistics 44.) Cambridge:
Cambridge University Press.
Kemrner, Suzanne
1992 Grammatical prototypes and competing motivations in a theory of
linguistic change. In: Garry W. Davis and Grant L. Iverson (eds.),
Explanation in Historical Linguistics, 145-166. (Current Issues in
Linguistic Theory 84.) Amsterdam: John Benjamins Publishing
Company.
1977 Syntactic reanalysis. In: Charles N. Li (ed.), Mechanisms o/Syntac-
tic Change, 57-139. Austin: University of Texas Press.
Lewandowska-Tomaszczyk, Barbara
1985 On semantic change in a dynamic model of language. In: Jacek
Fisiak (ed.), Historical Semantics - Historical Word Formation,
297-323. (Trends in Linguistics. Studies and Monographs 29.)
Mayerthaler, Willy
1981 Morphologische Natiirlichkeit. (Linguistische Forschungen 28.)
Wiesbaden: Athenaion.
Rosch, Eleanor
1977 Human categorization. In: Neil Warren (ed.), Studies in Cross-Cul-
tural Psychology 1,3-49. London: Academic Press.
Stampe, David
1979 A Dissertation on Natural Phonology. New York: Garland.
1987 On phonological representations. In: Wolfgang U. DressIer, Pho-
nologica 1984, 287-299. Cambridge: Cambridge University Press.
Wurzel, Wolfgang U.
1983 Thesen zur morphologischen NatUrlichkeit. ZeitschriJt for German-
istik 4: 196-208.
Cognitive semantics and diachronic semantics:
the values and evolution of classes
Fran~ois Rastier
For the centennial of the publication of

Breal's Essai de semantique.
The theory of prototypes is the best known contribution of cognitive

semantics to lexical semantics. Yet, due to the very fact of its univer-
salism, it does not succeed in defining the specific lexical classes of
different languages, nor in accounting for diachronic evolution.
We shall nevertheless take up the problem of qualitative inequali-
ties among lexical classes. We shall thus distinguish two types of
prototypes: (i) the terms for generic usage, which are neutral with re-
spect to associated values, and (ii) the paragons, terms which usually
carry a positive value. We shall then deal with the question of quali-
tative inequalities in diachrony by showing how the social evolution
of norms of valuation leads to rehandling the structure of elementary
lexical classes (taxemes); we shall finally give a detailed account of
an example dealing with the history of denominations for face in
French since the 15th century.
Besides from the fact that structural semantics and cognitive se-
mantics are not unified movements, they remain difficult to compare
because their objects and objectives - none of which have been
reached yet - are somehow different. In the field of diachrony, the
contributions of structural semantics have been, to tell the truth, no-
torious for a long time (see Coseriu 1964) and if the cognitive issue
managed to stimulate it (see the works of Koch 1995 and Blank
1997), it remains that diachronic cognitive semantics (Sweetser
1990, for instance) has not or not yet shown any visible theoretic or
practical progress: most of the time, it does not do much more than
retrace well known problems (see Nyckees 1997).
11 0 Fran~ois Rastier
1. The pan chronic problem of qualitative inequalities
At the beginning of the thirties, Hjelmslev set the problem of

qualitative inequalities among linguistic categories by distinguishing
intense terms and extense terms. I The main merit of the cognitive
theory of prototypes will have been to restate this problem among
lexical classes. Yet, the concept of prototype does not by itself
account for these inequalities. Indeed, it introduces two types of
fuzziness: one through its own definition which varies from one
author to the other, and the other from its use to describe matters of
graduality. Moreover, it must be founded. 2 Why does a given term
become prototypical or ceases to be such? This question is never
asked and nothing in the theory of prototypes gives us a way of
knowing how prototypes are born, grow and disappear.
Hjelmslev's problem of qualitative ineqUality of lexical units in-
side the taxeme must be considered in synchrony as well as in dia-
chrony in a panchronic perspective taking into account the structure
of lexical classes. Whereas in a synchronic description, discrete re-
presentations are preferred, continuous representations with thresh-
olds are necessary in diachrony. The panchronic perspective thus
requires to articulate two kinds of representations: discrete and con-
tinuous. If in synchrony the relations inside lexical classes can be
characterised by discrete semic oppositions, the gradual nature of
diachronic evolution can be represented by dynamic models which,
without contradicting the semic analysis, identify sememes as char-
acteristic zones within dynamic evolutions (see infra, 2.2.).
1.1. Some weak points of cognitive semasi%gy
1.1.1. The problem of semantic classes
The main shortcoming of cognitive semantics is the weakness of its

theory of semantic or conceptual classes. This is firstly due to the
vagueness of Rosch's concept of category, to the theorization of this
vagueness (see the cue validity) and to the absence of a defined point
Cognitive semantics and diachronic semantics 111
of view on its constitution: from which point of view, according to

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

crucial point of methodology. Whereas for Rosch and the psycho-
logists who came after her categories are classes of concepts (and
their putative corresponding objects), the study of which should be
submitted to an onomasiological method,s for cognitive semantics
they are classes of acceptations which are studied through a semasio-
112 Franfois Rastier
logical method. Taking a signifier as invariant, it tries to cross-define

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

tributions, Breal had introduced a diachronic structural principle, 8
which Saussure transposes in synchrony in the theory of value. This
differential principle will precisely enable the constitution of semic
analysis and its use to account for the different phases of evolution.
The semic analysis is then considered in context, since the new
meanings are described from the new contexts (a semic analysis
which would reduce itself to analysing words out of context would
be illegitimate because normative). Yet, if the polysemy of accepta-
tions, a contingent fact in synchrony, can be enlightened by diachro-
nic studies which show how one managed to pass from one accepta-
tion to another, this does not entail that it should be the preferential
object of diachronic lexicology.
1.1.3. The factors of diachronic changes
Factors are not causes, and the description generally limits itself, and
wisely so, to the how of evolution; but we shall see further that one
cannot avoid the question of why.
a) The four operations. - Since the beginnings of lexical semantics,

four operations have persistently qualified changes in meaning:
extension and restriction, metaphor and metonymy. These very oper-
ations which are used to describe the diachronic relations (for in-
stance in Darmesteter 1887) are also used to describe the synchronic
relations between acceptations from Reisig (1839) to Cledat (1845)
up to Robert Martin (1992). They are furthermore linked to a sema-
siological perspective, because they link two states (synchronic and
diachronic) of a signified and they take the signifier as invariant.
They can be grouped by two: extension and restriction are logical
concepts which belong to the classical theory of ideas,9 metaphor
and metonymy belong to the restricted rhetoric, theorized by the
grammarians of the 18th century (Dumarsais 1730 in particular),
precisely to save the literal sense and with it the own identity of
concepts.
114 Fran~o;s Rastier
(i) Logico-reforential operations. - Extension and restriction

belong to referential semantics 10 which is applied to objects accord-
ing to an unverifiable principle of quantitative variation very well-
known under the name of law of Port-Royal. These operations rest
upon (i) a discrete ontology, in such a way that the designated
objects can be counted, and (ii) a logic of classes, in such a way that
relations of inclusion can be described.
(ii) Tropes. - It is known that the use of tropes in diachronic
semantics, even when it claims to have its roots in Lakoff (Sweetser
1990), is the most traditional thing one can find: it pre-exists even to
Breal's semantics, since it takes an important place in Darmesteter
(1877), and already earlier in Reisig (1839). The description through
tropes supposes a deviation, or at least a variation between the literal
sense and the figurative sense. Yet, except if we repeat Saint Paul,
one cannot pretend that the ancient meaning is literal and the new
one figurative, nor a fortiori that the evolution of languages departs
from nature, as Dumarsais regretted in a famous page about cata-
chresis. 11
The use of rhetorical concepts in diachronic semantics is not self-
evident because tropes are used there to name relations that must
still be described and explained. Moreover, if one accepts that the
trope is a textual form, the relations which are in this way decon-
textualized are not tropes, strictly speaking. We consider metaphor
and metonymy as critical points of semantic forms, but it is not
enough to name them to consider them described. 12 What is the
difference between the units they unite? In other words, what is the
metaphorical or metonymical orientation? What motivates these
paths? To answer these questions, we shall study hereunder the rela-
tive evaluations of the terms matched, especially their respective po-
sition in relation to the evaluative thresholds inside the taxeme.
b) Two evolutions. - Studying the relative valuation of the units

matched seems to be for us a means to set the problem of evolution
as a whole without resorting to logics or rhetorics. In a study on
paragons (1991: 198-202), we have extracted a law of panchronic
valuation which accounts for two complementary evolutions:
(i) The evolution through extension goes from the higher valued
term to the lower ones: pecunia (Latin for 'cattle') has extended to
mean 'richness'.
(ii) In the same way, restriction goes from the less valued to the
most one: frumentum, which in Latin meant 'cereal', became fro-
ment (,wheat') in French (the most valued cereal); a form of this
valuation remains in the familiar acceptation of ble ('dough').
Viande (with the generic meaning of 'food' in Old and Middle
French) "becomes" viande ('meat') to designate food par excel-
lence. 13 This applies also to synchrony (diatopical): in Marseille, for
instance, you can hear J'ai un enfant et deuxfilles 'I have a child and
two girls', enfant being restricted to 'boy', eminently valorized in
this Mediterranean city.
Qualitative inequalities between marked terms (or paragons) and
unmarked terms are thus linked to two principles of distribution and
panchronic summation. Extension is hence a distribution of the posi-
tive evaluation from the paragon, and restriction the summation to-
wards the paragon. The law of panchronic valuation expresses in this
way the relations between intense and extense zones, the restriction
towards the highly valued consisting in a passage from the extense
zone to the intense zone, and the extension from the highly valued
bringing about the opposite movement. 14
Intense zone
Marked sememe
pancbronic pancbronic
restriction extension
Unmarked sememe Unmarked sememe
Extense zone
Figure I. Law ofpanchronical valuation

Examples: synchronic restriction, see enfant; diachronic restric-

tion, see viande; synchronic extension, see gagner son bifteck 'to
earn one's living'; diachronic extension, see egeria, hoover.
Let's return to our four diachronic operations.

(i) Extension and restriction. - We can now reverse in tenns of
valuation the law of Port-Royal (which nevertheless is only a law of
quantitative and not qualitative variation): the more an intension is
valued, the more extensions it has.
(ii) Tropes. - If we agree on this point, metonymy (and synec-
doche which it is often confused) does not obey to other rules: it is
from this point of view only a particular case of extension. It is the
most priced part or associated unit which extends its name to the
whole or the set. The transfonnation of proper names into "common
nouns", extremely widespread and witnessed everywhere, goes in
the same direction: egerias and Hercules, Hoovers and pes are ever-
yday examples. Admittedly, these valued tenns become unmarked
and get neutralized by their very use, and other words come and re-
place them in a never ending process without any direction - but not
without showing some regularities.
The usage of the concept of metaphor answers a totally different
problem, the one of changes in semantic domains. But, since tax-
ernes are included in domains, any change of domain leads to an al-
teration of the taxeme. There was a time when, in Old French
ouailles (from the Latin oviculas) meant 'ewes' in the domain
AGRICULTURE and 'faithful companions' in the domain RELIGION. In
modem French, only the last acceptation survives. The evolution
process can be described in the following way: the evangelical meta-
phor of the good shepherd introduces, among others, a picture of the
domain AGRICULTURE into the domain RELIGION. As every metaphor,
it shows two effects: in the classeme, it virtualizes the generic seme
and actualizes a new generic and afferent seme; in the semanteme, it
reshuffles the hierarchy of attributes (cf. Rastier 1987: ch. VII).
2. The taxeme as semantic form
We have used the concept of semantic form according to our hypo-

thesis that semantic perception 1S pertains to pattern recognition and
not to computation.
2.1. Qualitative inequalities inside the taxeme
That qualitative inequalities inside lexical classes can be described

through intensity and extensity (Hjelmslev) is a fact that we shall re-
late with the hypothesis of fundamental aesthetics: languages do not
articulate descriptions (as the objectivist tradition would like it to
do) but evaluations. In particular, qualitative inequalities inside lex-
ical classes seem to be linked to social evaluations which evolve in
history. This is not surprising if we recall that in sfite of their name
natural languages are indeed cultural productions. I
Structural semantics, or at least the impoverished image given of
it by most manuals, presents nevertheless one deficiency: the ab-
sence of qualitative inequalities. Here are two ways to overcome it:
a) To distinguish evaluative zones inside a taxeme (by drawing
one's inspiration from descriptions given formerly by Coseriu 1968).
It is a means to break with denotation as no metrics can distinguish
the big from the huge or the cold from the icy.
b) To take account of diachronic variations of the signified: (i) by
variations of thresholds inside the taxeme, (ii) by changes of taxeme
(iii) and finally by changes of domain.
2.2. Elements ofa morphodynamic model
We shall use for this purpose a model taken from the theory of dif-
ferential varieties and dynamic systems. From Thorn's and Zeeman's
works in particular, it has been applied to the domain of speech and
case-frames by Petitot (1983, 1985), to semasiological semantics (on
118 FrQn~ois Rastier
the polysemy of Fr. encore by Victorri and Fuchs 1996), to the struc-
tural analysis by Piotrowski (1997).17
Continuist modelization presents one important advantage: the
gradual and the discrete can be described as particular cases of the
continuous and not the reverse. Hence, without suggesting any
strong hypothesis on the continuous character of the "semantic
space", neither on the spatial nature of cognitive schemes, we admit
that the semantic discretization consists in isolating outstanding
points or zones on dynamics. We make the assumption that semantic
evolution can be represented on a gradual process with thresholds:
the discrete elements (like semes and phemes) result from the cap-
ture of discontinuities (see the phenomenon of categorial perception
discovered by Liberman).
A morphodynamic model is characterized by the functional
coupling of an external space (or substrate space or also space of
control) with the internal states of a system S. The singularities of
the internal space are projected as discontinuities on the external
space.
Pass
Attractor
Figure 2. Dynamic system S

As we have chosen an onomasiological perspective, we consider

as semantic space the set of the sememes of a taxeme,18 for instance
the one of levels of temperature: 'icy', 'cold', 'cool', 'lukewarm',
'hot'.19
acceptability threshold acceptability threshold

I I
Paradoxal zone Doxalzone I Paradoxal zone
I I
I lukewarm
I
Figure 3. Dynamic representation of a taxeme
Regrettably, this representation ignores the context, because de-

termining the main attractor and the position of qualitative and ac-
ceptability thresholds depends on the context (cool is neutral for a
beer, not for a bath2~.
The taxeme displays three main basins of attraction separated by
two absolute maxima: we call doxal zone the big basin of attraction
delimited by two absolute maxima and located below the acceptabil-
ity thresholds; and paradoxal zones the two zones located beyond
these thresholds.
Each sememe corresponds then to one local attractor. All the vari-
ous contexts attested for it define its basin of attraction (in synchro-
ny). The bottom of the basin of a sememe corresponds to the zone of
semantic stability, in other words to its meaning (prevailing accepta-
tion, or in other words set of semes inherent to the sememe). The
slopes of the basin, zones of instability, correspond to usages. The
basin broadens or narrows as the number of contexts grows or dimi-
nishes. The marked or intense terms correspond to narrow basined
attractors and the unmarked or extense terms to wide basined attrac-

tors.
The slope of the basins vary whether the sememe has more or less
usages in the corpus. Two neighbouring sememes are separated by
passes whose altitude is also variable: they are low when the se-
memes count many similar contexts, and high when they only count
a few or none at all. Parasynonyms are separated by low passes
which do not cross any qualitative threshold; the distinction, which
is always possible, consists in lowering the closest qualitative thresh-
old.
Discretization
1.
qualitative
threshold
'visage' 'face'
2.
'visage' 'face'
Marking
I 'visage'l I 'face' I
Figure 4. Discretization and marking
The differential principle of structural semantics expresses itself

by a covariance of the basins of attraction of sememes. Any local
distortion may have consequences on the global form and neigh-
bouring sections of local forms. 21 Yet, as the global determines the
local, even if an important local perturbation may modify the global
organization, a taxeme shows a certain resistance to the distortions

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

perior qualitative threshold (corresponding to the paragon, e.g.
bifteck).
• The attractor furthest to the absolute maximum, which is also the
one whose basin of attraction is the most widespread and/or the
deepest because it accepts the highest number of contexts (we call
it generic neutral, e.g. rue).
To these two attractors correspond the two main and con-

tradictory definitions of the prototype as defined by Rosch's theory.
acceptability threshold acceptability threshold

I
~
'~ , endoxalpaili paradoxal paili ",-'"
'..,. ",-
qualitative
threshold
generic neutral
Figure 5. Thresholds
The relationships between part-whole inside the global basin of

the taxeme can also be described as relationships between these two
outstanding points, in a way which does not have recourse to the
concept of inclusion, but to the one of relationships between minima
and maxima. .
A "logical" representation can indeed norm and codify a taxeme
but not represent the dynamic it has created nor the ones that make it
evolve. According to the modelization we have put forward (1987),
forces are inhibitions, activations and propagations. These opera-
tions suppose "potential differences" which are valuations.
2.3. Evolutive paths: endoxal and paradoxal
We distinguish two kinds of differential thresholds, qualitative

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

I
I I
I.----------------~
street 2
Figure 6. Extension from the neutral term
The same endoxal evolution can take place several times in

history. For instance passer in Latin ('sparrow') becomes pajaro
('bird') in Spanish, passereau in French (OFr. passe(re) ~ passerel
> passereau 'passerine', which designates all kinds of small birds);
in a renewed way, pia[, which in the first half of this century meant
'sparrow' has extended its meaning to 'bird' in familiar French.
Whereas the law of panchronic valuation expressed the relation-
ships between extense and intense zones, this form of evolution ex-
presses the relationships inside the extense zone: being established
inside a same zone, they characterise themselves by the conservation
of evaluation.
2.4. Force and form
The description of taxemes as sets of static relations has indeed to its

credit simplicity and economy: these relations can be characterized
as logical relations, and this brought quite a success to descriptions
through semantic features (actually used by defenders as well as by
opponents of structural semantics).
But the description of forms has no relevance if one cannot ac-
count for their evolution. Cognitive semantics acknowledges abstract
forces, especially of course in its analyses of actancy when, fol-

lowing Talmy, it enacts a kind of transcendental mechanics which
owes a lot to Aristotelian physics (see the concept of impetus in Tal-
my 1988). Forces and semantic forms must nevertheless be articula-
ted. The morphodynamic description may account for forms in terms
of forces. To say the least, these two aspects, form and force, are
complementary: a force is felt and measured by the distortions it en-
tails; a stabilized form results from an always momentary balance of
forces.
The effect of forces can be understood in two ways: the move-
ment of critical points; the concomitant distortion of the "normal"
sections of the form. The sections that exceed a threshold gain a
potential of disturbance. From this point of view, the edges of the
evaluative zones are privileged spots of disturbance. Beyond a
certain level, peripheral disturbances modify the main basin of the
taxeme.
2.5. The conditions ofform evolution
In morphodynamics, regular points and singular points are distin-

guished. As a form is better recognized by its singular points rather
than its regular points, some of the relationships which are char-
acterized, by analogy with perception, as relationships form/content
can be described or reformulated as relationships between the regu-
lar sections of the form and its singular sections. For instance, at the
textual level, we have described isotopies as products of the Gestalt-
ist law of good continuation: they are as such regular portions oftex-
tual forms, and appear then as semantic backgrounds. On the con-
trary, allotopies are singular points and certain tropes introduce
qualitative discontinuities by a disruption of isotopy. 24
At the lexical level, the isotopic contexts of a sememe keep a re-
gularity, whereas allotopic contexts create singularities which may
sustainably disturb the basin of attraction of the sememe. This point
must be qualified according to the typology of sememes: sparse with
low associated values, sememes are compatible with a large number
of contexts. For instance the one of grammatical morphemes, which,

as a general rule, does not present any generic feature of domain, can
show a great deal of isotopic contexts - and this is the reason why
the same grammatical morphemes are found in all discourses: hence
the diachronic evolution of grammatical morphemes is slower than
the one of lexemes.
2.6. Value and values
Let's admit that forms (here semantic forms) are inhibited move-
ments. As a normed set of evaluations, a doxa is precisely composed
of prescriptions and inhibitions which ensure a synchronic or dia-
chronic stability to semantic configurations. We retrieve here
Barthes's intuition that the lexicon is a frozen doxa (1984: 129) but
we pluralize it: frozen doxas. For instance in French there is no
taxeme of levels of temperature: the sememes expressing levels of
temperature are organized in different taxemes according to con-
texts. The same goes for sizes: 5 ft 10 is considered normal for a
man, tall for a woman, and giant for a child. This size is hence
respectively below and beyond the qualitative threshold, then be-
yond the acceptability threshold.
Whereas cognitive semantics, after having reinvented the Kantian
oversimplicity, is looking for the descriptive categories around trans-
cendental aesthetics as a prior frame to any percefstion, we have used
the term aesthetics in a more restricted meaning. S Therefore the his-
toric project of diachronic semantics leads us rather to look amongst
social evaluations to find the forces which shape and distort the
lexicon. Thus only the structural principles of semantic organization
pertain to language; relevant categories pertain to the specific lan-
guage considered; but the particular organization of its configura-
tions pertains to evaluative norms which are subject to variations.
The changes in evaluation resulting from a crossing of a qualitative
threshold or mainly an acceptability threshold introduce modifica-
tions that can be both fast and long lasting.
We clearly understand that everything depends upon the position

of the evaluative thresholds (whether they are qualitative or accepta-
bility thresholds). According to the position of these thresholds, the
structure of the taxeme changes; and this position does not depend
on language but on doxa. Therefore, every stabilized semic fonnula
results from an undefmed series of contextual relations, i.e. from an
interpretative tradition which varies with discourses and social prac-
tices.
One cannot talk about doxas, as systems of valuation, without
mentioning the social values which they reify. Without trying to play
with words, one can wonder if the value (relevant linguistic differ-
ence) is not ultimately founded upon the values, in the social sense
of the tenn, including the ethic and aesthetic judgements. 26 The law
of distribution and the absence of perfect synonyms enable us to out-
line a positive answer and to second it with examples: an icterus is
more posh than a jaundice, ethylism (especially if it is social) than
alcoholism, etc. In this case, interpreting an occurrence is not or is
not anymore a simple operation of relating it to a type, but of posi-
tioning it in a taxeme, and of locating it in one of the evaluative
zones of this taxeme.
3. Implementation: the denominations of FACE in French
For our proposition of illustration, a study of the joint evolution of

the words face and visage (figure) in French, we rely on Vaugelas
(1647), F. Brunot (1905) and especially Renson (1962).27
3.1. The different steps of this evolution
According to Renson (1962: I, 227)face is first attested in the 12th

century with the meaning of 'visage' 'face'. In fact, it means essen-
tially the surface of the face and in particular the cheeks, as it ap-
pears in contexts like cent foiz Ii baise de randon / Les ueuz la face
et le menton 'one hundred times he kisses at random / eyes, face and
chin' (Roman de Thebes, 6378). At the time, the plural faces with
the meaning of 'cheeks' is attested. 28 This acceptation justifies the
abundance of colour adjectives (vermaille, palie, etc.).
Yet, from the 13th to the 15th century, the number of contexts of
face grows substantially,face meaning then, apart from the physical
aspect of the figure, its expression (feature Imoral/): for instance in
Greban (doulce et tant prudente, benigne) or in Molinet (ayre use et
furibonde, humble et fort accointable; cf. Renson 1962: I, 229). This
evolution is general and also noticed for visage. It is evidence of the
psychologization of literature (which in fact constitutes the best part
of our documents).
But, as soon as the 12th century, as has been stressed by J. Trenel,
a new class of contexts appears with the use of face in the trans-
lations of the Bible (with the meaning of presence: la face de Dieu
[which translates the Hebrew plural p 'nim Elohim] and also of sur-
face: la face de la terre 'the face of the earth'). The face of God is
without any precise features (see Renson 1962: 1,233). This new ac-
ceptation, translated literally from Hebrew, shows perhaps a relation
with the spiritualization of the acceptation which describes the hu-
man figure: thus we find in Arnoul Greban adjectives like dampnee,
digne and saintissime, or in Molinet angelique.
On the opposite, visage is more rarely found in religious contexts,
and in the assumption of a reciprocal reinforcement of neighbouring
semes, it can be said then thatface shows the afferent feature Ireligi-
ous/, brought to the forefront in certain contexts.
It remains that in the 16th century face presents two acceptations,
for God and for the human beings, the first being ameliorative and
the second neutral, liable to physical as well as moral uses, and shar-
ing with visage many common contexts. The configuration is then as
follows: 'face 2' (/human/) and visage share a common general basin
of attraction and are only separated by a low pass, whereas 'face l'
(/divinel) lies beyond.
qualitative
-r~+-~----~~------~~~--~:~~d
I I I~icall
W i
;l1li
fuce 1
ldivim'
Figure 7. Fr.face, end 15th c.
Yet, at the end of the century appears the expression face du

grand Turc, which designates a totally different body part; thus Sieur
Tabourot des Accords wrote : son mari qui etait tout nud sur le tict,
avait laface du grand Turc tournee de ce cote la 'her husband who
was stark naked on the bed had the face of the Grand Turk turned
that way' (Escraignes dijonnaises, 42 VO). Must we recall the estab-
lishment of diplomatic relations by Fran~ois the First with the
"Sublime Porte" and suppose that the wars between Christendom
and Islam could well make of this expression the ludicrous and dis-
creetly blasphematory reverse of the divine face? Whatever the case
may be, the word face itself, according to Ferdinand Brunot and in
spite ofVendryes's reluctance to admit that this could be the reason
of it, finds itself marked with pejoration. In 1627, MIle de Gournay
feared that the "new critics" might decide suddenly to refuse to "es-
crire face, ... generalement refusee du nouveau jargon, Parce qu' on
dit la face du Grand Turc" [write face ... generally rejected from the
new lingo, because one says the face of the Grand Turk] (L 'Ombre,
1627: 958, cit. in Livet, 1895: 303). Twenty years later, Vaugelas
already emphasized that one "would not any longer dare say face for
visage except for certain expressions" (1647: Preface, IX). Hence its
sudden rarefaction according to Renson's counting: from 23% of the
130 Fram;ois Rastier
designations of figure in the 16th century, it drops to 3% in the 17th

century.29
Thanks to the afference of the feature lpejorative/, the acceptation
'face 3' is now separated from visage by an acceptability threshold.
Now visage covers the doxal zone, surrounded by two antipathetic
and paradoxal acceptations of face, one religious and the other one
infamous. This configuration is unstable, not by its form, but most
certainly because 'face l' and 'face 3' share the same signifier.
acceptability acceptabily
threshold threshold
qualitative
.L -1- .L threshold
1 1 1
~I ~I~.- ______________~~~
face I face 2 face 3
/divine/ !human! /obscene!
Figure 8. Fr.face, end 16th c.
The reduction of the basin offace benefits then to figure, which,

as soon as the middle of the 16th century, by a specialization of the
meaning of 'exterior shape', was used for 'shape of the human face' ,
and later, in the 17th century, was also used for the facial expression
or look (Fr. mine, 1662), and came to replace visage and face in the
current usage (see Brunot: 1905, s.v.).
Cognitive semantics and diachronic semantics 13 I
acceptability threshold acceptability

I I threshold
qualitative
threshold
' ..
face I
.,' I .. .,1
figure
1
...I.
face 3
Idivinel lobscenel
Figure 9. Taxeme of Fr.Jace (end 17th)
In today's French,figure has become a neutral and generic term,

liable to the most diverse uses; face keeps its feature lpejorative/, as
shown by the insulting phrase face de .... Finally, visage, reserved to
the written language or to an elevated spoken style, has gained a fea-
ture lameliorativel (widely used in advertisements for cosmetic pro-
ducts). Visage keeps the feature /human! but acquires a new accepta-
tion (although already very common in Montaigne) as it is suitable
for the aspect of very diverse objects (socialisme a visage humain,
1968), which anticipates perhaps, as formerly for 'face 1', its disap-
pearance from the class of the denominations for human face.
~
I
I
I
+---+----1--="'=
I
~ ~I
face I visage figure face 3
Figure 10. 20th c.
3.2. Discussion
3.2.1. About the unity of the taxeme
It can be objected thatface inface du grand Turc belongs to another

taxeme, the taxeme of the designations of the posterior. Yes, but
'face 2' remains a pejorative designation for visage: "Caesar be-
lieved that long and lean figures (visages) were real faces (faces) of
conspirators" (Voltaire, letter to d' Argental, February 11th 1764).30
Furthermore, it could be objected that, strictly speaking, 'face l'
belongs to the religious domain. Fine, but the human visage is also
designated there by face: the expression face-a-face (face to face)
comes thus from religious texts describing the encounter of man and
God.
Moreover, expressions likeface d'abbe (literally "face ofa rever-
end") 'red and illuminated face' (Cotgrave, 1611; Oudin, c. 1640),
face de careme 'pale and pallid face' (Panckoucke, 1749), are al-
ready attested in the 17th century in Bois-Robert (Epftres) or Racine
(Les Plaideurs). These two antithetic acceptations of face persist in
religious and moral contexts. Littre places side by side face de re-
prouve (literally 'face of outcast') 'sinister and frightening physio-
gnomy', and avoir une face de predestine (literally 'having the face
of a predestined') 'having a full, rosy and serene face' .31
To accoWlt for these opposed acceptations offace, the 17th cen-
tury lexicographers used the concept of style. Richelet (1680), the
ftrst to relate the condemnation offace, said: "This word is still used
in solemn and majestic poetry but not in gallant, lively poetry". Fu-
retiere (1690) claims on the contrary that ''the wordface is only used
for ftgure in mockery for a face that is too thick or too large" ("le
motface pour visage ne se dit plus guere en ce sens qu'en raillerie
d'un visage qui est trop gros ou trop large"). The Dictionnaire de
I 'Academie, in its second edition, will suggest a synthesis by distin-
guishing the "serious style when talking of God" and the "familiar
style: uneface rejouie, enluminee ('ajoyful, illuminated face')".
The theoretic problem raised by the notion of style should not be
underestimated; in fact, we admit that a taxeme belongs to one and
only one semantic domain. In so far as taxemes are classes of seme-
mes which constitute a basis for term selection within a given prac-
tice, and where domains reflect different practices, the sememes of a
given taxeme pertain to the same domain.
In today's French, 'face l' remains in religious texts, since, in
spite of the aggiornamenti, the diachrony of religious discourses
does not follow the same evolution as the other discourses grouped
under the enigmatic name of langue generale [general language]. 32
This diachronic autonomy of a discourse is not surprising: in medi-
cine, face still means 'visage', or more precisely 'the front part of
the head'. More generally, it can be pointed out that the different dis-
courses and the different social practices that they reflect are moving
aroWld differentiated diachronies: the evolution of a language, and in
particular of a lexicon, obeys to very various types of temporality.
Thefaces d'abbe and thefaces de predestines, which looked very
much alike by their florid complexion, have disappeared from to-
day's French: no ameliorative contexts are fOWld where face would
mean 'visage'. Thus, in an expression like face d'ange 'angel's
face', the pejorative feature of 'face' is propagated to 'angel' and not
the ameliorative feature of 'angel' which contradicts the one of
'face': see e.g. Teierama about the trumpeter Chet Baker who died
134 Fran~oisRQStier
of an overdose: "La mort donna des ailes a une face d'ange" [death
gave wings to an angel's face]. The disappearance of the ameliora-
tive acceptation offace, which appeared in the 12th century with the
translations of the Bible, is perhaps an indication of a secularization
of society, where the terms taken from the Scriptures are not any
longer a pledge of what the Academy called a "serious style". It
indicates also that lpejorativel has become an inherent feature of
'face 2'.
3.2.2. From syntagms to phrases
We must point out here a deficiency of the lexicographical approach:

traditionally linked to an ontology of the concept and the isolated
word, it deals only with simple lexical items and neglects complex
lexical items. Yet, in the religious discourse, naturally prone to
formulas through its links with the ritual, complex lexical items like
face de Dieu are particularly important.
In fact, for textual linguistics (and I believe that lexicology should
be founded on such linguistics), the minimal semantic unit is the
syntagm. Words are units deprived of their preferential contexts, de-
contextualized artefacts both of ontology and lexicography. In this
perspective, simple lexical items are defined as strongly integrated
syntagms (fact actually confirmed by linguistic evolution: each word
is a fragment of a myth). Yet, the diachronic status of syntagms and
complex lexical items are different:
• A non integrated syntagm like face du grand Turc can be an ini-

tiator: semantic innovation can create a local disruption, which,
amplified by other factors, will lead to a revision of the taxeme
(creation or deletion of an attractor, or simple distortion of the
global basin).
• Integrated to a complex lexical item, like face-a-face, a simple
lexical item does not retain the properties of its original taxeme:
its sememe is redefined in a new taxeme and is not revised by the
disruptions of the original taxeme. Thus, as Vaugelas pointed out:
"Pour les personnes, on dit encore regarder en face, reprocher en

face, soustenir en face, resister en face mais toujours sans l'article
la" [for people, we still say regarder en face, reprocher en face,
soustenir en face, resister en face but always without the definite
article la] (1647: 60). And indeed,face-a-face has survived with-
out any ameliorative feature (although it was drawn from transla-
tions of the Bible), nor subsequent pejorative feature; just like vis-
a-vis, although vis owes probably its obsolescence to some form
of ostracism if not some taboo (see below 3.2.3.).
3.2.3. The two networks of content and expression
If the norms of the doxa may account for the evolution of the signi-
fieds, they also participate in the evaluation of the signifiers, and the
signs fmd themselves caught into two networks: (i) an onomasiolo-
gical network which accounts for the evolution of the signs accord-
ing to the evaluative variations of the signifieds; (ii) a semasiological
network which takes into account the evaluative variations of the
signifiers. The evolution of a sign obeys to these two networks of
constraints.
Yet, and this is worth being noticed, the same evaluative forces
are at work in the onomasiological and semasiological networks: in
the first case, we have seen that the proscription concerning the face.
of God has extended to the human face. For the second, let us con-
sider the case of a prohibition concerning the signifier and let's take
the example of vis which belonged in Old French to the designations
of the visage. The dropping of the last consonant has transformed it
into a homophone of vit ('prick').
The homophony with vit was obviously not lost on facetious
minds. 33 The opposition between noble and evil parts of the body
was sufficient to inhibit vis in favour of visage. 34 Thus the same kind
of prohibition may contribute to inhibit a signifier (vis) or a signified
('face' 2). And in both cases, the controversial sign is eliminated
from the taxeme.
4. Conclusion
4.1. The historic contingence and the ideological conditions
Let us imagine that the hilarious Sieur Tabourot, being or not the in-
ventor of it, had never published this face du grand Turc, then the
designations of visage might perhaps have been changed. Without
returning to Plekhanov's concluSions on the role of the individual in
history, it is enough to extend to languages the well known fact in
the dialectology of small communities that a lexical innovation can
always be attributed to one person. In the written languages, it is not
rare that literature innovates and that its innovations are taken up in
the oral language (see in Chinese the four character expressions).
Nevertheless, for an innovation to be taken up, it must use salient
semantic categories: it is useless to recall, at the end of the 16th cen-
tury, the disturbance caused by the wars of religion, the expansion of
scepticism and philosophical dissoluteness. During the following
century, the censorship of the social, pious or academic good taste,
contributed, as we know, to normalize the language and to put an
end to semantic ambiguities: hence face was sacrificed on the altar
of decency.
4.2. The anthropological foundation of diachronic evolutions
We have seen that the evolution of taxemes can be described as a

succession of distortions. Among the forces that are instrumental in
it, ameliorations and pejorations play an antithetic role. The more a
content is highly valued, the narrower its basin and the more it can
be distorted. The zones neighbouring the acceptability thresholds are
unstable, the pejoration of a very valorized term (as was the case for
face) is enough to disturb a complete arrangement. Thus, the dog-
matic value that was attached to face enabled to transform a new
context into a blasphemy (as if faith was closer to blasphemy than
indifference).
More generally, as doxal contradictions correspond to social con-

tradictions, a conflict between groups can express itself by a conflict
between doxas, and a paradiastole is always possible (example: your
heroes are murderers).
We can nevertheless retain antithetic evolutions. Valuation is ex-
pressed through: an activation, the broadening of the attraction basin
(and therefore of the contexts), the acquisition of a typical value, the
use in an elevated style. On the opposite, tabooand prohibition are
expressed by: the active inhibition of the singularity and of the rela-
ted points, the narrowing of the attraction basin, the quantitative
rarefaction of form, and often a euphemistical shift and the use of a
sloppy style. 35
One wonders then, in the continuation of what we said about
value and values, if the founding of semantic oppositions does not
lie in the doxal opposition between the valorized and the devalor-
ized, and, ultimately, between the prohibited and the prescribed. If
this is the case, semantics, be it cognitive or not, can only have an
anthropological foundation, articulated upon ethnology and history.
Notes
1. Cf. e.g. 1971: 164.

2. Unless you believe; like Rosch following Berlin and Kay, that prototypes
are founded by nature. Surprisingly, for these authors, nature is achronic
(and yet, for instance, the mammoth is not any longer a prototype of big
game animal).
3. One can indeed wonder if her insistence on fuzzy membership does not
come from an insufficient defmition of the categories (which do not cor-
respond to any lexical classes listed by linguists).
4. In fact, nature is only a naturalization, i.e. a naturalization of doxa or the
ideology of common sense.
S. One knows that semasiology takes the signifier as invariant and considers
the problem of polysemy as fundamental, whereas onomasiology starts
from the signified and considers synonymy as prevalent. From the thesis
that synonyms exist in any language, advanced by Abbe Girard (after Pro-
dicos) and pursued by the synonymists of the 18th century, was born to-
day's differential semantics (see n. 9).
6. This ontological conception of meaning can be found in Guillaume with the

theory of the potential signified, as in the theory of prototypes. The main
discordance between Guillaume's approach and the cognitive approach lies
in the fact that the potential signified is an abstract form, whereas for
cognitivist semanticists (furthermore as opposed to Rosch), it is a better
specimen.
7. One knows that the transcendental theories of meaning, like Guillaume's or
Culioli's which was born from it and, lastly, cognitive semantics, describe a
time that is not the time of history, whereas diachronic semantics is a histo-
rical discipline.
8. The concept of value explains as well the law of distribution, which Breal,
developing in a historical perspective the synonymists' fmdings defmes as
follows: "We call distribution the intentional order as a result of which
words which should be synonymous and which were such in fact, have
nevertheless taken different meanings and cannot any longer be used one
for the other" (1897: 22). He concludes: "The history of language is a
series of distributions" (1897: 22).
9. With which the semic analysis bears incidentally some relation. See Breal
about "our fathers from the Condillac School" (1897: 277).
10. And for us the reference is an effect, not a starting point (see Rastier 1991,
ch. VII, 1994, ch. 11).
11. Anxious to found in nature the linguistic evolution, Sweetser makes her
own the legend, long ago cleared up by Meillet and Benveniste, that the
concrete meaning comes first in comparison to the abstract meanings.
12. Metaphor and metonymy, emblems of the restricted rhetoric since Jakobson
has strangely enough coupled them, are neither symmetric nor converse;
and their choice is furthermore problematic: why exclude the other tropes?
13. To the contrary and complementarily, as we will see (Fig. 1), bifteck came
to mean 'food' by extension from the valorized form.
14. This law has perhaps anthropological foundations; at least according to
Louis Dumont, the hierarchy subordinates the includer to the included, the
extense term to the intense term, the unmarked to the marked (see 1992: 11-
24).
15. We consider language not as the reflection or the transposition of percep-
tive forms, but as an object of perception (hence our proposals for a theory
of semantic perception): the link between language and perception lies
here. We leave open the question of the incidence of semantic perception
on the other forms of perception.
16. The diachronic structural semantics and the cognitive semantics, in so far as
they take into account general conditions of the capture of meaning and
fmd probably a common foundation in Husserl's phenomenology (especial-
ly the concept of qualitative discontinuity presented in the third Recherches
logiques 1901), will perhaps give us a way to articulate the cognitive (as
transcendental) and the cultural by taking into account cultural and histori-
cal conditions of cognition as languages force and articulate it.
17. Due to lack of space, we refer you to his excellent presentation, 1997: 167-
210.
18. And not the whole series of contexts associated to one expression (as Vic-
torri and Fuchs [1996] did for instance in a semasiological perspective).
19. Representational theories of meaning do not succeed in accounting for qua-
litative inequalities. Yet, distinguishing evaluative zones inside the taxeme
gives us a means to break with the representational theory of meaning,
since no metric enables us to distinguish the big from the huge and the cold
from the icy.
20. See Rastier 1996. Among the levels of acidity of a wine (weak, soft, cool,
vivid, nervous, sour, green), the same level, measured in pH, and express-
ible in abstracto by nervous, will be categorised as vivid (non pejorative)
for a young wine, and by sour (pejorative) for a wine that aged.
21. In traditional semic analysis, for instance, the introduction of a sememe in a
taxeme modifies the semic composition, i.e. the relational structure of the
other sememes.
22. The problem raised by the sememes which would mean one thing and its
opposite (as the famous addad of the Arabic tradition) is in our opinion
only raised in cases of antiphrasis or syllepsis, and therefore in context.
23. Thus the Annuaire des rues de Paris (Directory of Parisian streets) contains
not only streets but also boulevards, avenues, courts, etc.
24. In this way, we reformulate, without linking it to any zero level, the pro-
blem of the deviation which preoccupies stylistics.
25. The human universe is not made of knowledge on the one hand and emo-
tions on the other hand. This omnipresent distinction, up to nowadays cog-
nitive sciences, reiterates without any foundation the archaic separation of
heart and reason. Undoubtedly, the neutrality of information is only a mod-
em artefact which agrees with the persistent prejudice that language is a
simple ideographic instrument for the use of rational thought. Let's agree
that the human universe is made of social and individual appreciations,
which are the object of fundamental aesthetics. Fundamental aesthetics per-
tains to linguistics when it takes as object the linguistic material itself. At
the morphological level, all languages contain appraising morphemes (see
for instance the affix -acci- in Italian). At the immediately superior level,
the lexicon of languages swarms with evaluations, and acceptability thresh-
olds structure the elementary lexical classes. All the more for phraseologi-
cal units, very numerous in any text, which reflect and propagate a social
doxa. At the level of the sentence, it can be considered that any predication
is an evaluation. At the textual level at last, narrative analysis for instance
has underlined many times the importance of modalities qualified as thy-

mic. To put it shortly, fundamental aesthetics defines the semiotic substra-
tum on which arts of language are built, and remains well short of philoso-
phical aesthetics. It has nothing in common with any aesthetic, poetic or
stylistic function.
26. This question may take Saussure as an authority when he defmes two in-
dissoluble aspects of value: the internal value, which draws its principle
from differential semantics, and the external value, for which he gives the
example of the coin. This counter value remains metaphorical, but the arti-
culation of the two types of value raises a problem (see Piotrowski 1997): it
is in our opinion the correlation between linguistic valuations and social
values (of which the economical exchange values are only an isolated case,
exemplary because normed).
27. My thanks to Evelyne Bourion for having passed to me the elements of this
document.
28. Example: D 'andeus ses oiz ses faces moille 'Both his eyes wet his face'
(Benoit, Ducs de Normandie, 5114; six similar examples are found in this
piece) to be compared with Plure de ses oils, si li moille saface 'His eyes
weep and soak his face' (Chanson de Guillaume, 478).
29. One example: Corneille still uses it in the sense of visage in Medee (1635),
but uses it only from then on in religious contexts. Vaugelas may fmd that
''usage put it out of use" (1647: Preface, IX) for ridiculous, extravagant and
very insulting reasons. He recognizes that "qu'en meme temps que je con-
damne la raison pour laquelle on nous a oste ce mot dans cette significa-
tion, je ne laisse pas de m'en abstenir" [whereas I condemn the reason why
this word has been taken away from us with this meaning, I can only ab-
stain from using it]. The first edition of the Dictionnaire de l'Academie
(1694), already behind usage, simply defmesface by visage; but, if the fifth
edition (1798) confirms this, it adds that "Dans le serieux, il ne se dit en ce
sens qu'en parlant de Dieu" [seriously, it is only said with this meaning
when referring to God].
30. Even if the comparison with the posterior remains possible, in Zola for
instance where the eaters in L 'Assommoir "avaient des faces pareilles a des
derrieres" [had faces like bottoms] (ch. VII). We make sure to distinguish
the metaphor from its expression: it would have been preferable to associ-
ate the word visage to a comparison of the face and the bottom. Renson
quotes ['autre visage in Voiture and in various authors luckily forgotten vi-
sage sans nez 'face without nose', gros visage 'big face', visage a rendre
un lavement 'face to bring up a lavement' (1962: 1,210). But these occur-
rences remain isolated and were not adopted by usage.
31. The opposition of these two expressions can already be found in the Dic-
t;onna;re de ['Academ;e (1835). Victor Hugo unites brilliantly the two ac-
ceptations: "Le voyage qu'ils [mes parents morts] font est profond et sans
bomes, / On le fait a pas lents, parmi des faces momes, / Et nous le ferons
tous" (Hugo, Feuilles d'automne, 6). In this context, 'face 2' appears in the
religious domain (a funeral).
32. The evaluative norms belonging to this kind of discourse have saved face
from infamy. Using an understatement, the Jesuit authors of the Diction-
naire de Trevoux (1721) notes on expressions containing the word face that
"these phrases (face-a-face, etc.) are imitated from the Scriptures ... This
makes the use of certain terms more acceptable". The discourse of Law is
probably not subject to taboo. At least, the taxeme of designations of the
face does not know the same diachrony in the religious discourse than in
others.
33. One of Molinet's ballads, by and large obscene in view of its rhymes ob-
ligingly itemised in the manuscript for absent-minded readers: "Madame,
j'ai sentu les fa~ons / Du feu d'amour, puisque je vis / Les yeux plus aspres
que faucons / De vostre gent et plaisant vis" (cf. Dupire 1936: 866-867).
34. The competition of visage, suggested by Renson, is all the more unconvin-
cing since visage is derived from vis; the monosyllabic character of vis has
also been put forward, but we can fmd monosyllabic words which are quite
durable. Like face, vis fmds itself banished because visage is traditionally
opposed to parts of the body considered as evil. This theme, as we know,
has been developed by Freud, who takes up Schopenhauer's polarization
("the head and the genitals are, so to speak, the opposed poles of the
individual", excerpt from Le monde, in Insultes, Monaco, Le Rocher, 1988,
p.29).
35. Euphemization has certainly an anthropological generality: for instance the
name of the living body is extended to the dead body (the word cadavre
'corpse' is not used during a funeral). The use of a neutral term like corps
'body' (ex. levee du corps 'funeral' is not limited to an extension: we thus
avoid the pejoration of the dead body).
References
Barthes, Roland
1984 Le bruissement de la langue. (Essais critiques 4.) Paris: Seuil.
Blank, Andreas
manischen Sprachen. (Beihefte zur Zeitschrift fUr Romanische Phi-
lologie 285.) TQbingen: Niemeyer.
Berlin, Brent and Paul Kay
1969 Basic Color Terms. Berkeley: University of California Press.
Breal, Michel
1897 Essai de semantique. Paris: Hachette [new edition Brionne: Mont-
fort, 1983].
Brunot, Ferdinand
1905-1937 Histoire de la langue franfaise des origines a 1900, 12 Volumes.
Paris: Annand Colin.
Chamfort, Sebastien-Roch Nicolas
1968 Maximes et pensees, caracteres et anecdotes. Edited by Jean Dagen.
Paris: Flammarion.
Cledat, Louis
1895 Les lois de la derivation des sens appliquees au fran~is. Revue de
Philologie 9: 49-55.
Coseriu, Eugenio
1964 Pour une semantique diachronique structurale. Travaux de lin-
guistique et de litterature 1111: 139-186.
1968 Les structures lexematiques. In: W. Theodor Elwert (ed.), Probleme
der Semantik, 3-16. (Beihefte zur Zeitschrift rur fran.zOsische
Sprache und Literatur, N.F. 1.) Wiesbaden: Steiner.
Darmesteter, Emile
1877 De la creation actuelle de mots nouveaux dans la langue franfaise
et des lois qui la regissent. Paris: Vieweg.
1887 La vie des mots etud;ee dans leurs significations. Paris: Delagrave
[new edition Paris: Champ Libre, 1979].
Dumont, Rene
1992 Anthropologie, totalite et hierarchie. In: Anthropologie et philo-
sophie. Paris: Centre Pompidou, 11-24.
Dumarsais, Cesar C.
[1730] 1988 Traite des tropes. Edited by Fran~oise Douay-Soublin. Paris: Flam-
marion.
Dupire, N. (ed.)
1936 Les Faictz et dictz de Jean Molinet, 3 Volumes. Paris: Picard.
Hjelmslev, Louis
1971 Essais linguistiques. Paris: Presses Universitaires de France.
Husserl, Edmund
[1901] 1963 Recherches logiques. Volume 3. Elements d'une elucidation pheno-
menologique de la connaissance. Paris: Presses Universitaires de
France.
Katz, Jerrold J. and Jerry A. Fodor
1963 The structure ofa semantic theory. Language 38: 170-210.
Koch, Peter
1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine
kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46.
Lakoff, George
1987 Women, Fire, and Dangerous Things. What Categories Reveal
about the Mind. ChicagolLondon: University of Chicago Press.
Langacker, Ronald
1986 An introduction to cognitive grammar. Cognitive Science XII: 1-40.
Livet, Charles-Louis
1895-1897 Lexique de la langue de Moliere comparee cl celle des ecrivains de
son temps. 3 Volumes. Paris: Imprirnerie Nationale.
Martin, Robert
21992 Pour une logique du sens. Paris: Presses Universitaires de France.
Nyckees, Vincent
1997 Pour une archeologie du sens figure. Languefran~aise 113: 49-65.
Petitot, Jean
1983 Les catastrophes de la parole. Paris: Maloine.
1985 Morphogenese du sens. Paris: Presses Universitaires de France.
1996 Les modeles morphodynamiques en perception visuelle. Visio I: 65-
73.
Piotrowski, David
1997 Dynamiques et structures en langue. Paris: Editions du CNRS.
Pottier, Bernard
1974 Linguistique generale. Theorie et description. (Initiation a la lin-
guistique B, 3.) Paris: Klincksieck.
Rastier, Fran~ois
1987 Semantique interpretative. Paris: Presses Universitaires de France.
1991 Semantique et recherches cognitives. Paris: Presses Universitaires
de France.
1994 Semantique pour I 'analyse. With collaboration of Marc Cavazza
and Anne AbeiIle. Paris: Masson.
1996 Chamfort: le sens du paradoxe. In: Ronald Landheer and Paul J.
Smith (eds.), Le paradoxe en linguistique et en litterature, 119-143.
Geneve: Droz.
Reisig, Karl
1839 Vorlesungen aber lateinische Sprachwissenschaft. Leipzig: Lehn-
hold.
Renson, Jean
1962 Les denominations du visage en fran~ais et dans les autres langues
romanes. Paris: Les Belles Lettres.
Rosch, Eleanor
1978 Principles of categorization. In: Eleanor Rosch and Barbara Lloyd
(eds.), Categorization and Cognition, 27-48. HiIlsdale: Erlbaum.
Sweetser, Eve
1990 From Etymology to Pragmatics. Cambridge: Cambridge University
Press.
Tabourot des Accords, Jehan
1922 Escraignes dijonnaises (after Cioranesco XVlth). Dijon.
Talmy, Leonard
1988 Force dynamics in language and cognition. Cognitive Science 12:
49-100.
Vaugelas, Claude Faivre de
[1647] 1934 Remarques sur la langue fran~aise. Edited by Jeanne Streicher.
Paris: Droz.
Victorri, Bemard and Catherine Fuchs
1996 La polysemie. Paris: Hermes.
Section 11
Descriptive categories
Losing control:
grammaticization, subjectification,
and transparency
Ronald W. Langacker
My purpose is to explore a common type of semantic change involv-

ing attenuation in the degree of control exerted by an agentive sub-
ject. Attenuation figures crucially in many cases of grammati-
cization. When carried to extremes, it results in subjectification
(Langacker 1990b) and in the transparency of highly grammaticized
forms (Langacker 1995a).
1. Background
As background, I must first introduce a few basic notions of cogni-

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

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

overall scope overall scope
I immediate scope immediate scope
I
I
~
I
%~ "_.lm
Figure 1. Scope and profiling
The conception of a relationship presupposes that of its partici-

pants, which are accorded varying degrees of focal prominence. It is
useful to think of this metaphorically in tenns of primary and secon-
dary spotlights which can be directed at various elements of the
Losing control: grammaticization, subjectijication, and transparency 149
scene onstage. The primary focal participant is termed the trajector

(tr). When there is a secondary focal participant, it is called a land-
mark (lm). In a construction (a larger symbolic assembly), an expres-
sion which specifies a relational element's trajector is thereby identi-
fied as its subject, and one which specifies its landmark is an object
(if it profiles a thing) or a non-subject complement. As a fully gener-
al characterization, therefore, subject and object status are claimed to
reside in the focal prominence directed at onstage elements, rather
than being defined in terms of any particular semantic role. As pri-
mary focal participant, the trajector is the initial point of access - the
"starting point" - in conceptualizing the profiled relationship
(MacWhinney 1977; Gernsbacher and Hargreaves 1992; Chafe
1994; Tomlin 1995; Langacker 1997,in press).
I must next explain what is meant by subjectivity and subjectifica-
tion (Langacker 1985, 1990b, 1995b). As I myself use these terms,
the crucial factor is vantage point, and in particular the relative posi-
tions of the subject and object of conception (which are of course to
be distinguished from the grammatical subject and object). The sub-
ject of conception is an implicit locus of consciousness ("perspective
point") which apprehends the object and - if it is only the subject -
is not itself apprehended. The subject's activity determines the
nature of the conceptual experience but does not per se figure in its
"content". To the extent that an entity functions as the subject or
object of conception, it is said to be subjectively or objectively con-
strued. It is construed with maximal subjectivity when it remains off-
stage and implicit, inhering in the very process of conception with-
out being its target. It is construed with maximal objectivity when it
is put onstage as an explicit focus of attention. Thus profiling and
trajector/landmark status imply a highly objective construal.
To forestall possible confusion, I should note that Traugott also
speaks of subjectivity and subjectification (cf. also: Traugott, this
volume). For her, the latter is "a very general tendency toward grea-
ter pragmaticization of meaning" (1986: 540). It includes at least
three overlapping kinds of changes: (i) the shift from externally
based descriptions to internally grounded assessments, e.g. boor
'farmer' > 'crude person'; (H) extension to textual and metalinguistic
150 Ronald W. Langacker
uses, as when observe develops the meaning 'state'; and (iii) in-
creased involvement of speaker judgment, as in while 'during' >
'although' (1989: 34-35). In Traugott's inclusive sense, subjectifi-
cation may just affect the domain in which a property or relationship
is manifested. For me, however, it is a matter of perspective. More-
over, only a particular entity (not an expression overall) is said to be
subjectively or objectively construed. Whether boor means 'farmer'
or 'crude person', for example, its profiled referent is the onstage fo-
cus of attention, hence objectively construed, whereas the speaker re-
mains an implicit locus of judgment and is thus construed subjecti-
vely. This is not to deny that the speaker's attitude becomes more
evident in the later meaning, the primary content being evaluative,
but the speaker carries out this evaluation without becoming the fo-
cused target of conception. There is of course no point in trying to
decide between Traugott's version of SUbjectivity and my own, since
both will figure in an overall account of grammaticization. The issue
is purely terminological.
The speaker (S) and hearer (H) function as conceptualizers, i.e.
subjects of conception, with respect to the meanings of linguistic ex-
pressions. They are normally offstage and subjectively construed.
With a simple noun like dog, for instance, they have only a concep-
tualizing role - indicated by the dashed arrows in Figure 2(a) - and
thus remain external to its scope. A deictic element such as an article
or demonstrative makes a specification concerning the speaker's and
hearer's apprehension of the profiled entity, thus bringing them with-
in the expression's overall scope. Nonetheless, being referred to only
tacitly, they are offstage and subjectively construed, as shown for the
dog in Figure 2(b). It is however possible for the speaker and hearer
to go onstage and even be profiled as a focused object of conception,
notably in the case of pronouns like I,you, and we, as seen in 2(c).
Losing control: grammalicization, subjectification, and transparency 151
(a) dog (b) thedog (c) we

overall scope overall scope overall scope
immediate scope immediate scope immediate scope
~ (®--@)
Figure 2. Positions of the conceptualizers in relation to scope
2. Subjectification
In previous works (1990b, 1991), I characterized subjectification in

terms of replacement: some relationship within the objective situa-
tion under description is replaced by a comparable but subjectively
construed relationship inherent in the very process of conception.
However, as suggested by Verhagen (1995) and Harder (1996: 352),
I have come to believe that this subjective component is there all
along, being immanent in the objective conception, and simply re-
mains behind when the latter fades away. This revised notion of sub-
jectification is represented abstractly in Figure 3.
Initial Configuration Attenuation Subjectification

overall scope overall scope overall scope
immediate scope immediate scope immediate scope
m>9-t9 >'?t9
tr lm
\ ;/ \ I/ , I
\ II
I
\
I I \ I I
\ , ::. \ , ::. '. I,
T \1'
@ T~ T(b
Figure 3. Attenuation in degree of subject control
In the initial configuration, the solid arrow in bold depicts an ob-

jectively construed, profiled relationship. The direction of the arrow
indicates some kind of objective asymmetry which motivates the
choice of trajector; the trajector is in some sense more active - typi-
callyas an agent, experiencer, or mover. The solid arrow labeled T
stands for processing time. The import of the diagram is that the
conceptualizer, e, does some kind of mental scanning from the tra-
jector to the landmark in conceiving of the profiled relationship (Le.
the trajector functions as a "starting point" or initial point of access
in conceptualizing it). The dashed arrows represent the conceptu-
alizer's mental activity, carried out through processing time. In the
second diagram, the bold arrow representing the profiled relationship
is given in dashed lines to indicate attenuation, involving both the
objectively conceived relationship and the trajector's role in it. Al-
though the conceptualizer continues to carry out the same or a com-
parable mental scanning as in the initial configuration, the objective
situation now offers less motivation for it. We will be investigating
the progressively more tenuous objective basis for both the relation-
ship and for the trajector's selection as primary focal participant.
Finally, the last diagram indicates the full disappearance of any
objective basis for the conceptualizer's mental scanning. A relation-
ship is still established between the trajectorand landmark, but the
Losing control: grammaticization. subjectification. and transparency 153
basis for it now resides exclusively in the conceptualizer's activity.

This relationship is subjectively construed because it inheres in the
process of conceptualization itself, rather than being an onstage
object of conception. To the extent that the same mental operations
figured in the initial configuration, we can say that the subjective
relationship was immanent in the objective one. Observe that subjec-
tification per se need not have any effect on the choice of focal parti-
cipants (trajectorllandmark alignment). It merely removes any objec-
tive basis for selecting the trajector as initial point of access.
A first example is the semantic extension relating the two senses
of across exemplified in (l):
(1) a. The child hurried across the street.

b. There is a mailbox right across the street.
In (1)a, the trajector of across (instantiated by the child) has that

status by virtue of being a mover, successively occupying all of the
points along a spatial path traversing the static landmark. This is
shown in the first diagram of Figure 4. In (l)b, on the other hand, the
trajector of across - the mailbox - is static, as seen in the second
diagram; it occupies only a single position vis-a.-vis the landmark
(equivalent to the final position in (l)a). This illustrates attenuation
with respect to both the objective relationship and the basis for the
choice of trajector. The profiled relationship is less inclusive and
less dynamic in this second sense of across, and the trajector does
not stand out as primary focal participant by virtue of moving.
In conceiving of the trajector following the objective spatial path
in (1)a, the conceptualizer necessarily scans mentally along the same
path as an inherent aspect of tracking the subject's motion. Although
objective motion is lacking in the derived sense of across, we see
from the second diagram in Figure 4 that the conceptualizer nonethe-
less follows the same path subjectively in locating the trajector with
respect to a reference point (R). Once the objective motion is strip-
ped away, the subjective mental scanning becomes more apparent.
Observe that this fading away of the objective motion does not in-
herently involve any change in trajector/landmark alignment; only
the extent of the trajector's role is modified. Instead of successively

occupying all of the locations constituting a spatial path, it occupies
only the final location.
across across'
immediate scope immediate scope
Im lm
>
~ R
\
@"
\IIJ-P
'.
,
"'A
\ \ 1I /
, d 11 '
T T
'c'" I,
\\'11,
;)
©
Figure 4. An example of subjectification
Although I consider this to be a case of subjectification, I have

left a dashed arrow to show that an objective basis for the conceptu-
alizer's scanning along the spatial path has not disappeared entirely;
usually some vestige of it remains. It might appear that the transition
connecting the two senses of across would have to be discrete: the
trajector either traverses all the locations constituting the spatial
path, or occupies only the final location, but there is no intermediate
sense in which, say, it traverses only a portion of the path. But the
extent of the motion is only one dimension of possible variation -
other dimensions involve the status of the motion as well as who
does the moving. Differences along these parameters provide a series
of cases that are intermediate between that of profiled, objective mo-
tion by the trajector, and at the other extreme, that of purely subjec-
tive motion by the conceptualizer imposed with no objective basis.
Such a series is given in (2).
Losing control: grammaticization, subjectiflcation, and transparency 155
(2) a. The child hurried across the busy street. [profiled objec-
tive movement by trajector]
b. The child is safely across the street. [static location result-
ing from unprofiled, past, actual movement oftrajector]
c. You need to mail a letter? There's a mailboxjust across the
street. [static location as goal of unprofiled, potential, fu-
ture movement of addressee]
d. A number of shops are conveniently located just across the
street. [static location as goal of potential movement by a
generalized or generic individual]
e. Last night there was a fire across the street. [static loca-
tion, no physical movement necessarily envisaged at all]
Only in (2)a is movement along the entire spatial path put in pro-
file. This constitutes an important and discrete difference: since the
other sentences profile only a single, static configuration, they are
imperfective (or "stative") and take the verb be. There nevertheless
remains a vestige of objectively construed motion. It undergoes a
change in status from profiled to unprofiled, and from actual to po-
tential to generic. The mover also changes, from being an objectively
construed participant (the trajector) to being subjectively construed
(the unmentioned addressee), and from a specific mover to a gene-
ralized or unspecified one. It is only in the last example, (2)e, that
the conception of physical movement may be entirely absent, leaving
only subjective motion by the conceptualizer, who ·mentally traces
along the path in order to specify the trajector's location.
Thus I do not envisage attenuation and eventual full subjectifica-
tion as occurring in a single step. It is more likely a gradual evolu-
tionary process involving small steps along a number of possible pa-
rameters. lbis will usually result in the coexistence of alternative
values at a given diachronic stage, with gradual shifts in preference
being responsible for changes that in retrospect appear to be discrete
(cf. Heine 1992).
Attenuation can be observed with respect to at least four parame-
ters (the grouping is somewhat arbitrary). We have already noted
change in status: from actual to potential, or from specific to generic.
1S6 Ronald W. Langaclcer
A second parameter is change in focus, i.e. the extent to which par-

ticular elements stand out as focus of attention, notably in terms of
profiling. We saw that actual motion by the trajector is profiled in
(2)a, but unprofiled in (2)b, which designates only the final locative
configuration resulting from that motion. Full elimination, exempli-
fied by the absence of any objective movement in (2)e, might be
thought of as the extreme case of defocusing. A third kind of at-
tenuation is a shift in domain, e.g. from a physical interaction to a
social or experiential one, as in the evolution of modals (Sweetser
1982, 1990). The final parameter is change in the locus of activity or
potency. This is illustrated in (2) by the change in mover: from a fo-
cused onstage participant (the trajector) to an offstage one (the ad-
dressee), or from a specific mover to a non-specific, generalized one.
3. Grammaticization
Attenuation figures in many cases of grammaticization. Let us start

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

ment through space by the subject OR subjective movement
through time by the conceptualizer]
Losing control: grammaticization, subjectification, and transparency 157
be going to be going 10'

overall scope overa scope
immediate scope immediate scope

space
Im Im
cmJtrl-
Ir
>
I
\\ \, I'
I I
\ \I , I
\\111
\1\IIII
I
I
-
T '<'ft >
©
Figure 5. Subjectification as part of grammaticization
In the initial sense, the trajector follows a spatial path, at the end
of which he intends to initiate some activity, which constitutes a re-
lationallandmark. (This landmark is specified by the infinitival com-
plement, and the trajector by the subject noun phrase, at higher le-
vels of grammatical organization.) The movement of course takes
place through time (the arrow labeled t). In the future sense of be go-
ing to, the conceptualizer traces a mental path along the temporal
axis and situates the infinitival event downstream in the flow of time
relative to some reference point. As we saw before with across, there
is no change in trajector. The trajector does however have a dimin-
ished role in the profiled relationship: since it no longer moves
through space, its activity is confined to whatever it does in the land-
mark event. While this may seem peculiar, it is actually quite com-
mon (see Langacker 1995a). Moreover, it is unproblematic in cogni-
tive grammar because the trajector is characterized as primary focal
participant, not in terms of any particular semantic role.
This clearly qualifies as subjectification as defined above. In fol-

lowing the subject's movement through space, which unfolds
through time, the conceptualizer implicitly traces a mental path
through time - the same path which stands alone as the profiled tem-
poral configuration in the future sense of be going to. The conceptu-
alizer's subjective motion through time is immanent in the concep-
tion of the subject's objective motion through space, and remains
behind when the latter fades away. Once more, the heavy dashed
arrow in the second diagram of Figure 5 indicates that the relation-
ship is not purely subjective; some vestiges remain of the original
objective basis for the conceptualizer's mental path. We can discern
a number of intennediate stages:
(4) a. Sam was going to mail the letter - but he never reached the
post office.
b. Sam was going to mail the letter - but he never got around
to it.
c. IfSam isn't careful he's going to fall off that ladder.
d. Something bad is going to happen - I just know it.
e. It's going to be summer before we know it.
In the physical motion sense of (4)a, the subject does not just
move but also has the intention to carry out the infinitival event at
the end of the spatial path. This intention perseveres in many non-
motion uses, as seen in (4)b. The trajector thus continues to be a lo-
cus of activity tending toward realization of the infinitival process.
The activity is however attenuated by virtue of having lost its physi-
cal component; only its mental aspect remains. We can regard this as
either a change in focus (an extreme case, resulting in the full ab-
sence of physical motion) or else a change in domain (from physical
to mental/experiential).
Intention is a sort of potency directed toward realizing the envis-
aged event. One prevalent kind of attenuation involves progressive
diffusion in the locus of potency. We see this in examples like (4)c,
where the future event is conceived as being accidental; the subject
does not act with the specific intent of bringing it about. He is none-
theless still plausibly attributed a certain amount of responsibility for

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

(4)d-e where the subject no longer has any role in bringing about the
infinitival event - we have the situation I call transparency: anything
eligible to be the subject of the infinitival complement is also eli-

gible to be the subject of the entire expression, as seen by the so-
called "dummy" or "expletive" subjects in (5). This is the kind of
situation that was dealt with in transformational grammar by positing
a rule of Subject-to-Subject Raising.
(5) a. There is going to be another storm tonight.

b. Tabs are going to be kept on all the dissidents.
I have argued elsewhere (1995a) that raising rules are superfluous

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

and shouting at them. [source of volitional physical action]
b. Edward frightened the other hunters by accidentally firing

his rifle. [source of non-volitional physical action]
c. Edward frightened the priest by believing in satan. [locus
of mental attitude]
d. Edward frightened the children by being so ugly. [locus of
property reacted to by others]
e. Edward frightened his parents by not being among the chil-
dren getting off the bus. [mere association with circum-
stance reacted to by others]
Edward has a fully agentive role in (6)a, but in (6)b his agentivity
is attenuated through absence of volitionality. In (6)c Edward's par-
ticipation shifts from the physical to the mental domain, and changes
from a temporally bounded action to a steady-state attitude. Also, the
person being frightened now has a greater proportion of the respon-
sibility for the fright being induced; holding a belief is not frighten-
ing per se, but only if someone finds its content objectionable. In
(6)d-e the subject's responsibility diminishes still further, while the
object's increases. In (6)d, Edward is merely the passive and unwill-
ing locus of a static property. And in (6)e, even that is lacking - it is
only due to someone else's expectations that Edward is even associ-
ated with the fright-inducing circumstance.
This is not yet a case of full transparency. One cannot, for exam-
ple, say (7):
(7) *Tabs frightened civil libertarians by being kept on all the

dissidents.
Thus even in (6)e there is some tenuous respect in which Edward

is still held responsible for inducing the fright. But it may be quite
tenuous indeed. This illustrates the great flexibility we usually have
in construing agentivity or responsibility; the individual put in focus
as trajector or subject may in fact be only metonymically related to
the actual locus of potency (Langacker 1995a). The full transparency
characteristic of highly grammaticized forms (like be going to, as
162 Ronald W. LangacJcer
seen in (5)) merely carries the diffusion of responsibility to its ulti-

mate conclusion.
The verb have is an especially interesting example but can be
mentioned here only in passing (cf. Brugman 1988). As a main verb
taking NP complements, have shows a broad range of uses with
varying degrees of attenuation in subject control:
(8) a. Be careful- he has a knife! [source of immediate physical

control]
b. I have an electric saw (but I seldom use it). [source of
potential physical control]
c. They have a good income from investments. [locus of ex-
perience, abstract control]
d. They have three children. [locus of social interaction, gen-
eralized responsibility]
e. He has terrible migraine headaches. [passive locus of ex-
perience]
f. We have some vast open areas in the United States. [loca-
tional reference point, diffuse locus of potential experience]
Yet some vestige of subject involvement always remains. Even in

examples like (8)f (analyzed in Langacker 1993), where a general-
ized subject functions mainly as a locative reference point, there is
still a vague notion of possible experiential consequences. Hence the
main verb have is non-transparent. The auxiliary verb have in the
perfect construction is fully transparent, however:
(9) a. There may have been a serious breach ofsecurity.

b. Tabs should have been kept on those dissidents all along.
This construction represents a considerably more advanced stage

of grammaticization and no longer implies any necessary subject in-
volvement. (For detailed analysis, see Langacker 1990b, 1991: 211-
225; Carey 1994.)
Losing control: grammaticization, subjectification, and tramparency 163
5. Modals
The English modals are likewise highly grammaticized and for the
most part transparent, as seen in (9). They have evolved from main
verbs with meanings like 'want', 'know how to', 'have the strength
to', etc. Observe that such verbs have two crucial properties both re-
flected in grammaticized modals: they are force-dynamic (Sweetser
1982, 1990; Talmy 1988), and the action serving as target of the
force vector, i.e. the event expressed by the verb's complement, re-
mains potential rather than being actual. The profiled relationship
involves some kind of effectiveness or potency tending toward reali-
zation of the type of action expressed by the complement, but no ac-
tual instantiation of that action is implied. I cannot go into the details
of either the synchronic analysis of the modals or their diachronic
evolution (see Langacker 1990b, 1991: 269-281). Here I will simply
note that their development illustrates the attenuation of subject con-
trol, in that the locus of potency is no longer identified with the sub-
ject.
Consider first the root or deontic interpretations of modals. As
noted by Talmy and Sweetser, root modals generally convey force-
dynamic relationships in the domain of social interaction. This shift
from physical to social force constitutes attenuation in regard to do-
main. Moreover, the source of potency is no longer identified with
the subject, but is implicit and subjectively construed. It may be the
speaker but need not be, as seen in (10). It is not necessarily any
specific individual, but may instead be some nebulous, generalized
authority. In other words, the source of potency is highly diffuse.
(10) a. You may not see that movie - I won't allow it!
b. You must go home right away - your wife insists!
c. Passengers should arrive at the airport two hours before
their flight.
d. You must not covet your neighbor's wife.
Nor is the subject necessarily the target of the potency, which is

also diffuse. Although the modal force may be directed at a specific
individual- be it the subject, the addressee, or some third party - we

see from (11) that this is not always the case. The force is simply
directed toward realization of the target event, to be apprehended by
anyone who might be in a position to respond to it. Because the sub-
ject is usually not the source of potency, and need not be its target,
the root modals exhibit transparency.
(11) a. The next patient can come in now.

b. This fence must be painted by tomorrow - you had better
get busy.
c. This fence must be painted by tomorrow - you had better
tell Harry.
d. There may not be any alcohol served at the party.
e. Tabs must be kept on those dissidents!
What about epistemic modals, as in (12)?
(12) a. They should be able to find what they need

b. Tabs will probably be kept on all the dissidents.
c. There may be some rain tonight.
Epistemic modals are maximally diffuse in regard to the source

and. target of potency, hence transparent. I have described their po-
tency as inhering in the evolutionary momentum of reality itself, as
assessed by the speaker/conceptualizer: given how reality has been
evolving up through the present, what is the likelihood of it continu-
ing to evolve in such a way as to "reach" the target process? This is
basically equivalent to saying that the conceptualizer carries out a
mental extrapolation of ongoing reality, projecting into the future,
and senses the degree of force impelling this mental extrapolation in
the envisaged direction, or the degree of resistance encountered in
projecting it through to the target. In other words, the force dynamics
are inherent in the conceptualizer's mental activity, hence sub-
jectively construed in a strong sense.
Losing control: grammaticization, subjectijication, and transparency 165
6. Be-Auxiliaries
Let us next consider the frequent path of grammaticization whereby

verbs of motion or posture evolve into auxiliary verbs roughly gloss-
ed as 'be'. A case in point is Spanish estar, whose etymological val-
ue was 'stand'. Like English be, it takes various types of relational
expressions as complements, including adjectives, prepositional
phrases, and active participles:
(13) a. Esta enformo. 'He is ill.'

b. Esta en la cocina. 'He is in the kitchen.'
c. Esta trabajando. 'He is working.'
What exactly is the meaning of an auxiliary verb like Spanish

estar or English be? Elsewhere (1987a, 1987b) I have argued that the
characteristic property of a verb or a finite clause - the property that
distinguishes them from non-verbal relations like adjectives, prepo-
sitions, and participles - is a subjective factor, namely whether the
conceptualizer follows the profiled relationship'S temporal evolution
sequentially or construes it in summary or holistic fashion. A be-type
auxiliary verb embodies this subjective factor, but does not specify
any particular relationship for its profile, i.e. the objectively con-
strued relationship it follows through time is highly schematic. Be-
type verbs can thus be used in constructions like those in (13), in
which they combine with a non-verbal element to form the complex
head of a finite clause. When the schematic but sequentially viewed
relationship profiled by the auxiliary is equated with the specific re-
lationship profiled by the non-verbal complement, it lends to the
complement the sequential viewing required for the head of a finite
clause.
The historical evolution leading from a motion or posture verb to
a be-type auxiliary clearly involves attenuation. Let us consider it in
regard to a participial construction, as in (13)c. The initial configura-
tion is shown in the first diagram of Figure 6. In this construction,
the onstage profiled relationship is that of the trajector maintaining a
certain posture. Since 'stand' is a verb, the relationship is scanned
sequentially by the conceptualizer (note the arrow labeled T, for pro-

cessing time). The import of using 'stand' in this participial con-
struction is that maintaining the posture accompanies and perhaps
even renders possible another activity, carried out at the same time
by the same individual. This is shown diagrammatically by the dot-
ted correspondence line equating the two trajectors, as well as the
double arrow, which indicates that maintaining the posture enables
the participial activity to occur. The resulting expressions are rough-
ly comparable to (14)a. Expressed by the participle, the accompany-
ing activity is offstage and unprofiled - (14)a profiles the standing,
not the looking - and since sequential viewing requires a high degree
of focus, the offstage activity is viewed holistically, in summary
fashion. (This holistic construal is one semantic contribution of the
participial inflection; 'another - ignored here - is the "internal per-
spective" imposed by a progressive construction [Langacker
1995b].)
(14) a. He stood there looking over the fence.

b. The clock stood ticldng on the table.
c. The cup was leaking.
d. It was raining.
e. Tabs were being kept on all the dissidents.
The second diagram in Figure 6 represents attenuation with re-

spect to both the nature of the profiled process and its role in effect-
ing the accompanying activity. In (14)b, for instance, stand is at-
tenuated due to being predicated of an inanimate subject. Whereas a
person exerts muscular control in order to maintain a vertical orien-
tation, a clock stands passively, merely by virtue of having a certain
shape. The effective relationship between stand and the participial
activity is also greatly attenuated; presumably the clock will tick in
any orientation.
stand + PARTICIPLE stand' + PARTICIPLE

Qr 11
> immediate >
"
t
Q-~1fpe
t \ I I
\ I I
\ /
I
I.
T
be + PARTICIPLE
overall scope
be + PARTICIPLE
overall scope
immediate
tr scope
t~
: immediate
\ '/
T
T
Figure 6. Evolution of a be-type auxiliary verb
Carried to its extreme, attenuation of subject control erases even

the orientational specifications of a verb like stand. Nothing then
remains of any specificity - all that is left onstage is the highly sche-
matic notion of the trajector being involved in some wholly unspeci-

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

passive participial complement. In (15), we find a series of expres-
sions representing an evolutionary path not unlike the one leading to
the progressive be. In each example, get is a finite verb, hence it
profiles a process viewed sequentially, whereas the passive particip-
Losing control: grammaticization. subjectification. and transparency 169
ial complement is non-finite and viewed holistically. The evolution

of get involves progressive attenuation in both the nature of the pro-
filed relationship and the degree of control exercised by its subject.
The result is that get comes close to being just a passive auxiliary,
like be, serving only to provide the sequential viewing required for
the head of a finite clause. It has not gone quite that far, however.
(15) a. Sue got (herself) appointed to the governing board.

b. Ralph got fired again.
c. All my books got stolen.
d. Another bank got robbed last night.
In all the examples, the main clause subject - the trajector of the
finite verb get - is also the trajector of the complement. And since
the complement is based on a passive participle, its trajector is the
same element which functions as landmark (or patient) of the verb
stem from which the participle derives. Our interest, though, lies in
the trajector's role in the main clause relationship. In (IS)a, the tra-
jector and subject (Sue) is construed as a volitional agent who man-
ages to bring about the participial event, and who also, secondarily,
is an experiencer who enjoys the benefits of its occurrence. This vo-
litional construal is almost necessary with the reflexive herself, but it
is at least possible in the simpler construction without the reflexive.
That is our starting point, sketched in Figure 7(ar
170 Ronald W. Langaclcer
(a) get (b) gel'

Im Im
tr
O·~ 'CF-Q
(c) get"
Im
tr
0+ 0-0
Figure 7. Attenuation in the get-passive
More typical are examples like (15)b, where the subject is not a
volitional agent with respect to the participial complement. The sub-
ject may still be responsible in some way - Ralph may have been
fired because of what he did or did not do, or simply because his
boss recognized certain undesirable qualities in him - but not neces-
sarily. He is still however an experiencer in regard to the participial
event, as shown by the dashed arrow in Figure 7(b). But under any of
these construals, the subject's primary role is that of passive under-
goer of the complement event. Its external, specifically main clause
role has been attenuated and now ranges from some indeterminate
kind of responsibility to merely that of an experiencer.
If the main clause subject in (15)b is the locus of experience, this
cannot be true in (15)c, where the subject is inanimate. We have fur-
ther attenuation in the degree of subject control, for the experiencer
is no longer the subject per se, but rather an individual associated
with the subject, the possessor in this example. Moreover, even fur-
ther diffusion is possible, as the locus of the experience need not be
overtly specified or clearly delimited. Thus (15)d does not imply any
particular experiencer, which may be construed as the owners or em-
ployees of the bank, or perhaps just lawful members of society who
feel menaced by the rising level of crime. Hence in Figure 7(c) the
experiencer is not identified with the trajector, nor is it put in focus
as a specified, profiled participant.
The passive be in English merely adds sequential viewing to the
participial complement. The passive get may come close to this, but
so long as any vestigial notion of experience remains, the relation-
ship it profiles is not merely a schematic representation of the parti-
cipial relationship. Moreover, get-passives approximate but do not
quite achieve full transparency:
(16) a. ?A lot of headway got made last night.

b. ? ?Tabs got kept on all the dissidents.
c. ?*It got claimed that there are wombats on Venus.
d. *There got claimed to be wombats on Venus.
(Observe that all of these sentences are well formed if get is

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

b. There promises to be a bright future for all of us.
In the former, as in (17)a, the subject has to be capable of voli-

tion, planning, and communicative interaction. By contrast, (17)b
merely expresses the speaker's assessment based on an overall ap-
preciation of the current situation, hence there are no restrictions on
the subject. The overall pattern documented here suggests that the
so-called "raising" constructions are best seen as the limiting case of
"equi" constructions, representing the extreme situation where atten-
uation of subject control (or its counterpart in the object-raising con-
structions) results in transparency (cf. Newman 1981, 1982; Lang-
acker 1995a).
8. Conclusion
Let me conclude by noting some limitations of the work described

here. In articulating some general patterns observable in the evolu-
tion of numerous grammatical constructions, I have naturally been
unable to examine anyone construction in full depth and detail. In
particular, I have not attempted to determine how many distinct con-
ventional values have to be attributed to a particular element or con-
struction at a given historical stage to account for the variety of con-
textual interpretations it displays. I should also emphasize that this
work has not been based on serious historical investigation. In pre-
senting series of examples representing progressive degrees of at-
tenuation, I have not intended to suggest that they necessarily corre-
spond to the actual order of diachronic development, which clearly
has to be established in its own terms. On the positive side, I have
offered a unified account of diverse phenomena usually considered
in isolation from one another. Attenuation in subject control has
been shown to be a pervasive, multifaceted phenomenon that plays a
major role in certain kinds of grammaticization, with important con-
sequences for synchronic analysis and description.
References
Brugman, Claudia
1988 The syntax and semantics of HAVE and its complements. Ph.D. dis-
sertation, Department of Linguistics, University of California,
Berkeley.
Carey, Kathleen
1994 Pragmatics, subjectivity and the grammaticalization of the English
perfect. Ph.D. dissertation, Department of Linguistics, University of
California, San Diego.
Chafe, Wallace
1994 Discourse, Consciousness, and Time: The Flow and Displacement
of Conscious Experience in Speaking and Writing. Chicago /
London: University of Chicago Press.
Farrell, Patrick
1995 Lexical binding. Linguistics 33: 939-980.
Gernsbacher, Morton Ann and David Hargreaves
1992 The privilege of primacy: Experimental data and cognitive explana-
tions. In: Doris L. Payne (ed.), Pragmatics of Word Order Flexibil-
ity, 83-116. (Typological Studies in Language 22.) Amsterdam /
Philadelphia: Benjamins.
Harder, Peter
1996 Functional Semantics: A Theory ofMeaning, Structure and Tense in
English. (Trends in Linguistics. Studies and Monographs 87.) Berlin
/ New York: Mouton de Gruyter.
Heine, Bernd
1992 Grammaticalization chains. Studies in Language 16: 335-368.
1985 Observations and speculations on subjectivity. In: John Haiman
(ed.), Iconicity in Syntax, 109-150. (Typological Studies in
Language 6.) Amsterdam / Philadelphia: Benjamins.
1986 Abstract motion. Proceedings of the Annual Meeting of the Berkeley
Linguistics Society 12: 455-471.
1987a Foundations of Cognitive Grammar. Volume 1, Theoretical Pre-
requisites. Stanford: Stanford University Press.
1987b Nouns and verbs. Language 63: 53-94.
1990a Concept, Image, and Symbol: The Cognitive Basis of Grammar.
Gruyter.
1990b Subjectification. Cognitive Linguistics 1: 5-38.
1991 Foundations of Cognitive Grammar. Volume 2, Descriptive Appli-
cation. Stanford: Stanford University Press.
174 Ronald W Langacker
1992 Prepositions as grammatical(izing) elements. Leuvense Bijdragen

81: 287-309.
1993 Reference-point constructions. Cognitive Linguistics 4: 1-38.
1995a Raising and transparency. Language 71: 1-62.
1995b Viewing in cognition and grammar. In: Philip W. Davis (ed.), Alter-
native Linguistics: Descriptive and Theoretical Modes, 153-212.
(Current Issues in Linguistic Theory 102.) Amsterdam /
1995c A note on the Spanish personal "a". In: Peggy Hashemipour,
Ricardo Maldonado, and Margaret van Naerssen (eds.), Studies in
Language Learning and Spanish Linguistics in Honor of Tracy D.
Terrell, 431-441. New York: McGraw-Hill.
1997 A dynamic account of grammatical function. In: Joan Bybee, John
Haiman and Sandra A. Thompson (eds.), Essays on Language
Function and Language Type Dedicated to T. Givon, 249-273. Am-
sterdam / Philadelphia: Benjamins.
in press Dynamic conceptualization in grammatical structure.
MacWhinney, Brian
1977 Starting points. Language 53: 152-168.
Newman, John
1981 The semantics of raising constructions. Ph.D. dissertation, Depart-
ment of Linguistics, University of California, San Diego.
1982 Predicate adjuncts. Australian Journal ofLinguistics 2: 153-166.
Ruwet, Nicolas
1991 Syntax and Human Experience. Chicago / London: University of
Chicago Press. [Ed. and trans. by John Goldsmith.]
Sweetser, Eve E.
1982 Root and epistemic modals: Causality in two worlds. Proceedings of
the Annual Meeting ofthe Berkeley Linguistics Society 8: 484-507.
1988 Grammaticalization and semantic bleaching. Proceedings of the
Annual Meeting of the Berkeley Linguistics Society 14: 389-405.
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects
of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cam-
bridge: Cambridge University Press.
Talmy, Leonard
49-100.
Tomlin, Russell S.
1995 Focal attention, voice, and word order. In: Pamela Downing and
Michael Noonan (eds.), Word Order in Discourse, 517-554.
(Typological Studies in Language 30.) Amsterdam / Philadelphia:
Benjamins.
Traugott, Elizabeth
1986 From polysemy to internal semantic reconstruction. Proceedings of
the Annual Meeting ofthe Berkeley Linguistics Society 12: 539-550.
1989 On the rise of epistemic meanings in English: An example of subjec-
tification in semantic change. Language 65: 31-55.
1993 The conflict promises/threatens to escalate into war. Proceedings of
the Annual Meeting o/the Berkeley Linguistics Society 19: 348-358.
this volume The rhetoric of counter-expectation in semantic change: A study in
subjectification.
Verhagen, Arie
1995 Subjectification, syntax, and communication. In: Dieter Stein and
Susan Wright (eds.), Subjectivity and Subjectivisation: Linguistic
Perspectives, 103-128. Cambridge: Cambridge University Press.
The rhetoric of counter-expectation in semantic
change: a study in subjectification
1. Introduction
My aim in this paper is to exemplify how strategic language use to-

gether with cognitive processes that constrain choices and under-
standing of inferences leads to semasiological change over time. 1 In
particular, I will discuss how "rhetorical uses" of a "linguistic de-
vice" (Kay 1990) such as in fact give rise to changes of a highly reg-
ular kind.
The groundwork for this study was laid in syntactic-semantic in-
vestigations of the development of discourse markers like anyway,
besides, indeed, in/act (Traugott 1995a, in press). In these studies I
identified a structural path of change from:
i) Verb Adverb (VAdv); this is clause-internal, oblique, often a

manner adverb and it typically occurs clause-finally,
to ii) Sentence (or I[nflectional] P[brase] Adverb (lPAdv); this has
scope over the clause and occurs either after Comp[lementi-
zer] or adjacent to the tensed verb,
to iii) Discourse Marker (OM); this occurs outside of Comp and has
scope over the following complex structure. In short:
(1) VAdv> IPAdv > OM
A constructed example to clarify the distinctions is:
(2) a. There is no basis in fact. (VAdv).

b. If in fact (IPAdv) there is no basis in fact (V Adv) / If there
is in fact (IPAdv) no basis in fact. (V Adv).
178 Elizabeth Closs Traugott
c. In fact (DM) if in fact (IPAdv) there is no basis in fact.

(VAdv).
As (2) shows, old and new uses can (and typically do) coexist
synchronically in a relationship known as "layering" (Hopper 1991).
My main purpose was to stress that:
a) Semantic change is regular not only in well-known domains
such as space > time, deontic > epistemic, but also in other domains
like manner or spatial adverbial> discourse marker. Evidence for the
latter is provided by Hanson (1987) on modal adverbs like probably,
possibly, Powell (1992a) on stance adverbs like actually, generally,
loosely, precisely, Brinton (1990, 1996) on a variety of pragmatic
markers including Old English hWa!t 'what!, listen up!', anon, and
bifel, gelamp 'happen' constructions, and Jucker (1997) on well.
b) Recruitment to the new syntactic position2 was not possible
until a prior semantic change had occurred permitting use of the
form in that position (see also Bybee, Perkins and Pagliuca 1994 on
evidence that semantic leads syntactic change).
c) The changes should be considered cases of grammaticalization
(see also Onodera 1995, Brinton 1996) although they do not fulfill
the usual criterion of increased bonding or syntactic scope reduction
(see Lehmann 1995 [1982]).
From a cognitive semantic point of view, what is striking about
the adverbs is that both contemporary and historical examples, espe-
cially in their JP Adv use, exhibit characteristics of the rhetorical stra-
tegy often called "counter-expectation". Furthermore, they are typi-
cally used in contexts where counter-expectation is redundantly
marked, e.g. modals, or adversatives like although. When a speaker
expresses counter-expectation, he or she expresses beliefs or points
of view contrary to his or her own or the interlocutor's expectations
regarding the states of affairs under discussion (see Konig 1986,
1991; Heine, Claudi and Hiinnemeyer 1991; Fraser and Malamud-
Makowski 1996; Schwenter 1997; also, from a somewhat different
perspective, Anscombre and Ducrot 1989). 3 This can provisionally
be summarized as: 4
The rhetoric ofcounter-expectation in semantic change 179
(3) V2 is not expected given VI.
Counter-expectation is a matter of point of view and is therefore

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

b. Frankly. Bill spoke.
In (4a) subjectivity is minimal, but not absent: past tense is an-

chored in the speaker, as is choice of unmarked topic structure, etc.
By contrast frankly in (4b) explicitly draws attention to speaker's
subjective point of view or stance to what is being said.
While subjectivity is pervasive in language use, only a subset of
elements in language serve to express it explicitly, and often only in
certain constructions (e.g. frankly clause-initially in (4b) but not
clause-finally as in (4a)). These elements come to do so through the
process of subjectification which I characterize as follows: If the
meaning of a lexical item or construction is grounded in the socio-
physical world of reference, it is likely that over time speakers will
develop polysemies that are grounded in the speaker's world, wheth-
er reasoning, belief, or metatextual attitude to the discourse. Sub-
jectification, then is the semasiological development of meanings
associated with a form such that it comes to mark subjectivity expli-
citly.s I return to this point in section 5, where I also discuss the
relationship of this view of subjectification to Langacker's.
2. The distinction between JP Advs and DMs
There is currently far from common agreement about how to distin-

guish adverbs, despite extensive work on syntactic and semantic dis-
tinctions by, among others, Greenbaum (1969), lackendoff (1972),
Allerton and Cruttendon (1974), Emst (1984), Quirk et al. (1985),
and McCawley (1988). Most contemporary analyses of adverbs re-
cognize a multiplicity of syntactic positions for adverbs in English,
but make a major distinction only between what I am here calling
VAdvs (adverbs of direction, manner, etc. that are oblique argu-
ments of the verb), and one other. This other class of adverbs is usu-
ally characterized as "modal", "evidential", "stance" (e.g. Biber and
Finegan 1988), or as "pragmatic" (e.g. Fraser 1996). Distinctions in
meaning within either class are treated as homonymous (Fraser) or
unitary and monosemous, with variations on a core meaning depen-
dent on context (Sperber and Wilson 1995). Here I distinguish
IP Adv and DM on grounds not only of the distributional differences
sketched in Section 1 but also of pragmatic differences to be outlined
below. Furthermore, I regard the distinctions as polysemous; poly-
semy may be pragmatic (see Horn 1985; Sweetser 1990) as well as
semantic. Methodologically and theoretically it is not possible to in-
vestigate semantic change on the assumption of homonymy, because
if the forms are homonymous they are by definition not related, and
therefore no historical connection between uses can be postulated. It
is also impossible to investigate them on the assumption of mono-
semy since the meanings that license new contextual uses cannot be
accounted for sufficiently explicitly.
DMs have been a major topic in pragmatics and discourse analy-
sis since the publication of Schiffrin's groundbreaking (1987) book
entitled Discourse Markers, in which she analyzed a variety of prag-
matic markers such as y'lcnow, I mean, but, oh, then. More recently
there has been a tendency to bifurcate the domain into:
a) those markers that signal relationships between clauses;
Schiffrin (1990) terms them "discourse deictics", while Fraser
(1988, 1996) calls them "discourse markers",6
The rhetoric of counter-expectation in semantic change 181
b) other pragmatic markers that serve primarily interpersonal

functions, such as hedging, or turn-taking cues, e.g. well,
y'lcnow.
DMs are here taken to be adverbs that: "signal a comment speci-
fying the type of sequential discourse relationship that holds between
the current utterance ... and the prior discourse" (Fraser 1988: 21-
22). They are "metatextual,,7 in that their function is to comment on
the text being constructed; they are also "procedural"s in the sense
that they are primarily pragmatic and non-truth-conditional. Being
deictic, they "point to" the speaker's metatextual attitudes, and also
guide hearers to those attitudes.
3. A synchronic account of in fact
3.1. An earlier analysis
Fraser cites in fact in several of his writings on DMs, but classifies it

in different ways. For example, Fraser (1988) lists in fact under two
kinds of DM function (topic resumption and elaboration) while Fra-
ser (1996) lists it only once (elaboration). Fraser and Malamud-
Makowski (1996: 871) list it under contrastive DMs and say that it:
"signals that U2 is to be interpreted as a denial of the explicit propo-
sition asserted with U I, with the added condition that the assertion of
VI is not the direct responsibility of the speaker ofV2"."
Among their examples of in fact are:
(5) a. I thought the rock I found was granite. In tact. the entire
quarry was quartz.
b. A. The paint is purple.
B. In fact, it's mauve.
(Fraser and Malamud-Makowski 1996: 871-872)
Being clause-initial in (5), and without a complementizer, con-

trastive in fact occurs in the same surface position as elaborative in
fact, as can be seen by constructing (5b) as a monologue:
(5) c. The paint is purple. In fact it's mauve.
Here in fact would probably be construed not as a denial of the

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

tically distinct uses of in fact.
First, there is the VAdv, which is clause-final and, in (6), con-
trasts with infeeling (other contrasts in the data include in law, and
sensationalism ).
(6) Humanity, comfortably engaged elsewhere in the business of

living, is absent in fact but everywhere present in feeling.
(United Airlines Magazine, 5/1997)
A second type, the IPAdv, is illustrated by (7), where in fact

follows the finite verb and contrasts with e.g. however:
(7) It's almost as if he were avoiding the evening's moment of

truth. The moment is, in fact, about 45 minutes, the main body
of the play, during which Mr. Quinton ... holds the stage and
acts up a primeval cataclysm all by himself.
(New York Times, 7/29/1996, review of Phaedra)
In U 1 the frame It is as if, the scalar almost, and the semantically

negative avoiding all serve to warn that U 1 is problematic, dissonant,
and contrary to the speaker's view. This dissonance is further
highlighted by the reorientation of "moment" from something nor-
mally thought to be very short to a lengthy 45 minutes. Examples
like (7) show that U 1 may indeed be the responsibility of the speaker
- speakers set up false scenarios and then show that the assumptions
manifest in them are wrong, inappropriate to the occasion, etc. We
may characterize this as follows:
(8) IPAdv infact signals that U2 is to be interpreted as a denial of

one or more manifest assumptions in U 1.
The third (DM) use of in fact typically occurs clause-initially; it

may also occur after a finite verb. Unlike IPAdv infact, it is used in
a semantic set with in other words, I mean, but not however. A typi-
cal example is:
(9) Polling isn't The World's Oldest Profession. although around

election time it might seem like it. In fact, once upon a time,
way back in the first third of this benighted century, modern
polling wasn't yet even a gleam in the eye of a small-town
Iowa kid named George Gallup. Today ...
(United Press International, 10/8/1990)
As (9) shows, the main function of DM in fact is elaboration. Yet

it does also signal some level of contrast. As in (9), it is often used
after a negative statement. It serves to confirm the inference from
that negative statement (which itself counters a prior expectation),
and to provide an elaboration which is a clarification or reformula-
tion of the negative statement. Here the elaboration is presented as a
justification of the denial in the preceding sentence. The DM use of
infact can be characterized as:
(10) DM infact signals that U2 is an elaboration ofUI, and a more

precise formulation than U I.
4. A brief history of in fact
In fact has its historical beginnings in the early sixteenth century. In

the earliest uses we fmd fact as a lexical noun (Stage 0, the initial
stage before change has set in). My first example of it is in a prepo-
sitional phrase is:
(11) Stage 0: Full Lexical N

A sort of naughty persons ... ,
Have practised dangerously against your state,
Dealing with witches and with conjurors,
Whom we have apprehended in the ract,
Raising up wicked spirits from under ground.
(c1592 Shakespeare, 2 Henry VI,ILi. 163-170)
Fact begins to appear in the Helsinki Corpus by the late 17thC in

a bare prepositional phrase as the object of in (Stage I, the initial
construction which later split into V Adv and IP Adv). Here it is fa-
vored in coordinate constructions, either as a member of a taxono-
mically contrasted set as in (12a) or in an explicitly contrastive con-
text like (12b):
(12) Stage I: VAdv

a. But it is evident in fact and experience that there is no such
universal Judge, appointed by God over the whole World,
to decide all Cases of temporal Right.
(1671 Tillotson, Sermons, p.II.ii.445 [HC])
b. This company, therefore, have always enjoyed an exclusive
trade in {act, though they may have no right to it in law.
(1776 Smith, Wealth ofNations, Bk 5, p.701)
In these kinds of contexts, in fact came to be endowed with evi-

dential meanings ('in practice/reality/actuality') that assess some-
thing as physically accessible, empirically attested, real and true.
(12a) cites fact (action, performance) and experience (as opposed to
speculation) as sources of evidence, (12b) cites it as the pragmatic
and experiential locus of daily transactions, as opposed to abstract
legal right. Such uses, as well as independent developments in the
philosophical and scientific construals of 'fact', are likely to have
contributed to the recruitment of in fact to the class of adverbs that
are potentially epistemic modals, specifically those that signal epi-
stemic certainty.
By the mid-18thC in fact is used as an IP Adv in clause-initial
post-complement position. Often it occurs in the already contrastive
environment of but as in (13a) or in highly rhetorical contexts, as in
(13b) which is a rhetorical question, followed by a denial:
(13) Stage 11: IPAdv

a. Every particle eludes the grasp by a new fraction, like
quicksilver, when we endeavour to seize it. But as in fact
here must be something which terminates the idea of every
finite quantity ...
(1739-40 Hume, Treatise Human Nature, Bk. 1, p. 44 )
b. If this be their true interest, have they in fact pursued it?
Has it not, on the contrary, invariably been found, that ... ?
(1787 Federalist Papers, p. 31)
186 Elizabeth Closs Traugon
As these examples indicate, the IP Adv use points to the truth of p

despite prior contrary expectations regarding it.
The first example in my data of the third, DM, stage, occurs in a
novel by Jane Austen, in dialogue. In the new use, in fact functions
at the discourse level to express the speaker's attitude to the appro-
priateness of the discourse itself. In this sense it serves a metatextual
function and is entirely procedural. In (14a) in fact introduces justi-
fication of what has just been said, in other words, self-corrective
elaboration which is contrastive not in terms of truth but of appro-
priateness of expression. Similarly in (14b) infact signals that what
follows strengthens and specifies what precedes (lead inevitably to):
(14) Stage Ill: DM

a. "I was wrong, .. he continued, "in talking of its being broke
to you. I should not have used the expression. In fact, it
does not concern you - it concerns only myself.
(1815 Austen, Emma, Vol. 3, Chap. 10, p. 393)
b. Thus in various ways ethical questions lead inevitably to
psychological discussions; in fact, we may say that all
important ethical notions are also psychological.
(1874 Sidgwick, History of Ethics, Chap. 1, p. 4)
Note that in both instances the context explicitly refers to lan-

guage use (expression, discussion).
The data suggest that the epistemic meaning of in fact originally
arose in contexts of lexical contrast, specifically as an alternate in
lexical fields concerning human action as it pertains to judgment,
whether divine or human, but in any case where action is taken as
evidence. In these contexts in fact is used to refer to physical acts as
contrasted with thought, belief, etc. Once understood as epistemic, it
could be recruited to the class of epistemic JPAdvs. Initially, this re-
cruitment takes place in contexts that render in fact redundant, such
as contrastive but in (13a), or in the question and conditional in
(13b). Here it eventually takes on the force of contrast and can be
understood as adversative in less explicitly adversative contexts, and
eventually even out of context.
5. Implications for historical semantics
Cognitive linguists have tended to favor a metaphorical view of se-

mantic change (see Sweetser 1990; Heine, Claudi, and HOnnemeyer
1991), though recently there has been growing interest in conceptual
metonymic change, also known as change via context-induced infe-
rencing (see Traugott and Konig 1991). Ifwe consider the history of
in fact, a metaphorical interpretation is not ruled out at the earliest
stage. On the metaphorical view, physical deeds come to be under-
stood as mapped onto anything that is observable. On the other hand,
a metonymic inference from doings is that they are observable be-
cause they are actions (not thoughts, opinions, etc.). Good metony-
mies are often metaphorical as well, so there is no difficulty in look-
ing at the change either way. However, when we consider later
changes involving the use of in fact to signal counter-expectation at
the propositional level and then metatextual rejection of earlier for-
mulations, etc., an account from metaphor becomes less plausible.
An interpretation based on processes of inferencing from the con-
text, i.e., from the linear progression of utterances and the redUndan-
cies in them, is required, together with processes of subjectification
that lead speakers to semanticize (conventionalize) those inferences
that saliently express their attitude. 9
Among semantic changes, subjectification has been of special in-
terest. Although already mentioned in Breal's work (1964 [1900]:
Chapter 25), SUbjectivity did not come to be of major topic of re-
search until Benveniste raised the question whether a language could
still function and be called a language unless it were "marked so
deeply by the expression of subjectivity" (1971: 225), and distin-
guished between the syntactic subjectl"sujet d'enonce" and the
"speaking subject"I"sujet d'enonciation" (see also Lyons 1982,
Langacker 1985, this volume). A historical view of subjectification
has been developed by e.g. Traugott (1982,10 1989, 1995b), Lang-
acker (1990), papers in Stein and Wright (1995).
Although Langacker's and my own views on the ubiquity of sub-
jectivity coincide, we are using the term "subjectification" to refer to
considerably different phenomena, and from different perspectives
188 Elizabeth Closs Traugotl
on theory and methodology. The differences are therefore not "pure-

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

like must. Other prime examples are the development of perfonna-
tive uses of locutionary verbs such as promise, recognize, insist (ul-
timately derived from spatial and mental tenns), and especially of
the IP Adv and OM meanings of adverbs like in fact discussed here.
Thus:
VAdv > IPAdv > DM

Least subjective ---------i.. Most subjective
Figure 1. Adverbials: Path of subjectification
While the first stage (V Adv > IPAdv) might conceivably be con-
strued as a case of "loss of control" by the subject of the event-struc-
ture that VAdv modifies, the second (IP Adv > OM) cannot. In both
the IPAdv and OM cases in fact is grounded in the speaker's
perspective. The scope of that perspective is the clause in the case of
the IPAdv, of the relationship between successive discourse units in
the case of the OM. Similarly, subjectification effects in morpho-
logical domains (see e.g. Company 1993) cannot be identified with
loss of control.
Subjectification involves speakers recruiting fonns with appro-
priate meanings to externalize their SUbjective point of view. This is
an activity that draws on cognitive principles but takes place in the
context of conununication and rhetorical strategizing. As Stubbs
(1986) has pointed out: "whenever speakers (or writers) say any-
thing, they encode their point of view towards it: whether they think
it is a reasonable thing to say, or might be found to be obvious,
irrelevant, impolite, or whatever" (Stubbs 1986: 1). Stubbs's key
word here is "encode".
It has been suggested (Oiewald 1993, Keller 1995) that subjecti-
fication defined as preemption of old meanings to encode and ex-
ternalize speaker subjectivity is really objectification. From the point
of view developed here, which focuses on speakers' use of old struc-
tures with new functions, such a suggestion confuses objectivity
with making explicit and manifest features of the discourse situation.
The particular semantic change undergone by in fact discussed in

this paper is an example of the development of a class of elements
that Talmy has called "force-dynamic logic-gaters [that] limn out the
rhetorical framework, ... direct the illocutionary flow, ... specify the
logical tissue" (Talmy 1988: 88-89). If in fact is indeed a "logic-
gater", it might seem to be objective. However, as I have shown, it is
highly subjective, expressing speaker's alternative points of view to
some prior discourse in the case of the IP Adv, and to the discourse
being negotiated in the case of the DM. In neither use does it add
anything to the truth-value of what is said, or express a logical rela-
tionship. However, because it has the appearance ofa logical and ob-
jective marker, it can be said to "locate the claim ... in the commun-
ally available realm of intersubjectivity" and therefore is partially
intersubjective (Powell 1992b: 349), although its prime function is
subjective.
Does objectification then occur? Indeed. In semantic change it is
the highly conscious and deliberately interventive sort of change that
comes about when ordinary words are preempted for technical or le-
gal purposes, e.g. linguists' use of "competence", "relevance",
"quantity", or the redefinition of the word "harassment" following
on the V.S. Civil Rights Act of 1964. But these are cases of
"external", "non-natural" change, not of "internal", "natural" change,
which is the central concern of historical cognitive semantics.
Notes
1. Thanks to Lisbeth Lipari, whose work on/act in 1994 provided the original
impetus for work on this construction, and especially to Scott Schwenter,
for numerous comments, bibliographical suggestions, and inspiration.
The main historical data base is The Helsinki Corpus 0/ the English Lan-
guage (HC); (see e.g. Rissanen et al. 1993). Stanford Academic Text Ser-
vices made access to this and a variety of other computerized corpora pos-
sible.
2. The position is "new" with respect to the construction not the grammar.
Both the IPAdv and a position related to OM (left-most "Topic") have been
available from Indo-European times (see Kiparsky 1995).
3. Another way to conceptualize this is as a ''polyphony'' (see Nslke 1992,

working in the tradition of Anscombre and Oucrot 1989). For example, in
connection with negation, N0lke says not "induces (sic) a polyphonic struc-
ture into the sentence consisting of two (incompatible) viewpoints" (N0lke
1992: 190; italics original).
4. U = utterance. U I and U2 are not necessarily adjacent.
5. As a semasiological change, it concerns the histories of individual lexical
items. It has nothing to do with choice of styles, which may be more "objec-
tive" or "subjective" at different times (Adamson 1995); nor with the
choices made by certain groups for use in certain registers, in renewing the
lexicon or expanding it for new political, academic, legal, medical or other
purposes (Marchello-Nizia, in press). The claim is only that if a lexical item
is available in the system, over time it is likely to acquire more subjective
meanings.
6. The term "discourse marker" has also sometimes been applied to elements
of a conjunctive rather than adverbial type, such as and, but. The latter are,
however, more commonly called "discourse connectives", cf. Blakemore
(1987) on and, but, so, etc. Pons Borderia (1995) discusses the various ter-
minological distinctions.
7. As Dancygier (1992) points out, this term is preferable to "metalinguistic"
for expressions of attitude to the text being constructed; this allows "meta-
linguistic" to be used for corrections, and what Horn (1995) has called
"metalinguistic negation".
8. The term was introduced into relevance-theory by Blakemore (1987) to
characterize pragmatic markers that "guide the inferential phase of compre-
hen.sion" (Sperber and Wilson 1995: 12-13). Despite adopting this term, I
do not endorse the view that procedurals are primarily addressee-oriented.
9. Traugott (in press) presents a model of this process.
10. The term "subjectification" is not used in Traugott (1982), but the "expres-
sive" tendency of language identified there is roughly equivalent to subjec-
tification as defmed here.
References
Adamson, Sylvia
1995 From empathetic deixis to empathetic narrative: Stylisation and (de-)
subjectivisation as processes of language change. In: Stein and
Wrigbt (eds.), 195-224.
Allerton, D. J. and Alan Cruttenden
1974 English sentence adverbials: Their syntax and their intonation in
British English. Lingua 34: 1-30.
Anscombre, Jean-Claude and Oswald Ducrot

1989 Argumentativity and informativity. In: Michael Meyer (ed.), From
Metaphysics to Rhetoric, 71-87. (Syntbese Library 202.) Dordrecht:
Kluwer.
Benveniste, Emile
[1958] 1971 De la subjectivite dans le langage. In: id., Problemes de Iinguistique
generale, 258-85. Paris: Gallimard. (trans. by Mary Elizabeth Meek
as Problems in General Linguistics, 224-246. Coral Gables, FL:
University of Miami Press).
Biber, Douglas and Edward Finegan
1988 Adverbial stance types in English. Discourse Processes 11: 1-34.
Blakemore, Diane
1987 Semantic Constraints on Relevance. Oxford: Blackwell.
Breal, Michel
[1900] 1964 Semantics; Studies in the Science of Meaning. New York: Dover
Publications. (trans. by Mrs. Henry Cust).
Brinton, Laurel J.
1990 The development of discourse markers in English. In: Jacek Fisiak
(ed.), Historical Linguistics and Philology, 45-71. (Trends in
Linguistics. Studies and Monographs 46.) Berlin / New York:
Mouton de Gruyter.
1996 Pragmatic Markers in English: Grammaticalization and Discourse
Function. (Topics in English Linguistics 19.) Berlin / New York:
Mouton de Gruyter.
Bybee, Joan, Revere Perkins and William Pagliuca
1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Lan-
guages ofthe World. Chicago: University of Chicago Press.
Company, Concepci6n
1993 Old forms for new concepts: The recategorization of possessive
duplications in Mexican Spanish. In: Henning Andersen (ed.),
Historical Linguistics 1993, 77-92. Amsterdam: Benjamins.
Dancygier, Barbara
1992 Two metatextual operators: Negation and conditionality in English
and Polish. Proceedings of the 18th Meeting of the Berkeley
Linguistics Society: 61-75.
Diewald, Gabriele
1993 Zur Grammatikalisierung der Modalverben im Deutschen. Zeit-
schrift fUr Sprachwissenschajt 12: 218-234.
Emst, Thomas Boyden
1984 Toward an Integrated Theory of Adverb Position in English.
Bloomington: Indiana University Linguistics Club.
Ferrara, Kathleen
1997 Fonn and function of the discourse marker anyway: Implications for
discourse analysis. Linguistics 35: 343-348.
Finegan, Edward
1995 Subjectivity and subjectivisation: An introduction. In: Stein and
Wright (eds.), 1-15.
Fraser, Bruce
1988 Types of English discourse markers. Acta Linguistica Hungarica 38:
19-33.
1996 Pragmatic markers. Pragmatics 6: 167-190.
Fraser, Bruce and M. Malamud-Makowski
1996 English and Spanish contrastive discourse markers. Language
Sciences 18: 863-881.
Greenbaum, Sidney
1969 Studies in English Adverbial Usage. London: Longman.
Hanson, Kristin
1987 On subjectivity and the history of epistemic expressions in English.
Papers from the 23rd Annual Regional Meeting of the Chicago
LinguisticSociety: 133-147.
Heine, Bemd, Ulrike Claudi and Friederike Htlnnemeyer
1991 Grammaticalization: A Conceptual Framework. Chicago: Universi-
ty of Chicago Press.
Hopper, Paul J.
1991 On some principles of grammaticization. In: Traugott and Heine
(eds.), Vol. I, 17-35.
Horn, Laurence R.
1985 Metalinguistic negation and pragmatic ambiguity. Language 61:
121-174.
Jackendoff, Ray
1972 Semantic Interpretation in Generative Grammar. (Studies in Lin-
guistics series 2.) Cambridge, MA: MIT Press.
Jucker, Andreas H. (ed.)
1997 The discourse marker well in the history of English. English Lan-
guage and Linguistics I: 91-110.
Kay, Paul
1990 Even. Linguistics and Philosophy 13: 59-Ill.
Keller, Rudi
1995 The epistemic weil. In: Stein and Wright (eds.), 16-30.
Kiparsky, Paul
1995 Indo-European origins of Gennanic syntax. In: Adrian Battye and
Ian Roberts (eds.), Clause Structure and Language Change, 140-
169. Oxford: Oxford University Press.
K6nig, Ekkehard
1986 Conditionals, concessive conditionals and concessives: Areas of
contrast, overlap and neutralization. In: Elizabeth Closs Traugott,
Alice ter Meulen, Judy Snitzer Reilly and Charles A. Ferguson
(eds.), On Conditionals, 229-246. Cambridge: Cambridge Universi-
ty Press.
1991 The Meaning of Focus Particles: A Comparative Perspective. Lon-
don: Routledge.
1985 Observations and speculations on subjectivity. In: John Haiman
(ed.), lconicity in Syntax. (Typological Studies in Language 6.)
Amsterdam / Philadelphia: Benjamins.
1990 Subjectification. Cognitive Linguistics 1: 5-38.
this volume Losing control: grammaticization, subjectification, and transparency.
Lehmann, Christian
[1982] 1995 Thoughts on Grammaticalization. Munich: Lincom Europa. (First
published as Thoughts on Grammaticalizalion. A Programmatic
Sketch. Vol. I. Arbeiten des K6lner Universalien-Projekts 48. K6ln:
Universitllt zu K6ln. Institut fUr Sprachwissenschaft.)
Lyons, John
1982 Deixis and subjectivity. Loquor, ergo sum? In: Robert J. Jarvella
and Wolfgang Klein (eds.), Speech, Place, and Action: Studies in
Deixis and Related Topics, 101-124. New York: John WHey.
Marchello-Nizia, Christiane
in press Language evolution and semantic representations: From "subjec-
tive" to "objective" in French. In: C. Fuchs and S. Roberts (eds.),
Language, Diversity, and Cognitive Representations. (Human Cog-
nitive Processing.) Amsterdam: Benjamins.
McCawley, James D.
1988 The Syntactic Phenomena of English. Chicago: University of Chi-
cago Press, Vol. 2.
N",lke, Henning
1992 Semantic constraints on argumentation: From polyphonic micro-
structure to argumentative macro-structure. In: Frans H. van
Eemeren (ed.), Argumentation Illuminated, 189-200. Amsterdam:
SICSAT.
Onodera, Noriko Okada
1995 Diachronic analysis of Japanese discourse markers. In: Andreas
Jucker (ed.), Historical Pragmatics, 393-437. (Pragmatics and Be-
yond 35.) Amstcrdam / Philadclphia: Benjamins.
Pons Borderia, Salvador

1995 A prototype approach to the concept of connective. MS, Stanford
University.
Powell, Mava Jo
1992a The systematic development of correlated interpersonal and metal in-
guistic uses in stance adverbs. Cognitive Linguistics 3: 75-110.
1992b Folk theories of meaning and principles of conventionality: Encod-
ing literal attitude via stance adverb. In: Adrienne Lehrer and Eva
Feder Kittay (eds.), Frames, Fields, and Contrasts: New Essays in
Semantic and Lexical Organization, 333-353. Hillsdale, NJ:
Lawrence Erlbaum Assocs.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik
1985 A Comprehensive Grammar of the English Language. London:
Longman.
Rissanen, Matti, Merja KytO, and Minna Palander-Collin (eds.)
1993 Early English in the Computer Age: Explorations through the Hel-
sinki Corpus. (Topics in English Linguistics 11.) Berlin / New York:
Mouton de Gruyter.
Schiffrin, Deborah
1987 Discourse Markers. (Studies in Interactional Sociolinguistics 5.)
1990 Between text and context: Deixis, anaphora, and the meaning of
then. Text 10: 245-70.
Schwenter, Scott A.
1997 The pragmatics of conditional marking: Implicature, scalarity and
exclusivity. Ph.D. dissertation, Stanford University.
Sperber, Dan and Deirdre Wilson
21995 Relevance: Communication and Cognition. Oxford: Blackwell.
Stein, Dieter and Susan Wright (eds.)
1995 Subjectivity and Subjectivisation in Language. Cambridge: Cam-
bridge University Press.
Stubbs, Michael
1986 'A matter of prolonged field work': Notes toward a modal grammar
of English. Applied Linguistics 7: 1-25.
Sweetser, Eve E.
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects
of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cam-
bridge: Cambridge University Press.
Tabor, Whitney
1993 The gradual development of degree modifier sort of and kind of A
corpus proximity model. Papers from the 29th Regional Meeting of
the Chicago Linguistic Society: 451-465.
196 Elizabeth Closs Traugolt
Tabny, Leonard
49-100.
Traugott, Elizabeth Closs
1982 From propositional to textual and expressive meanings: Some
semantic-pragmatic aspects of grammaticalization. In: Winfred P.
Lehmann and Yakov Malkiel (eds.), Directions for Historical
Linguistics: A Symposium, 245-271. Austin: University of Texas
Press.
1989 On the rise of epistemic meanings in English: An example of sub-
jectification in semantic change. Language 57: 33-65.
1995a The role of discourse markers in a theory of grammaticalization. Pa-
per presented at the 12th International Conference on Historical
Linguistics, Manchester.
1995b Subjectification in grammaticalization. In: Stein and Wright (eds.),
31-54.
in press Constructions in grammaticalization. In: Richard Janda and Brian
Joseph (eds.), A Handbook of Historical Linguistics. Oxford:
B1ackweIl.
Traugott, Elizabeth Closs and Bernd Heine (eds.)
1991 Approaches to Grammaticalization. 2 volumes. (Typological studies
in language 19.) Amsterdam / Philadelphia: Benjamins.
Traugott, Elizabeth Closs and Ekkehard KOnig
1991 The semantics-pragmatics of grammaticalization revisited. In:
Traugott and Heine (eds.), Vol. I, 189-218.
Synecdoche as a cognitive
and communicative strategy
Brigitte Nerlich and David D. Clarke
The urge 10 classify is a fundamental

human instincl; like the predisposition to
sin, it accompanies us into the world at
birth and stays with us to the end
(A. Tindell Hopwood)
1. Derming synecdoche
Until quite recently, research in rhetoric, literary theory and

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

rical shifts and changes in the status of synecdoche (for metonymy,
read Bonhomme 1987). We shall provide a short summary of the
main stages here, but it has to be stressed that this overview is not at
all exhaustive:
In Antiquity, namely in the work of Aristotle, synecdoche was
still studied as part of metaphor, or rather, the three main figures of
speech, metaphor, metonymy and synecdoche, were not yet strictly
differentiated.
'Metaphor' in a wider sense
~
metaphor
/~
metonymy - - - - - - - - synecdoche
Figure 1. Antiquity
In classical rhetoric (epistomised perhaps by Pierre Fontanier,

[182111827] 1968) scholars distinguished between three, sometimes
four, main tropes or master tropes: metaphor, metonymy, synec-
doche and irony (cf. also Burke 1969). Synecdoche itself was seen
as subsuming various subtypes. The kernel consisted for a long time
of two subtypes of synecdoches: the part for whole one and the ge-
nus for species one (and vice versa), with the part-whole subtype
being the most stable part of this typology. Classical examples for
the part-whole and genus-species subtypes of synecdoche are: sail
for ship and bread for food.
figures of speech
metaphor metonymy synecdoche
genus/species part! whole
Figure 2. Classical rhetoric
In the process of reducing all tropes to two, especially under the

influence of Roman Jakobson (1956, [1956]1983), that is to meta-
phor (based on similarity) and to metonymy (based on contiguity),
metonymy was seen as including the part-whole type of synecdoche,
and the genus-species type of synecdoche was largely forgotten.
figures of speech
metaphor metonymy
[including part/whole]
Figure 3. Jakobson
Synecdoche was rediscovered by the Groupe J.1 and declared the

master-trope by Tzvetan Todorov (1970) in the 1970s, an enterprise
that was however fraught with difficulties, as a new tenninology on-
ly served to hide the older conception of synecdoche in which it was
dressed up. Todorov for example gave as an example for the type of
200 Brigitte Nerlich and David D. C1arke
synecdoche of the part for the whole the one of voile for bateau,
which is, as we shall see, a metonymy, a point already made by
Michel Le Guern (1973: 29-38), Nobuo Sato (1979), and Adolphe
Nysenholc (1981). The Groupe Il were aware of the distinction be-
tween the part-whole type of synecdoche (exploiting what is now-
adays called partonomies or meronymies; cf. Tversky 1990; Cruse
1986), and the genus-species type of synecdoche (exploiting taxono-
mies), but ended up confusing them again. What the Groupe Il calls
particularising synecdoche, that is the figure of the part for the
whole, should, as we shall see, be regarded as part of metonymy. In
short, whereas synecdoche was absorbed by metonymy in the case of
lakobson, synecdoche absorbed part of metonymy in the case of the
Groupe Il. The discussion in France and else-where surrounding the
Groupe Il' s general rhetoric would deserve a separate article.
Only quite recently (but going back to some pre-lakobsonian
writers such as Nyrop 1913, Esnault 1925 1 and others) the ancient
kernel of synecdoche, it's "noyau stable", has been broken up in a
different way. One constituent (the part-whole subtype) has been at-
tributed to metonymy, whereas the genus-species subtype of the ker-
nel has been preserved to define synecdoche as a third member in a
triplet of essential tropes, namely metaphor, metonymy and synec-
doche. This last option has been most forcefully defended by Ken-
ichi Seto in his paper entitled "Metonymy and the Cognitive Tri-
angle" (Seto, in press a), as well as in his contribution to the Ham-
burg workshop entitled "On Distinguishing Metonymy from Synec-
doche" (Seto, in press b). Armin Burkhardt summarises this view
well when he writes:
Der alte Streit urn die Zuordnung der TeiVGanzes-Beziehung kann begra-
ben werden, sobald einsichtig wird, daB er sich nur einer Aquivokation ver-
dankt - kann doch pars sowohl 'Bestandteil' als auch 'Teilmenge' bedeu-
ten. Nur im letzteren Falle darf die parsltotum-Beziehung als eine quantita-
tive und damit synekdochische betrachtet werden, wahrend die Beziehung
des materiellen Bestandteils zum Ganzen in die Reihe metonymischer
Relationen geh6rt. (8urkbardt ms., 1995: 2-3)
[The old quarrel about the whole-part relation can be forgotten as soon as
we become aware of the fact that it is only based on an equivocation, for
pars can mean 'component' as well as 'subset'. But only in the latter case
should we regard the parsltotum-relation as a quantitative and therefore
synecdochical one, whereas the relation of a material component to the
whole belongs to the set ofmetonymic relations.)
To summarise, one can therefore say that metonymy is based on

qualitative, synecdoche on quantitative relations, that is on set-inclu-
sion. Metonymy is based on our world-knowledge about space and
time, cause and effect, part and whole, whereas synecdoche is based
on our taxonomic or categorical knowledge. Metonymy exploits our
knowledge of how the world is, synecdoche of how it is ordered in
our mind (Nerlich, in press).
This view is to some extent opposed to two other more modem
approaches to metaphor, metonymy and synecdoche, one written in
the French tradition of "general rhetoric", the other against it.
In their 1977 article "Metaphor, Metonymy and Synecdoche
(Re)visited", Peter Schofer and Donald Rice argued that:
Synecdoche is characterized by a semantic or referential relationship of

inclusion made possible by the fact that one of the signifieds is also a se-
mantic feature of the other signified. Whereas we eliminated all but causal
relationships from metonymy, synecdoche is expanded here to include not
only the part for the whole but also the container for the contained. (Scho-
fer & Rice 1977: 141)
However, this opinion can be regarded as a rather isolated

deviation from the view that part-whole synecdoches are metony-
mies, a conception which was by then becoming generally accepted
even in French speaking countries.
In his 1984 article on metonymy, Hugh Bredin distinguished be-
tween structural relations and extrinsic relations and claimed that
synecdochic relations are structural, and metonymic relations are
extrinsic ones (Bredin 1984: 54):
The theory I want to propound is that synecdochic relations are structural,

and metonymical relations are extrinsic - relations, in the one case, be-
tween partiCUlars and their parts, and in the other case between particulars
and other particulars. ... It is of great importance, though, to note that what
is taken to be the particular in any given case is dependent on the context,

and not necessarily upon some inherent nature in things. In other words, the
concept particular, as it is being used here, has a strongly epistemic
character - as perhaps it always has. For example, ''wheel'' is a synecdoche
for an automobile; but if a racing driver is given the nickname "Wheel",
this is a metonymy. In one case the particular is an automobile, and wheels
are part of it, structurally related to the automobile as part to whole. In the
other case, wheels are a particular, and are extrinsically related to the
driver, who is another particular.
We must also note, that, in the synecdochic relation of genus to species, the
concept of the genus is taken as a particular, even though it is a concept.
One reason is that, when we begin to analyze a concept into its constitu-
ents, its having those constituents is the having of certain properties: thus,
mortal has the property of including the extension of the concept man; it
may be said to instantiate the having of that property as a particular instan-
tiates a universal. (Bredin 1984: 54)
Relations
/~
structural extrinsic
sm~~. / \
simple
(metonymy) /
dependent
'\
similarity others
(metaphor)
Figure 4. Bredin
However, we would argue with Burkhardt and Seto that kind-of

relations or taxonomical relations or category relations are one thing,
and part-of relations or partonomical relations or real word relations
are quite another (cf. Seto, in press b). Kind-of relations are
exploited in synecdoche, part-of relations are exploited (amongst
others) in metonymy. Both types of relations are relations between

constituents or components, but in the case of metonomy these re-
lations link occurrences in the "real world", whereas in the case of
synecdoche the relations that are exploited are those which pertain
between 'components' of concepts or categories. The crucial differ-
ence therefore between the classical and the modem conception of
synecdoche ist that part-whole relations as relations in the real world
are seen to be the basis for metonymies and are therefore put on the
same level as relations between cause and effect, container and con-
tained and so on, whereas genus-species relations as relations based
on set-inclusion are considered to be the basis for synecdoches.
figures of speech
including part/whole genus/species
Figure 5. Esnault, Burkhardt, Seto, etc.
2. Derming the communicational function of synecdoche
After having looked at synecdoche from a semasiological point of

view, we would now like to turn to a more onomasiological perspec-
tive.
In the following we would like to argue that metaphor, metony-
my, and synecdoche, thus defined and delimited, can be regarded as
universal cognitive and communicative strategies. We would also
like to argue that all three "figures of speech" can be studied on
three levels of analysis: on the level of synchrony (or panchrony), on
the level of diachrony, and on the level of developmental linguistics
(see figure 6). On the synchronic level, synecdoche can be studied as
204 Brigitte Nerlich and David D. Clarlce
part of the systematic analysis of semantic relations, that is of tax-

onomy or hyponymy (''the lexical relation corresponding to the in-
clusion of one class in another", Cruse 1986: 88), as well as on the
conceptual level and on the level of discourse or actual language
use, to which we shall come shortly (cf. figure 7); on the diachronic
level synecdoche can be studied as bringing about particularisation
and generalisation or restriction and extension of meaning; on the
developmental level synecdoche can be studied as underlying small
children's overextensions (cf. Nerlich, Todd & Clarke, in press).
hyponymy particularisation overextension
Figure 6. The synchronic-diachronic-developmental interface
synecdoche
conceptual communicational
+
(based on categories)
(based on semantic relations) l

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

Now, whereas the function of metaphor in discourse is fairly well

understood, and the function of metonymy is gradually being un-
ravelled, the function of synecdoche in actual discourse is not so
easy to pin down. There is a general consensus that synecdoche is
the least figurative of the three figures of speech.
As Klinkenberg and Seto (amongst others) have pointed out, syn-
ecdoche seems to belong to the nonnal functioning of ordinary
language (Klinkenberg 1983: 291). Seto writes:
In advanced countries a TV means a color TV and a fridge means an elec-

tric refrigerator. That this is possible is, I suppose, due to the mechanism of
the genus-for-species synecdoche. This kind of synecdoche may be related
to the so-called maxim of quantity. If this is possible, it seems reasonable
to suppose that the genus-for-species synecdoche is not a special figure of
speech, but a very general semantic mechanism which governs a wide
range oflanguage use .... (Seto, in press a: 7).
Heinrich Lausberg had already pointed out in his Elemente der

literarischen Rhetorik that the genus-for-species type of synecdoche
is especially frequent "urn im Kontext die Wiederholung eines die
Art bezeichnenden Wortes zu venneiden (variatio ... ), wobei der be-
stimmte Artikel als ruckverweisendes Bekanntschafts-Signal '" auf-
tritt" [so as to avoid in a certain context the repetition of a word that
designates a species (variatio ... ), with the definite article serving as
a cataphoric signal of familiarity] (Lausberg 1971: § 194). Synecdo-
che would therefore be one of the mechanisms used to establish co-
herence in texts (Harweg 1979: 186-187), and could or should be
studied by text-linguists. An example, randomly chosen from a Sun-
day colour-supplement, demonstrates this mechanism nicely. In an
article on North Korea, we can read:
By the Arch of Triumph tiny girls in white ballet tutus went cartwheeling
across lay-bys, bigger girls practised gravity-defying gymnastics routines;
never had I seen youngsters so supple and fit. Here, 10,000 kids assembled
to march with silver spears; there, 20,000 headed for the 120,000-seat Kim
Il Sung sports stadium to rehearse a series of dazzling visual tricks with
multi-coloured shields. When, finally, we tore ourselves away, the gym-

nasts kept waving until we had turned the corner and were out of sight.
(The Observer, Life magazine, 7 January 1996, p.IO).
The writer can introduce "variation" by going up and down in

hyponymic or synecdochic space or by going sideways in choosing
from an array of (register specific) synonyms.
Having lost its most figurative part, the part-whole synecdoche,
to metonymy, synecdoche thus seems to have lost at one and the
same time its status as a figure of speech and become a general se-
mantic or stylistic mechanism.
The figurative nature of the genus-for-species synecdoche can
certainly be revived, as demonstrated in the famous Raymond Que-
neau-example quoted by the Groupe J..I. (1970: 103),2 but still, our
much reduced synecdoche is certainly the least figurative of our
three master-tropes: metaphor, metonymy and synecdoche (Nerlich,
in press).
However, all is not lost, as apart from its function of variatio in
ordinary language, synecdoche seems to have a more specific func-
tion in one class of extraordinary language use or in one specific
functional style or register. A style or register can be defmed in the
following way:
Stilarten, die sich aus den Funktionen derSprache (Verstllndigung, Mit-

teilung, Wirkung) als gebrauchsfertige (relrurrente) Verwendungsweisen
ergeben, z.B. der wissenschaftliche und der umgangssprachliche Stil, der
Verkehrs- und Geschllftsstil, der Stil Offentlicher Verlautbarungen, der pub-
lizistische und der ktlnstlerisch-literarische 8til. (Lewandowski 1979, I:
217)
[Styles which emerge from the various functions of language (understan-
ding, communication, effect) as readily usable (recurrent) ways of language
use, as for example the style of scientific discourse and ordinary language,
the style of social interaction and commerce, the style of official announce-
ments, the style of publishing and the artistic and literary style.]
It seems that synecdoche is particularly common in any type of

"official" register, such as "police-speak" or the style of public an-
nouncements - and here we might have a hint as to one of the most
common communicative functions of synecdoche. Exploiting the

fact that the genus-species type of synecdoche is always true (where-
as in the case of metaphor the truth-function of a word is under-
mined), and that it is easier and safer to use a generic term rather
than a specific one, one can hear utterances such as "I apprehended a
person/vehicle on the premises". Synecdoche is used here "as an in-
strument to avoid introducing too many and especially too specific
referents" (Dirven 1993: 22). Not using the generic term can indeed
have legal consequences.
These uses of synecdoche in police-speak, in aircraft-speak, bus-
speak etc., in officialese and legalese in general, may strike us as
somewhat odd, but they are generally not perceived as "figures of
speech". We are dealing here rather with stylistic or register varia-
tions, where register is the product of functional variation correlated
with contextual variation, as researchers in the Hallidayan linguistic
tradition are apt to say (cf. Halliday 1978: 23-31).
When we look back at the history of the concept of register itself,
we find that John Rupert Firth first introduced this notion under the
heading of 'restricted languages' and "looked to 'science, technolo-
gy, politics, commerce', 'industry', 'sport', mathematics', and 'me-
teorology', or to 'a particular form or genre', or to a 'type of word
associated with a single author or a type of speech function with its
appropriate style' or 'tempo' (1968: 106,98, 112, 118-120)" (Beau-
grande 1993: 8). According to Firth, linguistics can regard each
"'person' 'as being in command of a constellation of restricted lan-
guages, satellite languages' ... , but these are 'governed' by 'the
general language of the community' (1968: 207-208)" (Beaugrande
1993: 8). Spinning out Firth's thoughts, one could say that the re-
moter the satellite or register orbiting round the common core lan-
guage, the more effective its use in communication, the more impact
its use has on the hearer. This remoteness can be based on the pre-
valent use of metaphors as in poetic texts, or on the prevalent use of
synecdoches as in official texts. That is to say, metaphors deviate
from ordinary language by jumping or crossing conceptual bound-
aries, and synecdoches deviate from ordinary language by moving
above (or below) basic level concepts (Rosch 1973). Both break "or-
dinary" expectations.
Firth also claimed that language is important in creating persons
and said that:
The meaning of person in the sense of a man or woman represented in fic-

titious dialogue, or as a character in a play, is relevant if we take a sociolo-
gical view of the personae or part we are called upon to play in the routine
of life. Every social person is a bundle of personae, a bundle of parts, each
having its lines ... the continuity of the person, the development of persona-
lity, are paralleled by the continuity and development of language in a va-
riety of forms. (Firth 1950, cit. Matthiessen 1993: 24)
Now, to be able to live life to the full in the social community,

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

ham to Heathrow bus and asked by the steward to "put the towels in
the receptacle provided when using the toilet". When putting the
paper towel in the bin later on we looked at it with different eyes.
There is an element of euphemism inherent in this use of synecdo-
che, but it is based on a different mechanism than the one where we
replace bin men with waste disposal operatives for example.
In Figure 8 we have tried to correlate three types of stylistic
devices with three types of registers or styles. Metaphor is predomi-
nantly used in the poetic register, but it can also be used in ordinary
language. Metonymy is predominantly used in ordinary language but
it can also be used in poetic texts. Synecdoche can ,be used in ordi-
nary language, mostly so as to introduce variation and coherence
into texts, but it also characterises official styles. It is therefore the
least "figurative" of our figures of speech.
Stilmittel - stylistic devices
~+
poetic ordinary
[variatio]
Stilform - style. register
Figure 8. The main communicational functions/registers of

metaphor/metonymy/synecdoche-use
In conclusion one can say that the genus-for-species synecdoche

plays a vital part in language and life. On the conceptual level it re-
flects and exploits the order in our categories, on the linguistic or
structural level it exploits semantic relations, and on the communi-
cationallevel it brings order into texts and into social relations.
Notes
I. At the turn from the nineteenth to the twentieth century, French historical
semanticists or those writing in the French language, had started to include
part-whole relationships under the heading of metonymy. Examples are
Nyrop's (1913) work and, more specifically, Esnault's (1925) work, where
he points out that: "Le s~mantisme du mot metaphore, c'est Transfer; celui
de metonymie Changement (de denomination); celui de synecdoque An-
nexion" [The semantics of the word metaphor is Transfer; that of metony-
my Change; that of synecdoche Annexation] (1925: 29). "Synecdoque et
metonymie ont, en effet, ceci de commun qu'elles respectent le 'cosmos',
l' ordre constant des phenomenes nature Is, ceci de different et d' oppose,
que la synecdoque considere les etres par leur classification, leur 'exten-
sion', la m~tonymie par leur activit~, leur 'compr~hension'" [Synecdoche
and metonymy have, in fact, that feature in common that they respect the
'cosmos', the constant order of natural phenomena, they differ and are in
fact opposed to each other in so far as synecdoche considers phenomena by
their classification, their 'extension', whereas metonymy considers them
from the point of view of their activity, their 'intension'] (1925: 35).
2. "11 reprit son chemin et songeusement quant A la tete, d'un pas net quant
aux pieds, il termina sans bavures son itin~raire. Des radis l'attendaient, et
le chat qui miaula esperant des sardines, et Amelie qui craignait une com-
bustion trop accentuee du fricot. Le maitre de la maison grignote les vege-
taux, caresse l'animal et repond a l'etre humain qui lui demande comment
sont les nouvelles aujourd'hui: - Pas fameuses." [He continued on his
way and, dreamily as to his head, with a clear step, as to his feet, he fmish-
ed his itinerary without any hitches. The radishes were waiting for him, and
the cat mewed in hope of sardines, and Amelie, who feared a marked com-
bustion of the stew. The master of the house nibbled the vegetables, strok-
ed the animal and answered the human being, who asks him how his day
has been: - Not brilliant.]
References
Beaugrande, Robert de
1993 'Register' in Discourse Studies: A concept in search of a theory, in:
Mohsen Ghadessy (ed.), 7-25.
Bonhomme, Marc
1987 Linguistique de la metonymie (Sciences pour la communication
16.). FrankfurtlM. etc.: Peter Lang.
Bredin, Hugh
1984 Metonymy. Poetics Today 5/1: 45-58.
Burke, Kenneth
1969 A Grammar of Motives. Berkeley / Los Angeles: University of Cali-
fornia Press.
Burkhardt, Annin
1995, ms. Zwischen Poesie und Okonomie. Die Metonymie als semantisches
Prinzip. To appear in: Euphorion 1996.
Cruse, David A.
1986 Lexical Semantics. Cambridge: Cambridge UII:iversity Press.
Dirven, Ren~
1993 Metonymy and metaphor. Different mental strategies of conceptua-
lization. Leuvense Bijdragen 82: 1-28.
Esnault, Gaston
1925 L'imagination populaire: Metaphores occidentales. Paris: Presses
Universitaires de France.
Firth, John Rupert
[1950] 1957 Personality and language in society. In: Papers in Linguistics. /934-
/95/, 177-189. London: Oxford University Press.
1968 Selected Papers of J.R. Firth /951-/959. Edited by Frank R. Pal-
mer. London / Harlow: Longmans, Green and Co.
Fontanier, Pierre
[1821; 1827] Les Figures du discours. Paris: Flammarion.
1968
Ghadessy, Mohsen (ed.)
1993 Register Analysis: Theory and Practice. London: Pinter Publishers.
Groupe J.1 (Jacques Dubois, Francis Edeline, Jean-Marie Klinkenberg, Philippe
Minguet)
1970 Rhetorique generale. Paris: Larousse.
Halliday, M.A.K.
1978 Language as a Social Semiotic: The Social Interpretation of Lan-
guage and Meaning. London: Amold.
Harweg, Roland
2 1979 Pronomina und Textkonstitution. (Beihefte zur Poetica 22.) Mtln-
chen: Fink.
Hopwood, A. T.
1959 The development of Pre-Linnaean taxonomy. Proceedings of the
Linnaean Society ofLondon 170: 230-234.
Jakobson, Roman
1956 The metaphoric and metonymic poles. In: Roman Jakobson and
Morris Halle (eds.), Fundamentals of Language, 76-82. Graven-
hage: Mouton & Co.
[1956] 1983 Two aspects of language and two types of aphasic disturbances. In:
id., (eds.), Language in Literature. Edited by Krystyna Pomorska
and Stephen Rudi, 95-120. CambridgelMass. / London: The Belk-
nap Press ofHarvard University Press.
Klinkenberg, Jean-Marie
1983 Problemes de la synecdoque. Du semantique a l'encyclopedique. Le
Fran~ais Moderne 51: 289-299.
Lausberg, Heinrich
41971 Elemente der literarischen Rhetori/c. Eine EinfiJhrung fur Studie-
rende der kiassischen, romanischen, englischen und deutschen Phi-
lolog;e. Mtlnchen: Hueber.
Le Guem, Michel
1973 Semantique de la metaphore et de la metonymie. Paris: Larousse.
Lewandowski, Theodor, (ed.)
31979 Linguistisches Worterbuch. Volume 1. Heidelberg: Quelle & Mey-
er.
Matthiessen, Christian
1993 Register in the round: Diversity in a unified theory of register an-
alysis, in: Mohsen Ghadessy (ed.), 221-292.
Meyer, Bemard
1993 Synecdoques: Etude d'une figure de rhetorique. Volume 1. Paris:
L'Harmattan.
Nerlich, Brigitte
in press Synecdoche: A trope, a whole trope and nothing but a trope? In:
Neal R. Norrick and Armin Burkhardt (eds.), Tropical Truth.
Nerlich, Brigitte, Zazie Todd and David D. Clarke
in press 'Mummy, I like being a sandwich'. Metonymy in language acquisi-
tion. In: Klaus-Uwe Panther and Gtlnter Radden (eds.).
Nyrop, Kristoffer
1913 Grammaire historique de la langue fran~aise. Volume 4. Seman-
tique. Copenhagen: Gyldendalske Boghandel Nordisk Forlag.
Synecdoche os a cognitive and communicative strategy 213
Nysenholc, Adolphe
1981 Metonymie, synecdoque, metaphore: Analyse du corpus chaplinien
et theorie. Semiotica 34: 311-341.
Panther, Klaus-Uwe and Gtlnter Radden (eds.)
in press Metonymy in thougt and language. Proceedings of the Hamburg
workshop on 'Conceptual Metonymy '. Amsterdam: Benjamins.
Rosch, Eleanor
1973 On the internal structure of perceptual and semantic categories. In:
Terrence E. Moore (ed.), Cognitive Development and the Acquisi-
tion ofLanguage, 111-144. New York: Academic Press.
Sato, Nobuo
1979 Synecdoque, un trope suspect. Revue d'Esthetique 1-2: 116-127.
Schofer, Peter and Donald Rice
1977 Metaphor, metonymy, and synecdoche revis(it)ed. Semiotica 21:
121-149.
Seth, Vikram
1993 A Suitable Boy. London: Phoenix.
Seto, Ken-ichi,
1995 On the cognitive triangle: The relation between metaphor, metony-
my and synecdoche, ms.
in press a Metonymy and the cognitive triangle.
in press b On distinguishing metonymy from synecdoche. In: Klaus-Uwe
Panther and Gtlnter Radden (eds.).
Todorov, Tzvetan
1970 Synecdoques. Communications 16: 25-35.
Tversky, Barbara
1990 Where partonomies and taxonomies meet. In: Savas L. Tsohatzidis
(ed.), Meanings and Prototypes. Studies in Linguistic Categoriza-
tion, 334- 344. London: Routledge.
Laws of thought, knowledge and lexical cbange
Beatrice Warren
There are universal laws of thoughts

which are reflected in the laws of change
of meaning ... even if the science of
meaning ... has not yet made much ad-
vance towards discovering them.
(Jespersen 1925: 212)
The concern of the present contribution is the interpretative mech-

anisms which make it possible for words to take on meanings that
are not memorized. The first section of the paper will deal with top-
ics such as meaning and reference, meaning and categorization, cri-
terial and non-criterial features of meaning, meaning and context.
This will lead up to a presentation of a model of the process of lexi-
cal change. Along with current cognitive approaches, it will be sug-
gested that certain types of mental operations on certain types of
knowledge are involved in lexical change. The nature of these oper-
ations and these types of knowledge will be described in the section
that follows. A distinction is made between three main types of lexi-
cal change, viz. novel hyponymic senses, non-literal senses and ap-
pended senses.
1. Preliminaries
1.1. Meaning and reference
The advent of cognitive approaches to linguistic theory has brought

along certain metaphors for the cognitive activities involved in inter-
preting figurative uses of words, notably mental space, domain,
source, target, mappings, blends, projections. By means of these
216 Beatrice Warren
tenns we may describe in some detail how the mind goes about
creating a new meaning, in particular figurative meanings. Consider,
for instance, the interpretation of the compound land yacht. I We can
posit that we have two mental spaces representing our concept for
'yacht' (input 1) and for its counterpart on land (input 2). By means
of cross-space mappings which will connect water with land, skip-
per with driver, course with road, yacht with car, tycoon with owner,
a new mental space emerges, giving us the meaning 'impressive,
luxurious car'. Admittedly, these tenns are illuminating. They do
provide "a handle on concepts" as Lakoff and 10hnson (1980) have
taught us good metaphors do.
However, the theory of blending is not a sufficient explanation
for the interpretation of metaphors. Consider another metaphorical
compound, i.e. asphalt beach, which - in Swedish translation -
happens to be the title of a book. Presumably, all of us have mental
spaces representing asphalt and beach; all of us are capable of agile
cognitive activity. Yet, I doubt that one can work out the intended
interpretation. However, as soon as we are infonned that it is an
expression employed by New Yorkers for rooftops used for
sunbathing, we will no doubt be able to produce a rationale for the
compound by means of conceptual blending.
By this I wish to illustrate the importance of retrieving a referent
in interpreting words. Meaning cannot be created solely by
extracting meanings from morphemes, but must involve matching a
linguistic unit with a referent (or, as I will presently argue, a class of
referents). Lexical meaning, it has been said, is that which takes us
from a combination of phonemes to a non-linguistic entity, be it
actual or hypothesized. Lexical meaning without a referent is
unthinkable.
1.2. Meaning and categorization
It was just claimed that if there is no referent, there is no meaning.

The reverse is, however, not true. The traditional view is that words
Laws o/thought. knowledge and lexical change 217
may have referents without having meanings. Proper names, which

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

form categories and our ability to form meanings for words would
indeed be difficult. However, I do not think that all features of
meaning can be of this kind. There are features of meaning that do
not specify the kind of entity or phenomenon the speaker wishes to
218 Beatrice Warren
denote, but his or her attitude to it. We must distinguish between

features of meaning which are criterial and fix reference and which
we can connect with our ability to form categories and features of
meaning which do not fix reference and for this very reason express
speaker evaluation and which in my view cannot have truth values.
Words containing these non-criterial evaluative-attributive features
of meaning are of fundamental importance, i.e. words such as good,
bad, beauty, folly, wisdom, justice, right, wrong, difficult, easy, dan-
ger, safety, etc., etc. Without words containing such features, free-
dom of speech would be of little value, because we would hardly
have verbal means of expressing an opinion. Bolinger expresses this
point in the following manner: "... language is more than logic and
meaning is more than truth and to exclude other values is to insist
that language is nothing but a transmission belt for factual knowl-
edge" (1975: 160).
The distinction between defining and evaluative-attributive fea-
tures of meaning that I have just described is of course reminiscent
of the time-honoured distinction between denotation and connota-
tion. I hope, however, that I have made the distinction more precise
in that I have pointed out that evaluative features are definitely com-
municated not vague associations forming a background noise and
above all in that I have pointed out that features express speaker
evaluation because they do not fix reference. That is to say, it is up
to the speaker what entities or phenomena words containing such
features apply to. The question left unanswered is what is the source
of these features? I have elsewhere suggested (Warren 1992: 19) that
the mood of the whole situation in which a word occurs, i.e. more
precisely the character of the speaker, the manner in which and the
reason for which (s)he uses a word may all contribute to giving it its
expressive force. It is claimed that if paralanguage expresses the
speaker's dislike but his actual words express approval, we trust the
former source of information rather than the latter. For survival it is
often no use knowing what kind of thing we are faced with, if we do
not also know whether this thing is bad or good or indifferent. Eva-
luations must be as important as forming categories.
Laws o/thought, knowledge and lexical change 219
1.4. Meaning and context
In the last two decades or so, the assumption that words have stable
meanings which we retrieve ready-made as we come across them in
an utterance has been challenged by a number of linguists, probably
independently of each other. For instance, Bransford and McCarrell
(1974) suggest that words are cues or instructions to create meaning.
Fillmore thinks of sentences as "blueprints" of which the interpreter
constructs interpretations (1985), Sperber and Wilson (1986) wish to
replace the traditional code model with an inferential model of inter-
pretation. The Norwegian philosopher-linguist Rommetveit (1988)
insists that words do not have literal, basic, invariant meanings, but
meaning potentials. I myself have suggested (Warren 1992) that we
must distinguish between the meaning of a word out of context and
its meaning in context. The former type of meaning I refer to as dic-
tionary meaning. It is the meaning that the lexicographer would be
interested in. The latter type of meaning I refer to as contextual
meaning, which is the value we give a word in context. Further, I
have suggested that whenever we come across a word in context, we
contextualize its meaning and reference. This contextualization is a
matter-of-course process in interpretation. It involves a readiness to
adapt the dictionary meaning if necessary and if at all possible so
that it fits the context in question. The question is what modifica-
tions are possible. I will presently attempt to address this issue, but
first I wish to point out that the contextualization process is normal-
ly followed by a decontextualization process. This process may in-
volve new meanings being generalized further or, if they are deemed
to be of nonce value, scrapped altogether, which is probably fre-
quently the case. However, sometimes they may be considered use-
ful for some reason and be memorized. This memorized new mean-
ing may spread to a sufficiently large group of people to become a
new dictionary meaning. In other words, a distinction is made be-
tween a new contextual meaning, which is part of parole and the
work of a moment, and a new dictionary meaning, which is part of
/angue and the creation of which is gradual. The distinction between
220 Beatrice Warren
contextual and dictionary meaning is hardly controversial. Neverthe-

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

ter of the referent.
ii. In principle words with lexical meaning will not have unique
reference. Words with purely referential meaning can therefore
simply be considered to be category labels. To form a category
normally involves perceiving at least one common trait, disre-
garding dissimilar ones. Hence it is natural to think of mean-
ing as componential.
lll. We must distinguish between features of meaning which fix
reference. That is to say, those about which there is consensus
among the members of a language community as to what kind
of entities and phenomena they denote (criterial features) and
features of meaning concerning which the speaker decides
what entities or phenomena they are applicable to (evaluative-
attributive). Criterial features derive from man's need and abi-
lity to form categories; evaluative-attributive features derive
from man's need and ability to form judgements.
Laws a/thought, knowledge and lexical change 221
IV. We must distinguish between out-of-context meanings of

words which are static and meanings of words in context
which are flexible. The fact that word meaning is modifiable
allows precision with least effort.
2. A model of interpretation of words in context
As illustrated in Figure 1, knowledge of the dictionary meaning and!

or the context - both linguistic and non-linguistic - and knowledge
of worldly "things" will suggest a plausible referent (or some plau-
sible referents). The dictionary meaning (provided we have one) will
then be contextualized to confirm that it does fit or can be made to
fit the character of the favoured referent or referents, a process
which will finally yield some chosen contextual meaning and refer-
ent(s).2
Contextu- CONTEXTUAL
Plausible
alization REFERENT(S)
referent( s)
process and MEANING
Figure I. A model of interpretation of words in context
If we apply this model to the example asphalt beach above, we

see that we did not arrive at any good idea of its reference and mean-
ing because we did not have sufficient context to access our knowl-
222 Beatrice Warren
edge that people sometimes sunbathe on rooftops, which, had we

had, predictably would have led to the metaphorization of beach.
As already pointed out, it is nonnally the contextualization pro-
cess that gives rise to new meanings. The question is then: in what
ways can dictionary meanings be modified to yield new meanings.
Assuming that certain natural strategies representing universal cog-
nitive processes will be appealed to, we can predict that interpreters
will of course assume that the speaker has used a particular word be-
cause (s)he believes that it will enable interpreters to retrieve the in-
tended referent(s). Therefore there must be some connection be-
tween dictionary and contextual meaning and reference and this con-
nection must be deemed so self-evident that - given the particular
context - it can be worked out. It is to the description of these types
of connections that we now turn.
3. Creating byponyms
Lyons makes use of the tenn denotatum, which he defines as "the

class of objects, properties, etc. to which a word correctly applies,,3
(1977: 207). The members of the denotatum of a word are mental
entities, whereas the referents of a word are nonnally actual entities
connected to a particular context. Applying this distinction to my ap-
proach, we can say that the dictionary meaning of a word is that
which takes us to its denotatum, the contextual meaning of a word is
that which takes us to its referent or referents.
Whereas the denotatum of a word encompasses all possible mem-
bers in the set, a word in context often refers to one or some mem-
bers. If context suggests that the entities referred to must have some
property or properties over and above the features specified by the
dictionary meaning of the word in question, there will be a natural
addition of features of meaning creating a subcategory. If such a
subcategory is of lasting interest, such contextual meaning may be-
come conventionalized. Indeed, examples of conventionalized spe-
cialized meanings are not hard to find. Missile is not used to denote
Laws o/thought, knowledge and lexical change 223
any thrown object but often about ballistic robots. Drug often sug-
gests narcotics and when we say of a woman that she is liberated,
we do not simply mean that she is freed but that she is freed from
traditional ideas concerning woman's role in society.
In literature on semantic change, it is often pointed out that nar-
rowing of meaning is a common type of lexical sense shift and that
it is paralleled by widening of meaning. Whereas, needless to say, I
naturally accept that narrowing is natural in interpretation, I am not
so convinced that widening of meaning is a possible interpretative
strategy. Consider the following text:
In engineering it is rare to fmd iron used in its pure fonn. Generally the me-
tal is alloyed with carbon and other elements to fonn wrought iron, steels
and cast iron.
The definite article preceding metal signals to the reader that

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

with metal. This suggests that generalization of meaning is not a na-
tural interpretative mechanism. In other words, it seems that we can
constrain the reference of - e.g. - fruit in a context so that it refers
exclusively to, say, apples (and not oranges, pears or bananas), but
we cannot generalize the reference of apple to include oranges and
pears and bananas, etc.
224 Beatrice Warren
I have just suggested that one of the conditions for making a

word acquire some non-conventional reference and meaning is that
there is such a well-established connection between the non-conven-
tional reference and/or meaning and its conventional denotatum and
meaning that the encoder can trust his audience to access it. The
connection between a superordinate term and one of its hyponym is
of this kind. However, as I try to illustrate below, this link can be
used in one direction only.
that which is a kind offruitlmetal-+ viz. in this context "apple", "iron"

*that which apple/iron is a kind of -+ viz. in this context "fruit", "metal"
This may seem strange at first sight, but perhaps the explanation
is simply the following: if you are told to look for a credit card in a
purse which is in a handbag, it would be very strange to construe
this as an instruction to look for the card anywhere in the handbag
but not in the purse. Our asswnption must be that it is somewhere in
the purse.
Perhaps I should emphasize that I do not deny that generalization
of meanings occur. They demonstrably do. What I doubt is that ge-
neralization is an interpretative mechanism. In other words, genera-
lization must be the result of some interpretative mechanism and not
a mechanism itself.
4. Creating non-literal senses
The rule that I have just suggested, i.e. that the entity or phe-
nomenon referred to should be within the denotatum of the word in
question, is a rule which we all know can be violated and indeed of-
ten is. Again, provided there is some decipherable connection be-
tween the conventional and non-conventional meaning and refer-
ence, we have then what is recognised as a non-literal use of a word.
According to tradition, as is illustrated in Table 1, this connection
may be (i) one or more reminiscent properties, in which case we
Laws ofthought, knowledge and lexical change 225
have a metaphor, (ii) a polarized property, in which case we have re-

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

Connection Sense extension
• reminiscent property • metaphor (memory of a
computer)
• polarized property • reversal (afinefriend)
• (property intensity of which • (hyperbole, litotes)
need to be up- or downgraded)
• relation of concomitance • metonymy (hunchback)
As we have seen, the difference between metaphor and metony-

my is traditionally that metaphor involves relations of resemblance,
whereas metonymy involves relations of concomitance. I have else-
where emphasized that the interpretation of metaphors involves re-
trieval of properties, whereas the interpretation of metonyms in-
volves retrieval of relations and that this is the important difference
(Warren, in press a and b). Metonymy, 1 have also suggested, is basi-
cally an abbreviation device. Consider some well-known examples
of metonymy:
(1) She heard the piano from next door.

(2) The kettle is boiling.
(3) That's Churchill. (pointing to a sculpture)
226 Beatrice Warren
Paraphrasing the interpretation of these, we get:
(1) a. She heard (the sound produced) by the piano.

(2) a. The (water in the) kettle is boiling.
(3) a. That's (a sculpture representing) Churchill. 4
We see here that what is left out is that which is so closely con-
nected with what is mentioned that it goes without saying in a con-
text which requires its retrieval to make sense. What is left of the
noun phrase has greater information value relative to that which is
left out.
Let us now compare (4) and (5) below:
(4) Arthur is Hamlet in this play. (Hamlet = metonym)

(5) Arthur is a real Hamlet. (Hamlet = metaphor)
In (4) we have - according to my definition - a metonym, in (5) a

metaphor. The interpretation of Hamlet in (4) can be accounted for
by supplementing this noun phrase as in (5):
(4) a. Arthur is the actor representing Hamlet in this play.
It is not possible to account for the interpretation of Hamlet in (5)

in similar fashion. In order to interpret Hamlet in (5), we need to re-
trieve at least one property shared by Arthur and Hamlet, for in-
stance that Arthur is a brooding young man.
As we see from this particular example, as well as from example
(3), a metonymic relation can be iconic. The relation between the
sculpture of Churchill and the person Churchill could surely be said
to be one of resemblance. According to the traditional definition,
Churchill in (3) would consequently be a metaphor and so would
Hamlet both in (4) and in (5). Since, according to my analysis, we
employ different strategies working out interpretations for Hamlet in
(4) and (5), it seems reasonable that we separate them also termino-
logically. S
Laws ofthought. knowledge and lexical change 227
To strengthen my argument I ask the reader to consider one more

example, viz.:
(6) Arthur has hisfather's eyes.
the interpretation of which spells out as:
(6) a. Arthur has (eyes which resemble) his father's eyes.
Since it is possible to account for the interpretation of the noun

phrase his father's eyes simply by expanding it without having to re-
trieve shared features, I do not hesitate to think of it as a metonym in
spite of the resemblance relation.
So far I have contrasted my approach to the metaphor-metonymy
distinction with the traditional definition. Let me also contrast it
with the favoured current approach which is that metaphor involves
mapping across domains, whereas metonymy involves mapping
within a domain (Lakoffand Johnson (1980) and Lakoffand Turner
(1989), further developed by Croft (1993». I have some problems
with this: one is that I fail to see much difference between claiming
that metaphor involves mappings across domains and the traditional
view that metaphor involves seeing similarity in dissimilarity. And
if we are to see, for instance, ecstasy and a particular drug as belong-
ing to the same domain, is that not because experience has supplied
an associative link between them? And is that then radically differ-
ent from the traditional claim that metonymy is based on contiguity?
Another difficulty is the looseness of the term domain. How do we
determine whether entities belong to the same domain or not?
Personally I would have imagined that all handbags belong to the
same domain, but if I were to hear somebody say of a handbag
which this person found elegant that it is a real Guzzi, although it
manifestly is not, I would take a real Guzzi to be a metaphor and be
willling to form two domains: one with Guzzi handbags and one
comprising all other handbags. The point I am trying to make is that
since domains do not seem to be static and invariant constructions of
228 Beatrice Warren
our experience but adaptable to context (on this see Ungerer and
Schmid (1996», surely it is the interpretation we favour that induces
the formation of domains and not the domains that basically induce
interpretations. I concede, however, that the creation of domains en-
ables access to features intended to be communicated in this case
"elegant" .
Finally, I fail to see that viewing metaphor and metonymy in or
across domains can account for important differences between these
two different figures of speech. Viewing metonymy as basically an
abbreviation device, on the other hand, we can explain why, for in-
stance, metonyms comparatively often lack expressive force (as the
examples cited here demonstrate) and comparatively rarely supply
names for unnamed entities. Metaphors are different in this respect:
they either are used to name (at least on word level) or to have ex-
pressive force or both functions simultaneously. (For other differen-
ces between metaphor and metonymy which remain unaccounted for
by the domain approach, see Warren (in press a).)
I conclude this subsection by giving the promised explanation of
why hyperbole and litotes are within parenthesis in Table 1. As al-
ready pointed out, we have hyperbole if, in fitting the conventional
meaning of the linguistic unit to our favoured referent, we find that a
match is only possible if the value of some feature of meaning is de-
creased in strength. Similarly, we have litotes if we find that a match
demands that the value of some feature of meaning is increased in
strength. However, the encoder need not find any such down- or up-
grading necessary. In other words, there may be disagreement be-
tween encoder and decoder as to whether adjustment of the conven-
tional sense is called for. For this reason, it is somewhat problematic
to look upon hyperbole and litotes as invariably non-literal uses of
words.
5. Creating appendant meanings
Finally I would like you to consider the following three examples:
(7) It won't happen while I still breathe. [live]

(8) - Why didn't you pick the tomatoes I asked you for?
- They were still green. [unripe]
(9) He ought to know. He is a professor. [expert]
The fIrst example is offered by Halliday (1985: 319-320) as an

example of metonymy. If breathe in (i) is a metonym, then green in
(ii) and professor in (iii) ought to be, too. Yet none of these ex-
amples fIt my criteria for metonymy. It does not seem possible to
think of them as abbreviated noun phrases, nor as non-literal.
I concede, however, that just as so often is the case in metonymy,
we are induced to go from conventional to non-conventional sense
and reference because of concomitance relations. The difference is,
however, that in the case that I term metonymy, the concomitance
relation is between entities, whereas in the examples above, the con-
comitance relation is between propositions, which translates into an
antecedent-consequent relation. That is to say in these cases:
• if one breathes, then one lives

• if tomatoes are green, then they are probably unripe
• if somebody is a professor, then (s)he is probably an expert of
some kind
I therefore refer to senses arising from antecedent-consequent

relations as implications. Implications do not replace the senses
from which they derive. Consequently green, breathe, professor in
the examples above can be said to have two senses: the conventional
explicit sense and the appended implicit sense. Provided the impli-
cation is not absolute, this appended sense is cancellable as we see
in (10):
230 Beatrice Warren
(10) The tomatoes are green [unripe}. but strangely enough ripe
anyway.
This is not possible in the case of my version of metonymy:
(11) *The [water in the} kettle is boiling. but not the water.
Whereas implications will not cause violations of truth condi-

tions, metonymic senses will, since that which the predication con-
cerns is left out: it is not the kettle, but the water in it that is boiling.
Implied senses can, however, be conventionalized and no longer
be dependent on a particular standard sense to be elicited. It is pos-
sible to hang up on somebody by putting down the receiver; it is
possible to go to the toilet behind a bush; it is possible to go to bed
in a tent without a bed. Sense changes of this kind have not infre-
quently been described in the literature. In English the most famous
example is possibly that of how today's meaning of beads developed
from gebet. which used to mean 'prayer', into meaning 'balls of a
necklace': if one told one's prayers, one would be likely to tell (i.e.
count) the balls of the rosary (Stem (1965».
My main reason for insisting on keeping implications termino-
logically apart from the non-literal metonymic senses is that here we
have a case in which conventional and novel senses may peacefully
co-exist. The fact that we do not have to choose either a conven-
tional or a novel sense makes a gradual switch-over from a conven-
tional to a novel sense possible. One consequence of this is that
there may be remnants of the conventional sense affecting the novel
sense for a long time. It is often pointed out that the grammaticaliza-
tion process is gradual and that there tend to be constraints in the
new grammatical item connected with the original lexical sense
(Hopper and Traugott 1993). Implication can account for these fea-
tures of grammaticalization. Consider, for instance, be going to as a
future tense auxiliary, the meaning of which is said to be 'future cul-
mination of present intention or cause' (Leech 1977: 54).
Laws of thought, knowledge and lexical change 231
6. Summing up
This account of lexical change is not complete in that I have con-

centrated on changes dependent on context, these being by far the
most frequent ones. That is to say, I have not considered changes in-
duced by changes in our perception of the class-distinctive features
of the members of the sets. This is a process referred to as substitu-
tion by Stem (t"951) and linguistic conservatism by Ullmann (1957)
- conservatism because we keep the name although the definition
has changed: pens, for instance, are no longer parts of wings of
birds. Nor have disseminating processes such as (notably) analogy
been considered. Analogy may cause a novel sense to spread from
one word to others with which this word is related by means of, for
instance, synonymy, hyponymy, antonymy or derivation. There are
also other processes by means of which new meanings are created
that have been ignored. However, the main processes are the ones
that have been brought up.
The account is basically cognitive in spirit in that it is accepted
that in making words assume meanings, perception and category-
forming principles, context and world knowledge (be it frames,
scenes, schemas, cognitive categories or domains) are essential in-
gredients. I have, however, found weaknesses both in traditional and
cognitive approaches. That is to say:
1. The cognitive linguists' rejection of objectivism (on this, see

Lakoff (1988)), is not sufficient to account for the distinction
between criterial and non-criterial features of meaning, a dis-
tinction which is of fundamental importance in semantics.
11. Although I accept that widening of meaning may be a result of
interpretative strategies, there are reasons to doubt that it is an
interpretative mechanism per se.
111. I have not accepted that the crucial difference between meta-
phor and metonymy is that the former is based on iconic rela-
tions, whereas the latter is based on contiguity relations, or
232 Beatrice Warren
that the fonner involves mapping across domains, whereas the

latter involves mappings within a domain.
IV. I have not accepted a wide definition of metonymy but argued
that the tenn metonymy should be restricted to non-literal ref-
erential senses. I have pointed out that the literal types of met-
onymy, which I would tenn implications, are different from
the non-literal ones, mainly in that these senses are based on
antecedent-consequent relations between propositions and
occur originally as appended senses.
Notes
I. I owe the example to Gilles Fauconnier, who brought it up at a lecture

given in Stockholm in June 1996. For an account of his theories, see Fau-
connier (1985).
2. I have presented retrieval of some plausible referent and the contextuali-
zation process as being sequential. These two processes may, however, be
more or less simultaneous.
3. Note that this definition can only apply to purely referential words, since
words with attributive-evaluative features have no correct application, this
being the very feature that enables words to convey attitudes.
4. The manner of paraphrasing I have suggested is more precisely:
She heard that which the piano produced ...
That which is in the kettle ...
That's something which represents Churchill ...
This mode of paraphrasing demonstrates better how knowledge of context
and interpretative strategies collaborate to produce a plausible referent.
That which represents Churchill could, e.g., refer to a photo, a painting, an
actor, etc. depending on context.
5. Nunberg (1996) and Jackendoff(1997) consider similar examples as cases
of metonymy, although they prefer the tenn meaning/reference transfer.
References
Bolinger, Dwight
1975 Aspects ofLanguage. New York etc.: Harcourt, Brace and World.
Bransford, John and Nancy Mc Carrell

1974 A sketch of a cognitive approach to comprehension: some thoughts
about understanding what it means to comprehend. In: Waiter
Weimer and David Palermo (eds.), Cognition and the Symbolic
Processes, 189-229. Hillsdale, New Jersey: Lawrence Erlbaum
Associations.
Croft, William
Fauconnier, Gilles
1985 Mental Spaces. Cambridge, Mass.: The MIT Press.
Fillmore, Charles
1985 Frames and the semantics of understanding. Quaderni di Semantica
6: 222-254.
Halliday, M. A. K.
1985 An Introduction to Functional Grammar. London etc.: Edward
Arnold.
Hopper, Paul and Elizabeth Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
Jackendoff, Ray
1997 The Architecture of the Language Faculty. (Linguistic Inquiry Mo-
nographs 28.) Cambridge, Mass.: The MIT Press.
Jespersen,Otto
1925 Menneskehed, nasjon og individ i sproget. Oslo: Aschehoug.
Lakoff, George
1988 Cognitive semantics. In: Umberto Eco, Marco Santambrogio and
Patrizia Violi (eds.), Meaning and Mental Representations, 119-54.
Bloomington / Indianapolis: Indiana University Press.
Lakoff, George and Mark Johnson
1980 Metaphors We Live By. Chicago: University of Chicago Press.
Lakoff, George and Mark Turner
1989 More than Cool Reason: A Field Guide to Poetic Metaphor.
Chicago: The University of Chicago Press.
Leech, Geoffrey
1976 Meaning and the English Verb. London: Longman.
Lyons, John
1977 Semantics. 2 volumes. London / New York: Cambridge University
Press.
Nunberg, Geoffrey
1996 Transfer of meaning. In: James Pustejovsky and Branimir Boguraev
(eds.), Lexical Semantics. The Problem of Polysemy, 109-133.
Oxford: Clarendon Press.
234 Beatrice Warren
Rommetveit, Ragnar
1988 On literacy and the myth of literal meaning. In: Roger SIllj6 (ed.),
The Written World, 13-40. (Springer Series in Language and Com-
munication 23.) Berlin / Heidelberg: Springer-Verlag.
Rosch, Eleanor
1978 Principles of Categorizations. In: Eleanor Rosch and Barbara B.
Lloyd (eds.), Cognition and Categorization. 27-48, Hillsdale, New
Jersey: Lawrence Erlbaum.
Rubl, Charles
1989 On Monosemy: A Study in Linguistic Semantics. Albany: State Uni-
versity of New York Press.
Sperber, Dan and Deidre Wilson
1986 Relevance: Communication and Cognition. Oxford: Basil Black-
well.
Stem, Gustav
[1931] 1965 Meaning and Change 0/ Meaning. Bloomington: Indiana University
Press.
Ullmann, Stephen
[1951] 1957 Principles o/Semantics. Oxford: Basil Blackwell.
Ungerer, Friedrich and Hans-J6rg Schmid
1996 An Introduction to Cognitive Linguistics. London / New York:
Longman.
Warren, Beatrice
1992 Sense Developments. Stockholm: Almqvist & Wiksell.
in press a What is metonymy? In: The proceedings 0/ The International
Conference on Historical Linguistics. August. 1995.
in press b No more ham sandwiches, please. In: GOnter Radden and Klaus-
Uwe Panther (eds.), Metonymy in Language and Thought.
Amsterdam / Philadelpia: Benjamins.
Section III
Case studies
Intensifiers as targets and sources
of semantic change
Ekkehard Konig and Peter Siemund
1. How regular is semantic change?
In contrast to the established view on sound change and morphologi-

cal change, which seem to be governed by clear restrictions and
principles, the prevailing view on semantic change has been for a
long time and probably still is that "each word has its own history"
(1. Grimm, J. GiIlieron). This pessimistic view has been seriously
called into question by some recent studies, which have shown that it
is possible to formulate some pervasive generalizations about se-
mantic change. First of all, it has been shown that all semantic
changes are instances of a very limited set of possible processes,
such as metaphor, metonymy, ellipsis, narrowing, broadening, etc.
(Traugott 1985; KoniglTraugott 1988; Blank 1997; Koch 1997). Se-
condly, it is possible to formulate at least some generalizations about
the source domains and target domains involved in such changes, in
particular, with regard to metaphorical change. The structuring of
temporal domains in terms of spatial notions (SPACE -+ TIME) or,
more generally, the pervasive change from concrete to abstract are
cases in point. Thirdly, various attempts have been made to formu-
late some generalizations about possible directions of semantic
change. E. Traugott's hypothesis that meanings grounded in the
socio-physical world of reference develop into meanings grounded
in the speaker's world and further into markers of metatextual attitu-
de to the discourse is one of the best known examples (Traugott
1989, 1995).
That semantic change is far from erratic or random and is in fact
general and regular to a certain extent is particularly obvious if the
focus of investigation is on minor lexical classes (function words)
238 Ekkehard K6nig and Peter Siemund
rather than on the major word classes (N, V, Adj, P). The semantic
development of such function words does not only reveal some gen-
eral tendencies of semantic change within a particular language, but
also allows cross-linguistic generalizations to be made about typical
source and target domains and possible directions of change.
2. The meaning of intensifiers
Since this paper is centrally concerned with general processes of se-

mantic change involving intensifiers either as target or as source, our
analysis has to begin with a brief analysis of their meaning (cf. Ed-
mondsonIPlank 1978; KoniglSiemund 1996a, b). Intensifiers are ex-
pressions like the following:
(1) Germ. selbst; Russ. sam; It. stesso; Engl. x-self, Fr. x-meme,
etc.
There is no general agreement as to how such expressions should

be categorized. We find at least a dozen different labels in grammar
handbooks or specific studies. The reason for choosing the term in-
tensifiers is not that it is particularly illuminating, but simply that it
avoids misleading connotations, in contrast to terms like emphatic
reflexives. The morphological properties of such intensifiers differ
considerably across languages: Intensifiers may be invariant (Ger-
man, Mandarin, Japanese) and they may inflect for some or all of the
following features: person, gender, number and case (Turkic, Slavic,
Romance, English). In the latter case they manifest adjectival be-
haviour and typically exhibit agreement with a nominal co-constitu-
ent. Among their syntactic properties there is one that seems to be
the most reliable criterion for identifying them across languages: In
nearly all of the fifty languages investigated so far we found that in-
tensifiers are used as adjuncts to noun phrases, i.e. they combine
with a noun phrase to form another noun phrase:
Intensifiers as targets and sources o/semantic change 239
(2) a. I would like to talk to the Pope himself

b. The work of Picasso himself is what I came to see.
c. The work ofPicasso itself ...
Furthermore, there seem to be semantic criteria which clearly al-

low their identification across languages. Intensifiers seem to have at
least two uses: an adnominal one and an adverbial one. In addition to
functioning as adjuncts to noun phrases (cf. (2)), intensifiers can also
be used as part of a verb phrase:
(3) a. The President wrote that speech himself

b. I had a car like that once myself
As is shown by examples like (3a-b), many languages further dif-

ferentiate between two such adverbial uses: (a) an exclusive use,
roughly paraphrasable by 'alone', 'without help' and (b) an inclusive
use, paraphrasable by 'too'. Since the relevant contrast is not found
in all languages, however, it will play no role in the subsequent dis-
CUSSlon.
To describe the meaning of adnominal intensifiers simply as ex-
pressing 'emphasis', 'intensification' or 'unexpectedness' is clearly
not very illuminating. Examples like (2) show that such expressions
evoke alternatives to the denotation of the expression they combine
with. Since this property is generally associated with focusing (iden-
tificational focus), it suggests that these expressions should be re-
garded as focus-sensitive adjectives or particles. Perhaps the evoking
of alternatives is then a property of the focusing intensifier are as-
sociated with, and the co-constituent they combine with can be an-
alyzed as their focus. TIris is exactly the view that will be taken here.
What then is the real contribution that an adnominal intensifiers
makes to the meaning of a sentence? To answer this question for ex-
amples like (2a), we need to ask what the alternatives evoked in such
cases are. A plausible answer certainly is that they are people associ-
ated with the Pope in some way: his Cardinals, his secretary, his col-
laborators or clergy. An assertion of (2a) is thus tantamount to as-
serting that the speaker does not want to talk to any of those people
240 Ekkehard Konig and Peter Siemund
from the environment of the Pope. Generalizing from such exam-

ples, it is suggested in Konig (1991) that adnominal intensifiers as-
sociate a periphery of alternatives (Y) with a center (X), constituted
by the denotation of their focus. Building on this idea, Baker (1995)
has distinguished several types of such relations between a center
(characterized as 'discourse-prominent' by Baker) and a periphery,
which can be illustrated by examples like the following:
(4) a. Bill Clinton himselfwill sign the document.

b. The passengers got away injured. The driver himself died
on the scene of the accident.
c. Mary's husband looks after the children. Mary herselfhas a
regular job.
d. He was not particularly tall, a little taller than Jemima her-
selfperhaps, but his shoulders in the tweed suit were broad,
giving an air of authority, and he himself, if not exactly
heavy, was certainly a substantial man. [Antonia Fraser, A
Splash of Red, 1981: 88]
A center can be established on the basis of hierarchies in the real

world (cf. (2a), (4a». Intensifiers in their adnominal use can thus al-
ways be combined with expressions for people of high rank. In fact,
it seems that these are the examples that we think of first whenever
we are asked to produce examples with adnominal intensifiers. In
our modem egalitarian societies central roles are more typically tem-
porary, rather than permanent. A driver is central, in contrast to the
passengers, in all matters of driving and safety, a guide is central
during the time of an excursion, etc. (4b). Centrality may also be the
result of taking a certain person as point of departure for the identi-
fication of others, as in (4c). Note that the intensifier in this example
could not be combined with the expression Mary's husband. Finally,
a person may be used as the center of observation or perspective in a
narrative and become central in that sense. The fmal example (4d) is
a case in point. Following Baker (1995), the conditions for using ad-
nominal intensifiers can thus be summarized as follows:
Intensifiers as targets and sources ofsemantic change 241
(5) conditions for the use of adnominal x-self (Baker 1995)

a. X has a higher rank than Y on a real-world hierarchy
b. X is more important than Y in a specific situation
c. Y is identified relative to X (kinship terms, part-whole, etc.)
d. X is the subject of consciousness, center of observation, etc.
(logophoricity)
In German there are expressions often used in captions which

nicely illustrate the points made above:
(6) Maria selbdritt

Mary self-three
'Mary together with two persons'
In contrast to the adnominal use, the adverbial (exclusive) use of

intensifiers will not play a prominent role in the subsequent discus-
sion. We will therefore only say that sentences with exclusively used
adverbial intensifiers make an assertion which can be roughly repre-
sented by the relevant sentences without intensifiers, but with an in-
formation structure in which the agent subject is focused against the
rest of the sentence as background. Moreover, there is a presupposi-
tion to the effect that it is the denotation of the subject which profits
or suffers most as a result of the relevant action:
(7) a. The President wrote the speech himself.

b. [[the President]f wrote the speech] (assertion)
c. The President is primarily affected. (presupposition)
The expressions listed in (1) are the most prominent representa-

tives of intensifiers in the languages mentioned, i.e. they are the ones
most frequently used, exhibiting the fewest restrictions and the great-
est versatility in their use. The list given in (1) should not be taken to
suggest, however, that there is only one single intensifier in each lan-
guage. The following examples from German show that languages
typically have more than one intensifier, even if their number is very
limited in each case:
242 E/clcehard Konig and Peter Siemund
(8) (hochst)selbst, (hochst)personlich. eigen, leibha/tig, in Person,

von selbst, von sich aus, an sich, (von) al/ein
3. Processes of semantic change with intensifiers

as targets and sources
In this section only a brief overview of all changes involving intensi-

fiers as sources or targets will be given. Each of the changes listed
will then be discussed in detail in a subsequent section.
Intensifiers frequently derive from expressions for body parts.
Leibhaftig in Gennan is the clearest example we can find in that lan-
guage. Intensifiers may further develop into reflexive anaphors. The
development of reflexive pronouns in English as a result of a fusion
between personal pronouns and the intensifier self (him + self ....
himself) as well as similar developments in Afrikaans are clear ex-
amples.
(9) BODY PARTS .... INTENSIFIERS .... REFLEXIVE ANAPHORS
Intensifiers may also, so it seems, adopt the meaning of their fo-

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

EUPHEMISMS
A very different development is indicated by the polysemy found

in many languages for intensifiers. In addition to the adnominal use
discussed above, many languages use the same expressions pre-
nominally as scalar focus particles expressing unlikely instantiations
of values in certain contexts. In other words the relevant meanings

correspond to Engl. even.
(11) INTENSIFIERS'" SCALAR ADDITIVE FOCUS PARTICLES
In German selbst exhibits both uses, in contrast to selber, which

is only used as an intensifier:
(12) a. Der Papst selbstlselber wird uns besuchen.

'The Pope himself will come to visit us.'
b. Selbst der Papst ware hier ratios.
'Even the Pope would not know what to do.'
Especially the Romance languages show that intensifiers may be

involved in the development of demonstrative pronouns (Sp. ese).
From these demonstratives, definite articles and personal pronouns
may develop (Sardinian, Catalan).
(13) INTENSIFIERS'" DEMONSTRATIVE PRONOUNS'" DEFINITE ARTI-

CLES/PERSONALPRONOUNS
Finally, we also find that intensifiers frequently develop into ex-

pressions indicating type identity (Sp. mismo).
4. From expressions for body parts to intensifiers
A wide variety of languages for which the relevant connection is still

visible synchronically show that intensifiers often develop from ex-
pressions for body parts ('body', 'soul', 'head', 'eye', 'bone', 'per-
son' « persona 'mask'). The following examples illustrate this his-
torical and synchronic connection (cf. Moravcsik 1972):
(14) Arab. ayn 'eye'; Arab. nafs 'soul'; Amharic ras- 'head'; Geor-
gian tviton, tQVi 'head'; Germ. leibhaftig; Hausa ni dakaina 'I
with my head'; Hebrewetsem 'bone'; Hung. maga 'seed'; Jap.
zisin 'body'; Okinawan duna 'body'; Rum. insumi 'person';

etc.
Given such convincing cross-linguistic evidence, Grimm's (1967)

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

fiers are primarily used with names of high-ranking persons. In
Old English, for example, self was primarily used with nouns
like Crist, Haeland, God, Drihten, de%l, cyning, etc. (Farr

1905: 19).
11. In many languages at least some intensifiers can only be com-
bined with a human focus. Examples are Turk. kendi, Jap.
zisin, Mand. ziji, Bengali nije, It. in persona. In other words,
here we find a situation which is closer to the original selectio-
nal restrictions. In other languages intensifiers can be assumed
to have extended their territory to foci of all types.
111. Some languages seem to have developed new expressions
which recreate, as it were, the original selectional restriction,
i.e. that only combine with expressions for persons of exalted
rank: hochstpersonlich, hochstselbst in German, selveste in
Norwegian, sahsf in Arabic, in persona in Italian, etc. That
these new expressions are more complex than the original in-
tensifiers, whose combinatorial potential had been extended, is
to be expected. This fact finds a parallel in the frequent phe-
nomenon that new, more complex local prepositions are creat-
ed once the older ones are primarily used in a temporal sense.
5. From intensifiers to reflexive anapbors
A wide variety of languages use the same expression both as intensi-

fier and as a marker of co-reference in a local domain, i.e. as reflex-
ive pronoun. In Europe this phenomenon is relatively rare, but it can
be found in the Finno-Ugric and Celtic languages and in English.
Outside of Europe the double use of the same expression both as in-
tensifier and as reflexive pronoun seems to be the majority pattern. It
is found inter alia in Turkic languages, in Semitic languages, in
many Caucasian languages, in Indic languages, in Persian, Mandarin
and many others. For most of these languages it does not seem to be
possible to identify a direction of change. Clear evidence for such a
direction is, however, provided by West Germanic languages such as
English or Afrikaans. In Old English there were no reflexive pro-
nouns and personal pronouns did double duty as markers of disjoint
reference and as markers of co-reference in a local domain. After
246 Elclcehard K6nig and Peter Siemund
modest beginnings in Old English and even in Middle English the

intensifier self was more and more used in combination with object
pronouns to indicate co-reference unambiguously:
(15) a. Hannibal ... hine selfne mid atre acwealde [King Alfred's
Orosius 4 11.110.2; Bately, 1980]
'Hannibal killed himself with poison. '
b. Judas se arleasa]Je urne Ha!lend bellEWde for ]Jam Iyoran
sceatte ]Je he lufode unrihtlice aheng hine selfne [Admo-
nitio ad filium spiritual em 1 9.25, Norman 1848]
'Judas the disgraceful who betrayed our Lord for that
wicked money that he loved unrighteously hanged himself. '
These combinations of personal pronouns and intensifiers were

later fused into one word and developed into reflexive anaphors in a
process that is still not completely understood. At the same time, the
original monosyllabic intensifier self was replaced by such com-
pounded forms, so that intensifiers and reflexive anaphors are identi-
cal in form, even if not in distribution, in Modern English. To this
piece of positive evidence for a direction of change from intensifier
to reflexive anaphor we can add the general argument that reflexive
anaphors are more strongly grammaticalized than intensifiers and
can also for this reason be assumed to be the target of a change con-
necting these two classes of expressions.
What we have not considered so far is the question of how a fu-
sion of personal pronouns and intensifiers can lead to reflexive pro-
nouns, i.e. result in expressions that mark co-reference in a local do-
main. Recall that in Old English a sentence of the form (16) could
either express co-reference or disjoint reference of the third person
pronoun:
(16) he acwealde hine

'he killed himlhimself
Now it is quite plausible to assume that an interpretation of dis-

joint reference was the preferred or unmarked option. Most activities
denoted by verbs in a language are preferentially other-directed, i.e.

the standard assumption is that these activities are outward directed,
away from the agent. The opposite assumption seems to be the un-
marked choice only for a very small subset of verbs: verbs of
grooming, of defending, preparing seem to be self-directed in the
standard case. So, whenever a simple personal pronoun appeared in
a sentence like (16), the standard interpretation was one of disjoint
reference. In order to indicate the marked option unambiguously, the
intensifier could be used in Old and Middle English and was stan-
dardly used from Early Modem English onwards. The relevant
meaning of the intensifier seems to have been the one discussed for
the adnominal use above (cf. Konig/Siemund 1996c; see Keenan
1996 for a different account). When an intensifier was added to the
pronoun, the referent was characterized as center and opposed to a
periphery. In the absence of any contextual information of the type
listed in (5) the center was clearly the agent, the referent of the sub-
ject, as opposed to other persons towards the relevant activity could
be directed. For hine in (16) to be interpreted as disjoint from the
subject after the addition of self, alternatives to such a referent would
have to be given in the context.
So far our assumption has been that the path from body parts to
reflexive anaphors necessarily involves intensifiers as an intermedi-
ate stage. There is, however, clear evidence that such an intermediate
step is not necessary. Expressions for body parts may lead directly to
reflexive markers. In Basque, for example, the reflexive marker bu-
rua 'head' is not used as an intensifier and similar facts have been
reported from other languages. Moreover, it is a well-known fact that
inherently reflexive verbs or reflexive uses of verbs may develop
from constructions with objects denoting part-whole relations as a
result of metonymic change. The following English example from
the end of the 18th centurY, which clearly has a co-referential, re-
flexive interpretation, is an interesting case in point:
(17) Women who have lost their husbands' affection, are justly re-
proved for neglecting their persons, and not taking the same
248 EkJcehard K6nig and Peter Siemund
pains to keep, as to gain a heart ... [Mary Wollstonecraft, The

Wrongs o/Woman, 1798]
6. From intensifiers to prominent roles
Intensifiers in Balto-Slavic languages exhibit an interesting type of

polysemy. The expressions for intensifiers are also used in the sense
of 'husband', 'lord/lady of the house', i.e. they may denote impor-
tant, prominent social roles.
(18) Lith. pats 'self, 'husband', pati 'wife'; Latv. pats 'self, 'Lord
of the house'
Similar usages are also found in Latin, Classical Greek and Rus-
sian, even if this does not seem to be a matter of polysemy.
(19) a. Russ. sam skazal 'the master/the husband said'

b. Lat. ipse/ipsissimus dixit 'the master said'
c. ClassGreek autos 'the master', 'the lord of the house', 'the
teacher'
That such usage is not strictly a matter of polysemy, i.e. of la

langue, is pointed out by Benveniste in the following quotation:
L'emploi de ipse pour le maitre de maison est un simple fait de 'parole', iI

n'ajamais atteint le niveau de la 'Iangue'. (Benveniste 1966: 302)
A similar usage can also be found in Hibemo-English as the fol-

lowing examples show:
(20) a. It's himself is going to speak.

b. Herself isn't too good again.
c. How is himself?
d. Herself isn't here right now. (Le. the person salient in a spe-
cific context)
The relevant changes in Balto-Slavic as well as the incipient

changes or extended uses are clearly based on ellipsis. The intensi-
fier itself expresses the central role of a contextually given focus. In
Irish English himself or herself can simply refer to a person who is
salient in a given context. The conditions for such changes or ex-
tensions in usage are nicely summarized by Benveniste in the fol-
lowing quotation:
Pour qu'un adjective signifiant 'soi-meme' s'amplifie jusqu'au sens de

'maitre', une condition est necessaire: un cercle ferme de personnes, subor-
donne a un personnage central qui assume la personalite, I' identite com-
plc~te du groupe au point de la resumer en lui-meme, a lui seul, ill'incarne.
(Benveniste 1969: 91)
The relevant condition is a hierarchically structured social group

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

250 Elclcehard K6nig and Peter Siemund
7. Intensifiers and scalar focus particles
In a wide variety of languages intensifiers, or related expressions, are

also used in the sense of English 'even', i.e. as scalar focus particles,
which characterize their focus as an extreme, or maxirnaIly unlikely
value for the open sentence which constitutes their scope. In a sen-
tence like the following George is characterized as a highly unlikely
candidate for the predication 'x will go to that party':
(22) Even George will go to that party.
A further aspect of meaning that distinguishes scalar focus parti-

cles from intensifiers is the 'additive' implication that alternatives to
the value given will make the sentence true, i.e. other people than
George will go to the party. This use of intensifiers or related ex-
pressions as scalar additive focus particles can be found in languages
like the following:
(23) Germ. selbst (but not selber); Norw. selv; Dutch zelJs; Fr.
meme
As is shown by this list, the relevant polysemy is a wide-spread

phenomenon among the languages of Europe. What the exact path of
the relevant extension or change is, is completely unclear. The only
relevant observation that can be made here is that there are contexts
in which intensifiers and additive scalar particles are interchange-
able. In the relevant contexts the denotation of the focus must be
characterized by the context as an extreme case for a predication that
is also made true by other values. Note that in the case of intensifiers
these aspects of meaning come from the context. All the intensifier
does is establish the usual relation between center and periphery:
(24) a. In such basic issues the Pope himself would not know what
to do.
b. (This picture is very valuable) The frame itself would cost a
fortune.
Such sentences clearly allow a paraphrase with even.
(25) a. Even the Pope would not know what to do.

b. Even the frame would cost a fortune.
Note that the additive implication is due to the conditional mood

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

b. Even the President will address the meeting.
8. From intensifiers to expressions of definiteness
As pointed out in (l3) above, intensifiers may develop into demon-

strative pronouns and further into definite articles. Intensifiers may
also develop into expressions indicating type identity, personal pro-
nouns and other anaphoric expressions. The following observations
will mainly be based on the Romance languages (cf. Ranoth 1990;
Selig 1992).
That the Latin intensifiers ipse developed into a demonstrative
pronoun is well-documented in the modem Romance languages, cf.
Cat. eix, Pg. esse, Sp. ese. In view of the fact that Lat. ipse is the re-
sult of compounding the Latin demonstrative pronoun is with the in-
tensifier -pse, the subsequent development of ipse into a demonstra-
tive pronoun is certainly not entirely unexpected.
It is also a well-known fact that demonstrative pronouns can de-
velop into definite articles. This can be observed in German, where
the articles der, die, das were originally used as demonstratives, but
also within the group of Romance languages itself. Here the source
of most modem definite articles can be found in Lat. ille, cf. Fr. le,
la; It. il, la; Pg. 0, a; Sp. el, la. The underlying use of ille that allow-
ed for this development is that as a determiner, not as a pronoun. We
are here confronted with a clear process of semantic bleaching. De-
monstratives are deictic expressions of definiteness, and when these
expressions lose their deictic component and only retain the ability
to pick out a unique referent in a certain discourse domain or world,
they develop into definite articles. Hand in hand with this de-
velopment has gone the emergence of an obligatory definite/inde-
finite marking on nominals.
Sardinian and those dialects of Catalan spoken on the Balearic Is-
lands did not use ille as source for the development of the definite
article, but derived it from ipse instead, cf. Sard. su, sa; Cat. es, sa.
Interestingly enough, the relevant dialects of Catalan have another
set of definite articles, viz. those derived from ille (el, la). Their use,
however, is mainly restricted to adverbials (a l'hora de la mort) and
proper names (l'Havana, el Papa). There are even some minimal
pairs. Thus, s 'esglesia refers to a particular church as a building
Intensifiers as targets and sources o/semantic change 253
whereas l'Esg/esia denotes the (Catholic) Church as an institution.

Also, es senyor means 'gentleman' or 'owner', but el Senyor is re-
served for 'the Lord' (cf. Hualde 1992: 281; Segui i Trobat 1993:
35). To base the defInite article on ipse appears to have been the
original strategy in the linguistic domain where Catalan is spoken to-
day. Those derived from ille were superimposed during the 12th and
13th century. The original situation is preserved in a number of
place-names: Sant Joan D~i, Collserola, Sant Esteve Sesrovires
(Segui i Trobat 1993: 33).
Another possible development of demonstrative pronouns is that
of 3rd person personal pronouns. Although most of the modem Ro-
mance languages used Lat. ille as the source of this development (Fr.
U, elle; Pg. ele, ela; Sp. el, ella), Sardinian derived them from Lat.
ipse, cr. isse, issa (Campidanese issu). As in the case of defInite
articles, this development can also be assumed to be due to semantic
bleaching. Note, however, that the relevant source are demonstra-
tives in their use as pronouns, not as determiners. Again, the deictic
component is lost and what remains is the referential function. It ap-
pears noteworthy that Sard. isse may also be used as a respectful
form of address and that it is by and large restricted to animate refer-
ents (cf. section 6). Apart from Lat. ipse, Lat. ille too left its imprint
on the Sardinian pronominal system. The clitic pronouns found there
are clearly derived from this demonstrative and are in fact preferred
in direct/indirect object positions (cf. Jones 1993: 201):
(27) a. ?Appo datu su dinari a issos.

b. Lis appo datu su dinari.
'I gave them the money.'
This does not mean that isse could not be found in object posi-
tions. However, if it is used there, it always implies some element of
contrast:
(28) Appo vistu a issos, ma no 'a tie.

'I saw them, but not you. '
254 Elckehard K6nig and Peter Siemund
Another possible path of development that intensifiers can take is

the one to expressions of type identity (Pg. mesmo, Sp. mismo). In
the French expression meme, we even find formal identity between
Lat. ipse and idem. In the course of this development, the Lat. encli-
tic pronominal intensifier -met became prefixed to ipsu (mismo <
me-ismo < medipsimu < -met ipsimu). What seems to have been
grammaticalized here is the discourse-anaphoric or discourse-deictic
function of ipse. This expression could regularly be used to pick up
information which had been established in the previous discourse.
Anaphoricity can be ~alyzed as the expression of identity, but this
is at first restricted to cases of token identity. When this expression
of identity is extended to also cover cases of type identity, we arrive
at the meaning of Sp. mismo in contexts like the following:
(29) Tengo el mismo coche que mi hermano.

'I've got the same car as my brother.'
In Sardinian we also find anaphoric uses of the pronominal forms

derived from Lat. ipse. Still, this use of Sard. isse appears to be re-
stricted to those positions where it is not a co-argument to its antece-
dent. In co-argument positions the reflexive clitic si is used invaria-
bly. In contrast to the reflexive clitic, isse in itself is not confined to
the co-referential interpretation (cf. Jones 1993: 241):
(30) Juannej credet Id Gavinij I'at comporatu pro isseiljlk.

'John thinks that Gavin bought it for himlhimself.'
The co-referential interpretation can be forced by adding an inten-

sifier (e tottu (lit. 'and all') or mattessi 'same') to the anaphoric pro-
noun. This strategy to restrict the binding domain of an anaphoric
expression is strikingly parallel to what we find in Old English, Afri-
kaans or Frisian (cf. section 5).
(31) Juannej credet ki Gavinij l'at comporatu pro isse e tottu/isse

mattessij.
'John thinks that Gavin bought it for himself.'
Intensifiers os targets and sources ofsemantic change 255
9. Conclusion
However tentative some of the observations and asswnptions are

that were made above, the picture that emerges provides further evi-
dence for the view that semantic change is not as random and unpre-
dictable as it is often asswned to be. The development of minor lexi-
cal categories, in particular, exhibits a great deal of similarity across
languages and seems to follow one of a limited nwnber of possible
paths. The processes of semantic change exhibited by the develop-
ment of function words are partly those found in the development of
major lexical categories, but there is also a certain element of gen-
erality, predictability and unidirectionality not found in the develop-
ment of nouns, verbs or adjectives.
References
Baker, Carl Leroy

1995 Contrast, discourse prominence, and intensification, with special
reference to locally free reflexives in British English. Language 71:
63-101.
Bately, Janet (ed.)
1980 The Old English Orosius. Oxford: Oxford University Press.
Benveniste, Ernile
1966 Problemes de linguistique generale. Paris: Gallimard.
1969 Le vocabulaire des institutions indo-europeennes, Volume I. Paris:
Editions de Minuit.
Blank, Andreas
1997 Prinzipien des lexi/calischen Bedeutungswandels am Beispiel der ro-
manischen Sprachen. Tilbingen: Niemeyer.
Edrnondson, Jerold and Frans Plank
1978 Great expectations: An intensive self analysis. Linguistics and Phi-
losophy2: 373-413.
Farr, James
1905 Intensives and Refle:cives in Anglo-Saxon and Early Middle English.
Baltimore: J .H. Furst.
Grlmm, Jacob
1967 Deutsche Grammatik. Hildesheim: Georg Olms Verlagsbuchhand-
lung (reprint).
256 Ekkehard K6nig and Peter Siemund
Hualde, Jose Ignacio

1992 Catalan. London: Routledge.
Jones, Michael Allen
1993 Sardinian Syntax. London: Routledge.
Keenan, Edward L.
1996 Creating Anaphors. An Historical Study of the English Reflexive
Pronouns. UCLA, ms.
Koch, Peter
1997 Ein Blick auf die unsichtbare Hand: Kognitive Universalien und hi-
storische romanische Lexikologie. In: Thomas Stehl (ed.), Unsicht-
bare Hand und Sprechwahl. Typologie und Prozesse des Sprach-
wandels in der Romania. TUbingen: Narr.
KOnig, Ekkehard
1991 The Meaning of Focus Particles: A Comparative Perspective. Lon-
don: Routledge.
KOnig, Ekkehard and Elizabeth Traugott
1988 Pragmatic strengthening and semantic change: The conventionaliz-
ing of conversational implicature. In: Wemer HUllen and Rainer
Schulze (eds.), Understanding the Lexicon. TUbingen: Niemeyer.
KOnig, Ekkehard and Peter Siemund
1996a Selbst-Reflektionen. In: Gisela Harras (ed.), Wenn die Semantik
arbeitet. Festschrift for Klaus Baumgartner, 277-302. TUbingen:
Niemeyer.
1996b Emphatische Reflexiva und Fokusstruktur. In: Inger Rosengren
(ed.), Sprache und Pragmatik 40, 1-42. Lund: Lunds Universitet.
1996c On the development of reflexive pronouns in English: A case study
in grammaticalization. In: Uwe BOker and Hans Sauer (eds.), Ang-
listentag 1996 Dresden, Proceedings, 95-108. Trier: Wissenschaft-
licher Verlag.
Moravcsik, Edith
1972 Some cross-linguistic generalizations about intensifier constructions.
Papers from the Regional Meetings ofthe Chicago Linguistic Socie-
ty 8: 271-277.
Norman, H.W.
1848 The Anglo-Saxon Version of the Hexameron of St. Basil ... and the
Anglo-Saxon Remains of St. Basil's Admonitio ad filium spirilu-
alem. London.
Romoth, Susanne
1990 Die IdentiUJtspronomina in der Romania. Geneve: Droz.
Seguf i Trobat, Gabriel

1993 Iniciacio a la morfosintaxi catalana. Palma: Aina Moll.
Selig, Maria
1992 Die Entwick/ung der Nominaldeterminanten im Spatlatein. Roma-
nischer Sprachwandel und lateinische Schriftlichkeit. TObingen:
Narr.
Traugott, Elizabeth Closs
1985 On regularity in semantic change. Journal of Literary Semantics 14:
155-173.
1989 On the rise of epistemic meanings in English: An example of subjec-
tification in semantic change. Language 57: 33-65.
1995 Subjectification and grammaticalization. In: Dieter Stein and Susan
Wright (eds.), Subjectivity and Subjectivisation in Language, 31-54.
Cognitive ease and lexical borrowing:
the recategorization of body parts in Romance 1
Thomas Krefeld
o. The Romance challenge
True problems will never disappear - science loves them too much.
Romance Linguists, for instance, have been fascinated by the paral-
lelism in the evolution of different Romance languages ever since
the birth of their science; this parallelism has so much more appeal,
because it is essentially of a typological or structural order, without
always implicating the etymological identity of the linguistic mate-
rial. Whereas diachronic morphosyntax has accepted this descriptive
(and explanatory) challenge, semantics, specially lexical semantics,
have for the most part remained reluctant. The cognitive orientation
could though stir up this state of things.
1. The parts of the human body and the

Torso-Extremities-Model
From a cognitive point of view the categorization of those domains

that do not depend, at least in an indirect way, on the form of social-
cultural life, merits special curiosity. As a field of application, let us
take the designations of the parts of the human body excluding, of
course, the anatomical nomenclature, which would not be represen-
tative of the pre-scientific and naive lexical categorization of the
"natural" spoken language. Our languages conceive the human body
as a whole, more precisely as an organized group of high com-
plexity; this group consists of parts, some of which are conceived as
natural sub-groups, which themselves show a higher or lesser degree
of complexity.3
260 Thomas Krefeld
French, for example, distinguishes the following groups (corps)

and sub-groups (parties): tete 'head', cou 'neck', fronc 'torso',
extremites (membres); further, certain joints linking the sub-groups
themselves (hanche 'hip,)4 and other joints linking certain parts of
the sub-groups (as genou 'knee', cheville 'ankle', coude 'elbow',
poignet 'wrist'). I will provisionally call this type of categorization
the "Torso-Extremities-Model".
Figure 1. The Torso-Extremities-Model (French version)
From an onomasiologic point of view, the sub-categorization of

the constitutional parts (or sub-groups) depends on the following cri-
teria:
• the respective vertical position (superior vs. inferior);

• the direction of view (front, rear, lateral);
• the exposition to view in an immobile position (exterior vs. inte-
rior).
Cognitive ease and lexical borrowing 261
The respective vertical position is fundamental. We find it again

in the sub-categorization of all parts (all except the neck). The direc-
tion of view is important for the sub-categorization of the head, the
neck and the torso, but is quite insignificant with regard to the extre-
mities (in French, only jarret 'hollow of the knee' and talon 'heel').
The least important, but at the same time the most specific, is the
exposition to view, which concerns the neck (gorge 'throat') and,
only in a very marginal way, the extremities (aisselle 'armpit').
It is therefore interesting to see that the French type of sub-cate-
gorization, reflects to a certain extent, from joint to joint so to speak,
the articulation of the skeleton. For example, the lexical categoriza-
tion of the arm is analogue to that of the leg, according to the ana-
logy of their physical constitution. Compare:
• sub-group bras; comprising of coude 'elbow', poignet 'wrist',

main 'hand', doigt 'finger';
• sub-group jambe; comprising of genou 'knee', cheville 'ankle',
pied 'foot', orteils (doigts) 'toe'.
Noteworthy are the synonymous names of the toe (orteilldoigt):

the lexicalization of a specific designation (ortei/) shows that the
"anatomic principle" cannot be completely neglected. The existence
of the less specific term (doigt), however, shows the dominance of
the "organic principle". The difference of their organic functions is
obviously more important than their anatomic analogy. The leg is de-
fined as a sub-group just on account to its main function, which is
assuring the movements by involving the functions of all of its parts.
For the arm, though, this is not the case, because the foot (pied) and
the hand (main) do not at all have analogous organic functions. Due
to its functional importance, the hand must be considered a supple-
mentary sub-group, which is not the case with the foot. This is also
the reason why the word for the foot frequently refers to the entire
leg (Rum. picior, Germ. dial. Fuj3 'foot', 'leg'), whereas the name of
the hand very rarely seems to refer to the entire arm.
262 Thomas Krefeld
Note that the relative autonomy of the hand and the resulting
functional differences between the ann and the leg are, according to
the envisaged language, more or less evident. At first glance, one
might object that the lexical categorization of French (as well as of
Romance and already of Latin) would correspond to the skeletal
symmetry ("anatomic principle") because the articulations of the
hand and the foot have the same names: Lat. digitus means 'finger'
as well as 'toe', and so do Fr. doigt, It. dita, Rum. deget etc., which
can be, if necessary, specified as Fr. doigt du pied, It. dita del piede,
Rum. deget de la picior etc. However, one must take into account
the fact that the designations of the hand were transposed to that of
the foot, and not vice versa.
Above all, we ascertain the existence of individual designations
for each of the different fingers (in French: pouce 'thumb', index
'forefinger', medius 'middle finger' [not really popular], annulaire
'ring finger', auriculaire 'little finger'), the inside part of the hand
(paume 'palm') and the joints between the fmgers and the hand
(nceuds 'knuckles'). Some of these designations (those of the most
salient fingers) have been transposed to the articulations of the foot,
despite their having completely different organic functions:
• Lat. pollex 'thumb' and pollex pedis;

• Romansh polisch and polisch dil pei;
• Fr. pouce and pouce du pied;
• It. pollice and pollice del piede (as well as alluce);
• It. mignolo 'little finger' and mignolo del piede.
2. The etymological point of view: the non-Latin element
Looking at the fundamental importance of the body for human orien-

tation in the world, we are not surprised to note a strong Latin-Ro-
mance continuity in this lexical domain; it suffices to cite the quasi-
Pan-Romance designations of the EAR (type Lat. auricula), the NOSE
(type Lat. nasus), the EYE (type Lat. oculus), the LIP (types Lat. labi-
um, labrum), the TEETH (type Lat. dens), the TONGUE (type Lat. lin-
gua), the HAIR (type Lat. capillus), the SKIN (type Lat. pellis), the
STOMACH (type Lat. stomachus), the RIB (type Lat. costa), the HAND
(type Lat. manus), the FINGER (type Lat. digitus) etc.
However this essentially basic lexicon contains a pile of lexemes
of non-Latin origin, as e.g. in French: joue 'cheek' < pre-Lat. *ga-
bota (FEW 4: 9-10); nuque 'neck' < Lat. (medical) nuca < Arab.
nukha; echine 'vertebral column' < Frankish *skina; flanc < Fran-
kish *hlanka; hanche < Frankish "'hanlca; bras < Greek-Lat. brac-
(c)hium;jarret < Gallic "'garra;jambe < Greek-Lat. campa.
It has to be emphasized that French, which often plays a soloist
role, this time is in harmony with the other Romance languages. The
absence of Latin elements is not particularly specific to French; the
situation, e.g., of Romansh (whose lexicon is often archaic) and Itali-
an are both analogous, even despite the fact that Italian belongs to
what Damaso Alonso (1978) called the "Romania continua". Even
Rumanian, the other great soloist among the Romance Languages,
behaves in an identical manner. If the etyma of the loan words
integrated in French, do not appear in the three other languages, we
find other terms of equally non-Latin origin or, at least, Latin ele-
ments that are more or less isolated and regional and which can be
found nowhere else. If, inversely, French distinguishes itself by a
geographically isolated Latin lexeme, the other languages, in con-
trast, present borrowings.
The following list shows a sample of non-Latin elements among the
French names for body parts and their Italian, Rumanian and Ro-
manshs equivalents:
1. CHEEK: Fr. joue < pre-Lat. *gabota vs. Romansh gauta <
Gallic galta; It guancia < Longobardic *wankja; Rum. obraz
< Slavic obrazu.
2. NECK: Fr. nuque (It. nuca) < Arab. nukha vs. Romansh totonal
tatona < ? (without doubt pre-Latin); Rum. ceafa < perhaps
Turk. (cf. Alban. qafe).
264 Thomas Krefeld
3. VERTEBRAL COLUMN: Fr. echine < Frankish *skina vs. Ro-

mansh spina dorsola < Lat. spina; It. schiena < Long. *skena,
Rum. spina < Lat. spina.
3. FLANK: Fr.j1anc (It flanco) < Frankish *hlanka vs. Romansh
/omau (Surselv.) / lama (Engd.) derived from/om 'faim'.
4. HIP: Fr. hanche < Germanic *hanka vs. Romansh calun (Sur-
selv.)/ gialun, chalun (Engd.) < pre-Lat. *calon-I*galon- (cf.
HR I: 144); It. anca < Longobardic or Frankish *hanka; Rum.
~old < Transsilv.Germ. scholder.
5. ARM: Fr. bras, Romansh bratsch; It. iI braccio; Rum. bra! <
Greek-Lat. brac(c)hium.
6. NUCKLE: nlEud < Lat. nodus vs. Romansh dartugf' < Lat. *ar-
tuculum < articulum (cf. HR I: 242); It. nocca < Longobardic
knohha; Rum. incheietura degetelor.
7. HOLLOW OF THE KNEE: Fr. jarret (It. garretto, besides cavita
del ginocchio) < Gallic garra; also Romansh garlet (Surselv.)/
giarlet (Engd.) < Gallic *garrulu de *garra; 1 vs. Rum. scobi-
tura genunchiului.
8. LEG: Fr. jambe; Romansh comba (Surselv.)/chomma (Engd.);
It. gamba < Greek-Lat. kampe vs. Rum. picior < Lat. petiolus.
Given the heterogeneity of the origins - there are words of sub-

stratum languages (Gallic, pre-Latin), of superstratum languages
(different Germanic languages, Slavic) and of adstratum languages
(Arabic)8 - and considering the long duration of development, we
must ask ourselves why Latin and the Romance languages adopted
so many loan words to express this very fundamental conceptual do-
mam.
3. The motivation for the borrowings
The sociolinguistic situations, that is to say the situations of lin-

guistic contact to which this handful of words is submitted, are so
divergent that it would be absurd to imagine one sole historical ex-
planation. This doesn't exclude that certain borrowings nevertheless
can be explained from their cultural context, i.e. by the cultures of

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

body parts.
When we reconstruct the corresponding lexical field in Latin, we
immediately meet with a somewhat contradictory categorization:
1,2,3,4 Lat. membra

3,4 Lat. artus, Fr. membres
a. ARM and LEG missing in Latin b. EXTREMITIES and BODY PARTS in Latin
Figure 2. The idiosyncratic categorization of body parts in Latin
We have, at one end, the rather subtle anatomic opposition be-

tween two collective terms:
• Lat. membra 'constituent parts of the body, extremities and head

and torso' (G IT: 863-864);
• Lat. arlus 'articulated extremities' (G I: 603; cf. the latinism It.

arlo).
At the other end, there apparently weren't any specific tenns for
the sub-groups of the superior and inferior extremities; that is to say,
there is neither a word for the entire arm, nor for the entire leg. It is
obvious that the Torso-Extremities-Model is only superficially root-
ed in Classical Latin (by the existence of arlus), but in the course of
history it grew stronger and stronger, operating on two levels:
1. The opposition of membra vs. arlus is neutralized in the way that

membra takes the signification of arlus 'articulated extremities'.
The older sense of membra 'constituent body parts including
extremities and torso' looses its lexical representation; it is "de-
lexicalized" .
2. The two lexical "gaps", concerning the designations of the ann
and the leg, were filled in, in different eras, but in a similar way.
3.1. The ARM
With regard to Fr. bras, we observe that Greekppax[rov 'ann in gen-

eral' (Gemoll: 163) was borrowed by Latin at quite an early moment.
If this had not been the case, this originally Greek element wouldn't
have got its Pan-Romance character (G I: 859; REW: 1256).
I would like to emphasize the fact that this borrowing doesn't just
fill in a lexical gaplO, the loan word brac(c)hium is part of a rather
complicated process of re-categorization. The Latin lexicography
shows that the Greek import brac(c)hium designates the arm in gene-
ral, and the forearm in particular. In Romance, we find something
quite curious, because the same tenn not only stands for the arm in
general, but also for the upper arm in particular. Why is this so?
Classical Latin uses several names for designating, among others,
the upper arm:
268 Thomas Krefeld
• armus 'upper part of the upper arm and the shoulder, upper arm'
(G I: 582);
• (h)umerus 'bone of the upper arm; the section of elbow right to
the collarbone and the shoulder blade, i.e. upper arm and shoul-
der' (G I: 3294);
• lacertus 'muscles; muscles of the upper arm; the strong upper
arm, with many muscles' (G 11: 527).
We cannot, a priori, exclude the possibility that the loan word

brac(c)hium would have completed this paradigm. But looking at the
very weak presence of the three cited Latin words in the Romance
languages, at least as designations of the arm (or upper arm), II we
should rather ascertain that brac(c)hium absorbed their function, or
more exactly, a part of their function: the loan word only replaces
them to the extent as they refer to the arm. It is true that the chrono-
logy can't be reconstructed in a reliable way and, of course, brac-
(c)hium didn't make the three tenns disappear immediately; never-
theless, we can suspect, going by what we have said, that the three
Latin words might have disappeared, because they didn't refer to the
organic sub-group arm, nor to a constituent part of it, but to ana-
tomic sections that don't correspond at all to the Torso-Extremities-
Model, because the three obsolete Latin tenns designate at the same
time an upper part of the torso (the shoulder) and a part of the (up-
per) arm.
The most illustrative example in this respect is the case of (h)u-
merus that was best maintained: this word doesn't refer anywhere in
Romance to the shoulder and the arm (except the anatomy technique
tenn It. omero) but orily to the shoulder. This semantic restriction
manifests very clearly the adaptation to another type of categoriza-
tion.
All those Romance languages that didn't manage to adapt the
sense of (h)umerus, have abandoned it completely. An evident result
of this process is the missing of a tenn for the upper part of the torso.
The new "gap" is compensated for by extending the referential range
of a contiguous anatomic tenn, i.e. Latin spatula (> Fr. epaule, It.
spal/a; cf. G 11: 2746). This Latin word referred to the shoulder blade
in a very precise way, and therefore to a part of the skeleton linked to
the shoulder. 12 The example makes evident that it doesn't suffice to
note a metonymic change, that bases itself on a relation of contiguity
(shoulder blade and shoulder), but that we also have to explain why
other, equally contiguous tenns have not been metonymically ex·
tended. Remember that Lat. lacertus, which already meant 'upper
ann' and 'shoulder' would have been an "ideal" candidate for met·
onymy. It did not succeed - I should say: it could not succeed -
because the successful concurrent spatula had the advantage of
being in hannony with the categorization of the Torso·Extremities·
Model. Latin spatula designates a part of the torso and only of the
torso, whereas lacertus designates a part as well of the ann as of the
torso.
3.2. The LEG
In the domain of this sub·group, we have to deal with a situation

somewhat analogous to that of the sub·group ann. Classical Latin
presents an opposition between:
• femur 'thigh, femur (thighbone)' (G I: 2715) and

• crus 'lower part of the leg' (G I: 1774·1775).
From the Romance perspective, there are two things to remember:

firstly, neither the one nor the other tenn survived in Romance 13 and,
secondly, all the Romance languages have a word to designate the
entire leg, or the sub·group leg, as we prefer to call it. We therefore
observe the same type of re·categorization.
But unlike the re·categorization (of the designations) of the ann,
the Romance languages show different solutions. Iberoromance
chose a "satellite" word, i.e. a word that in Classical Latin already
designated a part of the leg and which was therefore in metonymic
270 Thomas Krefeld
relation with the two lost words: Lat. perna 'hip.I4 and primarily
'ham of the posterior thigh' (0 II: 1618-1619), which gave Cat./Pg.
perna, Span. pierna.
Rumanian is characterized by a Latin diminutive petiolus 'small
foot, small leg (of a lamb etc.)' (0 II: 1670) based on pes, which
gave picior. Let us remember that this Rumanian word means 'leg'
and 'foot'.
Sardinian, Corsican and certain Mediterranean Italian dialects
adopted the feminine form anca of the Latin adjective ancus 'bene
(REW: 446; 0 I: 422).
However, in Central Romance, it is still a borrowing dating back
already to the Latin period, Oreek KaJlml 'curve, inflection' (Oemoll:
406; 0 I: 947) that gave Fr. jambe, It. gamba, Romansh comba/
chamma etc. The FEW, s.v. cam ba, states that this word was a tech-
nical term of veterinary medicine designating the 'ankle joint of a
horse, hock' (FEW II: 119). But one must say that the FEW's com-
mentary is not apt to instruct us about the process of borrowing be-
cause it doesn't offer any outline of the semantic background against
which it unfolds.
Without any doubt, Wartburg is right to discuss the case of camb
in the context of other Romance words that took the meaning of
'leg'. He resumes that the words that substituted crus are all taken
from the coarse names of the animal body. IS Thus, his reconstruction
focuses only on the "vulgar" aspect. In his Einfohrung in Prob-
lematik und Methodik der Sprachwissenschaft (1962), he continues
with comments on the same example, saying that "this apparition of
slangy (in Oerman, burschikos), familiar and expressive words, in
general characterizes the late empire" (1962: 117).
In my opinion, this quite stereotypical argumentation is, despite
its explanatory power, not very helpful in this case. Firstly, it makes
an abstraction of the Rumanian and Sardinian solutions, further, it
doesn't take into acco~t that Classical Latin perna also signifies
'hip' and, maybe, 'leg' and, at last, it doesn't consider the fact that
the Latino-Oreek camba was far from being exclusively a term of the
veterinary medicine. The word had the very general sense 'bent',
already attested in Latin in Plautus (0 I: 947) and thus is a synonym
of the adjective ancus, which lead, as we have seen, to other Ro-

mance designations of the leg (Sard., Cors. anca 'leg') and which
doesn't seem to be vulgar. Elsewhere, a technical veterinary term by
itself has no rude connotation. It could, eventually though, receive
this connotation but on the condition that it leaves the technical do-
main. On the level of the same variety, vulgarity (or expressivity)
and technicality are mutually exclusive.
Finally, one must not forget that a name for the entire leg was
missing in Classical Latin, just as it was for the arm, and that a lex-
erne signifying 'bent' would actually lend itself ideally, to "mend-
ing" the lexical network.
3.3. TheHJP
I have now to treat the most common of the loan words in this
domain, which is a Germanic designation of the hip. Walther von
Wartburg discussed the problem in his already cited Einfohrung
(1962: 117-118); his explanation can be summarized as follows: the
Latin designations of the inferior and superior parts of the leg (crus
andfemur) have disappeared; in the case offemur, it resulted from
the homonymie genante with jimus 'animal waste, manure', caused
by the fact thatjimus had changed its declension under the influence
of its synonym stercus (fimus > femus, -oris). Thus, the formal
changing of a word designating 'animal waste, manure' would have
made the designation of the thigh obsolete. It would result a lexical
gap, provisionally filled by the metonymic extension of the designa-
tion of the contiguous part: the Latin coxa 'hip' would have taken
the additional meaning of the thigh (> Fr. cuisse, It. coscia), desig-
nating the segment of the leg from the hip right to the knee. Unlike
the disappearance of femus, the semantic change of coxa wouldn't
end with an empty case, but only with a situation of "semantic dis-
tress" (in German, Notlage), released finally by the integration of the
Germanic *hanka 'hip' (> Fr. hanche, It. anca) , that was "heard
from time to time from the mouths of Germanic soldiers and colo-
272 Thomas Krefeld
nists" (118). Two aspects of this rather mechanical explanation give

rise to the following critical remarks:
1. Lat./emus, -oris 'animal waste, manure' plays a marginal role,

mainly because it has survived only in Old French, Occitan
and Catalan (REW: 3311). It isn't very "elegant" to suppose
that a word of a quite reduced spatial distribution would have
set off a more or less Pan-Romance process: Lat. coxa didn't
conserve anywhere (except Engd. cossa; cf. HR 11: 640) the
meaning of 'hip' and its supposed Germanic substitute *hanka
is again spread out over a vaster territory (Italy, the Iberian
Peninsula).
2. It is not very convincing to assume the metonymic extension
of coxa and to pretend at the same time that this metonymy is
immediately replaced because it brought "confusions" (Un-
klarheiten) with it. One would at least have to specify of what
nature these confusions could be; surely they are qualitatively
of a type different from those much less probable confusions
that provoked the homonymy between *femus, -oris 'animal
waste, manure' and/emur, -oris 'thigh': In this last case, all
metonymy appears excluded; there is hardly a linguist who
will be able to imagine a context in which a risk of confusion
could happen.
In fact, the metonymy between hip and thigh is so trivial that it

isn't worth searching for a justification and, in terms of the Romance
results, it probably already existed in the Latin of the Roman Em-
pue.
Still, it remains curios that coxa has completely lost the broad
sense 'hip' and 'thigh' (corresponding to the sense of Lat. perna
which underwent a similar modification). The best witness to the ob-
ligatory nature of this semantic restriction is the situation of Ro-
mansh, where reflexes of Lat. coxa are attested in a closed geogra-
phic area. The semantics are now strictly alternative and the word
means either 'thigh' (queissa in Surselvan) or 'hip' (cossa in Enga-
dine dialects), but has never both meanings (HR 11: 640).
It goes without saying that this type of alternative semantic reduc-

tion is not due to an unavoidable tendency towards more lexical pre-
cision: Rum. picior 'leg' and 'foot' has shown us that Romance lexi-
calization, in this domain as in others, also worked in the opposite
direction. The Rumanian word (of Latin origin) integrates two senses
in metonymic relation, although in Classical Latin, two different
forms (crus vs. pes) already existed.
The case of coxa (i.e. the alternative semantic reduction and the
parallel integration of Germanic *hanka) is unsuitable for illustrating
the traditional rule: phonic accidents may provoke semantic confu-
sions. We prefer to see a confirmation of re-categorization as
sketched out above. The existence of a lexical category such as Lat.
coxa that designates the section of the human body beginning at the
hip and reaching to the knee doesn't fit into the Torso-Extremities-
Model. According to this categorization, the hip is a part of the sub-
group of the torso, whereas the thigh is a constitutive part of the sub-
group of the leg.
4. Conclusion: recategorization and cognitive ease
From an anatomical point of view, the Latin way of categorizing the

human body (see above fig. 2a and 2b) might be called ingenious;
certain terms refer to very complex organic compounds as the UPPER
ARM-SHOULDER- or THIGH-HIP-categories. This categorization, never-
theless, suffers from a fundamental disadvantage: it does not corres-
pond with the perceptive saliency of the arms and the legs in visual
appearance of the human body. Arms and legs are without any doubt
cognitively privileged. It suffices to recall children's drawings (fig.
3a and 3b) which in their first stages (see fig. 3a) even omit the
torso.
274 Thomas Krefeld
a. Very early stage b. Latter stage
Figure 3. Perceptive saliency of the arms and the legs in children's drawings
Based on this, the presence of many non-Latin lexemes of com-

pletely heterogeneous origins is no longer surprising; it has to do
with lexical categories that have appeared in Romance languages
completely by chance, more precisely, by the contingent existence of
a more or less important bilingual minority. These lexical categories
succeeded in being borrowed by the unilingual majority because
their concepts are imposed by perceptive experience. However, the
re-categorization evidently wasn't introduced all at once by means of
a specific loan situation (or of one specific language of contact); it
occurred thanks to the contribution of many languages, step by step
along history's path. The cognitive perspective couldn't justify the
appearance of a bilingual population, but it enables us to explain the
integration of borrowings, i.e., its acceptation by a community of
non-bilingual native speakers. In other words, these borrowings are
neither luxury nor necessity borrowings but moreover borrowings of
cognitive ease.
Notes
1. I am grateful to Max Krefeld for drawing the figures 1, 2a, 2b.

2. For rich lexical material cf. Zauner 1894.
3. For further philosophical details see Husserl 1992: 227-300.
4. Curiously enough, there is no word for the SHOULDER JOINT, Fr. aisselle
meaning only 'armpit'.
5. Romansh is the name for the Swiss varieties of Rhaeto-Romance. The
etymologies are due to Gamillscheg, HR, Tiktin, REW and DELL
6. This word doesn't have the meaning 'joint' which is lexicalized by lisUra
(Engd.; cf. HR I: 438; < 1) and by giugadira (Surselv.; cf. HR I: 368) de-
verbal noun. of giugar 'to play'.
7. The type Fr. ja"et/Romansh g(i)arletlIt. ga"etto has many meanings; it
still designates the Achilles tendon (Romansh), the heel of the horse (Itali-
an) where the jointure "bends the behind leg" at the mammiferous ungu-
lates. (Robert: 1042)
8. The "substrata" just as the "superstrata" are at the time of the borrowing
evidently adstrata of Latin, resp. of Romance; the terminology isn't
particularly happy since the senses of the two terms oscillate between the
relative chronology (before vs. after the romanization) and the socio-
political status (dominating vs. dominated speech communities).
9. The terms of Luxuslehnwort 'luxury borrowing' and Bedurfnislehnwort
'necessity borrowing' have been coined by the Swiss dialectologist Emst
Tappolet; cf. Tagliavini 1998: 214.
10. This explanation goes perhaps for the case of the loan words designating
the cheek (as Fr. lajoue), that doesn't seem to have had a proper name in
Latin: the meaning of Lat. gena wasn't simply 'check', but rather 'part of
the face between the forehead, the temple and the chin, including the eyelid
and the eye-socket' (Georges I: 2913)
11. er. Lat. armus > Rum. (dialectal) arm 'cuisse des animaux', ORum. 'Goin-
ture) of the hip'; cf. Tiktin I: 214-215; REW: 4822; the relatively well
established Lat. (h)umerus survives, but exclusively in the signification
'shoulder' (Rum. umar, It. omero, Sp./Pg. hombro; cf. REW: 4232).
12. The fact that there exists in Latin a lexicalized metaphoric use (the word
also designates a type of spoon) is without importance in our context.
13. The REW doesn't give any reflex.
14. Unfortunately, G: 1618, is not very precise in its defmition.
15. "Therefore some coarse words came into use for 'leg', taken from animal
bodies" (FEW 11: 119); cf. also Wartburg (1962: 117).
276 Thomas Kreleld
References
Alonso, Damaso
1978 Die Ausgliederung der westromanischen Sprachen. In: Reinhold
Kontzi (ed.), Zur Entstehung der romanischen Sprachen, 163-186.
Dannstadt: Wissenschaftliche Buchgesellschaft.
Bemardi, Rut
1994 Handw6rterbuch des Rdtoromanischen: Wortschatz aller Schrift-
sprachen, einschliesslich Rumantsch Grischun, mit Angaben zuf
Verbreitung und Herlcunft. 3 Volumes. ZUrich: Offizin. (= HR)
Cortelazzo, Manlio and Zolli, Paolo
1979 Dizionario etimologico della lingua italiana. 5 Volumes. Bologna:
Zanichelli. (= DELl)
Gamillscheg, Emst
1934 Romania Germanica. Sprach- und Siedlungsgeschichte der Germa-
nen aul dem Boden des alten R6merreichs. 3 Volumes Berlin /
Leipzig: de Gruyter.
Gemoll, Wilhelrn
91965 Griechisch-deutsches Schul- und Handw6rterbuch. MUnchen /
Wien: Freytag.
Georges, Karl E.
81976 Ausftihrliches Lateinisch-deutsches Handw6rterbuch. 2 Volumes.
Hannover: Hahn [1913]. (= G)
Husserl, Edmund
1992 Logische Untersuchungen. Hamburg: Meiner (= Gesammelte Schrif-
ten 3.).
Meyer-LUbke, Willielrn
3 1935 Romanisches etymologisches W6rterbuch. Heidelberg: Winter.
(=REW)
Robert, Paul
1984 Le petit Robert. Dictionnaire alphabetique et analogique de la lan-
guefram;aise. Paris: Le Robert.
Tagliavini, Carlo
21998 Einftihrung in die romanische Philologie, TUbingen / Basel: Francke
[1973).
Tiktin, Hariton
21986 Rumdnisch-deutsches Worterbuch. 2., ilberarbeitete und ergtlnzte
Auflage von Paul Miron. Wiesbaden: Harrassowitz.
Wartburg, Walther von

21962 Einfohrung in Problematik und Methodik der Sprachwissenschaft.
TObingen: Niemeyer.
Wartburg, Walther von
1928ss. Franzosisches etymologisches Worterbuch. Leipzig / TObingen /
Basel: Zbinden. (= FEW)
Zauner, Adolf
1894 Die romanischen Namen der K6rperteile. Romanische Forschungen
14: 339-530.
Cognitive aspects of semantic change
and polysemy: the semantic space HAVE/BE
Peter Koch
1. The semantic space HAVE/BE
At least since Lyons's famous (1967) article, linguists have been

interested in the interlacements of possessive, existential, and loca-
tive predicates from a cross-linguistic perspective. Localist and/or
generative approaches, accounting for the relevant interlacements in
terms of a unitary underlying structure, 1 have been challenged by
typological and/or cognitive approaches which, while acknowledg-
ing the lexical diversity of natural languages, nevertheless make cau-
tious semantic generalizations, where cross-linguistic insights enable
us to do so (cf. Bickerton 1981; Wilson 1983; Hengeveld 1992; Hei-
ne 1997; Feuillet 1998). The point I want to stress here is the impor-
tance of the diachronic evidence from semantic change and - indi-
rectly - from polysemy,2 which allows us to formulate cognitively
valid generalizations even in lexical semantics.
In the area HAVE/BE, as I term it for the sake of convenience,
Bickerton (1981: 244-246) proposes a kind of "semantic space" like
Figure 1 in order to account for the joint lexicalization of different
semantic categories on the bases of adjacency:
OWNERSHIP LOCATION
POSSESSION EXISTENCE
Figure 1. The semantic space HAvE/BE (after Bickerton 1981: 245)

280 Peter Koch
In the meantime, this fourfould schema has turned out to be nei-

ther sufficient nor fully adequate for providing all the adjacencies we
need (cf., e.g., Wilson 1983). Furthermore, the theoretical basis of
'adjacency' itself is not quite clear. Is it a relation of adjacent storage
in our mind (whatever that may mean)? Is it a relation between cate-
gories of the same conceptual taxonomy (so that from a diachronic
perspective we could really speak of "extension" from one category
to the other)? Is it a relation of elements of the same conceptual
frame (so that we could speak of conceptual "contiguity": cf.
Note 8)? Above all, does it always represent the same kinds of rela-
tionship (so that OWNERSHIP: POSSESSION = OWNERSIllP : LOCATION
etc.)?
In Koch 1993, I proposed a refined onomasiological grid (Fig-
ure 2) for the semantic space HAVE/BE. First of all, I distinguish four
conceptual categories, which, albeit in varying combinations and
constellations and with differences in terminology, regularly occur
in publications on central conceptual distinctions in verbal and
predicate semantics (cf. Lyons 1967; Pottier 1974: 109-116; Clark
1978; Bickerton 1981: 245; Koch 1981: 260-276; Hagege 1982: 46;
Wilson 1983; Hengeveld 1992: 73-126; Heine 1997; Feuillet 1998:
670-673}:3
• with only one participant x:

1. ASCRIPTION, as in example (1);
2. EXISTENCE; as in example (2);
• each with two participants x and y:
3. possession, as in example (3); x = the possessed;
y = the possessor;
4. location, as in example (4): x = the located; y = the locality.
(I) Engl. John'A.Is ill. ASCRIPTION

(2) Engl. THERE ARE many unhappy peop/ex. EXISTENCE
(3) Engl. Johny HAS a book.,.. POSSESSION
(4) Engl. The book.,. WAS on the tab/ey. LOCATION
Cognitive aspects ofsemantic change and polysemy 281
I would like to stress that the (partly displaced) "adjacency" of

these four categories in Figure 2 is only a matter of graphic repre-
sentation. The actual semantic relationships between the four do-
mains (or parts of them) have to be specified on the conceptual level,
as we will see in sections 2-5.
On the basis of cross-linguistically observable lexical facts, there
are two further conceptual distinctions to be made, which cut across
the four main categories represented in Figure 2:
BOUNDED UNBOUNDED
... ,
X,Y \ X,Y
LOCATION
x x
rhematlex thematic x
Figure 2. The semantic space HAVEiBE (according to Koch 1993)
First, there is the distinction between the (inherent) thematicity

and the rhematicity of participant x. Thus, for LOCATION we can dis-
tinguish, with respect to participant x, THEMATIC LOCATION from
RHEMA TIC LOCATION.
(4) a. Engl. The bookx WAS on the table y. THEMATIC LOCATION

(x = thematic)
282 Peter Koch
b. Engl. THERE WAS a bookx on the tabley. RHEMATIC LOCATION

(x = rhematic)
But as illustrated by Figure 2, this distinction applies to POSSES-

SION and to EXISTENCE as well. 4
At first glance, the connection between thematicity/rhematicity
and certain predicative concepts instead of sentences seems a bit
misleading, because we are used to assigning degrees of thematicity
or rhematicity rather to elements of a sentence (e.g. subject, direct
object etc.). Nevertheless, inherent degrees of thematicity or
rhematicity attributed to participant slots are essential for differen-
tiating between semantically similar verbs and for relating them to
each other (cf. Koch 1981: 93-95, 341-342, 352-356, 1991: 297-301;
Oesterreicher 1991). This is conclusively demonstrated, for instance,
by Sp. estar (THEMATIC LOCATION) as opposed to haber (RHEMATIC
LOCATION):
(5) a. Sp. Ellibrox ESTABA sobre la mesay. THEMATIC LOCATION

'The book was on the table.' (x = thematic)
b. Sp. HAB1A un librox sobre la mesay. RHEMATIC LOCATION
'There was a book on the table. ' (x = rhematic)
In the final analysis, the synchronic situations in Spanish (5a/b)
and in English (4a/b) are not so different. In my view, English has
nowadays two distinct verbal expressions is, i.e., to be (THEMATIC
LOCATION) and there is/there are (RHEMATIC LOCATION), even if the
second has developped diachronically from the first (cf. also section
5).
Obviously, the inherent thematicity-rhematicity reliefs of verbs
are only valid for unmarked usage. Marked departures from the rule
are always possible (cf. [5c]), but this does not impair the inherent
thematicity-rhematicity relief of the verb in question:
(5) c. Sp. Librosx, no HAB1A sobre la mesay. (x * rhematic !)

'Books were not on the table.'
The second distinction I want to introduce concerns the bounded-

ness vs. the unboundedness of the predication, i.e. its limitation in a
spatial or temporal respece
(2) a. Engl. THERE ARE many lionsx in Africa.

BOUNDED EXISTENCE
b. Engl. THERE ARE many unhappy peoplex.
UNBOUNDED EXISTENCE
(3) a. Engl. Johny has a bookx in his hand. BOUNDED POSSESSION
b. Engl. Johny has a bookx· UNBOUNDED POSSESSION
(6) a. Sp. Albertox esta palido.
'Albert has a pale complexion.' BOUNDED ASCRIPTION
b. Sp. Albertox es palido.
'Albert is (being) pale. ' UNBOUNDED ASCRIPTION
The conceptual distinctions represented in Figure 2 provide us

with a cognitive map6 of the semantic space HAvE/BE, that allows us
to retrace characteristic paths of semantic change recurring polyge-
netically - that is, independently of each other - in different lan-
guages of the world and in different epochs of linguistic history. I
will allude here to several recurrent patterns and paths (sections 2
and 3) and then later focus on some of them (sections 4 and 5).
2. From LOCATION via BOUNDED POSSESSION to POSSESSION
We can take as our first example a path that is extremely well-

known to language typologists. It leads from RHEMA TIC LOCATION to
BOUNDED POSSESSION and then to POSSESSION in general (cf., e.g.,
Lyons 1967: 394; Clark 1978: 114; Hagege 1982: 48; Heine 1997:
50-53; for the examples cf. Freeze 1992: 576-577):7
(7) a. Russ. Ha CTOJI~ GblJIA KHHI'Sx.

'There was a book on the table. ' RHEMATIC LOCATION
284 Peter Koch
b. Russ. Y MCHRy GblRA KHHI'8x.

'I had a book.' POSSESSION
(8) a. Finn. Poydiillii y ON kynii x •
'There is a pencil on the table. ' RHEMA TIC LOCATION
b. Finn. Liisallay ON miesx.
'Lisa has a husband. ' POSSESSION
(9) a. Hindi Kamree-meey aadmiix HAI.
'There is a man in the room. ' RHEMA TIC LOCATION
b. Hindi Larkee-keey paas kuttaax HAI.
'The boy has a dog.' POSSESSION
(10) a. Tagalog MAY gerax sa Ewropay.
'There is a war in Europe. ' RHEM. LOCATION/EXISTENCE
b. Tagalog MAY relosx a naanaiy.
'Mom has a watch. ' POSSESSION
A relationship of BOUNDED POSSESSION (x, y) is prototypically

linked, i.e., is contiguous, to one of LOCATION (x, y) within the same
frame (what I have with me, is, e.g., in my hands, in my pocket etc.).
So, the first step of change must be a metonymicaf one. Via a fig-
ure-ground effect we slip from RHEMATIC LOCATION to RHEMA TIC
BOUNDED POSSESSION (p. 291, Figure 3: arrow 0).9 The second step,
then, is an extension of meaning lO from RHEMA TIC BOUNDED POSSES-
SION to RHEMATIC POSSESSION in general (arrow 8).
This last step of extension of meaning, necessary for reaching the
target concept in the examples (7b) - (lOb), is part ofa more general
phenomenon of extension from BOUNDED POSSESSION to POSSESSION
in general. To name just one further, well attested type of examples:
a very special, but surely prototypical kind of BOUNDED POSSESSION
is HOLDING SOMETHING IN ONE'S HAND. There are numerous cases of
extension of meaning from this prototype to the whole category of
POSSESSION (cf. Buck 1988: s.v. 11.11 HAVE; Koch 1991: 291-292;
1995: 30-31):11
(11) (lE . • /cap- 'to grasp, to seize' » Goth. haban 'to hold' > 'to
have' (cf. Engl. have; Germ. haben)
(12) (lE. *ghabh- 'to grasp, to seize' » Lat. habere 'to hold' > 'to
have'
(13) Lat. tenere 'to hold' > Sp. tener, Sard. tennere, Southern It.
tenere 'to have'
(14) Fr. tenir 'to hold' > Guad.Creole (ti)ni 'to have'
(15) ChSl.j~i 'to take' in relation to im-i5-ti 'to have'
3. POSSESSION, EXISTENCE and LOCATION
A very widespread change consists of transfonning verbs of POSSES-

SION into verbs of EXISTENCE and further on into verbs of
(RHEMATIC) LOCATION. We find the most famous example of this
change in the evolution of Latin into the Romance languages. Take,
for instance, the meaning 'to have' = RHEMATIC POSSESSION of Lat.
habere (cf. also [12]):
(16) a. Lat. Marcusy librumx HABET.

'Marcus has a book. ' RHEMATIC POSSESSION
In preclassical and postc1assical Latin, we have several attesta-

tions of lat. habere as an impersonal expression of RHEMA TIC
BOUNDED EXISTENCE, as for example in (16b), and of RHEMATIC LO-
CATION, as in (16c) (cf. also ThLL: s.v. habere, III C 2, p. 2461, 78-
2462,5):
(16) b. OLat. Animae pauxillumx in mey HABET. (Naevius, before

200 B.C., cit. Manoliu Manea 1985: 111)
'There is very little life in me. '
RHEMA TIC BOUNDED EXISTENCE
c. VulgLat. Misi tibe ... chiloma. Entroy HABET collyramx I
(2nd Century A.D., cit. Cugusi 1981: 724)
'I sent you a box [... ]. There is a loafofbread in it [... ].'
RHEMATIC LOCATION
286 Peter Koch
It is well known that many Romance languages have continued to

maintain these senses of habere. Sp. haber (which, by the way, to-
tally abandoned the original possessive sense) may serve as an ex-
ample:
(17) a. Sp. En el puebloy HABiA muchas casasx.

'In the village, there were many houses. '
RHEMATIC BOUNDED EXISTENCE
b.=(Sb) RHEMATIC LOCATION
Another instance is Fr. il y a (examples (18a) and (18b», which

has a somewhat more complicated history (cf. Kawaguchi 1991:
134-177):
(18) a. Fr. Dans le village, IL Y AVAIT beaucoup de maisons.

'In the village, there were many houses.'
b. Fr.IL Y A VAIT un livrex sur la tabley.
In Romance languages yet another sense developed, based on the

Latin innovation, namely RHEMA TIC UNBOUNDED EXISTENCE:
(17) c. Sp. HAY mucha gentex infeliz.

'There are many unhappy people.'
RHEMA TIC UNBOUNDED EXISTENCE
(18) c. Fr. IL YA beaucoup de gensx malheureux.
'There are many unhappy people. '
RHEMA TIC UNBOUNDED EXISTENCE
There is overwhelming evidence that all these semantic changes

are not idiosyncratic facts of Latin-Romance lexical history. We may
compare other completely independent· instances of analogous
changes in different languages (cf. Clark 1978; Bickerton 1981: 66-
67; Buchholz 1989; Hengeveld 1992: 159-160; Heine 1997: 95, 137-
138):
(19) a. Mod. Or. (colI.) 0 £1tVPO,y EXEI eva yA.7fyopo

aVToK(vTJ1:0x.
'Spiros has a fast car.' RHEMATIC POSSESSION
b. Mod.Gr. (coll.) E'aVTO TO 01t(f'ly EXEI Tpexoupevo
vepox.
(lit. 'in this the house it-has running water')
'In this house there is running water.'
c. Mod.Or.(coll.) ET7fV paA.(T(ay EXEI 1taVTeA.OVla
Ka( 1ta1tOUTOlax.
(lit. 'in the suitcase it-has trousers and shoes')
'There are trousers and shoes in the suitcase. '
RHEMATIC LOCATION
(20) a. Pg. JOQOy tinha uma maquina de escreverx.
'Joao had a typewriter.' RHEMATIC POSSESSION
b. Braz.Pg. Um dia nQo tinha mais ratos x.
'One day, there were no more mice.'
(21) a. Germ. Martiny hat eine Schreibmaschinex.
'Martin has a typewriter.' RHEMATIC POSSESSION
b. Southern Germ. Es hat kein Brotx mehr.
'There is no more bread. ' RHEMA TIC BOUNDED EXISTENCE
(ex. 22-26 cf. Bickerton 1981: 66-67)
(22) Guy.Creole Dem getl wan umanxl we get2 gyal-pikni x2.
(23) Haw.Creole Getl wan wahinixl shi get2 wan datax2.
(24) Hait.Creole Gel you famxl Id ge2 you pitit-jix2.
(25) Papiamentu Tim u muhexI cu tim un yiu-muhex2.
(26) Sao Tome Creole Tel ua mwalaxl ku te2 ua mina-mosax2.
'There is a woman who has a daughter.'
Verb\: RHEMATIC EXISTENCE
Verb 2 : RHEMATlC (INALIENABLE) POSSESSION
288 Peter Koch
(27) Guad.Creole Nil saxl Id niz, nil saxl Id pa nh

(Poullet et al. 1984: s.v. ni)
'There are those who have got something and those who
haven't got anything.'
nil: RHEMA TIC UNBOUNDED EXISTENCE
ni2: RHEMA TIC POSSESSION
(28) a. Bulg. HMAMr.y) cpHa JIHB8,I(1l" rOJIllMa.
'I have a large meadow. ' RHEMA TIC POSSESSION
b. Bulg. ilPC3 JIHTOTOy HMA MHoro HJIOPOBCx'
(Buchholz 1989: 333)
'In the summertime there are many fruits.'
RHEM. BOUNDED EXISTENCE
c. Bulg. CIlMO OrHJ/HOB8..,. HRMAIUE TIlM.
(Buchholz 1989: 334)
'Only Ognjanov was not there.' RHEMA TIC LOCATION
(29) a. Nubi acmayfii ma ya/a x tinin. (Heine 1997: 13 7)
'I have two children.' RHEMATIC POSSESSION
b. Nubi aldlxfii naa? (Heine 1997: 13 7)
'Is there food?' RHEMA TIC BOUNDED EXISTENCE
I think that by taking a cognitive perspective we can try to retrace

and understand the different conceptual steps involved in these se-
mantic changes.
4. From bounded possession to bounded existence
Even if we take into account the rather well attested Latin material in
ex. (16b) and (16c), several questions arise. Besides other problems
that I do not consider insunnountable,12 it is not entirely clear which
was the first sense to develop from an initially possessive sense of
habere: RHEMATIC EXISTENCE or RHEMA TIC LOCATION? Our oldest
attestation of the impersonal Latin verb habere (16b) rather ex-
presses RHEMA TIC BOUNDED EXISTENCE. The second oldest attesta-
tion (16c) clearly expresses RHEMATIC LOCATION.
Cognitive aspects of semantic change and po/ysemy 289
Let's suppose that the passage from RHEMATIC POSSESSION to

RHEMA TIC BOUNDED EXISTENCE was actually the first step of seman-
tic change (Figure 3: arrow 0). An argument for this hypothesis
could be that verbs like have are the simplest and most typical verbs
which serve to introduce entities as existing in the universe of dis-
course (e.g. Once upon a time, there was a Idng who had three
daughters. They aI/lived in an old castle in the forest . ... ).
In this context, note the Modem Greek personal variant (19d) of
(19b):
(19) d. ModGr. Auro ro U7f{nyEXElrpeXOUjievo vepox.

lit. 'This house has running water.' BOUNDED EXISTENCE
In Latin, we can hypothesize an analogous sentence like (16e) as

opposed to the ''traditional'' version (16d)(cf. Bassols de Climent
1948; Garcia Hemandez 1992: 167):
(16) d. Lat. Domi est multum vinum x •

'In the house there is much wine.'
e. Lat. Domus habet multum vinum x •
lit. 'The house has much wine.'
How could we explain that in (16e) an expression for POSSESSION

like habere directly replaced an expression of EXISTENCE like esse in
(16d)? Perhaps it would be easier to relate (16e) to (16f):
(16) f. Lat. Dominus habet multum vinum x•

'The master has much wine.' RHEMATIC POSSESSION
The step from (16f) to (16e) can easily be accounted for as a case
of metonymy. In a given frame, comprising MASTER, HOUSE and
WINE, we observe a figure-ground effect from MASTER-WINE to
HOUSE-WINE (cf. Note 8). But at this point (16e) could have been
290 Peter Koch
reinterpreted as a metaphorl3 replacing (16d), i.e. as a metaphor that

conceptualizes BOUNDED EXISTENCE in terms of BOUNDED POSSES-
SION. When a given object exists in a given PLACE, the PLACE can be
regarded as a CONTAINER that is similar to the POSSESSOR of the ob-
ject.
Metaphor always involves a conceptual leap, but interpreting
PLACES as POSSESSORs is a very bold leap, and syntactically,
expressions of PLACES, such as participanty in (19d) and in (16e) are
non-prototypical subjects of transitive verbs. The same problem can
be observed in well-known English examples like (31) with respect
to (30):
(30) Eng!. Johny HAS moneyx in his pocket.

RHEMA TIC BOUNDED POSSESSION
(31) Eng!. The tree(y) HAS a nestx in ity.
The metaphorical and syntactic "shock" provoked by sentences

like (19d) or (16e) can be attenuated by also indicating, within the
same clause, the fact of LOCATIVE BOUNDEDNESS of the EXISTENCE of
x. The locative phrase may be co-referential with the subject (31).
But the locative phrase also may be introduced by switching from
a personal construction such as (19d) or (16e) (where the subject
participant y still has the form of a POSSESSOR) to an impersonal con-
struction like (19b) (where the subject participant y has simply "eva-
porated"14 in favour of a new participant y that again expresses the
LOCA TIVE BOUNDEDNESS of the EXISTENCE of x). In Latin, we arrive
at a construction like (16g) that virtually can have arisen from a
blend of types (16d) and (16e):ls
(16) g. Lat. Domi habet mu/tum vinum x'

'In the house there is much wine.'
According to solutions such a (19b) or (16g), the PLACE-

POSSESSOR metaphor frequently and polygenetically provides for im-
personal expressions of EXISTENCE (Figure 3: arrow .. ,
corresponding to ex. [19b], [20b], [21 b], [22]-[26], [27], [28b], and
[29bD. It remains to be seen whether or not the reinterpretation of a
metonymy as a metaphor like in (16e), followed by a blend with a
construction like (16d), is a factor in analogous processes in other
languages, too, or if a direct - and very bold - metaphor from
RHEMA TIC POSSESSION to BOUNDED EXISTENCE is the ordinary way.
BOUNDED UNBOUNPED
X,Y
rhematlex themaflex
Figure 3. Recurrent paths of semantic change in the space HAVE/BE
5. From EXISTENCE to LOCATION and vice versa
Many languages lexicalize RHEMATIC LOCATION as opposed to THE-

MATIC LOCATION. French is one of these languages:
292 Peter Koch
(18) b. (see above) RHEMA TIC LOCATION

c. Fr. Le livrex ETAIT sur la tabley.
'The book was on the table. ' THEMATIC LOCATION
On the other hand, it is striking how many of these languages

have identical expressions for RHEMATIC EXISTENCE and for RHEMA-
TIC LOCATION (cf., e.g., [2] and [4b] , [17a] and [17b], [18a] and
[18b], [19b] and [19c]).We could doubt ifit is possible at all to dis-
tinguish these two categories (cf. F euillet 1998: 706-707). In my
opinion, the close proximity of the concepts of RHEMA TIC LOCATION
and RHEMATIC EXISTENCE, which is supported by psychological, phi-
losophical, and linguistic observations (cf. also Lyons 1967: 390-
391; Bolinger 1977: 99; Holenstein 1980: 32; Bogacki 1988: 24-25),
can, and must, be interpreted in terms of contiguity.
As we have already seen, the BOUNDEDNESS of EXISTENCE proto-
typically, though not always, is due to a LOCATIVE limitation. In
such a prototypical frame of BOUNDEDNESS of EXISTENCE, there
necessarily is a strong contiguity between EXISTENCE and LOCATION.
Thus, if you have at your disposal an expression for BOUNDED
EXISTENCE like Lat. habere in (16b/g), it easily slips via a figure-
ground effect into an expression of LOCATION (16c). This is a
metonymical effect (Figure 3: arrow e) (cf. Note 8). But in order to
grasp this effect, we have to distinguish between two figure-ground
constellations: figure = RHEMATIC LOCATION/ground = RHEMATIC
EXISTENCE on the one hand, and figure = RHEMA TIC EXISTENCE
/ground = RHEMATIC LOCATION on the other.
This twofold perspectivization is confirmed by several further lin-
guistic observations (cf. Koch 1993: 181-183 for the first and second
points):
First, there are languages such as Spanish, that use largely the
same lexical unit for RHEMATIC LOCATION and RHEMA TIC EXISTENCE
(Sp. haber), but that in some specific cases use the same lexical unit
for THEMATIC and RHEMATIC LOCATION (Sp. estar), completely ex-
cluding the latter verb, however, from the realm OfRHEMATIC EXIST-
ENCE:
(3) c. Sp. Delante del edificioy ESTABAN el cura y dos monaguil-

losx. (Koch 1993: 182)
'In front of the building, the priest and two choirboys
were standing. ' RHEMA TIC LOCATION
In such a case, the distinction between RHEMA TIC LOCATION and

RHEMATIC EXISTENCE is of indeniable importance.
Secondly, there seem to be languages, like German, that in a very
clear-cut way, i.e., by using different lexical units, mark the distinc-
tion between RHEMATIC (and THEMATIC!) LOCATION (Germ. seinllie-
gen etc.: [33a] and [33b]) and EXISTENCE (Germ. es gibt: [33c]):
(32) a. Germ. Auf dem Tisch y WAR/LAG ein Buch x.

'There was a book on the table. ' RHEMA TIC LOCATION
b. Germ. Das Buch x WAR/UG auf dem Tischy.
'The book was on the table. ' THEMA TIC LOCATION
c. Germ. In AJrikay GIBT ES L6wenx.
'There are lions in Africa. ' RHEM. BOUNDED EXISTENCE
In such a case, the distinction between RHEMATIC LOCATION and

RHEMATIC EXISTENCE is of crucial importance.
Thirdly, the contiguity between EXISTENCE and LOCATION seems
to be so strong that we not only observe metonymies from RHEMA TIC
EXISTENCE to RHEMATIC LOCATION (as described above), but also
metonymies in the opposite direction, as, e.g., in English and Italian.
How does this come about? First of all, these two languages de-
veloped particular expressions for RHEMATIC LOCATION by integrat-
ing into the verb for THEMATIC LOCATION an originally deictic ele-
ment, that is co-referential with the thematic locative participant y
(Figure 3: arrow e). Compare for Italian esserci (33b) in relation to
essere (33a):
(33) a. It. n librox ERA sui tavolay.

'The book was on the table. ' THEMA TIC LOCATION
294 Peter Koch
b. It. C 'ERA un IibrOx sui tavolay.

(Cf. for English: there is (4b) in relation to to be (4a».

Compared with the originally neutral verb of LOCATION «4a),
(33a», the new verbal expression «4b), (33b», integrating Engl.
therellt. ci by a phraseological device, designates a restricted type of
LOCATION, namely RHEMATIC LOCATION, whereas the verb Engl. to
bellt. essere restricts its meaning to THEMATIC LOCATION. Now, on
diachronic grounds it is obviously the novel RHEMA TIC LOCATIVE
expression that by way of metonymy also became an expression of
BOUNDED EXISTENCE in these languages (Figure 3: arrow CD) and not
VIce versa:
(33) c. It. In Africay Cl SONO molti leonix •

'There are many lions in Africa.' BOUNDED EXISTENCE
(Cf. for English: there is/are (2a».

If we did not distinguish in principle RHEMA TIC EXISTENCE and
RHEMATIC LOCATION, we would not be able to adequately retrace the
opposite diachronic paths in this semantic space:
RHEMATIC BOUNDED POSSESSION English, Italian
.~ t.
RHEMATIC LOCATION
......•.•..
•................
e.g. French, Greek
•
LOCATION
Figure 4. From RHEM. BOUNDED EXISTENCE to RHEM. LOCATION and vice versa
Let us now consider BOUNDED EXISTENCE. This is a special, but

prototypical case of EXISTENCE in general. Often speakers, relying
heavily on the context, do not even make the boundedness of
EXISTENCE explicit (cf. Hengeveld 1992: 97). Thus, (20b)/(21b)
could, according to the context, mean that 'there were/is no more
micelbread in the house, in the village, in the country, on earth etc.'
Put the other way around, any EXISTENCE is bounded in a certain
sense. Even UNBOUNDED EXISTENCE is BOUNDED by the universe of
discourse (cf. Bolinger 1977: 99). So it is not surprising, that by ex-
tension of meaning, expressions for BOUNDED EXISTENCE - whatever
their origin may be (cf. Figure 4) - are extended to EXISTENCE in
general (Figure 3: arrow 8):
(19) e. ModGr. EXEI1rOA,A,O';, t5VCT'f'VXIUjJEVOV,

a v(}ptJ1rO V,x. (cf. [19bD
(lit. 'it-has many unhappy people')
'There are many unhappy people. '
RHEMATIC EXISTENCE
(28) d. Bulg. HMA rq,lfHOB8m XOpllx. (Buchholz 1989: 332)
(cf. [28b])
'There are strange people. ' RHEMATIC EXISTENCE
(32) d. Germ. ESGIBTviele ungliickliche Menschenx. (cf. [32cD
'There are many unhappy people.' RHEMATIC EXISTENCE
(33) d. It. c't molta gentex in/elice. (cf. [33c])
'There are many unhappy people.' RHEMA TIC EXISTENCE
cf. also: (2b) in relation to (2a) for English;

(17c) in relation to (17a) for Spanish;
(18c) in relation to (18a) for French;
(27): for Guadeloupe Creole (nil).
296 Peter Koch
6. Final observations
Lexical change within the semantic space HAVE/BE is not arbitrary.

We have demonstrated this partly by referring to ample cross-lin-
guistic evidences elaborated by others (esp. Heine 1997) and partly
by carrying out a detailed analysis of apparently recurrent processes
of change in this area, giving us insight into the nature of "semantic
space". It would certainly be interesting to collect more cross-lin-
guistic data that would corroborate the recurrent nature of the meta-
phors, metonymies, and extensions of meaning interrelating the con-
cepts of POSSESSION, EXISTENCE and LOCATION.
In a rather conventional way, we followed grosso modo the dia-
chronic succession of changes, asking ourselves "in which direction"
a given semantic class of lexical units moved. This may have been a
convenient mode of presentation, but we have to realize that it is to-
tally artificial in relation to what really happens in language change.
Speakers do not intend to change anything in language (cf. Coseriu
1958), so lexical items are not "directed" anywhere. On the contrary,
speakers use lexical units to communicate in the most efficient and
successful manner possible that which they want to express. If
speakers affect semantic change (and they do), they affect it not by
providing existing words with novel meanings (semasiological per-
spective), but instead indirectly by expressing things through other
and/or new words (onomasiological perspective).
So if there really are regularities and "invisible-hand processes"
(cf. Keller 1994) in semantic change and if they are, at least to a
large extent, by-products of the speakers' activities, then we have to
read the results of our investigation "backwards" in order to give
them methodological significance (cf. Koch 1997 and in press a).
Very tentatively, then, we can make the following statements based
on our material: 16
A new expression for EXISTENCE in general is created with more
than random probability by extension of meaning from an ex-
pression for BOUNDED EXISTENCE (Figure 3: arrow 8).
Cognitive aspects of semantic change and polysemy 297
A new expression for (RHEMATIc) BOUNDED EXISTENCE is created

with more than random probability either by metonymy from an ex-
pression for RHEMATIC LOCATION (arrow 0) or by metonymy, re-
interpreted as metaphor, from an expression for RHEMA TIC POSSES-
SION (arrow").
A new expression for RHEMA TIC LOCATION is created with more
than random probability either by metonymy from an expression for
BOUNDED EXISTENCE (arrow 8) or by a phraseological device in-
volving restriction from an expression for LOCATION (arrow 0).
A new expression for RHEMA TIC POSSESSION is created with more
than random probability by extension of meaning from an
expression for RHEMATlC BOUNDED POSSESSION (arrow $).
A new expression for RHEMA TIC BOUNDED POSSESSION is created
with more than random probability by metonymy from an
expression for RHEMATIC LOCATION (arrow 0).
It would perhaps be possible to demonstrate that these paths of
semantic change (that must not be used at all) reappear polygeneti-
cally in the history of different languages and should be explained
by certain cognitive universals underlying our conceptualization of
POSSESSION, EXISTENCE and LOCATION.
In any case, the activation or non-activation of these types of se-
mantic change and the resulting interlinguistic differences in the arti-
culation of the semantic space HAVE/BE constitute features highly re-
levant to language comparaison. These features contribute to what I
would like to call 'lexical typology' (cf. Koch, in press b).
Notes
1. Cf. Lyons (1967); Langacker (1968); Clark (1978); FoleyNan Valin

(1984: 47-53); Kawaguchi (1991); Freeze (1992); Kayne (1993: 4-7). See
the critical survey in Heine (1997: 214-222).
2. At least since the seminal work of Breal (1921: 143-144) we know that
with respect to semantic change polysemy is only the other (synchronic)
side of the coin. Consequently, the old and the new meaning of a lexical, or
grammatical, unit can coexist for a while, but at some moment in
298 Peter Koch
diachrony the old meaning (or sometimes the new meaning) can be
abandoned (cf. Koch 1991: 293, 1994a: 203-209; Blank 1997: 114-130).
Even though Cognitive Semantics and Grammaticalization Theory have
"rediscovered" these fundamental facts (Wilkins 1996: 267-270; Heihe
1997: 82-83), many linguists still seem perplexed when faced with
phenomena of this kind.
3. I do not take into consideration here a possible, and probably necessary,
additional category EQUATION (e.g. Athens IS the capital of Greece),
because it will not be relevant for the processes I want to analyze. - I am
also aware of the internal, possibly prototypical, structure of a large
category of POSSESSION, including PHYSICAL POSSESSION, INANALlENABLE
POSSESSION, etc. (cf. Taylor 1995: 202-203; Heine 1997: 33-41; cf. also
Koch 1981: 314-317, 359). For the present purpose, I can exclude all kinds
of INANALlENABLE POSSESSION, the only relevant sub-category being what
I call BOUNDED POSSESSION (as exemplified by ex. [3aJ and [30J and as
represented in Figure 2). - I do not go into details concerning the verbal or
non-verbal character of the predication in cases where the categories of the
semantic space HA vF1BE are expressed by copulas or by zero (cf.
Hengeveld 1992: 26-30; on the other hand Feuillet 1991, 1998: 664).
4. For the category of ASCRIPTION, inherent thematicity of participant x seems
to be generalized (cf. example [1]), which does not exclude marked uses of
ASCRIPTION predicates as in Engl. Stupid he isn't. - Until now, the thema-
ticity-rhematicity conditions have not been accounted for systematically in
studies concerning the semantic space HAVF1BE. However they are often in-
directly involved in the description of single categories or parts of this area.
Thus, corresponding to THEMATIC vs. RHEMATIC POSSESSION we fmd: "pre-
dicat d'appartenance" vs. "predicat de possession" (Benveniste 1960: 196-
197); "possess2 " vs. "possess." (Clark 1978); "ownership" vs. "possession"
(Bickerton 1981: 245); "non-presentative possessive" vs. "presentative
possessive" (Hengeveld 1992: 125-126); "'belong'-constructions" vs.
"'have'-constructions" (Heine 1997: 29-33). Corresponding to THEMATIC
vs. RHEMATIC EXISTENCE we fmd: "modele Cogito, ergo sum" vs. "modele
il y a + GN" (Feuillet 1998: 730-707; cf. also Koch 1993: 180-181).
Corresponding to THEMATIC vs. RHEMATIC LOCATION, we fmd "non-
presentative locative" vs. "presentative locative" (Hengeveld 1992: 125-
126, with a clear-cut distinction from "existential"), or "locative/situative"
vs. "existential" (Clark 1978; Hagege 1982: 46, 49; Freeze 1992: 553).
Note that in the latter, very current. terminology the distinction between
RHEMATIC LOCATION and EXISTENCE proper - as in (2) -fades away unduly
(cf. section 5 for more details).
5. For the category of LOCATION, inherent boundedness of the predication is
logically necessary (cf. ex. [4]). - Until now, the inherent boundedness of
Cognitive aspects of semantic change and polysemy 299
the predication has not been accounted for systematically in studies con-
cerning the semantic space HAvE/BE. However they are often indirectly in-
volved in the description of single categories or parts of this area. Thus,
corresponding to BOUNDED vs. UNBOUNDED EXISTENCE we fmd: "presenta-
tive locative" vs. "existential" (Hengeveld 1992: 97, 125f.). Roughly cor-
responding to BOUNDED vs. UNBOUNDED POSSESSION we fmd: "physical
possession" vs. other types of "possession" (Heine 1997: 34; cf. above
n.3). Corresponding to BOUNDED vs. UNBOUNDED ASCRIPTION, we fmd:
"enonce [8 localisation temporelle et modale] et situe par rapport au locu-
teur" vs. "enonce hors de toute localisation temporelle et modale et hors de
la subjectivite du locuteur" (Benveniste 1960: 160, cf. also 167); "attribut"
vs. "essence" (Hagege 1982: 48-49); "qualification incidentielle" vs. "qua-
lification essentielle" (cf. Feuillet 1998: 725; cf. 711-725).
6. For the principle of semantic or cognitive maps, cf. Anderson (1982); By-
bee (1985: 195-(96); Croft et a1. (1987); Haspelmath (1997: 59-62). Cf.
however the critical remarks above concerning the interpretability of "adja-
cencies".
7. In Heine's 100 language sample, this designation type for POSSESSION
covers 20.9% (cf. Heine 1997: 75).
8. As for this understanding of metonymy and contiguity in relation to proto-
typicality and figure-ground effects, cf. Koch (1995: 40-41, 1999); Blank
(1997: 235-243). (What Traugott and Konig [e.g. 1991] call "pragmatic
strengthening" or "conventionalization of a conversational implicature" is
a kind of metonymy.) Note that the notion of prototypicality that we have
to apply here is an onomasiological one (cf. Koch (996). In this sense,
contiguity is an external relation between distinct categories. So, it would
not be legitimate to denominate a metonymical process, like the one
described, as a "metonymical extension" of the category RHEMA TIC
LOCATION.
9. Just as I use "RHEMATIC LOCATION" for "LOCATION with a rhematic partici-
pant x", so also "RHEMATIC BOUNDED POSSESSION" for "BOUNDED POSSES-
SION with a rhematic participant x" etc.
10. As for this understanding of (true) extension to a whole category, a process
that is virtually, but not necessarily, in relation to prototypicality, cf. Koch
(1995: 30-31); Blank (1997: 200-206).
11. According to Heine (1997: 75), the so called "'action'-type" for the desig-
nation of POSSESSION covers 13.6% of the lOO-language sample. The term
"action", however is misleading in this context, because the immediate
conceptual basis of these designations for POSSESSION is HOLDING IN ONE'S
HAND. (The designation for HOLDING IN ONE'S HAND may in turn go back
to a word designating the action of GRASPING/TAKING; but for some
languages, as far as we know, this is not at all certain: cf. L. tenere 'hold'
300 Peter Koch
as antecedent in ex. [13]-[14]. Anyway, it is not the concept of ACTION as

such that motivates new designations for POSSESSION, but the - in my view
prototypical - case of BOUNDED POSSESSION).
12. First of all, Greek influence as been invoked (cf. Svennung 1932: 84;
Luque Moreno 1978: 138-140; Garcia Hernandez 1992: 166-167). If this
should be correct, we would have to count the Greek and the Romance at-
testations of the POSSESSION ~ EXISTENCE pattern as only one occurrence.
But this would not detract from the broad polygenetic evidence for this pat-
tern. Secondly, a different meaning of L. habere has been considered the
source of the existential meaning, namely 'to be in a state' (with possible
Greek parallels in this case as well). This explanation cannot be excluded a
priori on lexicological grounds, so that it would be possible to interpret the
syntactically ambiguous noun phrase animae pauxillum. in (16b) not as the
object of an impersonal verb habet, but as the subject of a personal form
habet (cf. Luque Moreno 1978: 138-146; Garcia Hernandez 1992: 170;
Manoliu Manea 1985: 111; indeed, habet is also attested with
unambiguous subject participants x: cf. ThLL: s.v. habeo, III C 2, p. 2461,
65-77). But even on this assumption, one has to postulate that the
impersonal type domiy habet mulrum vinum.=oo, lit. 'in the house it-has
much wine', is the result of a contamination between the personal type
domiy habet mulrum vinum._s, lit. 'in the house is much wine', and the
personal type domusy-S habet mulrum vinum._oo , lit. 'the house has much
wine' (Luque Moreno 1978: 145). Thus, it seems impossible to totally
disregard the role of the latter, metonymicaVmetaphorical type on a clearly
possessive basis that is much closer to the etymological meaning of habere
(cf. ex. [12]) and fits perfectly into the polygenetic pattern illustrated by
ex. (19) ff.
13. For the understanding of metaphor cf. Lakoffllohnson (1980); Lieber
(1992: 14-82); Koch (1994a, 1995: 39-40); Blank (1997: 157-171). Since
metaphor always involves a conceptual leap, the (presumed) similarity is
an external relation between distinct categories. So, it would not be
legitimate to denominate a metaphorical process, like the one described, as
a "metaphorical extension" of the category POSSESSION. - As for
reinterpreting metonymies as metaphors, cf. Taylor (1995: 138); Koch
(1997: 234).
14. For the "evaporation" of highly thematic subject participants in the genesis
of impersonal verbal expressions cf. Koch (l994b).
15. Interestingly, our example (16c), although it goes one step further (and ex-
presses RHEMATIC LOCATION), is ambigous with respect to types (16e)/(31)
and (16g), since the participant y is anaphorical and therefore implicit.
16. It would be more accurate to begin each statement conditionally: "If a new
expression X is created ... ".
References
Anderson, Lloyd
1982 The 'Perfect' as a universal and as a language particular category.
In: Paul Hopper (ed.), Tense and Aspect: Between Semantics and
Pragmatics, 227-264. (Typological Studies in Language 1.) Amster-
dam I Philadelphia: Benjamins.
Benveniste, Emile
1960 "Etre" et "avoir" dans leurs fonctions linguistiques. In: Emile Ben-
veniste, Problemes de linguistique generale. Volume 1, 187-207.
Paris: Gallimard.
Bickerton, Derek
1981 Roots of Language. Ann Arbor: Karoma.
Blank, Andreas
manischen Sprachen. (Beihefte zur Zeitschrift fUr Romanische Phi-
lologie, 285.) Tiibingen: Niemeyer.
Bogacki, Cbristophe
1988 Les verbes a argument incorpore. Langages 89: 7-26.
Bolinger, Dwight
1977 Meaning and Form. (English Language Series 11.) London I New
York: Longman.
Bassols de Climent, Mariano
1948 Origen de la construcci6n impersonal del verbo habere. Revista de
estudios cltisicos 3: 215-225.
Breal, Michel
s1921 Essai de semantique (Science des significations). Paris: Hachette.
Buchholz, Oda
1989 Zu Konstruktionen mit unpersonlich gebrauchtem 'haben' in den
Balkansprachen. Zeitschrift ftJr Phonetik, SprachwissenschaJt und
Kommunikationsforschung 42: 329-338.
Buck, Carl D.
31988 A Dictionary of Selected Synonyms in the Principal Indo-European
Languages. Chicago / London: University of Chicago Press.
Bybee, Joan L.
1985 Morphology: A Study of the Relation Between Meaning and Form.
(Typological Studies in Language 9.) Amsterdam I Philadelphia:
Benjamins.
302 Peter Koch
Clark, Eve V.
1978 Locationals: existential, locative, and possessive constructions. In:
Joseph H. Greenberg (ed.), Universals of Human Language.
Volume 4. Syntax, 85-126. Stanford: Stanford University Press.
Coseriu, Eugenio
1958 Sincronia, diacronia e historia. El problema del cambio lingilistico.
Montevideo: Universidad de Montevideo.
Croft, William, Hava Bat-Zeev Shyldkrot and Suzanne Kemmer
1987 Diachronic semantic processes in the middle voice. In: Anna Giaca-
lone Ramat, Onofrio Carruba and Giuliano Bernini (eds.), Papers
from the 7th International Conference on Historical Linguistics,
179-192. (Current Issues in Linguistic Theory 48.) Amsterdam /
Cugusi, Paolo
1981 Gli ostraka latini dello Wadi Fawikhir. Per la storia del latino. In:
Letterature comparate. Problem; e metodi. Studi in onore di Ettore
Para tore, 11, 719-753. Bologna: Patron.
Feuillet, Jack
1991 Phrases nominales et phrases a verbe "etre". EUROTYP Working
Papers nO 1.
1998 Typologie de "etre" et phrases essives. In: Jack Feuillet (ed.), Ac-
tance et Valence dans les Langues de I 'Europe, 663-767. (Empirical
Approaches to Language Typology. EUROTYP 20-2.) Berlin / New
York.
Foley, William and Robert D. Van Valinjr.
1984 Functional Syntax and Universal Grammar. (Cambridge Studies in
Linguistics 38.) Cambridge: Cambridge University Press.
Freeze, Ray
1992 Existentials and other locatives. Language 68: 553-595.
Garcia HemAndez, Benjamin
1992 Nuevos verbos impersonales en latin tardio e influencia griega. In:
Maria Iliescu and Wemer Marxgut (eds.), Latin vulgaire - latin
tardif ///, 159-172. Tiibingen: Niemeyer.
Hagege, Claude
1982 La structure des langues. (Que sais-je? 2006.) Paris: Presses Uni-
versitaires de France.
Haspelrnath, Martin
1997 Indefinite Pronouns. Oxford: Clarendon.
Heine, Bemd
1997 Possession. Cognitive Sources. Forces and Grammaticalization.
(Cambridge Studies in Linguistics 83.) Cambridge: Cambridge Uni-
versity Press.
Hengeveld, Kees
1992 Non-verbal Predication. Theory, Typology, Diachrony. (Functional
Grammar Series 15.) BerlinlNew York: Mouton de Gruyter.
Holenstein, Ebnar
1980 Von der Hintergehbarkeit der Sprache. (suhrkamp taschenbuch wis-
senschaft 316) FrankfurtlM.: Suhrkamp.
Kawaguchi, Junji
1991 ArolR et les problemes de la localisation en fran~ais. Tokyo:
France Tosho.
Kayne, Richard S.
1993 Towards a modulary theory of auxiliary selection. Studia
Linguistica 47: 3-31
Keller, Rudi
21994 Sprachwandel. Von der unsichtbaren Hand in der Sprache. (Uni-
TaschenbUcher 1567.) T6bingen I Basel: Francke.
Koch, Peter
1981 Verb - Valenz - VerjUgung. Zur Satzsemantik und Valenzfranzosi-
scher Verben am Beispiel der Verfilgungsverben. (Reihe Siegen 32.)
Heidelberg: Winter.
1991 Semantische Valenz, Polysemie und Bedeutungswandel bei romani-
schen Verben. In: Koch and Krefeld (eds.), 279-306.
1993 HABEN und SEIN im romanisch-deutschen und im innerromanischen
Sprachvergleich. In: Giovanni Rovere and Gerd Wotjak (eds.),
Studien zum romanisch-deutschen Sprachvergleich, 177-189. (Lin-
guistische Arbeiten 297.) T6bingen: Niemeyer.
1994a Gedanken zur Metapher - und zu ihrer Alltiiglichkeit. In: Annette
Sabban and Christian Schmitt (eds.), Sprachlicher Alltag. Linguistik
- Rhetorik - Literaturwissenschaft. Festschrift ftJr Wolf-Dieter
Stempel, 201-225. T6bingen: Niemeyer.
1994b Depersonnalisation (et repersonnalisation). L'information gramma-
ticale 62: 9-11.
1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: eine
kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27-46.
1996 La semantique du prototype: semasiologie et onomasiologie? Zeit-
schrift ftJr franzosische Sprache und Literatur 106: 223-240.
1997 La diacronia quale campo empirico della semantica cognitiva. In:
Marco Carapezza, Daniele Gambarara and Franco Lo Piparo (eds.),
Linguaggio e cognizione, 225-246. (Societa di Linguistica Italiana
37.) Roma: Bulzoni.
1999 Frame and contiguity. On the cognitive bases of metonymy and cer-
tain types of word formation. In: Klaus-Uwe Panther and GUnter
304 Peter Koch
Radden (eels.), Metonymy in Language and Thought, 139-167 (Hu-

man Cognitive Processing) Amsterdam I Philadelphia: Benjamins.
in press a TREE and FRUIT: A cognitive-onomasiological approach. Studi
Italiani di Linguistica Teorica ed Applicata.
in press b Lexical typology. In: Martin Haspebnath, Ekkehard Konig, Wulf
Oesterreicher and Wolfgang Raible (eels.), Typology and
Universals. An International Handbook. Berlin I New York: de
Gruyter.
Koch, Peter and Thomas Krefeld (eels.)
1991 Connexiones Romanicae. Dependenz und Valenz in romanischen
Sprachen. (Linguistische Arbeiten 268.) TObingen: Niemeyer.
Lakoff, George and Mark Johnson
1980 Metaphors We Live By. Chicago: University of Chicago Press.
1968 Observations on French possessives. Language 44: 51-75.
Liebert, Wolf-Andreas
1992 Metaphernbereiche der deutschen Alltagssprache. Kognitive Lin-
guistik und die Perspektiven einer Kognitiven Lexikographie. (Euro-
pliische Hochschulschriften I, 1355.) FrankfurtlM.: Lang.
Luque Moreno, Jesus
1978 En tome al sintagma "haber impersonal + sustantivo" y sus
origenes latinos. Revista espaflola de lingilistica 8: 125-147.
Lyons, John
1967 A note on possessive, existential and locative sentences. Foun-
dations of Language 3: 390-396.
Manoliu-Manea, Maria
1985 Tipologia e historia. Elementos de sintaxis comparada romanica.
(Biblioteca romamca hisparuca 11, 337.) Madrid: Gredos.
Oesterreicher, Wulf
1991 Verbvalenz und Informationsstruktur. In: Koch and Krefeld (eels.),
349-384.
Pottier, Bemard
21985 Linguistique generale. Theorie et description. (Initiation a la lin-
guistique B, 3.) Paris: Klincksieck.
Poullet, Hector, Sylviane Telchid and Daniele Montbrand
1984 Dictionnaire des expressions du creole guadeloupeen. Fort-de-
France: Hatier.
Raible, Wolfgang
1981 Von der Allgegenwart des Gegensinns (und einiger anderer Rela-
tionen). Zeitschriftfi1r Romanische Philologie 97: 1-40.
Svennung, Josef
1932 Wortstudien zu den splJtiateinischen Oribasiusrezensionen.
Uppsala: Abnqvist & Wiksell.
Taylor, John R.
21995 Linguistic Categorization. Prototypes in Linguistic Theory. Oxford:
Clarendon.
ThLL=
Thesaurus linguae Latinae. 8 vol. Leipzig: Teubner, 1936-66.
Traugott, Elizabeth C. and Ekkehard KOnig
1991 The semantics-pragmatics of grammaticalization. In: Elizabeth C.
Traugott and Bemd Heine (eds.), Approaches to Grammaticaliza-
tion, I, 189-218. (Typological Studies in Language 19.) Amsterdam
/ Philadelphia: Benjamins.
Wilkins, David P.
1996 Natural tendencies of semantic change and the search for cognates.
In: Mark Durie and Malcobn Ross (eds.), The Comparative Method
Reviewed, 264-304. Oxford: Oxford University Press.
Wilson, Bob
1983 An examination of cross-linguistic constraints on the lexicalization
of predications of ownership, possession, location and existence.
University of Hawaii Working Papers in Linguistics 15/2: 1-18.
List of contributors
Andreas Blank
Philipps-Universitllt Marburg
Institut fur Romanische Philologie
Marburg, GERMANY
David D. Clarke
University of Nottingham
Department of Psychology
Nottingham, UNITED KINGDOM
Dirk Geeraerts
Katholieke Universiteit Leuven
Departement Linguistiek
Leuven, BELGIUM
Peter Koch
Eberhard-Karls-Universitat Tilbingen
Romanisches Seminar
Tilbingen, GERMANY
Ekkehard K6nig
Freie Universitat Berlin
Institut fur Englische Philologie
Berlin, GERMANY
Thomas Krefeld
Ludwig-Maximilians-Universitat Milnchen
Institut fur Romanische Philologie
Milnchen, GERMANY
Ronald W Langacker
University of California, San Diego
Department of Linguistics
La Jo11a, California, USA
308 Conributors
He/mut Ludtke
Christian-Albrechts-Universitat Kiel
Romanisches Seminar
Kiel, GERMANY
Brigitte Nerlich
University of Nottingham
Department of Psychology
Nottingham, UNITED KINGDOM
Franfois Rastier
Universite Paris IV, Sorbonne
Centre de Linguistique Franc;aise
Paris, FRANCE
Peter Siemund
Freie Universitat Berlin
Institut fiir Englische Philologie
Berlin, GERMANY
John R. Tay/or
University of Otago
School of Languages, Linguistics
Dunedin, NEW ZEALAND
Stanford University
Department of Linguistics
Stanford, California, USA
Beatrice Warren
Lund University
Department of English
Lund, SWEDEN
Index
acceptability threshold 120, 122, contrary to the speaker's view 184

124,127,128,131,137,140 contrast 21, 23, 24, 26, 27, 28, 30,
adverbial 178 31,32,33,35,37,41,42
adverb 178, 179, 180, 181, 186, control 147, 148, 152, 160, 162,
189, 190 163, 166, 168, 169, 170, 171,
agent 152, 169, 170 172
agentive 148, 161 counter-expectation 177, 179, 188
agentive subject 148 creativity 12, 50, 51
agentivity 161, 162 cross-over 52, 54, 55, 56, 57
appended senses 216,230,233 cross-space 216
attenuated 158, 161, 166, 167, 170
attenuation 3, 148, 152, 153, 156,
159, 160, 162, 163, 166, 168, defInite article 244, 253, 254
169,170,171,172,189,291 designation 20, 23, 24, 25, 26, 27,
attractor 120, 121, 122, 123, 135 30,38,39,41
auxiliarization 58 discourse 3,33,58,64, 77, 81, 112,
128, 134, 135, 142, 180, 181,
187, 190, 191, 198, 204, 205,
basin 120, 121, 122, 123, 124, 126, 207,238,253,255,290,296
129,131,135,137,138 discourse marker 178, 180, 181,
blend 216,291,292 182, 183, 184, 187, 189, 190,
body parts 7, 30, 242, 243, 244, 191
247,259,266,267 domain (conceptual) 6, 7, 11, 39,
40, 41, 42, 72, 81, 83, Ill, 112,
117, 118, 127, 133, 134, 142,
categorization 6, 216, 217, 218, 150, 156, 159, 161, 163, 178,
260,262,266,269,273,274 181, 216, 228, 229, 232, 233,
cognition 198 238, 246, 247, 253, 254, 255,
cognitive ease 259, 273, 275 260, 263, 265, 266, 270, 271,
cognitive semantics 198 273,281
collusion 55,58 dynamic systems 118
concept 19, 20, 22, 23, 27, 39, 40, dysphemism 82
41,42
conceptual blending 217
contiguity 6, 7, 8, 52, 74, 79, 200, efficiency 3, 8, 63, 64, 65, 66, 72,
225, 228, 232, 269, 280, 293, 73, 78, lOO, 101, 103, 104, 105,
294 106
310 Index
euphemism 65, 82, 209, 243, 250 homonymy 65, 69, 70, 101, 181,
explanation (of semantic change) 272
92,98, 100, 101, 103, 104, 106 honorific pronouns 243
expressivity 3,8,63,65,66, 70, 73,
78,83,106
extension of meaning 3, 7, 74, 82, Inflectional Phrase Adverb 178,
76, 96, 153, 204, 223, 232, 285, 179, 180, 181, 182, 183, 184,
296,297,298 185, 186, 187, 189, 190, 191
instance (of a schema) 20, 21, 25,
41
feature 32,34,37,41 intensifier 4, 10,237,238,239,240,
figure-ground effect 285,290,293 241, 242, 243, 244, 245, 246,
figures of speech 7, 198, 204, 205, 247, 248, 250, 251, 252, 253,
208, 210, 229 255,257
focus 238,240,241, 243, 244, 246, interpretative mechanism 216, 224,
250,251,252,257 225,232
form categories 217, 218, 221 interpretative strategy 223,232,233
frame (conceptual) 6, 7, 8, 39, 72,
75,76,77,118,127,232
functional motivation (of semantic language change 8, 10, 38, 50, 52,
change) 98, 99, 101, 102, 103, 61,62,63,64,65,66,72,297
104, 105, 106 lexical change 2, 8, 10, 50, 104,
106,215,216,232
linguistic sign 19, 22, 23, 26, 34,
generalization 218, 224, 225 39,42
genus-for-species 199, 200, 201, luxury borrowing 265, 375
204,206,207,210
grammaticalization 7, 9, 49, 57,
147, 148, 151, 156, 157, 160, meaning (criterial and non-criterial
163, 165, 168, 169, 172, 179, features ot) 216, 218, 221, 232
231 metaphor 2,3,4, 7, 8, 9, 10,51,52,
grammaticization see grammatica- 62,63,64,65,70,71,72,73,82,
lization 83, 84, 99, 114, 115, 117, 141,
grammaticized 148, 162, 163, 168 159, 188, 198, 199, 200, 201,
group (of concepts) 260, 261, 262, 202, 204, 205, 207, 208, 210,
269,270,273 216, 217, 225, 226, 227, 228,
229, 232, 234, 290, 291, 292,
298
hearer-oriented strategy 3, 8, 64, 65, metaphorical 38, 115, 141, 149,
103, 104, 105, 106 159,188,238,291
homonymic clash 53, 54, 69, 70, metaphorization 222
101
I"du 311
metaphor-metonymy distinction 180, 181, 243, 249, 2.51, 279,

228 280,298
metonYm see metonymy possession 11, 76, 280, 281, 283,
metonymic 59, 80, 115, 188, 201, 284, 285, 286, 288, 289, 290,
202, 227, 231, 245, 248, 269, 291,292,297,298
270,272,273,285,293 pragmatic strengthening 3,8, 189
metonymy 2, 3, 4, 7, 8, 9, 10, 31, profile 40,41, 148, 149, 155, 165,
51,52,64,65,72,75,80,82,83, 166, 168, 169, 171
99,114,115,117, 139, 198, 199, profiled 40, 149, 150, 151, 152,
200, 201, 202, 203, 204, 205, 153, 154, 155, 156, 158, 159,
206, 207, 210, 225, 226, 227, 160, 163, 165, 166, 168, 169,
228, 229, 230, 231, 232, 233, 171
235, 269, 272, 273, 290, 292, profiling 6,8,20,39,41, 148, 149,
295,297,298 150, 156
motivation for borrowing 266 prototype 2,3,6, 10, 18,20,26,27,
28,30,31,32,33,64,72,76,77,
110, 111, 113, 123, 138,285
Natural Lexicology 98, 106 prototype theory 91, 93, 94, 95, 97,
Natural Phonology 92, 98, 102, 98,100
103, 104, 105 prototypical 6, 7, 10,21,32, 34, 77,
necessity borrowing 265,275 78,81, Ill, 285, 291, 293, 295
neutralisation 30,31 prototypical change 76
prototypicality, prototype effects
91,92,93,98,99, 101, 105, 106
objectification 190, 191
onomasiological 10, 11, 28, 67,
112, 113, 120, 136, 204, 265, qualitative threshold 121, 122, 123,
280,297 124, 127
onomasiology 9, 112, 138
raising 160, 171, 172, 174

paragon 110, 115, 116, 123 recategorization 259, 266, 268, 270,
part-of-relation 203 273,275
part-whole-relation 123, 199, 200, reflexive anaphor 243, 246, 247
201,202,203,206,242,248 restriction of meaning 76, 81, 269,
perceptive experience 275 273
perceptive saliency 274 rhetoric 7, 114, 139, 177, 190, 191,
perfonnance 50,51 198,199,200,201,202
polysemous 30, 33, 52, 113, 114,
119, 138, 181
polysemy 2, 30, 31, 32, 33, 34, 38, semantic change 2, 3, 5, 6, 7, 11,
51, 52, 55, 65, 68, 70, 75, 80, 37,49,61,62,63,66,67,68,69,
312 Index
71, 72, 75, 77, 78, 79, 80, 81, 83, Synecdoche 3, 7, 10, 197, 198, 199,
84, 92, 93, 95, 96, 97, 99, 100, 200, 201, 202, 203, 204, 205,
101, 106, 107, 108, 148, 177, 206,207,208,209,210
179, 181, 188, 189, 191, 223,
237, 238, 243, 256, 272, 257,
279, 280, 284, 290, 292, 297, taxeme 110, 111, 112, 115, 117,
298 118, 120, 122, 123, 124, 125,
semantic extension see extension of 126, 127, 128, 133, 134, 135,
meaning 136, 137, 140, 142
semantic primitive 29 torso·extremities·model 260, 261,
semantic space 4, 297, 280, 282, 267,269,273
284,295,297,298 trajector 149, 150, 152, 153, 154,
semantic restriction see restriction 155, 156, 157, 158, 160, 162,
of meaning 166, 168, 169, 171
semasiological 67, 91, 94, 98, 112, transparency 101, 103, 104, 105,
113, 114, 118, 136, 140, 204, 147, 148, 160, 161, 162, 164,
297 168,171,172, 174
semasiology 9, 111, 138 transparent 162, 163, 164, 171
sentence meaning vs. utterance
meaning 21
signified 19,21,22" 41 unprofiled 155, 156, 166
signifier 19,21,22,25,41 utterance meaning see sentence
similarity 21,56,62,79,200,228 meaning 21
speaker·oriented strategy 2, 64, 65,
78, 103, 104, 105, 106
strategies working out interpreta· valuation 110, 115, 116, 124, 125,
tions 227 128, 138
sub·categorization 261 value 109, 110, 113, 126, 127, 128,
sub·groups (conceptual) 260, 261, 137,138, 139, 141
267 Verb Adverb 178, 183, 185, 186,
subject 148, 149, 150, 151, 152, 190
154, 156, 157, 158, 159, 160,
161, 162, 163, 164, 166, 168,
169,170,171, 172 whole·part·relation see part·whole·
subjectification 3, 6, 8, 9, 10, 147, relation
148, 150, 152, 153, 154, 156, widening of meaning see extension
157" 158, 168, 173, 177, 180, of meaning
188,189,190
subjectivity 8, 179, 180, 188, 190
synecdoche 117
9
I
7113110
11
166149
ISBN 311 0166143

www.deGruyter.com
I
9 783110 166 1 49
ISBN 311 0166143

www.deGruyter.com

(Cognitive Linguistic Research) Andreas Blank, Peter Koch - Historical Semantics and Cognition (1999, Mouton de Gruyter) PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

(Cognitive Linguistic Research) Andreas Blank, Peter Koch - Historical Semantics and Cognition (1999, Mouton de Gruyter) PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Cognitive Linguistics Research

@ Printed on acid-free paper

Library of Congress Cataloging-in-Publication Data

Historical semantics and cognition I edited by Andreas Blank,

Die Deutsche Bibliothek - Cataloging-in-Publication Data

Historical semantics and cognition I ed. by Andreas Blank ; Pe-

The papers collected in this volwne evolved from a symposiwn that

matic framework better than the papers originally presented in Ber-

Marburg / Tubingen Andreas Blank / Peter Koch

Introduction: historical semantics and cognition 1

Section I: Theories and Models

Cognitive semantics and structural semantics 17

Why do new meanings occur? A cognitive typology 61

Diachronic prototype semantics. A digest 91

Cognitive semantics and diachronic semantics: 109

Section ll: Descriptive categories

Losing control: 147

The rhetoric of counter-expectation in semantic change: 177

Synecdoche as a cognitive and communicative strategy 197

Laws of thought, knowledge and lexical change 215

Section DI: Case studies

Intensifiers as targets and sources of semantic change 237

Cognitive ease and lexical borrowing: 259

Cognitive aspects of semantic change and polysemy: 279

List of contributors 307

Andreas Blank and Peter Koch

Cognitive linguistics has had considerable influence on the develop-

1.1. The first group of articles deals with fundamental theoretical

On the ground of his more general model of linguistic change (cf.

1.2. The second group of contributions develops categories for the

"subjectification" in her understanding is the rise of a new sense

1.3. In the third group, theoretical options and categories related to

reveal modes of how we can conceptualize fundamental relations

In each of the contributions to this volume, fundamental topics of

2. Cognitive models and approaches

2.1. European structural semantics has pleaded for a strict theoretical

(l) We arrived at the village. Unfortunately, the church was

The papers brought together in this volume show that it is neces-

2.2. The verbalizing of extralinguistic entities is always related to the

and/or target domains. While most work in cognitive linguistics

2.3. Information relevant to meaning organizes not only in categories

2.4. Specific properties of prototypically organized categories and

speakers intentionally or accidently perceive or reinterpret a given

3. Recent issues in diachronic linguistics

and hearer-oriented linguistic strategies (~ Geeraerts) and, in a

During the last two decades, grammaticalization has been a major

for invisible-hand processes. Once a word or a syntagmatic construc-

5. Two perspectives in semantic investigation:

Traditional synchronic and diachronic semantics distinguish between

With regard to a "unified theory of language change", Liidtke pre-

Combining the onomasiological approach with a well-founded

tegies for verbalizing these concepts, and, moreover, to establish empirical

Theories and Models

This contribution is not specifically about historical semantics. My

types is unfortunate, in that there is much more to Cognitive Gram-

1. The Saussurian heritage

Not the least of the achievements of Ferdinand de Saussure was to

guistic sign", the symbolic association of a signifier (an "acoustic

some respect(s), and can thus be assimilated to it, as a marginal

of utterance meanings that an expression may have on specific oc-

virtue of the structuring of sound by the semiotic system that is a

between singular, dual, and plural. Yet the signification ("value") of

Figure 1. The Saussurean sign

In my opinion, efficie~.9:~9:. ..~:c~ressi:Y.~t.L~~.P-ot !!~.~..:~mpa