1panter Klaus Uwe Radden Gunter Editors Metonymy in Language PDF

Metonymy in Language and Thought
HUMAN COGNITIVE PROCESSING is a forum for interdisciplinary research on the

nature and organization of the cognitive systems and processes involved in speaking and
understanding natural language (including sign language), and their relationship to other
domains of human cognition, including general conceptual or knowledge systems and
processes (the language and thought issue), and other perceptual or behavioral systems
such as vision and non-verbal behavior (e.g. gesture). ‘Cognition’ should be taken
broadly, not only including the domain of rationality, but also dimensions such as
emotion and the unconscious. The series is open to any type of approach to the above
questions (methodologically and theoretically) and to research from any discipline,
including (but not restricted to) di¤erent branches of psychology, arti¼cial intelligence
and computer science, cognitive anthropology, linguistics, philosophy and neuroscience.
It takes a special interest in research crossing the boundaries of these disciplines.
EDITORS
Marcelo Dascal (Tel Aviv University)
Raymond Gibbs (University of California at Santa Cruz)
Jan Nuyts (University of Antwerp)
Editorial address: Jan Nuyts, University of Antwerp, Dept. of Linguistics (GER),
Universiteitsplein 1, B 2610 Wilrijk, Belgium, e-mail: nuyts@uia.ua.ac.be
EDITORIAL ADVISORY BOARD

Melissa Bowerman (Nijmegen); Wallace Chafe (Santa Barbara, CA)
Philip R. Cohen (Portland, OR); Antonio Damasio (Iowa City, IA)
Morton Ann Gernsbacher (Madison, WI); David McNeill (Chicago, IL)
Eric Pederson (Eugene, OR); François Recanati (Paris)
Sally Rice (Edmonton, Alberta); Benny Shanon (Jerusalem)
Lokendra Shastri (Berkeley, CA); Dan Slobin (Berkeley, CA)
Paul Thagard (Waterloo, Ontario)
Volume 4
Klaus-Uwe Panther and Günter Radden (eds)
Metonymy in Language and Thought

Metonymy in
Language and
Thought
Edited by
KLAUS-UWE PANTHER
GÜNTER RADDEN
University of Hamburg
JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
TM The paper used in this publication meets the minimum requirements of
8
American National Standard for Information Sciences — Permanence

of Paper for Printed Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data

Metonymy in language and thought / edited by Klaus-Uwe Panter, Günter Radden.
p. cm. -- (Human cognitive processing, ISSN 1387-6724 ; v. 4)
Papers presented at a workshop held June 23-24, 1996, Hamburg University.
Includes bibliographical references and index.
1. Metonyms Congresses. 2. Cognitive grammar Congresses. I. Panther, Klaus-Uwe, 1942-
II. Radden, Günter. III. Series.
P301.5.M49M48 1999
401’.41--dc21 99-23468
ISBN 90 272 2356 4 (Eur.) / 1 55619 204 5 (US) (alk. paper) CIP
© 1999 – John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. • P.O.Box 75577 • 1070 AN Amsterdam • The Netherlands
John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA
Acknowledgments
The present volume evolved from a workshop on metonymy held at Hamburg

University on June 23 and 24, 1996. This conference was probably the first
international meeting of scholars from Europe, North America and Asia which
was exclusively devoted to the study of metonymy in language and thought. We
would like to express our thanks to all of the participants of the conference, both
presenters and audience, for making the workshop an intellectually stimulating
and revealing event. The workshop was generously funded by the Deutsche
Forschungsgemeinschaft and the Hansische Universitätsstiftung, Hamburg,
whose financial contributions are gratefully acknowledged. For their help in
organizing the workshop and preparing the manuscript we are indebted to our
secretary Elisabeth Himmler and our student assistants Elizabeth Matthis,
Anatol Stefanowitsch and Sebastian Ross-Hagebaum who devoted many long
hours to the conscientious preparation of the volume. Without their dedication
and support this volume would not have been possible. We also benefited
greatly from the insightful comments given to us by two anonymous reviewers.
Finally, we would like to thank the editors of the series Human Cognitive
Processing for accepting this volume, in particular Jan Nuyts, who kindly
assisted us in all editorial matters.
Contents
Introduction 1
Klaus-Uwe Panther and Günter Radden
Part I: Theoretical Aspects of Metonymy
Towards a Theory of Metonymy 17

Günter Radden and Zoltán Kövecses
Speaking and Thinking with Metonymy 61

Raymond W. Gibbs, Jr.
Metonymy and Conceptual Integration 77

Gilles Fauconnier and Mark Turner
Distinguishing Metonymy from Synecdoche 91

Ken-ichi Seto
Aspects of Referential Metonymy 121

Beatrice Warren
Part II: Historical Aspects of Metonymy
Frame and Contiguity: On the Cognitive Bases of Metonymy and

Certain Types of Word Formation 139
Peter Koch
Co-presence and Succession: A Cognitive Typology of Metonymy 169

Andreas Blank
viii Contents
Metonymic Bridges in Modal Shifts 193

Louis Goossens
Metonymy in Onomastics 211

Olaf Jäkel
Part III: Case Studies of Metonymy
Grammatical Constraints on Metonymy: On the Role of the Direct

Object 233
Richard Waltereit
Putting Metonymy in its Place 255

Paul Pauwels
Conversion as a Conceptual Metonymy of Event Schemata 275

René Dirven
Opposition as a Metonymic Principle 289

Christian Vosshagen
Metonymic Hierarchies: The Conceptualization of Stupidity in

German Idiomatic Expressions 309
Kurt Feyaerts
The Potentiality for Actuality Metonymy in English and Hungarian 333

Klaus-Uwe Panther and Linda Thornburg
Part IV: Applications of Metonymy
“Mummy, I like being a sandwich”: Metonymy in Language

Acquisition 361
Brigitte Nerlich, David D. Clarke, Zazie Todd
Recontextualization of Metonymy in Narrative and the Case of

Morrison’s Song of Solomon 385
Anne Pankhurst
Contents ix
List of Contributors 401
Subject index 405
Author index 417
Metonymy and metaphor index 423

Introduction
Klaus-Uwe Panther and Günter Radden

1. Background
Eighteen years after Lakoff and Johnson’s (1980) seminal work on the role of
metaphor in conceptualization, which sparked a vast amount of research in
cognitive linguistics, it has become increasingly apparent that metonymy is a
cognitive phenomenon that may be even more fundamental than metaphor.
We believe that the contributions give a fair view of the state of the art in
metonymic research, although we are also aware of the fact that a great many
questions about metonymy still remain unanswered, some of which will be
addressed below.
The cognitive understanding of metaphor and metonymy is certainly at
variance with both naive and traditional scholarly views, which have strongly
been influenced by centuries of rhetorical and literary studies. The cleavage
between literal and figurative language, which was taken for granted by
traditional rhetoric and linguistics, has recently been challenged by Gibbs
(1994: 24–79; and this volume). Still, we owe the first basic insights into the
nature of tropes to Greek, Roman and medieval scholars, modern literary critics
and linguists. Many different classifications of tropes have been proposed,
starting with Aristotle, who subsumed metonymy and synecdoche under
metaphor, and more recently by the Groupe de Liège or Groupe µ, which
subsumed metaphor and metonymy under synecdoche (see Schofer and Rice
1977). Some of these ideas on metonymy definitely have a modern, cognitive
tinge. Various contributors to this volume (Koch; Blank; and Nerlich, Todd and
Clarke) link their cognitive approach to metonymy to this rhetorical tradition.
2 Klaus-Uwe Panther and Günter Radden
The authors of the contributions to this volume have different theoretical

backgrounds and are affiliated with different disciplines: linguistics, psycho-
linguistics, psychology and literary studies. Many of them share the assumption
that metonymy is a cognitive phenomenon underlying much of our ordinary
thinking and that the use of metonymy in language is a reflection of its
conceptual status. The conceptual framework within which metonymy is
understood in most of these contributions is that of scenes, frames, scenarios,
domains or idealized cognitive models (ICMs). Within these models, a met-
onymic link may be established between two conceptual entities in the broadest
sense. This view supersedes the traditional assumption of metonymy as having
primarily a referential function, a view which was still held by Lakoff and
Johnson (1980).
The papers read at the conference and collected in this volume address a
wide range of topics related to metonymy. The papers have been grouped into
four parts. Part 1 deals with theoretical aspects of metonymy as a cognitive
process. Part 2 investigates historical aspects of metonymy within a cognitive
framework. Part 3 contains a number of case studies on selected metonymies
or aspects of metonymy. Part 4 explores the notion of metonymy in its
application to language acquisition and literary criticism.
2. Contributions to the volume
2.1. Theoretical aspects of metonymy
Three papers address the role of metonymy in language and thought from a
broader theoretical perspective. The issue of the conceptual nature of me-
tonymy is investigated by Günter Radden and Zoltán Kövecses in their
paper “Towards a theory of metonymy.” Their approach is based on the notion
of idealized cognitive model (ICM) as proposed by Lakoff (1987). Metonymy
is understood as a conceptual process in which one conceptual entity, the
‘target,’ is made mentally accessible by means of another conceptual entity,
the ‘vehicle,’ within the same ICM. In principle, either of the two conceptual
entities related may stand for the other, i.e., metonymy is basically a reversible
process. There are, however, a number of cognitive principles which govern
the selection of a preferred vehicle. These principles lead to natural, or
‘default’ cases of metonymy and often escape our awareness. These principles
Introduction 3
may, however, be overridden by factors such as style, taboo or politeness and

may lead to the creation of expressive, or ‘non-default’ cases of metonymies.
In his contribution on “Speaking and thinking with metonymy,”
Raymond Gibbs lays the foundations for the study of metonymy in ordinary
language as well as literary discourse. He situates metonymy in a larger
cognitive context and adduces evidence for the conceptual basis of me-
tonymy. Thus, the Gricean notion of conversational implicature can be seen as
being metonymically motivated; metonymic reasoning may also contribute to
the establishing of coherence by means of ‘conceptual anaphors.’ Finally,
Gibbs also shows how metonymy is operative in discourse and leads to a
better understanding of contextually determined reference, indirect speech
acts, and colloquial tautologies.
Gilles Fauconnier and Mark Turner explore another conceptual aspect
of metonymy. In their paper “Metonymy and conceptual integration,” the
authors investigate the interaction of conceptual blending and metonymy.
They demonstrate that Lakoff’s and Kövecses’ unidirectional model of con-
ceptual metaphor does not account for expressions such as smoke is coming
out of his ears, which can only be understood as resulting from ‘blending’ the
source and the target domains, since, literally, there are no ears in the source
domain and there is no smoke in the target domain. Fauconnier and Turner’s
approach has far-reaching consequences for the theory of metaphor in that it
may very well turn out that most metaphors involve conceptual integration.
Also, conceptual entities may be metonymically linked in a blended space. In
the well-known symbolic representation of death as The Grim Reaper, the
input elements ‘scythe,’ ‘cowl,’ and ‘skeleton’ are conceptually integrated.
Thus the blend ‘shortens’ the metonymic distance between originally non-
contiguous conceptual entities.
Ken-ichi Seto’s paper “On distinguishing synecdoche from metonymy”
argues for a clear conceptual distinction between metonymy and synecdoche.
Despite the current interest in these tropes, they have not yet been defined in a
satisfactory fashion. According to Seto, the reason for this lack of precision
resides in the confusion between taxonomies and partonomies. Taxonomies
involve ‘kind of’ relations in a hyponymically-structured conceptual domain
(e.g., a ham sandwich is a kind of food), whereas partonomies involve ‘part of’
relations in the physical world (as in an arm is a part of the body). Seto calls
the former relations, which are defined by semantic inclusion, C-relations
(category relations); the latter relations, which are constituted by spatio-
temporal contiguity between entities in the physical world, are called E-

relations (entity relations). Seto proposes to reserve the term ‘synecdoche’ for
C-relations and ‘metonymy’ for E-relations. On the basis of this distinction, he
develops classifications of metonymic and synecdochic relations.
In her paper “Aspects of referential metonymy,” Beatrice Warren detects
structural parallels between referential metonyms on the one hand and noun-
noun compounds and denominal verbs on the other hand. These three construc-
tions have in common that they involve two referents and an implicit link
connecting them. In noun-noun compounds, the referring item is explicit,
whereas it is implied in metonyms. According to Warren, metonymy is basically
an abbreviation device and, apart from finding the implicit referring item, its
interpretation involves retrieving a relation. Metonymic relations are restricted
to a small number of recurrent links. This makes the interpretation of metonymy
less demanding, or even effortless, when compared to the open-ended interpre-
tation of metaphor, which involves working out several matching links (of
similar attributes) between the conventional and intended referents.
2.2. Historical aspects of metonymy
The outstanding role of metonymy in triggering linguistic change has long

been recognized and led to various classifications of types of metonymy, some
of which are described in the papers below. The cognitive paradigm in
linguistics offers a new analytic tool for analyzing historical data. Four papers
are devoted to the operation of metonymy in historical processes of language.
In his paper “Frame and contiguity: On the cognitive bases of metonymy
and certain types of word formation,” Peter Koch investigates metonymically-
induced changes of meaning. He makes use of the conceptual networks
provided by frame theory. Frames may be seen as conceptual gestalts and, in
metonymic changes of meaning, a new sense is highlighted as the ‘figure’ while
the old sense serves as its ‘ground’ within the frame. Figure/ground effects also
account for the origin of metonymy in discourse. Koch distinguishes between
three types of ad hoc metonymic innovation inducing a metonymic change:
hearer-based inferential innovations such as the interpretation of ‘fireplace’ as
‘fire,’ speaker-based ‘imprecise’ innovations such as the sense of ‘hip’ for
‘thigh,’ and expressive innovations such as ‘skull’ for ‘head.’
Andreas Blank presents further arguments for a frame-theoretical ap-
proach to metonymy. His paper “Co-presence and succession: A cognitive
Introduction 5
typology of metonymy” critically examines the classifications of metonymy

that have been proposed. He claims that most typologies of metonymy are
defective in that a number of metonymies do not fit into any of the categories
proposed. He argues that all types of metonymic changes can be subsumed
under two major types: relations between entities that are co-present within a
frame, and those that are successive within one frame or two related frames.
His typological model of metonymy comprises three levels of abstraction: the
two domains of co-present and successive contiguity at the highest level,
schematic types of contiguity at the intermediate level and concrete linguistic
metonymies at the lowest level.
A specific problem of semantic change is studied by Louis Goossens in his
paper “Metonymic bridges in modal shifts.” He investigates the conceptual shift
of the English modal must from a deontic to an epistemic meaning. This general
shift, which also applies to other modal verbs, has been accounted for in two
different ways: as a metaphorical mapping from the sociophysical world onto
the epistemic world (Sweetser 1990), or, especially in grammaticalization
studies, as a shift triggered by context-induced inference. Both synchronic and
diachronic data on the usages of must suggest that the shift from deontic to
epistemic is a gradual process, which can be accounted for by metonymic
bridges.
The metonymic approach to historical linguistics is also relevant to the
discipline of onomastics. In his paper “Metonymy in onomastics,” Olaf Jäkel
investigates the cognitive motivation underlying naming patterns as evi-
denced in the etymologies of German surnames. Apart from a number of
surnames which are not motivated or whose motivation is obscure, surnames
are coined by means of three principal patterns of naming: genealogy, profes-
sion and metonymy. Metonymic naming strategies make use of three types of
metonymy: ‘utensil metonymy’ (IMPORTANT UTENSIL FOR PERSON ) as in
Bohnsack ‘beanbag,’ ‘quality metonymy’ (SALIENT QUALITY FOR PERSON )
as in Wunderlich ‘strange,’ and ‘location metonymy’ (PLACE OF ORIGIN OR
RESIDENCE FOR PERSON ) as in Langacker ‘long field.’
2.3. Case studies of metonymy
A variety of case studies investigate the operation of metonymy on various

linguistic and conceptual levels. One paper is concerned with metonymy in
grammar (Waltereit); three papers look at selected lexical and semantic me-
tonymies (Dirven; Pauwels; Voßhagen); one paper studies the metonymic

structure of a particular concept in its cultural context (Feyaerts), and another
paper investigates the degree of exploitation of a particular metonymic prin-
ciple in two genetically unrelated languages (Panther and Thornburg).
Richard Waltereit’s contribution “Grammatical constraints on met-
onymic reference: On the primacy of the direct object” demonstrates that
metonymy also plays an important role at the level of grammatical relations.
First, metonymic transfer can involve the insertion of a participant into a given
thematic role such as Le 53 est rentré ‘No. 53 is back,’ where the number
stands for a hotel guest. Second, it can also involve a transfer of semantically
contiguous thematic roles as in Papa va balayer ta chambre ‘Daddy will
sweep your room’ vs. Papa n’a pas encore balayé les débris de verre ‘Daddy
hasn’t swept up the broken glass yet,’ where the container (the room) is
semantically contiguous with the object contained (the broken glass).
Waltereit argues that, with regard to metonymic transfers, the direct object has
primacy over the subject and other grammatical relations. The metonymically
privileged status of the direct object is caused by three factors: it is semanti-
cally opaque and, hence, allows for a number of thematic roles to fit into the
direct object slot; it is the argument that is semantically closest to the verb,
which entails that its referential autonomy is somewhat weakened; and it is
syntagmatically closer to the subject than other (oblique) arguments.
In his lexical-semantic study “Putting metonymy in its place,” Paul
Pauwels investigates the metonymic structure of four related verbs: put, set,
lay, and place. Pauwels’ corpus-based investigation shows that the majority of
examples were not of the traditional nominal or referential kind. In his corpus,
metonymy often seems to function as a ‘euphemistic avoidance strategy.’ But
it can also serve as a focusing strategy, which, in extreme cases, may result in
dysphemism. The most frequent metonymic type Pauwels encounters in his
corpus is based on a relation of inclusion, where a more general concept stands
for a more specific concept, or vice versa.
In his paper “Conversion as a conceptual metonymy of basic event
schemata,” René Dirven investigates the phenomenon which is generally
known as conversion or zero-derivation, in particular, the conversion from
nouns to verbs (e.g., author vs. to author). He shows that the process of
conversion is typically found in three event schemata: the action schema, the
location and motion schema, and the essive schema. Conversion is regarded as
a process in which one participant in the event schema is metonymically focused
Introduction 7
upon, but the whole event is conceptually involved. For example, in the action
schema the participants patient (fish), instrument (hook), and manner (pearl
fishing) are most frequently converted into new verbs (to fish, to hook, to fish
pearls, respectively). Five participant types are regularly exploited to yield new
verbs. In conclusion, Dirven raises the question if this selectivity is the result of
sociocultural saliency or rather a matter of linguistic preference.
Another semantic study, Christian Voßhagen’s paper “Opposition as a
metonymic principle,” focuses on antonymy as a metonymic relation. This
metonymy shows up in irony, where usually a positive concept metonymically
stands for a negative concept, and in some conventionalized lexical items such
as terribly in It was terribly amusing. As a rule, the metonymy applies to
evaluative concepts, which are semantically scalar but are reinterpreted as
complementary.
In his paper “Metonymic hierarchies: The conceptualization of stupidity
in German idiomatic expressions,” Kurt Feyaerts studies the metonymic
structure of everyday expressions of stupidity. For example, an expression
such as Du bist wohl nicht von hier? ‘You are not from here, are you?’
exemplifies the metonymic folk model OUTGROUP ORIGIN FOR STUPIDITY .
Feyaerts shows that metaphorically and metonymically organized hierarchies
have major structural characteristics in common. Higher-level metonymies
tend to be cross-culturally valid, while lower-level metonymies are more
culture-specific.
Klaus-Uwe Panther and Linda Thornburg emphasize the importance
of a cross-linguistic comparison of conceptual metonymies. In their paper “The
POTENTIALITY FOR ACTUALITY metonymy in English and Hungarian,” they
analyze the extent to which this metonymy is exploited across two genetically
unrelated languages, English and Hungarian. They explore its operation in
seven conceptual domains: sense perceptions, mental states and processes,
hedged performatives, indirect speech acts, (extralinguistic) actions, character
dispositions, and acquired skills. In some of these domains, the POTENTIALITY
FOR ACTUALITY metonymy is much more productive in English than in
Hungarian. The most striking contrast between the two languages emerges in
the domain of sense perceptions: whereas English systematically exploits the
metonymy in sentences such as I can taste the vanilla (for I taste the vanilla),
Hungarian systematically excludes the metonymy and resorts to a non-modal
construction in the indicative mood. The authors also discuss the relationship
between Gricean maxims, conversational implicatures and metonymy.
2.4. Applications of metonymy
Two contributions are devoted to the significance of metonymy in language

acquisition and literary criticism. The role of metonymy in language acquisi-
tion is investigated by Brigitte Nerlich, David D. Clarke and Zazie Todd.
Their paper “‘Mummy, I like being a sandwich’: Metonymy in language
acquisition” may in fact be the first study on the production and understanding
of metonymy in this field, whereas studies on the production and understand-
ing of metaphor in language acquisition proliferate. In child language the use
of metonymy serves two different functions: it is a means of extending the
known stock of words to cope with increasing communicative needs and of
exploiting ‘natural pathways of meanings’ creatively. The former use of
metonymy is a pragmatic strategy which leads to ‘compelled’ overextension.
Compelled overextensions are typically found up to the age of 2;5, while by
the age of four children start producing metonymy for creative purposes. This
use of metonymy is referred to by the authors as ‘creative metonymical
shrinking.’ Children’s comprehension of metonymy is empirically studied
using a group of 2–3 year-olds and a group of 4–5 year-olds.
In her paper “Recontextualization of metonymy in narrative and the case
of Morrison’s Song of Solomon,” Anne Pankhurst explores the function of
metonymy in narrative fiction. In the novel analyzed, the reader has to activate
metonymic strategies in order to understand the impact of an apparently
simple object, an earring. The earring serves several metonymic functions: it
is, in particular, a means of identifying its wearer and, at a macro-structural
level, holds together different episodes of the novel. Pankhurst argues that the
complex use of metonymy in this narrative cannot be accounted for by a single
theory. The most promising approach to understanding the complex processes
of reference and recontextualization in this world of fiction are provided by
Riffaterre’s functional view of metonymy and Gibbs’ metonymic models of
thought.
3. Perspectives for future research
The papers collected in this volume certainly contribute to a deeper under-

standing of the conceptual nature and function of metonymy. At the same
time, the contributors are aware of the fact that many aspects of metonymy are
Introduction 9
still poorly understood. Of the many remaining problems to be solved we will

single out two and briefly discuss them here:
i(i) the nature of metonymic shift;
(ii) the pragmatic function of metonymy.
3.1. The nature of metonymic shift
We assume that metonymy is not, as has often been taken for granted, merely
a matter of the substitution of linguistic expressions but a cognitive process
that evokes a conceptual frame. The notion of ‘conceptual frame’ is meant
here as a cover term for what is variously called ‘domain,’ ‘idealized cognitive
model’ (ICM), ‘schema,’ ‘scenario,’ ‘script,’ etc. in the cognitive-linguistic
literature (cf. also Blank, this volume; and Koch, this volume). The ‘substitu-
tion view’ of metonymy claims that the name of one thing is used in place of
that of another thing to which it is related. As will be shown below, this view
has serious draw-backs. Following Langacker (1993: 30), we assume that
“metonymy is basically a reference-point phenomenon [...] affording mental
access to the desired target.” Let us consider the conceptual frame of a
straightforward case of metonymy as exemplified in:
(1) The first violin has the flu.
The concept ‘the first violin’ is part of a knowledge structure that it evokes. As
a musical instrument, a violin is immediately associated with a violinist as the
player of that instrument. Moreover, the first violinist is defined as a member
of a larger group of musicians, the symphony orchestra. Among the musicians
of the orchestra, the first violinist is the most outstanding member. Finally, our
knowledge of orchestras includes, among other things, the notion of music and
its representation in scores. The predication has the flu as well as the attribute
first trigger a non-literal interpretation of the noun phrase the first violin. Thus,
the metonymic reading in (1) involves a shift from the instrument to the
musician as the most readily available element in the frame. Through this
metonymic shift, the reference point (‘the first violin’) is backgrounded and
the desired target (‘the first violinist’) is foregrounded. This conceptual shift is
reflected in grammatical form: thus the second sentence of (2a), in which she
anaphorically refers back to the target, is a felicitous continuation of (1),
whereas the second sentence of (2b), in which the pronoun is coreferential
with the reference point, is not:
(2) a. The first violin has the flu. She cannot practice today.
b. # The first violin has the flu. It is a Stradivarius.
But now consider a situation described by the following sentence:
(3) My ex-husband is parked on the upper deck.
The expression my ex-husband evokes a rich mental script involving mar-
riage, divorce, etc., all of which, however, do not seem to play a role in the
metonymic interpretation of this utterance. In contrast to (1), it is not the
conceptual frame of the noun phrase that is exploited for the metonymic
interpretation, but the predicate is parked on the upper deck. The predicate
helps identify the target of the metonymic shift, i.e., ‘my ex-husband’s ve-
hicle.’ Concomitantly, the metonymic reference point (‘my ex-husband’) is
foregrounded while the conceptual target (‘my ex-husband’s vehicle’) is
backgrounded. This analysis is corroborated by the linguistic fact that an
anaphoric pronoun cannot refer back to the target expression as in (4a), but
only to the reference-point expression as in (4b):
(4) a. # My ex-husband is parked on the upper deck. It has a California
license plate.
b. My ex-husband is parked on the upper deck. He is taking the
bus today.
On the basis of pronominal facts as in (4), Nunberg (1995: 111) claims that in
sentences such as (3) it is not the subject that is used metonymically but the
predicate, which “contributes a property of persons, the property they possess
in virtue of the locations of their cars.” This type of analysis thus postulates
that the metonymic shift is not achieved through the noun phrase but involves
a “predicate transfer” (for a critique of Nunberg’s theory cf. Kleiber 1995). It
is possibly more plausible and intuitively more satisfying, however, to view
the metonymy in (3) as an instance of referential shift, i.e., to understand my
ex-husband in the sense of ‘my ex-husband’s vehicle.’ We suggest that the
choice of the pronoun might be governed by a general cognitive principle
according to which humans take precedence over non-humans (see also
Radden and Kövecses, this volume). This principle would account for the fact
that the human entity in the frame seems to be foregrounded irrespective of
whether it is the reference point or the target.
A further point in need of clarification relates to the relationship among
the elements in the frame. In the case of an artifact as in example (1), the user
Introduction 11
of the artifact is so tightly integrated into the frame that the metonymic reading
has become lexicalized and is listed as a separate sense in dictionaries. In
contrast, with the exception of well-known individuals such as Shakespeare,
Mozart and Einstein, who are closely associated with their artistic or scientific
products, humans do not seem to be consistently tied to a frame which leads to
lexicalized metonymic senses. Thus it is highly unlikely that ex-husband
would have ‘car’ as one of its conventional senses. This will even hold for
human nouns such as car-dealer, which explicitly contains the concept of
‘car’ as an integral part of its frame.
We believe that both reference point and target are always present as
elements of the conceptual frame, but are highlighted to different degrees.
This can be shown by the following minimal pair which exemplifies two ways
of highlighting frame elements:
(5) a. The harpsichord has the flu. His part has been taken over by the
grand piano.
b. The harpsichord has the flu. Its part has been taken over by the
grand piano.
In contrast to (2a), in which only the human target can be foregrounded, the
sentences under (5) seem to allow the foregrounding of either the human
performer or the instrument. The possessive pronoun his in (5a) anaphorically
refers to the musician who is metonymically targeted by the harpsichord,
whereas the pronoun its in (5b) is grammatically congruent with the reference-
point expression, but conceptually relates to the part assigned to the harpsi-
chord in the score.
The car-parking situation described in (3), however, does not lend itself
to similar highlighting of either the reference point or the target as in (5). It is
much more difficult to foreground the target when the reference point is
human and the target is non-human. It seems, however, possible to say (6), in
which the anaphoric pronoun they highlights the cars and not their owners:
(6) ? Myex-husband and his girlfriend are parked next to each other.
They are both Fords.
In this sentence, the noun phrases my ex-husband and his girlfriend are
metonymically interpreted as ‘my ex-husband’s car’ and ‘his girlfriend’s car,’
respectively, i.e., there is a referential shift from HUMAN to NON - HUMAN .
The discussion thus far has looked at one area in which metonymic
highlighting is reflected in grammar. There are, of course, other grammatical

phenomena such as number and gender agreement which may be adduced as
further evidence that certain elements of a frame are given more prominence
than others. These issues shall, however, not be pursued here (for discussion
on this point see Nunberg 1995; and Kleiber 1995).
3.2. The pragmatic function of metonymy
An issue which has received relatively little attention in the discussion of

metonymy concerns the pragmatic function of metonymy in conversation:
why is metonymy used at all? Why is ‘literal’ language not the prevailing
means of communication? Part of the answer may lie in Sperber and Wilson’s
(1995: 158) principle of relevance: “Every act of ostensive communication
communicates a presumption of its own optimal relevance.” Sperber and
Wilson (1995: 153) hypothesize that a linguistic expression is optimally
relevant if it produces maximal contextual effects with a minimum of process-
ing effort.
As an illustration of the principle of minimal processing effort, consider a
situation in which nurses talk about their patients as in (7a) and (7b):
(7) a. It’s time for my gall bladder’s medication.
b. It’s time for Randolph’s medication.
Even when the patient’s name is known as in (7b), the metonymic noun phrase
my gall bladder may provide the easiest access to the targeted referent. To the
personnel in a hospital, information about a patient’s ailments is in general
more relevant than other attributes, such as the patient’s bald-headedness, his
level of education, etc. Hence, for the medical staff the metonymic wording
such as (7a) may be the most efficient way of identifying a patient. To persons
outside the hospital context, however, such metonymic reference represents
neither an economical nor an appropriate referential shortcut.
The significance of the principle of maximal contextual effects is illus-
trated by example (8a), which is routinely understood in the sense of (8b):
(8) a. The Chicago Bulls were able to nail down their fifth champion-
ship in game 6.
b. The Chicago Bulls nailed down their fifth championship in
game 6.
Introduction 13
Literally, sentence (8a) states the ability of the well-known basketball team to
win the championship, but metonymically this utterance implicates their
actually winning the game. The metonymy involved may be described as a
metonymic shift from POTENTIAL TO ACTUAL (see Panther and Thornburg,
this volume). Why should a sports commentator choose the metonymic predi-
cation were able to nail down rather than nailed down, which, after all, seems
to be the more economical wording of the two? Again, an explanation may be
provided within the framework of relevance theory. Sentence (8a) triggers
more contextual effects, i.e., pragmatic implications, than sentence (8b). Both
utterances convey the actuality of winning the championship, which is expli-
citly stated in (8b), but only conversationally implicated in (8a). Yet, the
metonymic wording in (8a) has the advantage of communicating additional
information: in stating ability, the predicate were able to strongly implicates
the notions of ‘effort,’ ‘difficulty,’ and ‘positive achievement,’ none of which
is present in (8b). The greater length of (8a) is thus more than compensated for
by the number of desirable contextual effects that it triggers.
As examples (7) and (8) demonstrate, a metonymic expression is hardly
ever completely equivalent in its pragmatic force to its ‘literal’ counterpart.
Thus, these data provide more evidence against the traditional ‘substitution
view’ of metonymy. In conclusion, this view of metonymy as a means of
providing maximal contextual effects with a minimum of processual effort
certainly opens new avenues of future research on the role of this as well as
other figurative modes of thought.
References
Gibbs, Raymond W., Jr.

1994 The Poetics of Mind: Figurative Thought, Language, and Understanding.
Cambridge: Cambridge University Press.
Kleiber, Georges
1995 Polysémie, transferts de sens et métonymie intégrée. Folia Linguistica 29:
105–132.
Lakoff, George
1987 Women, Fire, and Dangerous Things: What Categories Reveal about the
Mind. Chicago: The University of Chicago Press.
Lakoff, George, Mark Johnson
1980 Metaphors We Live By. Chicago: The University of Chicago Press.
Langacker, Ronald W.
1993 Reference-point constructions. Cognitive Linguistics 4: 1–38.
Nunberg, Geoffrey
1995 Transfers of meaning. Journal of Semantics 17: 109–132.
Schofer, Peter, Donald Rice
1977 Metaphor, metonymy, and synecdoche revis(it)ed. Semiotica 21: 121–147.
Sperber, Dan, Deirdre Wilson
1995 Relevance: Communication and Cognition. Second edition. Oxford and
Cambridge, MA: Blackwell.
Sweetser, Eve
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of
Semantic Structure. Cambridge: Cambridge University Press.
Part I
Theoretical Aspects of Metonymy

Towards a Theory of Metonymy
Günter Radden
Zoltán Kövecses
Eötvös Loránd University, Budapest
1. The cognitive view of metonymy
The aim of this paper is to propose a conceptual framework of metonymy as a

cognitive process.1 Unlike metaphor, metonymy has always been described in
conceptual, rather than purely linguistic, terms. In analyzing metonymic rela-
tionships, traditional rhetoric operated with general conceptual notions such
as CAUSE FOR EFFECT, CONTAINER FOR CONTENTS, etc. Still, metonymy was
mainly seen as a figure of speech, i.e., it was basically thought of as a matter of
language, especially literary or figurative language. This view of metonymy is
reflected in standard definitions, which tend to describe metonymy as “a
figure of speech that consists in using the name of one thing for that of
something else with which it is associated” (Webster’s Third New Interna-
tional Dictionary). These kinds of definition thus claim that metonymy oper-
ates on names of things, involves the substitution of the name of one thing for
that of another thing and assumes that the two things are somehow associated.
The cognitive view of metonymy espoused here makes different assumptions:
(i) Metonymy is a conceptual phenomenon;
(ii) Metonymy is a cognitive process;
(iii) Metonymy operates within an idealized cognitive model.
18 Günter Radden and Zoltán Kövecses
1.1. Metonymy is a conceptual phenomenon
Metonymy is claimed to be not just a matter of names of things, but essentially

a conceptual phenomenon. As already pointed out by Lakoff and Johnson
(1980: Ch. 8), metonymy, like metaphor, is part of our everyday way of
thinking, is grounded in experience, is subject to general and systematic
principles and structures our thoughts and actions (cf. also Gibbs 1994: 324–
333 and this volume). Lakoff and Johnson’s example of the metonymy in
She’s just a pretty face illustrates the general conceptual nature of metonymy.
We derive the basic information about a person from the person’s face. In our
culture, this is reflected in the tradition of portraits in painting and photogra-
phy. The conceptual metonymy THE FACE FOR THE PERSON is therefore part of
our everyday way of thinking about people.
The conceptual nature of metonymy is even more clearly manifested in
the structure of categories. In his discussion of metonymic models, Lakoff
(1987: 79–90) demonstrates that a member of a category may stand for the
whole category and thereby account for prototype effects. These salient
members may not even have a name so that the metonymic transfer merely
operates at the conceptual level. His example of the stereotypical subcategory
‘housewife mother’ illustrates this point: We tend to think of the category
‘mother’ in terms of this stereotypical member even if the submember remains
unnamed. Since most categories have prototypical structure, we may conclude
that basically all categories have metonymic structure.
The use of metonymic expressions in language is primarily a reflection of
general conceptual metonymies and is motivated by general cognitive prin-
ciples. We claim that all metonymies are ultimately conceptual in nature and
that many, if not most, metonymies do not even show up in language.
1.2. Metonymy is a cognitive process
The traditional view defines metonymy as a relationship involving substitu-

tion. This view is reflected in the notation generally used for stating met-
onymic relationships, namely, X STANDS FOR Y. In the above example of She’s
just a pretty face, the name face is thus taken to be a substitute expression for
person, so that the sentence is assumed to mean ‘she is just a pretty person.’
But this cannot be the whole meaning since She is a pretty person does not
mean that she is pretty ‘all over,’ but it suggests that, most importantly, she has
a pretty face. This can be seen in the oddity of a sentence expressing a counter-
expectation: ?She is a pretty person but does not have a pretty face. The two
metonymies, THE FACE FOR THE PERSON and THE PERSON FOR THE FACE, thus
complement each other: A person’s face evokes the person and a person
evokes the person’s face. Metonymy does not simply substitute one entity for
another entity, but interrelates them to form a new, complex meaning.2 To use
Warren’s (this volume) example: “We do not refer to music in I like Mozart,
but to music composed by Mozart; we do not refer to water in The bathtub is
running over, but to the water in the bathtub.” Metonymic relationships should
therefore more adequately be represented by using an additive notation such
as X PLUS Y, as suggested by Radden (in print). For the sake of simplicity, we
will keep the traditional formula X FOR Y with the proviso, however, that the
metonymic process is not understood to be one of substitution.
The metonymic process consists in mentally accessing one conceptual
entity via another entity. This is the cognitive explanation Langacker (1993:
30) offers for metonymy. He conceives of metonymy as a reference-point
phenomenon in which one conceptual entity, the reference point, affords
mental access to another conceptual entity, the desired target.3 We will refer
to the reference-point entity as the ‘vehicle’ and the desired target simply as
the ‘target.’ In the example of She’s a pretty face, the ‘pretty face’ serves as
the vehicle for accessing the ‘person’ as the target; in the reverse description,
She’s a pretty person, the ‘person’ serves as the vehicle for accessing the
person’s ‘pretty face’ as the target. In either construal, both the vehicle and the
target are conceptually present. However, one of them is seen as being more
salient than the other and is therefore selected as the vehicle.
1.3. Metonymy operates within an idealized cognitive model
The notion of ‘contiguity’ is at the core of most definitions of metonymy.4

Traditional approaches locate contiguity relationships in the world of reality,
whereas cognitive approaches locate them at the conceptual level.5 Lakoff
and Johnson (1980) think of contiguity in terms of the whole range of
conceptual associations commonly related to an expression, Lakoff (1987)
accounts for metonymic contiguity within the framework of idealized cogni-
tive models (ICMs), Croft (1993) deals with contiguity relations in terms of
encyclopedic knowledge representation within a domain or domain matrix,
Blank (this volume) and Panther and Thornburg (this volume) describe the
network of conceptual contiguity by using the notion of frame and scenario,
respectively.
While all of these models are comparable with respect to claiming a

cognitive basis, we believe that Lakoff’s (1987) framework of ‘idealized
cognitive models’ (ICMs) may capture metonymic processes best. The ICM
concept is meant to include not only people’s encyclopedic knowledge of a
particular domain but also the cultural models they are part of. The ICM
notion is not restricted to either the world of reality, the world of conceptual-
ization or the world of language but, as will be shown in Section 2, may cut
across these ontological realms. ICMs and the network of conceptual relation-
ships characterizing them give rise to associations which may be exploited in
metonymic transfer.
The impact which ICMs may have on metonymic (and metaphorical)
transfer shall be illustrated by way of the changes of meaning which the word
hearse underwent in the history of English. The semantic history of hearse
may have proceeded in the following steps.6 In medieval farming, the word
originally denoted a triangular harrow with pins and was then metaphorically
applied to a triangular frame for supporting candles at church services. The
new ‘candle-frame ICM’ evoked the functionally most salient part of it, the
candles. Our general knowledge of the ‘candle ICM,’ in its turn, gave rise to
the metonymic focus on the process of burning. In the Middle Ages, candles
were made of wax, were very expensive and were only lit for special occa-
sions. This Medieval ‘candle-burning ICM’ explains why the burning of
candles came to be metonymically associated with a special liturgical occa-
sion, Tenebrae, the Holy Week before Easter. The Medieval ‘Tenebrae ICM’
accounts for a further metonymic step. In the church service of the Holy
Week, all candles were gradually extinguished to commemorate the darkness
at Christ’s crucifixion. The burning candle was a metaphor for man’s life, and,
as an entailment, its extinction a metaphor of man’s death. The whole candle-
burning event was thus metonymically restricted to its final part, the extinction
of the candle. The ‘crucifixion ICM’ was then metonymically extended to
people’s death in general. The ‘death ICM’ accounts for the metonymic
highlighting of a salient part surrounding people’s death, the funeral. The
‘funeral ICM’ involves several parts, many of which were described by the
word hearse: the dead body, the coffin, the bier, the tomb, the funeral pall, the
framework supporting the pall, and the carriage for carrying the coffin.
Among these parts, the moving carriage eventually appeared to be the most
salient element of the ‘funeral ICM.’ The sense development of hearse from
‘harrow’ to ‘vehicle for conveying a dead person to the place of burial’ is
predominantly the result of different types of metonymic processes which
operated within cultural or general ICMs.

The metonymic stages in the sense development of hearse involved, not
only things, but also events: the burning of the candle, the extinction of the
candle, death, the funeral and the procession. This is to be expected in view of
the many possible relationships which may hold in an ICM. Metonymic
processes are thus not restricted to reference;7 they occur at the purely
conceptual level (categorization, linguistic reasoning), at different levels of
language (lexis, morphology, syntax, discourse), in different linguistic func-
tions (reference, predication, speech acts), and as a linkage interrelating
different ontological realms (concepts, forms, and things/events). In order to
be able to describe such diverse phenomena in a unified way, we will adopt the
widest possible view of metonymy.
1.4. Theoretical issues of metonymy
On the basis of the three cognitive properties of metonymy discussed above,

we will define metonymy as follows:
Metonymy is a cognitive process in which one conceptual entity, the vehicle,
provides mental access to another conceptual entity, the target, within the
same idealized cognitive model.
This working definition is useful in that it allows us to raise further important

empirical and theoretical issues. We believe that, amongst others, the follow-
ing questions need to be addressed in developing a theoretical framework of
metonymy.
A first question we need to ask is where do we find metonymy? Accord-
ing to the above definition, metonymy may occur wherever we have idealized
cognitive models. We have ICMs of everything that is conceptualized, which
includes the conceptualization of things and events, word forms and their
meanings, and things and events in the real world. We will refer to these types
of conceptualization as ‘ontological realms.’ ICMs are not restricted to a
single ontological realm, but may also interrelate ontological realms. For
example, people tend to see a close relationship between the two entities
which establish a sign: the concept of a thing and its name. This ICM cross-
cuts two ontological realms and, as will be shown below, leads to metonymy.
The notion of metonymy which follows from the conceptual definition given
above thus has much wider application than that of traditional approaches.
A second question which needs to be addressed relates to the ‘mental
bridge’ which allows the conceptualizer to access the desired target. This
question concerns the nature of the relationship between the vehicle and one
or more targets. Metonymy tends to make use of stereotypical, or idealized,
relationships within an ICM. Thus, certain places tend to be associated with
events which typically occur at the place. For example, the expression to go to
bed may, depending on the situation, evoke the metonymic targets ‘to go to
sleep,’ ‘to have sex’ or ‘to be sick.’ All these events are stereotypically
associated with beds, irrespective of the particular context that triggers the
situationally relevant target. More generally, we may describe the conceptual
relationship between space and event as one that is entrenched and may be
exploited by metonymy. The question that needs to be answered here is what
types of conceptual relationships in an ICM may give rise to metonymy.
A third question pertains to the choice of vehicle and target. Thus far, we
have only considered conceptual relationships between two entities either of
which may become the vehicle or the target as shown in the examples of She’s
a pretty face and She’s a pretty person. Unlike metaphorical mappings, which
tend to be unidirectional, metonymic mappings are in principle reversible.
This was already implicitly noticed in traditional approaches by listing both
directions of a metonymic relationship such as CAUSE FOR EFFECT and EFFECT
FOR CAUSE, GENUS FOR SPECIES and SPECIES FOR GENUS, etc. Such theoretically
possible alternatives have to be distinguished from the speaker’s choice of a
particular vehicle as the ‘entry point’ into the ICM. We therefore need to ask if
there are any preferred metonymic construals and, if this is the case, what
‘cognitive principles’ govern the selection of one type of vehicle entity over
another. Such precedence principles were already hinted at by Langacker
(1993: 30). To the extent that there are such preferred routes, these will define
the unmarked, or ‘default,’ cases of metonymy.
Given that there are such default routes, we need to ask, as a fourth
question, if there are any principles that determine the choice of a vehicle other
than by default construal. If this is the case, these metonymic construals yield
marked, or ‘non-default,’ instances of metonymy.
The following four sections of this paper will be devoted to finding answers
to these central questions which, for convenience, are summarized below:
ii(i) What are the ontological realms in which ICMs and metonymic relation-
ships may occur? (Section 2);
i(ii) What are the types of conceptual relationships that may give rise to
metonymy? (Section 3);
(iii) Are there any conceptual entities that can better direct attention to an
intended target than others? If this is the case, are there any cognitive
principles that govern the selection of such ‘default’ cases of metonymy?

(Section 4);
(iv) Are there any principles that override the preferred default routes and
yield ‘non-default’ cases of metonymy? (Section 5).
2. Ontological realms in which metonymy occurs
The following three ontological realms are distinguished for the present
purpose: the world of ‘concepts,’ the world of ‘forms,’ in particular, forms of
language, and the world of ‘things’ and ‘events.’ These realms roughly
correspond to the three entities that comprise the well-known semiotic triangle
as developed by Ogden and Richards (1923: 11): thought, symbol and refer-
ent. The interrelations between entities of the same or from different ontologi-
cal realms lead to various ICMs and possibilities for metonymy.
An important distinction has to be made between ICMs which interrelate
entities of different ontological realms within the same semiotic unit and ICMs
which interrelate entities of different semiotic units within the same ontological
realm or realms. The former situation of interrelated ontological realms gives
rise to two ICMs: the pairing of a concept and a form establishes a sign and may
be described as ‘Sign ICM’; the pairing of a thing or event and a sign, form or
concept establishes a referential situation and may be described as ‘Reference
ICM.’ In as far as these ICMs lead to metonymy, the metonymies will be
described as ‘sign metonymy’ and ‘reference metonymy,’ respectively. The
latter situation of interrelated semiotic units involves concepts, typically in
conjunction with forms. These ICMs will be referred to as ‘Concept ICMs,’ and
a metonymy based on a Concept ICM will be described as ‘concept metonymy.’
Figure 1 illustrates the semiotic relationships which lead to the sign metonymy
(1) and three types of reference metonymies (2)–(4) on the one hand and one
type of concept metonymy (5) on the other hand. The arrows indicate the
direction of the metonymic mapping which will be discussed below.
Concept ConceptA ConceptB
(3) (2) (1) (5)
Thing/Event Form FormA FormB

(4)
Figure 1. Sign, reference and concept metonymies
2.1. Sign ICMs and sign metonymies
The Sign ICM unites a form and one or more concepts. Thus, the word form
dollar or the dollar sign $ are intimately linked with the ‘currency denomina-
tion of dollar,’ ‘currency,’ or ‘money’ in general. As a rule, the form met-
onymically stands for the concept it denotes.
(1) FORM FOR CONCEPT: dollar for ‘money’
The very nature of language is based on this metonymic principle, which
Lakoff and Turner (1989: 108) describe as WORDS STAND FOR THE CONCEPTS
THEY EXPRESS. Since we have no other means of expressing and communicat-
ing our concepts than by using forms, language as well as other communica-
tion systems are of necessity metonymic. It is also for that reason that we fail
to notice the metonymic character of language. This general FORM FOR CON-
CEPT metonymy has to be distinguished from a FORM FOR CONCEPT metonymy
involving specific signs (see (47)).
The Sign ICM only seems to lead to the FORM FOR CONCEPT metonymy,
but not to the reverse metonymy of CONCEPT FOR FORM. A metonymic situation
of this kind is difficult to imagine. It may be said to occur in the ‘tip of the
tongue’ experience, in which we have a certain concept but cannot think of the
corresponding word-form, or in the context of foreign-language learning,
where learners must find a form in the foreign language for a concept with
which they are familiar.
2.2. Reference ICMs and reference metonymies
Real-world entities may be related to signs, concepts or forms. We thus have

three types of Reference ICMs and possible metonymies as shown in Figure
1a. In all three types of reference metonymies, the metonymic target is the
real-world thing or event.
The standard situation of reference involves signs, i.e., form-concept
units, which stand for a thing or event referred to. We thus have the me-
tonymy:
(2) FORM-CONCEPT FOR THING/EVENT: word cow for a real cow
Strictly speaking, the sign does not refer to the world of reality but to our
mental model of reality. Frawley (1992: 20ff) illustrates the mismatch be-
tween these two worlds by comparing the extensional and intensional mean-
ings of the verb to punch. In the extensional world of reality, an event of
punching involves a series of subevents: folding one’s fist, moving one’s arm,
bringing it into contact with an object, and recoiling it. A punching event thus
has duration. Intensionally, however, to punch is a punctual verb and, as such,
cannot be used to describe a durational event as in ??It took five minutes to
punch him. The projected world we refer to by means of a sign thus only partly
matches reality.8 We do, however, firmly believe that words refer to the
extensional world so that metonymy (2) has psychological validity.
Signs can only be said to stand for things or events they refer to — the
reverse metonymic situation, THING/EVENT FOR FORM-CONCEPT, is hard to
imagine. A thing or event may evoke the word denoting it, but it certainly
cannot be argued that when we see a cow and the word cow comes to our
mind, the thing ‘cow’ metonymically stands for the sign cow.
The above folk theory of reference has two variants in which either the
concept or the form of a sign is focused upon. Lakoff (1987: 168f) describes
the former situation as “reference via meaning,” and the latter as “doctrine of
direct reference.”
According to the Reference-via-Meaning ICM, “words have inherent
meanings (called intensions) and designate objects by virtue of those mean-
ings” (Lakoff 1987: 168f). In this view, the meaning associated with the word
cow is assumed to stand for any cow in the world of reality — in contrast to the
set-theoretic account, in which ‘cow’ denotes the set/class of cows.
(3) CONCEPT FOR THING/EVENT: concept ‘cow’ for a real cow
The reversal of this metonymic relationship, THING/EVENT FOR CONCEPT, may
occur in special situations. One such situation would apply to people who are
associated with certain outstanding properties. For example, the former En-
glish soccer player Bobby Charlton may stand for the concept of sportsman-
ship. Another situation might be that of a game of chess in which a missing
piece is replaced by an object such as an eraser or a button. At first sight, this
situation might be seen as involving things in reality only, i.e., THINGA FOR
THINGB. This is, however, not the case. In assigning an eraser the function of,
say, a white rook, we have the thing stand for the concept of a rook.
The Direct-Reference ICM and the type of referential metonymy in-
volved most clearly apply to the use of proper names for persons of that name.
The name John Smith directly refers to the bearer of this name. In our folk
theory of language, the Direct-Reference ICM also has a much wider met-
onymic application. Stephen Tylor (1978:168) points out that in our common-
sense view of language words are names of things, not names of classes:
“There are words and the things we speak of. The word ‘cow’ is the name of
the object ‘cow.’ The word ‘cow’ stands for the object ‘cow’ as a substitute for
it, and we know the meaning of the word ‘cow’ because we have a prior
knowledge of the object it stands for.”
(4) FORM FOR THING/EVENT : word-form cow for a real cow
Also Ogden and Richards (1923) draw attention to people’s tendency to see an
inherent connection between words and referents: Words are seen as part of
the thing and may even provide power over reality in superstition, magic and
spells.
The reversal of this metonymic relationship is hardly possible. A situa-
tion that comes closest to the THING/EVENT FOR FORM metonymy might be that
of language learning, where a child or student points to an object and wants to
be given its name.
2.3. Concept ICMs and concept metonymies
Concept metonymies involve a shift from ConceptA to ConceptB which may,

but need not, be accompanied by a shift in form. The two concepts form part of
the same ICM and are related to each other in some specific way. The
following four types of concept metonymies, (5)–(8), may be distinguished.
(5) FORMA-CONCEPTA FOR FORMB-CONCEPTB: bus-‘bus’ for bus drivers-
‘bus drivers’
The type of metonymy in (5) is the one commonly associated with metonymy.
It involves the relationship between two form-concept pairings as shown in
Figure 1b. In Lakoff and Johnson’s (1980: 38) example The buses are on strike,
the conventional pairing of a word-form and its associated concept, buses, is
used in place of another conventional form-concept unit, bus drivers. Since the
word bus does not have as one of its lexicalized senses ‘bus driver,’ the
metonymic shift is due to the pragmatics of the particular situation and our
knowledge that both buses and bus drivers belong to the same ICM. They are
related to each other by the relationship of control: CONTROLLED FOR CONTROL-
LER. The controlled entity, buses, is used to stand for the controlling entity, bus
drivers.
The metonymic relationship may also appear in its reverse form. For
example, the Control ICM may also give rise to the metonymy CONTROLLER
FOR CONTROLLED, as in I am parked over there.
(6) FORM-CONCEPTA FOR CONCEPTB: mother-‘mother’ for ‘housewife-

mother’
The metonymic situation in (6) differs from (5) in that the target concept is not
linked to a name. There may be different reasons for using this metonymy:
The language may lack a word for the particular concept, the speaker may not
be able to find a conventional name for the concept, or the speaker may not be
aware of the different concepts. This is most likely to be the case in everyday
communication. As already mentioned in Section 1.1., the word mother makes
us think of housewife mothers, i.e., the Mother ICM evokes the stereotypical
member of the category of ‘mothers.’ The metonymic relationship here is,
therefore, CATEGORY FOR A MEMBER OF THE CATEGORY. It is not clear at the
moment if the form of a category is required in order to perform the met-
onymic shift or if this metonymy can also operate at the purely conceptual
level, i.e., if, in thinking of a category like ‘mother,’ we are really thinking of
its member ‘housewife mother.’
The reversal of the metonymic relationship (6) is unlikely to occur. It
would apply to a situation in which an unnamed concept such as the stereo-
typical ‘housewife mother’ stands for the named concept, i.e., the word
mother. However, the specific metonymic relationship is reversible, i.e., we
have a type of metonymy in which A MEMBER OF A CATEGORY stands for THE
CATEGORY (see (16b)).
(7) FORMA-CONCEPTA FOR FORMA-CONCEPTB: White House-‘place’ for

White House-‘institution’
The metonymic situation in (7) applies to polysemy, in which two senses of a
word-form are relatable within the same ICM. Polysemy is a common way in
which metonymic concepts manifest themselves in language (see Lakoff 1987
and Taylor 1995). The metonymic concepts of a polysemous expression are
lexicalized, i.e., they are, in contrast to the metonymic situation of (5),
independent of a pragmatic situation. Thus, the expression White House is
lexically polysemous between the senses of the ‘building’ and the ‘executive
branch of the US government.’ The concepts of place and institution located at
that place are so closely interconnected in this ICM that the former is routinely
used to stand for the latter. The metonymy PLACE FOR INSTITUTION thus
accounts for our understanding of The White House did not intervene in the
sense of ‘the American government did not intervene.’
Metonymic polysemies seem to take a preferred direction, but they are in
principle reversible. Thus, an institution may also be used to stand for a place.
For example, Let’s have an oyster dish at Central Station illustrates the
metonymy INSTITUTION FOR PLACE.
(8) FORMA-CONCEPTA FOR FORMB-CONCEPTA: UN for United Nations
In the metonymic situation in (8) the form of an expression changes while the
concept roughly remains the same, which might also be indicated by a
notation such as CONCEPTA/B. This metonymy applies to reductions of form as
in the abbreviation UN, acronyms such as NATO, clippings such as exam for
examination, modifications of form as in the euphemism What the heck are
you doing? for What the hell are you doing? and substitutions by pro-forms
and equivalence in translation, in which the language changes but the content
is intended to be preserved. In all these cases, the form of language changes
but, ideally, not its conceptual content.
The reversal of this metonymic relationship does not occur freely. In the
case of abbreviations, this would imply that we understand a full expression
such as United Nations as standing for its abbreviated form, UN. In the case of
translation equivalence, however, either of the languages involved may of
course serve as the vehicle or target.
2.4. Summary
We can now summarize the observations made about the interaction of

ontological realms and metonymy.
First, on the basis of the three ontological realms, we distinguished three
basic types of ICMs. The Sign ICM and the Reference ICM interrelate entities
of different ontological realms, the Concept ICM interrelates entities of differ-
ent semiotic units. All of these ICMs give rise to metonymy, and there is, a
priori, no reason why the notion of metonymy should be restricted to one of
the types of concept metonymy only.
Second, sign and reference metonymies demonstrate a preferred, or even
exclusive, direction. This also holds true for concept metonymies (6), (7) and
(8), which also display a preferred direction. Concept metonymy (5), in which
one sign stands for another sign, is the only type of metonymy that is in
principle reversible; it is also the only type of metonymy that is characterized
by a variety of conceptual relationships between vehicle and target — the
other types of metonymic relationships carry very little additional content
apart from interrelating vehicle and target. Both these factors have certainly
contributed to focusing on this type of metonymy.
Third, the direction of metonymization allows us to recognize a hierarchy
for the preferred choice of an ontological vehicle. With sign metonymies (1),
the form serves to access a concept; with reference metonymies, the form-
concept unit (2), the concept (3) or the form (4) serves to access a thing or
event in reality; and with concept metonymy (6), the form-concept unit
accesses a concept. Three of the ontological metonymies, (5), (7) and (8), do
not display a preferred directionality because they operate at the same onto-
logical level. We therefore find the following hierarchy for the choice of a
metonymic vehicle:
form > form-concept > concept > reality
The cognitive motivation for these directionalities will be investigated in
Section 4.
Since this paper aims at providing an overall cognitive framework of
metonymy, which finds its richest variability in metonymies involving lan-
guage, its focus will mainly be on metonymies which most visibly, but not
exclusively, manifest themselves in the shape of linguistic expressions.
3. Types of metonymy-producing relationships
Conceptual relationships within an ICM which may give rise to metonymy will
be called ‘metonymy-producing relationships.’ Thus, the conceptual relation-
ship that holds between a container and the thing(s) contained may produce the
metonymies CONTAINER FOR CONTENTS and CONTENTS FOR CONTAINER. How-
ever, not all relationships within ICMs can produce metonymies. One example
of this is the ICM of the human face. A face has several parts that are closely
related in space, yet do not lead to metonymy. For example, the nose cannot
stand for the mouth metonymically and vice versa, i.e., I hit him in the nose will
not be understood to mean ‘I hit him in the mouth.’ On the other hand, we
understand the utterance Can you answer the door? in a metonymic sense, in in
which the door is meant to refer to a person standing behind the door. Both
examples involve a relationship of spatial proximity, but only the conceptual
distinctness between objects and people in the latter example leads to me-
tonymy. Metonymy may only arise when the intended target is uniquely
accessible and “the addressee’s attention is directed to the intended target”
(Langacker 1993: 30). The greater the conceptual contrast between vehicle and
target, the better is a relationship suited to be exploited metonymically. Thus,
a whole ICM is generally conceptually distinct enough from its parts to allow
metonymic processes to operate freely from whole to part or part to whole.
The distinction between whole and part is in fact of paramount impor-
tance for metonymic processes. Given that our knowledge about the world is
organized by structured ICMs which we perceive as wholes with parts, we
suggest that the types of metonymy-producing relationships may be subsumed
under two general conceptual configurations:
i(i) Whole ICM and its part(s)
(ii) Parts of an ICM
Configuration (i) may lead to metonymies in which we access a part of an ICM
via its whole or a whole ICM via one of its parts; configuration (ii) may lead to
metonymies in which we access a part via another part of an ICM. This, of
course, implies that the whole ICM is still present in the background.
The following typology of metonymy-producing relationships and me-
tonymies is not meant to be exhaustive. It includes those types that are most
frequently listed in classifications of metonymies and seem to reflect the most
entrenched metonymic routes.
3.1. Whole ICM and its part(s)
The relationship between a whole and a part typically applies to things and
their parts, where the notion of ‘thing’ is to be understood here in the
schematic sense of Langacker (1991). The prominent status which whole-part
relationships have in our cognition has been demonstrated by Tversky and
Hemenway (1984), who found that, in tasks involving attributes of basic-level
items, subjects predominantly listed parts as attributes. This may have to do
with the conceptually autonomous status which we attach to things. Things, in
particular physical objects, are typically conceived of as forming a gestalt with
well-delineated boundaries and as internally composed of various parts.
‘Whole-part configurations’ are also assumed to underlie the Scale ICM, the
Constitution ICM, the Event ICM, the Category-and-Member ICM, the Cat-
egory-and-Property ICM, and the Reduction ICM.
(i) Thing-and-Part ICM. This ICM may lead to the two metonymic
variants:
(9) a. WHOLE THING FOR A PART OF THE THING:
America for ‘United States’
b. PART OF A THING FOR THE WHOLE THING:
England for ‘Great Britain’
People often speak of America but mean one of its geographical parts, the
United States; conversely, people, especially foreigners, often speak of
England but mean Great Britain, including Wales and Scotland.
The WHOLE FOR PART metonymy is widely found in situations which
Langacker (1993: 31) describes as “active-zone/profile discrepancies,” where
an entity’s active zone is defined as comprising “those portions of the entity
that participate most directly and crucially in that relationship.” For example,
in He hit me or The car needs washing, the profiled whole things he and the
car may be said to stand for the active-zone parts ‘his fist’ and ‘the car’s
body,’ respectively. Even clearly separate things may be conceptualized as
active-zone parts of an over-arching whole ICM. Thus, it does not strike us as
unnatural to speak of lighting the Christmas tree for ‘lighting the candles on
the Christmas tree.’ The use of a whole entity for an active-zone part comes so
natural that it usually does not appear to us, including most scholars of
metonymy, as metonymic.9
The PART FOR WHOLE metonymy has traditionally been given special
attention and classified as a type of its own under the name of ‘synecdoche.’10
Synecdoches are less ubiquitous than WHOLE FOR PART metonymies and,
hence, more likely to be noticed. This applies to deliberate metonymic usages
such as Those are cool wheels you have there as well as to the widespread use
of body parts such as hand, face, head or leg for a person.11 In such situations,
the entity that is understood to be most crucially involved in the ICM is
metonymically highlighted. Especially more abstract things such as ‘hearing,’
‘intellect’ or ‘control’ tend to be metonymically expressed by one of their
concrete parts: Thus, ear may stand for ‘hearing,’ brain for ‘intellect’ and
hand for ‘control’ as in Things got out of hand for ‘things got out of control.’
(ii) Scale ICM. Scales are a special class of things and the scalar units are
parts of them. Typically, a scale as a whole is used to stand for its upper end
and the upper end of a scale is used to stand for the scale as a whole:
(10) a. WHOLE SCALE FOR UPPER END OF THE SCALE:
Henry is speeding again for ‘Henry is going too fast’
b. UPPER END OF A SCALE FOR WHOLE SCALE:
How old are you? for ‘what is your age?’
The expression speed defines the whole scale of velocity but we locate the
velocity in (10a) at, or even beyond, the upper end of the scale. Conversely,
mention of the positive end of the scale in (10b) evokes the whole scale, and it
is only for the purpose of achieving special effects that the negative end of a
scale may be used as in How young are you?
(iii) Constitution ICM. This ICM involves matter, material or substances
which are seen as constituting a thing. Properly speaking, such ‘substance-
things’ do not have parts but are constituted by their very substance. Sub-
stances are, amongst other things, characterized as being unbounded and
uncountable.12 A substance or material may be metonymically conceived of
as an object and is then construed as bounded and coded as a count noun as in
(11a). Conversely, an object may be conceived of as substance-like and is then
construed as unbounded and coded as a mass noun as in (11b).
(11) a. OBJECT FOR MATERIAL CONSTITUTING THE OBJECT:
I smell skunk.
b. MATERIAL CONSTITUTING AN OBJECT FOR THE OBJECT:
wood for ‘forest’
(iv) Event ICM. Events may be metaphorically viewed as things which may
have parts. Thus, one can speak of the theoretical and the practical ‘parts’ of a
driving test. As with things, an event as a whole may stand for one of its
subevents and a subevent may stand for the whole event.
(12) a. WHOLE EVENT FOR SUBEVENT: Bill smoked marijuana.
b. SUBEVENT FOR WHOLE EVENT: Mary speaks Spanish.
The event in (12a) involves as some of its subevents lighting a marijuana

cigarette, taking it to one’s lips, inhaling the smoke, etc. The inhaling part is
probably felt to be the central and most important subevent and the one that is
metonymically meant. This is exactly the reason why Clinton needed to
exclude that part when he argued that, as a young man, he smoked marijuana
but did not inhale.
The habitual event in (12b) is understood to refer not only to Mary’s

spoken command of a language, but also to include the skills of comprehen-
sion, reading and writing. Many subevents serve as conventionalized expres-
sions for an Event ICM as a whole. This applies in particular to successive
subevents which metonymically highlight one of these temporal phases (see
Lakoff 1987: 78f).13 In They went to the altar, an initial subevent stands for
the whole Wedding ICM, and in Our teacher had 100 essays to grade, a final
subevent stands for a larger Event ICM involving reading, correcting and
eventually grading students’ papers.
Metonymies also operate in the Auxiliary system. Sentence (12b) illus-
trates the way metonymy affects the tense system. Habitual events occur in
past, present and future time, but are described by use of the Present Tense. If
we assume that the Present Tense ideally locates events in present time, its use
for habitual events is metonymic. Another time metonymy is found in the use
of the Present Tense for future events as in I am off for ‘I will be off,’ in French
J’arrive for ‘I am coming,’ or in the robber’s threat The money or you’re a
dead man, where the present moment figures prominently for the future event.
We thus have the following PART FOR WHOLE time metonymies:
(13) PRESENT FOR HABITUAL: Mary speaks Spanish.
(14) PRESENT FOR FUTURE: I am off for ‘I will be off’
Metonymies may also affect an event’s grounding in reality or potentiality.
Thus, we use simple tenses in describing both a real event such as The teacher
is angry with me and a potential event as in He is an angry person. The latter
situation with angry as an attributive adjective does not describe a person’s fit
of anger at the present moment but his or her disposition to get angry
potentially.14 This metonymic relationship also occurs in its reverse form in
which a potential event is described as real. As Gibbs (1994), Thornburg and
Panther (1997), and Panther and Thornburg (this volume) have shown, certain
conditions of a speech act may be highlighted to stand for the intended speech
act as a whole. For example, in using can in Can you pass the salt? the speaker
highlights the precondition of ability for a directive speech act. Since such
speech acts with can convey the notion of potentiality, Panther and Thornburg
describe this metonymy as POTENTIALITY FOR ACTUALITY. These two reverse
metonymies may be stated as follows:
(15) a. ACTUAL FOR POTENTIAL:
He is an angry person for ‘he can be angry’
b. POTENTIAL FOR ACTUAL :

I can see your point for ‘I see your point’
(v) Category-and-Member ICM. A category and its members stand in a kind-
of relation. As shown by Seto (this volume), kind-of relations, on which
‘synecdoche’ is based, need to be distinguished from part-of relations, on
which metonymy is based. The relations of taxonomy and partonomy, how-
ever, tend to be confused. This is reflected in the German term for ‘subset,’
Teilmenge, literally ‘part-set.’ Taxonomic hierarchies may also be metaphori-
cally seen as part-whole structures in which “[e]ach higher-order category is a
whole, with the immediately lower categories being its parts” (Lakoff 1987:
287). We, therefore, feel justified in analyzing Category-and-Member ICMs
as instances of the whole-part configuration.
(16) a. CATEGORY FOR A MEMBER OF THE CATEGORY:
the pill for ‘birth control pill’
b. MEMBER OF A CATEGORY FOR THE CATEGORY:
aspirin for ‘any pain-relieving tablet’
A special type of this metonymic relationship is that between a generic type
and a specific token:
(17) a. GENERIC FOR SPECIFIC: Boys don’t cry.
b. SPECIFIC FOR GENERIC: The/A spider has eight legs.
The sentence in (17a) is a generic statement about boys, but it might be used in
the specific situation of a boy’s crying and, hence, is also understood specifi-
cally. Conversely, specific tokens may be used to stand for generic types. In
the situation of ‘generic reference’ expressed in (17b), the definite article the
or the indefinite article a is used to refer to spiders in general. As pointed out
by Norrick (1981: 35), “any specific instantiation of a class calls forth the
whole class.” A single violin may stand for the class of violins and a musical
note may stand for the musical key system as such. In language, the clearest
cases of the SPECIFIC FOR GENERIC metonymy are found in the use of proper
names as common expressions as in Gibbs’ (this volume) example He’s going
to OJ his way out of marriage. At a more general level, this metonymic
relationship also underlies our interpretation of proverbs. As shown by Lakoff
and Turner (1989: Ch. 4), proverbs such as Blind blames the ditch describe a
particular situation but convey a general understanding, which again is ap-
plied to a particular situation at hand. Lakoff and Turner analyze proverbs as
instances of the metaphor GENERIC IS SPECIFIC. Since both the specific and the
generic levels belong to the same ICM, however, we prefer to analyze them as
instances of the metonymy SPECIFIC FOR GENERIC.
Subtypes of this metonymy are AN INDIVIDUAL (AS A TYPICAL MEMBER OF
A CATEGORY) FOR A CATEGORY as in every Tom, Dick and Harry and SPECIFIC
CASE FOR GENERAL RULE, which “holds between laws and their concrete
instantiations generally” (Norrick 1981: 37). This metonymic relationship is
not, as claimed by Norrick, irreversible. One can easily think of situations in
which a general rule is invoked in describing a specific situation as in I always
leave my umbrella at home when it rains.
(vi) Category-and-Property ICM. Properties may either be seen meta-
phorically as possessed objects (PROPERTIES ARE POSSESSIONS) or metonymi-
cally as parts of an object. If categories are intensionally defined by a set of
properties, these properties are necessarily part of the category. Categories
typically evoke, and metonymically stand for, one of their defining or other-
wise essential properties and, conversely, a defining or essential property may
evoke, or stand for, the category it defines (cf. Norrick 1981).
(18) a. CATEGORY FOR DEFINING PROPERTY: jerk for ‘stupidity’
b. DEFINING PROPERTY FOR CATEGORY: blacks for ‘black people’
Most of the members of a category which Lakoff (1987: Ch. 5) discusses in
connection with metonymic models are much rather characterizable as salient
properties. For example, an ideal husband is characterized by the property of
faithfulness as in Wilbert doesn’t go to topless bars; he is my husband, while a
stereotypical husband is, amongst other things, characterized by the property
of dullness as in Wilbert will stay at home; he is my husband. Some categories
conventionally stand for specific properties such as Judas for ‘treacherous’ or
Cadillac for ‘the best of.’ Also, in referring to an upcoming star in linguistics
as a second Chomsky, we have in mind his or her intellectual brilliance.
Stereotypical properties are also evoked in our interpretation of ‘collo-
quial tautologies’ such as Boys will be boys. Since a tautology is literally
uninformative, it can only be meaningfully interpreted in the sense of a
stereotypical property associated with the category. The tautology in Boys will
be boys may, depending on the context, mean ‘boys are unruly’ or ‘boys are
cute and adorable’ (Gibbs 1994: 345–351). All these examples are instances
of a WHOLE FOR PART metonymy, which may be characterized as in (19a);
conversely, salient properties as parts of a category may stand for the category
as a whole as in (19b):
(19) a. CATEGORY FOR SALIENT PROPERTY:

Boys will be boys for ‘unruly’
b. SALIENT PROPERTY FOR CATEGORY:
How do I find Mr. Right?
(vii) Reduction ICM. A final type of a PART FOR WHOLE metonymy is found in
the reduction of the form of a sign, which was already alluded to under (8)
FORMA-CONCEPTA FOR FORMB-CONCEPTA. Its specific variant may be described
as:
(20) PART OF A FORM FOR THE WHOLE FORM: crude for crude oil
The reduction of forms may involve sophisticated metonymic chains. For
example, the abbreviation tg stands for a longer abbreviated form, tgif, which
represents the whole expression Thank God, it’s Friday; and even this excla-
mation may be seen as a part of the whole target sense: ‘it’s the weekend —
let’s enjoy ourselves.’
As already stated in Section 2.3., this ICM does probably not come in its
reverse form. In our folk model of language, only a complete form seems to
capture the essence of a concept properly. Form ICMs are thus different from
all the other Whole-Part ICMs in that the whole cannot metonymically stand
for one of its parts.
3.2. Parts of an ICM
This configuration relates conceptual entities that function as parts with

respect to a whole ICM. While the relationship between a whole and its parts
typically applies to things, the relationship between parts typically applies to
entities within an event. Events are constituted by relations and participants,
and PART FOR PART metonymies tend to build on the interaction between a
relation and one of the things participating in the relation. Metonymy also
arises when a relation is construed as a thing or a thing is construed as a
relation. In English, such conceptual recategorizations are coded morphologi-
cally: relations which are recategorized as things show up as derived nominals
(production), and things which are recategorized as relations show up as
derived verbs (to beautify) or, without a derivational morpheme, as conver-
sions (to tiptoe) (see Dirven, this volume).
The Part-Part configuration applies to various parts of ICMs and includes
the Action ICM, the Perception ICM, the Causation ICM, the Production
ICM, the Control ICM, the Possession ICM, the Containment ICM, the
Location ICM, the Sign and Reference ICMs and the Modification ICM.
(i) Action ICM. Action ICMs involve a variety of Participants which may
be related to the predicate expressing the action or to each other. There are,
thus, specific relationships such as those between an INSTRUMENT and the
ACTION, the RESULT of an action and the ACTION , etc., all of which are parts of
the Action ICM. These relationships may in turn be instantiated as specific
types of metonymy. The Action ICM includes the following types of met-
onymic relationships, the first four of which are reversible:
(21) a. AGENT FOR ACTION: to author a new book; to butcher
the cow
b. ACTION FOR AGENT: writer; driver
(22) a. INSTRUMENT FOR ACTION: to ski; to hammer
b. ACTION FOR INSTRUMENT: pencil sharpener; screwdriver
(23) a. OBJECT FOR ACTION: to blanket the bed; to dust the room
b. ACTION FOR OBJECT: the best bites; the flight is waiting
to depart
(24) a. RESULT FOR ACTION: to landscape the garden
b. ACTION FOR RESULT: the production; the product
(25) MANNER FOR ACTION : to tiptoe into the room
(26) MEANS FOR ACTION: He sneezed the tissue off the
table.15
(27) TIME FOR ACTION: to summer in Paris
(28) DESTINATION FOR MOTION: to porch the newspaper
(29) INSTRUMENT FOR AGENT: the pen for ‘writer’
With the exception of (29), INSTRUMENT FOR AGENT, all the Action metonymies
listed above involve predicates either as the vehicle or the target and typically
also involve changes of their word class: Participants are converted into verbs
and predicates are nominalized. Noun-verb conversion and nominalization
can therefore be seen as two complementary morphological processes leading
to the two types of reversible metonymies. What makes these morphological
derivations special types of metonymy, however, is their conflation of vehicle
and target: the original word class describes the metonymic vehicle and the
morphologically recategorized form expresses the target. The conflation pro-
cess echoes the incorporation of elements in the verb described by Talmy
(1985). The situations analyzed by him, however, are not metonymic. The
verb to swim as in to swim into the cave semantically incorporates the

elements of ‘motion’ and ‘manner,’ as does to tiptoe in to tiptoe into the room.
The ‘manner’ element is, however, only present in the semantics of to swim,
but it is visibly present in the morphological base of to tiptoe. As is well-
known, the processes of conversion and nominalization may result in mean-
ings different from their bases. Thus, an act of hammering need not be
performed with a hammer nor does a bite need to be bitten. The metonymic
relationship operates at a more schematic level. But to the extent that the base
form of the derivation is still recognizable, such morphological derivations
may reasonably be treated as instances of metonymy.
The metonymic relationships listed in (21)–(28) are typically achieved by
word-class changes, but are not restricted to these morphological processes.
For example, the RESULT FOR ACTION metonymy may also arise within the
same word class. Thus, the verb to win in Win a fortune! describes a result but
is meant to metonymically invoke the action of gambling.
(ii) Perception ICM. Perception plays such an outstanding role in our
cognitive world that it merits an ICM of its own. Since perceptions may also
be intentional, the Perception ICM may cross-classify with the Action ICM.
This applies to the metonymies INSTRUMENT/ORGAN OF PERCEPTION FOR THE
PERCEPTION as in to eye someone and MANNER OF PERCEPTION FOR THE PERCEP-
TION as in She squinted through the mailbox. Non-intentional perceptions may
produce the following reversible metonymies:
(30) a. THING PERCEIVED FOR PERCEPTION:
There goes my knee for ‘there goes the pain in my knee’
(Lakoff 1987: 511)
b. PERCEPTION FOR THING PERCEIVED: sight for ‘thing seen’
(iii) Causation ICM. Cause and effect are so closely interdependent that one of
them tends to imply the other. Moreover, their complementarity probably
accounts for the fact that people often confuse causes and effects. In principle,
the causation ICM may give rise to reversible metonymies:
(31) a. CAUSE FOR EFFECT:
healthy complexion for ‘the good state of health bringing
about the effect of healthy complexion’
b. EFFECT FOR CAUSE:
slow road for ‘slow traffic resulting from the poor state of
the road’
Effects more readily serve as a metonymic vehicle than causes, which is

evidenced most clearly in the following subtypes of EFFECT FOR CAUSE me-
tonymies:
(32) STATE/EVENT FOR THING/PERSON/STATE CAUSING IT:
She was my ruin.
(33) EMOTION FOR CAUSE OF EMOTION:
She is my joy ‘she makes me be happy’
(34) MENTAL/PHYSICAL STATE FOR OBJECT/PERSON CAUSING IT:
You are a pain in the neck ‘you give me pain’
(35) PHYSICAL/BEHAVIORAL EFFECT FOR EMOTION CAUSING IT:
She was upset ‘something made her upset’
A causal metonymy may also be seen in situations in which an action or a
motion brings about, or is accompanied by, typical sounds which together
establish an ICM.
(36) SOUND FOR EVENT CAUSING IT: The car screeched to a halt.
Here, the screeching noise results when the car brakes are applied. Similar
metonymic situations are illustrated in The train whistled into the station; The
fire trucks wailed out of the firehouse and She rang the money into the till.
Causal metonymies also permeate the field of perception. A percept may
stand for its cause (37a) and a cause may stand for the percept (37b):
(37) a. SEEING SOMETHING DONE FOR MAKING SURE THAT IT IS DONE
See that he gets his money. (Lakoff 1987: 437)
b. ACT OF FORMING A PERCEPT FOR PERCEPT
to take a look (Norvig and Lakoff 1987: 204)
(iv) Production ICM. Production ICMs involve actions in which one of the
Participants is a product created by the action. The production of objects
seems to be a particularly salient type of causal action. The Production ICM
leads to various types of metonymic relationships in which the thing produced
tends to be the intended target:
(38) PRODUCER FOR PRODUCT: I’ve got a Ford for ‘car’
Due to our close association of artists with their artistic productions and
inventors with their inventions, the metonymies ARTIST FOR HIS WORK as in
They are playing Mozart tonight and INVENTOR FOR THE THING INVENTED as in
macadam establish particularly common subtypes of the PRODUCER FOR PRO-

DUCT metonymy. A producer and the thing produced are conceptually differ-
ent enough to warrant clear identification of their roles. This also applies to an
instrument used for producing something or the place of production:
(39) a. INSTRUMENT FOR PRODUCT:
Did you hear the whistle? for ‘its sound’
b. PRODUCT FOR INSTRUMENT:
to turn up the heat for ‘the radiator’
(40) PLACE FOR PRODUCT MADE THERE:
china, mocha, camembert
(v) Control ICM. This ICM includes a controller and a person or object
controlled. It gives rise to reversible metonymic relationships:
(41) a. CONTROLLER FOR CONTROLLED: Schwartzkopf defeated Iraq.
b. CONTROLLED FOR CONTROLLER: The Mercedes has arrived.
Control ICMs seem to be naturally expressed by using the CONTROLLER FOR
CONTROLLED metonymy as in (41a), in which Schwartzkopf stands for the US
Army that did the fighting. Making the same statement using the CONTROLLED
instead of the CONTROLLER, as in The US Army defeated Iraq, does not evoke
the controller reading. The CONTROLLED FOR CONTROLLER metonymy seems to
apply only to situations in which the thing controlled is particularly salient or
the controller is unknown as in (41b).
The notion of control blends into that of possession. For example, the
user of an object is at the same time in control of the object used and possesses
it. This situation gives rise to the metonymy OBJECT FOR USER OF THE OBJECT,
as in Lakoff and Johnson’s (1980) example Mrs. Grundy frowns on blue jeans,
where the expression blue jeans stands for wearers of blue jeans.
(vi) Possession ICM. The Possession ICM may lead to reversible metony-
mies:
(42) a. POSSESSOR FOR POSSESSED: That’s me for ‘my bus’
b. POSSESSED FOR POSSESSOR:
He married money for ‘person with money’
There is a clear preference for choosing the possessor as the vehicle and an
alienably possessed object as the target. The POSSESSOR FOR POSSESSED me-
tonymy is well-entrenched and hardly noticeable. Compare expressions such
as I am parked over there (for ‘my car’); You have a flat tire (for ‘your car’) or
Bill is in the Guinness Book of Records (for ‘his name’). This is also reflected
in the use of anaphoric pronouns, which, if they can be used at all, refer to the
human vehicle as in Bill is in the Guinness Book of Records; he is on page 7
and not to the target as in #Bill is in the Guinness Book of Records; it is on page
7. Conversely, anaphoric pronouns in POSSESSED FOR POSSESSOR metonymies
refer to the human target as in Many big names have turned up and he was one
of them and not to the vehicle as in #Many big names have turned up and it was
one of them. POSSESSED FOR POSSESSOR metonymies apply to both alienable
possession (The ham sandwich had a side dish of salad) and inalienable
possession (appendicitis for ‘patient’), and in both cases they are somehow
felt to be marked.
(vii) Containment ICM. The image-schematic situation of containment is
so basic and well-entrenched that it deserves to be treated as an ICM of its own
among locational relations. As a rule, we are more interested in the contents of
a container than in the mere container so that we commonly find metonymies
which target the contents via the container as in (43a) rather than the reverse
metonymic relationship as in (43b).
(43) a. CONTAINER FOR CONTENTS: The bottle is sour for ‘milk’
b. CONTENTS FOR CONTAINER:
The milk tipped over for ‘the milk container tipped over’
(Norrick 1981: 58)
(viii) Location ICMs. Places are often associated with people living there,16
well-known institutions located there, events which occur or occurred there
and goods produced or shipped from there (see (40)). Hence, we find the
following metonymies:
(44) a. PLACE FOR INHABITANTS:
The whole town showed up for ‘the people’
b. INHABITANTS FOR PLACE:
The French hosted the World Cup Soccer Games for ‘France’
(45) a. PLACE FOR INSTITUTION:
Cambridge won’t publish the book for ‘Cambridge Univer-
sity Press’
b. INSTITUTION FOR PLACE:
I live close to the University.
(46) a. PLACE FOR EVENT:

Waterloo for ‘battle fought at Waterloo’
b. EVENT FOR PLACE:
Battle, name of the village in East Sussex where the Battle
of Hastings was fought
The relationship between places and people living there may also be, and often
has been, seen as a situation of containment, which is supported by the use of
the container preposition in in expressions such as live in New York. In this
view, the metonymic relationship in (44) would be treated as a metaphorical
extension of the container metonymy (43). The metonymic relationship in
(46) comprises important events which occurred at a particular place as well as
activities typically performed at a given place as its setting. Relating places
with what is typically done there is part of our cultural knowledge. It allows us
to interpret the mention of the place in I was behind the wheel all day in the
sense of the activity typically performed at that place, namely ‘I was driving
all day.’ This subtype of metonymy may more adequately be described as
PLACE FOR ACTIVITY PERFORMED AT THAT PLACE.
(ix) Sign and Reference ICMs. As shown in Section 2, Sign and Refer-
ence ICMs lead to metonymies cross-cutting ontological realms. In sign
metonymy, a (word-)form stands for a conventionally associated concept; in
reference metonymies, a sign, concept or (word-)form stands for the real
thing. In each case, one part of an ICM stands for another part of the same
ICM.
Sign metonymy may also apply to particular instances of the relationship
between the form and content parts of a sign and, just like the general FORM
FOR CONCEPT metonymy, this specific metonymy is irreversible:
(47) WORDS FOR THE CONCEPTS THEY EXPRESS:

a self-contradictory utterance
In (47), we understand the word form utterance “as referring to the conceptual
content expressed by the utterance” (Lakoff and Turner 1989: 108). This
metonymy also accounts for the compound expression four-letter word, which,
however, involves additional metonymic steps. The formal property of ‘four
letters’ stands for the category of ‘swear words’ and these types of words stand
for the concept expressed by them. Since the expression four-letter word may
also be used for swear words which have more or less than four letters such as
asshole or bastard, metonymy (16b), MEMBER OF A CATEGORY FOR THE CAT-
EGORY, also applies here. These members are, however, less central to the
category of swear words than genuine ‘four-letter words’ such as shit and fuck.
(x) Modification ICM. This ICM mainly applies to variant forms of a sign
apart from reduction. Reduced forms were accounted for by the PART FOR
WHOLE metonymy (20); variant forms stand in a Part-and-Part relationship of
an ICM. More specifically, we may distinguish between genuine cases of
modification as in (48) and substitution as in (49), both of which seem to be
unique to language and only operate in one direction:
(48) MODIFIED FORM FOR ORIGINAL FORM: effing for fucking
(49) SUBSTITUTE FORM FOR ORIGINAL FORM:
Do you still love me? — Yes, I do.
3.3. Summary
The discussion of a typology of metonymy-producing relationships in this

section can now be summarized.
First, the ICMs which include metonymy-producing relationships may be
subsumed under two high-level conceptual configurations: (i) whole ICM and
its part(s) and (ii) parts of an ICM. The whole-part configuration typically
gives rise to metonymies involving things, while the part-part configuration
primarily applies to metonymies involving predications in events and states.
This may have to do with the conceptually autonomous status which we attach
to things as opposed to the dependent status which we attach to relations,
which require the presence of things to be related.
Second, the conceptual relationships which hold in a particular ICM may
be exploited metonymically. A number of conceptual relationships only admit
metonymization in one direction. This applies to some relationships of Action
ICMs (e.g., (25) MANNER FOR ACTION), special situations of Causation ICMs
(e.g., (36) SOUND FOR EVENT CAUSING IT) and ICMs involving (word-)forms
(e.g., (48) MODIFIED FORM FOR ORIGINAL FORM). But also relationships which
lead to reversible metonymies such as (31a) CAUSE FOR EFFECT and (31b)
EFFECT FOR CAUSE tend to display a preference for one direction. For example,
we routinely speak of containers when we mean their contents as in He picked
up the glass and poured it into the pitcher, but situations in which we speak of
the contents when we mean the container as in I’ll put the roses in water are
very rare. It is claimed that the metonymy CONTAINER FOR CONTENTS, i.e., the
choice of a container as a reference point, is more natural than the metonymy
CONTENTS FOR CONTAINER, i.e., the choice of the things contained. Similarly,
metonymy (41a), CONTROLLER FOR CONTROLLED, is more natural than me-
tonymy (41b), CONTROLLED FOR CONTROLLER. Thus, Schwartzkopf defeated
Iraq evokes the metonymic reading in which Schwartzkopf stands for the US
Army that did the fighting. The US Army defeated Iraq, however, does not
evoke the controller reading. It is these metonymic preferences which will be
explored in the following section.
4. Principles governing the selection of the preferred vehicle
The choice of vehicle and target in default cases of metonymy appears to be

motivated or restrained by cognitive principles. The nature of such principles
was pointed out by Langacker (1993: 30) in a very relevant observation on the
function of metonymy:
Metonymy allows an efficient reconciliation of two conflicting factors: the
need to be accurate, i.e., of being sure that the addressee’s attention is directed
to the intended target, and our natural inclination to think and talk explicitly
about those entities that have the greatest cognitive salience for us.
The former factor relates to communicative aspects and will be described here
in terms of communicative principles, the latter factor pertains to cognitive
aspects and will be described in terms of cognitive principles.
Section 4.1. will be concerned with cognitive principles of relative sa-
lience. Since metonymic reference points are cognitive in nature, it is only
natural to expect that these principles have a cognitive basis. Since metony-
mies also play a major role in communication, we may also expect communi-
cative principles to contribute in determining the selection of a vehicle
expression. Section 4.2. will briefly examine the issue of communicative
principles of relative salience. The principles themselves are assumed to have
the status of preferential tendencies and will be stated in the form of X OVER Y.
4.1. Cognitive principles
Some initial ideas concerning cognitive principles which determine the rela-
tive salience of an entity were developed by Langacker (1993: 30): “Other
things being equal, various principles of relative salience generally hold:
human > non-human; whole > part; concrete > abstract; visible > non-visible;
etc.” Some of the semantic constraints which Cooper and Ross (1975) found
to determine the fixed ordering of bipartite expressions are also relevant for
default metonymies.17 The cognitive principles which have been identified
here relate to three general determinants of conceptual organization: human
experience, perceptual selectivity, and cultural preference, which, however,
tend to interact and overlap in several ways.
4.1.1. Human Experience

Our basic human experiences are derived from our anthropocentric view of
the world and our interaction in the world. In this world, humans take prece-
dence over non-humans, things are looked at from a subjective rather than
objective point of view, concrete objects are more salient than abstract enti-
ties, things we interact with are selected over things we do not interact with,
and functional things are more important to us than things which are non-
functional.
(i) HUMAN OVER NON-HUMAN. This principle accounts for the default cases
of the production, control and possession metonymies, namely (38) PRODUCER
FOR PRODUCT (I’ve got a Ford), (41a) CONTROLLER FOR CONTROLLED
(Schwartzkopf defeated Iraq) and (42a) POSSESSOR FOR POSSESSED (I have a
flat tire).
(ii) SUBJECTIVE OVER OBJECTIVE. Each person’s experience is necessarily
subjective and leads to a subjective view of the world. The principle SUBJEC-
TIVE OVER OBJECTIVE determines the use of metonymy (30b) PERCEPTION FOR
THING PERCEIVED as in What a beautiful sight for ‘thing seen.’ This principle
may also account for the unidirectionality of the reference metonymies (2)
FORM-CONCEPT FOR THING/EVENT and (3) CONCEPT FOR THING/EVENT. Here, we
access the world of ‘objective reality’ by means of our subjective world of
concepts.
(iii) CONCRETE OVER ABSTRACT. Our basic human experience relates to
concrete physical objects, which have more salience for us than abstract
objects. The cognitive principle CONCRETE OVER ABSTRACT accounts for why
we speak of a book written in a careful hand for ‘carefully written’ or having
one’s hands on something for ‘controlling something.’ Body parts make
particularly ‘good’ objects, and we routinely access various abstract human
domains by reference to our body. Special subcases of the CONCRETE OVER
ABSTRACT principle may be described as BODILY OVER EMOTIONAL (heart for
‘kindness’), BODILY OVER ACTIONAL (hold your tongue for ‘stop speaking’),
BODILY OVER MENTAL (brain for ‘intellect’) and BODILY OVER PERCEPTUAL
(good ear for ‘good hearing’). Since concrete objects are visible and abstract
things invisible, the principle also entails VISIBLE OVER INVISIBLE, which is
reflected in metonymies such as to save one’s skin for ‘to save one’s life.’
Visibility, in its turn, accounts for the default metonymy (43a) CONTAINER FOR
CONTENTS, since things in a container are typically hidden inside the container
and, thus, invisible. At a more general level, the CONCRETE OVER ABSTRACT
principle also accounts for metonymies (1) FORM FOR CONCEPT, and (47)
WORDS FOR THE CONCEPTS THEY EXPRESS, in which the concrete visual or
acoustic shape of a sign stands for the abstract concept denoted by the
(word-)form.
(iv) INTERACTIONAL OVER NON-INTERACTIONAL. Our experience of the world
mainly derives from our interacting in it. Entities we interact with form better
reference points than entities with which we do not interact. We often interact
with parts of a whole so that this principle provides a default motivation for
PART FOR WHOLE metonymies. For example, the part we interact with in
driving is the steering wheel so that we speak of sitting behind the wheel. We
mainly use our hands in interacting with the world and hence speak of hand-
on demonstration, we use our fingers in typing on the keyboard and thus speak
of having the world at our fingertips when we log into the Internet. Our
interaction with things is closely related to their function.
(v) FUNCTIONAL OVER NON-FUNCTIONAL. As shown by Tversky and
Hemenway (1984), we attach particular salience to functional parts such as the
engine of a car. There is a vested human interest in seeing things, in particular
artifacts, from a functional point of view. Thus, the part of a TV-set that appears
to be the most functional is its tube so that we speak of spending the whole day
in front of the tube; the parts of a car that are functional to its driving are the
wheels, the motor and the steering wheel so that we speak of a 24-wheeler, a
motorway and sitting behind the wheel. Parts which are not important in driving
such as the doors, the windshield wipers, or the fenders are, of course, highly
unlikely to be selected as metonymic reference points for the car.
4.1.2. Perceptual selectivity

A number of cognitive principles are relatable to perceptual salience, which,
of course, also determines the way we experience the world as human beings.
Our perceptual apparatus is geared towards things in our immediacy and
presence, things which are large and big, and things which form good gestalts,
have clear boundaries and are specific instances. These foci of perceptual
selectivity can be stated in the following principles of cognitive preference.
(i) IMMEDIATE OVER NON-IMMEDIATE. This cognitive principle accounts for
selecting stimuli in our spatial, temporal and causal immediacy. The me-
tonymy in I’ll answer the phone for ‘I’ll answer the person speaking at the
other end of the line’ is motivated by spatial immediacy. Metonymies (13)
PRESENT FOR HABITUAL as in I always take the 9 o’clock train and (14) PRESENT
FOR FUTURE as in I am off for ‘I will be off’ are motivated by temporal
immediacy. Metonymy (33) EMOTION FOR CAUSE OF EMOTION, as in She is my
joy for ‘she makes me be happy’ is motivated by causal immediacy. Here, the
emotion as an effect of some stimulus serves as the metonymic vehicle. In
general, effects are more perceptible and affect us more than causes. The
immediacy principle also accounts for many emotion metonymies in which
physiological and behavioral responses produced by emotions are used to
stand for the emotions themselves as in He got cold feet for ‘he became
frightened’ (see Kövecses 1990).
(ii) OCCURRENT OVER NON-OCCURRENT. This principle reflects our preferen-
tial concern with real, factual, and occurrent experiences. It accounts for
metonymy (15a) ACTUAL FOR POTENTIAL in expressions such as He is an angry
person or This is a fast car. Here, the occurrent senses of the words angry and
fast as found in their predicative usages stand for their non-occurrent, poten-
tial senses.
(iii) MORE OVER LESS. More of something is usually more salient perceptu-
ally than less of something. This principle accounts for the naturalness of
using expressions denoting the upper, but not the lower, end of a scale for the
whole scale as in How tall are you?, where tall refers to any size. In the social
and political domains, size is related to power and dominance, which may be
seen as metaphorical sizes.
(iv) DOMINANT OVER LESS DOMINANT. This principle explains the metonymic
use of the biggest and most powerful country or part of a country for a larger
geographical unit as in (9b) England for ‘Great Britain,’ Holland for ‘the
Netherlands’ and Russia for the former ‘Soviet Union.’ This principle may
also account for the traditional acceptance of masculine forms in a generic
sense as in mankind, postman or you guys.
(v) GOOD GESTALT OVER POOR GESTALT. A powerful organizing principle for
our perception is the tendency to see whole gestalts rather than the sum of their
parts. The gestalt-perceptual principle also applies to the selection of a pre-
ferred vehicle in metonymy and accounts for the wide-spread use of humans
and whole objects for their active-zone parts. Metonymy (9a) WHOLE THING
FOR A PART OF THE THING as in The car needs washing is thus well-motivated.
Unlike the perceptual gestalt of things, the gestalt of events is exclusively
conceptual in nature. The borderlines of events tend to be much fuzzier than
those of things. A subevent of a larger event may no longer be understood as a
part but may gain the status of an independent event gestalt of its own. As a
result, the metonymy SUBEVENT FOR THE WHOLE EVENT as in I am going to the
doctor for ‘I am seeing a doctor for medical treatment’ is widespread and well-
motivated.
An essential requirement of any gestalt is that it has clearly delineated
boundaries. Consequently, the gestalt-perceptual principle entails as its
subprinciple:
(vi) BOUNDED OVER UNBOUNDED. The productive metonymy (11a) OBJECT
FOR MATERIAL CONSTITUTING THE OBJECT allows us to construe bounded things
as substances as in We had chicken today. Its reverse metonymy (11b)
MATERIAL CONSTITUTING AN OBJECT FOR THE OBJECT as in I sent you an e-mail is
much less productive.
(vii) SPECIFIC OVER GENERIC. This cognitive principle relates to our prefer-
ence for good gestalts as well as for concreteness, immediacy and occurrence.
Specific and definite instances form better gestalts than general and unspecific
entities. The cognitive principle underlies metonymy (17b) SPECIFIC FOR GE-
NERIC and its subtypes. At a purely conceptual level, this principle accounts for
people’s tendency to generalize. For example, O.J. Simpson’s verdict of ‘not
guilty’ was taken by many Americans as a verdict for all black people.
4.1.3. Cultural preferences

The notion of ‘cultural preference’ as used here is mainly meant in the sense of
the preferential status given to particular members of a category. Lakoff’s
work on metonymic models has shown that certain members of a category are
more salient than others with respect to certain dimensions. These dimensions
are more or less strongly determined by a given culture.
(i) STEREOTYPICAL OVER NON-STEREOTYPICAL. Stereotypes probably provide
the best cases of culture-bond concepts. We already came across the impact of
stereotypes on metonymy in connection with categories such as ‘housewife’
and colloquial tautologies as in Boys will be boys.
(ii) IDEAL OVER NON-IDEAL. Ideals provide powerful organizing principles
in a culture. Ideal cases are social constructs and defined with respect to
desirability. Some concepts are inherently defined as desirable by a culture
such as ‘ideal love’ (see Kövecses 1988), others are represented by a paragon
like Babe Ruth for ‘ideal baseball players’ (Lakoff 1987). Also, negative
categories may have ideal examples that can stand for the whole category.
When we say You are a Judas, we use an individual who is a betrayer par
excellence in our culture to stand for betrayers and betrayal in general.
(iii) TYPICAL OVER NON-TYPICAL. Typical members of a category are often
picked out when a category as a whole is described. For example, one may talk
about the symptoms of sneezing and coughing in referring to a cold as in
You’ve got a bad cough.
(iv) CENTRAL OVER PERIPHERAL. The cultural impact of centrality is nicely
illustrated in Feyaerts’ (this volume) study of the conceptualization of stupid-
ity in German. Expressions such as You are not from here, are you? demon-
strate that people who are characterized as stupid live on the periphery of a
linguistic culture. Here, our folk understanding of inferior mental abilities is
based on the metonymy OUTGROUP ORIGIN FOR STUPIDITY.
(v) INITIAL OR FINAL OVER MIDDLE. In our conception of events, an initial or
final phase may be seen as being more important than the central phase. To
pull the trigger for ‘to shoot’ focuses on an event’s initial phase, to sign a
contract for ‘to make a contract’ focuses on an event’s final phase. The
etymologies of creed and mass provide nice historical illustrations of the two
aspects of this principle: creed derives from the first word of the Apostles’
Creed, Credo in unum Deum ‘I believe in one God,’ while mass for ‘service’
goes back to a formula said at the end of medieval church services, Ite, missa
est (contio) ‘go now, the meeting is dismissed’ (Ullmann 1972: 219).
(vi) BASIC OVER NON-BASIC. This principle applies to simple and well-known
‘ground’ routines or affairs as in Lakoff’s (1987: 88f) generators and submodels
and in our preference for basic level categories. The use of expressions such as
I’ve told you a hundred times for ‘several times’ may be said to exemplify a
metonymy BASIC FOR NON-BASIC because a hundred is a basic number.
(vii) IMPORTANT OVER LESS IMPORTANT. This principle accounts for the use
of stage for ‘theater’ as the most important part of the Theater ICM, or the
expression speaking a language for ‘knowing a language’ or the identification
of a capital city with a country.
(viii) COMMON OVER LESS COMMON and
(ix) RARE OVER LESS RARE. Common members of a category are culturally
given reference points and may be used metonymically such as aspirin for any
pain-relieving tablet, while rare members stand out because of their unique-
ness as in Lakoff’s (1987) example of a DC-10 crash, which people general-
ized to the extent that they refused to fly any DC-10.
It is, without doubt, possible to identify more such cognitive principles
which, however, partly overlap with the ones discussed above. Among these
we would probably have to list UNEXPECTED OVER EXPECTED, NEW OVER OLD,
and TRADITIONAL OVER NON-TRADITIONAL.
4.2. Communicative principles
At least two principles seem to contribute to determining the default selection

of a metonymic vehicle: the principle of clarity and the principle of relevance.
4.2.1. The principle of clarity

The communicative principle that ensures maximal ease of accessing the
intended target via a metonymic vehicle may be stated in preferential terms as
CLEAR OVER OBSCURE. This principle is, of course, reminiscent of Grice’s
(1975) maxim of manner, which, amongst other things, requires the speaker to
avoid obscurity. It might be assumed that clarity in communication is best
guaranteed by use of literal speech. Instances of metonymy which have a high
degree of cognitive motivation, however, do not seem to require any more
effort in directing the addressee’s attention toward the intended target. Active-
zone metonymies are highly motivated by the WHOLE FOR PART metonymy
and, hence, are understood clearly and effortlessly. In Langacker’s example
The dog bit the cat, we effortlessly supply ‘the dog’s teeth’ as the intended
target. This metonymy is so natural that it cannot even be replaced by the
expression of the intended target: *The dog’s teeth bit the cat. Here, the
metonymic mode of expression is clearer and more ‘accurate’ than the literal
one. In a vague expression such as They spent the night together, however, the
addressee cannot clearly access the intended target and so communicative
success is not guaranteed.
4.2.2. The principle of relevance

Sperber and Wilson’s (1995: 158) principle of relevance, according to which
“every act of ostensive communication communicates a presumption of its
own optimal relevance,” also applies to the use and interpretation of me-
tonymy. As a communicative principle of preference, it may be stated as
RELEVANT OVER IRRELEVANT. This principle, which could be described more

precisely as SITUATIONALLY MORE RELEVANT OVER SITUATIONALLY LESS REL-
EVANT, underscores the situational and contextual relevance of a metonymic
vehicle. As a rule, a cognitively salient vehicle is also relevant to the situation
at hand. It is only when the principle of relevance is in conflict with one or
more of the cognitive principles that its impact comes to the fore. As pointed
out in the Introduction to this volume, this is the case with in-group talks by
nurses about their patients or waitresses about their customers. Thus, the much
discussed metonymic example of the ham sandwich in reference to a customer
is well-motivated by the principle of relevance since, to the waitress, the food
served provides the best reference point for identifying a customer in the
Restaurant ICM.
4.3. Competing motivations
In light of the previous section, it can be suggested that the more cognitive
principles apply, the greater the cognitive motivation of a metonymy. As a rule,
more than one cognitive principle does in fact apply to a particular case of
metonymy. For example, the metonymy ARTIST FOR HIS WORK as in We are
reading Shakespeare (for ‘Shakespeare’s plays’) is motivated by a bundle of
cognitive principles: HUMAN OVER NON-HUMAN, CONCRETE OVER ABSTRACT, and
GOOD GESTALT OVER POOR GESTALT. Most instances of metonymy, however, are
not ‘fully’ motivated in the sense that not all cognitive principles converge.
Rather, we have a continuum of motivation ranging from fully motivated default
metonymies to weakly or unmotivated non-default metonymies.
Consider again Lakoff and Johnson’s example The buses are on strike for
‘the bus drivers are on strike.’ Since passengers ‘interact’ with the buses and
buses are more relevant to them than their drivers, the metonymy is motivated
by the cognitive principle INTERACTIONAL OVER NON-INTERACTIONAL and the
communicative principle RELEVANT OVER IRRELEVANT, but it is inconsistent
with the cognitive principle HUMAN OVER NON-HUMAN. The metonymy in I’ll
answer the phone is consistent with the principle IMMEDIATE OVER NON-
IMMEDIATE, but is in conflict with the principle HUMAN OVER NON-HUMAN. The
metonymic expression paper for ‘essay on a subject’ is motivated by the
principle CONCRETE OVER ABSTRACT, in particular, VISIBLE OVER INVISIBLE, but,
since paper is prototypically a mass noun, the principle BOUNDED OVER
UNBOUNDED is reversed. In all these cases, conflicting motivations decrease
the naturalness of the overall motivation of the metonymy.
4.4. Summary
In sum, there appear to be a number of cognitive and communicative prin-

ciples that determine the default routes to intended target entities. The signifi-
cance of these principles lies in the fact that they help us understand why we
select certain vehicle entities to access a target and why certain vehicle-to-
target routes have become conventionalized in the language. Many, or even
most, of the default cases of the metonymies listed in the preceding two
sections could be shown to be motivated by these principles. We are also able
to offer an explanation for the hierarchy of metonymic vehicles cross-cutting
ontological realms, which was presented in Section 2.4. as form > form-
concept > concept > reality. Forms rank over concepts due to the principle
CONCRETE OVER ABSTRACT, form-concepts occupy a natural intermediate posi-
tion between the two, and concepts rank over reality due to the principle
SUBJECTIVE OVER OBJECTIVE .
As a rule, several cognitive and communicative principles are involved in
a given instance of metonymy. They may be, and often are, in conflict with
each other and in this way decrease the naturalness of the metonymy. The
following section will briefly investigate the nature of such overriding factors.
5. Overriding factors
The use of metonymy may be motivated by a speaker’s expressive needs or a

given social situation. In the former situation, an individual speaker may want
to achieve a rhetorical effect by deliberately violating one or more of the
cognitive principles in his or her use of metonymy. The latter situation is
governed by social norms and usually involves violating the communicative
principle CLEAR OVER OBSCURE. Since the cognitive and/or communicative
principles are overridden deliberately, the resulting metonymy is usually also
felt to be figurative, or, non-default.
5.1. Rhetorical effects
Along with other figurative modes of thought, metonymy is commonly used

to produce rhetorical effects as in humor, jargon, literature, persuasion, slang,
poetry and the like. The rhetorical effects tend to derive from violations of
default cognitive and communicative principles. For example, Chaucer’s

depiction of ‘fright’ as Myn herte is soore afright and the slang expression get
out of my face for ‘leave me alone’ violate the principle of GOOD GESTALT OVER
POOR GESTALT in that a body part is used to stand for a person. The aesthetic
effect of the metonymies in The pen is mightier than the sword derives from
the deliberate reversal of the cognitive principle HUMAN OVER NON-HUMAN.
Shakespeare’s wording Let pride marry her and the journalist’s description
Many American lives were lost for ‘many Americans died’ both violate the
principle of CONCRETE OVER ABSTRACT.
5.2. Social-communicative effects
Social considerations in a communicative situation may require speakers to

override some cognitive or communicative principles. This is especially obvi-
ous in the use of metonymy-based euphemisms, in which considerations
involving the hearer’s ‘face’ prevent the speaker from using a clear expres-
sion.18 This is the case in the British English expression redundancies for
‘dismissals.’ The euphemistic word redundancy refers to a precondition that
may lead to a worker’s dismissal. The metonymy deliberately reverses the
cognitive principle CENTRAL OVER PERIPHERAL and, since the target is not
clearly identifiable, also violates the communicative principle CLEAR OVER
OBSCURE. The expression They did it for ‘they had sex’ also overrides the
principle CLEAR OVER OBSCURE, but here it is due to its violation of the
principle SPECIFIC OVER GENERIC.
The euphemistic expressions to go to the bathroom and to wash one’s
hands (for ‘to urinate/defecate’) focus on the initial and final phases of a
complex event and are, thus, motivated by the conceptual principle INITIAL OR
FINAL OVER MIDDLE. Since going to the bathroom and washing one’s hands are,
however, only tangentially concerned with the central and relevant activity,
the principles CENTRAL OVER PERIPHERAL, RELEVANT OVER IRRELEVANT as well
as CLEAR OVER OBSCURE are violated. The two expressions are no longer
experienced as instances of metonymic vehicles. Thus, to go to the bathroom
is no longer associated with its spatial meaning ‘to transport oneself to the
bathroom,’ but evokes the target sense directly in expressions such as The dog
went to the bathroom on the living room rug.19 Metonymic expressions which
are no longer felt to mystify a taboo topic tend to be replaced by new non-
default metonymies. This happened to the originally euphemistic word toilet,
which was replaced by bathroom and restroom, which in their turn were
supplanted by expressions such as facilities and comfort station. Violations of
the clarity principle also abound in officialese as in equal opportunity em-
ployer and other areas that are prone to using jargon.
These types of ‘vivid’ metonymy have traditionally been studied in
rhetoric and literary criticism. In the cognitivist view of metonymy presented
here they now appear as non-default cases, in which cognitive principles and/
or communicative principles are deliberately overridden. Since the primary
goal of this paper is to isolate the principles which determine default cases, the
issue of non-default metonymies shall not be explored any further.
6. Conclusion
We have attempted to offer a relatively comprehensive and integrated outline

of metonymy from a cognitivist point of view. As stated in the beginning, we
believe that such a view involves doing at least the following:
ii(i) identifying the ontological levels at which metonymy occurs;
i(ii) specifying the types of relationships that hold between elements in a
metonymic relationship;
(iii) establishing those cognitive principles that explain the most convention-
alized or most ‘natural’ vehicle-to-target routes;
(iv) discovering the conditions under which non-default routes can be se-
lected.
The paper argues that metonymy is not restricted to language but is a cognitive
process which operates within the same idealized cognitive model. Since
ICMs may cross-cut ontological realms, we may also expect to find me-
tonymy-producing relationships in and cross-cutting the three ontological
realms of concepts, forms and things/events. We have been able to identify
eight ICMs which give rise to ‘ontological metonymies.’
The metonymy-producing relationships were subsumed under two general
conceptual configurations: whole ICM and its part(s) and parts of an ICM. The
former configuration typically gives rise to metonymies involving things, the
latter primarily applies to metonymies involving predications. A small number
of conceptual relationships only admit metonymization in one direction; the
majority of conceptual relationships, however, lead to reversible metonymies.
Generally, one of these metonymic construals is conceptually preferred.

A number of cognitive and communicative principles govern the default
selection of the preferred metonymic vehicle. The cognitive principles pertain
to the areas of human experience, perceptual selectivity and cultural prefer-
ences. The communicative principles include those of clarity and relevance.
These cognitive and communicative principles may be overridden for
expressive or social reasons. Non-default metonymies, which arise through
such overriding factors, violate one or more of the default cognitive and
communicative principles, in particular the principle CLEAR OVER OBSCURE.
We do not claim that we have carried out this project fully. On the
contrary, what we have done here is just the beginning and we are certain that
there are others who do not necessarily agree with us concerning either the
details of our proposals or even our general claims. However, we believe that
the four issues raised are important for understanding the nature of metonymy.
If we can come to some consensus regarding these issues, we can address
further ones that have not been discussed in the cognitive linguistic literature
in the detail or depth they deserve. Among these are the following:
1. What is the precise nature of the relationship between metonymy and
other figurative modes of thought, in particular metaphor?
2. What is the function of metonymy, especially in comparison to that of
metaphor?
3. Why do people utilize metonymy?
4. What are the restrictions on the use of metonymy? To use Taylor’s (1995:
123) example, why does the PRODUCER FOR PRODUCT metonymy not apply
to Mary was delicious, meaning by Mary the cheesecake which Mary
made?
5. What is the relationship of metonymic speech to literal speech?
To be sure, there are answers to each of these questions in cognitive linguis-
tics, and some of them have been touched upon in the previous pages.
However, in light of what we have found, it seems to us that much more can be
done. And, indeed, the main goal of this paper was to point out some of the
deeper issues concerning metonymy and bring these issues to the attention of
the cognitive linguistic community.
Notes
1. This paper elaborates upon an earlier version which appeared in Cognitive Linguistics 9–
1 (1998). We are grateful to our anonymous reviewers for their insightful and helpful
comments. We also wish to express our thanks to Elizabeth Mathis and Gary Palmer for
their valuable linguistic contributions to this paper.
2. This view corresponds to Dirven’s (1993: 14) characterization of metonymy as opposed
to metaphor: “[...] in metonymy the two domains both remain intact, but they are seen to
be in line, whereas in metaphor only one domain, viz. the target domain is kept, and the
other domain, viz. the source domain disappears, so to speak.”
3. The attraction of Langacker’s cognitive explanation of metonymy lies in the pervasive-
ness of reference-point phenomena in language structure, in particular possessive expres-
sions. The view of metonymy as a reference-point phenomenon is, however, not
unproblematic. The process of first making mental contact to a reference point before
accessing the target should take longer than that of accessing a conceptual entity directly,
which, however, has not been confirmed experimentally in terms of processing time
(Gibbs 1993).
4. See the discussion of the notion of ‘contiguity’ in Koch (this volume). The notion of
contiguity is also present in cognitive definitions as in Croft’s (1993: 347) definition of
metonymy as “a shift of a word meaning from the entity it stands for to a ‘contiguous’
entity.”
5. For similar views see Lakoff and Johnson (1980), Lakoff (1987), Croft (1993) and Blank
(this volume).
6. The word hearse derives from Latin (h)irpicem ‘large rake used as a harrow’ and Samnite
(h)irpus ‘wolf,’ i.e., the pins of the harrow were metaphorically related to the wolf’s teeth.
7. In distinguishing metaphor from metonymy, Lakoff and Johnson (1980: 36) and Lakoff
and Turner (1989: 103) view metonymy as primarily having a referential function, as
opposed to the function of understanding associated with metaphor. But this view of
metonymy is at odds with the many other areas in which metonymy is found.
8. Frawley (1992: 19–20) illustrates the difference between the extensional world and the
mentally projected world by the well-known case of tautologies as in George Bush is
George Bush. This sentence is meaningfully interpretable in an attitudinal sense, which
may be made explicit by attitude predicates such as wonder, doubt, be uncertain, etc. as in
I wonder if George Bush is George Bush. Gibbs (1994: 345–351) has shown that such
tautologies are metonymic in that a category is used to refer to specific salient parts or
attributes. This analysis is also adopted here (cf. 3.1.(vi)).
9. Cf. Langacker (1993: 31): “it [the active-zone/profile discrepancy] has to be regarded as
natural and expected rather than pathological.” Similarly, Herskovits (1985: 363) points
out in her discussion of the salience principle that the metonymic shift to a salient part
“has come to be so natural that it is hard to think that it involves any special process such
as metonymy.” Langacker (1991: 191) even argues that “It is in fact quite difficult to find
convincing examples [...] where all aspects of the designated entity participate equally in
a relationship.”
10. Cf. the careful distinction between metonymy and synecdoche made by Seto (this
volume).
11. On metonymies relating to the hand, see Kövecses and Szabó (1996).
12. See Langacker (1991: Ch. 3). Prototypical substances are characterized by internal
homogeneity and divisibility, prototypical things are characterized by internal heteroge-
neity and integration of its parts into a whole.
13. For a more general distinction between metonymies based on co-presence and succes-
sion, see Blank (this volume).
14. The metonymy ACTUAL FOR POTENTIAL also accounts for the development of the possibil-
ity sense of the modals can and may. The epistemic sense of can may have developed
from the original sense of ‘know’ via ‘knowledge how to do X’ and ‘being able to act’ to
‘increased possibility of acting.’ In a similar chain of sense development, the original
sense of may ‘might, power’ may have led to the modal sense of possibility. Studies on
grammaticalization have convincingly explained the steps involved in these semantic
changes by implicature and pragmatic strengthening.
15. Cf. Lakoff in the discussion on “Semantic Accommodation” in Cogling, May 7, 1994.
16. Objects and animals may, of course, also be associated with a place. A nice example of
metonymic association is the proper name Canary Islands, which goes back to the name
Canaria given to it by the Romans on account of the many dogs seen there and which
later on provided the name for the bird canary, which the Spanish found on the islands.
17. The following semantic constraints identified by Cooper and Ross (1975) correspond to
the cognitive principles as used here: Here and Now correspond to IMMEDIATE OVER NON-
IMMEDIATE, Singular corresponds to SPECIFIC OVER GENERIC, Animate and Agentive corre-
spond to HUMAN OVER NON-HUMAN, and Count corresponds to BOUNDED OVER UNBOUNDED.
Possibly also the remaining semantic constraints are relevant for metonymy.
18. Violation of the communicative principle CLEAR OVER OBSCURE amounts to the same thing
as flouting Grice’s maxim of manner.
19. Cf. Morgan (1978: 263), who analyzes this example, which goes back to Robin Lakoff, as
conventionalized conversational implicature.
References
Cooper, William E., John Robert Ross

1975 World order. Papers from the Parasession on Functionalism, University of
Chicago: Chicago Linguistics Society, 63–111.
Croft, William
1993 The role of domains in the interpretation of metaphors and metonymies.
Cognitive Linguistics 4: 335–370.
Cruse, D. Alan
1986 Lexical Semantics. Cambridge: Cambridge University Press.
Dirven, René
1993 Metonymy and metaphor: different mental strategies of conceptualization.
Leuvense Bijdragen 82: 1–28.
Frawley, William
1992 Linguistic Semantics. Hillsdale, NJ: Lawrence Erlbaum.
1993 Processes and products in making sense of tropes. In A. Ortony (ed.),
Metaphor and Thought. Cambridge: Cambridge University Press. Second
edition. 252–276.
1994 The Poetics of Mind: Figurative Thought, Language and Understanding.
Grice, H. Paul
1975 Logic and conversation. In P. Cole, J.L. Morgan (eds.), Syntax and Seman-
tics. Vol. 3: Speech Acts. New York: Academic Press, 41–58.
Herskovits, Annette
1985 Semantics and pragmatics of locative expressions. Cognitive Science 9:
341–378.
Jongen, René
1985 Polysemy, tropes and cognition or the non-Magrittian art of closing curtains
whilst opening them. In W. Paprotté, R. Dirven (eds.), The Ubiquity of
Metaphor (Current Issues in Linguistic Theory 29). Amsterdam and Phila-
delphia: Benjamins, 121–139.
Kövecses, Zoltán
1988 The Language of Love. Lewisburgh: Associated University Press.
1990 Emotion Concepts. New York: Springer-Verlag.
Kövecses, Zoltán, Péter Szabó
1996 Idioms: A view from cognitive semantics. Applied Linguistics 17: 326–355.
Kövecses, Zoltán, Günter Radden
1998 Metonymy: Developing a cognitive linguistic view. Cognitive Linguistics
9: 37–77.
Lakoff, George
1987 Women, Fire and Dangerous Things: What Categories Reveal about the
Lakoff, George, Mark Turner
1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: The
University of Chicago Press.
1991 Foundations of Cognitive Grammar. Vol. II: Descriptive Application.
Stanford, California: Stanford University Press.
Morgan, Jerry L.
1978 Two types of convention in indirect speech acts. In P. Cole (ed.), Syntax
and Semantics. Vol. 9: Pragmatics. San Diego: Academic Press, 261–280.
Norrick, Neal R.
1981 Semiotic Principles in Semantic Theory. Amsterdam and Philadelphia:
Benjamins.
Norvig, Peter, George Lakoff
1987 Taking: a study in lexical network theory. Proceedings of the Thirteenth
Annual Meeting of the Berkeley Linguistics Society. Berkeley, California:
Berkeley Linguistics Society, 195–206.
Ogden, Charles A., Ivor A. Richards
1923 The Meaning of Meaning: A Study on the Influence of Language upon
Thought and of the Science of Symbolism. London: Routledge and Kegan
Paul.
Radden, Günter
in print How metonymic are metaphors? In A. Barcelona (ed.), Metaphor and
Metonymy at the Crossroads. Berlin, New York: Mouton de Gruyter.
Talmy, Leonard
1985 Lexicalization patterns: semantic structure in lexical forms. In T. Shopen
(ed.), Language Typology and Syntactic Description. Vol 3. Cambridge:
Cambridge University Press, 36–149.
Taylor, John R.
1995 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press. Second edition.
Thornburg, Linda, Klaus-Uwe Panther
1997 Speech act metonymies. In W.-A. Liebert et al. (eds.), Discourse and
Perspective in Cognitive Linguistics. Amsterdam and Philadelphia:
Benjamins, 205–219.
Tversky, Barbara, Kathleen Hemenway
1984 Objects, parts and categories. Journal of Experimental Psychology 113:
169–193.
Tylor, Stephen A.
1978 The Said and the Unsaid: Mind, Meaning and Culture. New York: Aca-
demic Press.
Ullmann, Stephen
1972 Semantics: An Introduction to the Science of Meaning. Oxford: Blackwell.
Webster’s Third New International Dictionary of the English Language
1971 Edited by Philip B. Gove. Chicago et al.: Merriam Co.
Speaking and Thinking with Metonymy

University of California, Santa Cruz
1. Introduction
The impulse to speak and think with metonymy is a significant part of our
everyday experience. Traditionally viewed as just one of many tropes, and
clearly subservient in most scholars’ minds to the master trope of metaphor,
metonymy shapes the way we think and speak of ordinary events and is the
basis for many symbolic comparisons in art and literature. Consider, as one
example, the French novelist Honoré de Balzac. Like many 19th century
fiction writers, Balzac provides wonderful examples of metonymic descrip-
tions in which the concrete depiction of some object or person stands for or
represents larger objects or domains of experience. For instance, in his novel
Père Goriot, Balzac opens with a long description of a boarding-house owned
by Madame Vauquer (Balzac 1951: 33):
Madame Vauquer is at home in its stuffy air, she can breathe without being
sickened by it. Her face, fresh with the chill freshness of the first frosty
autumn day, her wrinkled eyes, her expression, varying from the conven-
tional set smile of the ballet-dancer to the sour frown of the discounter of bills,
her whole person, in short, provides a clue to the boarding-house, just as the
boarding house implies the existence of such a person as she is.
The possibility of inferring something about the boarding-house from the

description of just a few of the details of Madame Vauquer’s face reflects a
common metonymic mapping whereby the person stands for an object or
place. Beyond this, as Balzac explicitly points out, our understanding of the
boarding-house allows us via metonymy to recognize something about Ma-
dame Vauquer’s philosophy of life.
62 Raymond W. Gibbs, Jr.
Our ability to draw metonymic inferences, where we infer whole from

parts or parts from wholes, is one of the special characteristics of the poetics of
mind (Gibbs 1994). My purpose in this paper is to briefly note some of the
important features of metonymy in speaking and thinking. I have four specific
aims. First, I illustrate some of the diversity of linguistic forms for metonymy
in everyday and literary language. Second, I demonstrate how metonymy is
not just a type of language but reflects a significant form of human cognition.
Third, I describe the importance of psychologically-real processing principles
for a cognitive theory of metonymy. Finally, I argue that metonymic prin-
ciples underlie several other linguistic phenomena, especially conversational
implicature, indirect speech acts, and colloquial tautologies.
2. Distinguishing metonymy from metaphor
At first glance, metaphor and metonymy appear to be similar, because each

describes a connection between two things where one term is substituted for
another. Scholars disagree as to whether metonymy is a type or subclass of
metaphor (Genette 1980; Levin 1977; Lodge 1977; Searle 1979) or whether
metaphor and metonymy are opposed because they are generated according to
opposite principles (Bredin 1984; Jakobson 1971). Yet metaphor and me-
tonymy can best be distinguished by examining how each makes different
connections between things. In metaphor, there are two conceptual domains,
and one is understood in terms of another. For example, in
(1) The creampuff was knocked out in the boxing match.
the term creampuff metaphorically refers to a boxer because the boxer is
viewed as ‘soft’ and easy to defeat. On the other hand, metonymy involves
only one conceptual domain, in that the mapping or connection between two
things is within the same domain, or within the same domain matrix (Croft
1993). Thus, in
(2) We need a new glove to play third base.
the term a new glove refers to a person who would play third base in a baseball
game. One general, but not perfect, rule-of-thumb for distinguishing metaphor
from metonymy is to employ the ‘is like,’ or ‘X is like Y’ test. If an expression
makes sense in the ‘X is like Y’ form, then it has metaphorical meaning. For
instance, the sentence The boxer is like a creampuff makes sense, and thus is
metaphorical, while The third baseman is like a glove does not, and thus is
metonymic.
Metonymy is related to synecdoche in that both tropes exploit the rela-
tionship of larger entities and lesser ones. Synecdoche substitutes the part for
the whole as in
(3) They’re taking on new hands down at the factory.
where the term hands stands for men. Metonymy also substitutes the token for
the type or a specific instance, property, or characteristic for the general
principle or function. For instance, in
(4) They prefer the bullet to the ballot box.
the term bullet represents armed conflict, while ballot box refers to peaceful
democratic processes.
Some linguistic expressions involve both synecdoche and metonymy.
For instance, consider
(5) General Schwartzkopf had 400,000 fatigues at his command.
Here fatigues, the specific type of uniform worn by common soldiers, refers to
soldiers, reflecting a part-whole relationship, or to warlike power, reflecting a
token-for-type relationship. Moreover, many linguistic statements reflect both
metaphor and metonymy (cf. Goossens 1990, this volume). A significant
challenge for scholars is to discover some of the complex ways that metonymy
interacts with other tropes, both within individual statements and across
longer stretches of discourse.
3. The diversity of metonymy
One problem with most discussions of metonymy is that metonymy is too

often seen as simply one kind of linguistic phenomenon. But metonymy is
quite diverse and exhibits itself in a variety of forms in language. Of course,
metonymy is best known for its place in literature such as in Balzac’s novels.
David Lodge (1981) has argued that it is possible to map the literary history of
the 20th century based on the movement back and forth between literature that
is essentially metaphoric (e.g., James Joyce and T.S. Eliot) and literature that
is basically metonymic (e.g., W.H. Auden and George Orwell) (cf. also
Pankhurst, this volume). Moreover, our ability to understand poetry depends
critically on the recognition of metonymy. For instance, consider

Shakespeare’s Othello where Iago pledges total loyalty to Othello. He calls
upon the stars to
Witness, that here Iago doth give up
The execution of his wit, hands, heart,
To wrong’d Othello’s service! (III. iii 465–466)
Wit, hands, and heart are metonyms that stand for the familiar tripartite
division of a human into mind, body, and soul. Like realistic novelists or
biographers, poets, such as Shakespeare, rely heavily on synecdochic detail to
evoke scene, characters, and cultural experience. The poet Philip Larkin, to
take another example, evokes the past glories of race horses in the following
stanza from “At Grass” (Larkin, 1988).
Silks at the start: against the sky
Numbers and parasols: outside,
Squadrons of empty cars, and heat,
And littered grass: then the long cry
Hanging unhushed till it subside
To stop-press columns on the street.
Our understanding and appreciation of this poem depends on our ability to

think metonymically, to recognize, for example, that silks at the start refers to
jockeys atop their mounts at the starting gate. Students often experience
difficulty understanding poetry precisely because they cannot figure out the
referents of metonymic terms.
Although metonymy is primarily studied as a mode of discourse in
literature and poetry, metonymy is an ubiquitous feature of everyday speech.
One of the most important discoveries in the study of metonymy comes from
cognitive linguistic analyses into the systematic character of metonymy in
conventional language. Consider the following expressions:
(6) Washington has started negotiating with Moscow.
(7) The White House isn’t saying anything.
(8) Paris has dropped hemlines this year.
(9) Hollywood is putting out terrible movies.
These examples do not occur one by one, but reflect the general cognitive
principle of metonymy where people use one well-understood aspect of
something to stand for the thing as a whole or for some other aspect of it
(Lakoff and Johnson 1980). All of the above expressions relate to the general
principle by which a place may stand for an institution located at that place.
Thus, a place like Wall Street stands for the particularly salient institutions
located at that place, namely the stock exchange and major banks. Various
metonymic models in our conceptual system underlie the use of many kinds of
figurative and conventional expressions (e.g., OBJECT USED FOR USER, CON-
TROLLER FOR CONTROLLED, THE PLACE FOR THE EVENT, and THE PLACE FOR AN
INSTITUTION LOCATED AT THAT PLACE).
Finally, there are thousands of ordinary verbs that are based on me-
tonymy. Consider the main verbs in the following sentences (cf. Dirven, this
volume):
(10) The librarian shelved the books.
(11) The scientists eyeballed the data.
(12) The maid dusted the table.
(13) The pilot dusted the crops.
(14) He boycotted the story.
(15) We can’t stomach any more articles on metonymy.
(16) He toed the line.
People experience little problem interpreting these expressions, despite the
fact that, for example, one could, in principle shelve books by hitting books
with shelves. But our deep background knowledge of the typical relationship
that shelves have with books allows us to figure out exactly which salient act
one does when shelving books, eyeballing data, and so on.
In other cases, we can create novel verbs based on our knowledge of
particular actions associated with specific people, called ‘eponymous verbs.’
Consider the case of a friend of mine who noted about another person we both
knew
(17) He’s going to OJ his way out of his marriage.
I immediately understood the rather complex meaning of this expression by
drawing a metonymic inference about the salient acts associated with OJ (for
the famous American football player and murder suspect O.J. Simpson),
namely that one can murder one’s wife to get out of the marriage (a comment
that my friend intended sarcastically about the other person we knew). Again,
our ability to think metonymically ‘on the fly’ enables us to make immediate
sense of these novel phrases.
4. Evidence on metonymy in thought
Although metonymy is clearly understood as a special linguistic form, or

trope, metonymy is now widely recognized as a particular type of mental
mapping, again whereby we conceive of an entire person, object, or event by
understanding a salient part of a person, object, or event (see Croft 1993;
Fauconnier and Turner, this volume; Feyaerts, this volume; Radden and
Kövecses, this volume). Under this view, people think in metonymy, as
illustrated in the way we use and understand language, and engage in a variety
of cognitive tasks.
Some of the best evidence in support of metonymic models of thinking
comes from experimental studies of prototype effects in cognitive psychology
(Lakoff 1987). Extensive research demonstrates that participants judge certain
members of categories to be more representative of those categories than other
members. For instance, robins are judged to be more representative of the
category BIRD than are chickens, penguins, and ostriches. Desk chairs are
judged to be more representative of the category CHAIR than are rocking chairs,
beanbag chairs, barber chairs, or electric chairs (Rosch 1978). The most
representative members of any category are termed prototypical members and
these often ‘stand for’ or represent the entire category.
To give another example, consider some prototype effects for the MOTHER
category (Lakoff 1987). People in our culture view housewife-mothers as
better examples of mothers than non-housewife-mothers. This effect is due to
metonymic reasoning where a salient subcategory (e.g., ‘housewife mother’)
has the recognized status of standing for the whole category. Various other
subcategories of mother such as ‘stepmother,’ ‘birth mother,’ ‘adoptive
mother,’ ‘foster mother,’ and ‘surrogate mother’ deviate from the central case
of the prototypical housewife-mother stereotype. The housewife-mother met-
onymically stands for the entire category of mothers in defining how people
reason about mothers and motherly behavior.
Everyday dialogue provides additional evidence for the conceptual basis
of metonymy. Consider the following example:
(18) A: How did you get to the airport?
B: I waved down a taxi.
Speaker B means to inform the listener A that ‘I got to the airport by hailing a
taxi, having it stop and pick me up, and then having it take me to the airport.’
How does a listener infer that B actually found a taxi to take him to the airport?
Traveling from one place to another involves a series of actions where people
find some vehicle to take them to the desired location, get into the vehicle, ride
in it to the destination, arrive and get out. An idealized cognitive model (ICM)
of this series of events includes the following (Lakoff 1987):
Preconditions: You have access to the vehicle.
Embarkation: You get into the vehicle and start it up.
Center: You drive (row, fly, etc.) to your destination.
Finish: You park and get out.
End point: You are at your destination.
It is conventional to use one part of this idealized model to evoke the entire
model. Thus, people can simply mention either the Precondition, Embarka-
tion, or Center to stand for the entire series of events that make up the travel
scenario. In the above brief exchange, speaker B mentions a Precondition (i.e.,
getting access to a taxi by hailing one) to represent the entire travel scenario.
Other possible responses that might work equally well specify other parts of
the idealized model, such as:
(19) I drove my car. (Center)
(20) I called my friend. (Precondition)
(21) I hopped on a bus. (Embarkation)
(22) I stuck out my thumb. (Embarkation)
By metonymically mentioning a subpart of the travel scenario to stand for the
whole scenario, speakers get listeners to draw the right inference about what is
meant. It is interesting, and significant, to note that many cases of conversa-
tional implicature are understood via metonymic reasoning. Consider the
following brief exchange from Grice (1975: 51):
(23) A: Smith doesn’t seem to have a girlfriend these days.
B: He has been paying a lot of visits to New York lately.
Grice argued with this example that what B said only expresses part of what he
meant by his utterance. Successful interpretation of B’s remark demands that
the listener make this inference about what the speaker meant. Grice called
this kind of inference a conversational implicature. Thus, although B simply
stated a fact about Smith’s recent visits to New York, B likely intended for A
to understand that Smith has, or may have, a girlfriend in New York. Follow-
ing Grice, we can say that B implicates the proposition just mentioned by
virtue of what is said along with various background knowledge and beliefs
shared with A, including maxims of conversation that participants in talk-
exchange are mutually expected to observe (Grice 1975).
In many instances, our ability to draw implicatures requires us to see how
a speaker’s utterance metonymically refers to a whole organized sequence of
activities. Thus, suggesting one part of a likely scenario (e.g., men often go to
other places to be with their girlfriends) activates a whole scenario and implies
other unstated parts (e.g., that Smith actually has a girlfriend these days).
Different psychological research, unknowingly, provides good evidence
that people reason metonymically when understanding language. Consider the
following pair of utterances (Gernsbacher 1991):
(24) I need to call the garage (where my car was being serviced).
They said they’d have it ready by five o’clock.
Note that there is a plural pronoun in the second sentence, but not in the first.
But antecedents of pronouns must agree in person, number, and case. None-
theless, Gernsbacher (1991) has shown that people rate as more natural and
are faster to understand the above pair of sentences with ‘conceptual
anaphors’ than they do pairs of sentences with appropriate singular pronouns.
This is so because the singular entity mentioned (garage) metonymically
stands for some conceptual set (the people working at the garage). Plural
pronouns are natural and easy to understand precisely because of our ability to
think metonymically about people, places, events, and objects. We see that the
mention of the subpart metonymically stands for the whole event frame.
Finally, our ability to conceptualize of people, objects, and events in
metonymic terms provides the basis for much of the way we reason and make
inferences during text processing. Many studies show that people metonymi-
cally infer entire sequences of actions having only read some salient subpart in
a story. Consider the following simple tale.
(25) John was hungry and went into a restaurant.
He ordered lobster from the waiter.
It took a long time to prepare.
Because of this he only put down a small tip when he left.
When people hear this brief episode, they presumably activate their knowl-
edge of the activities normally associated with eating in a restaurant and use
this information to fill in the gaps to make the story coherent. This type of
knowledge, called scripts by Schank and Abelson (1977), consists of well-

learned scenarios describing structured situations in everyday life. A number
of experiments show that people automatically infer appropriate script-related
actions when these are not explicitly stated (Abbott, Black, and Smith 1985;
Bower, Black, and Turner 1979; Gibbs and Tenney 1980; Graesser, Woll,
Kowalski, and Smith 1980). This work on script-based text processing illus-
trates the importance of metonymic models in everyday thought. People’s
knowledge in long-term memory for coherent, mundane series of events can
be metonymically referred to by the mere mention of one salient subpart of
these events. We see that mention of the subpart metonymically stands for the
whole event. In some versions of script theory, the most salient part of the
script (i.e., its metonymic representation) is explicitly encoded as a ‘script-
header’ in memory such that activation of the header accesses all the informa-
tion encoded in the entire script (cf. Schank and Abelson 1977). This inference
of a whole script from mere mention of a part facilitates our being able to
assume unstated propositions about what writers mean and to make meaning-
ful sense of seemingly anomalous and disconnected statements in texts.
I have reviewed only a small part of the linguistic and empirical evidence
demonstrating how people conceptualize of many people, objects, and events
via metonymy which allows individuals to reason appropriately about what
speakers intend to communicate. It is especially important to recognize how
many of the linguistic examples I have considered here would not be normally
judged as instances of metonymy by linguists and literary critics. Nonetheless,
people utilize metonymic schemes of thought to reason appropriately about
what is meant. In this sense, then, we must acknowledge a distinction between
processing metonymic language (e.g., understanding utterances like Paris has
dropped hemlines this year) and metonymic processing of language (e.g.,
understanding the gaps in narrative by inferring some rich source of informa-
tion, like a script, from the simple mention of some salient part of that
knowledge).
5. How metonymy is understood
I have described some of the forms that metonymy takes in language and
suggested that people understand many kinds of language, and not just me-
tonymy per se, because of their ability to think metonymically about people,
objects, and events. Let us now consider in a bit more detail how it is that
people understand metonymic language.
Consider the following metonymy (Nunberg 1979):
(26) The ham sandwich is getting impatient for his check.
Understanding this expression requires that a process of ‘sense creation’ must
operate to supplement ordinary ‘sense selection.’ One model, following the
Gricean view, assumes utterances like (26) violate maxims of truthfulness.
This ‘error recovery model’ assumes that sense creation is initiated only after
conventional meaning has been found to be in error. On the other hand, a
‘concurrent processing model’ claims that sense creation and sense selection
operate simultaneously, perhaps in competition with each other.
To test these models, Gerrig (1989) had participants read stories ending
with statements like The horse race was the most popular event. In conven-
tional context, this phrase referred to a standard race between horses. In
innovative context, this phrase referred to a unique situation where snails
competed in a race that was the length of King Louis’s horse. Readers took
roughly the same time to read this statement in both kinds of context. Thus, the
error recovery model seems incorrect. Instead, readers seem to be creating and
selecting metonymic meanings at the same time (i.e., the concurrent process-
ing model). So, understanding metonymy does not require that listeners
realize that this expression violates specific maxims of conversation.
The ability to quickly employ both sense selection and sense creation
processes is nicely illustrated in the following passage. One of the best
examples of metonymy I know of comes from the American satirist Erma
Bombeck, who once wrote in one of her newspaper columns about her
daughter’s difficulties finding a suitable roommate (Clark 1983). Consider
what Bombeck says as she quotes her daughter:
We thought we were onto a steam iron yesterday, but we were too late. Steam
irons never have any trouble finding roommates. She could pick her own pad
and not even have to share a bathroom. Stereos are a dime a dozen.
Everyone’s got their own systems. We’ve just had a streak of bad luck. First,
our Mr. Coffee flunked out of school and went back home. When we replaced
her, our electric typewriter got married and split, and we got stuck with a girl
who said she was getting a leather coat, but she just said that to get the room.
It seems odd, literally speaking, to talk about steam irons having trouble
finding roommates or electric typewriters getting married. Traditional theo-
ries of parsing will fail to handle many of these phrases even though we, for
the most part, can easily understand what Bombeck’s daughter is saying
(Clark 1983). Consider the sentence Steam irons never have any trouble
finding roommates. Most parsers will search their lexicons for the sense of
steam irons that is intended, namely ‘a person who owns a steam iron,’ and
will fail to find anything like this meaning. But the fact that we think about and
talk about people in terms of steam irons, stereos, Mr. Coffee and so on,
reflects the common metonymic mapping whereby we use a salient aspect of
an object or event to stand for the thing or event as a whole.
Clark and Gerrig (1983) have proposed that many ‘contextual expres-
sions’ such as Our electric typewriter got married and split are understood via
the following goal hierarchy:
1. Bombeck wants readers to recognize that she is using electric typewriter
to denote an object used for typing.
2. Bombeck wants readers to recognize that her assertion about getting
married is the kind of thing that she has good reason to believe that on this
occasion we can readily compute uniquely on the basis of our common
ground such that this kind of thing has something to do with electric
typewriters.
3. Bombeck wants readers to recognize that she is using electric typewriter
to denote the individual (one of her daughter’s roommates) who owns an
electric typewriter.
By inferring the lowest subgoal, (1), from the fact that the speaker is using the
noun electric typewriter, we are to infer the next subgoal up, (2), from the fact
that electric typewriter is being used in connection with the verb married.
Finally, we can infer the highest subgoal, (3), from our understanding of (2).
Key among this list of goals is the assumption that the listener/reader will
interpret the speaker’s current utterance given what the speaker and listener at
that moment mutually know (called their ‘common ground’).
Only by considering our common ground with Bombeck can we recognize
when electric typewriter is to be construed as having a ‘stands-for’ relationship
in which certain people, places, events, and things may stand for other people,
places, events, and things. Many of the inferences required to understand what
is meant in this passage are fundamentally metonymic in that each of these
contextual expressions requires readers to understand that the object mentioned
(e.g., steam irons, stereos, Mr. Coffee, and electric typewriter) stands for the
people who own these items (i.e., PEOPLE FOR THEIR POSSESSIONS). Many
contextual expressions, but not all, will be readily understood when interpreted
in light of conventional metonymic mappings such as OBJECTS USED FOR THEIR
USERS, PEOPLE FOR THEIR POSSESSIONS, THE PLACE FOR AN INSTITUTION LOCATED
AT THAT PLACE, or PRODUCER FOR THE PRODUCT. These ‘stands-for’ relationships
reflect pre-existing patterns of metonymic thought that substantially constrain,
in many cases, the kinds of inferences listeners are likely to draw to make sense
of what speakers say. Some psycholinguistic research has shown that people can
determine without great difficulty the appropriate referents for metonymic
expressions in discourse (Gibbs 1990).
6. Indirect speech acts and colloquial tautologies
Metonymic reasoning, where people infer wholes from parts and parts from
wholes, is also important in acts of reference where speakers make requests of
listeners (cf. Panther and Thornburg, this volume). Consider the following
types of requests:
(27) Can you shut the door?
(28) I’d like the door shut.
(29) Would you mind shutting the door?
(30) How about shutting the door?
(31) It’s getting cold in here.
Indirect speech acts arise when speakers plan a social transaction in which the
speaker exchanges something with the appointed addressee for the desired
information or action. To do this, the speaker must first assess what reasons there
may be for the addressee not to give the desired information. The speaker will
then formulate an utterance to deal with the greatest potential obstacles. By
picking out salient obstacles, even ones that are more apparent than real,
speakers assume that listeners can metonymically infer the entire sequence of
actions that must occur for the transaction of goods to be completed.
Research from both naturalistic and laboratory studies has shown that
people formulate their requests to best specify the main potential obstacles for
addresses (Francik and Clark 1985; Gibbs 1986). Thus, in a situation in which
the ability of the addressee to lend the speaker a blue sweater is salient,
speakers prefer the expression Can you possibly lend me your blue sweater?
over the request Would you mind lending me your blue sweater? Reading-time
experiments show that readers are faster to comprehend indirect requests that
specify addressees’ projected obstacles (Gibbs 1986). Thus, what makes some
indirect requests ‘conventional’ is the appropriateness of the sentence form in
matching the obstacles present for addressees in a social context. These
studies emphasize the importance of metonymic reasoning in people’s use and
understanding of indirect speech acts.
Another version of metonymy that has become quite colloquial for speak-
ers is to refer to aspects of people, objects, and events through tautological
statements. Consider the following brief exchange between two parents
(Gibbs and McCarrell 1990). A mother asks her husband Did the children ever
clean up their rooms? The father shakes his head and responds Well, boys will
be boys. At first glance, the father’s response to his wife’s question seems
nonsensical. The phrase Boys will be boys is true by virtue of its logical form
alone (as a nominal tautology) and, superficially, contributes no new informa-
tion to the conversation. But the utterance Boys will be boys is readily
interpretable and most listeners would agree that the father intended to convey
a particular meaning, something like ‘boys will be unruly and it is often
difficult to get them to do what you want.’ Nominal tautologies are found with
surprising frequency in everyday speech, literature (e.g., Gertrude Stein’s
famous line A rose is a rose is a rose), and advertising (e.g., Motor oil is motor
oil). These expressions are metonymic because the speaker refers to a general
category (e.g., boys) to refer to specific salient parts or attributes of that
category (e.g., unruly behavior) (Gibbs and McCarrell 1990). Interpreting
colloquial tautologies requires metonymic reasoning. This ability to infer parts
from whole also underlies the interpretation of expressions like
(32) The New York Times is late for the President’s press conference.
where the speaker means that a reporter representing the New York Times is
late for the news conference.
7. Conclusion
My main argument in this paper has been that metonymy is a significant part
of how people ordinarily think and speak. Although certain individuals, such
as great poets and writers, often create spectacular poetic examples of me-
tonymy, virtually all people are quite familiar with many conventional forms
of metonymy in both language and thought. I have suggested that several
sources of evidence, ranging from work in linguistics to psycholinguistics,
highlight the idea that people experience little difficulty thinking of, speaking
of, and understanding metonymic language. At the same time, evidence from
cognitive psychology and psycholinguistics raises the possibility that people
also reason metonymically (i.e., inferring wholes from parts and parts from
wholes) to solve problems, and draw conversational inferences about lan-
guage that, strictly speaking, is not metonymic. The proper study of me-
tonymy surely extends beyond looking at metonymic language alone. In this
sense, the approach I advocate in this chapter extends beyond looking at
metonymy as a lexical phenomenon, and seeks to discover the ways that
patterns of metonymy in language reflect patterns of metonymic thought.
More generally, my review of research from cognitive psychology suggests
that different forms of metonymic thinking can be found in experimental
situations which did not set out to study metonymy at all. I believe that
psychologists, and others, would benefit greatly from examining a wide
variety of research findings in light of possible links to metonymic thought.
The enthusiasm exhibited for metonymy in this article, and in the contri-
butions of this volume, is encouraging and should greatly boast metonymy’s
reputation from its present status as a secondary trope below metaphor. I must,
nonetheless, raise a cautionary note. Linguists, philosophers, literary theorists,
and psychologists, like myself, must be careful not to assume that a direct link
exists between metonymy in language and metonymy in thought. People may,
for instance, comprehend conventional metonymic language without neces-
sarily drawing metonymic mappings given our experience with many familiar
forms of metonymy. Additional psycholinguistic studies must be conducted to
determine how and when conceptual metonymies are inferred when under-
standing both metonymic and non-metonymic language. Furthermore, me-
tonymy scholars must be careful not to assume that particular mental
processes must operate for metonymy to be successfully understood. Thus,
many scholars have incorrectly conjectured that metonymies can be under-
stood only after some violation of a Gricean maxim has been recognized, an
idea that is not consistent with the available psycholinguistic data. Most
generally, as cognitive scientists from a variety of disciplines move forward to
further pursue the topic of metonymy, we must acknowledge the limitations of
our respective methodologies in drawing conclusions about the role that
metonymy has in thought and language. What is certain, though, is that

cognitive scientists from all disciplines must provide appropriate theoretical
explanations for how people speak and think with metonymy.
References
Abbott, Valerie, John B. Black, Edward E. Smith

1985 The representation of scripts in memory. Journal of Memory and Language
24: 179–199.
Balzac, Honoré de
1951 Père Goriot. Trans. M. Crawford. Harmondsworth: Penguin.
Bower, G., J. Black, T. Turner
1979 Scripts in memory for texts. Cognitive Psychology 11: 177–220.
Bredin, Hugh
1984 Metonymy. Poetics Today 5: 45–48.
Clark, Herbert H.
1983 Making sense of nonce sense. In F. Flores d’Arcais, R. Jarvella (eds.), The
Process of Understanding Language. New York: Wiley, 297–332.
Clark, Herbert H., Richard J. Gerrig
1983 Understanding old words with new meaning. Journal of Verbal Learning
and Verbal Behavior 22: 591–608.
Croft, William
Francik, Ellen P., Herbert H. Clark
1985 How to make requests that overcome obstacles to compliance. Journal of
Memory and Language 24: 560–568.
Genette, Gerard
1980 Narrative Discourse: An Essay in Method. Ithaca, NY: Cornell University
Press.
Gernsbacher, Morton Ann
1991 Comprehending conceptual anaphors. Language and Cognitive Processes
6: 81–105.
Gerrig, Richard J.
1989 The time-course of sense creation. Memory and Cognition 17: 194–207.
1986 What makes some indirect speech acts conventional. Journal of Memory
and Language 25: 181–196.
1990 Comprehending figurative referential descriptions. Journal of Experimen-
tal Psychology: Learning, Memory, and Cognition 16: 56–66.
New York: Cambridge University Press.
Gibbs, Raymond W., Jr., Nancy S. McCarrell
1990 Why boys will be boys and girls will be girls. Journal of Psycholinguistic
Research 19: 125–145.
Gibbs, Raymond W., Jr., Y. Tenney
1980 The concept of scripts in understanding stories. Journal of Psycholinguistic
Research 9: 275–284.
Goossens, Louis
1990 Metaphtonymy: the interaction of metaphor and metonymy in expressions
for linguistic action. Cognitive Linguistics 1: 323–340.
Graesser, Arthur C., S. Woll, D. Kowalski, D. Smith
1980 Memory for typical and atypical actions in scripted activities. Journal of
Experimental Psychology: Human Learning and Memory 6: 503–515.
Grice, H. Paul
1975 Logic and conversation. In P. Cole, J. Morgan (eds.), Syntax and Semantics,
Vol. 3: Speech Acts. New York: Academic Press, 41–58.
Jakobson, Roman
1971 Two types of language and two types of aphasic disturbances. In R.
Jakobson (ed.), Selected Writings, Vol. 2. Cambridge, Massachusetts: The
MIT Press, 239–259.
Lakoff, George
Larkin, Philip
1988 Philip Larkin: Collected Poems. Edited with an introduction by A. Thwaite.
London: Faber.
Levin, Samuel R.
1977 The Semantics of Metaphor. Baltimore: John Hopkins University Press.
Lodge, David
1977 The Modes of Modern Writing. London: Arnold.
1981 Modernism, antimodernism, postmodernism. In D. Lodge, Working with
Structuralism. London: Routledge and Kegan Paul, 3–16.
Nunberg, Geoffrey D.
1979 The non-uniqueness of semantic solutions: polysemy. Linguistics and Phi-
losophy 3: 143–184.
Rosch, Eleanor
1978 Principles of categorization. In E. Rosch, B. Lloyd (eds.), Cognition and
Categorization. Hillsdale, NJ: Erlbaum, 28–49.
Schank, Roger C., Robert P. Abelson
1977 Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowl-
edge Structures. Hillsdale, NJ: Erlbaum.
Searle, John
1979 Metaphor. In A. Ortony (ed.), Metaphor and Thought. New York: Cam-
bridge University Press, 92–123.
Metonymy and Conceptual Integration
Gilles Fauconnier
University of California, San Diego
Mark Turner
University of Maryland
1. Conceptual integration
‘Conceptual integration’ — often called ‘blending’ — is a basic and pervasive

mental operation. It allows us to ‘blend’ two mental spaces to create a third
that is not merely a composition of the first two but instead has emergent
structure of its own. A typical conceptual integration network includes two
input spaces, a blended space, and a generic space. The generic space has the
structure taken as applying to both inputs. All conceptual integration networks
have a partial cross-space mapping between the two input spaces and selective
projection from the inputs to the blended space. The blended space inherits
some structure from the inputs and also has emergent structure of its own
obtained by elaboration and pattern completion. Detailed analysis of the
mechanisms of conceptual integration and of the optimality constraints that
guide it is provided in Fauconnier and Turner (1994, 1996, 1998a, 1998b),
Turner and Fauconnier (1995, 1998, in print), Coulson (1996, 1997), Turner
(1996), and Fauconnier (1997). Conceptual integration operates in many areas
— everyday meaning construction, conceptual change, metaphor and anal-
ogy, scientific discovery, counterfactual reasoning, grammar, action, and
78 Gilles Fauconnier and Mark Turner
design. In the present paper, we restrict ourselves to the interaction of blend-

ing with metonymy.
First, here are some informal examples of the process. Coulson (1997)
considers the case of children in a dormitory inventing a game, based on
basketball, in which a player must throw a crumpled ball of paper into a
wastepaper basket. This new game is a blend. One input is partial knowledge
of basketball and the other input is the dormitory situation with crumpled-up
paper, a wastepaper basket, and so on. The partial mental mapping connects a
ball to crumpled paper, a basketball basket to a wastepaper basket, and players
to children. In the new game, as defined by the blend, some properties are
projected from the ‘basketball’ input (throwing the projectile into the recep-
tacle), others are projected from the ‘dormitory’ input (the basket is on the floor,
not high up in the air; the ball has specific properties of crumpled paper; etc.).
Many other properties of the game will emerge from available aspects of the
context (particular ways of throwing, conventions for scoring, fouls, etc.). The
generic space is the more schematic situation of throwing some object into a
container. It is taken as applying to both of the inputs. This very simple example
illustrates central properties of conceptual integration, in particular the fact that
it is creative (a new activity, different from basketball and different from
throwing paper, is produced) and underspecified (there is more than one way to
project from the inputs and more than one possible emergent structure).
Now consider a more intricate linguistic example. In the 1990s, British
Prime Minister Margaret Thatcher — known as the Iron Lady — was popular
in the United States. It was common to encounter claims that what the United
States needed was a Margaret Thatcher. The response we are interested in is:
“But Margaret Thatcher would never get elected here because the labor unions
can’t stand her.” Thinking about this counterfactual requires bringing Marga-
ret Thatcher together with American electoral politics. We must imagine
Margaret Thatcher running for president in America, and we must develop
enough structure to see the relevant barriers to her being elected. Crucially, the
point of this reasoning has nothing to do with the objective fact that it is
impossible for Margaret Thatcher to be elected, since she is not a citizen of the
United States, is already head of state, has no interest in running, and so on.
The speaker’s point, right or wrong, is that the United States and Great Britain
are, despite the obvious similarities, quite different in their cultural and
political institutions and will not choose the same kinds of leaders.
This point is made by setting up a situation (the ‘blend’) that has some
characteristics of Great Britain, some characteristics of the United States, and

some properties of its own. For example, in the blend, someone who has not
yet been president but is running for president has already had the sort of
experience with labor unions that can only be had by a head of state in Britain.
The two inputs in this example are the political systems of the United States
and Britain, linked by a generic frame of Western democracy. The novel
integrated structure, or blend, has Margaret Thatcher campaigning in Illinois
and Michigan and hated by the American labor unions. In that blend, Margaret
Thatcher is defeated. Because the blend is connected to the rest of the network,
the relevant inferences project back to the inputs, yielding the all-important
conclusion that the speaker is stressing the disanalogy between the United
States and Britain, amounting to the claim that the United States may need a
certain kind of leader, but the intricacies of its electoral politics make it
impossible for that kind of leader to be elected. There is a rich tradition of
looking at counterfactual examples as simply constructing a possible world
that differs minimally from the existing world, but it is clear that this view is
inadequate in the case of the Thatcher counterfactual, since these kinds of
minimal changes (making Thatcher an American citizen, changing the Consti-
tution to allow her to run, and so on) are beside the point.
In the seemingly quite different realm of action and technological design,
consider computer interfaces. The most successful interface is known as the
‘desktop’ interface, in which the computer user moves icons around on a
simulated desktop, gives alphanumeric commands, and selects options from
menus. (The user also selects by pointing and points by manipulating a mouse
or by pressing keys.) The reason for the success of the interface was that
novices could immediately use it at a rudimentary level by recruiting from
their knowledge of office work, interpersonal commands, choosing from lists,
pointing, and moving and handling objects. The entire activity is coherent and
integrated, once learned. The user of the interface manipulates an integrated
structure that derives some of its properties from different inputs — office
work, commands, menus, bodily motion — but that has considerable emergent
structure of its own — pointing and clicking buttons are not at all part of office
work or choosing from lists, having two-dimensional little squares disappear
under other little squares is not part of giving commands or of putting sheets of
paper into folders. The user is manipulating this computer interface neither as
a set of separate modules nor as a template for constructing elaborate con-
scious analogies, but instead as an integrated form with its own coherent
structure and properties. From an ‘objective’ point of view, this activity is

totally novel — it shares no physical characteristics with moving real folders,
and it is novel even for the traditional user of a computer who has issued
commands exclusively from a keyboard rather than from a mouse. Yet the
whole point of the desktop interface is that the integrated activity is immedi-
ately accessible and congenial. The reason, of course, is that a felicitous blend
has been achieved which naturally inherits, in partial fashion, the right concep-
tual structure from various inputs, and then cultivates it into a fuller activity
under pressure and constraints from reality and background knowledge.1
Conceptual structure contains many entrenched products of previous
conceptual integration. In the Thatcher example, integration happens quickly
and looks unremarkable. In the desktop example, there has been laborious
design to develop an efficient blend involving novel computer hardware, but
once that blend was developed, users could work with it quickly, automati-
cally, and productively.
2. Metaphor and metonymy in blends
2.1. ANGER IS HEAT
It is possible to find in even the most studied of basic metaphors examples of

blending and of the way blending interacts with metaphor and metonymy.
Zoltán Kövecses (1986) and George Lakoff (1987) have provided an impres-
sive analysis of metaphoric understanding of anger. This analysis reveals the
required mapping between folk models of heat and folk models of anger. In this
mapping, a heated container maps to an angry individual, heat maps to anger,
smoke and steam (products and hence signs of heat) map to signs of anger,
explosion maps to extreme, uncontrolled anger, and so on. This is reflected in
conventional expressions: He was steaming. She was filled with anger. I had
reached the boiling point. I was fuming. He exploded. I blew my top.
Kövecses and Lakoff also note the important basis for this metaphor in
the folk theory that anger has physiological effects like increased body heat,
increased blood pressure, agitation, and redness in the face. The metonymy
linking emotions to their physiological effects allows expressions like the
following to refer to anger: He gets hot under the collar. She was red with
anger. I almost burst a blood vessel.

The ANGER IS HEAT metaphor provides one set of counterparts. The
metonymy linking emotions to their physiological effects provides another.
When combined, they provide the following system of correspondences:
Table 1. Conceptual correspondences in the ANGER IS HEAT metaphor and in the metonymy
linking emotions to physiological effects
SOURCE TARGET
‘physical events’ ‘emotions’ ‘physiology’
container person person
heat anger body heat
steam sign of anger perspiration, redness
explode show extreme anger acute shaking, loss of
physiological control
boiling point highest degree of emotion
This system of counterparts can be elaborated in various ways:

(1) God, he was so mad I could see the smoke coming out of his ears.
The ears are now mapped onto an orifice of the container in the source. Notice
that in this example, and also in the more conventional ones like He exploded,
the description of the emotion is presented as a physiological reaction of the
individual. Something is happening to his body, e.g., smoke is coming out of
his ears. But the ‘content’ of this physiological reaction is not obtained
through the metonymy in the target. It comes from the ‘source’ (which
concerns physical events pertaining to heated containers — smoke coming
out, explosion, etc.).
The phrase the smoke coming out of his ears does not describe anything
directly in the source (where smoke comes out of inanimate containers and
vessels like kettles, because they are on fire or because they are so hot that
their contents are burning) or in the target (where people’s physiology does
not include internal combustion). There is selective projection from both
inputs, leading to a novel frame in the blend: there are no ears in the source
domain and no smoke in the target domain, but the organizing frame of the
blend has both and they interact.
In the conceptual integration network model, Kövecses’ and Lakoff’s
important observation about the correlation of physiological reactions in the
target domain with the source domain of heat and fire can be reflected
theoretically. ‘Explosion’ cannot be a physiological reaction in the source

(where there is no physiology) or in the target (where there is in fact not much
heat), but it can be a physiological reaction in the blend, where a body can
explode from anger.
In the blended space, we find the people and their emotions projected
from a target input space; we find the corresponding physiological reactions
projected either from the source input of physical heat, explosion and boiling,
or from the target input of the body physiology linked to the emotions.
In the conceptual integration network that is based on the ANGER IS HEAT
metaphor, the following set of counterparts results:
Table 2. Correspondences in the ANGER IS HEAT conceptual integration network

SOURCE BLEND TARGET
Input Space 1 Blended Space Input Space 2 Input Space 3
‘physical events’ ‘emotions’ ‘physiology’
container person/container person person
orifice ears/orifice ears
heat heat/anger anger body heat
steam/smoke steam/smoke sign of anger perspiration, redness
explode explode show extreme anger acute shaking, loss of
physiological control
boiling point boiling/highest degree highest degree of
of emotion emotion
If the blend stood by itself, it could not be interpreted in the real world
because anger does not produce smoke or explosion. But in the network
model, the blend remains linked to the inputs. A sentence like He was so mad
I could see the smoke coming out of his ears directly identifies the blend, not
the inputs, but inferences in the blend can be projected back to the target input
spaces. For example, in the blend, we infer from the smoke’s coming out of his
ears that he is extremely angry and showing physiological signs of it, because
in the blend, smoke is a physiological sign of great anger. We project that
inference back to the target input space: he is extremely angry and showing
physiological signs of it. (We do not project back to the target input space the
specific nature of the physiological signs in the blend, where, e.g., physiology
includes emitting smoke.) Of course, the structure of the blend itself is highly
dependent on the conventional metaphorical mapping of heat to anger.
In addition, we find an explanation for the actual grammatical structure of
an expression like He exploded; I could see the smoke coming out of his ears.
Specifically, we see how vocabulary that picks out apparently incompatible
conceptual elements from quite different domains (smoke should not go with
coming out of his ears since smoke comes from fire but people cannot have
fire inside them) can be used in a clause to evoke a ‘single integrated scene.’
That single integrated scene is available in the blend, even though it is
unavailable from the source or the target. The blend provides a frame (seeing
somebody in an abnormal and dangerous state, with corresponding emotions,
etc.) not available in the source or target.
Next, the blend can have a life of its own, not fully determined by the
inputs. So, we can say, with some hyperbole:
(2) God, was he ever mad. I could see the smoke coming out of his ears
— I thought his hat would catch fire!
It is easy to see how this works: in the blend, the hat on fire is a sign of
even greater heat, hence even greater anger, emotions, etc. But there is no
counterpart for the hat in the source: the elaboration is in the blend, where the
frame of somebody on fire or at least very hot is used (not the boiling kettle
anymore), and the existing mapping operates toward the source (greater heat)
and toward the target (greater anger, but also greater loss of control, greater
social danger, etc.).
The analysis by Kövecses (1986) and Lakoff (1987) underscores the
essential role of physiological reaction metonymies in the formation of the
metaphorical system for emotions. The metonymic correspondences are in the
target — body heat, redness, etc. These map directly onto the blend, in the
sense that in the blend (but not in the target), the physiological reactions are
smoke, explosion, etc. This is done by mapping hot (in the target, for people,
with a certain physiology) to hot in the source (for containers with quite
different physical properties), and then from source to blend, where the new
set of physiological reactions is constructed.
2.2. The Grim Reaper
The representation of death as The Grim Reaper, a sinister skeleton-like

character holding a scythe and wearing a cowl, is a complex integration (cf.
Turner and Fauconnier 1995). The Grim Reaper arises by blending many
spaces: (1) a space with an individual human dying; (2) a space with an
abstract pattern of causal tautology in which an event of a certain kind is

caused by an abstract causal element: e.g., Death causes dying, Sleep causes
sleeping, Lust causes lusting, Sloth causes laziness, and so on (not surpris-
ingly, these abstract causal elements are frequently represented in world
literature as blended anthropomorphic agents with special causal powers); (3)
a space containing a prototypical human killer; and (4) a space with reapers in
the scenario of harvest.
This complex blend allows non-counterparts to be combined by exploit-
ing metonymic connections in the inputs. For example, in the cross-space
mapping, there is no counterpart connection connecting skeleton to reaper,
skeleton to killer, or skeleton to death. But death as a cause is associated with
skeleton as an effect, so there is a metonymic relation between them. There-
fore, the reaper, death, the killer, and the skeleton can be combined in the
blend. In the blend, the killer-reaper is combined with the skeleton in a way
that fits the frame in the blend (the killer-reaper is a person and a person has a
skeleton).
Similarly, death in the input space of human dying is metonymically
associated with priests: priests are stereotypically present at an event of death,
and their institution is concerned with death and afterlife. This metonymy
makes it possible to give The Grim Reaper the attire of a priest even though, in
the cross-space mapping, reapers and priests are not counterparts. The me-
tonymy between death and priests in the input is projected to a PART FOR
WHOLE metonymy in the blend. The cowl, for example, pulled over the head of
The Grim Reaper at once evokes both religious connotations of death and the
impression of death as mysterious, unknown, solitary, and set apart from daily
events in human society.
3. Metonymy and optimality principles for blending
In Fauconnier and Turner (1998a) we offer evidence for a set of optimality

principles on integration networks. These optimality constraints interact and
compete. Here, we can do no more than stipulate them and give some illustra-
tion of their relation to metonymy:
Integration: The blend must constitute a tightly integrated scene that can
be manipulated as a unit. More generally, every space in the network
should have integration.
Web: Manipulating the blend as a unit must maintain the web of appropri-
ate connections to the input spaces easily and without additional surveil-
lance or computation.
Unpacking: It is optimal for the blend alone to allow reconstruction of the
inputs, the cross-space mapping, the generic space, and the network of
connections between all these spaces.
Topology: For any input space and any element in that space projected
into the blend, it is optimal for the relations of the element in the blend to
match the relations of its counterpart.
Good reason: All things being equal, if an element appears in the blend,
there will be pressure to find significance for this element. Significance
will include relevant links to other spaces and relevant functions in
running the blend.
There is a last optimality constraint that is the crucial one in a discussion
of metonymy. It concerns the projection of metonymic links from inputs to the
blend:
Metonymy projection constraint: When an element is projected from an
input to the blend and a second element from that input is projected
because of its metonymic link to the first, shorten the metonymic distance
between them in the blend.
We saw above that blending can combine non-counterpart elements from
a single input, such as death, the cowl of the priest, and the skeleton of the
person who has died. The metonymic distance is large between abstract death
as the general cause of all deaths and the cowl worn by a certain kind of
participant in a ritual associated with particular deaths. But in the blend, the
metonymic connection is direct: the cowl is the attire of death. Similarly, the
skeleton that remains after the corpse has decomposed is a distant product of
death. But in the blend, the skeleton is actually a body part of death, indeed the
body part that provides the form of death. The fact that metonymy is preserved
at all in such blends can be viewed as a consequence of ‘topology’: a
metonymic relation in the input corresponds to a metonymic relation in the
blend. The ‘metonymy projection constraint’ additionally specifies that me-
tonymies get tighter under projection.
Satisfying the metonymy projection constraint is not a matter of blindly
projecting metonymic links. The internal integration of the blend provides
opportunities for some acceptable metonymies but not for others. Since death
is an active person in the blend, and active persons are known to have
skeletons (although they are not normally visible), the PART FOR WHOLE
metonymy connecting a body to its skeleton becomes available as the counter-
part of the distant metonymy in the input. Tightening metonymies as they are
projected from input to blend can help optimize integration in the blend by
creating a tighter, more easily manipulated unit.
Now consider some additional cases that show how metonymy projection
operates. Consider a cartoon illustration of a report that a powerful newspaper
company has succeeded in its hostile takeover of a weaker automobile com-
pany and will now sell off the automobile company’s assets, thereby eliminat-
ing the company. The cartoon shows a giant screw-type printing press
smashing a car. Obviously, the printing press represents the newspaper com-
pany, the car represents the automobile company, and the smashing of the car
by the printing press represents the elimination of the automobile company by
the newspaper company. This cartoon is a metaphorical blend: input one has a
stronger object and a weaker object; input two has the contest between the two
companies. The cross-space mapping is the basic metaphor that maps stronger
objects destroying weaker objects to winning and losing. The strong heavy
object is mapped onto the powerful newspaper company; the weaker object is
mapped onto the weaker automobile company. But in the blend, we find the
printing press as the strong heavy object and the car as the weak object. This is
an efficient exploitation of metonymic connections: the printing press is a
salient instrument of producing newspapers and cars are the salient products
manufactured by automobile companies. In the input with the companies, the
printing press is not an instrument of destruction, but it does have a force-
dynamic function that can be associated with a car-smashing machine of the
sort used in recycling automobiles. In the blend, the printing press is fused
with both the company and the car-smashing machine.
What is going on here? The blend must achieve three goals. First, given
that the cartoon is a visual representation, the blend must be concrete and
specific. Second, it must fit the frame that has the stronger object and the
weaker object. Third, these objects in the blend must be properly connected to
the companies in input two. The companies in input two, being abstract,
cannot in themselves provide the corresponding concrete elements in the
blend. The weaker and stronger objects in input one are concrete but not
specific, and so cannot in themselves provide the corresponding specific
elements in the blend. But we can exploit internal connections in the inputs to
make the elements in the blend adequate. The printing press and the car are
concrete, specific objects associated metonymically with the companies, and
they can be fit into the frame of the stronger object destroying the weaker
object. They fit this frame in part because a printing press is already framed as
having force-dynamic structure that would make it capable of destruction
even though it is not intended for destruction, and in part because we are
familiar with car-smashing machines. In the blend, the two elements are
simultaneously (1) the two concrete, specific objects; (2) a stronger object
destroying a weaker object; and (3) two companies.
Clearly, such a blend is creative. Not just any connections will do. There
has to be a search for elements that simultaneously satisfy a number of
constraints. The printing press and car have topology in the blend (namely, the
printing press crushes objects and the car is the patient of an event of crushing)
that their counterparts in input two do not have (the press in input two is an
instrument of making newspapers, not of crushing, and the car is a salient
product of the automobile company, not a patient of an event of crushing).
Additionally, the printing press and car in input two have no counterparts in
input one. Interestingly, the printing press and the car, which are the elements
that did not project their input-topology to the blend, end up being the only
objects in the blend. (That is, in the input with the companies, the printing
press prints and the car is a working, undamaged product; but this topology
does not project to the blend; in the blend the printing press does not print, but
rather crushes, and the car is not a working, undamaged product, but a crushed
object.) The cartoon is remarkable because it is a case where integration and
topology are maximized by recruiting special metonymic connections in input
two. Because the topologies of strong and weak object on the one hand and
competing companies on the other will match only at a very abstract level, we
find that in addition to the companies, objects metonymically connected to
them are projected to the blend in a way that closely matches and elaborates
the topology in input one of strong and weak objects.
This example emphasizes that conceptual projection is a dynamic process
that cannot be adequately represented by a static drawing. Once the concep-
tual projection is achieved, it may look as if the printing press has always
corresponded to the stronger object and the car to the weaker. But in the cross-
space mapping, the printing press and the car play no role; they have no
counterparts in input one. Rather, the cross-space counterparts are stronger
object and newspaper company, weaker object and automobile company.
Under metonymy projection from input two, the printing press in the blend
becomes the counterpart of the stronger object in input one, and the car in the
blend becomes the counterpart of the weaker object in input one.
This example also shows that identity is metonymy of zero distance. The
metonymic relation in input one between company and commercial product is
transformed into identity in the blend, where the printing press is identically
both a printing press and the newspaper company to which it is, in the input,
metonymically related as an instrument.
Suppose the cartoon now contained the newspaper magnate operating the
printing press to smash the car, which is being driven by the car magnate. Here
the blend structure becomes elaborate through the recruitment to the blend of
an additional adversaries-with-instruments frame in which adversaries fight
with opposing instruments, and in which the winning adversary has the
superior instrument. Now the printing press and car in input two have counter-
parts in the adversaries-with-instruments frame: in input two, the printing
press is a symbol of a capacity for productivity that is an instrument of
corporate competition, and the car is a product that is an instrument of
corporate competition; these instruments in input two are the counterparts of
the instruments in the adversaries-with-instruments frame. Now, the topology
of opposing instruments in the blend matches the topology of opposing
instruments in the adversaries-with-instruments frame. This frame has the
useful property of aligning superiority of instrument with superiority of adver-
sary. In this case, we see that exploiting special internal connections in input
two makes it possible to recruit a frame that makes topology much stronger in
the blend structure.
4. Metonymy in an integrated theory of cognitive phenomena
It is curious that blending — a fundamental, indispensable cognitive opera-

tion, routinely employed in a variety of domains, commonly involved in other
cognitive phenomena that have received extensive analysis — should have
received so little systematic attention in the study of cognition and language.
The routine and largely unconscious nature of blending may have helped it
escape scrutiny. The many well-known spectacular blends — sirens, mer-
maids, chimerae, space aliens, cybernetic organisms, marionettes, Bambi —
may have made blending seem merely exotic.
Given the structural and dynamic mechanisms of blending and the opti-
mality constraints that guide it, metonymy plays an important role in con-
structing conceptual integration networks. Therefore, the various basic
cognitive phenomena that are varieties of conceptual integration — e.g.,
framing, provisional category extension, analogy, metaphor, the construction
of counterfactual spaces, aspects of grammar — will interact systematically
with metonymy according to uniform principles.
Note
1. We are grateful to Dan Gruen for pointing out this example.
References
Coulson, Seana
1996 The Menendez Brothers virus: analogical mapping in blended spaces. In A.
Goldberg (ed.), Conceptual Structure, Discourse, and Language. Stanford:
Center for the Study of Language and Information, 67–81.
1997 Semantic Leaps: The Role of Frame-shifting and Conceptual Blending in
Meaning Construction. Ph.D. dissertation, UC San Diego.
Fauconnier, Gilles
1997 Mappings in Thought and Language. Cambridge: Cambridge University
Press.
Fauconnier, Gilles, Mark Turner
1994 Conceptual projection and middle spaces. UCSD Cognitive Science Techni-
cal Report 9401. San Diego. [Available from http://cogsci.ucsd.edu and
from http://www.wam.umd.edu/ ~mturn]
1996 Blending as a central process of grammar. In A. Goldberg (ed.), Discourse
and cognition: Bridging the Gap. Stanford: Center for the Study of Lan-
guage and Information, 113–130.
1998a Conceptual integration networks. Cognitive Science 22/2: 133–187.
1998b Principles of conceptual integration. In J.-P. Koenig (ed.), Discourse and
Cognition: Bridging the Gap, Vol. II. Stanford: Center for the Study of
Language and Information, 269–283
Kövecses, Zoltán
1986 Metaphors of Anger, Pride and Love: A Lexical Approach to the Structure
of Concepts. Amsterdam and Philadelphia: Benjamins.
Lakoff, George
Turner, Mark
1996 The Literary Mind. New York: Oxford University Press.
Turner, Mark, Gilles Fauconnier
1995 Conceptual integration and formal expression. Metaphor and Symbolic
Activity 10/3: 183–203.
1998 Conceptual integration in counterfactuals. In J.-P. Koenig (ed.), Discourse
and Cognition: Bridging the Gap, Vol. II. Stanford: Center for the Study of
Language and Information, 285–296.
in print A mechanism of creativity. Poetics Today.
Distinguishing Metonymy from Synecdoche
Ken-ichi Seto
Osaka City University
1. Introduction
The purpose of this paper is to address the essential nature of metonymy in

relation to some other closely related concepts, particularly ‘synecdoche.’1
Despite the amount of interest shown in metonymy during the past two
decades, no precise definition of metonymy has been offered yet; most defini-
tions are so vague that they might also cover other concepts such as synecdo-
che, metaphor, irony, etc. A typical definition of metonymy may be: A stands
for B with which A is closely associated. This definition is hardly revealing
because it may also hold true for metaphor and possibly any other rhetorical
terms where meaning transfer takes place. What needs to be clarified, how-
ever, is the meaning of ‘closely associated’ itself. (Note that ‘stands for,’
unless otherwise restricted, will not characterize, even in part, metonymy
either, nor perhaps anything else.)
The following definition of metonymy (1) is meant to be broad enough to
cover the rapidly expanding range of metonymy, linguistic or non-linguistic,
and, at the same time, be narrow enough to exclude any other related concepts:
(1) Metonymy is a referential transfer phenomenon based on the spatio-
temporal contiguity as conceived by the speaker between an entity
and another in the (real) world.
The statement in (2) is an abbreviated formulation of (1):
(2) Metonymy is an E(ntity)–related transfer.
92 Ken-ichi Seto
Several terms in (1) and (2) need explanation: among others, ‘E-related’ or ‘E-
relation,’ ‘contiguity,’ and ‘entity.’ In addition to these, there are two more
terms that are often used in defining synecdoche and metonymy: ‘whole’ and
‘part.’ Both these terms, which will inevitably come into play with the above
notions, also need to be precisely defined because their ambiguity has often
caused confusion and misunderstanding in the discussion of metonymy and
synecdoche. Not only this, but the ambiguity has contributed to the general
supposition that synecdoche is a (mere) subtype of metonymy. It will be
argued in this paper that synecdoche should be independent of metonymy.
Traditionally, synecdoche is defined as a relation in which a part stands for a
whole or a whole stands for a part and a genus for a species or a species for a
genus. I will reserve the notion of ‘synecdoche’ to a C(ategory)–related
transfer, while the notion ‘metonymy’ will be applied to an E(ntity)–related
transfer. Synecdoche, as used in this paper, will be defined as:
(3) Synecdoche is a conceptual transfer phenomenon based on the
semantic inclusion between a more comprehensive and a less com-
prehensive category.
This definition of synecdoche is abbreviated as:
(4) Synecdoche is a C(ategory)–related transfer.
Section 2 argues that the distinction between partonomy and taxonomy
provides a firm basis for the disambiguation of ‘whole’ and ‘part,’ and,
ultimately, for the all-important distinction between E-relation and C-relation.
Section 3 concentrates on the notion of ‘contiguity,’ the most important, but
ambiguous, term used to characterize metonymy. Section 4 defines the notion
of ‘entity,’ and identifies three different types: spatial, temporal, and abstract.
Section 5 describes various types and subtypes of metonymy according to the
three kinds of entities and other norms. Finally, Section 6, after discussing
synecdoche briefly, concludes that synecdoche, which is C-related, should be
independent of metonymy, which is E-related.
2. The ambiguity of ‘whole’ and ‘part’
It is of vital importance to distinguish the two different ways in which the

terms ‘whole’ and ‘part’ can be (and, in fact, have been) used. Their theoreti-
cal disambiguation is particularly urgent because (a) the whole-part relation in

a properly restricted sense will turn out to be one of the major subcategories of
the E-relation; (b) traditionally synecdoche has been regarded as a subtype of
metonymy that is characterized by ‘whole’ and ‘part’; (c) the fact that there is
no consensus about what the terms denote, at least in the discussion of
metonymy and synecdoche, may cast doubt on the validity not only of
traditional studies but of more recent cognitive linguistic treatments of them as
well; and (d) there are reasons to believe that the disambiguation of the terms
may prepare the way to a proper understanding of how our mind works in
general in the interpretation of what is happening in and around us.
Curiously enough, outside the discussion of metonymy and synecdoche,
the ambiguity of the whole-part relation has been rather clearly recognized in
terms of taxonomy and partonomy (meronomy) as in Tversky and Hemenway
(1984), Cruse (1979, 1986), and Tversky (1990). Although there remain
problems about the demarcation and the characterization of taxonomy and
partonomy (for more discussion, see Wierzbicka 1984), the basic distinction
between the two is clear enough: taxonomy is a ‘kind-of’ relation while
partonomy is a ‘part-of’ relation. In other words, taxonomy is the relation
between a more comprehensive category and a less comprehensive one, while
partonomy is the relation between an entity and its parts, such as the relation
between a table and its legs.
Nevertheless, these two relations can be confused. Consider Figures l and
2:
tree body
fir arm
Figure 1. Taxonomy (C-relation) Figure 2. Partonomy (E-relation)

94 Ken-ichi Seto
Figure 1 shows an example of taxonomy: a fir is a kind of tree. Figure 2

represents an example of partonomy: an arm is a part of the body.2 Partonomy
is based on real-world constitutive relations; taxonomy is concerned with
mental (re)classifications of categories. Whereas we have some, if not abso-
lute, freedom to taxonomically (re)classify categories, we are not free to
change constitutive relations in the world because the world is there just as it
is. Thus, the referent of an arm is, wholly or partially, connected physically
with the body of which it is a part. On the other hand, firs could conceivably be
reclassified outside the tree category at any time, without affecting the physi-
cal constitution of the world.
These two classification systems are so obviously different from each
other that it seems hardly possible to confuse them. However, the confusion is
frequent. Why? This is simply because the representations in Figures 1 and 2
(or any variations of these) look the same. Therefore, it might seem possible to
say that firs are a part of the category ‘tree,’ as one says that an arm is a part of
the body. And, in fact, not only might it seem possible, but the English
language (and perhaps any other language) does permit speaking of the fir
class as being a (proper) part of the tree category. But it should not be
overlooked that Figure 1 (or a tree-like variation) is a mere diagrammatic
representation of a class relation of which there is no corresponding form in
the world. This is a spatial metaphor. Since there is no natural form given to
the fir class, when we speak of it as being a part of the tree category, we are
thinking metaphorically. In contrast, Figure 2 reflects the real world body-arm
relation. Figure 1 depicts (metaphorically) a category; Figure 2 represents a
formal relation. To avoid theoretical confusion, the uses of ‘whole’ and ‘part’
should be restricted to partonomy, i.e., the whole-part relation in the world.
The point is: partonomy is based on the perception of contiguity in the real
world while taxonomy is based on the conception of categorical hierarchy
(i.e., hyponymy) in our mind.
This clear-cut distinction between partonomy and taxonomy is missing in
recent cognitive linguistics discussion, despite the cognitive significance of
drawing a sharp line between the two modes of knowledge. For example,
Lakoff and others define metonymy as follows:
(5) a. “We are including as a special case of metonymy what tradi-
tional rhetoricians have called synecdoche, where the part stands
for the whole [...]” (Lakoff and Johnson 1980: 36).
b. “[A] part (a subcategory or member or submodel) stands for the

whole category — in reasoning, recognition, etc. Within the
theory of cognitive models, such cases are represented by
metonymic models” (Lakoff 1987: 79).
c. “Metonymy is used primarily for reference: via metonymy, one
can refer to one entity in a schema by referring to another entity
in the same schema” (Lakoff and Turner 1989: 103).
(5a) shows that Lakoff and Johnson (1980) follow the traditional definition of
synecdoche, which, according to them, is a special case of metonymy. But,
why so ‘special’? So far as the examples dealt with there (e.g., We need some
new blood in the organization), the whole-part relation can be interpreted as
an ordinary referential relation, i.e., partonomy. (5b), on the other hand, tells
us that Lakoff extends the notion of metonymy to a categorical relation, and
(5c) shows that metonymy is, again, meant to be a referential relation, but with
the cautious qualification of ‘primarily’! This uncertainty in defining me-
tonymy in (5a)–(5c) may be resolved if a clear distinction between the
referential relation and the categorical relation is made. To recognize par-
tonomy as distinct from taxonomy would be a first step towards a properly
restricted sense of metonymy.
3. Partonomy and contiguity
The next step is to extend the notion of metonymy, since the referential
transfer based on partonomy is just one kind of metonymy. Therefore it is
necessary to clarify the sense of ‘contiguity’ introduced in (1) because par-
tonomy is just one of several contiguous relations. Again, the term ‘contigu-
ity’ has been used in a number of different ways. The best way to proceed is to
delimit the notion of contiguity to ‘spatio-temporal contiguity between two
entities in the world.’3 Contiguity in this sense covers a wider range of
relations than partonomy. For instance, what is the relation between a kettle
and the water in it? The water is not part of the kettle, but is just in contact with
it. Yet, the kettle can refer to ‘the water’ in it in sentence (6):
(6) The kettle is boiling.
This example shows that contiguity is a notion wider in range than the whole-
part relation.
96 Ken-ichi Seto
There is another reason for adopting contiguity as part of the character-

ization of metonymy and for rejecting the traditional concept of synecdoche as
in (5a). Compare (7a) with (7b):
(7) a. Little Red Riding Hood
b. Bluebeard
Both are metonymies, but there is a difference. Little Red Riding Hood refers
to the girl who always wears the hood so named. The hood is not a part of the
girl, but just in contact with her. Bluebeard, on the other hand, refers to the
man who kills his wives. The beard is a real part of him, not just a false beard
that he might wear. It is clear that the difference is factually negligible and
theoretically trivial. Nevertheless, traditional rhetoric would judge that (7a) is
a metonymy because the hood is not a part of the girl, and that (7b) is a
synecdoche because the beard, a part, stands for the killer, the whole. Consider
another example in (8):
(8) Your nose is running.
Your nose refers to the mucus running out of the nose. The question is: is the
mucus a part of, or just contiguous with, the nose? Here we run up against a
difficulty. We are obviously in an area of fuzziness, where a clear-cut distinc-
tion between part and non-part can no longer be upheld. All this shows that the
traditional distinction between metonymy and synecdoche is not only unsus-
tainable but also theoretically unrevealing.
4. What is an entity?
Just as crucial in the definition of metonymy is the characterization of ‘entity.’

An entity is simply defined as a bounded thing in the cognitive-linguistic
sense of a ‘bounded region’ (cf. Langacker 1987). For our purposes, one may
distinguish three major kinds: spatial, temporal, and abstract things as in (9):
(9) a. spatial entities (a dog, a river)
b. temporal entities (an earthquake, washing)
c. abstract entities (power, beauty)
‘Spatial entities’ are understood here in the sense of physical entities which
have spatial extension. Thus a prototypical spatial entity is a thing that is
bounded by a clear contour, located in a three-dimensional space, and is easy

to recognize as an organic whole such as a person, a dog, a bicycle, etc. This
last condition presupposes that the object is relatively constant as to its
perceptual properties. Some physical entities have less clear contours, such as
mountains, rivers, cities and countries, but usually we also recognize them as
discrete objects by imposing a contour on them, sometimes physically (e.g.,
by building a wall around a city) and sometimes mentally (by drawing a partly
arbitrary and partly motivated line separating one thing from another). And it
is known that our eyes can see a contour where there is no physical contour.
The fact that we usually sketch the outline first when drawing may have
something to do with this.
Contours which we impose on temporal entities are metaphorical in
nature. A temporal entity is bounded by a temporal frame. The term ‘frame,’
itself metaphorical, should give what occurs (or ‘takes place’) in time a
beginning and an end, making it possible to capture the event as an identifiable
whole. Given the status of entities, framed events can behave more or less like
spatial entities; one temporal entity can thus be next to (contiguous with)
another on the temporal axis, just as one spatial entity can be next to another in
the three-dimensional world. Suppose, for example, an earthquake occurs,
closely followed temporarily by another. Each is conceptualized as a unique
event with its own beginning and end so that it is possible to say that there
were two earthquakes in a row. The fact that we can enumerate temporal
entities, such as earthquakes, is grammatically parallel to the enumerability of
ordinary spatial entities.
An event, as a temporal entity, may have its own internal structure:
preceding situation (‘whatever precedes the following process’), process, and
ensuing situation (‘whatever comes out through the preceding process’). The
following sentences show how the process-sense of washing in (10b) extends
metonymically to the senses of preceding and ensuing situations, respectively,
in (10a) and (10c):
(10) a. I saw a pile of dirty washing in the corner. (preceding situation)
b. I must get the washing done tonight. (process)
c. Hang the washing out to dry. (ensuing situation)
This is a case of metonymic extension because, according to my theory,
washing in the sense of process, which is assumed to be basic, is, as a temporal
entity, contiguous with the other two temporal entities, which are the referents
98 Ken-ichi Seto
of the preceding and the ensuing situations.4 In the case of a temporal

metonymy we are dealing with the referential transfer between two contigu-
ous temporal entities.
The third kind of entity is abstract. An abstract entity is typically a salient
property of a thing. Although there are some properties perceptually so vivid
(e.g., ‘red,’ ‘redness,’ ‘height’) that it may hardly seem right to call them
abstract, the point here is that they can become abstract entities in the sense
that they are not bounded by either space or time. The existence of abstract
entities is, as it were, borne out by the ontological metaphor. For example, it is
only through metaphorical reification that power in (11) can be the comple-
ment of has:
(11) He has such a power.
What is important about the status of the three kinds of entities discussed
above is that they exist, whether concrete or abstract, as individuals, not as a
category. And when metonymy is in operation, these entities come into play as
individuals, one referring to another which is in contiguity in the world with
the referring entity. This is the essence of what the E-related transfer means.
5. Kinds of metonymy
The above discussion has shown that there are different kinds of metonymy,
depending not only on the kinds of entities (spatial, temporal, and abstract),
but also on the types of reference, i.e., the way one entity refers to another
(whole-part, container-contents, process-result, etc.). Figure 3 is a simplified
diagram of major types of metonymy classified in terms of E-relations.
metonymy
(E-relation)
spatial temporal abstract
whole- container- adjacency whole preceding- object-

part contents event- ensuing property
subevent
Figure 3: Types of metonymy
In what follows, we shall see how each major type organizes its subtypes and
how they are related with one another.
5.1. Spatial entities
The spatial metonymy provides a model for other types of metonymy. It has
two major subtypes: the whole-part and the container-contents type. The
whole-part subtype is represented by the relation between a physical object
and its components such as the body and its parts. The typical example of the
container-contents subtype is seen in the relation between a kettle and the
water in it.
5.1.1. The WHOLE-PART type

There are quite a number of whole-part relations to be found in the external
world. Metonymy exploits them, taking one entity (a whole or a part) to refer
to another (a part or a whole). However, it is not always the case that all the
whole-part relations guarantee a successful metonymy. The question, then, is
what kinds of whole-part relations can be good candidates for the metonymic
transfer. One can distinguish three major whole-part templates:
(12) a. OBJECT-COMPONENT (e.g., a handle is part of a cup)
b. ORGANIZATION-MEMBER (e.g., a juror is part of a jury)
c. OBJECT-MATERIAL (e.g., the cup is made of porcelain)
Among these whole-part templates, the object-component type, a relation

between concrete count nouns, is prototypical, lending itself to metonymic
exploitation quite easily:
(13) a. The windmill is turning. (‘vanes’)
b. He picked up the telephone. (‘receiver’)
The second subcategory, the organization-member type, is sometimes
subject to metonymic transfer:
(14) The committee have decided to raise membership fees for next
year.
(14) is metonymic if committee is to be interpreted to mean ‘committee
members’ as is suggested by the plural verb agreement. The grammatical
alternation between The government, who have…, are… and The government,
100 Ken-ichi Seto
which has…, is… may also be regarded as metonymic.

The third subcategory, the object-material type, is slightly different from
the other two in that the whole-part relation is partly temporal. Generally,
material is processed to become a product (object). Thus the metonymic
expression nylons for ‘stockings’ is potentially temporal just as well as stein
(from German Steingut ‘stoneware, earthen ware’) for ‘beer mug’ is, though
materials are often, but not always, obvious in the final products.
Here it may be worth mentioning Cruse (1986: Ch. 7), which along with
Winston et al. (1987) is probably the most penetrating analysis ever made of
the whole-part relation. But Cruse leaves some uncertainty with regard to the
classification of collectives. He identifies three subcategories: groups (e.g.,
tribe), classes (e.g., proletariat), and collections (e.g., heap). First, the group
category, which corresponds to our organization-member relation, includes:
tribe, team, cabinet, committee, family, orchestra, jury, squad, audience, etc.
(Cruse 1986: 175). What is the exact relation between a group and its mem-
bers? More generally, how is a group internally organized?5 Let us replace the
term ‘group’ with ‘organization’ because groups are largely restricted to
associations of human beings. Thus, a university may be an organization
consisting of a definite number of faculties.6 Clearly, each faculty is part of
the university, not a kind of university; the university as an organization and
its faculties no doubt stand in a whole-part relation, hence in an E-relation.
Although the university-faculty relation is abstract, this is only due to the very
nature of organizations. Accordingly, the same level of abstractness should
also apply to the relations observed in seemingly more human organizations.
For instance, a committee usually has an internal structure; it is not an
unstructured mass where all the members are simply put together. It may have
some subcommittees which are parts of the committee, not kinds of commit-
tee. Or a committee may consist of a chairperson, some specially nominated
members, and others, who are all partonomically organized just as the human
body parts are partonomically organized. It is for this reason that a special
organization member is called the head (of a school, a corporation, a tribe,
etc.), and another the right arm, an indication that there is a metaphorical
parallelism between a man-made organization and a concrete human body or
an organism like a tree.
Now we are in a position to answer the original question about the
relation between a group and its members. A group is a kind of organization
whose internal structure is partonomically built up, and its individual mem-
bers are minimum units of the structure. By a minimum unit it is meant that a
constituent member belongs to an organization, not as a person per se, but,
rather, as a role unit or a substitutable part. This reasoning may have support
from the sense of substitute as in:
(15) Coming on as a substitute, he scored four crucial goals for
Cameroon.
A substitute has the same function as spare parts of, say, a motor because both
are potential parts of a whole, one for a team, the other for a machine.
Although a substitute is a metaphorical part and a motor part is non-meta-
phorical, the question is not whether a given part is metaphorical or real, but, if
metaphorical, what it is compared to and how it is related with the thing to
which it is compared. The group-member relation is a metaphorical whole-
part, hence, an E-relation. Therefore, the entity-transfer based on the group-
member relation or, more generally, the organization-member relation, is
metonymy.
Still, there may be a certain complication, which brings us to the second
of the Cruse’s three categories: classes. Cruse (1986: 179) refers to cases
where, he supposes, “[a]ny taxonomy can be thought of in part-whole terms
[...] a class can be looked on as a whole whose parts are its sub-classes.” This
is a dangerous step that might jeopardize the very concepts of, and the basic
distinction between, taxonomy and partonomy. It has already been seen that
the congruency of Figure 1 and Figure 2 is an illusion. But it may still be worth
mentioning the difference between organizations and classes, building on
insight gained from the two paragraphs above, so as not to fall victim to
superficial similarities in the figures. Figure 4 shows an organization (a
European-style university) with subclasses superimposed on its members at
the same time.
The column represents an organization: a (specific) linguistics depart-
ment. The department, which itself is usually a part of a larger organization
such as a (specific) faculty, may consist of one professor, three lecturers, and
two assistants. They are all parts of the linguistics department, but at the same
time each of them belongs to one of the three academic ranks: professor,
lecturer, and assistant. These layers represent categories. The professor of the
linguistics department is an element of the category ‘professor.’ The category
‘professor,’ on the other hand, may belong to a supercategory ‘university
teacher.’ It follows, then, that the particular professor has a double function:
102 Ken-ichi Seto
professor
lecturer
assistant
university teacher
Figure 4. Organization and class
one as a part (member) of the linguistics department, the other as an element of

the category ‘professor.’ Therefore, the organization-member relation is a
whole-part relation, hence an E-relation, but the class-element (or the class-
subclass) relation is, unlike Cruse’s supposition, not a whole-part relation, but
a C-relation.7
The last of the three categories of collectives is collection. Collections,
which are typically inanimate, are exemplified by heap, forest, wardrobe
(‘collection of clothes’), library (‘collection of books’), and so on. Thus the
relation between forest and tree is a collection-member relation, which, in
turn, is a whole-part relation. Although it is not our purpose to classify
collectives, the E-relation and the C-relation may well offer a clear perspective
on the matter. For instance, how is crowd conceptualized? Crowd is a concept
fixed with respect to place and time, more specifically, co-presence and
simultaneity. Co-presence is equal to spatial contiguity, and simultaneity to
temporal contiguity, which, combined together, make the typical case of the
E-relation. Crowd is, then, an E-related concept, for persons away from a
crowd, either spatially or temporally, are not part of the crowd. Likewise,
mere spectators are not part of the demonstrators. Herd, flock, and (wolf) pack
are also E-related concepts. In contrast to these, cattle is a C-related concept
because a small herd of cows kept away from the main body are still cattle. 8
5.1.2. The CONTAINER-CONTENTS type

The CONTAINER-CONTENTS type is the second major category of the spatial
metonymy. The transfer usually goes from the container to its contents. This
type is different from the WHOLE-PART type because the container does not refer
to the container plus the contents, but only to the contents as far as the reference
is concerned. In the following example, the dam refers to ‘the water’ in it:
(16) The dam has dried up.
The reason why we treat the CONTAINER-CONTENTS type not only indepen-
dently of the WHOLE-PART type, but also separately from the other types of the
spatial metonymy, is that this type, along with the WHOLE-PART type, pen-
etrates deeply within a language and extends widely over languages, a fact
that points to the cognitive importance of the category. Again, the example in
(17) is a typical instance:
(17) The kettle is boiling.
This sentence is translatable into many other languages word for word.9
There are lots of examples of this metonymic type, from the most proto-
typical down to the marginal. Sentence (18) is another typical example:
(18) He drank three bottles.
The entities below are less prototypical as a container, but still they show the
same transfer pattern of reference:
(19) a. She (re)arranged the bookshelf/closet.
b. The cistern is running over.
c. The lecture hall burst into laughter.
It is important to note that here we are dealing, not with metonymy itself, but
with what might be called the metaphor-metonymy complex (MMC) (cf.
Goossens 1990). The question of defining the boundary of the CONTAINER-
CONTENTS metonymy is closely associated with the question of defining the
boundary of the CONTAINER-CONTENTS metaphor. The metaphoricity of the
‘in’-relation seems higher in the following metonymic example:
(20) The whole village rejoiced.
The relation between a village and the villagers is, literally, not container-
contents, but, still, the ‘in’-relation applies to the pair. When we talk of
104 Ken-ichi Seto
negotiations between Washington and Moscow, the same MMC may be in

operation (cf. Taylor 1995: 123).
The metaphorical ‘in’-relation has been well known since Reddy (1979)
to be so prevalent in our daily language that it is hard to avoid it (cf. Lindner
1981; Talmy 1983; Herskovits 1986; also Weinrich 1964). Whereas it is
beyond the present purpose to investigate the matter, it may be of some help
for future study to note a few more marginal examples:
(21) a. Teddy ate the whole plate.
b. I shall ask for questions from the floor.
c. We eradicated TB from the street.
d. The line ceased running at four p.m. on account of the heavy
snowstorm.
The notion of container may be extended to flat objects like plate and floor,
respectively, in (21a) and (21b), and perhaps further to street and line in (21c)
and (21d), respectively. Anyway, it seems that we are leaving the CONTAINER-
CONTENTS metonymy type and entering the miscellaneous category of me-
tonymy that is only characterized by contiguity, or the E-relation.
5.1.3. Other spatial relations

There are some recalcitrant examples that will resist a neat classification. For
example, what does the potatoes refer to in (22)?
(22) Turn down the potatoes.
The phrase must refer to ‘the gas under the pan which contains the potatoes.’
The potatoes stands in a CONTAINER-CONTENTS relation with the pan containing
them, and the pan is in contact with the (burning) gas, thus the potatoes is also
connected with the gas by contiguity, though indirectly in a chain of metony-
mies. Another example is:
(23) I had to go to the underground streets to find a vacant meter.
Strictly speaking, vacant modifies the parking space next to the meter. From
our viewpoint, the meter metonymically refers to the space adjacent to it.
Between the meter and the space is neither a WHOLE-PART nor a CONTAINER-
CONTENTS relation; it is a relation only characterized by the spatial contiguity
between the two entities (for discussion on modification, see Pustejovsky
1995; Pustejovsky and Bouillon 1995). And if Little Red Riding Hood is,
strictly speaking, not part of the girl who wears it (see example (7a)), it must
be classified otherwise, together with the following examples:
(24) a. The red cap who was on duty last night got drunk.
b. He looked at his wrist. “I’d better get back to work.”
The relative pronoun who clearly indicates that the red cap refers to a person
in (24a); in (24b) his wrist ought to refer to ‘his wristwatch.’ These examples
may suggest the possibility of setting up a new category of metonymy,
something like the possessor-possessed or, perhaps more generally, the CON-
TROLLER-CONTROLLED type, as some scholars in fact do (Lakoff and Johnson
1980; Bredin 1984; Radden and Kövecses, this volume). I shall not pursue this
issue here.
Instead, let us briefly consider the implications that the following sen-
tence, perhaps the most famous example of metonymy (Nunberg 1978: 22),
may have in our context:
(25) The ham sandwich is getting restless at table 20.
In a restaurant situation, one waiter might say (25) to another, intending to
refer to a customer who ordered a ham sandwich. What is, then, the relation
between the customer and the ham sandwich? Is this spatial? In a sense it is; in
another it is not. A ham sandwich means primarily something to eat, and its
relation with the customer may be spatial even before it is brought to the table.
However, a ham sandwich is also a name (form) that is conventionally
associated with the referent so named (this association may be, again, met-
onymic as claimed by Radden and Kövecses, this volume), and the name may
be functioning in (25) as a kind of label stuck on the customer who ordered the
dish. In the latter interpretation, the ham sandwich in (25) can be a property, in
fact, the most salient property in the restaurant situation, that characterizes the
customer in question most effectively to a server. Perhaps this is one of the
reasons that prompt some scholars to adopt broader definitions of metonymy.
Thus Taylor (1995: 123f) suggests that
the essence of metonymy resides in the possibility of establishing connections
between entities which co-occur within a given conceptual structure [and
notes that the] entities need not be contiguous, in any spatial sense. Neither is
metonymy restricted to the act of reference. On this broader view, metonymy
turns out to be one of the most fundamental processes of meaning extension,
more basic, perhaps, even than metaphor.
106 Ken-ichi Seto
Accepting this broader view of metonymy, however, may come at a cost

because it would cloud the issue of distinguishing metonymy from synecdo-
che, or the E-relation from the C-relation. Also, it might distract an analyst
from the fact that there are definitely recurrent patterns of metonymy. Still, it
is true that metonymy is not restricted to the spatial transfer of reference alone.
In the next section we shall see some kinds of temporal metonymy.
5.2. Temporal entities
Temporal metonymy divides into two categories: one is based on the relation
between a whole event and a subevent, and the other on the relation between a
preceding and an ensuing situation.
5.2.1. The type WHOLE EVENT-SUBEVENT

The whole EVENT-SUBEVENT metonymy is exemplified by:
(26) He is reading for the first degree.
Reading is part of studying, which, again, is (supposed to be) part of being a
university student. Reading is a subevent for the whole event of ‘being an
undergraduate student’ in (26). The notions of the whole event and the part
event(s) are a metaphorical extension of the spatial meanings of whole and
part; the term ‘event’ is meant to be broad enough to include processes,
activities, and states of affairs (situations). Thus, in the next example:
(27) She can hardly get out of bed.
the subject is in a situation of being hardly able to get out of bed, which, as a
subevent, metonymically refers to the whole event of ‘being seriously ill.’ In
contrast, being up and about will be a subevent of the whole state of ‘having
recovered from illness.’ One further characteristic of the metonymy of this
type is that the transfer from the whole to the part seems rare, with most
examples restricted to the PART FOR WHOLE kind. This is probably because the
subevent is more often perceptually and conceptually salient while the whole
event, being abstract, lacks the perceptual salience that the spatial counterpart
has.
Instead of adding examples of the SUBEVENT FOR WHOLE EVENT type, let us
turn to a somewhat different set of expressions which, nevertheless, might be
grouped into the same category:
(28) Semis roared past me, taking the curves at fifty.

The roar (sound) of the passing semitrailers may be thought of as a subevent of
the whole event, i.e., an event accompanying the event of ‘passing.’ Similarly,
hiss in (29) may be regarded as a WHOLE EVENT-SUBEVENT metonymy that
refers to the whole reporting event:
(29) “Be quiet!” she hissed.
Likewise, it is possible to use metonymically roar, bellow, and shriek, or even
nudge (‘say with a nudge’) as a (makeshift) reporting verb. Relevant to these
is:
(30) She grumbled all the way up the stairs.
In general, the sounds, manners, gestures, etc. that accompany events can all
be good candidates for the metonymy of this type.
5.2.2. The type PRECEDING SITUATION-PROCESS-ENSUING SITUATION

The PRECEDING-ENSUING type of metonymy is processual:
(31) I felt fiercely proud of my mother for standing up for her righteous
neighbors.
The preceding event of standing up, which literally means just ‘rising to an
upright posture,’ is often (but not always) a prerequisite for doing some
activity (the following event). Thus standing up in (31) metonymically im-
plies that the mother did something positive for her righteous neighbors. The
referential relation between the preceding event and the ensuing event, how-
ever, seems to be more complex than this example may suggest. Figure 5
shows that the basic structure of the processual temporal metonymy is, in fact,
tripartite, consisting of (a) a preceding situation, which by implicature may be
interpreted as causal, (b) a process, and (c) an ensuing situation, which, again
by implicature, may be interpreted as resultant:
108 Ken-ichi Seto
(v)
preceding (i) (iii)

process ensuing
situation
situation
(ii) (iv)
(vi)
Figure 5. Temporal metonymy
As the arrows in Figure 5 show, there are six different routes available for a
metonymic transfer: (i) PRECEDING SITUATION TO PROCESS, (ii) PROCESS TO
PRECEDING SITUATION, (iii) PROCESS TO ENSUING SITUATION, (iv) ENSUING SITUA-
TION TO PROCESS, (v) PRECEDING SITUATION TO ENSUING SITUATION, and (vi)
ENSUING SITUATION TO PRECEDING SITUATION. Among them, the transfers from
process to the preceding situation and from process to the ensuing situation are
particularly frequent as is shown in what follows. The preceding situation
means whatever precedes the process, and the ensuing situation means what-
ever comes out through the process, especially products and (new) states. As
to the preceding situation, four major factors need to be recognized: (a)
material, (b) agent, (c) instrument, and (d) place. Thus the prototypical preced-
ing situation involves an agent who (intentionally) does something some-
where, using some material and some instrument. Keeping all this in mind, let
us now turn to specific transfer patterns of temporal metonymy.
(i) PRECEDING SITUATION TO PROCESS. This transfer pattern is illustrated by
the sentences in (32):
(32) a. He took off the uniform at last.
b. Mrs. Djiak spooned coffee into the pot.
Taking off the uniform in (32a) could metonymically mean ‘retiring from an
organization’: it describes an event preceding that of retiring. Likewise, to
open one’s purse could mean ‘to spend money.’ Spoon in (32b) is a denominal
verb. Clark and Clark (1979) examine denominal verbs and note that the
instrumental verb, of which spoon is an example, is the most frequent and
productive type, followed by locatum, location, goal, and agent verbs. These
verbs can all be (re)interpreted as metonymic because, for example, spoon,
which is an instrument as a noun, means as a verb a process (‘to put coffee into
the pot’) in which the instrument is involved (cf. Dirven, this volume).
(ii) PROCESS TO PRECEDING SITUATION. This transfer pattern typically in-
volves deverbal nouns such as supplies, allurements, toothpick and walk in the
following sentences:10
(33) a. The only grocery stores are holes in the wall that also sell beer,
liquor, and school supplies. (material)
b. There are a lot of allurements in big cities. (agent)
c. The bartender chewed on a toothpick. (instrument)
d. The Milford Track is the finest walk in the world. (place)
Also derived nouns such as invitation belong to this group:
(34) Have you sent out the invitations yet?
Invitations refers to ‘invitation cards’ (instrument) that are used in inviting
people (process). Another subtype of (ii) is the case of washing (process)
referring to ‘clothes that need to be washed’ as in (10a). The basic semantic
function of -ing is assumed to indicate processes taken holistically (cf.
Langacker 1991: 26).
(iii) PROCESS TO ENSUING SITUATION. This type is also productive:
(35) a. I carried the sheaf of printouts to a crowded table in the
periodicals room.
b. Anyone who came to work with a three-day growth, shoulder-
length hair, and a dirty sport jacket hanging over jeans would
be instantly slung into a black hole.
Printouts is a conversion from the corresponding verb, and growth is a derived
noun; both mean the outcome of a particular process. Another subtype of the
PROCESS TO RESULT type is characterized by the -ing-form; unlike the -ing
subtype in (ii), it works the opposite way, from process to result:
(36) My furnishings are Spartan gleanings from police auctions and
resale shops.
If we add more common examples like painting, lettering, clipping, etc. to this
class, it will expand quickly.
(iv) ENSUING SITUATION TO PROCESS. Regular examples of this class are
found in denominal verbs such as to cash, to cripple, to group, etc. They refer
110 Ken-ichi Seto
back, as verbs, to certain processes through which the corresponding nouns

result. Sometimes it is difficult to decide which of the two senses involved, the
one of result or that of process, is more basic. For example, statement is
ambiguous between the result sense (‘what is stated’) and the process sense
(‘the process of stating’), knowledge is ambiguous between ‘what is known’
and ‘the knowing of it,’ perception is ambiguous between ‘what is perceived’
and ‘the perceiving of this,’ and so on. This phenomenon was originally
referred to by Ogden and Richards (1956 [1923]: 134) as “utraquistic subter-
fuge.” The same kind of ambiguity is observed in:
(37) a. The building took a long time. (process)
b. The building is not old. (result)
There seems to be no systematic way of deciding which sense is to be given
priority for building or some other instances of the process-result ambiguity
(cf. Wells 1977; Sorensen 1990; Warren 1995). By contrast, if, in an answer to
the question of whether someone was angry or not, (38) is given:
(38) He slammed the door very hard.
then the questioner would reason that the person was very angry. It may also
be taken as a case of the RESULT TO PROCESS metonymy, an instance of
metonymic reasoning, though it must be admitted that the distinction between
the RESULT TO PROCESS type and the RESULT TO PRECEDING SITUATION type is
not very clear.
(v) PRECEDING SITUATION TO ENSUING SITUATION. This metonymy involves
the transfer from an earlier stage to a result. Depending on the kinds of the
earlier stage (material, agent, instrument, and place), there are four subcatego-
ries. The MATERIAL TO RESULT metonymy is exemplified by:
(39) I squeezed her shoulder gently and offered to make her some eggs.
Some eggs (material) will mean, for example, an egg dish like an omelet
(result). This type, given appropriate contexts, provides a lot of examples:
linen (‘sheets,’ ‘towels,’ etc.), silver (‘cutlery’), glass (‘drinking vessel’),
glasses (‘pair of spectacles’), nylons (‘stocking’), silks (‘silk dresses’), malt
(‘whisky’), oils (‘oil paintings’), plastic (‘credit cards’), leathers (‘leather
clothes’). It may be possible to add to the list some more examples such as the
Yellow Pages, the White Pages, and perhaps also White Paper and paper as in
I read a paper at the conference. Examples of the INSTRUMENT TO RESULT
metonymy are also easy to find: the pen as in live by the pen or The pen is
mightier than the sword, and the brush as in the brush of Turner. As for
examples of the PLACE TO RESULT (product) metonymy, there are Bordeaux
(‘wine’), Scotch (‘whisky’), and so on, where place names seem to function
partly as agents because they are closely connected with the local people who
have inherited a special knowledge of the products there. This may be related
to the fact that there are as many examples of the AGENT TO RESULT metonymy,
which is exemplified by:
(40) a. That was a truly beautiful Picasso.
b. I found myself whistling Mozart under my breath.
Not only can persons be the cause (agent = producer) of valuable products, but
companies can also be agents:
(41) a. I pulled the cover from my mother’s old Olivetti.
b. I eat anything. Cold pizza, McDonald’s, you name it.
(vi) ENSUING SITUATION TO PRECEDING SITUATION. Representative examples
of this transfer pattern are:
(42) a. She is my pride and joy.
b. The news was a great satisfaction to all of us.
c. Self-complacency is the death of the artist.
Pride and joy in (42a) mean what causes them. Likewise, if the basic sense of
satisfaction is assumed to be ‘the resultant state of being satisfied,’ a (great)
satisfaction in (42b) is a metonymic transfer from it, meaning ‘what satisfies
one.’ Finally, death in (42c) means ‘cause of death,’ just as death (‘bullets’) in
Death fell in showers.
5.3. Abstract entities
It has already been seen in (11) how a property can behave as an entity. Some
properties may be essential in the sense that they are inherent attributes that
characterize the whole entities in question; others may be accidental in the
sense that their saliency is highly dependent on context. Clearly, there is a
gradience of properties from the most essential to the most accidental. The
essential end is usually conventionalized and wholly or partly lexicalized,
which is exemplified by:
112 Ken-ichi Seto
(43) a. The Prime Minister is here to see you, Your Majesty.

b. She was considered a great beauty in her youth.
The case of abstract nouns (Majesty, beauty, etc.) standing for concrete nouns
has been traditionally considered metonymy, but no reason has been offered
as to why this semantic transfer can be regarded as a type of metonymy, nor as
to how it can be related to other more obvious types of metonymy. In order to
treat systematically the ABSTRACT FOR CONCRETE type, it should be integrated
into the PROPERTY FOR OBJECT metonymy, for what is important about Majesty
and beauty is not that they are abstract nouns, but that they denote essential
properties of their respective referents. Whether a given expression is abstract
or not is not directly relevant to the conditions of successful metonymy.
Essential properties are, however, typically realized by adjectives. Some
are given an independent lexical status; others, semi-independently, are given
separate lexical senses within the same lexical items whereby they, as entities,
refer to the objects that are characterized by the properties. Drunks and
multinational in (44) are such examples:
(44) a. These old drunks eat up a lot of family time.
b. You’re all too naive to manage a multinational.
Some properties such as usual in (45a) and nonprofits in (45b) are less
essential and more contextual, so that smaller dictionaries like the Longman
Dictionary of Contemporary English (DCE) contain no entries for them:
(45) a. “Your usual, boys?” she asked, grabbing a set of clean steins.
b. Arcadia House, like most nonprofits, survives on a shoestring
of grants and donations.
Still less essential properties would have very little or no chance of being
listed in dictionaries. In (46) a three-o’clock means ‘a meeting at three
o’clock’:
(46) I’ll catch up with you downstairs, Charlie, to go over the Nether-
lands proposal. And Luke, we have a three-o’clock, don’t we, to
discuss the Bloomington plant.
The next example shows how an accidental property can stand for the object
of which it is a property:
(47) When I mounted the rickety steps I saw Mrs. Polster drinking
something murky-brown out of a corrugated glass. It might have

been instant iced tea, but it looked as though it had been mixed with
transmission fluid. [....] She finished the brown-murk and pulled a
pitcher from the left side of the chair. (Sara Paretsky, Guardian
Angel)11
The brown-murk, which refers to ‘something murky-brown,’ is the only
contextual clue to describe the mysterious fluid that might have been instant
iced tea. Of course, it is no less impossible to give brown-murk or murky-
brown a definite lexical sense that describes the apparent soft drink than to
give three-o’clock in (46) the lexical sense ‘three o’clock meeting.’ Some
properties are thus so context-dependent that, out of context, they would lose
the E-relation they have with their referents.
In summary, metonymy involves an E-related transfer, a referential trans-
fer based on the contiguity between one entity and another in the (real) world.
There are three kinds of E-relations: spatial, temporal, and abstract. The
spatial E-relation has two major subtypes (the whole-part and the container-
contents) and one miscellaneous class that defies any further classification
(e.g., turn down/off the soup). The temporal E-relation has two major sub-
types: the whole event-subevent is static and the earlier-later is processual,
and the latter is, more precisely, the earlier preceding situation-process-
ensuing situation relation. The abstract E-relation is the property-object rela-
tion, where a lexically or contextually salient property, given the status of an
entity, stands for the object that it characterizes. All these relations are linked
by one common feature: contiguity in the world as conceived by the speaker.
This is the essence of the E-relation on which metonymy is based. And it
should not be confused with the C-relation that provides the cognitive basis
for synecdoche, to which I shall turn next.
6. Synecdoche and the C-relation
Traditional rhetoric has assumed that synecdoche describes the WHOLE FOR
PART (or, more specifically, the PART FOR WHOLE) kind of metonymy, without
distinguishing between the WHOLE-PART and the GENUS-SPECIES relations. Con-
sequently, no meaningful distinction has been drawn between metonymy and
synecdoche. More fundamentally, this is because the C-relation has never
been clearly distinguished from the E-relation. Synecdoche is now defined as
114 Ken-ichi Seto
the C-related transfer, i.e., the categorical transfer based on the semantic
inclusion between a more comprehensive and a less comprehensive category.
Thus defined, synecdoche becomes an internally consistent and externally
independent category. However, questions remain about the status of synec-
doche. Is it a category that should be comparable to metonymy, or is it a far
smaller category of minor importance? Moreover, whereas metonymy is rich
in variety as has been demonstrated in Section 5, synecdoche in the sense
defined above may look simple and poor in variety, because taxonomy ex-
hausts the C-relation. It is true, indeed, that synecdoche has no more than two
subtypes, but this fact should not in the least reduce the significance of
synecdoche. The categorical transfer based on taxonomy is, in fact, so preva-
lent in our daily language that its importance in the cognitive system could not
be overemphasized. The ubiquity of synecdoche in this sense is such that
actual examples are easy to overlook, with the consequence that synecdoche
has scarcely received the attention it deserves.12
Logically, there are two types of synecdoche: one is the transfer from a
less comprehensive category (species) to a more comprehensive (genus); the
other is the transfer from a more comprehensive (genus) to a less comprehen-
sive (species). The SPECIES TO GENUS transfer is exemplified by walkman: it
was and is the trademark for Sony’s personal stereo, but the once proper name
Walkman (species) has become a common noun, walkman (genus), for any
personal stereo. Likewise, the proper name Spa became a common noun, spa.
Another kind of example is pencil case where pencil (species) means ‘writing
instruments’ (genus). Bread as in to earn one’s (daily) bread is a similar
example. The transfer pattern from SPECIES TO GENUS is also observable in
verbs:
(48) If you wanted to bury, burn, or ship refuse in the Chicago area, you
had to cut him in on the action.
Ship is an instrumental denominal verb, but the referential range of the
instrument has been extended from marine vessels to any means of transporta-
tion.
The second subtype of synecdoche is the transfer from GENUS TO SPECIES:
(49) a. I have a temperature. (‘fever’)
b. I got a ticket again. (‘traffic ticket’)
c. My, my, you’re certainly a sight. (‘terrible sight’)
These are all general nouns used to convey more specific senses than they
literally mean. We do not always have to specify the details of things; when
the situation permits, we tend to use words with more general meanings that
require less effort. This phenomenon may have something to do with the
maxim of informativeness, the second of the Gricean four maxims (cf. Grice
1975). An extreme case of this sort is observed in:
(50) Now that he’s been promoted, he thinks he’s really somebody. (‘an
important person’)
Some -able adjectives also show the GENUS TO SPECIES pattern of synecdoche:
(51) Naperville is ringed by genteel tract houses on sizable lots.
The adjective sizable means ‘rather large,’ though whatever has a size, regard-
less of whether it is large or small, can be, literally, size-able. Readable,
considerable, etc. are of this type, too.
Perhaps somewhat harder to notice are verbal examples of the GENUS TO
SPECIES synecdoche. For instance, suppose one fashion model criticizes an-
other model’s manner of walking on the stage by uttering:
(52) She can’t walk.
Example (52) means ‘she can’t walk professionally (on the stage).’ Once it is
understood that walk in (51) is an example of the GENUS TO SPECIES synecdo-
che, the number of the examples of this class become virtually infinite; any
word can be used in a more specific sense than it normally means. Of course,
the opposite also holds true. The English language (and probably any other
languages) provides some metalinguistic devices to deal with the elasticity of
meaning of this kind: strictly speaking, loosely speaking, real (e.g., Mary’s
husband is a real bachelor), fake, etc. (Lakoff 1972; Taylor 1995) as well as in
a narrow sense and in a broad sense. What should be stressed here is that
practically all words (perhaps with a few exceptions) are capable of changing
their own categorical range (within limits) with no assistance from any of the
above devices. To limit ourselves to the GENUS TO SPECIES synecdoche, we
could say, for example, I can’t sleep or You can’t hit, meaning, respectively, ‘I
can’t sleep well’ and ‘you can’t hit (a ball) well.’ Or suppose a mother asks her
son to go and buy a dozen eggs, then she would expect him to come back with
a dozen chicken eggs, not a dozen reptile eggs. Sometimes professional people
may take advantage of this type of synecdoche. For example, stone for ‘jewel,’
116 Ken-ichi Seto
thus involving a potential customer in the specialist knowledge of the word.

All these examples show the same meaning transfer pattern: from a general to
a specific sense.
7. Conclusion
I have discussed why the essential nature of the term ‘metonymy’ has not yet
been agreed upon and given some suggestions for improving upon the present
situation. One obstacle which has prevented a proper understanding of me-
tonymy is the failure to distinguish the C-relation from the E-relation. The E-
relation, which is the contiguous relation between one entity and another in the
world, is quite different from the C-relation, which is the conceptual relation
between a more comprehensive (more inclusive) and a less comprehensive
(less inclusive) category. It must be stressed here again that a category and an
entity are cognitively two different things. It is important to note this because
confusing them is so frequent in the literature of metonymy (and synecdoche).
Another obstacle, closely connected with the first, is the failure to notice
the seeming ambiguity of the terms ‘whole’ and ‘part,’ which led to the
traditional view to treat synecdoche as a mere subtype of metonymy. In
people’s folk view, taxonomies (‘kind of’ relations) tend to be equated with
partonomies (‘part of’ relations). It has been argued in this paper that synecdo-
che, if it is to be a consistent category, should take only taxonomy, leaving
partonomy to metonymy because taxonomy is equivalent to the C-relation
while partonomy is one type of the E-relation. Therefore, it is concluded that
synecdoche, which is C-related, should be independent from metonymy,
which is E-related.
Notes
1. Part of this paper was first read at the 11th New Zealand Linguistic Society Conference
(Victoria University, 19 May, 1995). I would like to thank Claudia Brugman, Elizabeth
Mathis, Günter Radden, Jae Jung Song, Tomoko Tsujimoto, Kimihiro Yoshimura and
Beatrice Warren for their valuable comments and suggestions on earlier versions of this
paper. I am especially indebted to Brigitte Nerlich for a number of critical comments on
the manuscript. My special thanks are due to John Taylor for reading the entire draft and
making helpful suggestions and constructive criticisms.
2. A taxonomy might be better represented as a tree diagram and a partonomy as a ring in a

circle, but it is important to note that a tree diagram itself can also be interpreted in terms
of partonomy, just as a branch is a part of a tree in the world. On the other hand, a figure
showing a ring in a circle, even if it is intended to represent a partonomical relation, can
also be interpreted in terms of a taxonomy just as easily. The choice of figures or
representational formats does not affect the argument about Figure 1 and Figure 2. What
is at issue is a clear distinction between partonomy and taxonomy, and ultimately
between E-relation and C-relation.
3. This is one, presumably the most important, of the several different senses that Jakobson
(1956) gives to the term ‘contiguity.’ For some controversies concerning the character-
ization of metonymy in the classical paper, cf. Cooper (1986: 34–37), Dirven (1993: 10),
Warren (1995: 137f), and Nerlich (forthcoming).
4. Washing in (10a), which means ‘clothes to be washed,’ may also be seen as a ‘material’
cause. ‘Material’ can be one of the ‘causes’ which make the following process possible.
5. Note that I am not concerned with the relation between the group category and its
subcategories. Thus a subcategory tribe is C-related with its supercategory group (e.g.,
tribes are a kind of (human) group).
6. To repeat the same point, the category university may be classified into subcategories
such as national university, state university, private university, and so on. Of course,
other subcategorizations (e.g., women’s university) are also possible. All this is a C-based
classification.
7. Cruse (1986: 176) claims that proletariat-worker is a class-member relation, and that it is
an example, though admittedly marginal, of the whole-part relation. In our terms,
proletariat-worker is a class-element relation, which is necessarily C-related, not a
whole-part relation. Compare this with the relation between a (specific) labor union and
its members. The latter is an organization-member relation, hence an E-relation (e.g., a
union member is a part of the union). Also worth mentioning is that the existence of
sentences like Frying is part of/a type (kind) of cooking does not imply that sometimes the
distinction between the E-relation and the C-relation gets blurred, but instead confirms
that there are two different cognitive models. One model sees cooking as a series of
processes that are aligned along the temporal axis: Frying is one such process, i.e., a
(temporal) part of the whole processes (frying may come after chopping). The other
model sees cooking as a category of different cooking methods: Frying is one of the
possible cooking methods as opposed to, say, boiling, grilling, roasting, etc.
8. Note that the distinction between E-relation and C-relation does not correspond to the
traditional distinction between extension and intension because the latter terms are both
concerned with categories. Roughly speaking, extension is concerned with a class of
entities, and intension is concerned with the defining property of the class (cf. Lyons
1995: 81). However, while the C-relation has to do with categories, i.e., classes, the E-
relation has to do with the contiguous relation between individual entities, not classes.
9. Randomly chosen languages are: Japanese, Korean, Mongolian, Javanese, Turkish, and
Italian. Though it is too early to speak of universality (the German translation, for example,
is only marginally acceptable), the one-to-one correspondence among several different
languages in the relevant respect may hint that the container-contents type as a metonymic
template, not as a specific example, is a cross-linguistic phenomenon. This may be partly
118 Ken-ichi Seto
confirmed by the fact that all the examples in this section, like most of the metonymic
examples in the other sections, translate acceptably into Japanese word for word.
10. One may find it difficult to accept some of the following examples as instances of
metonymy because the nominalization process itself causes the shift in meaning. It may
be argued, however, that the meaning shift in the nominalization process itself is
metonymic. Thus supplies seems doubly metonymic: one from supply (v) to supplies (n);
the other from supply (n) as in a regular supply of fresh vegetables to supplies (n).
11. Most metonymic examples in this paper have come from Sara Paretsky’s V. I.
Warshawski series: Guardian Angel, Toxic Shock, Burn Marks, Tunnel Vision, and
Indemnity Only. A few examples are from dictionaries and other sources.
12. In diachronic semantics synecdoche in our sense has been known as ‘generalization’ and
‘specialization.’ Thus Geeraerts (1994) refers to specialization, generalization, me-
tonymy, and metaphor as “the classical quartet,” which I would like to call ‘the classical
trio’ in terms of synecdoche, metonymy, and metaphor. Nerlich (forthcoming) is clear in
appreciating the cognitive difference between metonymy and synecdoche along the line
of the present paper: “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, synecdo-
che of how it is ordered in our mind.”
References
Bredin, Hugh
Clark, Eve V., Herbert Clark
1979 When nouns surface as verbs. Language 55: 767–811.
Cooper, David E.
1986 Metaphor. Oxford: Basil Blackwell.
Cruse, D. A.
1979 On the transitivity of the part-whole relation. Journal of Linguistics 15: 29–
38.
1986 Lexical Semantics. London: Cambridge University Press.
DCE
1995 Longman Dictionary of Contemporary English. Essex: Longman.
Dirven, René
1993 Metonymy and metaphor: different mental strategies of conceptualisation.
Geeraerts, Dirk
1994 Historical semantics. In R.E. Asher (ed.), The Encyclopedia of Language
and Linguistics. Oxford: Pergamon Press, 1567–1570.
Goossens, Louis
Grice, H. Paul
Herskovits, Annette
1986 Language and Spatial Cognition: An Interdisciplinary Study of the Prepo-
sitions in English. Cambridge: Cambridge University Press.
Jakobson, Roman
1971 [1956] Two aspects of language and two types of aphasic disturbances. In
R. Jakobson (ed.), Roman Jakobson: Selected Writings. Vol. 2: Word and
Language. The Hague, etc.: Mouton, 239–259.
Lakoff, George
1972 Hedges: a study in meaning criteria and the logic of fuzzy concepts. CLS 8:
183–228.
1989 More than Cool Reason. Chicago: The University of Chicago Press.
1987 Foundations of Cognitive Grammar. Volume I: Theoretical Prerequisites.
1991 Foundations of Cognitive Grammar. Volume II: Descriptive Application.
Lindner, Susan
1981 A Lexico-Semantic Analysis of English Verb Particle Constructions with
OUT and UP. Bloomington: The Indiana University Linguistics Club.
Lyons, John
1995 Linguistic Semantics: An Introduction. Cambridge: Cambridge University
Press.
Nerlich, Brigitte
forthcoming Synecdoche: a trope, a whole trope, and nothing but a trope. In N.
Norrick, A. Burkhardt (eds.), Tropic Truth. Amsterdam and Philadelphia:
Benjamins.
1978 The Pragmatics of Reference. Bloomington: The Indiana University Lin-
guistics Club.
Ogden, Charles K., Ivor A. Richards
1956 [1923] The Meaning of Meaning. London: Routledge and Kegan Paul.
Pustejovsky, James
1995 The Generative Lexicon. Cambridge, Massachusetts: The MIT Press.
Pustejovsky, James, Pierrette Bouillon
1995 Aspectual coercion and logical polysemy. Journal of Semantics 12: 133–
162.
120 Ken-ichi Seto
Reddy, Michael J.
1979 The conduit metaphor: a case of frame conflict in our language about
language. In A. Ortony (ed.), 1993. Metaphor and Thought. Second edition.
Cambridge: Cambridge University Press, 164–201.
Sorensen, Roy A.
1990 Process vagueness. Linguistics and Philosophy 13: 589–618.
Talmy, Leonard
1983 How language structures space. In H.L. Pick, Jr., L.P. Acredolo (eds.),
Spatial Orientation: Theory, Research, and Application. New York, etc.:
Plenum Press, 225–282.
Taylor, John R.
1995 Linguistic Categorization: Prototypes in Linguistic Theory. Second edi-
tion. New York: Oxford University Press.
Tversky, Barbara
1990 Where partonomies and taxonomies meet. In S.L. Tsohatzidis (ed.), Mean-
ings and Prototypes: Studies in Linguistic Categorization. London:
Routledge, 334–344.
Tversky, Barbara, Kathleen Hemenway
1984 Objects, parts, and categories. Journal of Experimental Psychology: Gen-
eral 113: 169–191.
Warren, Beatrice
1995 Distinguishing between metaphor and metonymy. In G. Melchers, B. War-
ren (eds.), Studies in Anglistics. Stockholm: Almquist and Wiksell Interna-
tional, 137–151.
Weinrich, Harald
1964 Typen der Gedächtnismetaphorik. Archiv für Begriffsgeschichte 9: 96–104.
Wells, Rulon S.
1977 Metonymy and misunderstanding: an aspect of language change. In R. W.
Cole (ed.), Current Issues in Linguistic Theory. Bloomington: Indiana
University Press, 195–214.
Wierzbicka, Anna
1984 Apples are not a kind of fruit: the semantics of human categorization.
American Ethnologist 11: 313–328.
Winston, Morton E., Roger Chaffin, Douglas Herrmann
1987 A taxonomy of part-whole relations. Cognitive Science 11: 417–444.
Aspects of Referential Metonymy
Beatrice Warren
Lund University
1. Introduction
One of the most significant effects of Grice’s seminal paper in 1975 was that it
highlighted the importance of implicitness in human communication. There is
now general acceptance that the principle of compositionality is not sufficient
to account for the interpretation of utterances. To specify the conditions which
allow the unsaid to be communicated has therefore become an essential task of
linguistic theory of today. The current interest in metonymy can be seen as
concerned with a particular aspect of this very task.
Some examples of metonymy discussed in more recent literature are
given in (1)–(6) below. (The metonymic expression is in italics; the interpreta-
tion taken to be intended is within square brackets.)
(1) It won’t happen while I still breathe. [live]
(Halliday 1994: 340)
(2) A: How did you get to the airport?
B: I waved down a taxi. [A taxi took me there]
(Gibbs 1994: 327)
(3) She turned pale. [was frightened]
(Kövecses, quoted from Ungerer and Schmid 1996: 132)
122 Beatrice Warren
(4) “Oh dear,” she giggled , “I’d quite forgotten!” [she said while
giggling]
(Goossens 1990: 328)
(5) A customer at a parking lot handing a car key to the attendant:
This is parked out there. [the car to which the key belongs]
(Nunberg 1996: 110)
(6) She has her father’s eyes. [eyes like those of her father]
(Warren, in print)
In all these examples the speaker feels confident that the hearer will elaborate
on what has been said or on a certain part of it in a certain direction and that the
hearer will take these elaborations to be part of the conveyed message. The
examples also have in common that that which triggers the inference is a
salient subpart of a state, situation or entity: in order to live one must breathe;
getting from A to B in a taxi involves first of all getting it to stop; paleness
tends to accompany fright; keys will go to some object with a lock, etc. Our
ability to access an entire state, situation or object from the mention of some
part is seen by some as the hallmark of metonymic thinking (cf. in particular
Gibbs 1994: 319). It is this type of thinking that the speaker intuitively knows
is common to all of us and which makes it possible for her/him to trust that her/
his message will be understood in the intended manner.
However, a second look at the examples in (1)–(6) reveals that they are
not quite equivalent from a conceptual point of view. For instance, breathing
is a condition for living, whereas paling is not a condition for, but a common
effect of, fear and keys are neither results of nor conditions for cars. Also from
a linguistic point of view, there are differences. Some are clearly propositional
(notably (2)); others are clearly referential ((5) and (6)). Above all, some
violate truth conditions ((4)–(6)); others do not ((1)–(3)). Therefore, although
I am prepared to accept the possibility that there is a common cognitive basis
for the italicized expressions in (1)–(6), I suggest that there are different types
of metonymy which have different constraints and which behave linguistically
in different ways. Making subdivisions of metonymy should have the advan-
tage of providing homogeneous sets of examples, which obviously makes it
easier to work out the processes according to which the examples in question
are formed and deciphered. In this spirit, the focus of the present contribution
is the particular type of metonymy which I will refer to as referential me-
tonymy.
For anything to qualify as a referential metonym, the following applies:

(i) it should have a referent,
(ii) the intended referent is not explicitly mentioned1 but its retrieval depends
on inference,
(iii) inference is made possible because there is some connection between the
mentioned referent (the trigger) and the implied referent (the target)2
deemed so well known that in the context in question the former will
automatically suggest the latter.
A study of recent sense developments (Warren 1992), a project involving
idioms, as well as randomly collected instances have provided me with a
corpus of 120 predominantly English examples which satisfy the above crite-
ria. The analysis of these has inspired the hypotheses that will be presented in
the following.
2. The basic semantic structure of referential metonyms
It is possible to see parallels as to semantic structure between referential

metonyms and other linguistic constructions, in particular onomatopoeic
words, eponyms, adjective-noun combinations, noun-noun compounds and
denominal verbs (cf. Warren 1995: 137–150). Here I will restrict the compari-
son to referential metonyms, noun-noun compounds and denominal verbs.
These constructions have in common that they involve two referents which are
connected by an implicit link. In noun-noun compounds, the two referents are
explicitly mentioned, one of which is the head, i.e., the referring item (chair in
armchair, for instance, because it is a kind of chair, not a kind of arm). In
denominal verbs, the noun contained in the verb (the parent noun) will be
connected implicitly to the grammatical object in the case of transitive verbs
(as in to shelve books ‘put books on shelves’). In the case of intransitive verbs,
the parent noun will be connected to the grammatical subject (as in the cow
calved ‘the cow produced a calf’) or an adverb (as in we holidayed in Paris
‘we spent our holiday in Paris’). In referential metonymy we have a referent
which is mentioned and one which is implied and an implicit link connecting
these (as in wagtail ‘something that has a wagging tail’). It is the implicit
referent that is the head and the referring item.
If we investigate the nature of the links in these constructions, we will
124 Beatrice Warren
Table 1. Recurrent links in compounds, metonymy and denominal verbs

(a) Composition and cross-classification3
SOURCE-RESULT
(i) silver spoon ‘spoon made of silver’
(ii) silver ‘something made of silver’ [cutlery]
SUBJECT-SET
(i) girlfriend ‘friend who is a girl’ [fiancée]
(ii) don ‘somebody who is a don’ [Mafia leader]
(iii) to chaperon sb ‘to be the chaperon of someone’
(b) Causation
RESULT-CAUSER
(i) tear gas ‘gas which produces tears’ [gas used in riots, etc.]
(ii) ecstasy ‘that which produces ecstasy’ [a drug]
(iii) to calve ‘to produce a calf’ [to give birth to a calf]
CAUSER-RESULT
(i) hay fever ‘fever that is produced by hay’ [allergic catarrh]
(ii) hand (n.) ‘that which is produced by the hand’ [aid/applause]
(iii) to butcher sth ‘to do that which is done by butchers’ [kill to supply meat]
(c) Possession
PART-WHOLE
(i) armchair ‘chair which has “arms”’ [chair with armrests]
(ii) wagtail ‘something which has a wagging tail’ [bird]
(iii) to clothe children ‘to cause children to have clothes’
WHOLE-PART
(i) house roof ‘roof belonging to a house’
(ii) red pen ‘pen with something red in it’ [ink]
(d) Location in time and space
PLACE-OBJECT
(i) sea port ‘port which is at the sea’
(ii) walkabout ‘place where “walkabouts” are taken’ [road]
OBJECT-PLACE
(i) foxhole ‘hole which has foxes in it’ [abode of foxes]
(ii) date ‘that which occurs at a date’ [appointment]
(iii) to shelve books ‘to put books on shelves’
(e) Resemblance
COMPARANT-COMPARED
(i) clubfoot ‘foot which is like a club’ [talipes]
(ii) her father’s eyes ‘eyes which are like those of her father’
(iii) to mother sb ‘to behave like a mother does towards sb’
notice certain recurrent patterns involving composition and cross-classifica-

tion, causation, possession, location and resemblance. Some representative
examples are offered in Table 1. The first examples illustrate a compound noun
(i), the second a referential metonymy (ii), and the third a denominal verb (iii).
These recurrent patterns have been described in a number of studies. For
English compounds, cf., e.g., Bergsten (1911), Levi (1978), Warren (1978)
and Mellenius (1997); for denominal verbs, cf. in particular Clark and Clark
(1979), Leitner (1974) and Dirven (this volume). Metonymic patterns have
also been described, but in terms such as PART FOR WHOLE, WHOLE FOR PART,
CONTAINER FOR CONTENTS, GARMENT FOR PERSON, PLACE FOR INSTITUTION,
PRODUCER FOR PRODUCT, INSTRUMENT FOR RESULT, MATTER FOR ARTIFACT, etc.
For a detailed summary of metonymic patterns discussed in the literature, cf.
Nerlich et al. (this volume), Leisi (1985), Lipka (1988) and Radden and
Kövecses (this volume).
Table 1 requires some comments. First of all, it should be pointed out that
the paraphrases within single quotation marks are not meant to represent the
meanings of the examples. These are given within square brackets. My
hypothesis is that expressions of the type exemplified above are partially
implicit descriptions of that which the speaker wishes to denote. The para-
phrases attempt to represent the type of reconstruction that the speaker expects
the hearer to make in order to be able to match the expression with the
intended referent. Meanings, in the sense of definitions, are created by the
perception of class-distinctive features, which in turn requires that the referent
in question is seen as one of a kind. This explains inter alia why most noun-
noun compounds are not compositional. Not all holes that have foxes in them
are foxholes, for instance. As we will see, unlike compounds and denominal
verbs, referential metonyms do not, however, supply ‘names’ for the unnamed
very frequently.
Secondly, for reasons of clarity the table leaves out many details. Above
all, it ignores some relations. The justification for this has been that they could
possibly be considered as subtypes of those mentioned. The Causation rela-
tion, for example, subsumes the Purpose relation, alias GOAL-INSTRUMENT
pattern (as exemplified in ball bat, to knife somebody), the ACTIVITY-ACTOR
pattern (as in businessman) and the ORIGIN-OBJECT pattern, which is common
in reversative denominal verbs (dust the table (OBJECT-ORIGIN)) and pod peas
(ORIGIN-OBJECT)). The Resemblance relation subsumes a relation paraphrase-
able with the verb represent as in Alice was Cinderella in the play (Cinderella
‘the person who represents Cinderella’).
126 Beatrice Warren
Finally, as just pointed out, the relations described are not restricted to
compounds, denominal verbs and referential metonymy, but occur also in
other constructions. An important point to make in this connection is that not
all relations occur in all constructions. There appear to be constraints, some of
which are natural, whereas others are more intriguing. This, however, will
have to be the topic of another paper.
These omissions might be excused since the main purposes of Table 1 are
to reveal the basic semantic structure of referential metonyms and its striking
similarity to some other linguistic constructions. It is also meant to bring out
the similarity between the relations described and case endings, such as
essives, ergatives, instrumentals, possessive genetives, locatives, ablatives,
and consequently, deep case roles. Consider AGENT, INSTRUMENT, RESULT,
SOURCE, LOCATIVE and EXPERIENCER (a type of container or possessor). If the
interpretation of sentences involves assigning roles, i.e., working out unex-
pressed relations — as most linguists assume — then it is perhaps not very
surprising that we should find these same types of relations also in phrases and
other constructions involving implicit relations.
The problem facing linguists working with case roles is that it appears
impossible to give an exhaustive list of these. The same problem faces us here.
Although the 120 examples in my corpus can be taken to be based on one or
the other of the specified relations, arguably there are counterexamples.
Consider Nixon bombed Hanoi. The most natural papraphrase for the
metonym Nixon here appears to be ‘those who Nixon controlled or could
activate,’ rather than ‘those who Nixon represents.’ Consider also the much
quoted example The ham sandwich is waiting for his check and examples such
as The appendicitis is in ward 8 (cf. Dirven, this volume), which appear to be
counterexamples, although with some stretching it could also be maintained
that these are based on types of possession relations.4
Nevertheless, the regularity and ease with which referential metonyms
involving composition, causation, possession, etc. are formed and interpreted
and the difficulty of finding convincing counterexamples warrant the sugges-
tion that the relations in Table 1 are default relations. (This suggestion is also
in line with Langacker’s solution to the problem of case roles. That is, there
are certain role archetypes (cf. Langacker 1991: 236).) This would also
explain why they tend to be grammaticalized. In fact, we may reverse the
argument: if there are certain relations which are so frequently expressed that
they tend to be grammaticalized, they ought to be predominant. The reason
why these relations recur so regularly in linguistic constructions may be that

they are cognitively particularly salient. Psychologists and cognitive linguists
have indeed pointed out the great significance of ‘gestalt’ in human percep-
tion, which involves discerning bounded entities and phenomena and their
parts, constituent matter or position within a whole or an area, and they have
pointed out that such distinctions are reflected in the grammars of languages
(cf. in particular Langacker 1991). The importance of being able to establish
cause-effect relations among intelligent beings is also evident, as is the role of
causation in grammar (consider, for instance, ergative and instrumental cases,
or the fact that if there is an AGENT it is promoted to be the subject, or causative
verbs and conjunctions of cause, means and result). The resemblance relation,
however, is different. Whereas links of composition, possession, location and
causation presuppose co-occurrence or contiguity, links of resemblance are
not dependent on closeness in time and space. This difference has significant
consequences and probably explains the fact that, if there is a resemblance
relation between trigger and target, it has traditionally been thought of as the
basis of a different figure of speech, viz. metaphor. I will return to this issue in
the final section of the article. My immediate concern, however, will be the
function of referential metonyms.
3. The rationale of referential metonyms
As has just been pointed out, I see strong parallels between noun-noun
compounds and referential metonyms. Both constructions involve at least two
referents,5 and both involve an implicit link. The difference is that the refer-
ring item is explicit in compounds, whereas it is implied in metonyms.
Consider a few examples:
(7) I am bugged [the place I am at]
(8) We are just across the river
[the place we live/work in]
(9) Are you in the printer?6 [that which you have produced, i.e.,
your paper]
(10) The bathtub is running over [that which is in the tub, i.e., the wa-
ter]
(11) Answer the door! [the person at the door]
(12) I like Mozart. [that which Mozart has produced,
i.e., his music]
128 Beatrice Warren
The analyses of these examples are meant to demonstrate that referential

metonyms are basically abbreviations. What is left out is something that ‘goes
without saying’ in the context in question, but more importantly, that which is
retained has, in its particular context, the greatest information value. In the
examples above, it identifies the referent in question. It is important that we
realize that the traditional definition of metonymy, viz. “substituting for the
name of a thing the name of an attribute of it or something closely related”
(OED) is not correct in that no substitution is necessarily involved. We do not
refer to music in I like Mozart, but to music composed by Mozart; we do not
refer to water in The bathtub is running over, but to the water in the bathtub.
I have so far concentrated on the interpretation of the metonym itself and
ignored the interpretation of the utterance containing it. Also in the following
I will restrict my attention in this way, but will here at least mention the
distinction. Consider I am bugged in the sense of its paraphrase The place I am
staying at is bugged.7 Both describe the same type of situation and have the
same truth conditions. Yet their interpretations are not quite equivalent. In I
am bugged what is conveyed is not only that the place where the speaker
happens to be is bugged, but that this is something that is of particular
relevance to the speaker. Nunberg refers to this phenomenon as predicate
transfer. My explanation for this effect of metonymy is that giving the non-
referring part of the metonym the status of the grammatical subject makes it
the topic of the utterance, that which the comment is about. In other words, the
abbreviation may not only serve the purpose of saving words, but may also
serve to achieve the information structure intended by the speaker.
The rationale of metonyms in (7)–(12) then appears to be different from
that of compounds, which are mainly formed to name that for which we lack a
name. This is not to deny that there are metonyms that can be lexicalized and
become ‘names.’ Consider ecstasy, will (as in last will and testimony), glass,
box, etc. There are also examples of metonyms which have rhetorical force as
in she married money, the way of all flesh, heartthrob (for attractive girl) and
as in the headline Brains of British Museum crack Crystal Skull Riddle (the
last example is from Nerlich et al., this volume).
It is naturally these latter types of metonyms that attract our attention,
whereas the types exemplified in (7)–(12) tend to go unnoticed. In fact, it is
easy to overlook such examples of metonymy. Consider the metonymic shift
involved in the metalinguistic uses of words. That is to say, the fact that the
metalinguistic name for a word coincides in form with this word renders it a
natural metonym, possibly based on a compositional link. It is, in other words,

natural to refer to the object language word cat with the metalinguistic name
cat (as I just did), but few people think of this as metonymy.
Although referential metonymy has a potential as a naming and/or rhe-
torical device and has mainly been discussed in this capacity, it should not be
forgotten that it is a kind of semantic-syntactic construction, comparable to,
for instance, compounding and genitive constructions, which the language
user has to learn to interpret and form to reach full competence.8
4. ANTECEDENT-CONSEQUENT links and referential metonymy
One consequence of looking upon metonymy as basically a type of abbrevia-

tion is that utterances containing referential metonyms will be non-literal. This
is so, since the predication will superficially apply to that which is retained,
which, I remind the reader, is not the referring part of the metonym. However,
our interpretation shows that the predication applies in fact to the implicit
target as well as the trigger. Hence in (10) it is not the tub that is running over,
but the water in the tub.
If we now reconsider our very first example (renumbered as (13)):
(13) It won’t happen while I still breathe. [live]
(Halliday 1995: 340)
we will notice that there is no violation of truth conditions and also that it is
difficult to think of a paraphrase in line with the ones given in Table 1. Instead
the natural paraphrase connecting breathing with living is: if somebody
breathes, s/he will live. That is to say, we find an ANTECEDENT-CONSEQUENT
relation and we find that we have connected not entity to entity, but proposi-
tion to proposition. The fact that we have here an ANTECEDENT-CONSEQUENT
relation between trigger (breathe) and target ([live]) is consonant with the fact
that there is no violation of truth conditions in (13), since the truth of the
consequent depends on the truth of the antecedent.
Whereas in the examples of metonyms in Table 1 and in (7)–(12) the
trigger represents the modifier and the target the head which together pick out
a referent, breathe in (13) could not be looked upon as the modifier of [live].
Instead we have an expression which suggests two co-ordinate notions:
breathing and therefore living. Consider (14) and (15): (14) is an instance of
130 Beatrice Warren
referential metonymy; ecstasy could not be taken to refer to both an emotion

and a drug; (15), by contrast, involves an instance of propositional metonymy,
as a result dark does suggest both ‘not light’ as well as ‘closed.’
(14) He uses ecstasy. [the drug]
(15) The shops are dark [closed] on Sundays in this town.
ANTECEDENT-CONSEQUENT relations qualify as instances of concomitance rela-
tions, as do the equative, compositional, possessive, causal and locative
relations exemplified in Table 1. Concomitance relations are traditionally
considered to be criterial for metonymy. I suggest, however, that we make a
clear distinction between these two types of concomitance relations. ANTE-
CEDENT-CONSEQUENT relations give rise to propositional metonymy; the links
specified in Table 1 give rise to referential metonymy. The differences as to
paraphrasing, as to truth values and as to the relation between trigger and
target (i.e., subordination in one case and co-ordination in the other) indicate
that we are dealing with two distinct, although related, cognitive operations.
5. The resemblance relation and referential metonymy
As I have already mentioned, the difference between metaphor and metonymy

is traditionally said to be that metaphor is based on resemblance relations,
whereas metonymy is based on contiguity. I have also mentioned that resem-
blance relations are not dependent on closeness in time or space. What is
instead required is that at least one attribute of the target is perceived as
reminiscent of an attribute of the trigger. Another requirement for the resem-
blance relation in metaphor is that target and trigger are members of different
categories (or domains, as is the term that cognitive linguists would use). Now
let us reconsider the example given in Table 1 as an instance of referential
metonymy based on resemblance.
(16) Ann has her father’s eyes. [eyes like those of her father]
It is possible to interpret the italicized phrase without knowing in what respect
Ann’s eyes are like her father’s. Therefore, although I accept that resemblance
is normally involved in metaphor I suggest that, provided it is solely the
relation that takes us from the trigger to the target, it forms the basis of
metonymy. Normally, however, if there is a resemblance relation between
trigger and target, the hearer is expected to retrieve the feature or features that
form the basis of the similarity. Consider the interpretation of (17):
(17) Mary is the Cinderella in the family.
This involves working out some such features as ‘unjustifiably neglected’ and
possibly also ‘a person with qualities as yet not appreciated but which are
superior to her siblings (or the equivalent) and which will eventually raise her
above them in status.’ Compare this interpretation to that of Cinderella in (18),
which simply involves working out that Mary is the person representing
Cinderella in the play.
(18) Mary is Cinderella in the play. [the person representing Cinderella]
My suggestion concerning the crucial difference between referential me-
tonymy and metaphor is, therefore, that in the case of referential metonymy
the link between trigger and target is a relation (and one relation only),
whereas in the case of metaphor, it involves one or more attributes.
Working out several links (i.e., similar attributes) between trigger and
target may cause the interpretation to be much richer, but also less straightfor-
ward than is the case of metonymy. Indeed interpretations of metaphors often
vary with the interpreter, probably to a much greater extent than in the case of
metonymy. This indefiniteness may in turn make metaphors enigmatic and
interesting in a way that is not so common in the case of metonymy.
Perhaps the most important difference between metaphor and metonymy
is that metaphors often involve hypothetical thinking. When we interpret the
metaphor in (19), for instance, we see information as if it were a fluid seeping
through a container supposed to hold it (the White House).
(19) Information about the matter leaked from the White House.
Metonymy does not seem to involve hypotheses. Perhaps we can explain this
difference by pointing out that metonymy is based on relations which presup-
pose actual coincidence, whereas metaphor, which involves finding a match
for an attribute among all the mentally stored attributes, is freed from con-
straints of what could actually occur or coincide.
132 Beatrice Warren
6. Conclusion
I have argued that, if our aim is to gain as precise an understanding as possible

of the manner in which instances of metonymy are formed and interpreted, we
should distinguish between different types of metonymy. Referential me-
tonymy was singled out as the focus of the present contribution because of its
similarity to, above all, noun-noun compounds. Both involve two nominal
expressions, one of which is the modifier and the other the head and referring
item. Also, in both cases, there is an implicit link. This led to the suggestion
that there is a common basic structure of referential metonymy as displayed in
Table 2:
Table 2. Common basic structure of referential metonymy
Implicit referent Implicit link Explicit expression

(target and head) (trigger)
that which consists of silver
causes ecstasy
is in bathtub
etc.
The speaker expects the hearer to be able to fill in that which is left implicit,
which will result in a description of the intended referent.
I have also suggested that metonymic links are not quite as ad hoc as the
much quoted example The ham sandwich is waiting for his check may
suggest, but that there is a limited set of default relations which are of great
linguistic relevance (cf. Radden and Kövecses, this volume). This is evi-
denced by the fact that they are not restricted to metonymy, but are part of a set
of unexpressed relations that are important in the semantics of phrases and
sentences.
The illustration above also reveals that I look upon referential metonymy
as a kind of abbreviation having potentials as a naming and/or rhetorical
device.
I further argued that we should make a distinction between concomitance
relations between propositions (i.e., ANTECEDENT-CONSEQUENT relations) and
concomitance relations between entities (i.e., relations such possession, loca-
tion, causation, etc.). The former type of relation may be the basis of what is
sometimes referred to as pragmatic meanings of words which do not cause
violations of truth conditions and which may be cancellable. The latter type of
relation may be the basis of non-literal uses of (primarily) nouns. I have

suggested that the term referential metonymy should be restricted to sense
shifts based on the latter kind of relation.9
Finally, I suggested that the basic difference between metonymy and
metaphor is that the interpretation of metonyms involves retrieving a relation,
whereas the interpretation of metaphors involves retrieving at least one at-
tribute shared by the conventional and intended referents. This means that
even a resemblance relation can occasionally be metonymic, as in the phrase
her father’s eyes in Ann has her father’s eyes.
Notes
1. By “not explicitly mentioned” I mean that there is no morpheme which could be taken to
represent it. Printer, for instance, would not be considered a metonym, since the suffix -
er can be taken to formally — although opaquely — indicate the nominal status of the
expression, reflected in the interpretation ‘that which prints.’
2. The terms ‘trigger’ and ‘target’ are adopted from Fauconnier (1985).
3. These are the links involved in turning count nouns into non-count nouns and vice versa,
as in there is egg on the knife and two coffees, please. The former process is sometimes
referred to as ‘grinding’ and the latter as ‘chunking.’ Cf. Copestake and Briscoe (1996).
4. Nerlich et al. (this volume) see an unconventional metonymic relation in the compound
Treacle people juice, formed by her son Matthew. The connection between Treacle
people and juice is that Matthew was served a particular juice after having watched a TV
play called “The Treacle People.” In my view, the connection fits quite well with the
OBJECT -TIME pattern: ‘that which (practically) co-occurred with The Treacle People.’ The
expression is strange, however, in that the play will hardly affect the character of the
drink, i.e., be a salient subpart of it.
5. The basic structure of both compounds and metonyms can be expanded. Consider
compound within compounds (cowboy hat) and metaphors within metonyms or
metonyms within metaphors as in yellow belly in the sense ‘a native of the fens’: ‘sb who
is like sb who has a yellow belly (i.e., a frog).’ Cf. Warren (1992: 94–97).
6. This is a translation from Swedish, i.e., from Är det du som ligger i skrivaren?
7. There are, of course, other possible interpretations of (7), viz. ‘I am irritated’ or ‘I have
bugs on me.’
8. An increasing number of linguists do, however, approach metonymy in this spirit. Cf. in
particular Jackendoff (1997), Nunberg (1996), Fass (1997) and Copestake and Briscoe
(1996).
9. A terminological distinction of this kind is not infrequently made. Consider, for instance,
Stern (1965), who refers to sense shifts developed from the former type of relation as
permutations.
134 Beatrice Warren
References
Bergsten, Nils
1911 A Study of Compound Substantives in English. Uppsala: Almquist and
Wiksell.
Clark, Eve V., Herbert H. Clark
Copestake, Ann, Ted Briscoe
1996 Semi-productive polysemy and sense extension. In J. Pustejovsky, B.
Boguraev (eds.), Lexical Semantics: The Problem of Polysemy. Oxford:
Clarendon Press, 15–67.
Fass, Dan
1997 Processing Metonymy and Metaphor. Greenwich, Connecticut, etc.: Ablex
Publishing Corporation.
Fauconnier, Gilles
1985 Mental Spaces. Cambridge, Massachusetts: The MIT Press.
Gibbs, Raymond W.
Goossens, Louis
Grice, H. Paul
1975 Logic and conversation. In P. Cole, J. Morgan (eds.), Syntax and Semantics.
Volume 3: Speech Acts. New York: Academic Press, 41–58.
Halliday, Michael A.K.
1994 An Introduction to Functional Grammar. London, Melbourne, Auckland:
Edward Arnold.
Jackendoff, Ray
1997 The Architecture of the Language Faculty. Cambridge, Massachusetts: The
MIT Press.
1991 Concept, Image and Symbol. Berlin, New York: Mouton de Gruyter.
Leisi, Ernst
1985 Praxis der Englischen Semantik. Heidelberg: Winter.
Leitner, Gerhard
1974 Denominale Verbalisierung im Englischen. Tübingen: Niemeyer.
Levi, Judith
1978 The Syntax and Semantics of Complex Nominals. New York, San Fran-
cisco, London: Academic Press.
Lipka, Leonhard
1988 A rose is a rose is a rose: on simple and dual categorization in natural
languages. In W. Hüllen, R. Schulze (eds.), Understanding the Lexicon.
Tübingen: Niemeyer, 355–366.
Mellenius, Ingmarie
1997 The Acquisition of Nominal Compounding in Swedish. Lund: Lund Univer-
sity Press
Nunberg, Geoffrey
1996 Transfer of meaning. In J. Pustejovsky, B. Boguraev (eds.), Lexical Seman-
tics: The Problem of Polysemy. Oxford: Clarendon Press, 107–132.
OED
1989 The Oxford English Dictionary. Second edition. Oxford: Clarendon Press
Stern, Gustaf
1965 [1931] Meaning and Change of Meaning. Bloomington: Indiana Univer-
sity Press.
Ungerer, Friedrich, Hans-Jörg Schmid
1996 An Introduction to Cognitive Linguistics. London, New York: Longman.
Warren, Beatrice
1978 Semantic Patterns of Noun-Noun Compounds. Gothenburg Studies in En-
glish 41.
1992 Sense Developments. Stockholm: Almquist and Wiksell International.
1995 Distinguishing between metaphor and metonymy. In G. Melchers, B. War-
ren (eds.), Studies in Anglistics. Stockholm: Almquist and Wiksell Interna-
tional, 137–151.
1998 What is metonymy? In M. Hogg, L. van Bergen (eds.), Historical Linguis-
tics 1995. Amsterdam and Philadelphia: Benjamins, 301-310.
Part II
Historical Aspects of Metonymy

Frame and Contiguity
On the Cognitive Bases of Metonymy and

Certain Types of Word Formation
Peter Koch
University of Tübingen
1. Introduction1
It goes without saying that in a rhetorical perspective metaphor and metonymy

are usually considered to be close relatives of each other. And in fact, they
often get confused, not only by students, but — as to the application of these
two theoretical notions — even by scholars. In general, it is metonymies that
are reduced to metaphors. All in all, metonymy appears to be a kind of parente
pauvre of metaphor (cf. Bredin 1984: 45): metonymy seems less interesting,
less abstract and is supposed to demand a relatively minor intellectual effort.
Nevertheless, I am convinced that metonymy occurs much more frequently
than metaphor and tells us a great deal about our cognitive equipment. So it is
worthwhile studying its cognitive bases in some detail.
2. Ad hoc metonymy and metonymic polysemy
First of all, let me clarify a terminological problem that constantly arises in

connection with metonymy. On the one hand, the term ‘metonymy’ denotes a
rhetorical trope which is applied ad hoc to certain lexical material; compare
for instance:
140 Peter Koch
(1) Lat. iam proximus ardet Ucalegon

‘nearby, [the house of] Ucalegon is already burning’ [my
translation] (Verg. Aen. 2, 311f, cit. Lausberg 1973: § 568,
1c; my italics)
On the other hand, the term ‘metonymy’ is used to denote a specific semantic
relation between two senses of a given polysemous lexeme, as for instance:
(2) Eng. bar <sense1> ‘......’
<sense2> ‘......’
.......... ..........
.......... ..........
<sensei> ‘counter’
<sensej> ‘public house’ (metonymic sense in
relation to <sensei>)
To keep things clear, it would be better to call cases like Eng. bar (with respect
to its sensesi,j) ‘metonymic polysemy.’ Nevertheless, there is a close relation-
ship between ad hoc metonymy as illustrated in (1) and metonymic polysemy
as illustrated in (2): just as polysemy in general develops through ‘lexicaliza-
tion’ of a semantic change ultimately triggered by a contextual and/or expres-
sive ad hoc use of a word in discourse,2 metonymic polysemy also develops
through lexicalization out of ad hoc metonymic usage in discourse — or put
the other way around, metonymy constitutes an ad hoc innovation that can
potentially induce a ‘metonymic change’ in the meaning of the lexeme con-
cerned, which thereby becomes (metonymically) polysemous. I will come
back to this issue in Section 9.
3. Metonymy and contiguity
We have to look for the semantic basis of ad hoc metonymy in order to

account for metonymic polysemy as well. Traditionally, metonymy (and
especially ad hoc metonymy) belongs to the realm of rhetoric. To my knowl-
edge, the earliest definition of metonymy (Lat. denominatio) as a rhetorical
trope in its own right is to be found in the Rhetorica ad Herennium:
(3) Denominatio est, quae ab rebus propinquis et finitimis trahit
orationem, qua possit intellegi res, quae non suo vocabulo sit
appellata.
Frame and Contiguity 141
‘Denominatio [i.e., ‘metonymy’] is a trope that takes its expression

from near and close things and by which we can comprehend a
thing that is not denominated by its proper word.’ [my translation]
(Her. IV: 32, 43 = Anonymous 1894: 337; my italics).
By res propinquae et finitimae, the author of the Rhetorica ad Herennium
virtually introduces an associative or cognitive element into his definition,
namely the relation of closeness that might remind us of the σýεγγυσ,
constituting one of the three relations Aristotle sets up to account for the
process of remembering:
(4) διN καM τN φεξ& ησ θηρεýοµεν νοÞσαντεσ πN το& υ ν&
υν n eλλου
τινüσ, καM φ\µοßου n ναντßου n το& υ σýνεγγυσ. διJ το&
υτο
γßγνεται νÜµνησισ.
‘So we track down the sequence [of our ideas] by starting from the
present moment or from something else and from something simi-
lar or opposite or close [to it]. That is the way remembering comes
about.’ [my translation] (Aristotle, De memoria et reminiscentia,
451b: 18–22 = Aristotle 1975: 300; my italics).
In view of the fact that the theory of rhetorical tropes and the theory of
remembering are, prima facie, two very different fields of investigation, the
author of the Rhetorica ad Herennium was probably not in any way influ-
enced by Aristotle. And that is the way the story continued for a long time.
On the one hand, we have the classical rhetorical conception of me-
tonymy, more or less clearly involving a relation of closeness (or merely a
constant relation of some kind). In the form of the notion of κατÜχρησισ, the
rhetorical theory offers a device for coping even with lexicalized metonymic
(or synecdochical or metaphorical) change of meaning (cf. Lausberg 1973:
§§562 and 577; for metaphorical κατÜχρησισ see also Koch 1994: 203f, 207).
Not very surprisingly, several 19th century historical linguists (Reisig, Bréal,
Paul, Darmesteter, Nyrop, etc.) took up the rhetorical basis for describing
metonymic (as well as other types of) change of meaning (cf. Nerlich 1992;
Blank 1997a: 10–18). On the other hand, we have the modern thread of
English sensualist philosophy and associationist psychology (Hobbes, Locke,
Hume, James and John St. Mill, Bain, etc.) relying heavily on the relation of
‘contiguity’ and, to a lesser degree, on that of ‘similarity.’ This tradition does
not go directly back to Aristotle’s views presented in (4) — despite partly
analogous distinctions.3
142 Peter Koch
A first resolute attempt to bridge the by now traditional associationist

relations and the linguistic classification of semantic change (Wundt 1912:
459–627) unfortunately did not clearly outline just metonymy.4
Meanwhile, Kruszewski had applied the two associationist relations of
similarity and contiguity directly to the functioning of linguistic entities and,
thereby, prepared Saussure’s (synchronic) two-axes theory, which opposed
‘associative’ relations (i.e., those formed by pure similarity, in absentia) and
‘syntagmatic’ relations (i.e., those formed by contiguity, in praesentia) be-
tween linguistic entities (cf. Kruszewski 1884–90; Saussure 1916: 170–180;
Holenstein 1975: 143–145; Happ 1985: 37–52).
It was only Léonce Roudet (1921: 686–692), who, in a most straightfor-
ward account, took up associationist relations and Saussure’s two axes and
thereby established the systematics of semantic change as illustrated in Table
1.
Table 1. Systematics of semantic change (Leónce Roudet)

ideas words
(syntagmatic)
contiguity change by metonymy change by ellipsis
(associative)
similarity change by metaphor change by
analogical ‘irradiation’
To my knowledge, it is here for the first time that the cognitive component of
metonymy is explicitly stated in terms of associationist psychology (and in
accordance with the classical rhetorical tradition):
(5) Changements résultant d’une association par contiguïté entre les
idées. Tous les changements que l’on a appelés changements par
connexité ou par métonymie (devenue inconsciente) appartiennent
à cette catégorie. (Roudet 1921: 690)
Take our example (2). In a public house we normally find a counter across
which liquor and food are served. Thus, there is a spatial contiguity between
the ideas COUNTER and PUBLIC HOUSE. Their contiguity must have triggered the
metonymic change of meaning (from ‘counter’ to ‘public house’) that under-
lies the metonymic polysemy of Eng. bar ‘counter; public house.’
Now let us consider another example (cf. DHLF, s.v. prison).
(6) OFr. la prison ‘act of seizing,’ hence: ‘captivity,’ hence:

‘prison,’ hence: (MFr.) ‘penalty of imprisonment’
Between ACT OF SEIZING and CAPTIVITY there is a temporal and a causal
contiguity; and between CAPTIVITY and PRISON there is a spatial contiguity, as
there is also between PRISON and PENALTY OF IMPRISONMENT. These contiguity
relations are the cognitive basis of the corresponding metonymic changes (and
of the resulting synchronic metonymic polysemies we observe in OFr. la
prison ‘act of seizing; captivity; prison’ and MFr. la prison ‘prison; penalty of
imprisonment’).
Roudet’s diachronic considerations were not paid much attention to in the
discussions of the following decades, which concentrated on synchronic and
theoretical issues. A far more influential statement, including thoughts on the
functioning of metonymy in general (but not on metonymic change!), is due to
Roman Jakobson (1956), who, inspired by Kruszewski and Saussure, put
forward his two-axes theory with far-reaching overall correlations, paralleling
contiguity – syntagmatic axis – combination – Broca aphasia – metonymy on
the one hand, and similarity – paradigmatic axis – selection – Wernicke
aphasia – metaphor on the other hand.5 The oversimplifications inherent in
these correlations have been criticized on several grounds (cf. the criticism in
Holenstein 1974; Happ 1985: 12–17, 61–93, 127–139; Blank, this volume).
One point, however, has been differentiated in passing by Jakobson himself.
When interpreting different reactions to the stimulus hut in a verbal associa-
tion test, he considers neither ‘contiguity’ and ‘syntagmatic relation’ nor
‘similarity’ and ‘paradigmatic relation’ as coextensive, as shown in Table 2.6
Table 2. Association test: stimulus ‘hut’ (Roman Jakobson)

‘positional’ similarity ‘positional’ contiguity
(PARADIGMATIC relation) (SYNTAGMATIC relation)
‘semantic’ metonymy
CONTIGUITY (→ thatch, litter, poverty) (→ ... burnt out)
‘semantic’ tautology (→ hut) (→ ... is a poor little house)
SIMILARITY synonymy (→ cabin, hovel)
(CONTRAST) antonymy (→ palace)
metaphor (→ den, burrow)
In this way, Jakobson brings together the rhetorical-poetic notion of ‘me-

tonymy’ and the semantic-associative relation of ‘contiguity.’
144 Peter Koch
From a psychological and anthropological perspective, contiguity (and

similarity/contrast) seem to be absolutely fundamental associative relation-
ships (cf. also Raible 1981). But there remain some problematic points,
especially with respect to language. First, a ‘semiotic’ problem: if we apply
contiguity to metonymy, that is, to word semantics, which semiotic level do
we have to aim at? This problem will be dealt with in Section 4. Second, a
‘psychological’ problem: is contiguity a well-defined relationship and what is
it like? This question will be taken up in Section 5.
4. Contiguity as a conceptual relation
Let us begin with the semiotic problem. As we have already seen (cf. quota-
tion (5)), according to Roudet, metonymy is based on a contiguity between
ideas, that is — in more up-to-date terms — ‘conceptual’ contiguity. In
contrast to this, Table 2 reveals that Jakobson’s authoritative two-axes theory
does not deal with concepts, but with linguistic ‘signs.’ So in Jakobson’s
approach, metonymy seems to presuppose a (semantic) contiguity between
linguistic signs. Similarly, in the period following, many scholars accounted
for metonymy and metonymic change in terms of contiguity between the
‘senses’ of two ‘words.’
Consider for instance Ullmann’s systematics of the ‘essence’ of semantic
change, which only superficially seems identical to Roudet’s system (Table
1), by which it has indeed been inspired (cf. Ullmann 1962: 211–227; Blank
1997a: 19f, 35–38).
Table 3. ‘Essence’ of semantic change (Stephen Ullmann)

sense name
contiguity metonymy ellipsis
similarity metaphor popular etymology
As ‘sense’ probably corresponds to the Saussurian signifié, we would have to

explain metonymic change by contiguity between linguistic meanings, which
is not easy to imagine in abstracto. On a more technical level, contiguity
between ‘sememes’ of words has been described in terms of operations on sets
of semantic features, when treating, for example, rhetorical metonymies or
metonymic polysemy (cf. Dubois et al. 1970: 106ff; Martin 1992: 75–86).
According to this approach, in our polysemous example (2), we would have to

posit a semantic feature [counter] as figuring in the set of features that form the
sememe ‘public house’ of the word Eng. bar.
Things get even more complicated when we have to describe cases of
metonymic change in which the original meaning has not survived — a
situation that can occur, sooner or later, with any semantic change (cf. REW,
s.v. focus):
(7) CLat. focus ‘fireplace’ > VLat. focus ‘fire’ (cf. Fr. feu; It. fuoco;
Sp. fuego, etc.)
Would we have to posit, in this case, a feature [fireplace] as still functioning in
the sememic description of Fr. feu, etc.?
Obviously, this cannot be a realistic approach to metonymy. Following this
line of inquiry, we would be compelled to integrate into our linguistic descrip-
tion of a given lexical item all the information necessary to explain whatever
metonymy may occur in this lexical item in the course of — even future —
language history. In most cases of metonymy, I think, this method would yield
a far too powerful and at the same time an ‘overbred’ linguistic description.
Moreover, an intralinguistic solution for contiguity seems inappropriate
from the outset. The metonymy Eng. bar ‘counter; public house’ is possible
thanks to our knowledge of public houses and counters and not thanks to our
knowledge of the word bar. Similarly, the metonymy Lat. focus ‘fireplace;
fire’ was possible thanks to the Romans’ knowledge of fireplaces and fire. It is
not our knowledge of words (and their semantic features), but our knowledge
of the world that determines contiguities. So metonymy is not a problem of
linguistic structure, but a problem that concerns the relation between language
and the extralinguistic world. Contiguity has to be considered as constituting a
conceptual, extralinguistic and not an intralinguistic relationship (see Le
Guern 1973: 14, 25; Bredin 1984: 52f; Happ 1985: 129; Koch 1991: 284,
1993: 281; Blank 1993: 37, and this volume; also the notion cotopie sémio-
tique developed by Bonhomme 1987: 46).
5. Contiguity and frames
I now come to the psychological problem involved in contiguity relationships.

According to Umberto Eco (1984: 147), ‘contiguity’ is a wishy-washy notion:
146 Peter Koch
“[...] la contiguità è concetto abbastanza sfumato [...]” (cf. also Weinrich

1987: 107). Bredin (1984: 47) calls it “a convenient but unrevealing meta-
phor.” Originally, the term ‘contiguity’ (like Greek σýνεγγυσ in quotation
(4)) belongs to the conceptual domain of space. Applying it to other concep-
tual relations (like the temporal or causal one between ACT OF SEIZING and
CAPTIVITY in (6)), seems to involve a metaphor on the metaconceptual level:
by choosing exactly this term, we conceptualize, on the metaconceptual level,
different types of conceptual contiguity in terms of spatial contiguity. But is
this metaphor legitimate on psychological grounds? Or do we run the risk of
unduly spatializing cognitive relations of any kind?7
In my opinion, we can retain the term ‘contiguity’ because it actually
covers a coherent range of cognitive relations. But we are not bound to retain
the underlying spatial metaphor. An alternative model accounting for contigu-
ity as a unitary kind of relationship is the ‘frame’ model, elaborated in
cognitive psychology and linguistics in the last two decades (cf., for example,
Minsky 1975; Fillmore 1975, 1985; Tannen 1979; Barsalou 1992; Cordier
1993: 143–147; cf. also the notion ‘schema’ in Bartlett 1932: 197–214, 300–
304, 311–314). By recurring to frames, we can easily understand metonymic
phenomena because frames — and this is a point I would like to stress — are
non-linguistic, conceptual wholes. When acknowledging the latter fact, we do
not have to overproliferate linguistic-semantic descriptions only for the sake
of metonymies.
Contiguity is the relation that exists between elements of a frame or
between the frame as a whole and its elements. Thus, PUBLIC HOUSE constitutes
a frame (Figure 1), one of whose elements is the COUNTER (cf. (2)).
PUBLIC HOUSE
COUNTER
Figure 1. PUBLIC HOUSE frame

With regard to example (7), there is a FIREPLACE frame (Figure 2), one of
whose elements is obviously FIRE.
FIREPLACE
FIRE
Figure 2. FIREPLACE frame
Similarly, there exists a frame, say, DEPRIVATION OF LIBERTY (Figure 3) com-

prising different elements such as ACT OF SEIZING, CAPTIVITY, PRISON, PRIS-
ONER, PENALTY OF IMPRISONMENT, and so on, that easily accounts for the
metonymic changes of OFr. prison from ‘act of seizing’ to ‘captivity’ and
thereafter to ‘prison’ (cf. (6)).
DEPRIVATION OF LIBERTY
PRISONER
ACT OF
CAPTIVITY PRISON
SEIZING
PENALTY OF
IMPRISONMENT
Figure 3. DEPRIVATION OF LIBERTY frame

148 Peter Koch
Consider also the following, more complex example. In certain cultures at

certain times, we find a MARRIAGE frame (Figure 4), comprising elements such
as BETROTHAL, FIANCÉ(E), TRUST/ENGAGEMENT, SOMEONE who MARRIES1
(‘unites in matrimony’) the BRIDE and the BRIDEGROOM, who MARRY2 (‘take
each other as WIFE/HUSBAND’), MARRIAGE CONTRACT, VOW/OATH, WEDDING,
PRAYER, UNION OF WIFE AND HUSBAND, SET UP HOUSE, MOTHERHOOD, and so on.
On the basis of this frame, we can unitarily explain a series of quite
different metonymies in different languages:8
(8) a. Lat. sponsus,-a ‘fiancé(e),’ hence: ‘bride(groom)’ > VLat.
‘husband/wife’ (cf. Sp. esposo,-a; Fr. époux, -se)
b. Du. trouwen ‘to entrust s.o. to s.o.,’ hence ‘to betroth,’ hence
‘to marry1’ (cf. MHG truwen, NHG trauen)
c. OE weddian ‘to engage,’ hence ‘to marry2’
d. Lat. vota ‘vows’ > Sp. boda(s) ‘wedding’
Pol. ±lub ‘vow,’ hence: ‘marriage’
Goth. liuga ‘marriage’ (cf. OIr. lu(i)ge, Cym. llw, Br. lë ‘oath’)
e. (OHG ewa ‘law; contract’ >) MHG ë(we) ‘marriage contract,’
hence: ‘marriage’ (cf. NHG Ehe)
f. OFr. marier ‘to marry1’ = Eng. to marry ‘to unite in matri-
mony,’ hence: ‘to take s.o. as husband/wife’
g. Fr. mariage ‘marriage’, hence: ‘union of wife and husband’
and ‘wedding’
h. Lat. matrimonium (originally) ‘motherhood,’ hence: ‘mar-
riage’ (i.e., ‘union of wife and husband’)
> It. matrimonio ‘marriage,’ hence also ‘wedding’
> Sp. matrimonio ‘marriage,’ hence also ‘husband and wife’
i. VLat. casare ‘to set up house’ >? Sp. casar ‘to marry1’
j. Lat. oratio ‘prayer’ → Br. eured ‘wedding’
MARRIAGE
TRUST,
ENGAGEMENT
MARRIAGE
CONTRACT WEDDING
PARENTS
MARRY1 PRAYER
= UNITE IN MARRY2 VOW,
BETROTHAL MATRIMONY = TAKE AS OATH
WIFE/HUSBAND
WIFE/ UNION OF
BRIDE
FIANCÉ(E) HUSBAND WIFE AND
(GROOM)
HUSBAND
SET UP MOTHERHOOD
HOUSE
Figure 4. MARRIAGE frame
6. Frames and prototypes
But there still remain some problems with contiguity and even with frames.
The original paradigm of associationist philosophy and psychology (see Sec-
tion 3), which leans heavily on contiguity, takes a rather mechanistic ap-
proach.9 Whenever there is contiguity, there has to be an associative link. But
is not everything contiguous to everything else? So why do we not associate
everything with something or something with anything or everything? Frame
models seem to be more subtle because they represent non-accidental net-
works of contiguities. But even a frame model conceived as categorical
(something is or is not in a frame: this is inevitable, I think, in Artificial
Intelligence) would not serve our purpose. Quite on the contrary:
– We have to acknowledge (see (2)) that there are perhaps public houses
without a counter, which we would nevertheless call bars.
– We have to acknowledge (see (7)) that there are fireplaces without a
(burning) fire and that fire is not confined to fireplaces; and, nevertheless,
speakers of Vulgar Latin would have used the word focus to denote FIRE.
150 Peter Koch
– We have to acknowledge (see (8)) that there are marriages without a

wedding, husbands and wives that were never fiancé(e)s before, mar-
riages without motherhood and vice versa, etc. Nevertheless, all the
metonymic changes exemplified in (8) have taken place.
So then we have to admit that contiguity relations only hold for ‘salient’
members of the conceptual categories involved. Frames and the contiguity
relations constituting them have ‘prototypical’ character:10
– A prototypical public house has a counter, and so we call it bar.
– Prototypically, a fireplace has a fire burning in it, and for Roman people,
the prototypical fire may have been that of the fireplace; so they called it
focus.
– A prototypical marriage begins with a wedding; in certain cultures at
certain times, prototypical husbands/wives are engaged before they get
married, etc.
It is on the basis of such prototypical frames and contiguities that metonymy
works. This should not be misunderstood. Many prototype theorists speak of
metonymic ‘extension,’ reducing metonymic change to a prototypical effect
within one and the same category. I do not adhere to this kind of purely
semasiologically based, ‘extended’ prototype notion (cf. Fillmore 1982: 32f;
Jongen 1985: 126f; Lakoff 1987: 91–117, 416–461 and passim; Geeraerts
1988, 1997: 21f; Taylor 1989: 99–141; see the astute criticism of Kleiber
1990: 147–183; see also below, Section 7.1., and cf. Warren 1992: 123; Koch
1995: 37–40, 1996a: 129–131, 1996b: 231–234; Blank 1997a: 79–85, 1997b:
89–93). On the contrary, I would like to draw attention to the fact that, from an
onomasiological point of view, PUBLIC HOUSE, for instance, although belong-
ing to the same frame, is not a peripheral instance of the category COUNTER
(nor vice versa).11
So the two terms of a contiguity relation, for instance COUNTER and
PUBLIC HOUSE, are two distinct conceptual categories, both prototypically
structured in themselves (Figure 5). Strictly speaking, metonymic change
involves prototypicality only inasmuch as the underlying contiguity relation
would not hold but for the salient, prototypical members of one or both of the
two categories concerned, and it is then generalized to these categories as
wholes. Or, as Simon C. Dik (1977) puts it, these changes of meaning
presuppose a sort of ‘inductive generalization’ (cf. also Geeraerts 1997: 68ff).
PUBLIC contiguity
COUNTER
HOUSE
Figure 5. Contiguity relation between PUBLIC HOUSE and COUNTER
7. Frames and gestalts
When we introduce notions like ‘prototype’ and ‘salience’ into frame theory,
we recognize the ‘gestalt’ character of frames and contiguities.12 It is only by
leaning on salience effects within frames that we can avoid the aporias of a
mechanistic associationist approach or of categorical frame models.
More precisely, from a gestalt perspective, metonymy turns out to be a
‘figure/ground’ effect. Consider Figure 6 as a traditional example of a percep-
tual figure/ground constellation. In this figure, we can perceive a white cross
on black ground, but alternatively, it seems to be a black cross on white
ground (cf. Wittgenstein 1958: 207).
Figure 6. Figure/ground perception

152 Peter Koch
Coming now to the conceptual level, we can claim that every concept desig-
nated by a given lexical item appears as a figure in relation to (at least) another
contiguous concept that — for the time being — remains the ground within the
same frame. But at some moment, while we are using the same lexical item,
certain pragmatic, conceptual or emotional factors may highlight the ground
concept so that figure and ground become inverted. That is what we call
metonymy.
Thus, in one of the senses of OFr. prison (‘captivity’; see above (6)),
CAPTIVITY was the figure and PRISON one possible ground (Figure 7a). By
highlighting PRISON into the figure and backgrounding CAPTIVITY, prison was
acquiring the new metonymic sense ‘prison’ (Figure 7b).
According to Croft (1993: 348), we can interpret metonymy as a concep-
tual effect of domain highlighting within one domain matrix (opposing it to
metaphor as a conceptual effect of domain mapping across different domain
matrices).13 In the case of OFr. prison ‘captivity; prison,’ we shift, say, from
the domain (HUMAN) CONDITION to the domain LOCATION within the domain
matrix for prison. Note, however, two significant differences between the
domain-matrix approach and the frame approach proposed above:
1. Frame-internal relations, in as far as they represent contiguity relations,
exclude any similarity or taxonomic relations (relevant to metaphor and
extension/restriction of meaning only). For domain (matrices) this is not
so clear: the domain structure underlying the concept of the letter ‘T,’ for
CAPTIVITY PRISON
Figure 7a. Figure CAPTIVITY, ground PRISON

CAPTIVITY PRISON
Figure 7b. Figure PRISON, ground CAPTIVITY
instance, contains relations like COMMUNICATION – HUMAN BEINGS, i.e.,

contiguity relations, but also LIVING THINGS – HUMAN BEINGS, i.e., taxo-
nomic relations (cf. Croft 1993: 340–342). I think that the incorrect, but
widespread term ‘metonymic extension’ (see Section 6) results just from
an insufficient distinction between contiguity relations and taxonomic
relations (including ‘soft,’ prototype-based relations). The same holds, by
the way, for the confusion inherent in the traditional notion of ‘synecdo-
che’ (see Section 8).
2. In general, domain matrices are defined semasiologically in relation to an
existing lexical item (cf. also Croft 1993: 338), whereas frames, which
are relevant not only to metonymies but also to certain types of word
formation, can — and in fact, should — be defined onomasiologically, so
that even cross-over links within one and the same frame realized in
different languages (cf. Figure 4 and examples (8)), concepts which have
not yet been expressed, senses of a given word which do not yet exist, and
new words which have not yet been formed can all be provided for.
8. The range of metonymy
A frame-based interpretation of contiguity helps us to determine the notion of

‘metonymy,’ whose range is not clear in all respects. According to rhetorical
154 Peter Koch
tradition, we can distinguish metonymy from synecdoche, but the latter often
appears to be only a special case of the former (Lausberg 1973: §§572–577;
Bredin 1984: 45f). However, as Le Guern (1973: 36ff) has pointed out,
synecdoche is not, in reality, a unitary trope (see also Section 7.1.): traditional
synecdoche comprises (or rather: confuses) cases of taxonomic extension/
generalization (e.g., bread ‘foodstuff’) or restriction/specification (e.g., mor-
tal ‘man’) on the one hand, which have nothing to do with metonymy, and
cases of pars pro toto/totum pro parte on the other hand (e.g., roof ‘house’;
America ‘USA’).
In my opinion, we should integrate pars pro toto and totum pro parte into
metonymy (cf. also Ullmann 1962: 212; Schifko 1979: 247; Lakoff and
Johnson 1980: 36; Croft 1993: 350; Warren 1992: 64ff, 1995; Blank, this
volume; Seto, this volume). Separating metonymy and pars/totum ‘synecdo-
che’ would be artificial because the difference between pars/totum relations
and (other) contiguities is often not so easy to pin down. For instance, are the
relations COUNTER – PUBLIC HOUSE or FIRE – FIREPLACE to be considered pars/
totum relations or relations of location?14 Behind these possible differences
and uncertainties, we nevertheless perceive a fundamental constant: the pars/
totum tropes, like any metonymic trope, involve a figure/ground effect: in pars
pro toto, the totum — as a ground that becomes the figure — is a whole frame,
and the pars — as a figure that becomes the ground — is one of the concepts
of this frame (and vice versa for totum pro parte).
So we can retain our definition given in Section 5: contiguity is a salient
relation that exists between the elements (or sub-frames) of a conceptual
frame or between the frame as a whole and its elements. Consequently,
metonymy implies a contiguity-based figure/ground effect between elements
of a conceptual frame or between the frame as a whole and one of its elements
(or vice versa).
A new differentiation between two contiguity-based changes of meaning
has been suggested by Warren, who distinguishes ‘metonymy’ from ‘implica-
tion’ (cf. Warren 1992: 51–72, 101, 1995). Metonymy (in her narrow sense of
the term) is a non-literal use of a word that causes an abrupt shift of meaning,
restricted to a few, well-defined contiguity relations and — normally — to
nouns. Implication, on the other hand, is the gradual development of novel,
coexisting senses, based on if-then contiguities of any kind and occurring in
adjectives, verbs, and nouns. I reject this differentiation on the following
grounds.
First of all, it can easily be shown that metonymies (in this narrow sense)
are possible even with verbs and adjectives:
(9) Fr. descendre ‘to go down,’ hence: ‘to take down’ (with many
parallels in other languages; cf. Koch 1991: 294; Haspelmath
1993: 92–94, 101, 104, 112–120; see also our example Eng.
marry in (8f))
(10) Eng. sad ‘distressed,’ hence: ‘distressing’ (with many parallels in
other languages; cf. BDE, s.v. sad)
Secondly, we have to acknowledge that a frame-based figure/ground
effect is present in ‘implications’ as well as in ‘metonymies’ (in the narrow
sense). I would claim that the difference resides rather in the pragmatic,
referential and expressive conditions in which these metonymies (in the broad
sense) emerge in discourse, as we will see in the next section.
9. The origins of metonymy in discourse
In Section 2 I said that metonymic change of meaning is induced by ad hoc

metonymies. The first step is always a figure/ground effect in discourse. We
can imagine at least three possible types of ad hoc figure/ground effects for
metonymy in discourse.
Consider, first of all, a putative reconstruction of the ad hoc figure/ground
effect underlying (7) and illustrated in Table 4.
Table 4. Metonymic innovation through inference: FIRE
HEARER: Lat. “Incendamus focum!” FIREPLACE FIRE
initial interpretation (literally) figure ground

ground
‘Let’s light the fireplace!’
‘inverted’ interpretation ‘Let’s light the fire!’ > ground

ground > figure
This kind of frequently occurring figure/ground effect is due to a situational

context in which both interpretations are pragmatically indifferent, which may
in turn give rise to an almost surreptitious conceptual reinterpretation of the
156 Peter Koch
lexical item. Traugott and König call this process ‘pragmatic strengthening’ or
‘conventionalization of conversational implicatures/inferences.’ In this case,
it is the hearer (as a virtual speaker) who triggers the metonymic innovation
which induces a metonymic change later on (cf. König and Traugott 1988;
Traugott and König 1991: 193ff).
A similar analysis probably applies to one of the steps of semantic change
in our example (6), which may be represented as in Table 5.
Table 5. Metonymic innovation through inference: PRISON
HEARER: OFr. “Il est en prison.” CAPTIVITY PRISON
initial interpretation ‘He is in captivity.’ figure ground

ground
‘inverted’ interpretation ‘He is in prison.’ ground

> ground > figure
I suppose that this type of figure/ground effect is the one Warren calls
‘implication.’
A second type of ad hoc figure/ground effect is triggered by the speaker,
as illustrated in Table 6. A Latin speaker wants to refer to the thigh of a person,
but instead of using the ‘exact’ lexical item (Lat. femur), he imprecisely recurs
to another item (coxa) which designates a contiguous concept, namely HIP (cf.
Blank 1997a: 388f).
Table 6. Metonymic innovation through imprecision: THIGH
SPEAKER: HIP THIGH
referential means Lat. coxa ‘hip’ figure ground

ground
thing meant thigh x > ground

ground > figure
→(VLat. coxa ‘thigh’ (cf. Fr. cuisse)
The third type of ad hoc metonymy corresponds to metonymy as a rhetorical

trope.15 A Latin speaker, in a very emotional setting, wants to refer to the head
of a person he or she dislikes or is angry with. So the speaker does not use the
neutral word caput, but he chooses a more expressive lexical means, the word
testa ‘skull.’ It is the figure/ground effect between SKULL (as a salient part) and
HEAD (as the whole) that produces expressiveness here (Table 7).
Table 7. Metonymic innovation through expressiveness: HEAD
SPEAKER: SKULL HEAD
expressive means VLat. testa ‘skull’ figure ground
thing meant head x > ground > figure
→(VLat. testa ‘head’ (cf. Fr. tête)
Note the pragmatic differences between these three types of metonymic

innovation (irrespective of the fundamental figure/ground constant). In the
examples in Tables 4 and 5, it is the hearer (qua virtual speaker) who brings
about the innovation, whereas in the examples in Table 6 and in Table 7 it is
the speaker. In the cases described in Tables 4 and 5 and in Table 6, the
‘oddness’ of the innovation can be ‘repaired’ on the basis of Grice’s (1975)
conversational implicatures, whereas in the case described in Table 7 the
metonymic trope is meant to be striking and not intended to be ‘repaired.’
Later on, by conventionalization through lexicalization, it loses its striking
effect and, hence, its expressiveness.
10. Contiguity, gender change, and word formation
So far I have been looking at the role of contiguity in relation to metonymic

change of meaning, which, by producing metonymic polysemy, constitutes an
important contribution to lexical diversification. Now it is interesting to note
that contiguities within frames also enable us to explain other types of lexical
diversification.
Recall the frame DEPRIVATION OF LIBERTY (Figure 3) that comprised
concepts such as ACT OF SEIZING, CAPTIVITY, PRISON, PRISONER, PENALTY OF
IMPRISONMENT, etc. and allowed for an explanation of our example (6): as we
have seen, several elements of this same frame were designated one after the
other by the polysemous item Fr. la prison (with feminine gender). Now I
would like to draw your attention to the item li prisons (with masculine
gender), which in Old French enabled speakers to designate yet another
concept, namely PRISONER:
158 Peter Koch
(6’) OFr. la prison ‘captivity’ → li prisons ‘prisoner’

This new possibility of designation is based once again on a contiguity
relation within the same frame. However, it is not simply a case of change of
meaning, but the creation of a new lexical item identical with the original one
(la prison), except for the gender. We have to conclude that frames and
contiguities not only account for metonymic change and polysemy, but also
for other lexical processes such as gender change.
In this area, besides gender change, which is of minor importance, we
especially have to think of certain types of word formation. OFr. li prisons (6’)
has not survived in Modern French, but in competition with it, there existed
the item li prisoniers ‘prisoner,’ which has survived in Modern French:
(6”) OFr. la prison ‘captivity’ → li prisoniers ‘prisoner’ > MFr. le
prisonnier
Semantically, the item prisonnier belongs to a type of derivation called
‘Ausgriff’ by Gauger (1971: 66–74). In my analysis, this type of derivation is
also founded upon contiguity within frames: a new concept is ‘grasped’ (and a
new word is formed) by ‘reaching out’ from a contiguous concept designated
by a word which already exists.16 This is a most important insight. Frames and
contiguities represent a very fundamental cognitive principle underlying sev-
eral lexical processes which considerably differ from each other. The
processes are: metonymic change of meaning (without any morphological
change), gender change, derivation of the ‘Ausgriff’ type, and, as we will see
in a moment, even a certain type of composition.17
Let us consider, as a final example, the FRUIT-TREE frame (cf. Koch, in
print). Based on the contiguity between FRUIT and TREE, in Italian, we find
within this frame the following metonymic change of meaning and the result-
ing metonymic polysemy:
(11) a. It. limone ‘lemon’, hence: ‘lemon tree’
In Spanish and English, on the other hand, we have word formations that are
based on exactly the same contiguity between FRUIT and TREE:
(11) b. Sp. limón ‘lemon’ → limonero ‘lemon tree’ (derivation)
c. Eng. lemon → lemon tree (composition)
As (11c) shows, there exists in English a — by the way, very important —
type of nominal composition N1 + N2, in which the concept designated by N1
(LEMON FRUIT) and the concept designated by N1 + N2 (LEMON TREE) are

contiguous (whereas the relation between the concept designated by N2 (TREE)
and the concept designated by N1 + N2 (LEMON TREE) is that of taxonomic
subordination).
A purely semasiological approach (see Section 7.2.) may be able to
reconstruct examples like (11a), but it would not account for the fundamental
cognitive principle underlying examples (11a), as well as (11b) and (11c),
whose essential conceptual homology can be elucidated only by an onomasio-
logical frame approach.
11. Conclusion
In conclusion, when trying to grasp the mechanisms underlying metonymy, it

may be useful to integrate four different traditional and non-traditional para-
digms of ‘cognitive’ research: the associationist paradigm, gestalt theory,
frame theory, and prototype theory. In certain cases (see Section 9), it is
stimulating to take König and Traugott’s (1988) pragmatic strengthening
theory into account as well.
It is only the integration of these different approaches that enables us to
appreciate the nature and the relatedness of frames and contiguities. The
salient links between elements of a given frame — as constituting a prototypi-
cal conceptual gestalt — are what we call contiguity relations. Along these
contiguities, people can produce the figure/ground effects underlying metony-
mies. But frames and contiguities extend considerably beyond the realm of
metonymy: they also help us to understand certain types of gender change and
word formation.
So, metonymy is a very important — though not the only — rhetorical
and lexical device that gives us a clue to the fundamental role of frames and
contiguities.
Notes
1. I express my gratitude to Mary Copple (Berlin) for the stylistic revision of this paper.
2. Concerning the intimate link between semantic change and polysemy, cf. Bréal (1921:
143ff, 284–287), Werth (1974: 377f), Koch (1991: 283), Blank (1993: 31), for me-
160 Peter Koch
tonymy, see especially Schifko (1979: 248–251). The lexicalization steps for metaphor
are described in Koch (1994: 203–209); a general framework for lexicalization (and
delexicalization) processes within semantic change has been established in Blank
(1997a: 116–130). Note the important difference between ‘innovation’ and ‘change’
proposed by Coseriu (1958: 44–46).
3. Cf. Amin (1973: 19–81, especially 38). The first to explicitly take up Aristotle seems to
be the Scotsman Thomas Brown, who occupies a somewhat particular position with
respect to associationism (cf. ibid.: 72).
4. Cf. the critical outlines of Wundt’s approach in Roudet (1921: 681–686); Nerlich (1992:
77–80); Blank (1997a: 18f).
5. From Hjelmslev onwards, Saussure’s rather open concept of ‘associative’ relation is
replaced by the technical, more specific concept of ‘paradigmatic’ relation (EITHER/
OR-relation of substitution in a given syntagmatic environment); cf. Hjelmslev (1963:
33–40); Happ (1985: 52–59).
6. Cf. Jakobson (1956: 90f); Holenstein (1974: 81–91); Happ (1985: 75–79, 130–139).
According to Raible (1981), the distinction between semantic(-paradigmatic) contiguity
(hut – thatch) and (semantic-)syntagmatic contiguity (hut – burnt out) is essential for
preferences in human verbal associations.
7. Rastier (1997: 140–145) underscores the undeniable tendency of cognitive linguistics to
spatialize its objects of description.
8. Cf. BDE, s.v. marry; DCECH, s.vv. boda, casar, esposo; DE, s.v. Ehe; DELL, s.v. mater;
DHLF, s.vv. époux, marier; DSSPIL, s.vv. 2.33 marry, 2.34 marriage; REW, s.vv. casa,
sponsus, votum.
9. Cf. the critical observations in Amin (1973, especially 39–42, 111–115).
10. For prototype theory, cf. Rosch (1973); Fillmore (1975); Taylor (1989); Kleiber (1990);
Cordier (1993).
11. See also Geeraerts (1997: 74f). Concerning the difference between salience effects within
a frame and prototypicality effects within a category, cf. Cordier (1993: 122–124, 135–
149).
12. For gestalt theory in general, cf. for example Wertheimer (1922/23); Köhler (1947);
Metzger (1986). Amin (1973: 97–155, 201f) stresses the holistic character of gestalt
psychology as opposed to the mechanistic associationist approach, but he nevertheless
does not definitively exclude a synthesis of association and ‘gestalt’ (cf. also Raible
1981: 5f).
13. Cf. also Taylor (1989: 90). For the notions ‘domain’ and ‘domain matrix,’ cf. Langacker
(1987: 147ff). — As to the totally different character of metaphor, cf., for instance,
Bühler (1965: 342–350); Black (1954: 1977); Lakoff and Johnson (1980); Bredin (1984:
57); Liebert (1992: 28–82); Koch (1994: 209–214).
14. Cf. also Cordier (1993: 123). Within the type of contiguity called ‘co-presence’ by Blank
(this volume), there are several relations that can hardly be distinguished from pars/
totum: TYPICAL ASPECT — ACTIVITY, TYPICAL ASPECT — FRAME, TYPICAL ASPECT — OBJECT,
FUNCTION — OBJECT, OBJECT — PLACE. On the other hand, Blank’s distinction between
‘co-presence’ (including pars/totum) and ‘succession’ (excluding pars/totum) seems

rather clear-cut.
15. We could speak here of ‘everyday rhetoric’ (Alltagsrhetorik) in Stempel’s sense (cf., for
example, 1983); cf. also Koch and Oesterreicher (1996: 68–74, 79f); for the metonymy
presented in Table 6 (and its many polygenetic parallels), cf. Koch (1997: 231f, 236).
16. Different types of contiguity relations involved in certain Italian derivation processes
have been described by Schwarze (1995) in terms of several conceptual models that are in
part frame models (the ‘activity model,’ ibid.: 500–506; the ‘object constitution model,’
ibid.: 506–509, etc.)
17. Cf. also Schifko 1979, 252–257 (unlike Schifko, I would not apply, however, the term
‘metonymy’ even to cases of derivation (‘Ausgriff’) and composition).
Abbreviations for languages
Br. Breton MFr. Modern French

CLat. Classical Latin NHG New High German
Cym. Cymric OE Old English
Du. Dutch OFr. Old French
Eng. English OHG Old High German
Fr. French OIr. Old Irish
Goth. Gothic Pol. Polish
It. Italian Sp. Spanish
Lat. Latin VLat. Vulgar Latin
MHG Middle High German
References
Amin, Ismail
1973 Assoziationspsychologie und Gestaltpsychologie: Eine problemgeschicht-
liche Studie mit besonderer Berücksichtigung der Berliner Schule. Bern,
Frankfurt/M.: Lang.
Anonymous
1894 Incerti auctoris De ratione dicendi ad C. Herennium libri IV. Leipzig:
Teubner (reprint Hildesheim: Olms, 1966)
Aristotle
1975 Aristotle in Twenty-Three Volumes. VIII: On the Soul. Parva Naturalia. On
Breath. With an English translation by W.S. Hett. Cambridge, MA, Lon-
don: Harvard University Press.
Barsalou, Lawrence W.
1992 Frames, concepts, and conceptual fields. In A. Lehrer, E.F. Kittay (eds.),
162 Peter Koch
Frames, Fields, and Contrasts: New Essays in Semantic and Lexical Orga-
nization. Hillsdale, NJ, London: Erlbaum, 21–74.
Bartlett, Frederick Ch.
1932 Remembering: A Study in Experimental, Social Psychology. Cambridge:
Cambridge University Press.
BDE
1988 The Barnhart Dictionary of Etymology. Robert K. Barnhart, New York:
Wilson.
Black, Max
1954 Metaphor. Proceedings of the Aristotelian Society 55: 273–294.
1977 More about metaphor. Dialectica 31: 431–457.
Blank, Andreas
1993 Polysemie und semantische Relationen im Lexikon. In W. Börner, K.
Vogel (eds.), Wortschatz und Fremdsprachenerwerb. Bochum: AKS, 22–
56.
1997a Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romani-
schen Sprachen. Tübingen: Niemeyer.
1997b Il senso di una semantica dei prototipi e dei ‘frames’: osservazioni
decostruttive e ricostruttive. In M. Carapezza et al. (eds.), Linguaggio e
cognizione. Atti del XXVIII Congresso della Società di Linguistica Italiana,
Palermo, 27–29 ottobre 1994, Roma: Bulzoni, 89–103.
Bonhomme, Marc
1987 Linguistique de la métonymie. Bern: Lang.
Bréal, Michel
1921 Essai de sémantique (science des significations). Fifth edition. Paris:
Hachette.
Bredin, Hugh
Bühler, Karl
1965 Sprachtheorie: Die Darstellungsfunktion der Sprache. Second edition.
Stuttgart: Fischer.
Cordier, Françoise
1993 Les représentations cognitives privilégiées: Typicalité et niveau de base.
Lille: Presses Universitaires de Lille.
Coseriu, Eugenio
1958 Sincronía, diacronía e historia: El problema del cambio lingüístico. Uni-
versidad de Montevideo.
Croft, William
DCECH
1980–83 Diccionario crítico etimológico castellano e hispánico. 5 vol. Second
edition. J. Corominas, J.A. Pascual (eds.), Madrid: Gredos.
DE
1989 Duden. Etymologie. Herkunftswörterbuch der deutschen Sprache. Second
edition. G. Drosdowski, Mannheim: Dudenverlag.
DELL
1951 Dictionnaire étymologique de la langue latine: Histoire des mots. A.
Ernout, A. Meillet (eds.), Paris: Klincksieck.
DHLF
1992 Dictionnaire Historique de la Langue Française. 2 vol. Alain Rey, Paris:
Dictionnaires le Robert.
Dik, Simon C.
1977 Inductive generalisation in semantic change. In P.J. Hopper (ed.), Studies in
Descriptive and Historical Linguistics. Festschrift for W.P. Lehmann.
Amsterdam and Philadelphia: Benjamins, 283–300.
DSSPIL
1971 A Dictionary of Selected Synonyms in the Principal Indo-European Lan-
guages: A Contribution to the History of Ideas. Third edition. C.D. Buck,
Chicago: The University of Chigaco Press.
Dubois, Jacques, Francis Edeline, Jean-Marie Klinkenberg, Philippe Minguet, François
Pire, Hadelin Trinon
1970 Rhétorique générale. Paris: Larousse.
Eco, Umberto
1984 Semiotica e filosofia del linguaggio. Torino: Einaudi.
Fillmore, Charles J.
1975 An alternative to checklist theories of meaning. Proceedings of the Annual
Meeting of the Berkeley Linguistics Society 1: 123–131.
1982 Towards a descriptive framework for spatial deixis. In R.J. Jarvella, W.
Klein (eds.), Speech, Place, Action: Studies in Deixis, Related Topics.
Chichester: Wiley, 31–59.
1985 Frames and the semantics of understanding. Quaderni di Semantica 4: 217–
240.
Gauger, Hans-Martin
1971 Durchsichtige Wörter: Zur Theorie der Wortbildung. Heidelberg: Winter.
Geeraerts, Dirk
1988 Prototypicality as a prototypical notion. Communication and Cognition 21:
343–355.
1997 Diachronic Prototype Semantics: A Contribution to Historical Lexicology.
Oxford: Clarendon.
Grice, H. Paul
Happ, Heinz
1985 ‘Paradigmatisch’ – ‘syntagmatisch’: Zur Bestimmung und Klärung zweier
Grundbegriffe der Sprachwissenschaft. Heidelberg: Winter.
Haspelmath, Martin
1993 More on the typology of inchoative/causative verb alternations. In B.
Comrie, M. Polinsky (eds.), Causatives and Transitivity. Amsterdam and
Philadelphia: Benjamins, 87–120.
164 Peter Koch
Hjelmslev, Louis
1963 Prolegomena to a Theory of Language. Second edition. Madison: Univer-
sity of Wisconsin Press.
Holenstein, Elmar
1974 Die zwei Achsen der Sprache und ihre Grundlagen. In E. Holenstein (ed.),
1976. Linguistik – Semiotik – Hermeneutik: Plädoyers für eine strukturale
Phänomenologie. Frankfurt/M.: Suhrkamp, 76–113.
1975 Roman Jakobsons phänomenologischer Strukturalismus. Frankfurt/M.:
Suhrkamp.
Jakobson, Roman
1956 Two aspects of language and two types of aphasic disturbances. In R.
Jakobson, M. Halle, 1971. Fundamentals of Language. Den Haag, Paris:
Mouton, 67–96
Jongen, René
1985 Polysemy, tropes, cognition or the non-Magrittian art of closing curtains
whilst opening them. In W. Paprotté, R. Dirven (eds.), The Ubiquity of
Metaphor. Amsterdam and Philadelphia: Benjamins, 121–139.
Kleiber, Georges
1990 La sémantique du prototype. Catégories et sens lexical. Paris: Presses
Universitaires de France.
Koch, Peter
1991 Semantische Valenz, Polysemie und Bedeutungswandel bei romanischen
Verben. In P. Koch, T. Krefeld (eds.), Connexiones Romanicae. Dependenz
und Valenz in romanischen Sprachen. Tübingen: Niemeyer, 279–306.
1993 Kyenbé – tyonbo. Wurzeln kreolischer Lexik. In C. Foltys, T. Kotschi
(eds.), Berliner Romanistische Studien: Für Horst Ochse. Berlin: Institut
für Romanische Philologie der Freien Universität, 259–287.
1994 Gedanken zur Metapher – und zu ihrer Alltäglichkeit. In A. Sabban, C.
Schmitt (eds.), Sprachlicher Alltag: Linguistik – Rhetorik – Literaturwis-
senschaft. Festschrift für Wolf-Dieter Stempel. Tübingen: Niemeyer, 201–
225.
1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: eine kritische
Bestandsaufnahme. Romanistisches Jahrbuch 46: 27–46.
1996a Le prototype entre signifié, désigné et référent. In H. Dupuy-Engelhardt
(ed.), Questions de méthode et de délimitation en sémantique lexicale:
Actes d’EUROSEM 1994. Reims: Presses Universitaires de Reims, 113–
135.
1996b La sémantique du prototype: sémasiologie ou onomasiologie? Zeitschrift
für französische Sprache und Literatur 106: 223–240.
1997 La diacronia quale campo empirico della semantica cognitiva. In M.
Carapezza et al. (eds.), Linguaggio e cognizione. Atti del XXVIII Congresso
della Società di Linguistica Italiana, Palermo, 27–29 ottobre 1994, Roma:
Bulzoni, 225–246.
in print TREE and FRUIT: A cognitive-onomasiological approach. Studi Italiani di
Linguistica Teorica ed Applicata.
Koch, Peter, Wulf Oesterreicher

1996 Sprachwandel und expressive Mündlichkeit. Zeitschrift für Literaturwis-
senschaft und Linguistik 102: 64–96.
Köhler, Wolfgang
1947 Gestalt Psychology: An Introduction to New Concepts in Modern Psychol-
ogy. New York: Liveright.
König, Ekkehard, Elizabeth C. Traugott
1988 Pragmatic strengthening and semantic change. In W. Hüllen, R. Schulze
(eds.), Understanding the Lexicon. Tübingen: Niemeyer, 110–124.
Kruszewski, Mikolaj
1884–90 Prinzipien der Sprachentwicklung. Internationale Zeitschrift für allge-
meine Sprachwissenschaft 1: 295–307; 2: 258–268; 3: 145–187; 5: 133–
144, 339–360.
Lakoff, George
1987 Foundations of Cognitive Grammar. Vol. I: Theoretical Prerequisites.
Lausberg, Heinrich
1973 Handbuch der literarischen Rhetorik. Second edition. München: Hueber.
Le Guern, Michel
1973 Sémantique de la métaphore et de la métonymie. Paris: Larousse.
Liebert, Wolf-Andreas
1992 Metaphernbereiche der deutschen Alltagssprache: Kognitive Linguistik
und die Perspektiven einer Kognitiven Lexikographie. Frankfurt/M.: Lang.
Martin, Robert
1992 Pour une logique du sens. Second edition. Paris: Presses Universitaires de
France.
Metzger, Wolfgang
1986 Gestaltpsychologie: Ausgewählte Werke aus den Jahren 1950 bis 1982. M.
Stadler, H. Crabus (eds.), Frankfurt/M.: Kramer.
Minsky, Marvin
1975 A framework for representing knowledge. In P.H. Winston (ed.), The
Psychology of Computer Vision. New York: McGraw-Hill, 211–277.
Nerlich, Brigitte
1992 Semantic Theories in Europe 1830–1930: From Etymology to Contextual-
ity. Amsterdam and Philadelphia: Benjamins.
Raible, Wolfgang
1981 Von der Allgegenwart des Gegensinns (und einiger anderer Relationen).
Strategien zur Einordnung semantischer Informationen. Zeitschrift für Ro-
manische Philologie 97: 1–40.
166 Peter Koch
Rastier, François
1997 De l’antinomie spatio-temporelle aux zones anthropiques. In H. Dupuy-
Engelhardt, M.-J. Montibus (eds.), L’organisation lexicale et cognitive des
dimensions spatiale et temporelle. Actes d’EUROSEM 1996. Reims:
Presses Universitaires de Reims, 139–153.
REW
1992 Romanisches etymologisches Wörterbuch. Second edition. W. Meyer-
Lübke. Heidelberg: Winter.
Rosch, Eleanor
1973 On the internal structure of perceptual and semantic categories. In Th.E.
Moore (ed.), Cognitive Development and the Acquisition of Language.
New York: Academic Press, 111–144.
Roudet, Léonce
1921 Sur la classification psychologique des changements sémantiques. Journal
de Psychologie normale et pathologique 18: 676–692.
Saussure, Ferdinand de
1916 Cours de linguistique générale. Paris, Lausanne: Payot.
Schifko, Peter
1979 Die Metonymie als universales sprachliches Strukturprinzip. Grazer Lin-
guistische Studien 10: 240–264.
Schwarze, C.
1995 Grammatik der italienischen Sprache. Second edition. Tübingen:
Niemeyer.
Stempel, Wolf-Dieter
1983 ‘Ich vergesse alles’. Bemerkungen zur Hyperbolik in der Alltagsrhetorik. In
M. Faust et al. (eds.), Allgemeine Sprachwissenschaft, Sprachtypologie und
Textlinguistik. Festschrift für Peter Hartmann. Tübingen: Narr, 87–98.
Tannen, Deborah
1979 What’s in a frame? In R.O. Freedle (ed.), New Directions in Discourse
Processing. Norwood, NJ: Ablex, 137–181.
Taylor, John R.
1989 Linguistic Categorization. Prototypes in Linguistic Theory. Oxford:
Clarendon.
Traugott, Elizabeth C., Ekkehard König
1991 The semantics-pragmatics of grammaticalization revisited. In E. Traugott,
B. Heine (eds.), Approaches to Grammaticalization. 2 vol. Amsterdam and
Ullmann, Stephen
Warren, Beatrice
1992 Sense Developments: A Contrastive Study of the Development of Slang
Senses and Novel Standard Senses in English. Stockholm: Almquist and
Wiksel.
1995 What is metonymy? Paper read at the 12th ICHL, Manchester, August
1995.
Weinrich, Harald
1987 Zur Definition der Metonymie und zu ihrer Stellung in der rhetorischen
Kunst. In A. Arens (ed.), Text-Etymologie: Untersuchungen zu Textkörper
und Textinhalt. Festschrift für Heinrich Lausberg zum 75. Geburtstag.
Stuttgart: Steiner, 105–110.
Werth, Paul
1974 Accounting for semantic change in current linguistic theory. In J.M. Ander-
son, Ch. Jones (eds.), Historical Linguistics I: Syntax and Morphology,
Internal and Comparative Reconstruction. Amsterdam: North-Holland,
377–415.
Wertheimer, Max
1922/23 Untersuchungen zur Lehre von der Gestalt. Psychologische Forschun-
gen 1: 47–58.
Wittgenstein, Ludwig
1958 Philosophical Investigations. Second edition. Oxford, Cambridge:
Blackwell.
Wundt, Wilhelm
1912 [1900] Völkerpsychologie: Eine Untersuchung der Entwicklungsgesetze
von Sprache, Mythus und Sitte. Bd. II: Die Sprache. Third edition. Leipzig:
Engelmann.
Co-presence and Succession
A Cognitive Typology of Metonymy
Andreas Blank
University of Marburg
Metonymy may in fact be more common than

metaphor [...]. And not only is it not widely stud-
ied, but most accounts of it are unsatisfactory.
(Bredin 1984: 45)
1. Introduction
Over a decade ago, Hugh Bredin (1984) complained about the marginal status
of metonymy studies up to that time.1 And in fact, typologies of metonymy
from Fontanier (1968 [1830]) to Schifko (1979) or Lakoff and Johnson (1980)
lack consistency in that there are usually metonymies remaining that do not fit
into any category. Especially, the widely accepted tripartition into ‘spatial,’
‘temporal’ and ‘causal’ metonymies (cf. Ullmann 1962; Duchácek 1967;
Schifko 1979) is insufficient, as it does not provide an explanation for metony-
mies such as:
(1) OE sælig ‘blessed’ > ModE silly ‘stupid’
(2) OGr aggelos ‘messenger’ > ‘angel’2
(3) Arab al wazir ‘minister,’ ‘mayor’ > Sp alguacil ‘lower law-court
official’
The main concern of the following reflections is how metonymies can best be
classified. In Section 4 I will introduce a couple of older and more recent
170 Andreas Blank
typologies, some of which also include synecdoche or parts of this trope.3

First, however, I will investigate the psychological foundation of metonymy
(Section 2) and sketch its cognitive background (Section 3). These reflections
are intended to contribute to a typology of metonymy in light of cognitive
linguistics (Section 5).4 As we will see, a cognitive basis is necessary for a
better understanding of the metonymic process.
2. The psychological background of metonymy
In ancient rhetoric, metonymy was already seen as the transfer of a word to a

“closely related or neighboring thing” (Anonymous 1954, Rhetorica ad
Herrenium IV: 32, 43).5 A similar interpretation is found in modern rhetoric
manuals: Fontanier (1968 [1830]: 79) calls the different types of metonymy
“tropes by correspondence,” which designate an object by the name of another
object, Lausberg (1960: § 565) sees a “real relation” between the meanings of
the two words, and Mortara Garavelli (1988: 149) speaks of “a connection of
reciprocal dependency.”6
What kind of relation is this? What gives rise to metonymy? According to
Stephen Ullmann (1962: 218ff) the underlying relation is “contiguity of
senses,” i.e., an association between (intralinguistic) semantic features of two
words. In contrast with metaphor, says Ullmann (ibid.), “metonymy [...] arises
between words already related to each other.” This definition, however, raises
two serious questions:
1. How can semantic features or sememes ever be contiguous to each other?
The only possibility is to put them next to each other in a representation of
word fields, but if we call this contiguity we are confusing the way we
conceptualize things with the way they really are, or — as Coseriu might
put it — we are making a “transitus ab intellecto ad rem” (cf. Blank 1993:
34; for the quotation cf. Coseriu 1958: 9).
2. The second problem is that often there are not two words being related,
but only one lexicalized word which is metonymically used to give a
name to an object or an idea that has not been verbalized before, as in (2)
or in:
(4) L pecunia ‘cattle (as a kind of currency)’ > ‘money’
But even when a concept has already been verbalized, a new metonymy shows
Co-presence and Succession 171
another facet of the old meaning, as in (1) OE sælig ‘blessed’ > ModE silly
‘stupid,’ which of course, initially, was intended to be a euphemistic synonym
of OE dumb or contemporary synonyms. In all cases, the metonymy relies on
extralinguistic world knowledge (cf. Blank 1993, 1997a; Warren 1995; Koch,
this volume): cattle served as money in former times, angels are God’s
messengers, and the most positive thing to say about a stupid person is that he
or she is blessed by Jesus.7
Thus, the associative relation is rather between the two concepts BLESSED
– STUPID, MESSENGER – ANGEL or CATTLE – MONEY. It is important to point out
that this conceptual contiguity exists before the metonymic transfer is done
(cf. Section 3.1.). As Hugh Bredin (1984: 57) says, “[...] metaphor creates the
relation between its objects, while metonymy presupposes that relation.”
Ullmann’s interpretation goes back to a scarcely known article by Léonce
Roudet (1921), who appears to be the first to have brought the terms ‘contigu-
ity’ and ‘similarity’ (Roudet says ‘resemblance’) into historical linguistics.
Roudet (1921: 688f) distinguishes between semantic change due to an asso-
ciation “par contiguïté entre les idées,” “par ressemblance entre les idées” as
well as “changements résultant des rapports syntagmatiques entre les mots.”8
The last type of change is generally called ‘ellipsis,’ the second ‘metaphor’
and the first ‘metonymy’ (cf. Blank 1997b: 157–190, 230–269, 281–302).
Roudet’s view of an association between ‘ideas’ is more plausible than
Ullmann’s ‘contiguity of senses’ according to our conception of the psycho-
logical grounding of metonymy.9
More popular than Roudet’s lucid classification became Jakobson’s (1971)
application of the terms ‘contiguity’ and ‘similarity’ to linguistics (cf. Koch, this
volume). The major point of Jakobson’s paper is the construction of a parallel
between communicative strategies of aphasic speakers and both metaphor and
metonymy. He begins by stating that any linguistic sign can either be ‘com-
bined’ with other signs or be ‘substituted’ by others. The first “mode of
arrangement” is called “combination” and is based on the in praesentia
contiguity of signs in an utterance, the second is named “selection” and is based
on an in absentia similarity of concepts in a paradigm (1971: 74f). He further
states that one type of aphasia (Wernicke and amnesic aphasia) is characterized
by a loss of semantic knowledge, the mode of selection being disturbed. These
speakers often have recourse to syntagmatic strategies, e.g., they give a
paraphrase of the word they cannot find — or create a metonymy. The other
major type of aphasia (Broca) is characterized by a loss of syntactic knowledge.
172 Andreas Blank
Thus, syntagmatic combinations become difficult, while selection still works,

enabling speakers to give synonyms for the words they want to express and even
create new metaphors (cf. Jakobson 1971: 77–89). According to Jakobson, a
disorder in combination leads to strategies based on selection/similarity, which
he calls the “metaphoric way,” whereas a disorder in selection leads to strategies
based on combination/contiguity, i.e., the “metonymic way” (1971: 90).
Thus, by analyzing the strategies employed by aphasic speakers,
Jakobson creates a strong interdependence between paradigmatic association,
similarity and metaphor on one side, and syntagmatic association, contiguity
and metonymy on the other. However, in light of more recent studies on
aphasia, Jakobson’s view is clearly too simplistic (cf. Happ 1985: 72–89).10
Jakobson himself seems to have noticed that his twofold linear interpretation
was insufficient, as, in the last chapter of his article, this structure undergoes a
fundamental correction from one page to the next: here he not only distin-
guishes between different types of semantic similarity and contrast (such as
synonyms, antonyms, metaphors), but he also interprets metonymies such as
thatch, litter or poverty in the sense of ‘hut’ as a combination of “the positional
similarity [i.e., ‘selection’; A.B.] with semantic contiguity” (1971: 91). In
other words, Jakobson ends up by interpreting metonymy as relying on a
‘paradigmatic’ association by contiguity!
This correction means in every respect a return to Roudet’s threefold
typology.11 Both paradigmatic and syntagmatic relations in language can
theoretically be combined with the three psychological associations of simi-
larity, contrast and contiguity (cf. Blank 1993: 32ff; 1997b: 131–156).12
Contiguity can therefore be realized ‘syntagmatically’ between words (in a
complex word, a phrase or a sentence) and ‘paradigmatically’ as an associa-
tion between concepts in the mind. The first combination is the basis for
ellipsis (e.g., notebook computer > notebook), the second can give rise to
metonymy.13
3. The cognitive and communicative backgrounds of metonymy
3.1. Concepts, frames, scenarios and contiguity
In the preceding section, I claimed that the relation upon which metonymies
are based is a conceptual one, and that these concepts are contiguous to each
other. The nature of this conceptual contiguity has yet to be explained.

In metonymy, target and donator (or vehicle) concepts, rather than being
isolated, are quite often parts of greater conceptual networks that have been
described as ‘frames,’ ‘scenes,’ ‘scenarios,’ ‘domains,’ etc. (cf. Minsky 1975;
Fillmore 1975, 1977, 1985; Croft 1993; for the terminology cf. Andor 1985).
Frames, scenes or scenarios are static or dynamic mental representations of
typical situations in life and their typical elements.14 Concepts within frames
are related by ‘conceptual contiguity.’15 As will be shown later (Section 5.1.),
the difference between dynamic and static conceptualizations is of fundamen-
tal relevance for a cognitive typology of metonymy. The content and the shape
of a frame depends on our everyday experience, on our world knowledge.
Beings, things, processes, and actions that generally or ideally occur together
are represented in the mind as a frame. A frame is formed by an ‘inductive
generalization’ of extra-linguistic knowledge (cf. Dik 1977: 283; Koch, forth-
coming: Section 6.2, this volume). This does not imply that all the concepts of
a frame are to be realized every time it is activated, it is sufficient that they
usually are.
Figure 1 shows two typically related English frames, which allow several
observations important for the understanding of frames:
MORNING NOON
SUGAR PUB FAST
TOAST
DRINK
BUTTER BREAKFAST LUNCH
MILK PEAS
SALT
MARMELADE CHIPS
EGG DISH
KETCHUP
HAM STEAK
Figure 1. The English breakfast and lunch frames
Concepts within frames can build a complex network of contiguity (e.g.,

TOAST – BUTTER – HAM – EGG – MILK), but they are also related to the frame
itself (BREAKFAST) and to other contiguous frames (LUNCH). There are even
‘intersections’ (DRINK, SALT, KETCHUP). When a specific frame is opened or
accessed, all concepts that by convention belong to this frame are simulta-
neously activated.16
174 Andreas Blank
Furthermore, this example shows that frames and their contents are
entirely culture-dependent: they not only differ from one linguistic commu-
nity to another but they can also vary even within the same community. An
English breakfast, for example, typically includes bacon and eggs, buttered
toast, etc., whereas in Scotland porridge made of oats is an important part of
breakfast.17
Finally, these two frames show that even strong contiguity between
concepts in a frame does not automatically lead to metonymy.18 Frame-
relations and frame-networks only build the cognitive foundation necessary
for a metonymic transfer. In this case, a salient conceptual relation is high-
lighted (Croft 1993: 348) and the target concept is verbalized by the word
usually related to the concept that serves as donator. As Langacker (1993: 30f)
puts it, the donator concept serves as a ‘reference point’ for accessing the
target. The metonymic innovation may remain a hapax legomenon or lead to
‘lexicalized metonymy’ by semantic change when being adopted by a group
of speakers or the whole speech community.19
Thus, we can say that metonymy is a linguistic device based on salient
conceptual relations within a frame-network. In this frame-network, three
forms of conceptual contiguity are relevant for metonymy:20
1. Relations between concepts within a frame, as in (5):
(5) L praeco ‘messenger’ > OSp pregón, OPt pregão ‘message’
2. Relations between concepts and the superordinate frame, as in (6):
(6) OF travail ‘pain’ > MF ‘work’; ME travail ‘pain’ > ModE travel
‘journey’
3. Or even relations between related frames, as in (7):
(7) OF disner ‘to have the first meal of the day’ > ModF dîner ‘to have
lunch’ > ‘to have dinner.’
3.2. The communicative efficiency of metonymies and their

lexicalization
Metonymies show a clear-cut cognitive background, which enables us to

produce and understand them easily, because the only thing speakers have to
share are the same conceptual relations, common world knowledge about how
life is typically organized and how the ‘things of life’ are interrelated. This
makes metonymies very efficient tools for resolving different tasks in commu-
nication.21
As we have already seen, metonymies can express new concepts which
have not been verbalized before but which are closely related to other con-
cepts in a frame; this was the case in examples (2) and (4). We have also
learned that they can be used for euphemistic, attenuating speech. This caused
the metonymic transfers in (1), (6) and also in (3), where a more prestigious
expression was used in addressing an inferior person in order to please him. In
time, it became impossible for Spanish speakers to address the ‘lower law-
court official’ in any other way than by calling him alguacil. A third possibil-
ity offered by metonymy is to reflect changes in reality and, consequently, in
the conceptual representation of the real world. This was the case in (7).
When, in the 16th century, the upper classes of Paris became used to getting
up late, they shifted the first (and most important) meal of the day from the
middle of the morning to noon. The term dîner, however, was kept. The same
thing happened a century ago, as people increasingly had to stay away from
home all day in order to work and, consequently, moved the most important
meal to the evening.
Metonymy is further used to turn deverbal nouns into more prototypical
nouns, as in:
(8) L prehensio ‘act of seizing someone’ > F prison ‘captivity’ >
‘prison’
(9) F bâtiment ‘act of building’ > ‘building’
(10) L civitas ‘citizenship’ > ‘community of citizens’ > OF cité, I città,
Sp ciudad, Pt cidade ‘town,’ ‘city’
As in the above examples, we often find that names for localities are
created by metonymic processes (cf. Koch 1997).
Finally, we find that quite often two concepts are strongly related in the
speech context so that a single word can accidentally refer to either of them.
The shared speech context enables us to understand utterances like The ham
sandwich is waiting for his check (Lakoff and Johnson 1980: 35), where a
waiter refers to a customer by the dish the customer has ordered. When this
conceptual constellation satisfies the communicative needs of a defined group
of speakers or the whole speech community, the metonymy or the type of
176 Andreas Blank
metonymy (cf. Section 4) becomes conventionalized, as in (5) or in the

following examples:
(11) L parare ‘to prepare,’ ‘to arm’ > It parare, Pt parar ‘to ward off, to
repulse’
(12) a. L plicare ‘to fold’ > Rum a pleca ‘to leave’
b. L plicare ‘to fold’ > Sp llegar ‘to arrive’
(13) F desfier ‘to break the vassal’s oath’ > ‘to defy’
In (11) PREPARATION is considered a precondition for REPULSING an en-
emy; in (12a) FOLDING THE TENTS in the Balkan shepherd society was con-
nected with LEAVING, while in the marine culture of Spain FOLDING THE SAILS
was linked to ARRIVAL. In (13) we can easily imagine that an utterance like OF
je vos desfie, said by the vassal to his lord, pragmatically implied the illocution
of DEFYING this lord. Here a performative meaning is generated by me-
tonymy.22
Linguistic (and even non-linguistic) communication can be seen as a
process whereby people try to maximize their communicative success by
minimizing their linguistic effort (cf. Sperber and Wilson 1986: 46; Keller
1994: 142f). Metonymy is a response to both demands. It can support commu-
nicative success when we use euphemism and it minimizes linguistic effort
when we use the same word for concepts that are closely related.23 The
selection of the convenient donator concept for expressing the target in
question is governed by a number of cognitive principles. Generally, concepts
that more closely correspond to our knowledge, that are more salient and/or
more concrete, are preferred to less bounded, less important or more abstract
concepts (cf. Langacker 1993: 30f; Radden and Kövecses, this volume).
4. Typologies of metonymy
Having discussed some of the psychological and cognitive bases of me-

tonymy, we can come to a classification of the different types of metonymy.
Let us first consider some traditional and more recent typologies. One ap-
proach to classification that can be found throughout ancient and modern
rhetoric is to give more or less complex lists of types of metonymy, like
INVENTOR FOR THE THING INVENTED, CONTAINER FOR CONTENTS, etc. (cf., e.g.,
Quintilian, Institutionis oratoriae; Dumarsais 1818; Fontanier 1968 [1830];

Lausberg 1960; Lakoff and Johnson 1980; Bredin 1984).
Another traditional way of classifying has already been referred to in
Section 1: several types of metonymy are subsumed under a more general type
of contiguity, i.e., SPATIAL, TEMPORAL and CAUSAL CONTIGUITY. Typologies of
this kind have been proposed, for example, by Ullmann (1962), Duchácek
(1967) and Schifko (1979). As said above, the problem with this more general
typology is that there remain types of metonymy which do not rely on
temporal, spatial or causal contiguity. Ullmann (1962: 218ff) and Duchácek
(1967: 123–129) resolve this problem by simply listing those remaining cases,
such as KINSHIP RELATIONS, RELATIONS BETWEEN A PERSON AND AN OBJECT, AN
INDIVIDUAL AND A GROUP, A WHOLE AND ITS PARTS and AN ACTIVITY AND
RELATED PHENOMENA. Certainly, it has to be questioned whether such a par-
ticular list could ever be complete,24 but a more important problem seems to
be the following: temporal, spatial and causal relations are very general “types
of contiguity,” which comprise several more specific types; the rest of the list
contains, in fact, such specific types! Duchácek and Ullmann thus combine
different levels of abstraction. Schifko (1979: 245f) groups cases such as these
in an extended “causal dimension,” which comprises relations like CAUSE –
INSTRUMENT, AGENT – ACTIVITY, ACTIVITY – AFFECTED OBJECT, ACTIVITY –
PRODUCT.
It is interesting to note that the types of metonymy in Schifko’s “causal
dimension” rely on certain case roles (cf. Fillmore 1968) substituting for each
other by means of the metonymic process. The idea that metonymy is a case-
role substitution is elaborated on by Dan Fass (1991: 43f and Appendix B) and
can also be found implicitly in Warren (1992: 66). Fass analyzes the types of
metonymy given by Stern (1931) and Lakoff and Johnson (1980) and states
that substitutions can be bidirectional but that they lead to different results,25
and that one role (PATIENT) can even be substituted for an instance of the same
role (e.g., in PART-WHOLE relations). From the point of view of case role or
valence semantics, this is a valid approach. The problem with this theory is
that, in order to see the case-role effect, we need concrete sentences that must
contain verbs. Lexicalized metonymies, however, show this metonymic sense
without being syntactically actualized. Thus, we can only state that a case-role
substitution might have occurred during the past.26 Fass also does not give a
complete list of all metonymic types resulting from case-role substitution.
Warren (1992: 65ff, 1995: Section 3, this volume) distinguishes five
178 Andreas Blank
types of contiguity, each of which comprises several subtypes: COMPOSITION

(e.g., linen ‘type of fabric’ > ‘sheets’), CAUSATION (e.g., ecstasy ‘strong
emotion, exaltation’ > ‘drug’), POSSESSION (e.g., sceptre ‘emblem of regal or
imperial power’ > ‘sovereignty’), LOCATION (e.g., town ‘densely populated
area’ > ‘town dwellers’), REPRESENTATION (e.g., university ‘building’ >
‘staff’). These five types seem indeed to cover the field of metonymy except
for the ASPECT – ATTRIBUTE relation, which in Warren (1992: 51–63) is
separated from metonymy and called ‘implication.’ The necessity of this
distinction, however, remains opaque, given the fact that the psychological
foundation of both is conceptual contiguity, and so it seems more logical to me
to subsume ‘implication’ under metonymy. 27
In summary, we can say that there is a certain tendency to reduce the
different types of metonymy to a smaller number of more general types of
contiguity. What is generally lacking in these typologies is an integrated
cognitive framework that covers all possible contiguities. Without such a
framework as a basis, all lists of types of contiguity and of the corresponding
metonymies remain purely empirical.
5. A cognitive typology of metonymy
In this section, I shall try to integrate the theoretical issues on contiguity and
frames, as introduced in Section 3, into a typological approach to metonymy
different from the taxonomies presented above. Instead of listing types of
conceptual relation, the focus will be set on two rather abstract conceptual
frameworks that derive directly from the two fundamental ways of conceptu-
alizing real-life situations, viz. as static frames or as dynamic scenarios.
5.1. Co-presence and succession
A typology of metonymy somewhat different from those presented above has

been suggested by Marc Bonhomme (1987: 58–65), who distinguishes be-
tween ‘métonymies situatives’ and ‘métonymies actancielles,’ the latter also
working on the ground of case-role substitution. What is interesting in the
distinction he makes, is that the concepts related in the first group are spatially
and/or temporally ‘co-present’; they rely on the synchronism of their ele-
ments. The concepts in ‘métonymies actancielles’ stand in a causal, instru-
mental, final or consecutive relation, that is to say, they are ‘successive in

time,’ although Bonhomme does not explicitly characterize them in this way.
With this distinction Bonhomme does not describe actual metonymies,
but offers a framework to classify them, or to be more precise: to classify types
of contiguity according to the temporal status and the perspectivization of
their conceptual relation. This fits in very well with our understanding of the
cognitive background of metonymy. We have co-presence in a frame, tempo-
ral succession in a dynamic scenario, or even succession between two related
frames.
‘Co-presence’ and ‘succession’ are two very general issues that hang
closely together with two fundamental models of human conceptualization:
1. the ‘synchronic’ model, in which all aspects of a given situation or a
‘system’ are equally present and where time is excluded;
2. the ‘diachronic’ model, where the processual, consecutive character of
things and events is highlighted.
These two ‘domains of contiguity’ each contain a considerable number of
more specific types of contiguity. The latter involve conceptual relations
which give rise to concrete metonymies. At first glance, the choice of co-
presence and succession as superordinated domains of contiguity seems to be
arbitrary. I feel, however, that these two meta-conceptual models are philo-
sophically (cf. Aristotle, Categories 12 and 13) and cognitively well rooted
(cf. the distinction made in Section 3.1. between frames and scenarios). Their
main advantage consists in integrating all possible types of contiguity without
any remainder.
Co-present relations exist between the ACTORS (people, animals, institu-
tions) interacting in a frame, their ACTIVITY, INSTRUMENTS, TOOLS, AFFECTED
OBJECTS or PRODUCTS, the PLACE where an activity takes place, and the TIME at
which this activity usually occurs.
Another group of co-present relations are typical (essential or implicated)
ATTRIBUTES and ASPECTS of persons, objects and activities as well as distin-
guishable PARTS of persons, bodies, objects and activities. The latter are, in
rhetoric, usually subsumed under synecdoche as PART-WHOLE and WHOLE-
PART relations. But together with the implicated ASPECTS they are psychologi-
cally grounded in conceptual contiguity. The difference between this type and
other types of contiguity is the fact that here distinct concepts are not just
related, but that they ‘build’ the whole and that the whole ‘consists’ of them.
Furthermore, INDIVIDUAL REPRESENTATIVES of a COLLECTIVE BODY and finally
180 Andreas Blank
the FRAME as a whole are co-present types of contiguity.

Figure 2 shows the co-present conceptual relations determined so far.
FRAME, ACTIVITY and an important STATE OF THINGS appear in the center
because ACTIVITY or STATE often build the core of a FRAME and sometimes
coincide with the latter. The ensuing list gives examples of metonymic seman-
tic change based on co-present conceptual relations.
TOOL OBJECT PART

INSTRUMENT TYPICAL ASPECT
TIME FRAME
PERIOD FUNCTION
ACTIVITY
STATE INVENTOR
COLLECTIVE ACTOR PRODUCT
BODY AUTHOR
PRODUCER
PLACE
Figure 2. Co-presence
Examples:
(a) ACTOR – OBJECT: L praeco ‘messenger’ > OSp pregón, OPt pregão
‘message’; F courrier ‘postman’ > ‘mail’
(b) INVENTOR – PRODUCT: G Zeppelin ‘proper name’ > ‘long, cylindrical,
dirigible balloon filled with gas’28
(c) ACTOR – TYPICAL ASPECT: OE sælig ‘blessed’ > ModE silly ‘stupid’; OGr
aggelos ‘messenger’ > ‘angel’
(d) OBJECT – TYPICAL ASPECT: L pecunia ‘cattle (as a kind of currency)’ >
‘money’
(e) OBJECT – FUNCTION: E target ‘little round shield’ > ‘object to be aimed at
in shooting’
(f) ACTIVITY – OBJECT: L vestis ‘act of dressing’ > ‘clothes’
(g) ACTIVITY – TIME: It vendemmia ‘gathering of grapes’ > ‘vintage (season)’
(h) ACTIVITY – PLACE: F comédie ‘comedy’ > ‘theater (building)’
(i) STATE – PLACE: OF prison ‘captivity’ > ‘prison’
(j) TYPICAL ASPECT – ACTIVITY: F baiser ‘to kiss’ > ‘to make love’
(k) TYPICAL ASPECT – FRAME: ME travail ‘pain’ > ModE travel ‘journey’
(l) PLACE – OBJECT: L focus ‘fireplace’ > F feu ‘fire’

(m) TOOL – OBJECT: L lingua ‘tongue’ > ‘language’
(n) INSTRUMENT – ACTOR: F plume ‘pen’ > ‘author’
(o) PART – WHOLE: Pt palavra ‘word’ > E palaver ‘talk,’ ‘conference’; F âme
‘soul’ > ‘human being’
(p) PART – PART: L bucca ‘cheek’ > VL bocca ‘mouth’
(q) OBJECT – COLLECTIVE BODY: L civitas ‘citizenship’ > ‘community of
citizens’
(r) COLLECTIVE BODY – STATE: L civitas ‘community of citizens’ > OF cité
‘city’
Successive relations exist between a STATE and its PREVIOUS and CONSECUTIVE
STATE, between an ACTIVITY or a PROCESS and its PURPOSE or AIM, its CAUSE or
PRECONDITIONS, or its PRODUCTS or RESULTS. Other successive relations exist
between PERIODS, different PLACES and, last but not least, related FRAMES. The
conceptual types of these successive relations are shown in Figure 3. As
above, the figure is followed by a list of examples.
PREVIOUS STATE CONSECUTIVE STATE

STATE
PRECONDITION ACTIVITY CONSECUTIVE ACTIVITY
PROCESS
MATERIAL PRODUCT
CAUSE RESULT
PURPOSE INSTRUMENT
AIM TOOL
(ACTOR/AUTHOR /
PRODUCER/INVENTOR)
PERIOD 1 PERIOD 2 PERIOD 3
PLACE 1 PLACE 2 PLACE 3
FRAME 1 FRAME 2 FRAME 3
Figure 3. Succession
182 Andreas Blank
Examples:
(a) ACTIVITY – CONSECUTIVE ACTIVITY: L plicare ‘to fold’ > Rum a pleca ‘to
leave,’ Sp llegar ‘to arrive’
(b) ACTIVITY – RESULT: F desfier ‘to break the vassal’s oath’ > ‘to defy’; L
prehensio ‘act of seizing someone’ > F prison ‘captivity’ > ‘prison’; F
bâtiment ‘act of building’ > ‘building’
(c) ACTIVITY – AIM: F pour ‘for (exchange)’ > ‘in order to’
(d) CAUSE – AIM: It perché, Sp porque ‘because’ > ‘in order that’
(e) RESULT – CAUSE: L reus ‘accused’ > Rum rau ‘bad’
(f) PRECONDITION – ACTIVITY: L parare ‘to prepare,’ ‘to arm’ > It parare, Pt
parar ‘to ward off, to repulse’; E to realize ‘to make real’ > ‘to recognize’
(g) MATERIAL – PRODUCT: E horn ‘horn of an animal’ > ‘wind-instrument’
(h) INSTRUMENT – PRODUCT: F plume ‘pen’ > ‘style (of an author)’
(i) PERIOD – PERIOD: Sp. mañana, ME morwe, morrow ‘morning’ > ‘the
following day’
(j) PLACE – PLACE: OL *loukos ‘clearing, open space’ > L lucus ‘grove,’
‘wood’
(k) FRAME – FRAME: OF disner ‘to have the first meal of the day’ > ModF
dîner ‘to have lunch’ > ‘to have dinner’
5.2. Three levels of abstraction
The two lists of contiguity types are certainly incomplete, as one might always
find metonymies grounded on conceptual contiguities other than those men-
tioned here. Up to now, however, every type of metonymy I have found
belongs to one of the two superordinate domains of conceptual contiguity, as
the underlying concepts are either co-present or successive to each other.29
The typology of metonymy proposed here involves three levels of ab-
straction (cf. Figure 4): two superordinate ‘domains of contiguity’ (co-pres-
ence and succession), a principally open list of ‘types of contiguity’ or
‘contiguity schemas’ (or even ‘conceptual metonymies’; cf. Radden and
Kövecses, this volume), and finally — on the level of the language — concrete
metonymies.
In our everyday life we observe, experience and learn concrete contigui-
ties and we are subsequently learning and using metonymies that result from
this knowledge. For speakers it is important to know which types of contiguity
are salient and which are usually realized by metonymy. This knowledge is
not only necessary to understand existing metonymies but it is very useful for
producing new ones. The rather abstract ‘domains of contiguity’ are less
relevant for speakers than for the purpose of linguistic description. They are
kinds of ‘meta-frames’ which show the same level of abstraction as the
conceptual meta-frames behind some types of word formation.30 Meta-frames
generally highlight a more general aspect underlying and structuring different
sorts of frames, as shown in Figures 2 and 3.31
DOMAINS OF CONTIGUITY
CO-PRESENCE SUCCESSION
TYPES OF CONTIGUITY (SCHEMAS)
CONCRETE METONYMIES
Figure 4. A typology of metonymies
The ‘types of contiguity’ can best be compared to the ‘image schemas’

underlying metaphors (cf. Lakoff and Johnson 1980; Johnson 1987), which
are highly recurrent and conventionalized, but nevertheless build on an open
list. These ‘contiguity schemas’ are extremely important for the production
and understanding of concrete metonymies. The more embodied a certain type
of contiguity is in our conceptual knowledge, the more it can be used in
concrete metonymies and the higher the chances are that such a metonymy is
communicatively successful and perhaps becomes lexicalized within a spe-
cific group of speakers or the whole speech community (cf. Section 3.2.). In
certain cases, it is only the type of contiguity in question that becomes
conventionalized for a group of speakers, e.g., waitresses or waiters, as in the
case of ORDERED DISH – CUSTOMER. On the other hand, no special dish (not
even the famous ‘ham sandwich’) seems to have become lexicalized yet in the
sense of ‘customer.’
Some contiguity schemas have restrictions concerning the group of
speakers; others have restrictions on the objects that can be metonymically
expressed. As Taylor (1989: 123) points out, it is, for example, rather uncom-
184 Andreas Blank
mon to use the contiguity schema AUTHOR – PRODUCT for referring to a cake by
the name of the baker (in a sentence like Mary was delicious).32
The three traditional types of metonymy as mentioned in Section 4
(spatial, temporal and causal relations) are so to speak ‘transverse’ to the two
superordinate types of conceptual contiguity. In both domains we can find
spatial relations (e.g., PART – WHOLE vs. CONTIGUOUS PLACES), temporal rela-
tions (e.g., ACTIVITY – TIME WHEN THE ACTIVITY OCCURS vs. PREVIOUS STATE –
CONSECUTIVE STATE) and causal relations (e.g., ACTIVITY – AFFECTED OBJECT
vs. PROCESS OF PRODUCTION – PRODUCT).
6. Conclusion
In this paper I have tried to show that a theory and typology of metonymy first
of all has to define the cognitive background within which it works. Further-
more it has to be made clear what is understood by ‘contiguity.’ Considering
this, I come to the following conclusions:
1. Metonymy as a linguistic device is the transfer of a word to another
concept on the basis of conceptual contiguity between a donator and a
target concept. Any spontaneous metonymy can be adopted by the speech
community and thus become lexicalized.
2. Conceptual contiguity results from a relation between these concepts
within cognitive frames, between a concept and the frame itself or be-
tween two related frames.
3. The conceptual relations in these frames and between frames are repre-
sented as types of contiguity. Any concrete metonymy is rooted in such a
conventionalized contiguity schema.
4. All conceptual relations relevant to metonymy are either co-present or
successive in time. These two very fundamental aspects in human con-
ceptualization constitute meta-frames which contain typical convention-
alized contiguity schemas. The co-presence and the succession domains
are mentally stored abstractions from the schemas behind actual metony-
mies.
Frame theory and the three-level conception of metonymy make metonymy a
very powerful and efficient tool in communication and semantic change. To
highlight a given concept we only have to apply the appropriate contiguity
schema to the frame in question. In contrast to metaphor, no new relation
needs to be established (by the speaker) and interpreted (by the hearer), as
both frames and contiguity schemas already belong to our knowledge.
Notes
1. I express my gratitude to Mary Copple for the stylistic revision of this paper.
2. A semantic loan of a metonymy in Hebrew (m’l’k).
3. Indeed, most of the newer typologies of metonymy mentioned in this paper integrate
PART-WHOLE, WHOLE-PART and even PART-PART relations, which in rhetoric usually fall
within the bounds of synecdoche (cf. Dumarsais 1818: 115; Lausberg 1960: § 572; Le
Guern 1973; Garcia Arance 1979; Ruwet 1983). In fact synecdoche is a highly problem-
atic category as it comprises contiguity-based types like those already mentioned (e.g., E
soul ‘spiritual part of man’ > ‘human being’), and similarity-based transfers like E bread
‘food made out of flour or meal’ > ‘food,’ where the relation is between the more specific
and the general (cf. Nerlich, forthcoming: Section 5; Seto, this volume).
4. A cognitive foundation of metonymy and other contiguity-based phenomena in the
lexicon akin to mine is discussed in detail in Koch, this volume. Focusing in my
contribution on the cognitive roots of metonymy is nevertheless necessary for a better
comprehension of the typology suggested in Section 5.
5. “denominatio est quae ab rebus propinquis et finitimis trahit orationem, qua possit
intellegi res quae non suo vocabulo sit appellata.” (Rhetorica ad Herennium IV: 32, 43;
my italics).
6. “[...] désignation d’un objet par le nom d’un autre objet qui fait comme lui un tout
absolument à part, mais qui lui doit ou à qui il doit lui-même plus ou moins, ou pour son
existence, ou pour sa manière d’être” (Fontanier 1968: 79). — “Die Metonymie
verwendet [...] ein Wort in der Bedeutung eines anderen Wortes, das semantisch mit dem
verwendeten Wort in einer realen Beziehung steht” (Lausberg 1960: §565; my italics). —
“[...] entità qualsiasi mediante il nome di un’altra entità che stia alla prima come la causa
sta all’effetto e viceversa, oppure che le corrisponda per legami di reciproca dipendenza”
(Mortara Garavelli 1988: 149).
7. Allusion to The Sermon on the Mount (Matthew 5, 3).
8. A fourth category, changes resulting from “rapports associatifs entre les mots,” concerns
semantic change in word formation.
9. The same direction had already been taken some years earlier by Kristoffer Nyrop (1913:
80): “Une chose ne se présente jamais à l’état isolé, elle est toujours accompagnée de
différentes circonstances qui la complètent ou l’expliquent; elle provoque des associa-
tions d’idées, dues à des rapports de ressemblance, de contiguïté, de cause à effet etc.”
10. For semantic disorders in aphasia, cf. Gurd and Marshall (1993); for syntactic disorders,
cf. De Bleser and Bayer (1993); for a contemporary definition of the different types of
aphasia, cf. Huber et al. (1989: 107–132).
186 Andreas Blank
11. Jakobson’s interpretation of metonymy as relying on syntagmatic contiguity has, how-

ever, influenced several scholars studying metonymy, e.g., Le Guern (1973: 23–28), who
tries to explain metonymies as typicized kinds of ellipsis. To be seen in the same tradition
is Bonhomme’s (1987) division in ‘métonymies actancielles,’ which are grounded in a
kind of deep predicate structure. However, here again the syntagmatic relation is less
evident than the conceptual (paradigmatic) relation. Cf. also Section 5.1.
12. The other possibilities leading to semantic innovations are ‘paradigmatic contrast’ (E bad
‘not good’ > E Slang ‘very good’; see Voßhagen, this volume) and ‘formal (phonetic)
similarity’ of two words; combined with paradigmatic similarity or contiguity the latter
may lead to popular etymology (MHG kruoc ‘jug’ > ‘pub’ (NHG Krug) under the
influence of MLG kruch ‘pub’). Formal and syntagmatic contrast are the basis of all
semiotic systems. Cf. Raible (1981: 6–26); Blank (1997b:131–156).
13. Contiguity as an associative principle also works in several processes of lexical change
other than semantic change, as for example in word formation and in the creation of
idioms. Cf. Blank (1996, 1997a: 96f, 1998, forthcoming); Koch (forthcoming: Section
6.4., this volume).
14. In this sense — but only in this sense — a frame is a prototypical phenomenon (cf.
Fillmore 1977: 56; Koch 1995: Section 7), but, in contrast to Lakoff’s theory of ICMs
(1987: 77–90), I would not say that one more or less prominent concept in a frame could
be considered as a prototypical instance of this frame. In other words: a metonymic
transfer like OGr aggelos ‘messenger’ > ‘angel’ does not occur because an angel is the
prototype of a messenger, but because ‘being a messenger (of God)’ is a salient aspect of
angels.
15. It is obvious that in many cases this relation also exists between the ‘things’ themselves
(cf. the definition given in Section 2), but this is not a necessary condition.
16. This is — grosso modo — what Fauconnier (1984) calls an ‘espace mental,’ and, with a
slightly different accentuation, what Langacker (1984) describes as an ‘active zone.’
17. The famous Dr Johnson observed that oats is “a grain, which in England is generally
given to horses, but in Scotland supports the people” (Johnson, 1785, s.v. oats).
18. One of the rare metonymies in this domain is BritE tea ‘drink,’ ‘meal eaten after the
middle of the afternoon’ (cf. G (informal) Kaffee ‘breakfast’).
19. For a detailed description of the metonymic process, cf. Blank (1997b: 242f). For recent
approaches to the metonymic transfer, cf. Croft (1993); Langacker (1993: 29ff); Koch
(1995: 40f); Radden and Kövecses (this volume).
20. Radden and Kövecses (this volume, esp. Section 2) have a more general understanding of
‘frame’ (which they call, following Lakoff 1987, ‘ICM,’ e.g., ‘Constitution ICM,’
‘Containment ICM,’ etc.). The point they have in common with my interpretation is that
they too distinguish between conceptual relations within ICMs and between a whole ICM
and its part(s).
21. For further details see Blank (1997b: 375–404).
22. Koch (1993: 270f) calls this type of metonymic transfer ‘delocutive semantic change.’
23. After lexicalization, metonymic polysemy may cause misunderstandings, as was the case
for OSp pregón and OPt pregão. In this case, one of the two meanings tends to be
expressed by other means, as for example by derivation: NSp pregonero, NPt pregoeiro
‘messenger.’
24. Cf. Bredin’s (1984: 45) critique of this kind of typology: “For what is presented [...] is not
a true definition, but an enumeration of instances.”
25. E.g., Nixon bombed Vietnam is, according to Fass, an AGENT FOR INSTRUMENT substitution,
which highlights the agent responsible, while Exxon has raised its prices again is an
INSTRUMENT FOR AGENT substitution.
26. Indeed, the idea of case-role substitution in sentences could suggest that metonymy relies
on syntagmatic contiguity. This is clearly not the case: in sentences such as The buses are
on strike the syntactic position of the subject remains the same. Only the semantic role of
this subject has been replaced by another role taken out of the paradigm.
27. Warren’s typology is a good example of how underestimating the importance of the
underlying associations can lead to confusion in distinguishing different types of me-
tonymy, but also metonymy and metaphor: cf. Warren (1992: 64f): “[...] it is difficult to
see exactly what the difference is between metonyms and metaphors, or indeed between
metonyms and implications.”
28. As proper names do not have meanings, transfers based on this conceptual relation are
simple ‘namings’ and not semantic changes.
29. Take, for example, the ACTOR (INVENTOR, etc.) – PRODUCT relation: at the moment of initial
production it belongs of course to the co-presence domain, but later moves to the
succession domain. As is the case with much in language, it all depends on perspective.
30. Cf. the ‘conceptual models’ of Italian word formation as described in Schwarze (1988);
cf. also Blank (1997a: 96f, and 1998).
31. An example from Schwarze (1988: 438–443) is the ‘Tätigkeitsmodell’ [“model of
activity”], which introduces different types of affixation as being conceptually centered
(if not to say “derived”) around the concept ACTIVITY.
32. Gilles Fauconnier, however, has drawn my attention to the fact that in very specific
contexts — for example in a cookery contest — this metonymy would be easily under-
stood.
References
Anonymous
1954 Ad C. Herennium, de ratione dicendi (Rhetorica ad Herennium). London:
Heinemann.
Andor, József
1985 On the psychological relevance of frames. Quaderni di Semantica 6: 212–
221.
188 Andreas Blank
Aristotle
1958 Kategorien. Lehre vom Satz (Organon I/II). Translated and edited by Eugen
Rolfes. Hamburg: Meiner.
Blank, Andreas
1993 Polysemie und semantische Relationen im Lexikon. In W. Börner, K.
Vogel (eds.), Wortschatz und Fremdsprachenerwerb. Bochum: AKS-
Verlag, 22–56.
1996 Tyson est aux anges – Zur Semantik französischer Funktionsverbgefüge.
Zeitschrift für französische Sprache und Literatur 106: 113–130.
1997a Il senso di una semantica dei prototipi e dei frames: osservazioni deco-
struttive e ricostruttive. In M. Carapezza, D. Gambarara, F. Lo Piparo
(eds.), Linguaggio e cognizione. Rom: Bulzoni, 89–103.
1997b Prinzipien des lexikalischen Bedeutungswandels am Beispiel der roma-
nischen Sprachen. Tübingen: Niemeyer.
1998 Kognitive italienische Wortbildungslehre. Italienische Studien 19: 5-27.
forthcoming Pathways of lexicalization. In M. Haspelmath et al. (eds.), Language
Typology and Universals: An International Handbook. Berlin, New York:
Mouton de Gruyter.
Bonhomme, Marc
1987 Linguistique de la métonymie. Bern, Frankfurt: Lang.
Bredin, Hugh
Coseriu, Eugenio
1958 Sincronía, diacronía e historia: El problema del cambio lingüístico.
Montevideo: Universidad de la República.
Croft, William
De Bleser, Ria, Joseph Bayer
1993 Syntactic disorders in Aphasia. In G. Blanken et al. (eds.), Linguistic
Disorders and Pathologies: An International Handbook. Berlin, New
York: Mouton de Gruyter, 160–169.
Dik, Simon C.
1977 Inductive generalisation in semantic change. In P.J. Hopper (ed.), Studies in
Descriptive and Historical Linguistics. Festschrift for Winfred P. Lehmann.
Duchácek, Otto
1967 Précis de sémantique française. Brünn: Univ. J.E. Purkyne.
Dumarsais, César Chesneau
1818 Les Tropes. Edited and commented by Pierre Fontanier. Paris: Belin-le-
Prieur (Reprint Geneva: Slatkine 1967).
Fass, Dan
1991 Metonymy, case role substitution and sense ambiguity. In: D. Fass et al.
(eds.), Proceedings of the IJCAI-91 Workshop on Computational Ap-
proaches to Non-Literal Language (Sydney, Australia). Boulder, CO: Uni-
versity of Boulder, 42–51.
Fauconnier, Gilles
1984 Espaces mentaux: Aspects de la construction du sens dans les langues
naturelles. Paris: Éditions de Minuit.
Fillmore, Charles
1968 The case for case. In E. Bach, R.T. Harms (eds.), Universals in Linguistic
Theory. New York: Holt, Rinehart & Winston, 1–90.
1975 An alternative to checklist theories of meaning. Proceedings of the 1st
Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley
Linguistics Society, 123–131.
1977 Scenes-and-frames semantics. In A. Zampolli (ed.), Linguistic Structures
Processing. Amsterdam and Philadelphia: Benjamins, 55–81.
254.
Fontanier, Pierre
1968 [1830] Les figures du discours. Paris: Flammarion.
Garcia Arance, Maria del Rosario
1979 Semantica de la metonimia y de la sinecdoque. Valladolid: Universidad de
Valladolid.
Gurd, Jennifer M., John C. Marshall
1993 Semantic disorders in aphasia. In G. Blanken et al. (eds.), Linguistic Disor-
ders and Pathologies: An International Handbook. Berlin, New York:
Mouton de Gruyter, 153–160.
Happ, Heinz
1985 ‘Paradigmatisch’ – ‘syntagmatisch’: Zur Bestimmung und Klärung zweier
Grundbegriffe der Sprachwissenschaft. Heidelberg: Winter.
Huber, Walter, Klaus Poeck, Dorothea Weniger
1989 Aphasie. In K. Poeck (ed.), Klinische Neuropsychologie. Second edition.
Stuttgart, New York: Thieme, 89–137.
Jakobson, Roman
1971 [1956] Two aspects of language and two types of aphasic disturbances. In
R. Jakobson, M. Halle, Fundamentals of Language. Second edition. The
Hague, Paris: Mouton, 67–96.
Johnson, Mark
1987 The Body in the Mind: The Bodily Basis of Reason and Imagination.
Chicago: The University of Chicago Press.
Johnson, Samuel
1785 A Dictionary of the English Language: In Which the Words Are Deduced
from Their Originals, and Illustrated in Their Different Significations. 2
vols. Seventh edition. London.
Keller, Rudi
1994 Sprachwandel: Von der unsichtbaren Hand in der Sprache. Tübingen:
Francke.
Koch, Peter
1993 Kyenbé – Tyonbo: Wurzeln kreolischer Lexik. In Ch. Foltys, Th. Kotschi
(eds.), Berliner Romanistische Arbeiten. Für Horst Ochse. Berlin: Freie
Universität, 259–287.
190 Andreas Blank
1995 Der Beitrag der Prototypentheorie zur Historischen Semantik: Eine

kritische Bestandsaufnahme. Romanistisches Jahrbuch 46: 27–46.
1997 D’où viennent les substantifs spatio-locaux? In H. Dupuy-Engelhardt, M.-
J. Montibus (eds.), L’organisation lexicale et cognitive des dimensions
spatiale et temporelle. Actes d’EUROSEM 1996. Reims: Presses
Universitaires de Reims, 107–122.
forthcoming Ein Blick auf die unsichtbare Hand: Kognitive Universalien und
historische romanische Lexikologie. In Th. Stehl (ed.), Unsichtbare Hand
und Sprecherwahl: Typologie und Prozesse des Sprachwandels in der
Romania. Tübingen: Narr.
Lakoff, George
1984 Active zones. Proceedings of the 10th Annual Meeting of the Berkeley
Linguistics Society. Berkeley: Berkeley Linguistics Society, 172–188.
Lausberg, Heinrich
1960 Handbuch der literarischen Rhetorik. München: Hueber.
Le Guern, Michel
Minsky, Marvin
1975 A framework for representing knowledge. In P.H. Winston (ed.), The
Psychology of Computer Vision. New York: McGraw-Hill, 211–277.
Mortara Garavelli, Bice
1988 Manuale di retorica. Mailand: Bompiani.
Nerlich, Brigitte
forthcoming Synecdoche: a trope, a whole trope, and nothing but a trope? In N.
Norrick, A. Burkhardt (eds.), Tropic Truth. Berlin: Mouton de Gruyter.
Nyrop, Kristoffer
1913 Grammaire historique de la langue française. Vol. 4: Sémantique. Kopen-
hagen: Gyldendal.
Quintilian, M. Fabius
1886 Institutionis oratoriae. Edited by Ferdinand Meister. Leipzig: Freytag.
Raible, Wolfgang
1981 Von der Allgegenwart des Gegensinns. Zeitschrift für Romanische Philo-
logie 97: 1–40.
Roudet, Léonce
1921 Sur la classification psychologique des changements sémantiques. Journal
de Psychologie 18: 676–692.
Ruwet, Nicolas
1983 Synekdochen und Metonymien. In: A. Haverkamp (ed.), Theorie der Meta-
pher. Darmstadt: Wissenschaftliche Buchgesellschaft, 253–282 (Original:

1975, Synekdoques et métonymies. Poétiques 6: 371–388).
Schifko, Peter
1979 Die Metonymie als universales sprachliches Strukturprinzip. Grazer
Linguistische Schriften 10: 240–264.
Schwarze, Christoph
1988 Grammatik der italienischen Sprache. Tübingen: Niemeyer.
1986 Relevance: Communication and Cognition. Oxford: Blackwell.
Stern, Gustaf
1931 Meaning and Change of Meaning. Bloomington: Indiana University Press.
Taylor, John R.
1989 Linguistic Categorization. Oxford: Clarendon.
Ullmann, Stephen
Warren, Beatrice
1992 Sense Developments: A Contrastive Study of the Development of Slang
Senses and Novel Standard Senses in English. Stockholm: Almquist and
Wiksell.
1995 What is metonymy? Paper read ad the XIIth ICHL, Manchester 1995.
Metonymic Bridges in Modal Shifts
Louis Goossens
University of Antwerp
1. Introduction
1.1. Some background
This paper will be concerned with the development of the English modal must
from deontic modality to epistemic modality, more particularly with the
way(s) in which this development came about. This shift is not restricted to
English must, of course, or to English, for that matter. As Bybee et al. (1994:
Ch. 6) have pointed out, the more general shift is one from agent-oriented to
epistemic modality, and it is attested in modals in a large number of languages
across the world. The development can be interpreted as involving:
(a) a metaphorical shift (cf., for example, Sweetser 1990 for a mapping of the
socio-physical domain onto the epistemic domain),
(b) a context-induced interpretation (Heine 1995) or the conventionalization
of an implicature (Traugott 1989).
As Heine has observed, the two interpretations differ in level of abstraction:
(a) is a macro-level view of the change, (b) is to be situated at the micro-level.
1.2. Metaphor from metonymy
In Goossens (1990) and also in Goossens et al. (1995) I have argued that there
is considerable interaction between metonymy and metaphor. The most fre-
quent pattern is what I have called ‘metaphor from metonymy,’ which is
exemplified in (1).
194 Louis Goossens
(1) “Oh dear,” she giggled, “I’d quite forgotten.”

In the metonymic interpretation she giggled denotes a state of affairs (if you
wish, a scenario) in which the subject said something lightheartedly ‘while’
giggling. Giggled thus expresses the complex ‘giggled + said lightheartedly.’
In the metaphorical reading, on the other hand, we take it that she said
something lightheartedly ‘as if’ giggling; thus giggling is mapped onto light-
hearted linguistic action. But also in the metaphorical reading, we are aware
that it has the metonymic interpretation as a possible basis. This can be
represented as in Figure 1.
In the ellipse on the left, two domains, A and B, are intertwined in one
complex matrix domain. This symbolizes the metonymic interpretation. On
the right-hand side, the two domains are discrete: A (‘giggling’) is mapped
onto B (‘say in a lighthearted tone’). As the broken arrow signals, however,
there is still a conceptual link with the possibility that the two domains (or
actions) may be intertwined in one complex domain (the combination of the
actions of giggling and light-hearted speaking).
Applying this to the modal shift under discussion, the compromise be-
tween the context-induced and the metaphorical interpretation can be viewed
similarly. As Figure 2 indicates, there is a metonymic conceptualization in
which the deontic sense (D) and the epistemic one (E) are intertwined. The
metaphorical interpretation is shown on the right with the deontic domain
being mapped onto the epistemic one. But, at least initially, there must have
been an understanding that the two domains could be relevant together (as the
broken arrow indicates).
A B A B
Figure 1. Metaphor from metonymy
D E D E
Figure 2. From deontic to epistemic

2. Research question
With respect to English must, the question arises whether there are sufficient
grounds to consider the development of its epistemic sense as a metaphor from
metonymy. In the following quotation, Traugott and König clearly argue that
the change was induced from contexts which are open to both a deontic and an
epistemic interpretation and which would therefore give it a metonymic basis:
[...] must in the epistemic sense of ‘I conclude that’ derived from the obliga-
tive sense of ‘ought to’ by strengthening of conversational inferences and
subjectification. If I say She must be married in the obligation sense, I invite
the inference that she will indeed get married. This inference is of course
epistemic, pertaining to a state of affairs that is anticipated to be true at some
later time. (1991: 209)
Bybee et al. (1994: 196) generally favor the thesis that the overall change from
agent-oriented to epistemic “involves the conventionalization of an
implicature, by which the inferences that can be made from the meaning of a
particular modal become part of the meaning of that modal.” Also here, in the
terminology adopted, we find acceptance of the metonymic origin. In the
specific case of English must, however, they are opposed to such an origin:
Since the epistemic use of must arises in contexts with aspectual interpreta-
tions distinct from the obligation uses, it appears that metaphor may be at
work in this change. Metaphorical change involves a shift to a different
domain — in this case from the domain of social obligations and physical
necessities applied to an agent, to the epistemic domain that speaks of the
necessary conditions under which a proposition can be true. (Bybee et al.
1994: 201)
The argument focuses on the essential difference between a state of affairs that
combines with deontic must, which is typically dynamic or at least controlled,
whereas for epistemic must we need a ‘State,’ that is, a state of affairs
conceived of as non-dynamic and non-controlled.1 As I have pointed out in
Goossens (1987), the qualification of a controlled/dynamic nuclear predica-
tion by such aspectual notions as Imperfective/Progressive or Perfect gives it a
State-character.
Traugott and König’s example would seem to indicate that there are
contexts which are compatible with both the deontic and the epistemic inter-
pretation. But then their example is clearly a constructed one. It remains to be
demonstrated whether language use actually provides contexts in which the
196 Louis Goossens
two interpretations are simultaneously relevant, and more generally, what the
precise basis for the development may have been. In what follows, I will study
contexts which are relevant to a better understanding of this modal shift in
must, first in contemporary English (Section 3), then in older language stages
(Section 4).
3. Metonymic bridges for contemporary English must
3.1. The data
I have analyzed samples of must in context from three contemporary data

bases: Brown, Lancaster-Oslo-Bergen (LOB), and London-Lund (LOLU).2
The first two are written corpora (American English and British English,
respectively), the third is a spoken British English corpus. Table 1 arranges
these data into five categories:
Deontic/Necessity: This comprises instances which express an obligation
(deontic uses) as well as a more general necessity. The deontic uses include
instances in which the obligation is imposed/backed by the speaker (which is
the rule) and instances in which this is not the case, but we have not separated
them in our analysis.
Inferable necessity: Another label could be ‘objective epistemic’ or ‘objective
inference’ (must expresses an inference which is not defeasible; the speaker
draws a ‘logical’ conclusion; the context contains the necessary information
for the hearer to arrive at the inference presented by the speaker).
Transitional: These are indeterminate uses (especially with respect to the
difference between deontic and epistemic; the context permits the interpreta-
tion to go either way).
Epistemic: ‘Subjective epistemic’ or ‘subjective inferential’ uses; the infer-
ence is the speaker’s; the hearer cannot be assumed to have the (contextual)
information to arrive at this inference ‘objectively.’
Undecidable: The context does not permit any conclusive interpretation; these
are incomplete utterances from the spoken corpus; instances of this kind are
listed under ‘?’ in the last column of Table 1.
Table 1. Must in three contemporary data bases

Deontic/Necessity Inferable Transitional Epistemic ?
BROWN
(must, 69) 54 3 – 12 –
LOB
(must, 75) 48 5 7 15 –
(must not, 2) 2 – – – –
LOLU
(must, 121) 60 1 3 52 5
(must not, 29) 23 – – 4 2
Before we tackle the instances that are directly relevant to our purposes, allow
me to make a number of preliminary observations. For obvious reasons the
undecidable (‘?’) cases do not receive any further discussion. As far as the
deontic/necessity examples are concerned, they outnumber the epistemic ones
in all the samples (but more clearly in the written ones). Among them there were
23 instances of must not in the spoken data, but only two in the written samples
(more specifically in LOB). Of special interest are the four cases of must not
which I have assigned to ‘subjective epistemic,’ because we do not expect
epistemic readings of must to occur with not (cf. Coates 1983: Section 4.1.3).
They are all in the interrogative-negative (three out of four come in tags) and
have the effect of trying to elicit agreement of the addressee with an inference
of the speaker; (2) is a case in point. Neither this instance, however, nor the other
examples of this type, are relevant to the issue that concerns us here.
(2) I like to think of those days # and how tough it was # for the average
Englishman #. what a hard life they must have had # – and mustn’t
there be #. endless stories about this mansion # –3
It is the ‘inferable necessity’ and the ‘transitional’ uses that are our focus.
They will be discussed in the next subsection.
3.2. Bridges
As can be observed in Table 1, there were 19 instances in our sample (out of a

total of 296) which somehow provide us with bridges between the deontic and
198 Louis Goossens
the (subjective) epistemic uses. Of these, 9 are examples of ‘inferable neces-

sity’ while 10 are ‘transitional’ uses.
Let us first consider the instances in which both a deontic and an
epistemic (inferential) reading are relevant (the ‘transitional’ uses). As it
happens, they all occur in the British English samples, though I do not think
that any further significance should be attached to this. (3) and (4) exemplify
such transitional uses.
(3) (LOLU)
<A> and conversation. went like this. this sort of conversation
(m – have you noticed president #. that. (m – the boiled eggs
# at Sunday *breakfast* – always
<b> *(laughs)*
<A> hard – and president said – ah well – the simple truth is that.
if you’re going to boil eggs. communally # they must be
hard
(4) (LOB)
“Well, I cannot say that I approve. You must remember that I am
taking your aunt’s hospitality, and, if your plans go right, on
entirely false pretences.[...]”
In (3) they must be hard is compatible with an interpretation ‘one has to/must
make them hard’ (which is deontic) and with an epistemic reading ‘they will
of necessity be hard.’ The epistemic interpretation is objective rather than
subjective (the inference follows from the fact that the eggs are boiled commu-
nally, which is contextually also available to the hearer; it is in that sense not a
subjective/speaker inference). In (4), you must remember can be paraphrased
as ‘I want you to remember’ (deontic), but also as ‘I’m sure you will remem-
ber’ (subjective epistemic; in this case the inference is to be assigned to the
speaker).
On the other hand, there are bridging instances of a different kind, as
illustrated in (5) and (6).
(5) (LOLU)
what Fan didn’t realize not being a lawyer #. or a lawyer’s wife #-
that apparently the wife was put on probation # so that Fan didn’t
realize that # that she must also have been up before the court # you
can’t be put on probation # not unless you’re guilty
(6) (BROWN)
Rossi and Ferguson have been across the street, talking to the kid.
They’ve found some sort of new evidence, a bundle of clothes or
something, and it must link the kid even stronger to the crime. Why
won’t you accept facts? The two kids were together a lot, they [...]
Both (5) and (6) present an inference for which the evidence is available
contextually, so that the hearer can arrive at the same conclusion as the
speaker — although it remains the case that it is the speaker who draws the
conclusion (since he/she does the talking). Some five percent of the data have
to be accounted for in this way. Although these instances are not metonymic
between deontic and epistemic as the ‘transitional’ cases are, they may never-
theless be considered as stepping stones to the full-scale subjective epistemic
use which is typical of present-day English. The diachronic data to be dis-
cussed in Section 4 will confirm the transitional character of this type of usage
toward a full-scale (subjective) epistemic use of English must.
The following conclusions can be drawn from this brief discussion:
(i) Although the normal understanding of ‘epistemic’ implies that it is a
subjective inference of the speaker (this is, for instance, what is indicated by
‘subjectification’ in the above quotation from Traugott and König), we also
have to consider that inferences may be ‘objective,’ i.e., the inferential ground
is understood to be available to both speaker and hearer, as in instances (5) and
(6). Without claiming that these ‘objective epistemic’ instances provide us
with a metonymic bridge between the deontic and the epistemic usage, I
would like to hypothesize that somehow they must have played a part in the
shift to the full epistemic sense typical of contemporary epistemic must.
(ii) Occasionally an utterance/sentence with must is compatible with both
a deontic and an objective, or subjective, inferential reading. A condition for
such a ‘dual’ interpretation is that the state of affairs represented can be taken
to be both non-controlled or controlled (or at least controllable). This is the
case for remember in instance (4) (something may spontaneously come to
mind, or may be the result of a conscious effort to recall it); the normal
interpretation of be hard in (3) is that it is a State, but in the deontic reading
assigned to it above this is viewed as controllable (the subject will (have to)
treat them in such a way that they become/are hard). Uses like these could be
regarded as the metonymic bridge we have been looking for.
Armed with these synchronic observations let us now move on to some
diachronic data to see whether and how the rise of epistemic must confirms the
200 Louis Goossens
picture that both ‘inferable necessity’ and ‘transitional epistemic/deontic’

instances play a part in the development of an epistemic sense of English must.
4. ‘Metonymic bridges’: older language stages
4.1. The data
Using the categories that proved to be relevant in our investigation of the

present-day English data, we have looked at the instances of must (mote,
mooste, ...) in successive samples of the Helsinki Corpus of English Texts.
Obviously the Helsinki samples are too restricted to provide us with the very
first instantiations of the new development in focus here (which, moreover, may
be expected to have occurred in unrecorded spoken language anyway), but it
may reasonably be assumed that such an analysis will help us to understand how
obligational/necessity must came to develop an epistemic sense.
4.2. Beginnings of epistemic must
As it happened, the first clear uses of subjective epistemic must were found in
sample Early Modern English 3 (1640–1710). The four instances found are
listed here as (7)–(10).
(7) (CEDIAR3B4 , 457–461) (EMOE3)
That was opposed & spoke against with such vehemency by my L.
Clarendon (her owne Unkle) as putt himm by all preferments,
which must doubtlesse, [{have{] been greate, as could have ben
given him5
(8) (CEOFFIC3, 180–184) (EMOE3)
The manner of doing it, so unsought for and unexpected, must take,
with any good mind, more then the thing itself, and I am sure the
sense must always last with mee.
(9) (CEPRIV3, 1200–1204) (EMOE3)
[...] I did thinke never to imploy him, but he being all your tailors I
have altered my resolution; beleev I must not have a gown again
this seaven yeare this cost soe much; four pound four shillings the
outside cost me ready mony [...]
(10) (CETRAV3B, 342–346) (EMOE3)

[...] were put to such stress, that had not their Armado come to their
relief, they must have desisted their Enterprize: Upon arrival
whereof the new recruits gained so much on the Arab Governors,
that […]
In (7), (8) and (9) the epistemic reading is underscored by other epistemic
expressions: doubtlesse in (7), I am sure in (8), and beleev [= ‘I believe’] in (9).
(10), and (7) in its emended version, are counterfactual. For (10) the paraphrase
may be either would have desisted or would have had to desist; in the second
paraphrase the epistemic reading still contains a necessity ingredient, but that
does not invalidate the fact that the reading is also the expression of a subjective
inference. In all cases the combining state of affairs is non-controlled; in
instance (10) as a result of the qualification by the Perfect aspect.
In our search for stepping-stones towards this infrequent, but already
firmly established, usage of an epistemic sense in EMOE3, let us now first
provide a general survey of the uses of must in EMOE3, as well as in the four
samples before EMOE3, viz. EMOE2 (1570–1640), EMOE1 (1500–1570),
EME4 (1420–1500) and EME3 (1350–1420).
4.3. Must in five successive Helsinki samples
Table 2 gives the distribution of the different uses of must, classifying them into
the categories used for the analysis of the present-day items (deontic/necessity,
inferable necessity, transitional, and (subjective) epistemic). For the Middle
Table 2. Must in five successive periods (Helsinki Corpus)

Wish Deontic/ Inferable Transitional Epistemic
Necessity Necessity
ME3(1350–1420)
(Total: 63) 2 43 18 0 0
ME4(1420–1500)
(Total: 78) 4 70 4 0 0
EMOE1(1500–1570)
(Total: 88) 0 71 15 2 0
EMOE2(1570–1640)
(Total: 191) 0 174 15 2 0
EMOE3(1640–1710)
(Total: 173) 0 131 22 16 4
202 Louis Goossens
English data we need an additional category ‘wish.’ This use builds on the older
permission sense; it is clearly recessive in Middle English and disappears
completely in the EMOE samples. As such, it will not concern us any further.
A first observation is the absolute predominance of the category Deontic/
Necessity, which is much more pronounced than in present-day English. As
pointed out above, it also includes instances which as such do not express an
obligation. (11) and (12) illustrate this more general necessity use.
(11) (CMGAYTRY, 360–364) (ME4)
[...] for-thynkynge þat we hafe of oure syn, with-owtten will of
thoghte to turne agayne to it. And þis sacrament must haue thre
thynges: — Ane es, sorowe in oure herte þat we hafe synnede.[...]
(12) (CEBOETH2, 304–308) (EMOE2)
[...] All that is so, long must last & holde togither, as it is one, but
must needes perish & decay, whan so it leaves to be; [...]
In addition, note that in ME3, ME4 and EMOE1 instances, in which the
speaker can be assumed to be in authority, or at least strongly associates
himself/herself with some other authority/necessity source, are markedly
lower in frequency than those where this is not the case. Among those that do
not involve speaker authority, must is still regularly used in instances where it
is a past tense (is used in past time main clause contexts), a usage which
decreases as we come closer to present-day English.
The most important observations to be made, however, relate to the other
three categories (inferable necessity, transitional and epistemic). As was
pointed out in the preceding section, the fully (subjective) epistemic usage
does not come in until EMOE3 (four instances). The ‘transitional’ usage is
found from EMOE1 onwards, but is lacking in the ME samples. Inferable
necessity, however, is well established from our earliest sample (ME3) on-
wards, and remains so in all the subsequent ones. It is there that inferential
must finds its starting-point, as I will try to demonstrate in the next section.
4.4. Inferable necessity: a ‘bridge’ to inferential must
Let us first offer a few instances of ‘inferable necessity,’ one from each period.
(13) (CMBOETH, 375–379) (ME3)
But I have wel concluded that blisfulnesse and God ben the
sovereyn good; for which it mote nede be that sovereyne blisfulnes
is sovereyn devynite
(14) (CMCAXPRO, 348–352) (ME4)

The fruytes of vertue ben Immortall/ Specyally whanne they ben
wrapped in the benefyce of hystoryes/ Thenne it muste followe/
That it is mooste fayre to men Mortalle to suffre labours and payne/
for glorye and fame Jnmortalle
(15) (CESCIE1B, 729–733) (EMOE1)
Also the third square which is N.H.L.P. must of necessitee be equal
to the square of C.D. and F.B., bicause those lines be so coupeled
that euery couple are equall in the seuerall figures
(16) (CEEDUC2A, 503–507) (EMOE2)
I do fully see the evidence of all which you have said, and therefore
I must needs be perswaded of it. I do heartily thanke God for it, [...]
(17) (CEHAND3B, 312–316)(EMOE3)
and consequently the bud inoculated in the morning, must be more
likely to grow, having the whole days plenty sap to invite it to unite
with
All these instances of must present a conclusion of some sort, an inference
drawn objectively, rather than a subjective epistemic assessment of the com-
bining state of affairs. The state of affairs is a State in (13), (15), and (17), a
Process6 in (14) and (16). Note that supporting adverbials underscore the
inferential character: nede in (13), of necessitee in (15), therefore and needs in
(16), consequently in (17), where in addition the participial construction adds
a reason for drawing the conclusion.
Given my claim that ‘inferable necessity’ provides us with the interface
between the categories ‘deontic/necessity’ and ‘(subjective) epistemic,’ two
questions must be asked next: (i) How do we go from ‘deontic/necessity’ to
‘inferable necessity’? In other words, what is the contextual — the met-
onymic, if you wish — bridge between these two categories? (ii) What is the
transition like from ‘inferable necessity’ to ‘(subjective) epistemic’? We
tackle these questions in Sections 4.5. and 4.6., respectively.
4.5. From ‘deontic/necessity’ to ‘inferable necessity’
Within the category ‘deontic/necessity’ it is not instances of must expressing

obligation, but rather those in which must expresses a more general type of
necessity not rooted in the field of social pressure that must be our starting-
point. The following examples, again one from each sub-period, illustrate this
204 Louis Goossens
type of general necessity.

(18) (CMCTPROS, 567–571) (ME3)
Wherto and why burieth a man his goodes by his grete avarice, and
knoweth wel that nedes moste hy dye? For deeth is the ende of
every man [...]
(19) (CMSIEGE, 447–451) (ME4)
[...] for þer is no wattur ner þe cyty þen flem Jurdane. And when
þey fynd no freche watur, þey most nedys fle, and we wyl follo and
sley þeme all
(20) (CEFICT1B, 434–438) (EMOE1)
[...] rebuking a wyld roge because he went idelly about, he shewed
me that he was a beggar by inheritance — his Grandfather was a
beggar, his father one, and he must nedes be one by good reason.
(21) (=(12)) (CEBOETH2, 304–308) (EMOE2)
[...] All that is so, long must last & holde togither, as it is one, but
must needes perish & decay, whan so it leaves to be; [...]
(22) (CESERM3A, 565–569) (EMOE3)
But now a very acute and scholastical man that would argue that
God must needs have done whatever he fancies convenient for the
World should be done, might [...]
None of these expresses an obligation, at least in the sense that no clear
authority source is involved. The states of affairs range from State (in (20); but
‘being a beggar’ can be taken to be controllable as well as non-controllable) to
Process ((18) and (21)) and Action ((19) and (22), but note that in (22) this
Action is qualified by Perfect and that as a result the overall state of affairs
qualifies as a State).
Although none of these is a good instance of an (objective) inference
either, I would nevertheless like to argue that an inferential ingredient, be it to
varying degrees, can be read into them. In that sense they provide us with
stepping stones from ‘(deontic)/necessity’ to ‘inferable necessity.’ In all of the
instances we find the adverb needs (nedes, nedys, needes), which we also
found in some of the ‘inferable necessity’ cases. In (18) the general necessity
expressed is a subclause argument of ‘he knoweth’; it is as if this general
necessity is presented as inferable from our, as well as the matrix clause
subject’s, knowledge of the world. In (19) the necessity for them to flee can be
inferred from the absence of fresh water (mentioned in the preceding clause).
In (20) the necessity for the subject to be a beggar follows ‘by good reason’
from the fact that he belongs to successive generations of beggars, as is made
explicit in what precedes. Also (21) and (22) come in a context in which a
reasoning is presented.
General necessity instances like the ones discussed here can therefore be
taken to offer a natural transition between necessity and objective inference/
inferable necessity.
4.6. ‘Transitional’ uses: from (inferable) necessity to epistemic usage
As was pointed out in Section 4.3., instances which are transitional towards
(subjective) epistemic usage do not occur until the EMOE samples, and only
sparingly in EMOE1 and EMOE2. The following instances are representative
of the kind of transition we came across; the first (EMOE1) instance is from an
educational treatise, the second (EMOE2) from The Trial of Sir Walter Raleigh,
and the other two (EMOE3) from Pepys’s Diary and The Earl of Essex’s
Correspondence, respectively. I add a brief discussion to each instance.
(23) (CEEDUC1B, 71–75) (EMOE1)
A childe shall learne of the better of them, that, which an other daie,
if he be wise, and cum to iudgement, he must be faine to vnlearne
againe
Contrary to what the reader may have been led to expect on the basis of Section
4.4., (23) does not build on an inferable necessity, but rather on a general
necessity with a deontic tinge, since the speaker can be assumed to give his
backing to this necessity. Note in this respect that be faine to vnlearne againe
is a process that may come about in an uncontrolled as well as in a controlled
way. At the same time, the example can be interpreted epistemically, but the best
paraphrase in that interpretation is will rather than must. In contemporary
English, the instance with will would be predictive rather than inferential.
(24) (CETRI2B, 73–77) (EMOE2)
Nay, I will prove all: thou art a Monster; thou hast an (Ênglish^)
Face, but a (^Spanish^) Heart. Now you must have Money:
(Âremberg^) was no sooner in (Êngland^) (I charge thee (^Ra-
leigh^)) but thou incitedst (^Cobham^) to go unto him, [...]7
206 Louis Goossens
In (24), now you must have Money is part of an argumentation in a trial; now
indicates the next step in this argumentation. Must is used to express the
speaker’s inference (as such it is subjective), but the speaker later goes out of
his way to give the grounds for his inference, which gives the inference an
‘objective’ backing. In other words, ‘inferable necessity’ supports the
speaker’s inference.
(25) (CEDIAR3A, 312–315)(EMOE3)
We were full in discourse of the sad state of our times. And the
horrid shame brought on the King’s service by the just clamours of
the poor seamen. And that we must be undone in a little time.
This is a nice mixture of inferable necessity and epistemic use (again predic-
tive; here the paraphrase is would rather than will because this is reported
speech or, more specifically, erlebte Rede ‘semi-indirect speech’). A deontic/
general necessity reading is out of the question here.
(26) (CEOFFIC3, 735–739) (EMOE3)
[...] although in a matter so nice as this yr Excellency must needs
bee ye best Judge. I believe ye Parliament is like to sitt longer [...]
On the one hand, this may express the speaker’s (subjective) judgment, but
can also be interpreted as expressing a more objective necessity (in line with
the presence of the adverb needs). On the other hand we can take it as
expressing the speaker’s advice, which gives it a (speaker-backed) deontic
overtone.
Summing up, these transitional uses confirm that the epistemic usage of
must typically builds on inferable necessity must: this is illustrated by (24),
(25) and, be it less obviously, in (26). However, there are transitional cases
which exhibit other features. Besides the subjective epistemic ingredient, a
deontic interpretation proved possible in (23) and perhaps also in (26). Note
also that in three out of four instances, must is rather prospective and para-
phrasable in contemporary English by will/would.
The conclusion at this point seems to be that the rise of epistemic must is
rooted in the ‘inferable necessity’ uses, but that complexes of deontic and
epistemic uses may have played a supportive role. In the next section we add
another factor that appears to have been operative to complete this picture.
4.7. Subjectification
A factor signaled by Traugott and König (cf. the quotation in Section 2.) is
subjectification. For Langacker this is a crucial ingredient in the semantic
make-up of the present-day modal auxiliaries of English; it is this
subjectification that turns them into ‘grounding predications’ (Langacker
1990 and 1991: 6.3).8 Although a detailed account of how must develops into
a subjectified item is outside the scope of this paper, I would like to confirm
the basic correctness of the claim that subjectification played a part in the rise
of epistemic must. More particularly, it can be argued that it is subjectification
in the deontic area which paved the way for epistemic must, which, as we have
defined the category, is necessarily subjectified (it reflects the speaker’s
subjective inference).
An initial analysis of the deontic/necessity uses in our data revealed that,
although speaker-backed deontic uses are already firmly established in the
ME samples, it is not until we reach EMOE1 that those subjectified deontic
instances begin to outnumber the non-subjectified ones in this category, and
that in EMOE3, where the first clear instances of epistemic must occur, the
ratio is roughly 75% (for the speaker-backed deontic uses) as opposed to 25%
(for the ‘not-speaker-backed’ deontic/necessity instances). There is no doubt,
it seems to me, that the increase and, finally, the predominance of subjectified
uses in the deontic area, made the rise of an essentially subjectified epistemic/
inferential must possible.
5. Conclusion
(i) The rise of epistemic must took place against the background of a gradual
meaning development of must from a general necessity sense to the expression
of inferable necessity, which amounts to an objective, non-defeasible infer-
ence. Early instances which are transitional to epistemic must still show
elements of an objective inference.
On the other hand, another type of transition is found in instances where a
deontic and an inferential interpretation are possible simultaneously. In those
cases, however, must is prospective and parapharasable by present-day En-
glish will, rather than must. Although they must have contributed to the rise of
epistemic must and indicate that we cannot explain everything in terms of a
208 Louis Goossens
single pattern of contextual shift, they are less essential for our understanding
of how present-day inferential must developed. What is crucial, however, is
the marked increase in subjectification in the deontic senses. This was an
essential factor in the development of subjectified inferential/epistemic must.
(ii) With respect to the research question formulated in Section 2, it is
clear that our empirical findings speak strongly against a purely metaphorical
shift hypothesis. The shifts that we have observed are minimal and very
gradual. There is no evidence that there was at any given point in the develop-
ment of English must a single conceptual shift whereby an element from the
sociophysical domain was mapped onto the epistemic domain.
(iii) More importantly, at least from the point of view of a collective
volume whose focus is on the workings of metonymy, the view that must
extended its meaning by way of contextualizations in which the deontic and
the epistemic sense are intertwined (in other words, the view that the change
from deontic to epistemic must is a case of conceptual metonymy) cannot be
upheld either. The concatenation that we have witnessed is one that involves
shifts that do not reach a cognitive salience which can be interpreted as the
conceptualization of one ‘element’ in terms of another within the same (com-
plex) domain. Moreover, more than one type of minimal shift appears to have
contributed simultaneously.
(iv) Finally, we have emphasized the important contribution of a more
global shift, namely that of subjectification in the deontic area. This is a
conceptualization and grammaticalization operation which eventually af-
fected the inferential domain that had become part of the semantic make-up of
must. This process of subjectification is in need of further investigation, but
beyond the scope of what I set out to clarify in this paper.9
Notes
1. My terminology for the differentiation into states of affairs is in line with that of
Functional Grammar (cf. Dik 1989).
2. The samples were collected randomly by Ludo Lejeune. For details, cf. Lejeune (1995
and 1996).
3. The quotations from LOLU are presented in a somewhat simplified form. The diacritic
signs retained are #, which marks intonation units, * *, which signals overlap, and those
that indicate pauses of varying lengths.
4. The abbreviations for the different texts are taken over from the Helsinki Corpus. The
sources, in the order of their first occurrence in the examples quoted as (7)–(26), are as
follows: CEDIAR3B = Evelyn, The Diary of John Evelyn; CEOFFIC3 = Letters, Non-
private; CEPRIV = Letters, Private; CETRAV3B = Fryer, A New Account of East India;
CMGAYTRY = Gaytridge, Dan Jon Gaytridge’s Sermon; CEBOETH2 = Elizabeth,
Boethius; CMBOETH = Chaucer, Boethius; CMCAXPRO = Caxton, The Prologues and
Epilogues; CESCIE1B = Record, the Path-way … of Geometrie; CEEDUC2A = Brinsley,
Ludus Literarius or the Grammar Schoole; CEHAND3B = Langford, Plain and Full
Instructions to Raise All Sorts of Fruit-trees; CMCTPROS = Chaucer, The Parson’s Tale;
CMSIEGE = The Siege of Jerusalem in Prose; CEFICT1B = Harman, A Caveat … for
Commen Cursetors; CESERM3A = Tillotson, Sermons; CEEDUC1B = Ascham, the
Scholemaster; CETRI2B = The Trial of the Earl of Essex; CEDIAR3A = The Diary of
Samuel Pepys. For further details, see Kytö (1993).
5. The notation [{.....}] indicates an italicized emendation in the source text. Note that the
text makes sense with and without the emendation; without the emendation we have to
read been as the equivalent of Modern English be.
6. Again in the terminology of Functional Grammar, i.e., dynamic, but not controlled. Note
that in (16) the Process character is the result of passivization.
7. The notation (^.....^) in the Helsinki Corpus indicates a typographical shift from the basic
font in the source text.
8. For a critical discussion with respect to contemporary must, cf. Goossens (1996: 2.3.3)
9. This process of subjectification is in focus in Goossens, forthcoming.
References
Bybee, Joan, Revere Perkins, William Pagliuca

1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages
of the World. Chicago: The University of Chicago Press.
Coates, Jennifer
1983 The Semantics of the Modal Auxiliaries. London: Croom Helm.
Dik, Simon C.
1989 The Theory of Functional Grammar. Part I: The Structure of the Clause.
Dordrecht, Providence: Foris.
Goossens, Louis
1987 Modal shifts and predication types. In J. Van der Auwera, L. Goossens
(eds.), Ins and Outs of the Predication. Dordrecht, Providence: Foris, 21–
37.
for linguistic action. Cognitive Linguistics 1: 323–340. (Also in Goossens
et al. (1995), 159–174.)
1996 English Modals and Functional Models: A Confrontation. University of
Antwerp, UIA. Antwerp Papers in Linguistics 86.
210 Louis Goossens
forthcoming Patterns of meaning extension, “parallel chaining,” subjectification,

and modal shifts. In A. Barcelona (ed.), Metaphor and Metonymy at the
Crossroads. Berlin, New York: Mouton de Gruyter.
Goossens, Louis, Paul Pauwels, Brygida Rudzka-Ostyn, Anne-Marie Simon-Vandenbergen,
Johan Vanparys
1995 By Word of Mouth: Metaphor, Metonymy and Linguistic Action in a
Cognitive Perspective. Amsterdam and Philadelphia: Benjamins.
Heine, Bernd
1995 Agent-oriented vs. epistemic modality: some observations on German
modals. In J. Bybee, S. Fleischman (eds.), Modality and Grammar in
Discourse. Amsterdam and Philadelphia: Benjamins, 17–53.
Kytö, Merja (ed.)
1993 The Helsinki Corpus of English Texts. University of Helsinki, Department
of English.
1990 Subjectification. Cognitive Linguistics 1: 5–38.
Lejeune, Ludo (ed.)
1995 Databases for the Study of English Auxiliaries. Volume I: The Modals in
Outline. Wilrijk, Departement Germaanse UIA.
1996 Databases for the Study of English Auxiliaries. Volume II: The Modals in
Spoken British English. Wilrijk, Departement Germaanse UIA.
Sweetser, Eve
1990 From Etymology to Pragmatics. Cambridge: Cambridge University Press.
Traugott, Elizabeth C.
1989 On the rise of epistemic meaning: An example of subjectification. Lan-
guage 65: 31–55.
Traugott, Elizabeth C., Ekkehard König
1991 The semantics-pragmatics of grammaticalization revisited. In E.C.
Traugott, B. Heine (eds.), Approaches to Grammaticalization. Volume I.
Metonymy in Onomastics
Olaf Jäkel
University of Halle
1. Introduction
How are surnames generated in German and in other languages? Are there
principles or patterns after all in the realm of anthroponymy? Or can we at best
reconstruct the etymology of each single name, finally to establish overall
arbitrariness? Like all proper names, surnames belong to a special class of
linguistic expressions that is in marked contrast to appellatives. Unlike these,
their semantic function lies not in the categorization of objects or in descrip-
tive conceptualization, but in the naming and identifying of individuals.
With several people bearing the same first name in one community,
unambiguous reference to a certain person became a problem. The solution
was found in giving by-names, e.g., ‘James son of Zebedee,’ ‘John the
Baptist,’ ‘Richard Lionheart,’ ‘Erik the Red,’ or ‘Thomas Aquinas.’1 In by-
names like these lies the origin of surnames as we know them, as Adolf Bach
confirms: “Our surnames are by-names inherited from ancestors” (Bach 1952:
231; cf. Reaney 1967: 19f).2 The custom of hereditary surnames did not
spread over Germany until the twelfth century, taking almost five hundred
years until complete proliferation: “The custom of bearing a surname was not
generally established in Germany until the year 1600” (Gottschald 1982: 47).
In the field of onomastics, the notion of semantic motivation has always
been more than a mere working hypothesis. It is simply regarded as a “fact that
proper names, at least at the time of origin, are more or less motivated”
(Naumann 1987: 16). Any original motivation, though, may later vanish from
the conscious knowledge of the linguistic community.3 This is why, from
212 Olaf Jäkel
today’s point of view, most onomasticians see proper names as having etymo-
logical meaning rather than lexical meaning.4 It is this etymological meaning
of surnames that we will be dealing with in the following account of various
naming principles.
All surnames used as examples in this paper have been taken from the
Hamburg Telephone Directory of 1993/94 (Deutsche Bundespost Telekom
1992), which served as corpus material in this investigation. After a short
glance at surnames motivated by ‘genealogy’ or ‘profession’ (Section 2) we
will turn to metonymic strategies of surnaming after certain ‘utensils’ or
‘qualities’ (Section 3). Section 4 focuses on surnaming motivated by ‘location
metonymy,’ which shows a rich structure of subtypes that can be analyzed
profitably from a cognitive linguistic perspective. Section 5 presents a sum-
mary of the suggested taxonomy of motivated surnames. As a final point, the
evidence of this onomastic investigation is brought to bear on the controver-
sial discussion of the role of metonymy currently taking place among cogni-
tive linguists.
2. Motivated surnames
At the outset of an investigation of motivated names it must be noted that in

German there is quite a number of common surnames which have to be
regarded as semantically unmotivated:
(1) Grumt, Techert, Voelz, Guhr, etc.
Even with the help of etymological research, the history of these and many
other surnames remains pretty much in the dark. Though working on the
general assumption that all surnames were once motivated, onomasticians
find it impossible to reconstruct the original motivation of a number of
names.5 The reasons must be seen in multiple flaws of diachronic transmis-
sion involving articulation or pronunciation, auditory understanding, and
spelling.
This supplies us with a residual category for those names whose semantic
motivation is not provided by any of the naming strategies discussed in what
follows. As this investigation is mainly concerned with the ‘principal patterns’
of naming, we will deal with surnames which do not necessitate longwinded
reconstructions of etymological meaning. The naming patterns demonstrated
by these examples surely account as well for surnames whose etymology is

less transparent. Detailed analyses of such names, though, can be found in the
specialized onomastic literature (cf. References).
Probably the oldest strategy of motivated surnaming makes use of the
descent of the name-bearer. This ‘genealogical’ naming simply gives the first
name of the father as a surname to his son:
(2) Werner, Thomas, Stephan, Christopher, Paul, Heinz, Heinrich,
Karl, Franz, Jonas, Lorenz, Martin, etc.
This group also comprises some supposedly unmotivated names such as Arendt
(< Arnold) or Drews (< Andreas), in which the father’s first name gradually
became simplified for ease of articulation or otherwise corrupted (cf. Smith
1950: 57ff). In addition, this motivation can be made clearer by suffixing the
morpheme -sohn ‘son,’ which in some cases is reduced to -son or -sen:
(3) Jansen, Friedrichsen, Michelsen, Thomsen, Johnson, Andersson,
Svensohn, Paulsohn
These examples represent an old German tradition with mainly patrilinear
inheritance of names from fathers to both sons and daughters. In comparison,
surnames after the first name of the mother are extremely rare (cf., among
others, Matthews 1966: 70; and Reaney 1958: xvff). The Scandinavian lan-
guages, however, have an autonomous way of surnaming daughters by suffix-
ing the morpheme -dóttir to the father’s first name.6
The use of professions, occupations, or offices held as surnames has
produced some of the most widespread family names of today. The following
list of common examples may evoke a vivid picture of life in medieval
communities:
(4) Müller ‘miller,’ Maier ‘mayor,’ Schulze ‘bailiff,’ Vogt ‘sheriff/
bailiff,’ Schmidt ‘smith,’ Bauer ‘farmer,’ Fischer ‘fisherman,’ Jäger
‘hunter,’ Schäfer ‘shepherd,’ Hirt ‘herdsman,’ Schneider ‘tailor,’
Weber ‘weaver,’ Schuster ‘shoemaker/cobbler,’ Schreiner ‘joiner,’
Zimmermann ‘carpenter,’ Radmacher ‘wheelmaker,’ Böttcher
‘cooper,’ Seiler ‘ropemaker,’ Töpfer ‘potter,’ Bäcker ‘baker,’
Fleischer ‘butcher,’ Knecht ‘farm hand,’ Ochsenknecht ‘stable lad,’
Koch ‘cook,’ Pfeiffer ‘piper,’ Geiger ‘violinist,’ Fiedler ‘fiddler,’
Meister ‘master,’ Kriegsmann ‘mercenary,’ Kaufmann ‘merchant,’
214 Olaf Jäkel
Förster ‘forester,’ Köhler ‘charcoal burner,’ Ritter ‘knight,’ Richter

‘judge,’ Rath ‘councillor,’ Schreiber ‘writer/clerk,’ Glöckner ‘bell
ringer,’ Kirchner ‘sexton,’ Küster ‘sacristan’7
By means of this ‘professional’ naming strategy, a Thomas with the by-name
der Müller ‘the miller’ became a Thomas Müller, who would pass on this
surname to his descendants, no matter if they carried on his trade or changed
professions.8
3. Metonymic strategies
In addition to the genealogical and professional naming patterns exemplified

so far, there are also metonymic strategies of naming.9 These shall be our
prime concern in the remaining parts of this paper.
First of all, we find a host of surnames that are identical to common
meaningful nouns in the language:
(5) Wolf ‘wolf,’ Hase ‘hare,’ Hering ‘herring,’ Amsel ‘blackbird,’ Fink
‘finch,’ Hahn ‘cock,’ Panther ‘panther,’ Hasenbein ‘hare’s leg’
(6) Dorn ‘thorn,’ Kirsch ‘cherry,’ Wurzel ‘root/carrot,’ Kohl ‘cabbage’
(7) Nagel ‘nail,’ Keil ‘wedge,’ Hose ‘trousers,’ Dose ‘box/can,’ Boden
‘ground/floor/bottom,’ Stein ‘stone,’ Kreis ‘circle,’ Tischbein
‘table-leg,’ Mastbaum ‘mast,’ Bohnsack ‘beanbag,’ Wintermantel
‘winter coat’
(8) Montag ‘Monday,’ Winter ‘winter,’ Mai ‘May,’ Hundertmark
‘hundred marks’
These are names or compounds from the animal (5) or plant (6) kingdom, for
concrete objects (7) or abstract entities (8), whose use as surnames is in all
probability motivated metonymically. The immediate problem with names
like these is that in most cases the particular kind of metonymy or alternative
motivation is up to speculation. Is a person called Hering because of an
objective similarity (like slimness), or because the name-givers meant to insult
him, i.e., as a derogatory term, or as a term of endearment? Then the name
would not yet be motivated metonymically. Instead, it would have to be
regarded as a straight metaphorical comparison, drawn with more or less good
reasons, and with more or less friendly intentions. But what if the bearer
acquired his surname because he worked as a fishmonger, or because he was a
famous fisherman, or because he was a notorious lover of smoked herring, or
because he once had an accident or a unique encounter in which a herring
played a special part? Then, though in most cases the precise motivation might
be difficult or impossible to confirm (cf. Naumann 1987: 28; Bach 1952: 284;
and Dorward 1995: x), we would be fully justified in calling this metonymy
proper, namely of the type IMPORTANT UTENSIL FOR PERSON.
But the precise motivation of naming can also be more transparent. In the
following examples, adjectives have been converted into surnames:
(9) Klein ‘short,’ Groß ‘tall,’ Langer ‘long’ + deriv. morph. ‘person,’
Starke ‘strong,’ Schön ‘pretty,’ Treu ‘faithful,’ Alt ‘old,’ Jung
‘young,’ Schwarz ‘black,’ Braun ‘brown,’ Roth ‘red,’ Lustig
‘merry,’ Wunderlich ‘strange’
Most of these cases exemplify the metonymic mapping SALIENT QUALITY FOR
PERSON. The same sort of adjectives can also enter into compounds with the
morpheme -mann ‘man’:
(10) Dickmann ‘fat man,’ Hartmann ‘hard man,’ Klugmann ‘wise man,’
Bangemann ‘scared man’
Furthermore, salient qualities of the original name-bearer can even be ex-
pressed by regular synecdoche (11), or hyperbolically (12):
(11) Kahlkopf ‘bald head,’ Weißhaar ‘white hair,’ Langnese ‘long nose’
(Low German), Krummbein ‘crooked leg’
(12) Riese ‘giant,’ Wicht ‘midget’
Examples such as those in (11) name a salient part of the body (like ‘bald
head,’ ‘white hair,’ ‘long nose,’ ‘crooked leg’) that stands for a person as pars
pro toto. The hyperbolic terms (‘giant,’ ‘midget’) in examples (12) obviously
refer to the height of the original name-bearer.
4. Location metonymy
With utensil metonymy and sometimes also with quality metonymy, we are on
fairly slippery ground as far as the particular details of motivational diagnosis
216 Olaf Jäkel
are concerned. With another important metonymic relation, the reconstruction

of each particular motivation is far less speculative: PLACE (OF ORIGIN) FOR
PERSON. This ‘location metonymy’ proves an enormously productive pattern
in the creation of surnames. If a term originally designating a locality becomes
a surname, it probably referred either to the current place of residence or to the
place of origin of the name-bearer (cf. Reaney 1958: xiii; Cottle 1978: 17–21;
and Dorward 1995: viii).
To provide an overview of location-metonymic surnames we can sub-
categorize them according to their morphological structure (see Figure 1). The
majority of them are compounds whose head denotes a (geographical) land-
mark. The prefixed modifier can either define the precise location in relation to
that landmark — in most cases by means of a preposition, sometimes by means
of a cardinal point — or it specifies the landmark by means of an adjective or
another noun.
The landmarks serving as heads in those location-metonymic surname-
compounds cannot be subcategorized further on formal grounds. But to obtain
a general idea we can subdivide them into ‘natural’ versus ‘man-made’
localities (see Figure 2), which is not to be read as a clear dichotomy. The
subtypes given in both groups simply list the varieties that were found in the
empirical corpus study.
It is interesting to note that with the more supraregional of those man-
made landmarks, the necessity of compounding decreases. Towns or cities,
regions or counties as well as whole countries make good surnames even
without any morphological modification (the reason for this will be discussed
in Section 4.5.).
Compounds
Attribute/ Head
Modification
Location in relation Specification of Landmark

to landmark landmark
Preposition Cardinal point Adjective Noun
Figure 1. Morphology-based subcategorization of location-metonymic surnames

(Geographical) Landmarks
natural man-made
– Landscapes/areas – Farmland
– Lakes and rivers – Fords and dikes
– Woods and copses – Buildings
– Settlements
– Towns and cities

supraregional – Regions/counties
– Countries
Figure 2: Landmarks that can serve as heads in location-metonymic surname compounds
The structure emerging from these attempts at subcategorizing location-met-

onymic surnames will also guide our following presentation, which has sub-
sections dealing with Prepositions as modifiers (4.1.), Cardinal points as
modifiers (4.2.), Heads: natural landmarks (4.3.), Heads: man-made land-
marks (4.4.), and ends with Supraregional landmarks (4.5.).
4.1. Prepositions as modifiers
The finding that surnames can be formed by combining a (geographical)

landmark as head with a preposition as modifier allows a certain predictabil-
ity. Going through the limited number of prepositions, one is bound to find
appropriate entries in ordered lists of names (e.g., telephone directories). Most
of the following surnames clearly designate the place of residence of their
bearers. The first modifiers we find are the simple spatial prepositions an ‘at’
(13), auf ‘on’ (14), bei ‘at, by’ (15) and in ‘in’ (16) along with their historical
or dialectal variants:
(13) a. Amberg ‘at the hill’
b. Ambach ‘at the brook’
c. Amthor ‘at the gate’
d. Amende ‘at the end’
(14) a. Aufderheide ‘on the heath,’ Auf dem Kampe ‘on the open
plain,’ (Middle Low German)
b. Auf dem Garten ‘on the garden’
c. Uphus ‘on the house’ (dial.), Uphoff ‘on the farm/yard’ (dial.)
218 Olaf Jäkel
(15) a. Beiderwieden ‘by the pasture/meadow’ (dial.)

b. Beimgraben ‘by the ditch,’ Biedebach ‘by the brook’
c. Bienwald ‘by the wood’
d. Biedenweg ‘by the path’
e. Biederstädt ‘by the town’ (dial.)
(16) a. Imwalde ‘in the wood,’ Imholz ‘in the copse,’ Indenbirken ‘in
the birches’
b. Imhof ‘in the yard’
As heads of these compounds we find natural landmarks like landscapes or
areas (e.g., 13a, 14a), lakes or rivers (13b, 15b) and woods or copses (15c,
16a) as well as man-made landmarks like farmland (14b), certain paths (13d,
15d), buildings (13c, 14c, 16b), or settlements (15e).
Surname-compounds are also formed by the group of prepositions vor ‘in
front of’ (17), hinter ‘behind’ (18), and über ‘across’ (19). Mountains or hills
are clearly preferred as heads by this group (17a, 18a, 19a). The reason must
be that these prepositions require a point of reference with some highly salient
vertical dimension.
(17) a. Vormberg ‘in front of the hill,’ Vormstein ‘in front of the rock’
b. Vormwald ‘in front of the wood’
(18) a. Hinterberger ‘behind the hill’ + deriv. morph. ‘person’
b. Hinterland ‘behind the land/hinterland’
(19) a. Ueberberg ‘across the hill’
b. Bobenhausen ‘across the houses/upper houses’ (dial.)
While neben ‘next to, beside, by’ (20) is only sparsely used in surnames,
zwischen ‘between’ must be even more unusual, as it does not appear at all in
the corpus. Instead, mittel ‘middle’ (21) appears in combination with almost
any conceivable landmark:
(20) Nebendahl ‘next to dale/side-dale’ (dial.)
(21) a. Mittelberg ‘middle hill’
b. Mittelbach ‘middle brook’
c. Mittfeld ‘mid field,’ Mittelfeld ‘middle field’
d. Mittelstraß ‘middle street’
e. Mittendorf ‘mid village,’ Mittelsdorf ‘middle village,’
Mittelstädt ‘middle town’ (dial.)
Directional prepositions are also represented in the corpus, with zu ‘to, to-
wards’ (22) probably defining a place of residence ‘on the way to’ the
landmark in question, and von ‘from’ (23) in contrast referring to the place of
origin of the name-bearer (cf. the discussion in Section 4.5.):
(22) a. Zumsande ‘towards the sand’
b. Zumwalde ‘towards the wood,’ Zumholz ‘towards the copse,’
Zumbusch ‘towards the bush’
c. Zum Felde ‘towards the field’
d. Zurmöhle ‘towards the mill’ (dial.)
(23) a. von der Heide ‘from the heath’
b. von der Lippe ‘from the Lippe’ (river name)
Um ‘around’ is rare (24), probably because this preposition defines only a
very imprecise location. By contrast, außen ‘outside’ is quite frequent as a
modifier in compound surnames (25).
(24) Umland ‘around land/surrounding land’
(25) a. Außendorf ‘outside/outer village,’ Utendörfer ‘outer villages’
(dial.)
b. Utermöhle ‘outer mill’ (dial.)
c. Butenhoff ‘outside farm/yard’ (dial.)
d. Butendeich ‘outside the dyke’ (dial.)
The last example (25d) in particular goes to show how important the in/out-
orientation in relation to certain landmarks was at the time of surnaming. In
settlements owing their existence to huge dikes around low-lying land prone
to flooding, the unprotected area ‘outside the dike’ was the place of residence
of the underprivileged. Thus the surname Butendeich ‘outside the dike’ is
quite frequent in the southern parts of Hamburg by the river Elbe and espe-
cially on the island of Finkenwerder.
4.2. Cardinal points as modifiers
Very similar to prepositions, cardinal points can also function as modifiers in

compound surnames like the following examples:
(26) Nordbruch ‘north fault/quarry’
220 Olaf Jäkel
(27) a. Westenholz ‘west copse’

b. Westhoff ‘west farm/yard’ (dial.)
(28) a. Ostwald ‘east wood,’ Osterwald ‘eastern wood,’ Osterholz
‘eastern copse’
b. Osterfeld ‘eastern field’
c. Osterhus ‘eastern house,’ Osterhof ‘eastern farm/yard’
Ostendorf ‘east village’
(29) a. Sudholz ‘south copse’ (dial.)
b. Sudenfeld ‘south field’ (dial.)
Though woods and copses seem to be prominent (27a, 28a, 29a), almost any
natural as well as man-made landmark can serve as head in these compound
surnames. The function of the cardinal points may be either to specify the
landmark (e.g., ‘the eastern wood’), or to define a location in relation to it
(e.g., ‘to the east of the wood’). In either case, this once was the place of
residence of the original name-bearer.
In the following sections we will sort out compound surnames according
to the type of heads they have. We will disregard the names with prepositional
or cardinal point modifiers discussed so far, though in principle of course
these could be included. A subdivision on formal grounds into heads modified
by adjectives and heads modified by nouns (cf. the overview given in Figure
1) would not be of much use due to the openness of these word classes. In
principle, almost any adjective or noun can serve to specify the landmark
figuring as the compound’s head. One general finding of our study, though, is
that the adjectives employed most frequently as modifiers are the antonymic
pairs of groß ‘big’ versus klein ‘small’ and neu ‘new’ versus alt ‘old,’ along
with their historical or dialectal variants.
4.3. Heads: natural landmarks
Natural landmarks functioning as heads in compound surnames can be subdi-

vided into groups denoting landscapes or areas, lakes or rivers, and woods or
copses (cf. the overview in Figure 2). Of landscapes or areas, mountains (30a)
are most prominent in the corpus, but smaller hills (30b) and valleys (30c) are
also represented. On the other hand, meadows (30d) and open plains (30e) are
far less frequent, presumably because these do not really make salient land-
marks.
(30) a. Lüttenberg ‘small mountain’ (dial.), Stoltenberg name +

‘mountain,’ Stauffenberg name + ‘mountain,’ Kirchberg
‘church hill,’ Eulenstein ‘owls’ rock’
b. Wiesehügel ‘meadow hill’
c. Wiesental ‘meadow dale/valley,’ Hacketal ‘pick dale/valley’
d. Wetterau ‘weather meadow’
e. Feldkamp ‘field plain’ (Middle Low German)
At all times, lakes and rivers have provided important geographical orienta-
tion. Quite naturally they also contribute to location-metonymic surnaming.
The corpus yields brooks or streams (31a), but also lakes (31b):
(31) a. Graubach ‘grey brook,’ Kirchbach ‘church brook,’
Rendenbach name + ‘brook,’ Schnädelbach name + ‘brook’
b. Lüttensee ‘small lake’ (dial.)
Very often a distinctive copse or a wood provides the name to identify a
person living there:
(32) Grotewold ‘big wood’ (dial.), Maiwald ‘May wood,’ Rheinwald
‘Rhine wood,’ Buchwald ‘beech wood,’ Mickenhagen ‘small
wood’ < Middle High German hac ‘(enclosed) wood’
4.4. Heads: man-made landmarks
As mentioned above, the distinction between rather natural and rather man-
made landmarks should not be overrated. The only relevant criterion for
location-metonymic surnaming is that the locations should be proper land-
marks, i.e., distinctive, fairly unmistakable points prominent in the landscape.
As subgroups of man-made landmarks we find farmland, certain paths, fords
and dikes, buildings, and larger settlements (cf. Figure 2).
Farmland, which is clearly demarcated, can serve better than the open
plain (see above, examples 30d and 30e) to pinpoint residents. Quite frequent
as heads of surnames are fields (33a, b), pastures (33c), and gardens (33d):
(33) a. Grothefeld ‘large field’ (dial.), Grünfeld ‘green field,’ Sommer-
feld ‘summer field,’ Wüstefeld ‘desert field,’ Strohfeld ‘straw
field’
b. Langacker ‘long field,’ Übelacker ‘bad field’
222 Olaf Jäkel
c. Hagelweid ‘hail pasture’

d. Baumgarten ‘tree garden,’ Baumgart ‘tree garden’ (abbrev.),
Weingarten ‘vineyard’
Surnames can tell us that at the time of naming, those landmarks having to do
with river control were of utmost importance: dikes (34a), fords (34b), or
bridges (34c) are more frequent in the corpus than simple streets (34d).
(34) a. Grothendiek, Grottendiek both: ‘large dyke’ (dial.)
b. Steinfurth ‘rock/stone ford,’ Ossenfort ‘oxen ford’ (dial.)
c. Ossenbrügge ‘oxen bridge’ (dial.)
d. Kirchgäßner ‘church street/lane’ + deriv. morph. ‘person’
Certain houses (35a), (farm)yards (35b) and castles (35c) are buildings dis-
tinctive enough to supply further surnames:
(35) a. Backhaus ‘bakery,’ Grothaus ‘big house,’ Kirch ‘church’
b. Neuhoff ‘new farm/yard,’ Eickhoff ‘oak farm/yard’ (dial.),
Diekhoff ‘dyke farm/yard’ (dial.), Kirchhof ‘church yard,’
Grotheguth ‘large farm’ (dial.)
c. Oldenburg, Ollenburg both: ‘old castle’ (dial.), Nienburg ‘new
castle’ (dial.)
Depending on their size, castles might already figure as larger settlements.
Other members in this group of surname sources are villages (36a) and towns
(36b):
(36) a. Niendorf ‘new village’ (dial.), Rahmsdorf name + ‘village’
b. Hofstadter ‘farm/yard/court town’ + deriv. morph. ‘person,’
Hustedt ‘farm/yard/court town’ (abbrev.)
4.5. Supraregional landmarks
While most of the location-metonymic surnames discussed so far owe their

existence to rather regional landmarks, we have already moved into the
supraregional dimension with town names. For a town name to become a
surname distinguishing a ‘Dietrich (von) Bern’ from another ‘Dietrich (aus)
Buxtehude,’ both persons must have been named as incomers in a third place.
In fact, such naming after place of birth or origin is one of the most common
strategies: “In the Middle Ages, and in particular in towns, the custom spread
of naming inhabitants who had moved there after their origin” (Gottschald
1982: 48; cf. Smith 1950: 47; and Reaney 1967: 36ff). Thus, incomers could
be named after their place of birth, but also after their native region or county,
and even country or nationality.
From a distance, all these supraregional landmarks are distinctive enough
to serve as surnames without further specification. This is why there are hardly
any compounds in this group. However, the development of town names to
proper and lasting surnames usually included a phase in which they were
preceded by the preposition von ‘of, from’ (37a) (see also example (23)),
which in most cases was dropped later (37b) (cf. Reaney 1958: xiv f).
Furthermore, the derivational morpheme -er ‘person‘ could be suffixed to a
town name to create a surname for an incomer (37c).
(37) a. von Coelln
b. Berlin, Utrecht, Bern, Buxtehude
c. Bremer, Achner
While in German surnames comprising regional associations or counties are
quite common (38), reference to countries or nationalities (39) is rarer. Pre-
sumably this is because during the centuries of active surnaming, international
influx into the German provinces was low.
(38) Friese ‘Frisian,’ Bayer ‘Bavarian,’ Preuß ‘Prussian,’ Böhme ‘Bo-
hemian,’ Hesse ‘Hessian’
(39) Deutscher ‘German,’ Holländer ‘Dutch,’ Norweger ‘Norwegian’
(40) Ausländer ‘foreigner,’ lit.: ‘out-land-er,’ Undeutsch ‘un-German’
In this connection, the examples under (40) certainly are a reflection of the
mentality of the name-givers. If somebody is named Ausländer ‘foreigner’ or
even Undeutsch ‘un-German,’ the intention — then as now — is clearly not so
much to characterize the name-bearer precisely, but to brand him as excluded
from the community.10
The final example (41) provides confirmation ex negativo of the impor-
tance of location metonymy for surnaming:
(41) Wanderer ‘wanderer’
Here it is exactly the ‘non-locatability’ of the person named which to the
name-givers seemed significant enough to identify and distinguish him from
more settled first-namesakes.
224 Olaf Jäkel
5. Summary and conclusion: metonymic strategies in

onomastics
In our corpus study based on the 1993/94 Hamburg Telephone Directory, we

have investigated the most important surnaming patterns in German. Met-
onymic strategies in particular were found to play a leading role in the creation
of surnames. The following overview (Figure 3) presents a taxonomy of
naming strategies proposed as result of this investigation.
Before we come to some final considerations of the role of metonymy in
onomastics, the general naming patterns found in our investigation can be
summarized:
(a) Genealogical naming: after descent (patronymic)
(b) Professional naming: by reference to profession, occupation or office
(c) Metonymic naming:
Utensil metonymy: IMPORTANT UTENSIL FOR PERSON
Quality metonymy: SALIENT QUALITY FOR PERSON
Location metonymy: PLACE (OF ORIGIN OR RESIDENCE) FOR PERSON
The surnames generated according to these five patterns can be construed as
alternative ways of answering the same basic question ‘Which George do you
mean?,’ namely:
– ‘George, the son of so-and-so’ (genealogical)
– ‘George, the linguist’ (professional)
– ‘George (with the) straw hat’ (utensil-metonymic)
– ‘George (with) the paunch’ (quality-metonymic)
– ‘George from California’ (location-metonymic)
Surnaming
unmotivated motivated
(motivation not
recovered)
genealogical professional metonymic
Utensil Quality Location

metonymy metonymy metonymy
Figure 3. A taxonomy of the essential strategies of surnaming

To a great extent, our results are in line with the vast body of expert
knowledge represented in the standard onomastic literature. A slight diver-
gence in classification results from the fact that nowhere in the standard
literature is metonymy recognized as an important motivating principle in
surnaming.
The classification that has become the standard for German onomastics
comes from Adolf Bach (1952), who distinguishes surnames after ‘first
names’ (Rufnamen), ‘place of origin’ (Herkunft), ‘place of residence’ (Wohn-
stätte), ‘occupation and status’ (Beruf und Stand), and ‘nicknames’
(Übernamen). Among English onomasticians, the subdivision of surnames
into the four classes of ‘local,’ ‘occupational,’ ‘patronymical’ (or ‘relational’)
and ‘descriptive’ (or ‘nicknames’) is consensual since the work of Elsdon C.
Smith (Smith 1950: 44; cf. Reaney 1958: xii and 1967: 20; Cottle 1978: 9; and
Dorward 1995: vi).11 What is correct in these traditional classifications is
preserved in our suggested taxonomy of surnaming strategies, which also has
some additional advantages.
The incorporation of metonymy as a motivating principle in surnaming will
certainly not revolutionize onomastics. Yet it seems desirable to differentiate
between utensil metonymy, whose analysis remains speculative, and the much
clearer quality metonymy. In traditional onomastics, surnames of both patterns
are jumbled up in that huge and ill-defined residual category of ‘nicknames,’
which is usually left completely unanalyzed. Another nontrivial result is the
finding that surnaming after place of residence and after place of origin is based
on the same location-metonymic principle. Our discussion has shown the
immense productivity of this principle in the generation of surnames.
The surnaming patterns investigated here on the basis of German names
also account for English surnames (this was shown in Jäkel 1996). In fact,
English and German surnames do not only follow the same motivational
patterns, but they even share most of the morphological details discussed
above. Our taxonomy of naming strategies was even found to apply to Non-
Indo-European languages like Hungarian and Japanese.12
As regards the relevance of the results of our investigation for the
controversial discussion of the general role of metonymy currently taking
place among cognitive linguists, our claims are modest. With our results we
have supplied firm evidence to corroborate the view that metonymy — in
contrast to metaphor — is basically a ‘naming strategy.’ To distinguish
metonymy from metaphor, which has primarily an ‘explanatory function’ (see
226 Olaf Jäkel
Jäkel 1997), Lakoff and Johnson (1980: 36) state: “Metonymy, on the other
hand, has primarily a referential function, that is, it allows us to use one entity
to stand for another.”13 And that this referential function really is about the
‘naming of individuals’ can be confirmed by two quotes from Lakoff (1987).
There, metonymy is defined as follows: “With respect to naming, A stands for
B” (Lakoff 1987: 19). And in the context of his theory of ‘idealized cognitive
models’ Lakoff (1987: 85) explains: “Most metonymic models are, in fact, not
models of categories; they are models of individuals.”
Yet again, this should not be misunderstood. Metonymic structures can
definitely have cognitive status (this point is also made by Lakoff 1987: 90).
They just do not have it as strategies that enhance conceptual understanding or
offer conceptual explanations, but as principles for the naming of individuals.
Humans obviously have a strong cognitive need for distinctive yet economical
naming. Nowhere can this be seen more clearly than in the supposedly
antiquated domain of onomastic research. And that metonymic patterns of
various kinds are at work in exactly this of all domains is no coincidence: just
because of its general naming function there is a vast field of operation for
metonymy in onomastics.
Notes
1. It will be seen later that these by-names anticipate all known strategies of surnaming.
2. This and all remaining quotations from German sources are given in translation by the
author.
3. Cf. Fleischer (1962: 30): “At the moment of a name’s origin the relation is present, later
to congeal.”
4. This is stated explicitly by Debus (1966: 14) and at least implicitly held by many
onomasticians; for example, see Bach (1952: 206ff), Fleischer (1962: 30ff), Witkowski
(1964: 51), Cottle (1978: 10ff), and Dorward (1995: xvi). Cf. also Lyons (1977: 223).
5. See Matthews (1966: 71): “But there always remain at least three or four per cent about
which the honest investigator can only state, ‘I don’t know’.” Another problem which
cannot be considered here is the possibility of alternative reconstructions for one and the
same name. For a treatment of this issue see e.g., Bach (1952: 235) and Matthews (1966:
71f).
6. Note that in the Scandinavian tradition the surname does not carry the same weight as in
German. Its character is still more like that of a by-name. Concerning Icelandic in
particular, cf. Wiktorin (1991: 57): “Unlike our common practice, Icelandic gives more
prominence to the first-name. [...] It is only logical that in the Icelandic telephone
directory entries are arranged alphabetically according to first-names.”
7. Of course the spelling of these surnames may vary. Thus, Schmidt is more frequent than
Schmied, Böttger is found as well as Böttcher, and Köster alongside Küster, to name but
a few examples.
8. See Wimmer (1973: 27f), who analyses this development as the transition from an
appellative use of Müller to a proper name Müller.
9. That professional surnaming should not be counted among metonymic strategies is
confirmed by Norrick (1981). Although he proposes no less than 17 different metonymic
principles, he deems no special metonymic principle necessary if occupation stands for
person: “[...] the theory of reference [...] predicts that one can refer to a person by
identifying his role and conversely” (Norrick 1981: 100). This distinguishes professional
naming from those strategies discussed below as truly metonymic. However, other
authors who subscribe to a more extensive notion of metonymy (e.g., some of the
contributions to this volume) would notch up even professional and genealogical naming
patterns under metonymic strategies.
10. For a synchronic analysis of the concept ‘Ausländer’ and its current motivation see Jäkel
(1993).
11. Matthews (1966: 69) confirms this consensus: “[...] nearly all writers on the subject of
surnames have classified them into the four types of Locality, Relationship, Occupation
and Nicknames.” And Smith (1950: 45) even pronounces the universal status of this four-
part classification: “In general, surnames in all countries originate in one of the above
four ways if they are not consciously adopted.”
12. I thank Zoltán Kövecses for providing information on Hungarian surnames, and Ken-Ichi
Seto for his detailed explanation of Japanese surnames. According to his estimate, no less
than 90 % of Japanese surnames are of the location-metonymic type, which is much more
than in German or English.
13. But cf. Taylor (1989: 122ff), who seems to be taking the other side, though without
voicing any convincing examples or arguments to support his view. For more controver-
sial positions on the status of metonymy, see the other contributions in this volume.
References
Bach, Adolf
1952 Deutsche Namenkunde. Bd. I: Die deutschen Personennamen. Second
edition. Heidelberg: Winter.
Cottle, Basil
1978 The Penguin Dictionary of Surnames. Second edition. Harmondsworth:
Penguin.
Debus, Friedhelm
1966 Aspekte zum Verhältnis Name – Wort. Groningen: Wolters.
Deutsche Bundespost Telekom (eds.)
1992 Amtliches Telefonbuch der Deutschen Bundespost Telekom, Ortsnetzbereich
Hamburg, 1993/94 edition. Frankfurt a.M.
228 Olaf Jäkel
Dorward, David
1995 Scottish Surnames. Glasgow: Harper Collins.
Fleischer, Wolfgang
1962 Zur Frage der Namenfelder. In W. Fleischer (ed.), Name und Text.
Tübingen: Niemeyer, 25–42.
Gottschald, Max
1982 Deutsche Namenkunde: Unsere Familiennamen nach ihrer Entstehung und
Bedeutung. Fifth edition. Berlin, New York: Mouton de Gruyter.
Jäkel, Olaf
1993 Wer oder was ist eigentlich ein Ausländer? Anmerkungen zu einer
seltsamen Kategorie. In Pressestelle der Universität Hamburg (eds.), Uni hh
2: 44.
1996 Metonymy as a cognitive principle in onomastics. Paper presented at the
XIXth International Congress of Onomastic Sciences, University of Aber-
deen, Scotland, August 4–11, 1996.
1997 Metaphern in abstrakten Diskurs-Domänen: Eine kognitiv-linguistische
Untersuchung anhand der Bereiche Geistestätigkeit, Wirtschaft und
Wissenschaft. Frankfurt a.M.: Peter Lang.
Lakoff, George
Lyons, John
1977 Semantics. Volume I. Cambridge: Cambridge University Press.
Matthews, Constance M.
1966 English Surnames. London: Weidenfeld and Nicolson.
Naumann, Horst (ed.)
1987 Familiennamenbuch. Leipzig: Bibliographisches Institut.
Norrick, Neal R.
Benjamins.
Reaney, Percy H.
1958 A Dictionary of British Surnames. London: Routledge and Kegan Paul.
1967 The Origin of English Surnames. London: Routledge and Kegan Paul.
Smith, Elsdon C.
1950 The Story of Our Names. New York: Harper.
Taylor, John R.
Clarendon.
Wiktorin, Karl
1991 Island erfahren: Reiseinformationen. Eichstätt: Lundipress.
Wimmer, Rainer
1973 Der Eigenname im Deutschen: Ein Beitrag zu seiner linguistischen
Beschreibung. Tübingen: Niemeyer.
Witkowski, Teodolius
1964 Grundbegriffe der Namenkunde. Berlin: Akademie-Verlag.
Part III
Case Studies of Metonymy

Grammatical Constraints on Metonymy
On the Role of the Direct Object
Richard Waltereit
University of Tübingen
1. Introduction
While significant results have been achieved in the study of metonymy on a

conceptual basis (cf. Le Guern 1973; Lakoff 1987; Croft 1993; Nunberg 1995;
Blank 1997), relatively little work has been done on investigating metonymy
from a grammatical point of view (cf. Langacker 1993).1 In this paper I want
to venture into the theory of metonymy by asking whether or not there are
grammatical constraints on metonymy. In particular, the question will be
addressed as to whether there are arguments (subject, direct object, etc.) that
are particularly suited to metonymic phenomena. It will be argued that the
direct object (or, in terms of semantic roles, the ‘theme’) is, in a very particular
way, privileged in this respect. This is claimed to be true for French but it is
likely to extend also to other European languages.
The problem of grammatical constraints on metonymic reference has
been alluded to in Jackendoff 1992. He describes a fictive situation: we visit
Mme Tussaud’s wax museum in London and see Ringo Starr looking at his
own wax statue. Now let us imagine the narrator reporting the following:
(1) All of a sudden, I accidentally bumped into the statues, and *Ringo
toppled over and fell on himself. (Jackendoff 1992: 5)
As Jackendoff notes, the only pragmatically available understanding is that
the statue hits the real Ringo, but this reading is apparently blocked grammati-
234 Richard Waltereit
cally. Grammatically, the only possible reading is that the real Ringo hits the
statue. One has to consider the relation between the person (Ringo) and his
physical representation (the statue) a metonymic one. Now, the possibility of
metonymic reference relation is clearly dependent on the grammatical context,
as (1) shows. The nature and the form of this dependency are far from obvious.
In this paper, I hope to shed some light on these questions, but it will of course
be impossible to study the issue exhaustively.
As for the theory of metonymy, I assume that metonymy is a device for
reference that exploits a contiguity relation between two entities (cf. Koch,
this volume; as for non-referential approaches to metonymy, cf., e.g., Radden
and Kövecses, this volume; Feyaerts, this volume; Panther and Thornburg,
this volume; and Pankhurst, this volume). Contiguity can be a relation of
spatial vicinity, a PART-WHOLE relation, a CONTAINER-CONTENTS relation, an
AGENT-ACTION relation, etc. It seems less important to try to enumerate the
possible types of contiguity than to acknowledge that contiguity is a relation
of experiential ‘togetherness,’ where experience is to be understood in the
broadest sense. Given this assumption, contiguity can take virtually any form,
provided speakers construe a relation between the entities involved and take
the relation as communicatively relevant (Nunberg 1995). To the extent that
the respective contiguity belongs to a relatively stable relation, it can be
addressed as part of a frame (taking this notion as a cover term for what has
been called ‘domain,’ ‘script,’ ‘scene,’ etc.) (cf. Koch, this volume).2
2. Two levels of metonymy in verbal semantics
In order to clarify the notions that will be used throughout this paper, a basic
distinction with respect to metonymic relations in verbal semantics has to be
introduced. Consider (2) and (3):
(2) a. La soupe aux poissons n’a pas donné de pourboire.
‘The fish soup didn’t leave a tip.’
b. Michel n’a pas donné de pourboire.
‘Michel didn’t leave a tip.’
(3) a. Ils ont enfin servi la soupe aux poissons.
‘They finally served the fish soup.’
b. Ils ont enfin servi Michel.
‘They finally served Michel.’
Grammatical Constraints on Metonymic Reference 235
In (2a) we deal with a metonymy of the classical type: the subject noun phrase
does not actually refer to the fish soup, but of course to the guest who ordered
it. We grasp this immediately because the predicate n’a pas donné de
pourboire normally requires a human subject (as is the case in (2b)). The
selection restrictions of the predicate are therefore violated in (2a). In order to
get an appropriate referent for the actual subject the hearer has to search for
some entity that meets the ‘human’ requirement and that is at the same time
semantically contiguous with the actual subject noun phrase. There is clearly a
contiguity of the ‘guest’ and the ‘ordered dish’ in restaurant communication
that makes it easy to choose the guest as the appropriate referent. That is the
way metonymies usually work: the given noun phrase violates the verb’s
selection restrictions, and this violation has to be accommodated by contigu-
ity-based reference. The selection restrictions of the predicate tell the hearer
what kind of linguistic expressions would ordinarily be expected in the given
syntactic slot; the contiguity relation ensures that the expression will neverthe-
less be correctly understood. I will refer to this as a metonymy on the
‘insertional level’ (because here we are dealing with lexical insertions into
subcategorization frames).
Now, things are quite different in (3). The verb servir is polysemous, as is
its English counterpart to serve. It allows two different thematic roles in the
direct object position, or rather two different classes of objects: the ‘served
dish’ and the ‘served person.’ These are two semantically contiguous roles,
and their relation is the same as in (2). But this time, there is no metonymic
reference. It is really the dish and really the person, respectively, that are
served. In order to understand (3a) and (3b) it is not necessary to activate a
metonymic reading. Rather, the contiguity relation of the served dish and the
served person is lexically encoded. The two different thematic roles are part of
the verb’s lexical content. I will refer to this phenomenon as contiguity on the
‘role level.’ The verb permits two different thematic roles in the direct object
position without formally indicating this variation. This state of affairs entitles
us to regard the verb as polysemous. Polysemy means that a lexical item has
two separate but related meanings compatible with the same surface manifes-
tation. Furthermore, a polysemy is always a form of lexical ambiguity that is
not predictable from grammatical or other rules. A polysemy of a word is
always an individual, ‘accidental’ phenomenon; ultimately, it is the result of a
diachronic semantic change (this was already noticed by Bréal 1897).3
This leads us to the next crucial assumption: role-level contiguities arise
diachronically from insertional-level metonymies. The occasional metonymic

use is likely to be fixed later as a new meaning of the verb, when a metonymic
shift is no longer involved and the new meaning becomes lexicalized (cf.
Blank 1997: 103–130). This can be shown, for example, for the French verb
servir. On the one hand, we have the reading ‘servir quelqu’un de quelque
chose’ (with the beneficiary in the direct object position, corresponding to
example (3b)). This reading is very old and has been in use since Old French.4
On the other hand, we have the reading ‘servir quelque chose à quelqu’un,’
corresponding to example (3a), which has only been attested since the 17th
century. It is plausible to assume that the more recent reading has arisen out of
a metonymic use of the older reading.
In the remainder of this paper I will present some arguments supporting the
claim that on both levels of contiguity-based phenomena, the direct object has
a certain primacy over all other arguments, as is represented in the hierarchy (4):
(4) Direct object > subject > other
This hierarchy makes two claims: firstly, verbal polysemies of the kind
exemplified in (3) concern the direct object (if there is one), otherwise the
subject. In very few cases they may affect other arguments. Secondly, syntac-
tic constructions that inherently involve contiguity allow metonymic refer-
ence more easily for the direct object (if there is one) than for the subject. Note
that the hierarchy does not make the general claim that the object sustains
metonymic reference better than the subject. I do not intend to affirm that, say,
a subject cannot sustain metonymic reference if it appears in a transitive
sentence. Already the example (2a) makes obvious that this would not be the
case. The following sections will explain the hierarchy (4) in more detail.
3. Role-level contiguities: verbal polysemy
Let us begin with role-level phenomena. They crucially involve thematic

alternations insofar as the roles selected by the verb are modified. As will be
seen, any kind of polysemy-based thematic alternation observes the following
order: if there is a direct object, the polysemous alternation applies to the
direct object. If there is no direct object, the alternation nearly always applies
to the subject. In the entire French valence dictionary of Busse and Dubost
(1983), I have found only one counterexample to this rule.5
3.1. Frame-based alternations
Let me exemplify my hypothesis with the following data:

(5) a. Marc va balayer ta chambre.
‘Marc will sweep your room.’
b. Marc n’a pas encore balayé les débris de verre.
‘Marc hasn’t swept up the broken glass yet.’
Here we have a container-contents polysemy in the direct object: the things
that are swept up and the room that becomes thereby cleaned.
(6) a. Le public a bissé cette chanson.
‘The public called for an encore of that song.’
b. Les spectateurs ont bissé sans succès le fameux groupe de rock.
‘Without success, the spectators called the famous rock group
back for an encore.’
With this verb, too, the direct object allows semantic variation: the artist who
is supposed to give an encore and the song s/he is supposed to sing. These
roles are semantically contiguous, but neither of them activates a metonymy in
the sense that they have to be understood non-literally. Accordingly, the verb
bisser can be treated as polysemous.6 The same kind of alternation of roles in
the direct object position shows up in the following sentence pairs:
(7) Contiguity of game and player:
a. Elle a gagné la partie.
‘She has won the game.’
b. Elle m’a gagné.
‘She has beaten me at the game.’
(8) Contiguity of observation and looked-out object:
a. Elle a épié tes allées et venues.
‘She has spied upon you.’
b. Je passe mon temps à épier une occasion favorable.
‘I spend my time on the lookout for a good opportunity.’
(9) Contiguity of dangerous action and endangered thing:
a. Elle allait risquer une démarche fatale.
‘She was going to risk taking a fatal step.’
b. Vous devriez ne pas risquer votre réputation.

‘You should not risk your reputation.’
(10) Contiguity of sound and resounding room:
a. Ta voix résonne dans le couloir.
‘Your voice echoes in the corridor.’
b. Cette salle résonne trop.
‘This hall resounds too much.’
An extreme example comes from English to teach: to teach (mathematics/
Sunday school/little children/third grade) (Langacker 1987: 269f), where we
are confronted with the fourfold contiguity of the ‘school subject,’ the ‘institu-
tion,’ the ‘type of learner,’ and the ‘school grade’ in the direct object.
If there is no direct object but nevertheless two semantically contiguous
thematic roles in the verb’s argument structure, then this contiguity nearly
always applies to the subject, as is the case of the intransitive predicate in (10).
The roles of the sound and the resounding room are contiguous. This results in
a verbal polysemy with respect to the subject. The contiguity in all these cases
relates to the respective verb’s frame or domain (cf. Croft 1993; Koch 1997
and this volume). The sound and the resounding room are entities that belong
to the encyclopaedic knowledge activated by the verb résonner ‘to resound.’
Why do verbs have these polysemies? A semantic-pragmatic motivation
for the polysemy in, e.g., (9) could run like this: the ‘endangered thing,’ the
‘dangerous thing’ and an ‘actor’ are likely to co-occur whenever people speak
of some dangerous action. We may call this the ‘frame of the dangerous
action.’ Now, people may wish to verbalize only part of this frame. It is
therefore natural that languages should provide them with tools that enable
this partial construal. One of these tools is apparently verbal polysemy. The
verb allows two different construals, each selecting different entities from the
respective frame. The contiguity of the ‘dangerous thing’ and the ‘endangered
thing’ ensures successful communication and provides the motivation for the
polysemy. This is the overall picture emerging from examples (5) to (10). In
these examples, we are dealing with a contiguity relation between two kinds of
thematic roles. The contiguity relation in question is a very specific one, as
defined by the verb. ‘Specific’ means that the contiguity crucially relates to
particular properties of the frame activated by the verb. It does not, or not
primarily, relate to general properties of action verbs, process verbs, etc. As
the semantic nature of these alternations hinges to a large extent on the frame
embodied by the respective verb, we do best to call them frame-based alterna-

tions.
3.2. Locative alternations
Regularities of a slightly different kind are found with the locative and
‘swarm’-alternations that have received a lot of attention especially in the
generative literature (cf. Fillmore 1968; Anderson 1971; Baker 1988; Olsen
1994). These can always be reduced to some kind of ‘local’ relation. Locative
alternations are of the type illustrated in (11) and (12):
(11) a. Ils ont chargé du charbon sur le bateau.
‘They have loaded coal onto the ship.’
b. Ils ont chargé le bateau de charbon.
‘They have loaded the ship with coal.’
(12) a. Il faut substituer ce chiffre par un autre.
‘You have to replace this number with a different one.’
b. Il faut substituer ce chiffre à un autre.
‘You have to put this number into another number’s place.’
What is alternating in these alternations? The object position sustains two
different semantic roles. These two roles are clearly contiguous to each other,
they establish a (be it metaphorical) CONTAINER-CONTENTS relation. This is
obvious in (11), where the ship and the coal are actually regarded as a
container and its contents. It is less obvious but nevertheless plausible in (12).
The verb substituer ‘to substitute’ evokes the idea of a base from which the
substituted thing is detached as a profile (in the sense of Langacker 1987: 183–
189). The highlighted profile is part of the base: one can only substitute parts
in wholes. The PART-WHOLE relation is in turn a relation of ‘topological
inclusion,’ a CONTAINER-CONTENTS relation (Winston et al. 1987: 431–435).
Note that in these cases it is only possible to speak of polysemy with
respect to the direct object position. Only in the direct object position are there
alternating contiguous roles without any surface indication of the alternation.
The oblique objects (sur le bateau ‘onto the ship,’ de charbon ‘with coal,’
etc.) present alternating roles that are in a relation of contiguity, too, but
formally indicate the semantic variation by a varying preposition. The formal
marking of the oblique object prevents these from being semantically ambigu-
ous; and ambiguity would be a prerequisite for polysemy. The ambiguity

condition is indeed crucial for my argumentation. Only if an argument of a
given verb is ambiguous with respect to its thematic interpretation can it be
analyzed as the diachronic result of an insertional metonymy. Recall from
Section 2 that role-level contiguities are considered as lexicalizations of a non-
literal (metonymic) reading of a determinate thematic role. The contiguity of
the oblique objects (sur le bateau, de charbon, etc.) cannot be the direct
diachronic result of a metonymic interpretation, because their semantic rela-
tion to the predicate is formally marked and is therefore unlikely to result from
a non-literal reading.
3.3. The ‘swarm’-alternation
The so-called ‘swarm’-alternation owes its name to the famous example Bees
are swarming in the garden/The garden is swarming with bees. The point of
this alternation is that two thematic roles can be coded either in the subject
position or in a prepositional phrase. Interestingly, the contiguity relation
between the two roles is usually a CONTAINER-CONTENTS relation:
(13) a. L’eau déborde de la baignoire.
‘The water is flowing over the bathtub.’
b. La baignoire déborde d’eau.
‘The bathtub is overflowing with water.’
In (13) there is clearly a contiguity between the contents, the water, and the
container, the bathtub.
(14) a. La pluie ruisselait sur les murs.
‘The rain was streaming down the walls.’
b. Il ruisselait de cold-cream, de sueur et de vin. (Flaubert)
‘He was streaming with cold cream, sweat and wine.’
The CONTAINER-CONTENTS relation is less obvious here, but at any rate we are
dealing with a relation of local contact with some surface. The ‘container’ is,
so to speak, ‘flat.’
With both verbs, the subject can express both the contents and the
container. This flexibility has to be explained by polysemy. For a monose-
mous interpretation to be possible, we would have to find one label that would
appropriately characterize the selection restrictions of the subject. But this
turns out to be impossible. We cannot identify a reasonable criterion that

would integrate both, say, the container and the contents of the déborder-type
of verbs. We therefore have to assume a kind of polysemy; again, we have two
semantically contiguous roles available for the subject. Now this may appear
to challenge the claim that these kinds of polysemies always apply to the direct
object. However, it does not, because there is no direct object in these verbs
and my prediction was that if there is no direct object, the polysemy concerns
the subject, which here is indeed the case.
Again regarding these verbs, one meaning arose diachronically out of the
other one: ruisseler has been lexicalized with the ‘place’ as subject only since
1658. With déborder it was the other way round. Only since the end of the
19th century has the ‘contents’ been available as a possible subject of the verb.
Now let us take a look at some other verbs which superficially display the
same phenomenon:
(15) a. Les oiseaux piaillent dans l’arbre.
‘The birds are cheeping in the tree.’
b. Tout l’arbre piaille d’oiseaux.
‘The tree is cheeping with birds.’
(16) a. Il y a des souris qui couinent dans le grenier.
‘There are mice squeaking in the attic.’
b. Le grenier couine de souris.
‘The attic is squeaking with mice.’
Again, there is a CONTAINER-CONTENTS relation concerning the subject, but this
time we are not dealing with a polysemy. The (b) sentences in (15) and (16)
represent uses of the verb that are not lexicalized. They are highly marked; it is
very unusual to speak of a ‘squeaking attic.’ We will not find these uses in a
dictionary, yet they are possible and attested. They represent what I have
called insertional metonymies (cf. Section 2.1.). The subjects in (15b) and
(16b) apparently violate the selection restrictions. Yet, this violation is accom-
modated by a metonymic interpretation. Maybe one day the insertional me-
tonymy will become fixed as a polysemy, as is the case in the standard
‘swarm’-alternation (cf. (13) and (14)). This kind of alternation leads us to
phenomena involving reflexives and possessives, which will be considered in
the next section.
4. Insertional metonymies: syntactic constructions
One might think that there is no use looking for syntactic restrictions in the
area of insertional metonymies, because any constituent can in principle be
affected by a metonymic shift. This is true, of course. One would be surprised
to find, for example, that a dative phrase cannot be metonymic. And, of
course, I will not put forward such a claim. There are, however, plausible
arguments in support of the hypothesis that also in this domain the direct
object has certain privileges. Certain ‘marked’ syntactic constructions that
inherently involve contiguity relations allow metonymic reference for the
direct object but less naturally or not at all for other arguments. A construc-
tion, in this sense, is a specific arrangement of syntactic ‘slots’ which can be
filled from an open set of lexical items. It may be restricted to a certain class of
lexical items but its properties are not entirely deducible from that class. The
construction types involved here include reflexive cliticization and some types
of inalienable possession constructions.
4.1. Reflexive clitics
Reflexive cliticization, with the reflexive clitic pronoun being a direct object,
is a construction that is particularly suitable for metonymies. In (17) and (18) I
present some examples where the reflexive clitic allows metonymic reference
and a corresponding lexical direct object which does not:
(17) a. Pierre s’économise pour être en forme le jour du match. (Zribi-
Hertz 1978)
‘Pierre saves his energy in order to be in form on the day of the
match.’
b. *L’entraîneur l’économise pour qu’il soit en forme le jour du
match.
‘The coachi saves hisj energy for himj so that hej will be in form
on the day of the match.’
In (17a) the reflexive is clearly metonymic, because Pierre will surely not
save himself, but rather something about himself, i.e., his energy. This met-
onymic relation only works with the reflexive; we cannot have a non-reflexive
pronoun in the same slot. The asymmetry is the same in (18).
(18) a. Je ne me suis pas raconté dans ce roman.

‘In this novel, I didn’t relate myself.’
b. *Je n’ai pas raconté mon père dans ce roman.
‘In this novel, I didn’t relate my father’s life.’
This kind of metonymic reference only works with the direct object reflexive,
not with the indirect object reflexive.
Furthermore, there is a much cited phenomenon with certain psychologi-
cal verbs that do not permit direct object reflexivization (cf. Belletti and Rizzi
1988: 296; Grimshaw 1990: 152–158):
(19) a. *JeS meDO préoccupe (but cf. Mon fils me préoccupe)7
‘I’m worried.’
b. *Elle seDO choque.
‘She shocks herself.’
What is wrong here? In my view, the subject in these sentences has to be
understood metonymically. By saying je me préoccupe I mean that something
about me, perhaps my health, is worrying me. This should be a metonymy.
Now it seems that the metonymy on the subject cannot work because it is
somehow blocked by the direct object reflexive. This again strengthens the
idea that the direct object is privileged with respect to metonymy: if there is a
direct object in the reflexive construction, the other arguments cannot be
interpreted metonymically.
On the other hand, if the reflexive is an indirect object, the subject can
very well support metonymic reference:
(20) Il seIO plaît.
‘He is pleased with himself.’ (literally ‘He pleases himself’)
Here again, the subject has to be understood metonymically: it is always
something about that person that he is pleased with, e.g., his handsomeness.
But this time, the reflexive is a dative. It is therefore plausible to assume that
the metonymy of the subject works because there is no direct object to
intervene and block the metonymic interpretation.
The overall picture arising from this discussion is that in a canonical
reflexive construction there is a PART-WHOLE relation holding between the
subject and the direct object, the subject being the whole. This asymmetric
relation cannot be inverted.8 Inversion of the PART-WHOLE relation, as in je me
préoccupe, yields an ungrammatical utterance, as is represented schematically
in Figure 1:
A
B A
je me mouche *je me préoccupe

‘I blow my nose’ ‘I’m worried’
Figure 1. PART-WHOLE relations in reflexive constructions
The reflexive ‘object’ seems to be a particularly suitable candidate for a

metonymy, while the ‘subject’ of the reflexive construction seems to be a
rather reluctant one.
The PART-WHOLE relation being a contiguity relation, it becomes clear
now that the reflexive construction inherently involves contiguity. What is
more, the reflexive construction not only inherently involves the contiguity of
its referents, but also a contiguity of the reflexive and its antecedent in the
chaîne parlée. Reflexive elements enforce a syntagmatic co-occurrence of
referents; thus, their grammatical functioning relies on contiguity. Maybe it is
because they rely on contiguity for their grammatical and their referential
basis that they facilitate metonymic shifts more readily than other types of
pronouns and noun phrases.
4.2. Inalienable possession
Inalienable possession constructions are a topic of special interest to many

syntacticians and semanticists. In many languages, one finds the particular
phenomenon that possessive relations involving the ‘personal sphere’ (body
parts, clothes, personal belongings, sometimes even relatives, etc.) allow
codings that are unlike those in other possessive relationships:
(21) a. Jean a lavé les pieds de son fils.
‘Jean has washed his son’s feet.’
b. Son fils, Jean lui a lavé les pieds.
‘Jean has washed his son’s feet.’
(22) a. Jean a lavé le toit de sa voiture.
‘Jean has washed the roof of his car.’
b. *Sa voiture, Jean lui a lavé le toit.
‘Jean has washed the roof of his car.’
The syntax is the same in the (a) sentences and (b) sentences, respectively.
Both possessive relations (human being-feet vs. car-roof) may be coded in the
‘ordinary’ type of construction with the adnominal phrase, as (21a) and (22a).
But only with the BODY-PART relation in (21b) may the possessor (the son) be
coded in a dative phrase. Only the possessive relations that are intuitively
conceived of as ‘highly dense,’ ‘inalienable’ permit this marked construction
type. (The precise semantic restrictions of the construction in French cannot
be investigated here; cf. Spanoghe 1995; König and Haspelmath 1998 and
references cited therein for analyses.) Now, the relation between a possessor
and an (inalienable) possession is clearly a contiguity relation, because they
belong to the same domain (the human body): a foot, for example, is semanti-
cally contiguous to its bearer, because people know that people usually have
feet. This contiguity has, in this case, even been solidified as a grammatical
rule of French insofar as the possessor may be coded as a dative phrase. The
traditional term ‘inalienable possession’ is misleading, because it tells us little
about the nature of the relation in question. What is traditionally thought of as
inalienability boils down to a special instance of contiguity. Therefore, in-
alienable possession is another marked construction that inherently involves
contiguity.
The contiguity relation of the kind in question may give rise to a genuine
metonymy. There is a construction where the whole slides metonymically into
the syntactic slot initially reserved for a part. In French it is possible to say
(23a) as an alternative to (23b):
(23) a. Sylvie est jolie des yeux. (Frei 1972 [1939])
b. Les yeux de Sylvie sont jolis.
‘Sylvie’s eyes are beautiful.’
(23a) really says something about Sylvie’s eyes, but it is the entire person that
appears in the subject position. We are dealing with a kind of PART-WHOLE
metonymy in which a part (the eyes) is replaced by a whole (the entire person).
In Relational Grammar this is referred to as ‘possessor ascension’ (cf. Blake
1984).
In French and other contemporary European languages this construction
is only available for intransitive predicates. Only the subject can be affected
by such a metonymy, as König and Haspelmath 1998 have noted. So once
again, the hierarchy of arguments is confirmed: the subject is affected only if
there is no direct object. Now it would be nice, of course, to have examples
where possessor ascension applies to the direct object in order to have even
better evidence for the hierarchy. Here I draw again on the study carried out by
König and Haspelmath. There are languages where other arguments are
implied, and in most cases it is the direct object:
(24) a. Homeric Greek (König and Haspelmath 1998: 564)
tón rh’ ébalen kephalen hupèr oúatos
this.one PAST-hit head above ear
oxéi khalkoi
javelin:DAT of.bronze
‘He hit his head above the ear with the bronze javelin.’
b. Sierra Popoluca (Mixe-Zoque, Mexico) (Blake 1984: 437)
šéwan a-kuʔd-aʔy an-sék
John me-eat-indirect my-beans
‘John ate my beans.’
c. Korean (König and Haspelmath 1998: 562)
acessi-ka sikey-lul cwul-ul kochi-ess-ta
uncle-NOM watch-ACC chain-ACC repair-PAST-IND
‘The uncle repaired the watch-chain.’
(24b) literally says: ‘John ate me with respect to my beans.’ This is the same
kind of synecdochic relation as illustrated in the French example (23a), but
this time the direct object is concerned.
As König and Haspelmath report, this kind of construction involving
possessor ascension most often occurs with the subject or the direct object. All
their examples with possessor ascension affecting the subject contain intransi-
tive predicates. With the transitive predicates they report it is always an object,
never the subject, that is affected by the metonymy. This seems to provide
fairly good evidence for the claim that possessor ascension preferably applies
to direct objects.
4.3. A counterexample? The shift from agent to instrument
There is an alternation that seemingly provides evidence against the claim that
metonymies privilege the direct object. As is well-known, action verbs allow
the agent role to be replaced by the instrument role:
(25) a. Marie ouvre la porte avec la clé.
‘Marie opens the door with the key.’
b. La clé ouvre la porte.

‘The key opens the door.’
(26) a. Jean a facilement coupé le pain avec le couteau aigu.
‘Jean has cut easily the bread with the sharp knife.’
b. Le couteau aigu a coupé le pain facilement.
‘The sharp knife has cut the bread easily.’
These alternations metonymically exploit the contiguity of agents and instru-
ments. Their semantic relation is mediated by the general frame of actions
(actions always imply an agent and many of them also an instrument). And the
alternations always affect the subject. They cannot affect any other argument,
at least in active sentences, because in these, agents are always represented by
subjects (Givón 1984: 139f). Hence we are dealing with transitive sentences
that present a metonymically mediated alternation in the subject position. This
seems to challenge the claim of the direct object’s privileges with respect to
metonymy. However, I want to argue that my general claim is not contradicted
by these data, because these alternations can be shown to be a (systematic)
insertional metonymy which does not instantiate a particular construction;
therefore, they fall under the general caveat mentioned at the beginning of
Section 4.
How is this alternation to be classified with respect to the role vs.
insertion distinction that I have proposed in this paper? The alternation repre-
sents a contiguity of thematic roles in a broad sense, but it is not a role-level
contiguity in the sense of a verbal polysemy as defined above. It seems
implausible to analyze this alternation as a polysemy, given that it is appar-
ently so regular and so systematic. With any action verb, the agent can be
replaced by an instrument (provided the respective action actually implies an
instrument). Hence the alternation is predictable from the mere fact that the
verb is an action verb. Its predictability strongly suggests that it is not the
result of an individual semantic change. Rather, it is a general grammatical
process independent of individual verbs. Therefore, it cannot be considered an
instant of polysemy, according to the definition given in Section 2. Nor does
the AGENT-INSTRUMENT alternation instantiate a specific construction, because
the characteristics of the alternation seem to depend on the general properties
of action verbs, i.e., on the properties of a determinate lexical class. They do
not depend on a specific syntactic arrangement. To conclude, the AGENT-
INSTRUMENT alternation is a regular alternation that is neither an example of
polysemy nor a specific construction. Therefore it cannot be considered a

counterexample to my general claim.
5. Conclusion: Why the direct object?
The obvious question now is: why should the direct object — in an accusative
language like French — be so vulnerable to metonymic interpretation? I am
unable to offer a completely satisfying explanation, but here are some possible
suggestions.
Firstly, the direct object belongs — with the subject — to those argu-
ments that are semantically opaque. Oblique functions (dative, locative and
temporal adjuncts, etc.) are most often semantically transparent, i.e., their
preposition- or case-marking permits only one thematic role (or few of them).
The dative phrase usually has to be interpreted as an experiencer or a benefi-
ciary argument; locative prepositions clearly mark locative arguments, etc.
Subjects and direct objects, on the other hand, are wide open as to their
thematic interpretation, the exception being that the direct object cannot mark
an agent (cf. Givón 1984: 135–185). Clearly, ambiguities in verb arguments
are only possible within the range permitted by the overt marking of each
particular argument. It is therefore unexpected that verbs display a polysemy
that relies on an ambiguity of an oblique object. But subjects and direct objects
should easily facilitate contiguity-based polysemies, because these arguments
allow for a wide range of thematic interpretations.9 (The same can also be
seen in the examples of metonymy given in Fass 1991.) This point clearly
argues for the supremacy of the subject in relation to verbal polysemy, even
over the object. But perhaps the following argument makes us understand why
the subject is outranked by the object in the end.
The ‘outranking of the subject by the object’ only concerns transitive
sentences, because only in transitive sentences is there both a subject and an
object. The object in transitive sentences and the subject in intransitive
(unaccusative) sentences are most often ‘themes’ (according to common
inventories of thematic roles). It seems natural to assume that their sharing this
property should be relevant for an explanation of the hierarchy (4). Now, the
roles of individual verbs labeled ‘themes’ generally seem, in a very particular
sense, semantically more ‘specific’ than roles which are eligible for subjects in
transitives (agents, instruments, experiencers). Whereas it is possible to have
an idea of what an agent, an instrument or an experiencer is even without

thinking of particular verbs, the ‘theme’ is a very abstract notion whose
interpretation is maximally dependent on the respective verb’s content. Al-
ready Fillmore (1968: 24) recognized this problem and considered his “objec-
tive” case “the semantically most neutral case.” This means, in turn, that in
order to have an appropriate characterization of the theme of a given verb it is
necessary to provide a more detailed description of that role than would be
necessary for an appropriate characterization of, e.g., the actor of a given
verb.10 As an example, recall the alternation of risquer in (9). To describe its
subject, it was sufficient to speak of an ‘actor,’ which is a valid attribute for
any action verb. To describe its possible objects, however, it was necessary to
choose the more specific attributes ‘dangerous action’ and ‘endangered thing.’
These are specifications of what could also be called the theme of risquer.
That is, the semantic change yielding the polysemy of risquer could switch the
direct object from the entity ‘dangerous action’ to the contiguous entity
‘endangered thing.’ But no such specific contiguous entity would be available
for the ‘actor.’ In other words, the argument bearing the theme (direct object
or subject) can undergo semantic change more easily than arguments bearing
other roles. The semantic specificity of themes might be a reason for their
inclination to form contiguity-based polysemies.
Notes
1. I am indebted to Andreas Blank, Mary Copple, Ulrich Detges, Paul Gévaudan, Martin
Haspelmath, Hendrikje Scholl, and two anonymous reviewers for useful comments on
earlier versions of this paper.
2. The relevant literature on frames includes Fillmore 1975, 1976, 1977, 1985; Tannen
1979, Schank and Abelson 1977; Barsalou 1992.
3. As pointed out by an anonymous reviewer, it is sometimes difficult to determine whether
we are dealing with two alternating roles of one predicate or whether the different
interpretations merely reflect a polysemy of the entire predicate. In fact, both analyses are
normally possible. Therefore the decision is less an empirical than a theoretical one. It
corresponds to what is known as the ‘predicate-independent’ vs. the ‘predicate-depen-
dent’ approach to thematic roles. Given that this paper focuses on arguments, not on
predicates, it seems reasonable to opt for the predicate-independent approach, i.e., to
assume that different interpretations of a verb as in (3a) and (b) actually point to
alternations of thematic roles and not to differences of predicates, of which role alterna-
tions would only be ‘reflexes.’ But note that the polysemy approach adopted here forces
us to consider roles not in isolation, but as parts of the verb they are employed with. Only
verbs, by virtue of being lexical items, can be polysemous. Roles alone cannot, because
they are not lexical items. In short, the alternation in (3a) and (b) is a verbal polysemy
with respect to the direct object’s interpretation.
4. The historical data used in this paper are drawn from Le Robert Historique.
5. Cf., e.g., hériter de quelque chose/de quelqu’un ‘inherit sth./to be s.o.’s heir.’ In this case,
the contiguity of the inheritance and the heir applies to a prepositional object.
6. An anonymous reviewer objected that these alternations might simply boil down to
active-zone phenomena (Langacker 1984). Standard examples of active zones are We all
heard the trumpet or Don’t ever believe Gerald. The semantic content of their direct
objects does not precisely coincide with the entity designated by these objects, which
participates most directly in the process. A more ‘literal’ version would be We all heard
the sound of the trumpet or Don’t ever believe what Gerald says. Consequently, active-
zone phenomena overlap with metonymies (cf. Langacker 1993: 31). But there is a
crucial difference between active-zone phenomena and the alternations (5) to (10):
whereas active-zone examples are felt as non-literal speech and can be paraphrased by a
synonymous ‘literal’ sentence (as in the cited examples), frame-based alternations can-
not. Take examples (5a) and (b): the variants of balayer present slight but important
differences in meaning. Balayer ‘to sweep’ in (5a) has an incremental theme (the cleaned
room, place, etc. gets clean ‘step by step’) and this necessarily implies duration. The
variant (5b), however, has a holistic theme (things may be swept up ‘all in one go’) and it
does not necessarily imply duration. Accordingly, it is not always possible to replace the
variants one for each other (which again indicates that they actually differ in semantic
content). On the one hand, someone who has ‘swept up something’ has not always
thereby ‘swept a place.’ On the other hand, someone who has cleaned an already clean
place (i.e., without sweeping up anything), can nevertheless claim to ‘have swept it.’
7. This (ungrammatical) construction must, of course, not be confused with the impeccable
one se préoccuper de quelque chose ‘to worry about something.’
8. The PART-WHOLE relation is even iconically reflected in the reflexive clitic’s slight
morphological and phonetic weight, as was suggested to me by Peter Koch. Even on the
level of linguistic form, the reflexive’s antecedent is bigger than the reflexive itself, just as
the whole is bigger than one of its parts. Interestingly, the asymmetry between the reflexive
and its antecedent fades away as the reflexive marker grows in phonetic shape (which again
confirms the iconic nature of the part-whole relation in question). Préoccuper-type
constructions are sometimes better with a strong reflexive than with a reflexive clitic: je me
choque moi-même is better than je me choque ‘I shock myself’ (cf. Belletti and Rizzi 1988:
297f for analogous examples in Italian).
9. Languages may differ considerably with respect to the semantic range permitted by the
subject and the direct object. Clearly, arguments such as subject and direct object are
language-specific categories. The semantic spectrum of a given argument is therefore
highly dependent on particularities of the language’s grammar which could not be studied
here, e.g., the case system, word order rules, etc. The flexibility for metonymic reference
is also part of these language-specific properties. Cf. Hawkins (1986: 53–73) for differ-
ences of this kind between English and German, and Müller-Gotama 1994 for a cross-
linguistic study on semantic transparency in argument linking.
10. Koch 1981 offers an articulated theory on abstractness vs. concreteness of description of
thematic roles.
References
Anderson, Stephen
1971 On the role of deep structure in semantic interpretation. Foundations of
Language 7: 387–396.
Baker, Mark
1988 Incorporation: A Theory of Grammatical Function Changing. Chicago:
The University of Chicago Press.
Barsalou, Lawrence W.
1992 Frames, concepts, and conceptual fields. In A. Lehrer, E.F. Kittay (eds.),
Frames, Fields, and Concepts: New Essays in Semantic and Lexical Orga-
nization. Hillsdale: Erlbaum, 21–74.
Belletti, Adriana, Luigi Rizzi
1988 Psych-verbs and -theory. Natural Language and Linguistic Theory 6: 291–
352.
Blake, Barry J.
1984 Problems for possessor ascension: some Australian examples. Linguistics
22: 437–453.
Blank, Andreas
1997 Prinzipien des lexikalischen Bedeutungswandels am Beispiel der roma-
nischen Sprachen. Tübingen: Niemeyer.
Busse, Winfried, Jean-Pierre Dubost
1983 Französisches Verblexikon: Die Konstruktion der Verben im Franzö-
sischen. Second edition. Stuttgart: Klett.
Bréal, Michel
1897 Essai de sémantique. Paris.
Croft, William
Fass, Dan
1991 Metonymy, case role substitution and sense ambiguity. In D. Fass et al.
(eds.), Proceedings of the IJCAI-91 Workshop on Computational Ap-
proaches to Non-literal Language. Boulder, Colorado (Technical Report
CU-CS-SS0–91), 42–51.
Fillmore, Charles
1968 The case for case. In E. Bach, R.T. Harms (eds.), Universals in Linguistic
Theory. New York: Holt, Rinehart and Winston, 1–88.
1975 An alternative to checklist theories of meaning. Proceedings of the First
Conference of the Berkeley Linguistics Society. Berkeley, California: BLS,
123–131.
1976 Frame semantics and the nature of language. Annals of the New York
Academy of Sciences 280: 20–32.
1977 Scenes-and-frames semantics. In A. Zampolli (ed.), Linguistic Structures
Processing. Amsterdam: North-Holland, 55–81.

254.
Frei, Henri
1972 [1939] Sylvie est jolie des yeux. In A. Sechehaye (ed.), Mélanges de
linguistique offerts à Charles Bally. Geneva: Slatkine, 185–194.
Givón, Talmy
1984 Syntax: A Functional-Typological Introduction. Vol. I. Amsterdam and
Philadelphia: Benjamins.
Grimshaw, Jane
1990 Argument Stucture. Cambridge, Massachussets: The MIT Press.
Hawkins, John A.
1986 A Comparative Typology of English and German: Unifying the Contrasts.
London: Croom Helm.
Jackendoff, Ray
1992 Mme Tussaud meets the binding theory. Natural Language and Linguistic
Theory 8, 1–31.
Koch, Peter
1981 Verb-Valenz-Verfügung: Zur Satzsemantik und Valenz französischer Ver-
ben am Beispiel der Verfügungs-Verben. Heidelberg: Carl Winter Univer-
sitätsverlag.
1997 La diacronia quale campo empirico della semantica cognitiva. In F.
LoPiparo (ed.), Linguaggio e cognizione. Atti del XXVIII Congresso.
Rome: Bulzoni, 225–246.
König, Ekkehard, Martin Haspelmath
1998 Les constructions à possesseur externe dans les langues de l’Europe. In J.
Feuillet (ed.), Actance et valence dans les langues de l’Europe. Berlin:
Mouton de Gruyter, 525–606.
Lakoff, George
1984 Active zones. Proceedings of the Tenth Annual Meeting of the Berkeley
Linguistics Society. Berkeley, California: BLS, 172–188.
Le Guern, Michel
Le Robert Historique
1992 Dictionnaire historique de la langue française. Paris: Dictionnaires Le
Robert.
Müller-Gotama, Franz
1994 Grammatical Relations: A Cross-linguistic Perspective on Their Syntax
and Semantics. Berlin: Mouton de Gruyter.
1995 Transfers of meaning. Journal of Semantics 12: 109–132.
Olsen, Susan
1994 Lokativalternationen im Deutschen und Englischen. Zeitschrift für
Sprachwissenschaft 13: 201–235.
Schank, Roger C., Robert P. Abelson
1977 Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowl-
edge Structures. Hillsdale, NJ: Erlbaum.
Spanoghe, Anne-Marie
1995 La syntaxe de l’appartenance inaliénable en français, en espagnol et en
portugais. Frankfurt a.M.: Lang.
Tannen, Deborah
1979 What’s in a frame? Surface evidence for underlying expectations. In R.O.
Freedle (ed.), New Directions in Discourse Processing. Norwood, NJ:
Ablex, 137–181.
Winston, Morton E., Roger Chaffin, Douglas Herrmann
1987 A taxonomy of part-whole relations. Cognitive Science 11: 417–444.
Zribi-Hertz, Anne
1978 Economisons-nous: A propos d’une classe de formes réflexives
métonymiques en français. Langue Française 39: 104–128.
Putting Metonymy in its Place
Paul Pauwels
KVH Antwerpen
1. Characterizing Metonymy1
When Lakoff and Johnson (1980) put figurative language at the top of the
agenda, claiming that it involved basic cognitive processes rather than deviant
usage, their work gave an enormous boost to metaphor research, but not to the
study of metonymy. After all, they did call their work Metaphors We Live By,
and in devoting only a single chapter to metonymy, they seemed to imply it
was a minor process in comparison. Moreover, they mainly defined me-
tonymy in comparison with metaphor.
Metaphor is principally a way of conceiving one thing in terms of another,
and its primary function is understanding. Metonymy, on the other hand, has
primarily a referential function, that is, it allows us to use one entity to stand
for another. But metonymy is not merely a referential device. It also serves the
function of enhancing understanding. [...] determines which aspect we are
focusing on. (Lakoff and Johnson 1980: 36)
It is this focusing behavior which gives metonymy its special cognitive status.
Examples (1) and (2) illustrate what is meant by ‘focusing.’
(1) The Times hasn’t arrived at the press conference yet.
(2) We need some good heads on the project.
In (1), which follows the WHOLE FOR PART pattern (institution for member of
the institution), the metonym is said to focus on the status of the reporter thus
referred to. In (2), the (body)part which is selected to refer to the whole
person, focuses the intelligence of that person as a particularly relevant
256 Paul Pauwels
characteristic. The grounding of figurative expressions is, for Lakoff and

Johnson, more obvious in the case of metonymy than in the case of metaphor
since it “usually involves direct physical or causal associations” (1980: 39).
Metonymy is clearly the poor member of the family here: it only ‘focuses an
aspect,’ and it is also more obvious and less creative — hence presumably less
worthy of the researcher’s attention.
Lakoff (1987) approaches the difference between metaphor and me-
tonymy against the background of an already more elaborate cognitive frame-
work, using the notions of (cognitive) ‘domain’ (any sort of conceptualization
relative to which semantic structures are characterized) and ‘idealized cogni-
tive model’ (a kind of gestalt which gives structure to reality). For Lakoff, “a
metonymic mapping occurs within a single conceptual domain which is
structured by an ICM,” whereas “metaphoric mapping involves a source
domain and a target domain” (1987: 288). The “direct associations” (Lakoff
and Johnson 1980: 39) which characterize metonymy are here (re)defined in
terms of membership within a single cognitive domain. The central element
differentiating metaphor and metonymy seems to be the greater cognitive
distance (to put it in spatial terms) between the concepts involved in metaphor.
Taylor (1989) remarks that metonymy has received little discussion in
comparison to metaphor. He takes a broad view of metonymy which not only
encompasses conventionalized metonymy as in (1) and (2) above, but which
also accommodates situationally sanctioned examples like Lakoff and
Johnson’s famous ‘ham sandwich’ example and common contextual modula-
tions of meaning as in (3):
(3) a. He opened the door.
b. He walked through the door.
For him “the essence of metonymy resides in the possibility of establishing
connections between entities which co-occur within a given conceptual struc-
ture” (1989: 123f). In this view, it is not surprising that “metonymy turns out
to be one of the most fundamental processes of meaning extension, more
basic, perhaps, even than metaphor” (1989: 124).
In reaching this conclusion, Taylor is supported by Goossens (1990,
1995). Looking at diachronic evidence, Goossens comes to the conclusion that
metonymy is much more central in our conceptual apparatus than is usually
thought, and, on top of that, it is often at the basis of metaphor. The following
examples are taken from Goossens (1990):

(5) have a word in someone’s ear
(6) I could bite my tongue off.
In cases like (4), for example, it is not clear whether the domain of giggling
and that of (light-hearted) linguistic action are conceived separately. We are
still aware of the metonymic basis in the metaphorical interpretation, a fact
that Goossens (1990: 328) calls ‘metaphor from metonymy.’ In these cases,
the difference between metonymy and metaphor lies in the distinction be-
tween a ‘while’-interpretation and an ‘as if’-reading, respectively. In most
such cases it seems the ‘as if’ reading is most likely, although the metonymic
basis cannot be discounted completely. This is illustrated by (5): speaking
privately with someone may involve whispering into his/her ear, but the
expression can be used to conceptualize situations where this is not actually
happening. Another ‘blend’ Goossens distinguishes is ‘metonymy within
metaphor,’ where a shared element bridges the gap between the metaphor’s
source and target domain, functioning metonymically in the target domain, as
in (6). Tongue is used literally in the source domain, but in the framework of
the metaphorical usage (depriving oneself of one’s speech as some kind of
punishment) it functions as a metonym for the speech faculty. It should be
clear from the foregoing that, in actual usage, the line between metaphor and
metonymy is not always easy to draw.
Against this background, Croft (1993) can be read as an attempt to come
to a more narrowly defined concept of metonymy. Like Taylor, he discusses
the difference between metaphor and metonymy against the background of
other processes of meaning extension. He relies on Langacker’s notion of
‘domain matrix,’ i.e., the situation which occurs when the base against which
an expression is profiled refers to the intersection of several domains (which is
the case for most expressions). Lakoff’s (1987) distinction is rephrased as
follows:
Metaphor is a mapping between two domains that are not part of the same
matrix; [...] In metonymy, on the other hand, the mapping occurs only within
a domain matrix. [...] We will call this conceptual effect domain highlighting,
since the metonymy makes primary a domain that is secondary in the literal
meaning. (Croft 1993: 348)
At the same time, he notices a similarity between metonymy and other, non-
figurative, kinds of lexical ambiguity. Unlike Taylor (1989) Croft does want
258 Paul Pauwels
to differentiate between the two. The shift in (7a) and (b), for example, does
not involve figurativeness, since there is no ‘shift of reference,’ while there is
such a shift in (8a) and (b).
(7) a. This book is heavy. (in the literal sense, i.e., weighs a lot)
b. This book is a history of Iraq.
(8) a. The Times is owned by R. Murdoch.
b. The Times hasn’t arrived yet.
(7a) and (7b) highlight different primary domains of the concept book, viz. the
object-domain in (7a) and the contents-domain in (7b); the difference between
(8a) and (8b), on the other hand, lies in the fact that only (8a) highlights a
primary domain — focusing the newspaper as a company — while (8b)
highlights a secondary domain — focusing the newspaper as an entity which
can be represented by a correspondent. So far, the characterization is relatively
straightforward.
Croft then considers further examples. In a footnote he describes
Goossens’ example (4) of metaphor from metonymy as a case of domain
mapping, which is complicated by the fact that the source domain (sound) is
one of the domains “which happens to be in the matrix of the target” (speak-
ing) — hence the appearance of being metonymy (1993: 367). Here, Croft’s
argument is not as convincing. Within the domain matrix of linguistic action,
there obviously is a subdomain relating to the manner of speaking. In some
cases, this manner can indeed involve giggling (in a way which would not be
possible for say ‘barking,’ or indeed, ‘thundering’ which do not describe
human sound in their prototypical meaning — which is Goossens’ point). In
those cases, the use of giggle would simply be the highlighting of a secondary
domain, i.e., metonymy. In situations where the manner of speaking is con-
ceived in terms of giggling, i.e., speaking in a particular way as if it were
giggling, though actually no giggling is involved, there is domain mapping
and metaphor. The difference between such cases of metonymy and the more
central ones like (8b) above is twofold. First of all, these cases are situationally
invoked; there is no permanent link between the concepts involved. Secondly
— and this may relate to the first difference — the metonym deals with a
conceptualization expressed by a verbal expression which has a different
(probably more complex) kind of domain matrix than a conceptualization
involving a nominal expression. As things stand, Croft does not seem to want
to make room for overlap between metonymy and metaphor in his model.
In what seems to be a further move away from recognizing distinctness of
domains in metonymy, he even questions the relation between metonymy and
highlighting.
It may not be the case that domain highlighting within the domain matrix of a
word is involved in all cases of metonymy. In some cases the shift in domain
prominence in the matrix is quite subtle, and sensitive to the semantics of the
associated words. [...] [I]t appears that no domain selection is involved. (Croft
1993: 350)
In example (9), which contains a metonym of the type PART FOR WHOLE, there
is no novel domain selection:
(9) We need a couple of strong bodies for our team.
The domain of the body is subsumed under the domain of the human being, and
if one subscribes to the encyclopaedic view on meaning, this means there is no
domain selection, but a “subtle shift in domain prominence” (Croft 1993: 350).
Summarizing, Croft distinguishes five possible cases: distinct domains
and mapping between them (metaphor); distinct domains of which one hap-
pens to be in the matrix of the other, and mapping (metaphor); single domain
matrix and highlighting of a secondary domain (metonymy); single domain
matrix and a difference in domain prominence (metonymy); and single do-
main matrix and highlighting of a different primary domain (literal lexical
ambiguity).
Finally, using Langacker’s (1987: 300) distinction between ‘dependent
predications’ (which depend on other predications for their elaboration) and
‘autonomous predications’ (which elaborate the dependent ones), Croft offers
an explanation for the high incidence of nominal examples in most discussions
of metonymy. He suggests highlighting occurs on autonomous predications,
while mapping occurs on dependent ones. He takes care, however, to empha-
size that metonymy is not restricted to nouns — the autonomous predications
par excellence — but that also verbs can behave autonomously with respect to
adverbs, and that nouns can behave as dependent predications with respect to
other nouns.
In most of the above, the issue seems to be the way to distinguish between
the working of metaphor and metonymy. Lakoff and Johnson (1980), Lakoff
(1987) and also Croft (1993) draw the line by referring to the distinctness of
domains, while Goossens (1990, 1995) wants to accommodate overlap. The
260 Paul Pauwels
notion of domain matrix and the composition of domain matrices seem to play
a crucial role in all analyses.
In a recent paper, Langacker (1994) comments on the importance which
distinctness, as opposed to continuity, plays in conceptualization. In his view,
most phenomena which have been described as continuous will yield to an
alternative analysis in terms of discrete categories upon further research.
There may be an exception, though, which is highly relevant to the present
discussion:
An important aspect of linguistic semantics is our ability to construe one
structure against the background provided by another. [...] I wish to empha-
size a parameter that may well be continuous: the salience of the background
structure, i.e. its level of activation in the construal of the target. (1994: 17)
This could be taken to mean the discussion as to whether a particular expres-

sion is metonymic or metaphorical is in some cases academic since it depends
on the perceived salience of the domains involved, which is undoubtedly also
influenced by contextual factors, and not just by the nature of the expression.
2. Getting a grip on the issue
The present investigation takes as its starting point a corpus of utterances

containing one of four related verbs of manipulation, i.e., put, set, lay and
place. Out of an overall corpus of 3,125 examples based on the LOB Corpus
(Johansson et al. 1978) and the Leuven Drama Corpus (Geens et al. 1975),
some 220 examples were selected as relevant to metonymy. This represented
approximately 10% for put, set and place, and 28% for lay.
Using a corpus structured around one central domain (manipulation), it
was thought, would help us shed more light on the types of relationships
between the domains involved in these examples of metonymy.
Since, moreover, the corpus provides us with examples of utterances in
context, we can also try to gain more insight into the motivation for different
metonyms. In Pauwels and Simon-Vandenbergen (1995) it has been shown
that a lot of metaphors carry value judgments and that this may be the way in
which they ‘enhance our understanding.’ Here we will look for similar func-
tions and motivations for metonymy. Finally, taking a restricted corpus of
3,125 examples as our starting point has allowed us to cast the net wide. We
will also be exploring the boundaries of metonymy, and try to find elements in
the corpus which allow us to see what is more central and what is more
marginal to metonymy.
In our discussion we will refer to the notions of ‘cognitive domain,’
‘source’ and ‘target domain,’ ‘domain matrix,’ ‘mapping’ and ‘highlighting,’
and ‘dependent’ and ‘autonomous predications,’ as characterized above. We
will also make use of Langacker’s (1987: 231ff) distinction between ‘trajec-
tor’ (TR) and ‘landmark’ (LM) to describe the profiling of the state of affairs
in the sentence. In our corpus of manipulation verbs, the prototypical pattern is
a three-place predicate profiling two TR-LM relationships, as in (10).
(10) He put the book on the table.
(11) The constable put the spade in once.
(12) Adams was lighting and placing candles.
(11) and (12) are common alternatives. In (11) the secondary LM is not
expressed, or hidden, while in (12) the verb is used as a two-place predicate
profiling only one TR-LM relationship.
3. Variation in the corpus
A preliminary survey of the corpus already yields a quite varied set of

metonyms based on relationships of varying complexity. It is clear that
metonymic relations do not only occur at the level of the lexicalized concept,
but also at the level of the more abstract unit of a state of affairs.
(13) She continues to lay tea.
(14) She has put your tea out.
(15) (Croupier:) “Place your bets, please, ladies and gentlemen!”
Examples (13) and (14) can still be considered as a fairly straightforward case
of nominal metonymy, where the tension between the verbs lay and put which
conceptualize a manipulation of objects and tea which conceptualizes a drink
is resolved by invoking a PART FOR WHOLE metonymy. The drink which is
central to the meal which is being put on the table, is taken to refer to the
whole. The metonym has the function of focusing the status of the meal.
An example like (15) is already more complex. The relationship between
262 Paul Pauwels
the concept of bet and the actual object which is being placed is of a more
complex nature. The bet is created by the act of placing money in a certain space
on the table. The relationship can be described as that between the result of an
action and an object which is instrumental in achieving that result, with the
RESULT standing for the INSTRUMENT. By referring to the result, the metonym
focuses the function of the action. Similar examples were found in put on a
disguise, lay bait and place an order. In the latter, however, the physical act of
placing is already much less central in the process, so much so, even, that the
example would preferably be analyzed as a metaphor from metonymy, where
there is a mapping from the domain of manipulation to the domain of ordering
goods.
(16) Father? Put your heart into it. Love me, come on. Have a shot at
filial devotion.
(17) I’m putting it aside for my old age.
(18) a. I do put my foot down about discussing them.
b. Just the once. I had to put my foot down there.
Examples (16)–(18) provide further illustration of the interpenetration of
metonymy and metaphor. (16) is a case of metonymy within metaphor:
making an effort to feel is metaphorically described as putting in, i.e., an act of
manipulation, while heart refers metonymically to the effort in which it is
crucially involved. Examples (17) and (18), although metaphorical to all
intents and purposes, have a metonymic basis: in both cases, situations where
the describing and the described act co-occur are easily conceivable —
indeed, it is probable that such situations are at the basis of the expressions —
which makes them typical cases of metaphor from metonymy. Examples (17)
and (18) are different from the foregoing in another respect: they could also be
taken literally. In (17) and (18a), however, there is still an element inside the
sentence pointing to the need for a reinterpretation in figurative terms. In (17)
there is the prepositional phrase introducing the element of long-term plan-
ning, and in (18a) the activity is linked to linguistic action by a prepositional
phrase. In (18b), however, the figurative interpretation depends on the wider
context.
(19) The power of life and death — no need of other vices. If you’ve
once put on the black cap, everything else tastes like wax fruit.
(20) The board is laying a cable along a seven mile route in Surrey.
(21) The church has no views on drains, gasworks or bricklaying.
Examples (19)–(21) clearly demonstrate that also pure metonymy can be
context induced. In all three cases, the act of manipulation is a necessary part
of the action complex for which it stands; an ‘as if’ interpretation is ruled out.
The metonyms are at the same time referential and clarifying, since they focus
on the most visually salient aspect of the action complexes. In (20) there is
even a further metonym which, to put it in Croft’s terms, is induced by the
dependent predication (laying a cable) on the autonomous one (the board):
the board is the institution responsible for the action but is not the actual agent.
Most of the above examples (13)–(21) are characterized by a tension
between the concrete/specific and the abstract/general. In (13) and (15) there
are two factors contributing to this tension. First, there is the fact that the verbs
profile only one TR-LM relationship, which means there is an inherent
vagueness in comparison to the fully descriptive pattern (exemplified by (10)).
The less tangible character of the LM (bets, tea) outside the domain of
manipulatable objects is the second element. In (14), which is closer to the
fully descriptive pattern, the secondary LM is hidden, which again creates
vagueness, and the primary LM (tea) is again less tangible. In (19), a hidden
secondary LM combines with a primary LM which is marked as generic by the
definite article (the black cap), and in (20) the indefinite article introduces
indeterminacy into the primary LM (a cable), while a second metonym makes
the TR more abstract. (21) is probably the most striking example of the shift
towards abstraction, since the specific description has been turned into a
(necessarily) more general concept. The ‘exceptions’ are (16)–(18), where
metonymy and metaphor interact.
4. Salient source domains
This section focuses in more detail on the relationship between frequently

occurring source domains and their targets and describes the extent to which,
and the level of abstraction at which, there is interpenetration of domains. At
the same time the grouping together of similar expressions may provide
further insights into the motivation for such metonymies and their function.
264 Paul Pauwels
4.1. Manipulating body parts
A first salient source domain involves the manipulation of body parts. In all,
some 60 expressions containing references to hand, finger, fist, foot, knee,
eyes, face and heart were found.
(22) She has glasses of whisky standing around at vantage points, to
which she would put her lips when so disposed.
(23) He was unlikely ever to set foot in his kingdom again.
(24) a. I’d strangle any man as much as laid a hand on her.
b. The first person who sets a finger on anyone of his family...
c. Trouble is I’m practically certain I’ve never laid eyes on you,
mate. (set)
d. I’ll tie a lovers’ knot with his arms and throw him in the Tiber if
I ever lay my tentacles on him.
In (22) drinking is described in terms of an action preparatory to the drinking.
The effect is to draw attention to the behavior of the person, rather than to the
actual drinking. Whether she is just wetting her lips or whether she is drinking
hard is left to be inferred. The effect is euphemistic, but there is of course the
irony added by the first part of the sentence.
It is probably best to consider cases like (23) as metaphorical in present-
day usage; the means of entering a country are so diverse and the prototypical
case of bodily movement is now only used in a context of overstatement.
There is a similar connotation of overstatement about the examples in
(24). In (24a) and (b) assaulting someone is referred to in terms of physical
contact — even the slightest touch is prohibited. In (24a) the overstatement is
as much the effect of the intensifier as much as, as of the description lay a
hand (not even two hands); (24b) minimizes the extent of contact (not even a
complete hand) to achieve a similar effect. Seen in the context of (24a) and (b),
(24c) describes an even more tenuous kind of contact. The expression derives
its value from the tension between the laying of a body part which would seem
to indicate physical contact, and eyes which, in this case, denotes a greater
distance. In (24d) the overstatement runs in the other direction: rather than
minimize the extent of physical contact as in the threats (24a) and (b) or the
disclaimer (24c), the extent of physical contact is maximized for effect. Again,
there is a shift away from the metonymic origins — these are still clearly
present in an example like (24a), in so far as the hands are the prototypical
means of assault. (24c) and (24d), however, should be considered as pure
metaphor. In (24c) ‘seeing’ is grasped in terms of physical contact, with eyes
as a metonymic bridge (a case of metonymy within metaphor), while in (24d)
getting hold of someone with the aim to inflict physical harm is described in
terms of the octopus’s, usually unwelcome, attentions.
In sum, roughly 50 of the examples on this domain carry a connotation of
euphemism or, conversely, of overstatement. This connotation is often con-
textually supported, and the sentences often contain conditionals or markers of
exceptionalness (if ever, the first who, the first time, as soon as, (neg+) before,
once, everything, ...). The recipient domains in most cases refer to socially
unacceptable behavior such as excessive drinking, suicide/killing, fighting,
harassing, anger, or shame.
Again we note the tension between specific and general in most ex-
amples. In (23) there is a difference in ‘scale’ between the primary LM (foot)
and the secondary LM describing its target-location (kingdom). It could be
argued that this case is an exception; in other examples like in this house, in
this room the tension is less obvious. Still, the fact remains that literal
descriptions usually have as secondary LM either a physical object, a sub-
stance, or a surface. Similarly, in (24), there is a tension between the primary
LM (body part) and the secondary LM (person as a whole), while in literal
descriptions the focus is more narrowly on the contact between body parts as
in lay a hand on someone’s arm. It is this tension which is resolved in the
metonymic interpretation. In (22) the tension is between the specificness of
the description of the contact between the lips and a glass, and the frequency
suggested by the plural glasses in the antecedent.
4.2. Manipulating human beings
A second subset with approximately 20 examples describes human beings as

manipulated entities. It should be clear from the start that babies, children,
unconscious persons, bodies and willing adults can indeed be lifted, trans-
ported and laid, set, put and placed in a literal sense. In the examples,
however, the manipulation stands for a more elaborate kind of interaction.
Motivations for these uses can be found in euphemism or, conversely,
dysphemism.
266 Paul Pauwels
(25) I can lay a woman any time I want.

(26) Lucy’s laid him out in his dress uniform.
(27) a. She laid him out flat. (cold)
b. She was out cold but still breathing.
(28) I ought to put you across my knee. Try it. You take one step
towards me and I’ll...
(29) “But I always put my babies on pots straight away,” said old Mrs.
Harford, reprovingly.
In (25) the sexual act is referred to by means of the prototypical position of the
participants (i.e., lying down). On the one hand the expression is quite graphic
— which accounts for the selection of lay to function metonymically. On the
other hand, the use of the verb as a two-place predicate without profiling a
secondary LM results in a certain vagueness. At the same time the causative
aspect in the verb is stressed. The overall effect seems to be one of
dysphemism, which is probably a consequence of the focus on the physical
and causative aspects of the interaction. Vagueness and euphemism are appar-
ent in (26), where the preparation of a body for burial (not only a disagreeable
topic, but also one which is ‘shrouded’ in mystery for most people, one can
presume), is referred to by invoking the resultant state of a body stretched out.
Again the action of bringing the body in this state is a necessary part of the
complete process. Example (27a), though similar at first sight, is the result of a
completely different kind of meaning extension. Here, the usage is based on a
metaphorical usage of out (cold/flat), as exemplified by (27b). This usage is at
the basis of expressions like knock someone out, which is here being
metaphorized by means of lay someone out flat/cold, which focuses the result
rather than the means by which this result is arrived at. The causative aspect of
lay is again foregrounded by this usage. In how far the scene in (26) is invoked
in the sanctioning of (27a) is a moot point — that there are possible bridges is
obvious. The final two examples (28) and (29) are at the same time vague —
the really disagreeable bit is not mentioned — and explicit — there is no doubt
as to what is being referred to. They probably originated as euphemism, but
they can also be used as dysphemism.
In all of the above, the metonyms focus on one stage of a sequence which
is characterized by unity of place, time, and participants in the interaction —
not unlike Fillmore’s scenes. The events referred to (having sex, preparing a
body, spanking, toilet-training) are (or used to be) taboo topics. The subdo-
main of manipulation is not only selected because it avoids the taboo; there is
also the concreteness of manipulation which provides the necessary salience,
and the focus on the control exerted by one participant over another, which is
a central element in the event.
4.3. Manipulating instruments
In the following, a selection from 35 examples, more prototypical scenes of

manipulation are involved in metonymic patterns. A first set of expressions,
with lay as a central element, has ‘stopping’ as a target domain.
(30) It is not difficult to lay down The Conquest of Peru or Volume 2 of
The Cambridge Medieval History, once it has been taken up; but it
is not so easy to feel altogether happy about never taking it up again.
(31) Carpentry is an occupation that lends itself to being laid down at
will, either temporarily or permanently.
(32) Ron, demobbed from home guarding, gladly laying his rifle aside,
built me a fruit cage.
In (30), to stop reading is referred to by means of laying down a book — a
physical activity which indeed entails the more general complex of activities,
and which seems intrinsically bound up with it. This is in contrast with (31),
which should probably be considered as metaphor pure and simple: the very
general activity of stopping is here being conceived in terms of stopping to
manipulate. There still is a metonymic link in the background, since carpentry
is prototypically a manual job which requires manipulation of tools and objects.
In (32) as well, the link between source and target is less strong: soldiering and
rifle-carrying are of course related, in so far as one of the main aspects of being
a soldier (in the public mind) is the obligation to ‘take up’ arms, but it is not
inconceivable for a soldier never to carry arms. Moreover, there is no ‘entail-
ment’ relationship between source and target as is the case in (30).
(33) Plant plastic gnomes where you walk and even, perhaps, put pen to
paper and relate your experiences in the force.
(34) I’ll cook something, you can wash your hair, you can relax, we’ll
put on some records.
268 Paul Pauwels
(35) a. If that’ll be all sir, I’ll let her in and put a kettle on.
b. I’ll put on an egg.
A second set of examples, where put is the central verb, all involve the starting
up of a process. In (33) the act of writing is referred to by means of the
manipulation of the prototypical instrument involved, or rather the first move-
ment of that instrument. In so far as this expression is used to refer to ‘manual’
writing, it can be considered as metonymy. If it were used, consciously, to
refer to writing by means of other instruments, this would be a case of
metaphor from metonymy (cf. also Pauwels 1995: 150). Similarly, in (34), a
fairly simple act of manipulation which occurs at the beginning of a sequence
of activating a recordplayer and playing a record, is used to refer to the
complete sequence. In this case, the metaphorical potential is possibly greater,
for two reasons. For one, the classic turntable is on its way out and being
replaced by the CD-player, where the record is not put on in the literal sense.2
On top of that, this might activate another path via which the usage can be
sanctioned, i.e., the metaphorical put on as in put on the radio. Example (35a)
falls into the same pattern as the above, with both the potential of a metonymic
reading (‘on the stove’) and a metaphorical reading (‘electric kettle’). An
example like (35b) clearly demonstrates a further shift away from the literal,
descriptive usage, since the instrument involved in the boiling of the egg
(indeed, boiling, rather than any other way of cooking an egg, is referred to
thus) is not mentioned.
In cases like the above, the metonyms appear to be mainly referring
devices. Referring to a complex activity by means of its fairly concrete first or
last stage is basically a time-saving strategy. Rather than focus on a specific
aspect, or euphemistically avoid another (the main strategies encountered so
far), this type of metonymy tends toward imprecise language use, exploiting
Grice’s maxims, one of which states one need not be more specific/precise
than necessary. The fact that quite a few of these examples involve closely
scripted events undoubtedly plays a role here. The selection of lay to concep-
tualize ‘stopping’ and put to conceptualize ‘starting’ relates to the prototypical
meanings of these verbs (see further Section 6) — a full discussion is beyond
the scope of this paper.
Not all the examples of manipulating instruments are of the same type,
however. Example (36) is clearly much more closely linked with the examples
in the previous sections as far as the motivating factor behind the metonym is
concerned.
(36) “Put a pillow on his face and get him out of it,” said Paul to himself.
Although the actual act of putting a pillow on someone’s face falls into the
pattern of referring to the start of a scripted sequence of actions, there is, in this
case, an additional aspect of euphemism which provides further motivation for
the partial selection.
This is also true of (37), which is however even more complex, since
there is a complete interpenetration of metaphor and metonymy on different
levels.
(37) Wipe that smile off your face and get up, before I put you on the
governor’s report.
At the highest level, this is again a case of the first stage in a sequence standing
for the complete sequence: writing someone’s name down in an official
register is used to refer to the consequent punishment. However, the writing
down itself is metaphorized in terms of manipulation, where that which is
written down is referred to metonymically. The result of the complete se-
quence is a ‘demetonymization’ (cf. also Goossens 1990: 335) since in the end
the person is literally the affected entity.
5. Target domain linguistic action
In this section I will take a closer look at the nature of the link between source
and target domain from the perspective of the target domain. Of the examples
found in the corpus, quite a few, with put as the central element, have already
been discussed in some detail in Pauwels (1995). Also, examples from previ-
ous sections fit in here.
(18) a. I do put my foot down about discussing them.
(30) It is not difficult to lay down The Conquest of Peru or Volume 2 of
The Cambridge Medieval History, once it has been taken up; but it
is not so easy to feel altogether happy about never taking it up again.
(33) Plant plastic gnomes where you walk and even, perhaps, put pen to
paper and relate your experiences in the force.
(37) Wipe that smile off your face and get up, before I put you on the
governor’s report.
270 Paul Pauwels
(38) The draft was laid before parliament.

(39) Let me put your mind at ease and say that Gregory has certainly no
intention of wanting to marry me.
The source domain of (18a), i.e., manipulating body parts, is not only related
to the target domain of linguistic action (LA) by virtue of the coalescing in
time of the two actions. The body is used frequently in non-verbal communi-
cation, and it is this background — the communication domain — which
supports this metaphor from metonymy.
In (30) and (33), the specific subdomains of reading and writing centrally
involve the manipulation of instruments. The source domain is clearly in the
matrix of the target. This is also true of special cases of making official
statements, where the introduction of a topic to an audience is usually accom-
panied by the introduction of a written document at the same time, as in (38).
Since the addressee is often not only the listener, but also the undergoer of
an LA, it is not surprising that metonyms involving the person addressed/
affected are often embedded in LA metaphors. Especially metaphorizations
involving the source domain of manipulation are prime candidates, as is clear
from (37). (39) is slightly different, since the metonym does not highlight the
‘whole’ person, but rather the mind-‘part’.
In summary, there appear to be three types of metonymy from the target
domain point of view. There are two types of event-metonymy, one in which
source and target domain are both subsumed by the more general domain of
non-verbal communication, as in (18), and a second one where the source
domain is subsumed by the target domain of LA, as in (30), (33) and (38). The
third type are nominal metonymies embedded in LA-metaphors which acti-
vate a more general domain, as in (37), or a more specific domain, as in (39).
6. Metonymy, generalization and the extension of verbal

meaning
In quite a few of the examples of metonymy discussed above where the

metonym was invoked to resolve tension at the sentence level between the
dependent predication and the autonomous one, the origin of the tension could
be traced to a discrepancy in the scale or level of abstraction of the concepts
involved. Indeed, the most pervasive patterns of metonymy are those which
have traditionally been described as PART FOR WHOLE or WHOLE FOR PART.
Looking at, for example, Lakoff and Johnson’s list (1980: 38f) it is clear that
most of their cases can be construed as more specific kinds of these two basic
types. PRODUCER FOR PRODUCT, OBJECT USED FOR USER, CONTROLLER FOR
CONTROLLED, INSTITUTION FOR PEOPLE RESPONSIBLE, PLACE FOR INSTITUTION,
PLACE FOR EVENT can all be characterized by invoking a ‘container’ relation-
ship between two domains: a superordinate one and a subsumed one. In
analyzing these examples as metonymy, the stand-for relationship is taken to
resolve the semantic tension.
However, the tension can be, and, judging from the evidence, often is,
resolved in another way. Take example (40), which can be described as a
straightforward case of metonymy, where table stands for the objects which
are being laid on it.
(40) I’ll just lay the table. (LET/Ayckbourn)
The autonomous predication (table) is the locus of the effect induced by the
dependent one (lay). This is however not the full story: the preferred status of
goal-objects appears to have resulted in an accommodation of this fact in the
semantics of the verb, which has resulted in a generalization. Rather than
‘manipulate,’ lay can be taken to mean ‘arrange, fix, prepare or make ready.’ In
other words, the meaning of the dependent predication appears to have shifted
as well. Also, the syntax of the verb has been affected by this: rather than profile
the primary LM of the lay-prototype (the goal-object which is being moved, as
in example (12)), this two-place predicate only profiles the secondary LM of
that prototype (the location involved in the movement). That this pattern has
proved productive, should be clear from the list of examples under (41).
(41) set the table, set the alarm, set a fuse, set the thermostat, set a watch,
set the oven to the required temperature, ...
Frequencies in the corpus seem to indicate that set has evolved furthest in this
direction. There are some examples with lay in this pattern, but no examples at
all for put and place. For put, this could be explained by the fact that the
original prototype concerns bodily movement (pushing, shoving) rather than
manipulation. In this prototype, the secondary LM with respect to which the
movement can be characterized is the agent her/himself. The use of put to
profile manipulation of an object with respect to a location is a later, extended
use, and in this literal meaning put needs the presence of at least a hidden
272 Paul Pauwels
secondary LM, which can then not move up in the hierarchy, so to speak, to
take the place of the primary one. Place, finally, is a much more recent
addition to the language: set and lay are attested in OE, put from late OE
onwards, while place is a 15th century loan from French.
7. Conclusion
The previous sections have tried to present a survey of metonymy as it

occurred in a specific corpus. From the survey, it soon became clear that the
majority of the examples were not of the ‘traditional’ nominal kind. If any-
thing, it seems like most nominal metonyms occur inside metaphors. Me-
tonymy often seems to function as a kind of ‘avoidance strategy,’ for reasons
of euphemism perhaps. Conversely, it also serves as a ‘focusing strategy,’
which in extreme cases results in dysphemism. The relationship between the
domains involved in metonymy seems to be one of inclusion, with either a
more general concept standing for a more specific one, or vice versa. There
will always be a difference in scale or level of abstraction, which is not the
case for metaphor.
Finally, it should be clear that metonymy cannot be investigated without
looking at other processes of meaning extension to which it relates in actual
usage. Goossens’ (1990, 1995) work demonstrates how metonymy interacts
with, gives rise to, and is embedded in metaphor. Furthermore, it appears that
generalization and specification are also often related to metonymy — a
conclusion which should not surprise us, given the interaction between gen-
eral and specific inherent in metonymy.
Notes
1. This paper benefited from comments on an earlier version by L. Goossens.

2. In this actual example, taken from a corpus of English from the 1960s, the ‘CD-
interpretation’ is of course not open; present-day English however continues to refer to
‘putting on’ a CD; further illustrations of the generalization which has occurred are
provided by ‘put on a cassette’ and ‘put on a song.’
References
Croft, William
Geens, D., L.K. Engels, W. Martin
1975 Leuven Drama Corpus and Frequency List. PAL. KU Leuven.
Goossens, Louis
1990 Metaphtonymy: the interaction of metaphor and metonymy in figurative
expressions for linguistic action. Cognitive Linguistics 1: 323–340.
1995 From three respectable horses’ mouths: metonymy and conventionalization
in a diachronically differentiated data base. In L. Goossens et al. (eds.), By
Word of Mouth: Metaphor, Metonymy and Linguistic Action in a Cognitive
Perspective. Amsterdam and Philadelphia: Benjamins, 175–204.
Johansson, Stig, Geoffrey N. Leech, Hellen Goodluck
1978 Lancaster-Oslo/Bergen Corpus of British English, for use with digital
computers.
Lakoff, George
1994 The limits of continuity: discreteness in cognitive semantics. In C. Fuchs,
B. Victorri (eds.), Continuity in Linguistic Semantics. Amsterdam and
Pauwels, Paul
1995 Levels of metaphorization: the case of put. In L. Goossens et al. (eds.), By
Word of Mouth: Metaphor, Metonymy and Linguistic Action in a Cognitive
Perspective. Amsterdam and Philadelphia: Benjamins, 125–158.
Pauwels, Paul, Anne-Marie Simon-Vandenbergen
1995 Body parts in linguistic action: underlying schemata and value judgements.
In L. Goossens et al. (eds.), By Word of Mouth: Metaphor, Metonymy and
Linguistic Action in a Cognitive Perspective. Amsterdam and Philadelphia:
Benjamins, 35–70.
Taylor, John R.
Clarendon Press.
Conversion as a Conceptual Metonymy
of Event Schemata
René Dirven
Gerhard Mercator University of Duisburg
1. Definition
As a conceptual process of locating a reference point, metonymy belongs to

the wider set of strategies of finding a point in the common reference space
between a speaker and a listener that can serve as a bridge or link to the
intended referent (cf. Langacker 1991: 170ff). One possible strategy used for
this purpose in English is the genitive, as in I’ll take Dad’s car, whereby Dad
is the link known to the son and his girlfriend for identifying the car. In a
restaurant scene where the waiter speaks about one of the customers in
metonymical terms as the hamburger, we find the same communicative need
of identification as with Dad’s car. Only now the context does not provide an
identity marker, since the customer’s name is unknown and his place in the
restaurant may not be a salient one. An alternative would be to give tables and
customers numbers. But even then the waiter would not necessarily say (1a),
but might equally well say (1b).
(1) a. The customer at table 13 is complaining about his hamburger.
b. Table 13 is complaining.
If no numbers are available, the waiter can only take the most relevant feature
in the given link between the institutional context of the restaurant and the
reason why people go there, i.e., in order to get food. Since people tend to take
different dishes, these become a salient means of identifying customers. In
fact, in the case of the referencing use of metonymy as in the restaurant scene,
276 René Dirven
the hamburger ordered or eaten by the customer serves as the identifying

description. The only identifying link that the waiter has with the customer is
by means of the food he has ordered.
Notice that the very same process of referencing occurs in other institu-
tional contexts, too. Hospital nurses frequently refer to their patients in terms
of the organ affected or the illness they suffer from. Thus, one may hear things
like (2):
(2) a. The liver is still waiting outside.
b. I’ll now take the by-pass back to his room.
Although this metonymical strategy for the sake of referencing is ubiquitous
— or perhaps because it is so ubiquitous — we are no longer aware that we
apply it constantly. But it is always used in the context of picking out referents
in a somewhat amorphous or anonymous setting. For example, teachers use it
with their students (the striped sweater), spectators with referees (the black
one) or football players (the red one, the bald one), musicians (the violins, the
tuba), pigeons (various shades of gray, or, in some languages, colors), sales-
persons use it with customers in a flower shop, not at the counter but in the
back room (the fuchsia, the geranium woman).
But even if we already have a name for identifying an institution, the
tendency to refer to it by way of metonymy remains very strong especially
with employees referring to their workplace, which is often indicated by the
street or quarter. This is perhaps a sign of distancing and neutralizing emo-
tional ties and commitment signals which might be contained in using the real
name. Thus, Dutch teachers tend to refer in Dutch to their schools as the High
Street (Hoogstraat), Cattle Market (Veemarkt) or the village name so that we
may get metonymies such as:
(3) a. The High Street is now losing ground to the Cattle Market (i.e.,
the one school is losing pupils to the other).
b. The Cattle Market has gone co-educational now.
Whatever the communicative need triggering this referential use of met-
onymical expressions, it is clear from these examples that the syntactic/
semantic frame is determined by the intended referent (humans/institution)
and not by the supporting reference point (inanimate object, place, organ, or
disease, etc.). Hence we cannot say *The liver is still waiting outside, but it’s
nearly its turn, where the pronoun it is supposed to be coreferential with the
antecedent noun phrase the liver.
2. Application to conversion
Metonymy as illustrated in Section 1 operates at the discourse level. Given its

ubiquity, it can be expected to operate at all levels of linguistic structure. The
present paper will concentrate on the morphological level, more particularly
on conversion.
In its most productive form, conversion gives rise to new verbs. The
process is usually analyzed as a categorical change of a lexical item from the
status of a noun or an adjective to that of another word class. The adjective clean,
already used in its present meaning in Old English, was converted into the verb
to clean in the 15th century (OED). But situating the conversion process at the
word level is relatively naive and simplistic, as if such processes occur in
abstracto. It is now beginning to become clear that conversion is not merely a
process at the word level, but rather at the predicate-argument level, which we
shall henceforth call the ‘nucleus’ level. Thus to clean is a transitive verb
implying a case frame containing an agent, a patient, and possibly an instrument,
a manner and a result. This linguistic configuration can in fact be seen as an
iconic reflection of a conceptual configuration, in which an agent as the energy
source transmits energy to an object which is affected by the energy. This is
obviously the case in X makes the table clean, where the adjective clean denotes
the resultant state of the energy transmitted by the agent. Since the resultant state
is the most salient element in this whole action schema, it comes to stand for the
whole event in X cleaned the table. At the same time, the entire action schema
remains implicitly present in the resultant state, which now metonymically
denotes the action schema as such. This particular process of conversion may
also offer an explanation for the fact that all conversions from adjective status
to verb status are of this ‘resultant state’ type.
We may wonder whether this type of metonymical processing is still
relatable to the ‘referencing’ need of the HAMBURGER FOR CUSTOMER me-
tonymy discussed before. What, in other words, is similar and what is different
in the two metonymical sentences in (4)?
(4) a. The hamburger left without paying.
b. The waitress cleaned the table.
The reference-supporting use of the hamburger for the person who ordered
and ate a hamburger is based on a similar schema as in someone made the
table clean. But in the HAMBURGER metonymy the patient becomes the identity
278 René Dirven
feature for the agent, which can only be realized at the clause or sentence
level. Statistically, this referencing function seems to be intimately related to
the agent function, or at least to the subject function. Indeed, the HAMBURGER
metonymy is far less likely in direct object or indirect object position.
(5) a. ? I asked the hamburger to pay at the counter.
b. ? I showed the hamburger the way to the station.
This suggests that even the referencing type of metonymy cannot occur in
every possible context, but that it is highly constrained to one prototypical
configuration, i.e., one in which the agent/subject function, at least for hu-
mans, is salient.
Conversion differs from this agent/subject or referencing type of me-
tonymy in that it does not apply at the sentence or clause level, but rather at the
predicate-argument or nucleus level. Any of the participants in an action
schema or in any other event schema, except the agent/subject, can become
the bearer of the saliency feature in the appropriate configuration and then
serve as input for the conversion process. Thus in reporting a soccer game, we
can describe certain ways of handling the ball with the head as He headed the
ball into the goal, whereby the exact timing of the ball-to-head contact and the
exact force and direction given to the ball by the head are of paramount
importance. In the total action schema The player sends the ball into the goal
with his head the instrument participant with the head is so salient that it can
stand for the action itself. This does not mean, however, that the instrument
slot in the action schema disappears. Although instrument verbs like to kick, to
head, to hammer, etc. do not normally require an instrument, they can take one
— if it is sufficiently specified:
(6) a. He kicked the ball with his left foot.
b. He headed the ball with the back of his head.
c. He hammered the nail into the wall with his shoe.
This means that the conversion process does not necessarily delete the partici-
pant role which is denoted by the converted lexical item in the total action
schema. Since the type of metonymy that we find in the conversion process
takes place at the nucleus level, it cannot become a means of referencing as
metonymy at the clause level does; indeed, a nucleus is by definition not
grounded in communicative space.
Whatever differences there may be between these two processes of
metonymy (reference metonymy, event-schema metonymy), they do have one

thing in common, i.e., they single out a salient participant to become either the
identity marker for an unknown or unnamed referent (reference metonymy) or
the main designation for the event itself (event-schema metonymy).
3. Types of event schemata serving as input for conversion
It is highly interesting to note that, in traditional grammar, very few classes of

verb conversion were distinguished; in fact no more than four. Thus
Zandvoort gives the following “paraphrasing” relationship for the “meanings
expressed in regard to the nouns from which they are converted” (Zandvoort
1961: 267):
(7) a. to make use of or to treat with (to motor, to elbow, to gun)
b. to act as, to behave like (to ape, to dog, to mother)
c. (sometimes in the case of animals) to bring forth (to kitten, to
goat, to lamb)
d. to remove (to dust, to skin, to weed)
Marchand (1969: 367ff) also only distinguishes four major syntactic-semantic
classes of conversion, although he proposes many subclasses for each of his
main classes:
(8) a. Predicate–Subject Complement (to bully: to be a bully)
b. Predicate–Object Complement (to knight: to make someone a
knight)
c. Predicate–Adverbial Complement (to anger: to provoke to an-
ger)
d. Predicate–Object (to calve: to bring forth a calf)
Whereas Zandvoort’s analysis is still situated at the word level, Marchand’s
approach constitutes a major step forward in that he is implicitly performing
his analysis at the nucleus level, which — because of the lack of case-
grammatical categories — he labels with sentence-level categories like Sub-
ject, Object, Object Complement, Adverbial, etc.
It is only thanks to the insights of case grammar that it was possible for
researchers to handle the semantic relationships involved with converted
verbs in terms of semantic-conceptual roles such as patient, instrument, man-
280 René Dirven
ner, goal or source and essive roles (e.g., Kastovsky 1974, 1977; Leitner 1977;
Clark and Clark 1979). Elaborating on this previous research, I proposed five
classes of converted verbs based on underlying semantic roles holding be-
tween a putative predicate and a semantic role (Dirven 1986: 321ff):
(9) a. object verbs (to fish, to crew, to anger)
b. instrument verbs (to harpoon, to head, to veto)
c. manner verbs (to queue, to balloon, to spoon)
d. locative verbs (to bottle, to shelve, to record)
e. essive verbs (to author, to nurse, to knight)
As was already suggested in Section 2, these five categories can be seen as
labels which summarize a full-fledged event schema in which several partici-
pants operate on each other.
There are three canonical event schemata underlying conversions:
(i) the action schema, in which an agent acts upon a patient, often using an
instrument and involving a certain way or manner of doing this as in to
fish;
(ii) the location or motion schema, in which the agent may perform an action
aiming at a localized effect as in to bottle;
(iii) the essive schema, or the schema for ‘beingness,’ in which the status of
class membership or an attribute is assigned to an entity.
These three event schemata will now be analyzed in more detail, whereby the
semantic role selected is seen as the element responsible for the metonymy
underlying the event schema as a whole.
4. Metonymic focus within the action schema
The action schema conceptually synthesizes the flow of energy from an agent
to a patient via an instrument in a certain manner. In the instantiation of
someone trying very patiently to catch fish with a fishhook, each of these
semantic roles, except for the agent itself, can be focused upon, assume a
predicate’s relational function of linking the semantic roles, and in the process
highlight one aspect of the action schema, but always imply the whole of it.
Thus the patient itself, the instrument, and in some contexts, the manner, can
come to stand for the whole action schema as in (10):
(10) a. He was fishing (salmon).

b. He was luring fish.
c. He was fishing pearls.
In (10a) the patient fish is the metonymic focus of the whole action schema
and, as a converted verb, it normally does not require the patient to be further
specified, although this is always possible by way of specifying a particular
type of fish. It is also noteworthy that only the basic-level term fish can be the
input for a conversion, and that this is not possible with subordinate-level
terms such as salmon (*He was salmoning). Obviously, this has an experien-
tial motivation. When fishing, people do not always know what the catch
might be and hence this specific level could not be focused upon. In the few
cases where the type of fish to be caught is predictable as with salmon or carp,
one could imagine that in the jargon of some fishermen these conversions
could occur.
The instruments used to catch fish such as a lure, a hook, a harpoon, a net,
etc. are purposely and intentionally employed and the nouns describing them
are readily converted into verbs.
The way or manner in which fish are caught may be transferred to other
domains closely related to the same domain of trying to catch things in the
water (of the sea). The example in (10c), to fish pearls, is paraphrased in CED
as ‘to take pearls from the bottom of the sea like one takes or catches fish.’
Basically, no fishing as such is involved here so that only the manner or
circumstances in which fish are caught in the sea is of real relevance: just as
one must wait patiently for fish to bite and thus get caught and removed from
the water, one must search for pearls at the bottom of the sea patiently and then
remove the ‘catch’ from the water.
A theoretically important question is where to draw the dividing line
between metonymic uses of object, instrument, or manner verbs such as to
fish, to harpoon, to hook, and their metaphorical extensions. It is interesting to
note that in the expression to fish pearls, there is clear contiguity with the same
domain as in to fish herrings: in both cases we remain in the domain of
catching things in the sea. Once we leave this domain and apply the search to
people, their behavior, or to some abstract domain, we are in a totally different
domain. It is remarkable to note the different syntax (addition of a preposition)
that correlates with a usage in a more abstract domain.
(11) to fish for information/for a rich man
282 René Dirven
In these metaphorical uses of to fish the preposition of purpose for precisely

denotes the as yet unrealized accomplishment, i.e., the absent target or at least
the not yet spotted target, which is also found in to look for, to grope for, to
search for, to wait for, etc. And after one has fished long enough for a rich
man, one may hook him. These metaphors are very far away from the met-
onymic expression to fish pearls, in which the transitive verb plus a direct
object construction iconically reflects the closer proximity between the
agent’s action and the patient affected by the action.
5. Metonymic focus within the motion schema
In its simplest form, the motion schema comprises a moving patient and one or
more elements of the motion’s trajectory, i.e., source, path and goal. Given the
fact that in the human sphere of interest, goals are more salient than sources
(Ikegami’s 1987 goal-over-source principle), it need not surprise us that
locative verbs resulting from conversion processes are predominantly goal
verbs.
It is, moreover, not astonishing that the natural topography of the goal
also determines the type of arrival. The type of physical end-point of the
motion determines the specific meanings of each converted verb as
contextualized in the following sentences:
(12) a. The tide had gone out, leaving the boat stranded on the rocks.
b. Before going home, the fisherman beached his boat.
c. On the cruise we’ll first land in Casa Blanca.
d. The plane was forced to land in Cairo.
e. The submarine surfaced again.
f. The plane was grounded there for 24 hours by the hijackers.
The metaphoric expression leave somebody or something stranded, as in
(12a), historically relates to the now obsolete noun strand, meaning ‘the land
bordering a sea, lake, or river’ — i.e., the coastline or margin, be it rocks, sand,
etc. (still found in the name of a street in London, i.e., The Strand). Thus, for
example, a boat which has run aground during a storm has been left helpless
and unable to move. Since no successful arrival was achieved, the boat got
‘stranded.’ The type of arrival denoted by to beach is different, as (12b)
shows. The Dictionary of Contemporary English defines the meaning of the
verb as ‘to pull a boat onto the shore away from the water.’ Thus, if we leave a
boat on a beach, by running or hauling it there, we have ‘beached’ the boat. It
is only when we reach the land, as opposed to the sea or the air, that we have
‘landed’ as in (12c, d). In each of these instances, the physical strip or area of
land which constitutes the goal stands for the motion as a whole. Likewise to
surface in (12e) picks out the top of the water as its goal and to ground in (12f)
takes the earth seen in opposition to the air as its goal.
These instances of conversion may also show that in the conversion process
there is never a purely physical goal, but that the goal also stands for the
circumstances surrounding the scene. This also applies to their metaphorical
senses. Thus to strand evokes and reflects the shipwreck scene or a situation in
which a child is left unattended in a strange place. To beach brings to mind a
spontaneous situation in which, for example, the people jump out of the boat as
it reaches the beach, haul it up onto the sand anywhere away from the water and
just leave it there. To land suggests a purposeful arrival, either from the sea or
the air, as well as in the phrase ‘landing a job.’ To surface calls up an image of
a swimming or underwater-motion scene, as well as, metaphorically, the
coming to light of scandals or lies. To ground evokes a scene of forced contact,
which can also be noted in the metaphorical usage of ‘grounding a naughty
child.’ In other words, the use of metonymy is not restricted to the aspect of
motion as such, but it also encompasses the wider experiential scene of which
it is one specific facet. This wider experiential scene is in fact the real motivation
for locative conversion verbs — and probably for all conversions.
For locative verbs such as to bottle, to box, to can, etc., it is the scene of
food preservation that provides the context. Other more ‘abstract’ goals may
be shelter as in to house, to jail, or to harbor; display as in to bench or to field;
or else giving things a specific shape or form as in to bundle, to pile, or to slice.
Even purely abstract mental motion leading to a new artifact involves the
motion schema as in to book, to map or to register.
6. Metonymic focus within the essive schema
In the essive schema, a patient is assigned class membership status or an

attribute. In converted essive verbs such as to author, to volunteer, to nurse,
etc., we only find an implied remnant of class membership. In its prototypical
use, the noun nurse has the more general meaning of ‘a person, usually a
284 René Dirven
woman, who tends the sick, injured, or infirm’ or the more specialized
meaning of ‘a woman employed to breast-feed another woman’s child’
(CED). The six meanings of the converted verb to nurse can be neatly
associated with either of these two meanings:
(13) a. Mary nursed the sick soldiers (= to tend an injured person)
b. Mary nursed her father’s ailment (= to tend the wounds)
c. Cathy nursed the crying child in her arms (= to clasp)
d. Cathy nursed the baby five mornings per week (= to care for)
e. Gilly nursed the baby since the baby’s mother could not breast-
feed her child (= to breast-feed)
f. The baby nursed at the woman’s breast (= to suckle)
To put it in a somewhat simplistic way, we could say that one need not be a
nurse in order to do all the ‘nursing acts’ described in (13). Indeed, the verb’s
meaning is far more general than that of the corresponding noun. The two
domains covered by to nurse are ‘tending sick or injured people’ and ‘taking
care of young children, either as a care-giver or else in the breast-feeding role.’
The far richer meaning of the verb to nurse as compared with the noun nurse
also reveals another aspect of the metonymic nature of the conversion process.
Whereas the noun can only denote more or less fixed categories or classes
and their members, the verb is far more flexible and somehow breaks the
constraints imposed by the boundaries of the noun’s class membership mean-
ings. Thus the conceptual entity ‘a nurse’ can only, in her profession, tend sick
or injured people or breast-feed another woman’s child. But the looser con-
ceptual entity of ‘a person who nurses’ can do both: tend sick or injured people
or breast-feed young children. When nursing young children, she can be
imagined in a specific act such as holding a child in her arms, in her overall
assignment, or in her ‘technical’ role of breast-feeding a child. In fact, it is
again the various overall scenes that may be associated with her work as a
nurse that are conceptualized in the conversion process.
We can also combine the essive schema with the action schema: for
example, an agent may turn a patient into a member of a class. Thus, in to
knight, the action schema and the essive schema combine to express the
institutionalized act of elevating a person to knighthood. This is, however, a
very unproductive pattern.
The third type of essive relationship is that of assigning an attribute to a
patient. This relationship is only exceptionally expressed by nouns, the typical
word class for this purpose being adjectives. Here again, the essive schema
combines with the action schema as in to clean the table, which is a highly
productive pattern in English.
7. Why only these (five) classes of conversion?
As pointed out already, English descriptive linguistics traditionally posits a

small number of main classes of converted verbs. But a question that has — as
far as I know — never been asked is: why are there so few classes of converted
verbs and why do we have the ones we have?
Of the 16 semantic roles listed below, the nine italicized ones occur in
converted verbs:
(14) Event schemata and associated semantic roles
a. Action schema: Agent, Patient, Instrument, Manner
b. Experiencing schema: Experiencer, Stimulus
c. Possession schema: Possessor, Possession
d. Transfer schema: Recipient, Beneficiary
e. Location/Motion schema: Place, Source, Path, Goal
f. Essive schema: Class Membership, Attribute
(The ‘human roles’ in (b, c, d) could be subsumed under the
label dative role.)
The overall pattern may become clear as a result of this survey: the conversion
process is mainly applied to non-human participants such as patient (as
object), instrument, manner, locative and attribute. Although, in the class of
essives, the class-membership group only contains human entities such as to
author, to nurse, or to volunteer, they are all assigned to patient roles. In other
words, as long as human referents are not treated differently from non-human
referents they can become the input for conversion processes. But if human
referents are treated in their typically human agent and dative roles, they are
unlikely to become involved in a conversion process. These facts may find an
explanation in the cognitive principle of anthropocentrism. Since human
beings are already the focus of attention in most linguistic structures, they
cannot be focused upon again in the conversion process, at least not in the
agent or dative roles, which are prototypically human roles.
This generalization would explain why an expression such as to police a
286 René Dirven
district can never mean that people are turned into ‘police officers of a district’
(essive relation), but rather it can only involve the patient role, i.e., that you put
a sufficient number of policemen in a district so that it may become safe again.
Here the police are not understood as agents, but as patients, or instruments that
in an atmosphere of street violence may restore a neighborhood’s feeling of
safety.
The deeper explanation for the fact that typically human roles such as
agents or datives do not enter into conversion processes may have to do with the
clash between two cognitive principles, i.e., the principle of anthropocentrism
and the principle of metonymic focusing. Conceptually, it is indeed very
difficult to see how a human being could be part of some larger configuration
to which it could stand in a metonymic relationship. Whereas the opposite
relationships abound, i.e., a book being referred to by means of its author or a
customer being referred to by the food he ate, a metonymic relationship of
humans to some higher conceptual category is hardly imaginable and, hence,
agents and dative roles do not easily enter into conversion processes.
References
Clark, Eve V., Herbert H. Clark

CED
1979 Collins Dictionary of the English Language. London, Glasgow: Collins.
DCE
1995 Dictionary of Contemporary English. Harlow: Longman.
Dirven, René
1986 Can case grammar cope with conversion. Annales Universitatis Scientiarum
Budapestinensis Sectis Linguisica 17: 315–335.
Ikegami, Yoshihiko
1987 ‘Source’ vs. ‘goal’: a case of linguistic dissymmetry. In R. Dirven, G.
Radden (eds.), Concepts of Case. Tübingen: Narr, 122–146.
Kastovsky, Dieter
1974 Word-formation, case grammar, and denominal adjectives. Anglia 92: 1–
54.
1977 Problems of word-formation. In Ch. Gutknecht (ed.), Grundbegriffe und
Hauptströmungen der Linguistik. Hamburg: Hoffmann und Campe, 301–
335.
Leitner, Gerhard
1977 Zur Vorhersagbarkeit von Derivation: Teil von Nomina als Basen. In H.E.
Brekle, D. Kastovsky (eds.), Perspektiven der Wortbildungsforschung.
Bonn: Bouvien-Verlag Herbert Grundmann, 140–154.
Marchand, Hans
1969 The Categories and Types of Present-Day English Word Formation: A
Synchronic-Diachronic Approach. München: Beck.
OED
1994 The Oxford English Dictionary. Second Edition on Compact Disc for the
IBM PC. Oxford: Oxford University Press.
Zandvoort, Rainard W.
1961 A Handbook of English Grammar. Groningen: Wolters.
Opposition as a Metonymic Principle
Christian Voßhagen
1. Introduction
This paper discusses the notion that forms of language use in which something
is uttered to convey its opposite are metonymic. In this view, a conceptual
entity can be used to provide mental access to its opposite, which is closely
associated with it within a conceptual structure.
The basis of this discussion is the observation that opposites generally
form part of one conceptual domain, and that mappings which occur within
such domains involving opposites are influenced by factors that also influence
other metonymic mappings. Such factors include conceptual contiguity, sa-
lience, social factors, idealized structuring of conceptual domains, and the
highlighting function of metonymies for expressive purposes. These factors
will be discussed as motivations of metonymic mappings of one concept onto
its opposite in figurative language including euphemism, socially motivated
reversals, irony, and expressive use of negatives to convey positive evalua-
tions.
2. Opposition of meaning and metonymy
Speakers frequently use, and make sense of, expressions which involve saying
one thing and meaning the opposite. These expressions often contain adjec-
tives and seem to be used in two general ways. First, something positive can
stand for something negative, as in the following example:
290 Christian Voßhagen
(1) Y has been cheated by her friend X and says: “X is a fine friend.”
Here, the adjective fine conveys a negative, ironic meaning. Second, some-
thing negative can stand for something positive. A well-known example of
this is the adjective bad in its use in slang:
(2) bad
a. ‘eminently suitable or appropriate; excellent; wonderful [...]’
(Wentworth and Flexner 1967);
b. ‘a simple reversal of the white standard, the very best’ (Major
1971);
c. ‘good. Originally from the terminology of the poorest black
Americans, either as simple irony or based on the assumption
that what is bad in the eyes of the white establishment is good
for them [...]’ (Thorne 1990).
These processes lead to the following, related phenomena. First, expressions
can be ambiguous between a literal and its opposite meaning, as illustrated in
the following description given by an African American speaker (Folb 1980:
205):
(3) “Like ‘funky’ — it kin be really good or really bad. All depends.
You kin tell by d’ way dude say it — watch his face, move his body
certain way. Like, ‘He’s funky man!’ Move way from d’ dude —
he stink! ‘Hey, dude’s funky!’ Nigger be smiling, make d’ fist —
dude’s okay!”
Second, expressions can acquire a meaning that is the opposite of their
original meaning. The adjective terrific with its now commonly accepted
positive meaning is an example of this. This paper regards instances of
figurative speech of this kind as manifestations of a widespread metonymy: A
CONCEPT STANDS FOR ITS OPPOSITE.
In this view, using an expression in order to convey its opposite falls under
definitions of metonymy such as the one formulated by Lakoff (1987: 84):
– There is a “target” concept A to be understood for some purpose in some
context.
– There is a conceptual structure containing both A and another concept B.
– B is either part of A or closely associated with it in that conceptual
structure. Typically, a choice of B will uniquely determine A, within that
conceptual structure.
– Compared to A, B is either easier to understand, easier to remember, easier

to recognize, or more immediately useful for the given purpose in the given
context [….]
In this paper I assume that metonymy is conceptually grounded (cf. Lakoff
1987; Radden and Kövecses, this volume), and that linguistic metonymies
have to be regarded as manifestations of general conceptual metonymies. I
argue that (i) opposites belong to one conceptual domain (cf. Cruse 1986:
198ff) and that (ii) the relation between opposites is one of close mental
contiguity.
In linguistic semantics, opposition of meaning is often treated as a purely
lexical relation. Although, in contrast to this view, opposition is seen as a
conceptual phenomenon here, some insights from lexical semantics will be
taken up in this paper when they are compatible with my view and help to
clarify the arguments presented. The following section distinguishes between
two types of opposition: antonymy and complementarity. In contrast to many
lexical approaches, however, opposition is considered to be a conceptual
relation holding within one domain. Sections 4 to 7 discuss different aspects of
this conceptual view, addressing the general questions of how conceptual
contiguity between opposites is reflected in language and what conditions
give rise to the metonymic use of an opposite term for a more ‘literal’ term.
Section 4 is concerned with the concept of markedness. It discusses salience
and evaluation as perceptual and social factors determining markedness and
metonymic use of expressions. Section 5 briefly discusses how social and
pragmatic factors may influence the use of one term to stand for its opposite.
Section 6 proposes an idealized instance of opposition which serves as a
cognitive model that may underlie some forms of opposition of meaning and
influence metonymic mappings. Section 7 discusses the influence of emo-
tional factors on the use of opposition-metonymies for expressive purposes.
3. Semantic opposition in lexical and conceptual structure
Opposition is without doubt one of the most basic semantic relations (cf. Cruse
1986: 198; Deese 1970: 102). A vast number of lexical items are related by
opposition. In contrast to other relations distinguished in semantics, opposition
is readily comprehended by speakers and is among the first semantic relations
children learn to distinguish accurately (Landis, Herrmann and Chaffin 1987).
It is also a very basic associative relation. Word-association experiments,

as conducted by Deese (1962) and others, have consistently shown that words
which have an opposite are frequently associated with the opposite term (cf.
also Esper 1973; Aitchison 1987). In a similar way, substitution errors involv-
ing the utterance of another word than the intended one, committed both by
impaired and non-impaired speakers, often involve not only relations such as
synonymy or hyponymy, but frequently opposition of meaning (Fromkin
1973: 262; Garman 1990: 161). All these findings support the claim made
above that the relation between opposites is one of close mental contiguity.
However, relations of opposition such as antonymy are usually treated as
purely lexical relations, and the association of a word with its opposite is often
seen as resulting only from linguistic factors such as co-occurrence in spoken
and written texts. Such arguments do not explain why antonymic pairs should
ever co-occur so frequently in texts. Psycholinguistic research suggests that
the relation of opposition is based on conceptual rather than lexical association
(cf. Murphy and Andrews 1993). For example, speakers tend to give different
antonyms for a given lexeme depending on whether it is presented to them in
isolation or in different contexts. This finding is relevant in that it suggests that
knowledge about the relation is independent of the way opposites are repre-
sented linguistically. The conceptual view of metonymy adopted in this paper
implies that it is not the metonymic expression that provides access to its
opposite, but the concept it expresses that provides access to another concept.
As an illustration, consider the following example. Here, a negative
expression conveys a clearly positive meaning, but none of the senses can
count as the lexical opposite of wicked; still, the figurative use involves a clear
opposition between positive and negative evaluation:
(4) wicked
‘excellent in any way; potent; strong; capable’ (Wentworth and
Flexner 1967)
Among the classes of lexical items related by opposition, adjectives constitute
an important group. The examples given so far show that they are also used
figuratively.
A distinction between different kinds of adjectival opposition which is
also important for the present discussion is made in lexical semantics between
‘antonymy’ and ‘complementarity’ (cf. Lyons 1977; Lehrer 1990; Cruse
1986).1 Antonymy involves pairs such as big-small, good-bad, happy-sad,
terrible-wonderful and pertains to concepts that are gradable, denote different

degrees of intensity, and have a kind of ‘neutral area’ between them. They
may be said to constitute opposing portions of a scale. Figure 1 illustrates this
situation for the antonyms good and bad and the ‘neutral area’ (the simple
quotation marks indicate that the expressions represent concepts, not lexical
items):
0
‘bad’ ‘neutral’ ‘good’
Figure 1. Concepts of evaluative antonyms
The labels ‘neutral’ and ‘0’ imply that there is a portion of the scale where
neither of the concepts applies; thus, it is possible to say something is neither
good nor bad. For the same reason, negation of one pole does not necessarily
imply assertion of the other — it is possible to say something is not good, but
not bad either. The arrows at the ends of the scale indicate that both concepts
are graded and can be intensified; thus, it is possible to say something is very
good or very bad.
In the analysis that follows, I would like to suggest a differentiation
between two groups of antonyms: ‘evaluative antonyms,’ exemplified by
good-bad, and antonyms belonging to the domain of ‘physical measurement,’
exemplified by big-small. The difference between these groups can be seen in
comparing Figures 1 and 2:
0 ‘small’ ‘neutral’ ‘big’

Figure 2. Concepts denoted by physical-measurement adjectives
With evaluative antonyms, a lack of the quality of evaluation common to both

antonyms — represented by ‘0’ in the figures — coincides with the neutral,
intermediate value on a scale that leads from a high degree of negative
evaluation to a high degree of positive evaluation. The scale for physical
measurement antonyms, on the other hand, is unidirectional, a total lack of the
quality (in this case, size) marks one end of the scale, a large amount of it
marks the other.
Complementarity is exemplified in pairs such as married-unmarried or
dead-alive and involves concepts which are not gradable, do not express
degrees of intensity, and have no intermediary terms between them. Anything
which belongs to the conceptual domain of complementaries is either the one

or the other (cf. Cruse 1986). This situation is represented in Figure 3 for dead
and alive:
‘dead’ ‘alive’
Figure 3. Complementary concepts
From a logical perspective, there are only two values possible in the concep-
tual domain of, say, ‘organisms’ — not alive implies ‘dead,’ not dead implies
‘alive.’ This may be called a ‘two-valued organization’ of the domain.
4. Asymmetry of opposite pairs: salience, markedness, and

evaluation
The two groups of evaluative antonyms and physical measurement antonyms

have one property in common: they are asymmetrical. In the case of physical
measurement antonyms, this property has already been addressed: one end of
the scale constitutes an absolute zero point, whereas the other end corresponds
to a large amount of the denoted quality. This asymmetry has an interesting
consequence. Physical measurement adjectives denoting the greater amount
of the quality involved, such as big, are ‘semantically unmarked’ — they can
be used to denote the quality common to the pair. In asking How big is it?, a
speaker does not necessarily imply any notion about the size of the object in
question, whereas in asking How small is it?, there is a strong implication that
the object is relatively small. This preference for one member of the pair is
conditioned by perceptual salience (cf. Givón 1978; Cruse 1986); the more
salient member of the pair affords better mental access to the conceptual
domain than the less salient one. This situation is metonymic; an easy-to-
perceive and well-understood aspect of a conceptual domain — the more
salient one — stands for the domain as a whole (Lakoff 1987), or, more
specifically, THE POSITIVE END OF A SCALE STANDS FOR THE WHOLE SCALE
(Radden and Kövecses, this volume).
Thus, the phenomenon of markedness shows that opposite concepts are
closely associated within a conceptual domain, and that one of them can be
used as a reference point affording mental access to a target (Langacker 1993).
The target is, in this case, the conceptual domain that includes both concepts.
The systematic occurrence of this phenomenon in language supports the claim
made in the previous section that contiguity between opposites is not only
conditioned by lexical association. It can be argued that just as the adjective
denoting the larger amount of the quality is linguistically unmarked, the
corresponding salient concept is perceptually unmarked with regard to its
opposite, making it both “simpler” (Cruse 1986: 248) and “more basic”
(Lakoff 1987: 60).
Evaluative concepts show a different kind of asymmetry. Whereas the
markedness of physical measurement antonyms is perceptually motivated,
evaluative antonyms involve what may be said to be a socially motivated
markedness. The adjectives that express a socially desirable, positive concept
may be called unmarked, whereas those that express a negative, socially
undesirable concept may be regarded as marked. As an illustration of social
markedness, consider the following euphemism:
(5) pretty ear
‘an ear deformed from being hit repeatedly; a cauliflower ear’
(Wentworth and Flexner 1967)
The adjective pretty is used to denote something that may well be described in
terms of its antonym, ugly. Mention of the negative, socially marked term is
avoided, and its opposite is used instead to stand for the undesirable quality.
The motivating factor is not perceptual salience, but is closely connected with
social-pragmatic factors such as politeness.
In the following section, other social factors will be shown to motivate
metonymic mappings onto opposite concepts.
5. Opposition of attitudes: social factors and irony
‘Social markedness’ as discussed above sometimes coincides with formal or

morphological markedness. Consider the pair sane-insane: the negative,
marked member is also formally marked. But as can be seen in the following
example from African American slang, insane can nevertheless be used to
stand for the positive, unmarked concept:
(6) insane
‘positive, healthy state of mind’ (Major 1971)
Why do speakers avoid a positive, unmarked expression like sane and use a
negative, marked one to access the positive concept? A plausible explanation
results if conflicting social or cultural attitudes are considered as a motivating
factor. For example, the slang use of bad as illustrated in example (1) gains its
metonymic sense from what might be called a negative attitude. Dictionary
definitions speak of a ‘reversal’ and, more specifically, of the assumption that
what is bad for one social or ethnic group must be good for the other.
When social attitudes differ, linguistic expressions may take on different,
but related meanings that correspond to these attitudes. When the attitudes are
opposed, linguistic expressions may take on opposite meanings. Thus, the
adjective bad had a conventionally accepted negative, marked meaning for
white mainstream culture, and a positive one for many African American
speakers who opposed the attitudes of the dominant group. In this sense, the
metonymic use of expressions like bad, wicked, etc., is motivated by social
opposition.
A similar motivation seems to underlie other cases where the opposition
is not necessarily between ethnic groups, but between subculture and main-
stream culture, as in the following examples (all from Wentworth and Flexner
1967):
(7) nasty
‘excellent; “wicked” [...]’
(8) mean
‘[...] psychologically exciting, satisfying, and exhaustive; Mean —
the best; the greatest’
(9) evil
‘wonderful; specifically, thrilling, very satisfying [...] Implies that,
like sex, that which is really thrilling and satisfying is considered
sinful by puritanical people.’
As will be shown below, there are many expressions, especially in American
slang, that reverse attitudes of mainstream culture. Along with other factors,
this aspect of social evaluation seems to be a basic motivating factor for this
kind of metonymy.
There is another phenomenon that seems to be based on opposing atti-

tudes and which is commonly defined as ‘saying something and meaning the
opposite’: irony. Consider the following examples:
(10) big deal
‘anything important, satisfying, exciting, interesting, lavish [...]
Sarcastically, anything or anyone believed to be unimportant, unin-
teresting, or unimpressive [...]’ (Wentworth and Flexner 1967)
(11) big idea
‘an unwelcome suggestion, proposal, or action’ (Wentworth and
Flexner 1967)
(12) hot shit
‘wonderful; attractive; handsome; charming [...] Always used in the
negative [...] Thus, through ironical use, the term has come to
[mean] conceit ’ (Wentworth and Flexner 1967)
(13) fat
‘[...] Poor, slight , slim; usu. in “a fat chance” = little or no chance
[...]’ (Major 1971)
These expressions have acquired an additional, negative sense through ironi-
cal use. In the case of big deal and fat chance, the negative is even the
predominant, or more usual, sense of the expression according to 30 native
speakers of English I informally asked. Big, hot and fat thus have a conven-
tional ironic sense, with the positive standing for the negative.2
In many cases, irony involves uttering something positive where a nega-
tive judgment is to be expected or can be inferred. However, it clearly involves
more than just saying the opposite of what is meant, or meaning the opposite
of what is said. As convincingly argued by Sperber and Wilson (1995), irony
rather involves implying, or hinting at, an ‘incongruence’ between some
statement or social norm echoed or mentioned by the speaker and what is
actually the case in the pragmatic context of the utterance. This can be
achieved in very subtle ways and need not involve uttering a positive expres-
sion and letting it stand for a negative concept. For example, if a person steps
on the foot of another person in an elevator and this person politely, but
ironically asks,
(14) Could you step on your own feet?

it is difficult to determine what would be the opposite of this statement. A
possible opposite of this utterance, for example, Could you step on my feet
again?, would itself be ironic and therefore would not state the intended
meaning of the ironic utterance (14). In this case, there seems to be an
opposition between the polite request uttered by the speaker and a less polite
non-figurative utterance or speech act that could be expected in the speech
situation.
It is thus not possible to reduce irony to a form of metonymy. Irony is a
more complex phenomenon. In my opinion, it is better accounted for in
approaches like mention theory or pretense theory, which recognize that it
involves a reference not only to the referents a speaker is talking about, but
also to other persons who might really mean what is stated ironically (cf.
Sperber and Wilson 1995; Clark and Gerrig 1984). According to these theo-
ries, an ironic speaker ‘echoes’ what another person said or might have said, or
‘pretends’ to be a person who said or might say (and mean) it.
Nevertheless, an opposition of positive and negative attitudes is often the
basis of irony, just as in the case of the socially determined oppositions
mentioned above. Instances such as (9), (10) and (11) show that the ironical
meaning can be the exact opposite of the original, positive one, suggesting that
an opposition between the positive statement and the negative target concept is
at least a very typical form of irony. But it has to be emphasized that the crucial
factor here is a social-pragmatic opposition between attitudes, not a semantic
opposition between literal and target concept, which can be considered a ‘by-
product.’ As an illustration of this, consider the following verbal exchange from
Edward Albee’s Who’s Afraid of Virginia Woolf (1965: 32). The protagonists,
the married couple Martha and George, are constantly struggling throughout the
whole play, and in this scene they have just been having an open fight in the
presence of their guests, Honey and Nick. Now George and Nick are sitting
together in the living room, their wives have gone upstairs.
(15) George: MARTHA!
[No answer]
DAMN IT!
[...]
Martha: WHADD’YA WANT?
George [to Nick]: Isn’t that a wonderful sound?
Although it is plausible to interpret George’s last, ironic comment as conveying

something like ‘I think that is an awful sound,’ this interpretation also relies on
the background knowledge that he has been fighting with the person whose
utterance he is commenting upon, and that the negative attitude he communi-
cates with his utterance is not so much towards her tone of voice, but towards
her. Thus, the basic opposition here is between a positive attitude represented
by the literal utterance, and a negative attitude being communicated.
6. Two-valued orientation
The examples given so far show that both negative and positive concepts can
stand for their opposites. In euphemism and irony, positive concepts tend to
stand for their negative opposites; in expressions motivated by opposition of
social or cultural attitudes, negative concepts stand for their positive oppo-
sites. It is remarkable that such non-literal uses of language always seem to
involve evaluative oppositions such as between good and bad rather than
oppositions such as between big and small, which may be called denotative in
contrast. Opposition metonymies only seem to apply to evaluative (connota-
tive), not denotative oppositions.3 Thus, a more specific formulation of the
metonymic principle suggested here could be AN EVALUATIVE CONCEPT MAY
STAND FOR ITS OPPOSITE.
This further specification is necessary because evaluative opposition
often seems to be conceptualized in ‘non-semantic’ terms by speakers, which
sometimes results in an instance of opposition being seen as absolute, not
scalar. Hayakawa (1978: 211) described this phenomenon with the following
words:
In the expression “We must listen to both sides of every question,” there is an
assumption, frequently unexamined, that every question has two sides — and
only two sides. We tend to think in opposites, to feel that what is not good
must be bad and what is not bad must be good. [...] This penchant to divide the
world into two opposing forces — “right” versus “wrong,” “good” versus
“evil” — and to ignore or deny the existence of any middle ground, may be
termed the two-valued orientation.
Hayakawa was mainly thinking of the world of politics, but as the following
examples and observations illustrate, two-valued orientation is a much more
general phenomenon:
(i) Speakers often interpret the negation of one evaluative term as an asser-
tion of its opposite although the intermediary term is equally possible. In
answering the question Is it a good movie?, the reply with No tends to be
understood in the sense of ‘It is a bad movie,’ although it may be neither
good nor bad (cf. Lyons 1977: 278).
(ii) In irony, speakers often make use of the opposite, positive pole in order to
state something negative rather than using intermediary terms.
(iii) In euphemism, the positive opposite of an avoided expression is often
used instead of a neutral term. Consider the following examples from
Italian, French, and English, in which the use of euphemisms is motivated
by religious taboo:
(16) benedetto
‘maledetto’ (Kainz 1965 I: 255)
(17) sacred animal
‘cursed beast’ (ibid.)4
(18) a holy terror
‘(sl.) [...] mischievous, embarrassing child’ (Hornby 1974)
(iv) An expression that literally denotes something salient can be metaphori-
cally used to convey both something very negative and something very
positive, but nothing in between, which leads to an abrupt transition from
positive to negative in the dictionary entries:
(19) smash
‘A total failure [...] a popular success; [...] a hit [...]’ (Wentworth
and Flexner 1967)
(20) tough
‘great, wonderful, difficult, terrible’ (Major 1971)
(v) Two-valuedness can be seen in a very interesting way in the following
slang expression:
(21) laugh at the other side of one’s face
‘to cry; to change one’s mood from happy to sad [...]’ (Wentworth
and Flexner 1967)
Laugh stands for ‘cry,’ positive and negative emotions are treated as two sides
of one thing, i.e., the negative side is the other side of the positive rather than
detached or separated from it.
All these examples point to a general human tendency to think of evalua-
tive concepts in terms of opposites. This view is also reflected in the expres-
sions cited by Hayakawa and in folk notions such as the two sides of a story,
there are two sides to everything, there is a thin line between love and hate,5
or in the common metaphor to think in black and white rather than in shades
of gray.
Two-valued orientation in domains of evaluation seems to be a basic
conceptual phenomenon, an idealized kind of opposition. There is thus a
discrepancy between semantic and conceptual opposition. Semantically, an
opposition between gradable terms involves a whole scale, including a neu-
tral, middle ground; on the conceptual level, however, it may only involve the
two opposite terms. This state of affairs is represented in Figure 4:
‘good’ ‘bad’
Figure 4. Idealized opposition of evaluative concepts
This figure illustrates that, under certain conditions, the relation between
evaluative scalar concepts may be similar to that between complementary
ones (cf. Figure 3). Just as either of the complementary concepts ‘dead’ or
‘alive’ applies in the domain of organisms, either of the concepts ‘good’ or
‘bad’ applies in the evaluative domain. Therefore, the two-valued orientation
reduces the number of potential meanings of evaluative expressions: they are
either meant literally or are understood metonymically in the complementary
sense of their opposites. Such mappings seem to occur much more frequently
in domains of evaluation than in others. That is, whereas it would be very
strange to say big and mean ‘small’ in a purely descriptive, non-ironic context,
it is both possible to utter a positive term in order to convey a negative concept
(as in irony and euphemism), or to utter a negative term in order to convey a
positive concept, as in the socially motivated reversals discussed in Section 5.
A further form of metonymic use of negative concepts that seems to be
motivated by factors other than those mentioned so far will be discussed in the
following section.
7. The expressiveness of negative concepts
In Section 4, asymmetry was described as a general property of evaluative

conceptual opposition that is determined by social markedness. There is a
factor that reverses this asymmetry in evaluative conceptual domains, making
negative concepts suitable for the expression of positive concepts even when
there seems to be no social motivation, as in the positive use of bad.
As a first illustration, consider the following conventional expressions:
(22) It was terribly amusing.
(23) He was awfully pleased.
(24) That’s frightfully neat.
These examples show that originally negative expressions can be used adver-
bially to intensify positive ones without creating contradictions. The crucial
meaning component here seems to be not the negative value of the concepts,
but their intensity. They are comparable in meaning to very, although they
convey a higher degree of intensity. Cases like these can be described in terms
of Croft’s (1993) notion of ‘domain highlighting,’ which emphasizes the
communicative function of a metonymy which makes primary an aspect of
something that is secondary in the literal meaning. In these cases, it seems that
mention of the negative concept has the primary function of highlighting the
aspect of intensity connected with it.
As the following examples show, negative expressions can even be used
to express positive concepts:
(25) mad
‘[...] exciting; remarkable; pleasing; excellent’ (Wentworth and
Flexner 1967)
(26) frantic
‘exciting; satisfying; wonderful; cool [...]’ (Wentworth and Flexner
1967)
In these examples, the sense of ‘excitement’ is an element of the positive
meaning. I would like to suggest that the arousal that is connected with
negative concepts is paradoxically made use of when speakers want to express
how significant and/or exciting they regard a positive quality. In that sense,
the negative expression partly fulfills the function of an interjection like wow!
Note that this interpretation is inconsistent with an interpretation in terms

of ironic understatement, where the literal interpretation of the statement does
not figure in the intended, figurative interpretation, and where the pragmatic
function of the utterance is to convey the complete inappropriateness of the
literal interpretation. In the interpretation suggested here, an important ele-
ment of the literal meaning remains ‘active,’ or appropriate, in the figurative
interpretation: the significance or excitement that is attributed to the target
concept by the speaker.
Interestingly, this interpretation is consistent with an observation made
by Wilhelm Wundt (1912) in connection with a related phenomenon. People
sometimes use swear words in order to express affection. They often use them
as pet names, for example, in mocking speech between lovers. For Wundt, this
was “a transition in affective speech onto the negative emotional side that is
used in a positive sense for the sake of its greater intensity” (ibid.: 576; my
translation). Wundt’s observation about the intensity conveyed by such ex-
pressions is crucial because it implies that negative concepts are not employed
in this way although they are negative. On the contrary, they seem to be
employed because they are negative and therefore more intense or prominent.
Consider another example from Who’s Afraid of Virginia Woolf (Albee 1962:
41). The two couples are sitting together, and as before, the whole interaction
is dominated by the struggle between Martha and George.
(27) [George takes from behind his back a short-barreled shotgun, and
calmly aims it at the back of Martha’s head. Honey screams... rises.
Nick rises, and, simultaneously, Martha turns her head to face
George. George pulls the trigger.]
George: POW!!
[Pop! From the barrel of the gun blossoms a large red and yellow
Chinese parasol. Honey screams again, this time less, and mostly
from relief and confusion.]
You’re dead! Pow! You’re dead!
Nick [laughing]: Good Lord.
[Honey is beside herself. Martha laughs too... almost breaks down,
her great laugh booming. George joins in the general laughter and
confusion...]
Honey: Oh! My goodness!
Martha [joyously]: Where’d you get that, you bastard?
[...]
George [a trifle abstracted]: Oh, I’ve had it awhile. Did you like
that?
Martha [giggling]: You bastard.
[...]
George [leaning over Martha]: You liked that, did you?
Martha: Yeah... That was pretty good. [softer] C’mon... Give me a
kiss.
The expression bastard is clearly not intended in the sense of a swear word in
this dialogue: Martha giggles when she utters it and then expresses the wish to
be kissed. In this case, Martha uses this term to express affection. For Wundt,
the main motivation for using language in this way is expressiveness: “It is
only the wish to emphasize the emotion in the strongest possible way that
leads to this result” (Wundt 1912: 576; my translation).
In a very similar context, many native speakers of English find it per-
fectly natural that a mother says You brat! to her child in a tender tone of
voice. There does not really seem to be a conflict between the intended
positive message and the negative semantic content of the utterance. It may be
that a negative emotion concept such as anger or fear can afford mental access
to, and vivid expression of, a positive emotion concept exactly because it is
perceptually or conceptually more prominent and because it therefore allows
the speaker to focus on the intensity of the target concept.
Modern inquiry into the nature of emotions has found that every emotion,
positive or negative, has its own specific effects and intensity, and most
importantly, that certain negative emotions are experienced more intensely
than positive ones; in particular, physiological effects such as heart rate seem
to increase more in the negative emotions fear and anger than in the positive
emotion happiness (cf. Ekman et al. 1983): Following Wundt’s suggestion
noted above, it might be said that negative emotion concepts are, paradoxi-
cally, very suitable for the expression of highly positive concepts. There
seems to be a general difference between cases like these and cases in which
something positive stands for something negative. POSITIVE FOR NEGATIVE, as
in an utterance like Oh, excellent, you idiot!, always seems to involve an ironic
intention of the speaker. When something negative is used to stand for
something positive, this is not the case.
Among the numerous negative slang expressions that are used to stand
for something positive, there is a very interesting group in which expressions
originally connected with ‘fear’ are used to stand for something ‘pleasant.’
Consider the following entries from slang dictionaries:
(28) scare
‘[...] a pleasant surprise’ (Major 1971); “Wow, what a scare!”
(29) panicky
‘extreme pleasure or excitement’ (Major 1971); ‘excellent, very
satisfying or exciting’ (Wentworth and Flexner 1967)
(30) terrible
‘wonderful; great; the best; the most [...] “Terrible — the best; the
greatest.”’ (Wentworth and Flexner 1967)
In these instances the negative concept is used to afford mental access to its
positive opposite because it is more expressive in that it highlights the aspect
of excitement. I would like to argue that these are expressive metonymies,
which accounts for the motivation of the slang expressions mentioned earlier
as well as for the more conventional instances of opposition such as the
adjective terrific mentioned at the beginning.
8. Conclusion
I have tried to argue that opposition can be regarded as a conceptual met-

onymic relationship, suggesting that there are certain regular factors condi-
tioning the use of concepts to convey their opposites: conceptual contiguity,
salience, social factors, the existence of idealized conceptual structures which
people use in organizing the respective conceptual domains, and the function
of metonymies in highlighting certain aspects of these domains.
The relation of evaluative opposition cannot be accounted for in purely
semantic terms. The contiguity between opposites that provides the concep-
tual basis for metonymic mappings is not merely based on lexical association,
but on speaker’s encyclopedic and contextual knowledge. Such situational
and general knowledge also enables speakers to make sense of utterances that
would simply be contradictory in isolation, and explains why the choice of one
concept uniquely determines our understanding of another concept without
creating confusion. In this paper I have attempted to justify the claim that
cases in which one thing stands for its opposite fall under the definition of
metonymy referred to in the introduction of this paper. The discussion has

shown that especially the notions of speaker purpose and context mentioned in
this definition are crucial. The metonymic use and understanding of evaluative
concepts is highly influenced by such extralinguistic factors. For example,
social or interactional factors determine which negative concepts are to be
avoided and which euphemistic positive concepts are to be used instead. They
account for the fact that negative concepts are used to stand for positive ones
as a sign of protest against the attitudes of other social groups, and they allow
us to use positive concepts to stand for negative ones in irony in order to
convey disagreement. Moreover, perceptual and emotional factors account for
the use of negative emotional concepts in place of positive ones because of
their greater intensity and expressiveness.
The metonymic principle of opposition allows people to make sense of
such apparently contradictory utterances. Their two-valued orientation pro-
vides the disposition to conceive of scales in terms of polar opposites and the
conceptual contiguity between opposites provides the ground for metonymi-
cally accessing one opposite term via the other term. Our metonymic under-
standing of an expression as its opposite is triggered by extralinguistic factors
and gives rise to certain evaluative interpretations such as irony or intensity of
expressions.
Notes
1. Antonymy and complementarity are not the only kinds of lexical opposition that are
distinguished in lexical semantics, but they constitute two very basic groups and the
difference between them corresponds largely to the distinction in logic bewteen contra-
diction and contrariness (cf. Lyons 1977: 271).
2. The positive senses of these expressions are also metaphorically motivated: salient,
positive physical measurement antonyms have obviously acquired a positive sense
through metaphorical use.
3. This is not meant to deny that evaluative contrasts are also denotative; rather, evaluative
or connotative opposition exists in many conceptual domains in addition to denotative
opposition.
4. In the same context, Kainz (1965) points out that the word taboo itself could mean both
‘holy’ and ‘cursed’ in its original use by native Australians; the same holds for Latin
sacer.
5. This expression was used in a preview of a thriller on American TV.
References
Aitchison, Jean
1987 Words in the Mind. Oxford: Basil Blackwell.
Albee, Edward
1962 Who’s Afraid of Virginia Woolf. Harmondsworth: Penguin, 1965.
Clark, Herbert C., Richard Gerrig
1984 On the pretense theory of irony. Journal of Experimental Psychology:
General 113: 121–126.
Croft, William
Cruse, D.A.
1986 Lexical Semantics. Cambridge: Cambridge University Press.
Deese, James
1962 On the structure of associative meaning. Psychological Review 69: 161–
175.
1970 Psycholinguistics. Boston: Allyn and Baron.
Ekman, P., R.W. Levenson, W.V. Friesen
1983 Autonomic nervous system activity distinguishes among emotions. Science
221: 1218-1220.
Esper, Erwin A.
1973 Analogy and Association in Linguistics and Psychology. Athens: Univer-
sity of Georgia Press.
Folb, Edith A.
1980 Runnin’ Down Some Lines: The Language and Culture of Black Teenagers.
Cambridge, MA: Harvard University Press.
Fromkin, Victoria
1973 Speech Errors as Linguistic Evidence. The Hague: Mouton.
Garman, Michael
1990 Psycholinguistics. Cambridge: Cambridge University Press.
Givón, Talmy
1978 Negation in language: pragmatics, function, ontology. In P. Cole (ed.),
Syntax and Semantics. Vol. 9: Pragmatics. London: Academic Press, 69–
112.
Hayakawa, Samuel I.
1939 [1978] Language in Thought and Action. Fourth edition. New York:
Harcourt Brace Jovanovich.
Hornby, Albert S.
1974 Oxford Advanced Learner’s Dictionary of Contemporary English. Oxford:
Oxford University Press.
Kainz, Friedrich
1965 Psychologie der Sprache. 4 vols. Stuttgart: Ehmke.
Lakoff, George
Landis, T.Y., D.J. Herrmann, R. Chaffin
1987 Developmental differences in the comprehension of semantic relations.
Zeitschrift für Psychologie 195: 129–139.
Lehrer, Adrienne
1990 Polysemy, conventionality, and the structure of the lexicon. Cognitive
Linguistics 1: 207–246.
Lyons, John
1977 Semantics. 2 vols. Cambridge: Cambridge University Press.
Major, Clarence
1971 Black Slang. London: Routledge.
Murphy, George L., Jane M. Andrew
1993 The conceptual basis of antonymy and synonymy. Journal of Memory and
Language 32: 301–319.
1995 Relevance. Second edition. London: Basil Blackwell.
Thorne, Tony
1990 Dictionary of Contemporary Slang. London: Bloomsbury.
Wentworth, Harold, Stuard Flexner
1967 Dictionary of American Slang. New York: Crowell.
Wundt, Wilhelm
1912 Völkerpsychologie. Band 2. Leipzig: Engelmann.
Metonymic Hierarchies
The Conceptualization of Stupidity in German

Idiomatic Expressions
Kurt Feyaerts
University of Leuven (KUL - KULAK)
1. Introduction
One of the tenets of cognitive semantic is the focus on figurative language use
as a central topic of cognitive semantic research.1 As opposed to structuralist
linguistic approaches, where phenomena like metaphor and metonymy are
considered mere rhetorical or semantically peripheral devices, cognitive se-
mantics demonstrates that metaphor and metonymy are essentially ‘concep-
tual phenomena’ which show up only secondarily in linguistic expressions.
Up until now, the field of cognitive semantics has witnessed an overwhelming
interest in the role of metaphor as a conceptual mechanism structuring large
parts of our knowledge. Although the notion of metonymy was never entirely
absent, it was mostly relegated to the secondary status of being introduced or
mentioned in an essentially metaphorical context. However, recent years have
seen an increasing interest in the role metonymy plays in the structuring of our
conceptual system. Several studies seem to indicate that metonymy is catching
up to metaphor as a relevant area of study in the field of conceptual structure.
This article forms part of this renewed interest in metonymy. Its purpose
is to extrapolate the notion of ‘metaphoric hierarchies’ to metonymy, thus
showing that there are also large-scale metonymic structures to be found in
which several individual metonymies take part. I illustrate this with German
idiomatic expressions which all designate some kind of stupidity. It will
310 Kurt Feyaerts
become clear that most of the expressions in this domain involve a metonymic
relationship between source and target structure.
This paper is structured as follows. First, in Section 2, I will situate this
topic within the context of my larger research project. In Section 3, I will
briefly bring to mind some major characteristics of metaphoric hierarchies. In
Section 4, I will discuss some theoretical aspects of metonymy which are
related to the kind of material examined in this study. Finally, Section 5 will
concentrate on some examples of metonymic hierarchic structures as they
appear in this material.
2. Preliminary remarks
For a better understanding of the theoretical discussion about metaphor and

metonymy (Sections 3 and 4) as well as the analysis of the data (Section 5), it
is essential to first address a few general theoretical and methodological issues
of my research. Section 2.1. describes the nature of the data collected as well
as the methodological basis on which the corpus was put together. An elabo-
rated example illustrates the kind of semantic extension patterns that will be
analyzed in this study. Section 2.2. concentrates on the epistemological status
of the structures analyzed. Referring to observations of the cognitively in-
spired person perception psychology, I will argue that the object of a cognitive
semantic study can only be of a strictly conceptual, non-objectivist nature.
2.1. Background and material
Within the overall cognitive semantic framework as it appears, amongst

others, in the writings of Lakoff and Langacker, this article must be situated
against the background of my investigation into the conceptual structure of
STUPID in approximately 500 idiomatic German expressions. What allows
these expressions to be categorized as members of a single category is the fact
that they all profile a schematic concept, STUPID, which I define in a rather
broad way as ‘a negatively valued, non-pathological deviance from the norm
in the domain of mental abilities.’ For present purposes, I abstract from several
kinds of stupidity like ignorance, credulity, narrow-mindedness, etc., which
all represent different elaborations of this schematic notion.
The database consists of more or less fixed expressions, most of which
Metonymic Hierarchies 311
are collected from phraseological dictionaries. As a consequence of this

methodological approach, native speakers of German may not be familiar with
some of the material presented here. However, from a cognitive semantic,
usage-based point of view, it is important not to restrict the description to an
arbitrary set of familiar, well-established expressions. Instead, an adequate
description of the conceptual structure of a particular concept requires that as
many relevant expressions as possible be taken into account.2
The expressions in (1)–(3) represent examples demonstrating a met-
onymic extension to the target concept STUPID.
(1) Du bist wohl nicht von hier?
‘You are not from here, are you?’
(2) Du bist wohl aus der Paddengasse?
‘You are from the Toad Alley, right?’ (Street in Berlin)
(3) Ich komme nicht aus Buxtehude!
‘I don’t come from Buxtehude!’ (Town near Hamburg; ‘I am not
stupid’)
These expressions all instantiate the metonymic pattern OUTGROUP ORIGIN FOR
STUPIDITY. This low-level, conceptually rich metonymy elaborates a more
schematic extension pattern, namely A SALIENT PROPERTY FOR A LESS SALIENT
ONE.3 This schematic metonymy plays a central role in the conceptual struc-
ture of the target concept. It appears that a mental property like STUPID is most
often structured in terms of physical or social properties, which, in comparison
to the abstract feature STUPID, are easier and more ‘objectively’ perceived.
Regarding this particular conceptual metonymy, it has been pointed out
by Christie Davies (1987, 1990) that this extension shows up in many cultures.
She observes that the people who are the object of stupidity jokes are generally
(a) groups living on the periphery of a linguistic culture, where (b) people
speak a marked variety of the standard language used in that culture. Conse-
quently, ‘outgroup’ origin has to be understood as ‘not belonging to the central
group.’ An excellent example of this cultural pattern are the Belgians since
they are called stupid in two respects. This may be explained by referring to
Davies’ theory combined with the fact that Belgium is divided in two major
regions: Dutch-speaking Flanders and French-speaking Wallonia. On the one
hand, Flemings live in the southern part of the Dutch-speaking culture and
speak a variant of standard Dutch. Walloons, on the other hand, occupy the
312 Kurt Feyaerts
most northern part of the French-speaking culture and speak French with a
heavy regional accent. Belgians are consequently called stupid by the Dutch
as well as by the French. What makes the double use of this metonymic model
with respect to Belgians particularly interesting is the fact that it hides another
metonymy according to which, in both cases, the whole stands for one of its
parts. Accordingly, the Dutch as well as the French envisage in their stupidity
jokes the Belgians and not the Flemish or the Walloons, respectively.
Of particular interest for the present purpose is the observation that in this
cognitive-cultural model for stupidity, a functional relationship is established
between people’s origin on the one hand and their mental abilities on the
other. To put it in linguistic terms, in (1)–(3) both properties are ‘conceptual-
ized’ as being contiguous. In this case contiguity is realized as a kind of causal
relationship according to which stupidity is determined (or caused) by a
specific origin. In this construed relationship, stupidity is linked to another
property as its effect or manifestation.
2.2. Non-objectivist folk models
Before I continue the analysis one point needs to be made very clear. What I
am aiming at here is the description of a ‘folk model’ for stupidity as it appears
in linguistic expressions rather than the verification of these models from the
point of view of observing experts. Logically, one knows that this kind of
functional link between geographic origin and mental ability is not generally
‘true’ or justifiable. Instead, I aim to find the conceptual structures which are
activated (un)consciously in the spontaneous categorization of a person as
STUPID. In this description I thus abstract from people’s reflection over this
value judgment. Lakoff (1987: 85f) describes these culturally determined
conceptual patterns as ‘social stereotypes,’ which he classifies as a subcat-
egory of metonymically structured cognitive models:
[...] they define cultural expectations, they are used in reasoning and espe-
cially in what is called ‘jumping to conclusions.’ However, they are usually
recognized as not being accurate, and their use in reasoning may be overtly
challenged.
These social stereotypes function as cognitive models and they reflect culturally
established preferences and norm concepts which play a decisive role in
people’s judgments about typical properties of all kinds of personality types. As
an example of a contemporary American stereotype, Lakoff describes the
stereotypical politician as “conniving, egotistical, and dishonest” (ibid.). In this

article I will try to reveal some aspects of the German stereotype for stupidity.
Lakoff’s theoretical description of social stereotypes functioning as cul-
turally determined interpretation models is supported by observations of
person perception psychology. As a matter of fact, these observations provide
a general cognitive, culturally independent grounding for this kind of cultural
model.4 According to person perception theory we are inherently engaged in a
constant and mostly unconscious process of observing and comparing people
with each other and with ourselves. It is this continuous process of comparison
in which we use our own socially determined norm system as a frame of
reference that underlies our value judgments.5 Of great importance for the
present purpose is the observation that we always encounter another person
primarily in his/her most salient aspects. Obviously, most prominent are those
characteristics which can be sensorially perceived like general appearance,
behavior, clothing, color of skin, but also aspects like sex, age, language use
(lexical choice, accent, speed of speech, body odor, etc.). Also other objective,
more external personal ‘facts’ like place of residence, origin, profession,
membership in a social class, etc. belong to this category of salient character-
istics. Properties like these play a determinant role in our first, spontaneous
judgments about other people on the basis of which we often draw far-
reaching inferences regarding someone’s stable personality traits. This im-
plies that value judgments like for example X is stupid, honest, not reliable,
which aim at characterizing someone in terms of a stable personality trait, are
essentially to be interpreted as generalizations made by the conceptualizer on
the basis of salient, but mostly partial knowledge. In their description of the
perception process Bruner et al. (1958: 277) refer to this dynamic nature of
human observations:
In forming impressions of a person, we usually start with partial information
[...]. These circumscribed bits of evidence generate none the less a host of
inferences about a person, inferences that are made with varying degrees of
certainty. One ‘knows’ more about a person than what seems to be immedi-
ately connoted by the acts one has witnessed or the information one has
gained about him. To ‘know’ is not used here in the sense of ‘know correctly.’
This implies a conceptualizer who actively construes the conceptual structure

of his environment according to his/her interaction with it. As Heider (1988:
336) puts it: “perception is basically a constructive act rather than a receptive
or simply analytic one.”
314 Kurt Feyaerts
In sum, one can say that cognitively oriented person perception theory
does not provide any grounding for the specific linking of conceptually
elaborated properties with each other. Instead, its value for the present analy-
sis must be situated on a more schematic and structural level where it provides
experiential grounding for the conceptualized contingency between two or
more seemingly unrelated properties. Its specific importance for the present
purpose is its function as a grounding principle for the metonymic structure of
the target concept STUPID.
3. Metaphoric hierarchies
Before I concentrate on metonymy, I first bring to mind some principal

characteristics of conceptual metaphoric structures. As to its most fundamen-
tal characteristic, metaphor is defined as a ‘conceptual’ pattern underlying and
motivating individual lexical expressions. This pattern involves the mapping
onto each other of two concepts belonging to different knowledge domains, as
a result of which one concept (the target) is structured in terms of the other
concept (the source). The aspect which attracts my special attention in this
context is the integration of single conceptual metaphors into larger hierarchi-
cally organized structures.6 Take, for example, Figure 1, which focuses on
(parts of) the conceptual structure of the metaphor LIFE IS A JOURNEY. This
mapping can be described as an elaboration of at least three higher schematic
mappings. Two of these mappings (TIMES ARE THINGS and STATES/ATTRIBUTES
ARE THINGS) are characterized by what Lakoff (1993: 218) calls ‘duality
structure,’ which means that these metaphors can be realized in a location/
object pair (compare the duals (or pairs) TIMES ARE MOVING OBJECTS vs. TIMES
ARE LOCATIONS). In this example I draw special attention to the location
structure of both duals. With respect to the third schematic metaphor, EVENTS
ARE ACTIONS, I realize that LIFE and JOURNEY represent summary scannings of
an event (LIVING) and an action (TRAVELING) respectively and that strictly
speaking they do not elaborate this schema in a direct way, but in order to not
complicate the schematic representation unnecessarily I will maintain this
conceptual shortcut. (In my figures, full lines represent patterns of elaboration
and schematization, capitals indicate conceptual structures and italics are used
for linguistic instantiations.)
This representation of the LIFE IS A JOURNEY metaphor as part of a larger
T IM E S A R E T H IN G S E V E N T S A R E A C T IO N S S T A T E S / A T T R IBU T E S
A R E T H IN G S
T IM E S A R E T IM E S A R E S TAT ES AR E A T T R IBU T E S
E N T IT IE S L O C A T IO N S L O C A T IO N S AR E
P O S S E S S IBL E
O BJ E C T S
P A S S IN G O F P A S S IN G O F L IF E IS A C H A N G IN G C H A N G E IS
T IM E IS T IM E IS P E R S O N JOUR NEY S T A T E IS M O VEM E NT O F
M O V IN G E N T IT Y M O V IN G W IT H C HANGE OF O BJ E C T
R ES P EC T TO L O C A T IO N
L O C A T IO N
F U T U R E IS S e v e ra l e l a b o r a t i o n s : C H A N G E IS
E N T IT Y M O V IN G M OVEM ENT OF
TO WAR DS L IN E O F C O N D U C T I S A P A T H O BJ E C T
P R ES E NT (H e g o t o f f t h e ri g h t t rack )
T he tim e of (M IS )F O R T U N E IS A P A T H H e h as l o s t h i s
l eav i n g w i l l s o o n (S h e ran i n t o b ad l u ck ) l u ck
co m e
D Y IN G IS G O IN G T O A F IN A L
D E S T IN A T IO N
(H e m o v ed t o t h e o t h er w o rl d;
H e p as s ed aw ay )
Figure 1. Conceptual structure of the LIFE IS A JOURNEY metaphor
conceptual system allows for three major structural observations to be made

with regard to the overall metaphoric organization of our knowledge struc-
tures. A first characteristic concerns the different ‘levels of schematicity’ on
which a particular concept or mapping of concepts can be described. Lakoff
(1993: 222) observes this in terms of ‘inheritance hierarchies’ in which
“‘lower’ mappings in the hierarchy inherit the structures of the ‘higher’
mappings.” The relationships between the concepts in this hierarchy can be
basically determined in terms of schematization and elaboration. The lower a
concept or mapping is situated in a hierarchy, the more elaborated and specific
it is. On the lowest level, this process of specification finally results in
concrete linguistic instantiations. A second structural observation concerns
316 Kurt Feyaerts
the ‘patterns of overlap’ that show up throughout these hierarchies. This

means that a particular mapping may elaborate more than just one schematic
structure, thus incorporating an overlap between the schematic structures
which are involved. In this example, LIFE IS A JOURNEY elaborates both STATES
ARE LOCATIONS as well as TIMES ARE LOCATIONS.7 With regard to the overlap-
ping target concepts, life can be considered a state one is in (alive or dead) as
well as a period of time (compare: a lifetime). On the side of the source
domain, a journey implies directed movement along multiple locations to-
wards a final destination. Vis-à-vis these two observations, Lakoff (1993)
correctly points out that “metaphorical mappings do not occur isolated from
one another” (ibid.). Referring to the representation in Figure 1, one could
further specify this by saying that they are isolated from one another neither
vertically (in terms of schematization and elaborations) nor horizontally (in
terms of overlap). Finally, a third observation points out the different nature of
metaphorical mappings on different levels of abstraction. Lakoff (1993: 224)
indicates that “the metaphors higher up in the hierarchy tend to be more
widespread than those mappings at lower levels.” According to this correla-
tion, it seems that highly schematic metaphors are of a more general-cognitive
nature than lower level metaphors which tend to be more culture-specific.8
4. Defining metonymy
4.1. General characteristics
Before I look at similar metonymic hierarchies, I will explain my interpreta-

tion of metonymy as a conceptual phenomenon. It is not my purpose here,
however, to provide a full account of the various interpretations of metonymy.
In the relevant contemporary literature two major global approaches may be
distinguished. A first approach describes metonymy in terms of ‘contiguity’
and thus focuses on the nature of the relationship between the concepts
involved. It finds its origin in traditional structuralist theories, in which
linguistic meaning is seen as adhering to the objective reality. In this view, the
notion of contiguity is basically limited to an observable, real-world relation-
ship between two referents.
The second approach was developed only recently in cognitive semantics
and describes metonymy in terms of the conceptual range of the extension
involved. More specifically, it defines metonymy as a conceptual extension

taking place within the boundaries of a single domain matrix (cf. Croft 1993:
348, who speaks of ‘domain highlighting’) which is accompanied by a refer-
ential shift (compare, among others, Lakoff 1987; Croft 1993; Langacker
1993; Goossens 1995). A domain matrix is defined as the whole of knowledge
structures which are activated in multiple domains as the conceptual back-
ground of a profiled concept.
4.1.1. Metonymy as conceptualized contiguity

An aspect that attracts my special interest in the contemporary approach to
metonymy is the conceptual reinterpretation of the notion ‘contiguity.’ Taylor
(1989: 123f) observes rather vaguely that the “essence of metonymy resides in
the possibility of establishing connections between entities which co-occur
within a given conceptual structure.” The basic ‘conceptual’ nature of me-
tonymy is emphasized more strongly by Schmid (1993: 94; “konzeptualisierte
Kontiguität,” ‘conceptualized contiguity’) and Dirven (1993: 11ff). In this
respect, I fully subscribe to the following claim made by Dirven (1993: 14)
according to which
contiguity cannot be based on any form of objective or ‘natural’ contiguity.
This has the far-reaching implication that contiguity must be taken to mean
‘conceptual contiguity’ and that we can have contiguity when we just ‘see’
contiguity between domains.
This view describes metonymy in a radical departure from a mere rhetorical

trope as an experientially determined product of ‘conceptualization,’ i.e., as a
contiguity relationship between ‘interactional,’ non-inherent properties, enti-
ties, etc. This approach expresses the cognitivist view of reality as a domain
which does not exist independently of human understanding, knowledge and
belief.
Within this cognitive framework, I identify metonymy in terms of the
nature of the conceptual relationship rather than the range of the extension vis-
à-vis the boundaries of a domain matrix. More specifically, I oppose contiguity
as the essential feature of metonymy in the same way that a relation of similarity
is taken to characterize metaphor. There are different kinds of functional-
associative relationships like CAUSE-EFFECT, OBJECT-PROPERTY, CONTAINER-
CONTENTS, etc. which can be categorized as contiguous. It is not my purpose,
however, to present an extensive overview of the different kinds of contiguity
relations in an attempt to find a schematic definition for this concept (see,
318 Kurt Feyaerts
however, Radden and Kövecses, this volume). My focus will be restricted to

just a few relationships which are typical for my data. In this respect, I briefly
refer to argumentation theory where this kind of conceptual relationship is
denoted as ‘reasoning by association’ which “consists in unifying elements into
a single whole by bringing together elements which were previously regarded
as separate” (van Eemeren et al. 1996: 106).9 Interestingly, as prominent
examples of reasoning by association, van Eemeren et al. (1996) describe
argumentation models based on the relationship people draw between elements
like ‘cause’ and ‘effect,’ ‘a person X’ and ‘X’s actions,’ ‘the essential’ and ‘its
manifestations,’ etc. In the course of this article, it will become clear that this
kind of contiguity is of major importance for the conceptual structure of the
target STUPID.
4.1.2. The ‘domain matrix’ approach

What makes the second approach to metonymy somewhat problematic is the
use of the notion ‘domain matrix’ as the key element for the distinction
between metaphor and metonymy. According to this approach, metonymy is
defined in terms of a conceptual extension which does not cross the bound-
aries of a domain matrix. It must be clear that there can be no doubt about the
importance of the notion of domain matrix for a cognitive theory of meaning.
However, the question regarding which conceptual structures are activated as
the conceptual background for the meaning of an expression (a profiled
concept) essentially pertains to the ‘psychological reality’ of the conceptual
networks we use to represent the structure of our knowledge. Bearing in mind
the cognitive semantic tenet that meaning (semantic structure), which is in
principle equated with the rich conceptual structure, is experientially based
and consequently to a large extent individually determined, it is doubtful that
this conceptual structure can be described accurately from an external theo-
retical perspective.10 One must then try to determine the criteria that precisely
motivate which conceptual structures belong to a domain matrix and which
concepts do not. Drawing distinct boundaries around a domain (matrix)
always reflects an arbitrary intervention by a linguist. As such, this observa-
tion does not pose such a big problem since no linguistic research can ever
escape this kind of subjective interference. However, it does raise a method-
ological problem when the distinction between two conceptual phenomena is
based on the arbitrary delimitation of a domain matrix, especially when
independent support is not sought.11
Consider, for example, the conceptual relationship between the concepts

KNOWING and SEEING, which is traditionally described as a metaphoric one
(KNOWING IS SEEING). The question is, however, how can it be demonstrated in
a convincing way that a concept like VISION or, more schematically, SENSORY
PERCEPTION does not belong to the domain matrix of an expression which
profiles a concept like MENTAL PERCEPTION/KNOWING? The answer to this
question would have to explain that one does not experience sensory/visual
perception as causally or conditionally linked to knowledge. Although I do not
think that this would be an adequate representation of the conceptual struc-
tures that get activated as a conceptual background for KNOWING, finding a
correct answer to this question is not my major concern. What I am aiming at
is this: who can say whether and why a particular description of a domain
matrix is correct or not? Intuitively, I experience the conceptual relationship
between SEEING and KNOWING to be of a different order than, for instance, the
relationship between LIVING and TRAVELING, which essentially reduces to a
structural comparison (‘life is like a journey’). Although I admit that both
concepts can be structurally mapped onto each other, involving ontological,
image schematic and logical structures, I claim that this description does not
exhaust the conceptual relationship between both concepts. An important
aspect of this relationship is the causal-conditional contiguity of both experi-
ences, which indicates that a metonymic extension (PERCEPTION FOR RESULT OF
PERCEPTION) can be identified as well. This observation seriously questions the
hypothesis of two different domain matrices being involved in this extension
pattern. This example shows that on a theoretical level ‘domain matrix’ is a
notion too malleable to serve as a criterion for distinguishing important
conceptual phenomena.
Another problem which further questions the efficacy of ‘domain matrix’
as an adequate criterion and which develops within the logic of the domain
matrix approach itself, is the definition of metaphor as “a mapping between
two domains that are not part of the same matrix” (Croft 1993: 348). The
question raised by this definition is whether the mere crossing of a boundary
of a domain matrix by an extension represents an adequate and exhaustive
characterization of metaphor as a conceptual phenomenon. As a matter of fact,
one cannot exclude metaphoric mappings taking place within the boundaries
of a domain matrix. Consequently, my claim is that metaphor primarily
depends on something different than a crossing of domain boundaries. To
illustrate this, I take an example used by Goossens (1995) in his analysis of the
conceptual structure of LINGUISTIC ACTION:
320 Kurt Feyaerts

Goossens rightly distinguishes two possible interpretations for the verb to
giggle in this sentence. The first one is clearly metonymic (‘to say something
while giggling’) with a part referring to the whole, whereas the second one (‘to
say something as if giggling’) is mainly metaphoric. Goossens (1995) charac-
terizes this second interpretation as an example of ‘metaphor from metonymy’
because in the metaphoric reading, “the conceptual link with the metonymic
reading is still present” (ibid.: 164). In both interpretations, the domain (HU-
MAN) SOUND functions as the source domain. Notably, Goossens argues that in
the metaphoric reading an extension is taking place between two separate
domains which do not belong to a single domain matrix. I do not see any
reason, however, to conclude that in this interpretation the domain (HUMAN)
SOUND does not belong to the domain matrix of the profiled concept LINGUISTIC
ACTION. In my opinion, this auditory experience represents an essential aspect
of every linguistic utterance. Again, the question is, how can it be motivated
on independent grounds that SOUND is not activated as part of the conceptual
background of the meaning ‘speaking as if giggling.’
What makes example (4) particularly interesting is the reasoning used by
Goossens to demonstrate the metaphoric status of this extension. The crossing
of domain boundaries is presented as a secondary feature depending on the
nature of the conceptual relationship between source and target which is
described in terms of a comparison (‘as if’): “[a]nother way to interpret it is
that she said this as if giggling; hence [my emphasis; K.F.] there is a crossing
of domain boundaries, we have a metaphor” (ibid.: 164). This observation
seems to support my view that the primary and decisive criterion for the
distinction between metonymy and metaphor resides in the nature of the
conceptual relationship (contiguity vs. similarity). The metaphoric status of
this semantic extension is also clear without the reference to a domain bound-
ary appearing ex nihilo. This does not mean, however, that the notion of
domain matrix has no importance at all in this question. In fact, defining
metonymy in terms of conceptualized contiguity entails the criterion of an
extension taking place within a single domain matrix. The construal of two
concepts or domains as being contiguous implies that both structures “are seen
to be in line” (Dirven 1993: 14), which means that the restriction of a
metonymic extension to a single domain matrix is to be considered an inherent
effect of construing two conceptual entities as contiguous.
4.2. Hyperbolically used metonymies
Before I consider some examples of metonymic hierarchies, I want to draw

attention to a specific aspect of the expressions found in my material. Since
they all profile a negative value judgment, it is not surprising that most of these
expressions display a high degree of expressiveness. A stylistic/rhetorical
device which is often used in this material is hyperbole, which “involves the
exaggerated expression of a negative or positive appreciation of something”
(Geeraerts 1994: 2823). Hyperbolic expressions represent a typical example
of figurative language use.12 Compare for example the following expressions:
(5) Er kann nicht bis drei zählen.
‘He cannot count to three.’
(6) Sie glaubt, Gott heiße Gerhard.
‘She believes God’s name is Gerhard.’
(7) Er weiß nicht, daß es zwei Arten von Menschen gibt.
‘He does not know that there are two kinds of human beings.’
Expressions like these are clearly metonymic in nature. The concepts which
are involved in each of these expressions are associated by some kind of a
causal relationship according to which, for example, ‘a person’s inability to
count’ is presented as a prominent manifestation of the essential property of
‘stupidity.’ The schematic metonymy which is elaborated in these examples is
EFFECT FOR CAUSE. Hyperbolic expressions are commonly treated as typical
examples of figurative speech, which raises the question whether a distinction
should be made between ‘literal’ and ‘figurative’ metonymy. At first glance,
there seems to be a difference between expressions as in (5)–(7) and the
examples which are commonly presented as metonymies. It appears that the
examples in most studies on metonymy are mostly of the ‘literal’ kind, where
the semantic extension can be interpreted as an objectively verifiable, referen-
tially ‘realistic’ contiguity relationship. Compare, for example, the different
metonymies mentioned by Lakoff and Johnson (1980: 38f):
PART FOR WHOLE Get your butt over here!
PRODUCER FOR PRODUCT I’ll have a Löwenbräu.
OBJECT USED FOR USER The sax has the flu today.
CONTROLLER FOR CONTROLLED Nixon bombed Hanoi.
322 Kurt Feyaerts
INSTITUTION FOR PEOPLE RESPONSIBLE Exxon has raised its prices again.
THE PLACE FOR THE INSTITUTION The White House isn’t saying
anything.
THE PLACE FOR THE EVENT Remember the Alamo.
Expressions pointing to an imaginative, ‘non-realistic’ contiguity relationship
are only scarcely mentioned, although precisely in this kind of relationship the
conceptual status of a metonymic extension becomes very clear. The conclu-
sion that we are dealing with two different types of metonymy, i.e., ‘literal/
realistic’ vs. ‘figurative/imaginative’ metonymy, cannot be maintained. In
both cases, the conceptual-semantic relationship involves two interactional
(attributed) properties. Accordingly, even a contiguity relation which is expe-
rienced as being ‘realistic,’ inevitably pertains to a conceptually mediated
reality. This understanding is not unimportant for an adequate analysis of
metonymy as a conceptual mechanism. It shows that in both cases a contiguity
relationship, which is to be considered the conceptual basis of metonymy, is
involved.
Of central importance for an adequate analysis of examples such as in
(5)–(7) is the insight that phenomena like hyperbole and metonymy are
situated on two different semasiological dimensions. Metonymy, on the one
hand, concerns the contiguity relationship between two semantic structures
and thus basically pertains to the ‘conceptual-semantic’ level. Compare, for
example, the relationship between the semantic source structure ‘unable to
count to three’ and the target structure ‘stupid’ in (5). The observation that a
property like ‘unable to count to three’ is ‘not realistic,’ is of no relevance for
the purpose of this study, since the relationship I am interested in involves
nothing but attributed semantic structures. Hyperbole, on the other hand,
arises from the ‘inappropriate’ use of a linguistic expression in a particular
situation and thus primarily concerns a specific ‘referential’ aspect of a
semantic structure. To be more precise, the experience of an utterance being
hyperbolic has to be understood against the background of a narrow interpre-
tation of the concept of reference as ‘correct reference’: “if the meaning of the
expression is a true description of the thing it indicates, then the expression
refers correctly” (Geiger 1993: 278). In normal everyday communication, we
expect (most of the) utterances to be referentially true, but in using hyperboles,
we deliberately violate this communication principle for stylistic,
argumentational or other reasons. This narrow concept of reference is of no
interest to a semantic analysis, where the notion of truth does not play a
prominent role. Accordingly, in describing metonymy, my attention is drawn
by the relationship between two semantic structures, regardless of whether
these semantic values are ‘true’ or ‘real’ for a referent or not. Lyons (1977:
184) points out that the search for truth should be left to philosophers:
philosophers are professionally concerned with the explication of the notions
of truth, knowledge, belief and existence. The fundamental problem for the
linguist, as far as reference is concerned, is to elucidate and to describe the
way in which we use language to draw attention to what we are talking about.
In light of these observations, it does not seem appropriate to make a distinc-

tion between two kinds of metonymy. This distinction would implicate that a
phenomenon like hyperbole interferes with the semantic (metonymic) struc-
ture of an expression. I have tried to show that this is not the case. Therefore I
prefer to speak of ‘hyperbolically used metonymies.’
5. Metonymic hierarchies
On the basis of this characterization of metonymy, I can now turn to some large-
scale metonymic structures that generalize over several linguistic instantiations.
I will show that metonymic hierarchical systems display similar structural
characteristics to the ones I described for metaphoric hierarchies (see above).
First, I will discuss two specific examples and then I will present a schematic
picture of the conceptual structure of STUPID which will demonstrate the
structural role metonymy can play as a pervasive conceptual mechanism.
(8) Er steht da wie der Ochse vor dem Scheunentor.
‘He stands there like the ox in front of the barn-door.’
This example concerns the conceptual background of a non-conventionalized
expression which profiles stupidity in terms of a particular behavior: a person
who is doing nothing else than just standing there like an immobilized object
(NOT MOVING). More specifically, an expression such as (8) incorporates an
overlap between the two concepts DEVIANT BEHAVIOR and DEVIANT APPEAR-
ANCE. A deviant behavior in its turn elaborates both concepts DEVIANT SOCIAL
PROPERTY and DEVIANT PHYSICAL PROPERTY. On a higher schematic level, these
latter concepts are schematized as SALIENT DEVIANT HUMAN PROPERTY. This
324 Kurt Feyaerts
conceptual hierarchy is represented in Figure 2, which mainly focuses on the

internal structure of the source concept. Accordingly, the metonymic exten-
sion to the target is mentioned only once, namely on the level of the linguistic
expression (STUPID). The metonymic relationship involved here can be para-
phrased as ‘stupidity shows up in deviant behavior.’ (Broken arrows are used
to indicate metonymic extensions.)
With respect to the three structural characteristics I distinguished for
metaphoric hierarchies, it is obvious that conceptual metonymy can be de-
scribed on different levels of abstraction as well. Profiling stupidity as a
particular way of standing can be subsumed under the highly schematic
pattern in which a salient property stands for (expresses) a less salient one,
since physical and social properties can be perceived more easily than an
abstract mental property. As already mentioned above, conceptual/linguistic
metonymies also display patterns of overlap between higher schematic struc-
tures. Not surprisingly, the third characteristic applies to metonymically orga-
nized systems as well: metonymies of a higher, more schematic level tend to
be of a more universal, purely cognitive nature. Lower metonymies on the
other hand, appear to be more culturally restricted.
D E V IA N T H U M A N P R O P E R T Y
D E V IA N T S O C I A L P R O P E R T Y D E V IA N T P H Y S IC A L P R O P E R T Y
D E V IA N T BE H A V IO R D E V IA N T A P P E A R A N C E
N O T M O V IN G
E r s t eh t da w i e der O ch s e v o r dem S ch eu n en t o r
S TU P ID
Figure 2. Hierarchical structure of the DEVIANT BEHAVIOR FOR STUPIDITY metonymy

I want to illustrate this cultural dependence with a second example which

refers to a typical German social stereotype in which stupidity plays a major
role:
(9) Mantafahrer!
‘Manta-driver!’
This metonymic model gets its name from a particular kind of car (Opel
Manta) which can be characterized as having a sporty look, a low and
aerodynamic profile, a powerful engine, etc. Moreover, it is a relatively cheap
car which made it a big commercial success. At the same time, however, this
popularity has ‘degraded’ it to an ordinary car everyone can afford. To
(re)gain some exclusiveness many people equip their Opel Manta with fancy,
aggressive-looking accessories like chrome spoilers, muscle tires, rally seat
belts, high-beam headlamps, racing colors, etc. All these elements contribute
to a typical image of this car as an ‘ordinary car dressed up with kitschy
gadgets.’ What is particularly interesting in this context, is the cultural (Ger-
man) model in which this car is associated with a typical kind of people
owning and driving it. This social stereotype characterizes a Manta-driver as
macho, having a blond girlfriend or wife, belonging to a lower social class,
driving fast and aggressively, and also as not very intelligent but rather stupid.
Calling somebody a Mantafahrer thus can be interpreted in a particular
context as ‘he is stupid.’ This conceptual extension is clearly metonymic in
that salient properties like deviant behavior, appearance and possession (i.e.,
the car itself) are construed as manifestations of stupidity. Accordingly, these
concepts all take part in the source domain structure which serves as the
conceptual base for profiling the target STUPID. To put it another way, these
source concepts overlap in any stupidity expression that instantiates this
Mantafahrer-model. This model is represented schematically in Figure 3. This
hierarchy shows how, in its metonymic extension (broken arrow) onto the
target concept STUPID, the culturally specific Mantafahrer-concept elaborates
(full lines) several schematic concepts.
Figure (4) presents a global representation of the schematic conceptual
structure of the target domain. The difference between this and previous
figures is that in this case I have brought together the most important source
concepts in a single system. The many broken arrows indicate that STUPID is
essentially metonymically structured.13 Presented in this figure are only the
most schematic levels of the conceptual structure. Not included are several
326 Kurt Feyaerts
S ALIENT DEVIANT HUM AN P ROP ER TY
DEF IC IENT S OC IAL P R OP ER TY DEF IC IENT P HYS IC AL P ROP ER TY
DEVIANT BEHAVIOR DEVIANT AP P EAR ANC E DEVIANT P OS S ES S ION
M ANTAF AHR ER
S TU P ID
Figure 3. Hierarchical structure of the Mantafahrer-model
lower-level schemas which further elaborate parts of this structure. Notice that
I consider the target structure STUPID as an elaboration of DEFICIENT MENTAL
ABILITIES and that, normally, one would expect the abstract and undifferenti-
ated concept STUPID to appear right below this schema. In order to keep this
figure as clear as possible, however, I moved it to the bottom. The numbers
between brackets refer to instantiating expressions which correspond to ex-
amples mentioned below or elsewhere in the text.
An interesting observation concerns the complex source structure which
can be observed in (5). On the one hand, not being able to count to three
elaborates both mental as well as practical abilities. On the other hand,
however, this deficiency also relates metonymically to little children who are
not able to count properly so that this notion — albeit not very prominently —
also takes part in the conceptual structure of the source domain. It thus appears
that — in applying the theory of complex conceptual networks — a complex
source domain structure may consist of schematically-related, as well as
(metonymically) extended concepts. A similar conceptual pattern underlies
the expression in (10), which instantiates a deviant appearance (the most
prominent source structure) and at the same time metonymically refers to a
small (newborn) child (DEVIANT AGE) which typically has this kind of appear-
ance. Both conceptual relationships combine into the complex source struc-
ture which metonymically stands for the target STUPID.
de v i a n t h um a n p ro p e rt y
s ev i a n t s o c i a l p r o p er t y de v i a n t p h y s i c a l p ro p e rt y
de fi c i e n t m ar k e d de v i a n t de v i a n t de v i a n t de fi c i e n t
a b i l i t i es ag e b eh av i o r s t a t us a p p e a ra n c e b o dy p a rt
(1 3 )
def i c i e n t de fi ci en t de fi c i en t low o ut g ro up de fi ci en t de fi c i e n t
s o ci al p ra c t i c a l m e n t a l p ro fe s s i o n o ri g i n h e a d/ b r a i n s e n s o ry
ab i l i t i es ab i l i t i es ab i l i t i es (1 4 ) (1 - 3 ) (1 2 ) a p p a ra t us
(1 5 ) (7 ) (1 1 )
(8 )
(5 )
(9 )
(1 0 )
S T U P ID
Figure 4. General conceptual structure of STUPID
(10) Er ist nicht trocken hinter den Ohren.

‘He is not dry behind his ears.’
(11) Sie hat nicht alle Sinne beisammen.
‘She does not have all senses together.’
(12) Er hat Wasser im Kopf/Er hat ein weiches Gehirn.
‘He has water in his head.’/‘He has a weak brain.’
(13) Sie ist nicht so dumm wie sie aussieht.
‘She is not as dumb as she looks.’
328 Kurt Feyaerts
(14) Das ist noch unter dem Nachtwächter./Er ist ein geistiger
Untergefreiter.
‘That is even below the night watchman.’/‘He is an intellectual
“under-corporal”.’
(15) Er ist zu dumm, um einen Eimer Wasser umzustoßen.
‘He is even too stupid to knock over a bucket of water.’
Most of these examples speak for themselves, except perhaps for (13), which is
not an expression that profiles stupidity in a direct way. On the contrary, it
profiles a positive mental property. However, the image which is used confirms
the social stereotype according to which mental ability can be inferred from
somebody’s appearance (‘appearance as a manifestation of mental abilities’).
From a more profound analysis of the data, it follows that the prototypical
schematic source structure for STUPID is DEFICIENT HEAD/BRAIN (compare (12)).
More than the other source structures this schematic concept hides a complex
network of multiple interacting conceptual relationships. Correspondingly, this
conceptual source domain displays the highest instantiation rate.
With regard to a structural comparison between metaphorically and
metonymically organized conceptual hierarchies, it seems that both systems
have major structural characteristics in common. Although I am aware of the
different conceptual strategies which are involved in both mechanisms (cf.
Dirven 1993), Figure 4 clearly shows that generalized metonymy, just like its
metaphoric counterpart, operates on different levels of schematicity (‘first
similarity feature’). At the same time this broader focus allows one to identify
more easily the patterns of interplay and overlap between different metony-
mies in the hierarchy (‘second similarity feature’). In this respect I have briefly
discussed some interesting examples (compare (5), (8), (9), (10)) that profile
stupidity on the basis of a complex source domain. Such patterns of overlap
can be described as a complex structure situated somewhere in conceptual
space, defined by different relationships with other conceptual structures and
activated as (part of) the source domain matrix of the profiled concept.
The examples cited above display two possible patterns of overlap be-
tween concepts/metonymies. The first one resides in the elaboration of two
schematic structures, whereas the second one consists of a metonymic asso-
ciation with another concept. With this description, I apply Langacker’s
(1987) view on schematic networks as the organizing principle of our knowl-
edge structures. Accordingly, the conceptual background (domain matrix) of a
profile requires “specifications in numerous domains” (ibid.: 163). All these
specifications relate in one way or another to the conceptual entity which is

profiled by the expression.
The entity designated by a symbolic unit [its profile; K.F.] can therefore be
thought of as a point of access to a network. The semantic value of a symbolic
unit is given by the open-ended set of relations — simple and complex, direct
and indirect — in which this access node participates. (ibid.)
Not all domains/concepts which constitute the conceptual background of a

profile are activated with the same degree of intensity (cf. also Croft 1993:
344f). Although in this context I did not deal with that issue explicitly, I
mentioned that some concepts which are metonymically related to the source
structure may be less directly involved in the definition of the profiled concept
STUPID. Compare, for example, the association of a concept like MARKED AGE
(CHILD) with the conceptual structure which is directly instantiated by the
expression in (5). A similar observation holds for the association of a low
social class with the central aspects of the Mantafahrer-stereotype.
As to the ‘third similarity feature,’ finally, I have demonstrated that
specific, low-level metonymies like the German Mantafahrer-model tend to
be culturally more specific with a correspondingly restricted use. Higher-level
metonymic mappings on the other hand are more widespread in their use
because of their general cognitive, culturally unrestricted nature.
6. Conclusion
Conceived in a cognitive semantic framework, this article raises two major

points concerning the nature and structure of metonymy. First, it is argued that
on both a theoretical and methodological level, the notion ‘domain matrix’ is
not precise enough to be used as the primary criterion for the distinction
between metaphor and metonymy. The essence of metonymy resides in the
nature of the relationship that connects two concepts with each other. Accord-
ingly, I have defined metonymy in terms of a conceptualized contiguity
relationship between two entities. Hence, metonymy is categorized primarily
as a conceptual rather than a strictly referential phenomenon. Consequently,
the main interest of this contribution is to be situated on the level of conceptual
structure. In this perspective, the analysis of the data demonstrates the crucial
role metonymy plays in structuring our knowledge system. From the examples
provided here, it follows that metonymy represents a sophisticated and robust
330 Kurt Feyaerts
conceptual mechanism which must be treated on a par with metaphor.

It was also shown, second, that metonymically motivated idioms may
cluster into higher-level conceptual metonymies in a similar way that meta-
phoric expressions do. This comparison has led to the conclusion that both
types of conceptual organization have three major structural characteristics in
common. The extrapolation of the notion of multi-level hierarchies of concep-
tual patterns from the study of metaphors to the study of metonymies repre-
sents the specific theoretical contribution of this paper to the cognitive
semantic study of metonymy.
Notes
1. I wish to thank Dirk Geeraerts for valuable help with the preparation of the talk on which
this article is based, Günter Radden for helpful comments on an earlier version of this
paper and Jessica Dobratz for proofreading major parts of the text. This article presup-
poses a certain acquaintance with basic theses and terminology of cognitive semantics
and of the Lakovian theory of metaphor in particular.
2. Compare Kövecses (1989: 44): “each and every expression related to a concept has to be
examined if we wish to uncover the minute details of the concept,” but see also Geeraerts
and Grondelaers (1995: 174ff).
3. In this respect I refer to Langacker’s (1993: 30) characterization of metonymy in terms of
reference points: “[...] a well-chosen metonymic expression lets us mention one entity
that is salient and easily coded, and thereby evoke — essentially automatically — a target
that is either of lesser interest or harder to name. [...] Other things being equal, various
principles of relative salience generally hold: human > non-human; whole > part;
concrete > abstract; visible > non-visible; etc.”
4. The cognitive basis of person perception makes it a discipline of great interest for
cognitive semanticists: “In dealing with person perception we are concerned with it as it
occurs in the layman, as a process affecting ordinary human interaction, and not with the
person as an object to be diagnosed by the expert” (Tagiuri 1958: xii).
5. Cf. Heider (1988: 341): “Our main concern is not the exact measure of one’s own ability,
but right away the relation to others. If somebody would tell us exactly what our ability
measures, we still would ask: but how do others do?”
6. Cf. Lakoff (1993) and Kövecses (1995a) for recent studies in this respect.
7. Several examples illustrate that LIFE IS A JOURNEY also elaborates the object dual of both
schematic metaphors which are involved here. Compare, for instance, expressions like
He lost his life (CHANGE OF STATE IS MOVEMENT OF POSSESSED OBJECT) or His youth has
abandoned him (TIME PASSING IS MOVING ENTITY).
8. See also Kövecses (1995b: 193), who distinguishes between a generic, mainly physiologi-
cally motivated conceptual level and a specific, mainly culturally determined conceptual
level.
9. Argumentation theory provides a good basis for an adequate description of the concep-
tual relationships underlying these expressions. Crucially, the kind of argumentation (or
conceptual linking) involved here is situated on a rhetorical rather than a strictly logical
level.
10. Obviously, as this theoretical perspective is inherent in any linguistic research, no
linguistic analysis can ever claim to reflect the mental richness of a conceptual structure
in full detail.
11. I refer to Sandra and Rice (1995), where the issue of the psychological reality of
theoretical notions such as ‘domain’ is discussed in greater detail.
12. According to Dirven (1993: 18), figurative meaning “only arises — or at least can arise
— if the conceptual distance between the two (sub)domains or things referred to is large
enough.”
13. In order not to complicate the picture too much, not every relationship between source
and target or between one source concept and another can be explicated in one single
figure. For example, the contiguity relationship and potential metonymy (CAUSE/EFFECT)
among the source concepts DEFICIENT SENSORY APPARATUS, DEFICIENT HEAD/BRAIN and
DEFICIENT ( PRACTICAL) ABILITIES is not represented in this schema.
References
Bruner, J.S., D. Shapiro, R. Tagiuri

1958 The meaning of traits in isolation and combination. In R. Tagiuri, L.
Petrullo (eds.), Person Perception and Interpersonal Behavior. Stanford:
Stanford University Press, 277–288.
Croft, William
Davies, C.
1987 Language, identity and ethnic jokes about stupidity. International Journal
of the Sociology of Language 65: 39–52.
1990 Ethnic Humor around the World: A Comparative Analysis. Bloomington:
Indiana University Press.
Dirven, René
Leuvense Bijdragen 82, 1–28.
Eemeren, F.H. van, R. Grootendorst, F. Snoeck-Henkemans
1996 Fundamentals of Argumentation Theory: A Handbook of Historical Back-
grounds and Contemporary Developments. Mahwah, NJ: Lawrence
Erlbaum.
Geeraerts, Dirk
1994 Nondenotational semantic change. In R.E. Asher, J.M.Y. Simpson (eds.),
The Encyclopedia of Language and Linguistics. Vol. 5. Oxford: Pergamon
Press, 2823–2824.
332 Kurt Feyaerts
Geeraerts, Dirk, S. Grondelaers

1995 Looking back at anger: cultural traditions and metaphorical patterns. In J.
Taylor, R.E. MacLaury (eds.), Language and the Cognitive Construal of
the World. Berlin: Mouton de Gruyter, 153–179.
Geiger, R.A.
1993 Metacognitive aspects of reference: assessing referential correctness and
success. In R.A. Geiger, B. Rudzka-Ostyn (eds.), Conceptualizations and
Mental Processing in Language. Berlin: Mouton de Gruyter, 267–289.
Goossens, Louis
for linguistic action. In L. Goossens et al. (eds.), By Word of Mouth:
Metaphor, Metonymy, and Linguistic Action in a Cognitive Perspective.
Heider, F., M. Benesh-Weiner (ed.)
1988 The Notebooks. Vol. 2: Perception. München: Psychologie Verlags Union.
Kövecses, Zoltán
1989 Emotion Concepts. New York: Springer.
1995a American friendship and the scope of metaphor. Cognitive Linguistics 6:
315–346.
1995b Anger: its language, conceptualization, and physiology in the light of cross-
cultural evidence. In J. Taylor, R.E. MacLaury (eds.), Language and the
Cognitive Construal of the World. Berlin: Mouton de Gruyter, 181–196.
Lakoff, George
1993 The contemporary theory of metaphor. In A. Ortony (ed.), Metaphor and
Thought. Second edition. Cambridge: Cambridge University Press, 202-251.
1987 Foundations of Cognitive Grammar. Vol. I: Theoretical Prerequisites.
Lyons, John
1977 Semantics. Cambridge: Cambridge University Press.
Sandra, Dominiek, Sally Rice
1995 Network analyses of prepositional meaning: mirroring whose mind — the
linguist’s or the language user’s? Cognitive Linguistics 6: 89–130.
Schmid, Hans-Jörg
1993 Cottage und Co., idea, start vs. begin: Die Kategorisierung als Grundprinzip
einer differenzierten Bedeutungsbeschreibung. Tübingen: Niemeyer.
Tagiuri, Renato
1958 Introduction. In R. Tagiuri, L. Petrullo (eds.), Person Perception and
Interpersonal Behavior. Stanford: Stanford University Press.
Taylor, John
Clarendon Press.
The Potentiality for Actuality Metonymy
in English and Hungarian
Klaus-Uwe Panther
Linda Thornburg
Eötvös Loránd University, Budapest
1. Introduction
In a recent paper (Thornburg and Panther 1997) we postulated and exempli-

fied some general metonymies used by speakers of English, e.g., the metony-
mies FORM FOR CONTENT, EFFECT FOR CAUSE, POTENTIALITY FOR MOTIVATION,
NECESSITY FOR MOTIVATION and POTENTIALITY FOR ACTUALITY, the latter being
one of the most pervasive. A natural extension of this research is to investigate
the function and exploitation of metonymies across languages. In this paper,
we single out the POTENTIALITY FOR ACTUALITY metonymy and explore how it
functions in two genetically and typologically unrelated languages, English
and Hungarian.
1.1. Metonymy
Metonymy is traditionally regarded as a figure of speech that involves a

process of substituting one linguistic expression for another, i.e., metonymy is
viewed as a relation in which one linguistic expression ‘stands for’ another.
The best-known cases of metonymy in this traditional sense are expressions
that are used for the purpose of indirect referring. For example, there is a
convention that the referential noun phrase
334 Klaus-Uwe Panther and Linda Thornburg
(1) the White House

can be used to refer to the executive branch of the government of the United
States, a spokesperson for that branch, or even the President himself but is not
synonymous with any of these.
More recently, with the advent of cognitive linguistics, it has been
recognized that the traditional view of metonymy is too narrow and that
metonymy, like metaphor, is a conceptual tool that operates within ‘idealized
cognitive models’ or ‘domain matrices’ (e.g., Lakoff 1987; Croft 1993;
Kövecses and Radden 1998; Radden and Kövecses, this volume). In this view
metonymy is “[…] a cognitive process in which one conceptual entity, the
vehicle, provides mental access to another conceptual entity, the target, within
the same idealized cognitive model” (Radden and Kövecses, this volume).
The cognitive approach to metonymy is not necessarily incompatible
with the traditional view. At some level the vehicle and the target are equiva-
lent, although they are not semantically synonymous. On the one hand, the
metonymic substitution of one expression for another creates a relation of
pragmatic equivalence between the substituting and the substituted expres-
sion; on the other hand, in accordance with Radden and Kövecses (this
volume), we assume that both the vehicle and the target are conceptually
present when a metonymy is used. We thus consider metonymy to be both a
property of conceptual structure and of language use, with varying degrees of
conventionalization in the linguistic system.
In this chapter, we will focus on metonymic phenomena that have a non-
referential function. By way of example, consider the propositions expressed
in the assertions in (2a) and (2b):
(2) a. She was able to finish her dissertation.
b. She finished her dissertation.
Just as the noun phrase vehicle in (1) does not have the same meaning as any
of its metonymic targets, sentences (2a) and (2b) are not semantically synony-
mous. It is possible to assert (2a) and to deny (2b) without contradiction. Yet
on many occasions of use the speaker of (2a) pragmatically conveys the same
propositional content as that expressed in (2b). In this sense the statement (2a)
can be used to ‘stand for’ the statement (2b), the only difference being that in
the first case the speaker predicates the ability to finish the dissertation of the
subject she, whereas in the second case the speaker predicates the actuality of
finishing it. In pragmatic terms, (2b) is a (generalized) conversational
implicature induced by (2a) (see Grice 1975). In contrast to the referential

metonymy in (1), statement (2a) is an instance of what we call ‘predicational
metonymy’ and exemplifies the POTENTIALITY FOR ACTUALITY metonymy.
A second type of non-referential metonymy is found in the domain of
communicative actions. We call this type of metonymy ‘speech act me-
tonymy’ wherein one illocutionary act stands for another illocutionary act, as,
e.g., in the statement or assertion
(3) a. I don’t know where the bath soap is.1
which may metonymically stand for the question or inquiry
(3) b. Where is the bath soap?
Utterance (3a) has the direct illocutionary force of an assertion about what the
speaker does not know, but in many contexts it is used with the indirect
illocutionary force of a question (for the identification of speech acts in
conversation, cf. Panther and Thornburg 1998). Searle (1975) hypothesizes
that indirect speech acts involve the performance of two illocutionary acts: the
secondary illocutionary act relies on the literal force of the utterance whereas
the primary illocutionary act corresponds to its intended force, in the case (3a)
that of an inquiry. Searle’s claim that two illocutionary acts are performed in
an indirect speech act is just another way of saying that both the metonymic
vehicle (the secondary illocutionary act) and the target (the primary illocution-
ary act) are conceptually present in the speaker’s mind.
In conclusion then, it is obvious that the phenomenon of metonymy is not
restricted to its well-known referring function but is much more pervasive in
ordinary language use. In addition to referential metonymies, at least two
more types of metonymic relations exist and can be shown to be linguistically
relevant. To summarize, we propose the following pragmatic typology of
metonymic functions:
(4) a. propositional metonymies:
(i) referential,
(ii) predicational;
b. illocutionary metonymies.
All of these metonymic functions may co-occur in a single utterance, as in (5):
(5) I don’t know whether the first violin was able to pass her driver’s
test.
which contains the metonymic types presented in (6):

(6) a. propositional:
(i) referential: the noun phrase the first violin may stand for the
person in an orchestra playing the first violin;
(ii) predicational: the propositional content that The speaker
does not know whether the first violin was able to pass her
driver’s test may stand for the propositional content that
The speaker does not know whether the first violin passed
her driver’s test;
b. illocutionary:
the assertion I don’t know whether the first violin was able to
pass her driver’s test may stand for the question Was the first
violin able to pass her driver’s test?
In this paper, we concentrate on the operation of the POTENTIALITY FOR
ACTUALITY metonymy in the predicational and the illocutionary domains
across two languages.2
1.2. The scenario approach to metonymy
In accordance with recent cognitive approaches to semantics, we assume that

meaning, including propositional and illocutionary meaning, is fruitfully ana-
lyzed in terms of scenarios consisting of parts which can bear metonymic
relations to each other and the whole of the scenario (cf. Blank, this volume).3
Applying this part-whole analysis to speech acts in our prior paper (Thornburg
and Panther 1997; cf. also Searle 1969, 1975) we proposed, among other
things, the scenario for directive speech acts given in (7):
(7) Scenario for Directive Speech Acts
(i) the BEFORE: H can do A.
S wants H to do A.
(ii) the CORE: S puts H under a (more or less strong) obligation
to do A.
the RESULT: H is under an obligation to do A (H must/should/
ought to do A).
(iii) the AFTER: H will do A.
The scenario approach allowed us to show two things: (1) that indirect speech
acts (and related phenomena) yield to an explanation in terms of general

metonymies and (2) that their linguistic forms are highly motivated. To
illustrate, an indirect request such as that given in example (8):
(8) John, you will take out the garbage.
instantiates a ‘part’ of the Directive Scenario, namely the AFTER condition,
which metonymically stands for the CORE of the requestive speech act scenario
or the scenario as a ‘whole.’ Moreover, the form of (8) is motivated since the
prospective future action of the hearer is an integral component of that
scenario.
1.3. The state-of-affairs scenario
From the scenario given in (7) for the specific illocutionary type ‘directives’
we can abstract a more general State-of-Affairs Scenario. Speech acts are just
special instances of what we call ‘states of affairs,’ e.g., states, events, pro-
cesses, and actions, which, of course, include linguistic actions. We assume
that the distinction between BEFORE conditions, the CORE, the immediate
RESULT or EFFECT of the core, and the AFTER, which we have already posited for
the Directive Speech Act Scenario in (7), also holds for states of affairs. Thus,
we propose the scenario given in (9) for propositional contents that describe
existing (actual) states of affairs.
(9) The State-of-Affairs (SoA) Scenario
(i) the BEFORE: necessary preconditions: motivations, potentiali-
ties, capabilities, abilities, dispositions, etc., which
can bring about the SoA
(ii) the CORE: the existing/true SoA
the EFFECTS: necessary consequences immediately following
from the SoA
(iii) the AFTER: non-necessary consequences of the SoA
As an example of a state of affairs, consider a description of the act of opening
a window, such as in (10):
(10) Mary opened the window.
A necessary pre-condition (a BEFORE condition) for the existence of the state of
affairs described in sentence (10) is Mary’s ability to bring it about. An even
more basic BEFORE condition would be that it is at all possible (for anyone) to
open the window. Other BEFORE conditions which might be regarded as
specific subcases of the ability condition would be, e.g., that Mary has the
strength, the time, the patience, etc. to open the window (see Panther 1981). If
Mary is not able to open the window or is not strong enough or does not have
the time to open the window, then the window-opening simply does not come
about. Certain mental states also count as BEFORE conditions in our account,
e.g., the desire or intention to open the window, which can be regarded as
motivations for the action. The CORE of the state of affairs is the actual
performance of the action. The immediate EFFECT of a successful act of
opening the window is another state of affairs, that is, that the window is open.
An intended but not necessarily following consequence of opening the win-
dow (the AFTER) might be that fresh air flows through the window. An
unintended consequence might be that flies come through the open window.
In general, the AFTER is only loosely associated with the CORE, whereas the
EFFECT is a necessary part of it.4
The SoA Scenario can be metonymically exploited just like the more
specific Scenario for Directive Speech Acts. For example, in answering the
question (11a) with (11b) the answerer B may exploit the metonymy EFFECT
FOR CORE :
(11) a. A: Did you open the window?

b. B: The window is open.
Thus, in some contexts, interlocutor B might wish to convey that she actually
has opened the window although, literally, she only states the result of the
window-opening.
2. The POTENTIALITY FOR ACTUALITY metonymy in English and

Hungarian5
In Section 1, we exemplified a metonymic principle that is based on the

conceptual contiguity of the modality of possibility and its actualization. It is
an empirical question whether and to what extent the POTENTIALITY FOR
ACTUALITY metonymy is operative cross-linguistically, possibly as a universal
cognitive principle. As a first step towards answering this question we under-
took a comparative analysis of two genetically and typologically unrelated
languages, English and Hungarian. We investigated the linguistic relevance

and distribution of the POTENTIALITY FOR ACTUALITY metonymy in seven
conceptual and communicative domains: perceptual events, mental states and
processes, hedged performatives, indirect speech acts, (extra-linguistic) ac-
tions, character dispositions, and acquired skills. The analysis revealed inter-
esting contrasts between the two languages in the exploitation of the
POTENTIALITY FOR ACTUALITY metonymy. As will be demonstrated below, this
metonymy is used much more systematically in English than in Hungarian.
2.1. The sense perception metonymy (ABILITY TO PERCEIVE FOR ACTUAL

PERCEPTION)
The first conceptual domain that we analyze concerns perceptual events like
seeing, hearing, feeling, tasting, and smelling. All the English sentences from
(12)–(20), apart from examples (15) and (16), are about actually occurring
sense perceptions although, literally, they either question or assert what we
call the BEFORE condition of the Sense Perception Scenario, i.e., they question
or assert the ability for the particular sense perception. For example, in (12)
and (13) the speaker wants to know whether the hearer actually sees someone
or something. In (17) the speaker actually heard the sneering laughter. In (18)
the gambler actually has a sense of being lucky. In (19) and (20) the taste and
smell are actually experienced. In each of these English examples the ability
for the sense perception ‘stands for’ the actual perceiving. Thus the POTENTI-
ALITY FOR ACTUALITY metonymy is fully exploited for the expression of sense
perceptions in English.6
(a) see
(12) Can you see him?
Látod? (Keresztes 1992: 34)
see-2.SG.PRES.IND.DEF.
(13) Can you see well?
Jól látsz? (Keresztes 1992: 34)
well see-2.SG.PRES.IND.INDEF.
(14) I can’t see the movie screen while you have that hat on.
(Searle 1975)
Nem látom a mozivásznat a kalapodtól.
not see-1.SG.PRES.IND.DEF. the moviescreen-ACC. the hat-
2.SG.POSSESS.-FROM/WITH
(15) The whole town can be seen from the window.

Az ablakból az egész város látható. (Keresztes 1992: 106)
the window-FROM the whole town see-POSS.-PART. (‘…is available
for viewing’)
(16) It can be seen that…
Látható, hogy… (Keresztes 1992: 106)
see-POSS.-PART that (‘(It) is available for seeing that…’)
(b) hear
(17) I could hear his sneering laughter as her arms carried me off
through the fire of oblivion. I can hear it yet. (LOB)7
…hallottam… …hallom.
…hear-1.SG.PAST.IND.DEF... …hear-1.SG.PRES.IND.DEF.
(c) feel
(18) I’m on the right streak tonight, I can feel it. (LOB)
Ma este jó úton járok, érzem.
...feel-1.SG.PRES.IND.DEF.
(d) taste
(19) I can taste the vanilla.
Érzem a vanília ízét.
feel-1.SG.PRES.IND.DEF. the vanilla taste-ACC.
(e) smell
(20) I can smell the garlic.
Érzem a fokhagyma szagát.
feel-1.SG.PRES.IND.DEF. the garlic smell-ACC.
Although the POTENTIALITY FOR ACTUALITY metonymy is available in Hungar-
ian, it is not exploited for the expression of sense perceptions. All of our
informants agreed that the available modality markers would not occur with
these sense perception verbs in the given contexts. Consider sentences (12)–
(14): the verb lát ‘see’ is grammatically marked for person, number, tense,
mood and its conjugation type. The markers for the modalities of possibility or
ability, i.e., the affix -hat-/-het- or the verb tud, do not occur in expressions of
actual sense perception events. At first sight, examples (15) and (16) appear to
be counterexamples to what we have just claimed. The verb lát is affixed by

the ‘possibility’ morpheme -hat-; however, the verb appears in a participial
form without a human subject who actually experiences the sense perception.8
Sentences (15) and (16) are thus not metonymic; rather, they focus on the
ability or possibility of something being available to vision without met-
onymically implying that an actual event of seeing has occurred.
The sense perception verbs in (17) and (18) — Hungarian hall meaning
‘hear’ in (17) and érez meaning ‘feel’ in (18) — behave analogously to lát
‘see.’ Only slightly different are the sense perceptions of smelling and tasting
exemplified in (19) and (20), which are expressed with nominal forms for the
concepts ‘smell’ and ‘taste’ plus the verb érez ‘feel.’ Nevertheless, the verb
érez occurs without a modality marker.
To summarize these observations then, the POTENTIALITY FOR ACTUALITY
metonymy is systematically exploited for the expression of sense perceptions
in English, even conventionalized. For Hungarian, however, our findings
indicate the opposite, namely that the POTENTIALITY FOR ACTUALITY metonymy
appears to be excluded in this conceptual domain, suggesting an important
typological difference between the two languages. At this point we can only
state this contrast as an interesting linguistic fact about English and Hungarian
without venturing into any speculations about the reasons for this difference.
We now go on to examine predications involving mental processes and states,
which exhibit a similar contrast between the two languages.
2.2. The mental process/state metonymy (ABILITY TO PROCESS FOR ACTUAL

MENTAL PROCESS)
In the conceptual domain of mental processes and states, English again fully
exploits the POTENTIALITY FOR ACTUALITY metonymy, as can be seen in sen-
tences (21)–(24). The second sentence in (21) strongly implies that the
speaker is actually in a state of understanding. In (22) the speaker actually
remembers the past event. In (23) the speaker conveys his actual belief that
Steve is guilty and in (24) he actually imagines how it happened.
(a) see (= ‘understand’)
(21) I can’t let her down just like that, yet one day it will have to come.
I can see that now. (LOB)
…Most már látom.
…now already see-1.SG.PRES.IND.DEF.
(b) remember
(22) I can remember when we got our first TV.
Emlékszem, mikor…
remember-1.SG.PRES.IND.DEF. when…
(c) believe
(23) Mary can’t believe that Steve is guilty, but I can.
Mari nem hiszi, hogy Pista b²
unös, de én igen.
Mary not believe-3.SG.PRES.IND.DEF. that…
(d) imagine
(24) I can imagine how it happened.
El tudom képzelni, hogyan…
PRT able-1.SG.PRES .IND.DEF. imagine-INF . how…
For the limited data we have for Hungarian the situation is less clear. Although
the Hungarian sentences in (21) and (22) have no modality markers, (24) does
contain the modal verb tud ‘be able’ in the Present Indicative. Sentence (24)
thus instantiates the POTENTIALITY FOR ACTUALITY metonymy. Example (23) is
unclear because in the English sentence can is negated. Our Hungarian
informants opted for a translation without a modality marker but they also told
us that reversing the sequence of the clauses to I can believe that Steve is
guilty, but Mary can’t would allow the modality affix -het- ‘possible’ in the
relevant metonymic sense.
To conclude this section of our analysis regarding the mental process/state
verbs, the English examples strongly suggest that the POTENTIALITY FOR ACTU-
ALITY metonymy is again fully exploited to render the pragmatic effect of
actuality. However, for Hungarian, despite the fact that the modality markers
co-occur with some verbs of mental states/processes and apparently have the
relevant metonymic function, the POTENTIALITY FOR ACTUALITY metonymy
seems to operate in this conceptual domain only to a limited extent. We now turn
to what we call speech act metonymies; the first type we consider is ‘hedged
performatives,’ i.e., explicit performative utterances that are hedged by modal
or modal-like expression (cf. Fraser 1975; Bach and Harnish 1979).
2.3. Speech act metonymies
2.3.1. Hedged performatives

It is common in English to perform an illocutionary act by combining a
performative verb with modal ‘hedges’ such as can, must, would like to, etc.
Thus it is possible to actually assure someone that she will be paid by saying I
can assure you that you will be paid. It is possible to actually invite someone
to a party by saying I would like to invite you to my party. It is possible to
actually ask someone to leave by saying I must ask you to leave.9 The
existence and pervasiveness of hedged performatives in English and other
languages suggests an even more general metonymic principle than the POTEN-
TIALITY FOR ACTUALITY metonymy, which might be called the VIRTUALITY FOR
ACTUALITY metonymy, where ‘virtuality’ includes other modalities apart from
potentiality, e.g., necessity, obligation, and permission, and mental attitudes
such as desires or wishes.
We hypothesize that the class of performative verbs that are sensitive to
the POTENTIALITY FOR ACTUALITY metonymy usually denotes positive and/or
non-face-threatening illocutionary acts. Thus promising is usually regarded as
a positive speech act, whereas asking is a potentially face-threatening act.
Within the subclass of ‘positive’ performative verbs in English the POTENTIAL-
ITY FOR ACTUALITY metonymy is systematically exploited as can be seen in
(25)–(28).
(a) testify
(25) I can testify that, seen from the surrounding heights, it is a fairy-
land of lights… (LOB)
Tanúsíthatom, hogy a környez² o magaslatokról nézve…
testify-POSS.-1.SG.PRES.IND.DEF. that…
(b) recommend
(26) I can recommend this type of cage, as it is impossible for the birds
to throw out seed husks… (LOB)
Ajánlhatom ezt a fajta kalitkát…
recommend-POSS.-1.SG.PRES.IND.DEF. this-ACC. the type cage…
(c) give one’s word

(27) I can give you my word that he is not at home.
Szavamat adhatom, hogy nincs otthon.
word-1.SG.POSSESS.-ACC. give-POSS.-1.SG.PRES.IND.DEF. that…
(d) promise
(28) I can promise you I’ll be home.
Megígérhetem neked, hogy otthon leszek.
make-a-promise-POSS.-1.SG.PRES.IND.DEF. to-you…
In Hungarian the POTENTIALITY FOR ACTUALITY metonymy is also systemati-
cally exploited for hedged performatives. Performative utterances can be
realized in Hungarian, as in English, either by using the performative verb in
the 1st Person Present Indicative or by hedging the performative verb with a
modality morpheme. Thus the illocutionary act of testifying can be carried out
by using the form tanúsíthatom (as in (25)); it is possible to achieve the speech
act of recommending by using the modalized verb form ajánlhatom (as in
(26)); in the same vein, a speaker can felicitously give her word by saying
szavamat adhatom (as in (27)) or accomplish the speech act of promising by
uttering megígérhetem (as in (28)).
To summarize, in both English and Hungarian the ‘actuality effect’ of the
POTENTIALITY FOR ACTUALITY metonymy is very strong with hedged performa-
tives.
2.3.2. Indirect speech acts

We now turn to the pragmatic phenomenon which is usually referred to as
‘indirect speech acts’ (cf. Searle 1975). In a recent paper (Thornburg and
Panther 1997) we analyzed indirect speech acts as exemplifying PART-WHOLE
metonymies in which part of the speech act scenario, for example a prepara-
tory condition for an offer, as in (29), stands for the whole scenario. Here we
undertake to reduce speech act metonymies to more fundamental metonymic
principles. We will present data for ‘commissives,’ ‘directives’ and ‘impreca-
tions’ to see to what degree they can be analyzed as instances of the POTENTI-
ALITY FOR ACTUALITY metonymy. It will be seen that some interesting asymmetries
arise among these different speech act types.
Commissives. In examples (29) and (30) the modal construction I can VP
instantiates its systematic use in English for conveying indirect offers, a type
of commissive speech act. Although these sentences do not immediately strike

one as cases of the POTENTIALITY FOR ACTUALITY metonymy, the I can may
stand for future actuality, i.e., I can VP metonymically may stand for I will VP.
How does this metonymy come about? We propose that those contexts in
which it is assumed that the speaker not only can do the action but also wants
to do the action, lead to the inference that she will do the action. In such
indirect commissives the metonymic link between can and will is fairly strong
and normally not cancelable, as we have indicated in parentheses in (29).
We still have not answered the question why (29) and (30) are commis-
sives. Our reasoning is: given that S can do A in a context where S wants to do
A leading to the inference that S will do A, three components (two BEFORE and
one AFTER component) of the Commissive Scenario are present, which to-
gether suffice to metonymically evoke the Commissive Scenario as a whole.
(29) John, I can make breakfast while you get dressed (#but I won’t).
Jani, én megcsinálhatom a reggelit,…
John, I PRT.-make-POSS.-1.SG.PRES.IND.DEF. the breakfast-ACC.…
(30) When you come to Budapest, I can be your translator.
…lehet-ek/nék a tolmácsod.
…will-be-POSS.-1.SG.PRES.IND./COND.
translator-your
The same analysis appears to apply to the Hungarian examples in (29) and
(30): the modality affix -hat-/-het- is systematically used for conveying indi-
rect offers.10 Moreover, in both languages, the future actuality effect (and
thereby the offer interpretation) is canceled when it is assumed that the
speaker’s want is absent. The absence of the speaker’s want leads to a literal
interpretation of ability. Thus there are contexts in which it is possible both in
English and Hungarian to have utterances of the following sort: Of course, I
can make breakfast, but I won’t (because you behaved so badly), where the
implicated future actuality is explicitly canceled by the speaker.
Directives. In examples (31)–(35) we see the nearly grammaticalized use
of the modal construction can you/you can VP for conveying indirect requests.
By ‘nearly grammaticalized’ we mean that it is still possible to cancel the
request interpretation in certain contexts (cf. Searle 1975). Interestingly, in
these indirect requests, the metonymic link between can and will, i.e., between
potentiality and future actuality, is much weaker than in the commissives just
analyzed. Why should this be so? Perhaps the reason lies in the fact that
utterances like (31)–(35) normally occur in a context in which the speaker

wants the hearer to carry out the action and assumes the hearer’s ability to do
so. But because the speaker cannot safely assume the hearer’s want to carry
out the action nor the ensuing future actuality of the action, the metonymic
link between potentiality and actuality in directives is much weaker than in
commissives. Yet given the fact that both hearer’s ability and speaker’s want
are present, two parts of the Request Scenario are supplied and license a
request interpretation.
(31) Can/could you pass the salt?
a. Ide tud-od/nád adni a sót?
to-here able-2.SG.PRES.IND./COND.DEF. the salt-ACC.
b. Ideadhatnád a sót.
pass-POSS-2.SG.PRES.COND.DEF.
(32) Can/could you please open the window?
a. Ki tudnád nyitni az ablakot?
out-PRT. able-2.SG.PRES.COND.DEF. open-INF. the window-ACC.
b. Kinyithatnád ...
PRT.-open-POSS.-2.SG. PRES.COND. DEF.…
(33) I’ll take a shower and you can/could make breakfast.

Én lezuhanyozom, te addig megcsinálhat-od/nád a reggelit.
…you until-then PRT.-make-POSS.-2.SG.PRES.IND./
COND.DEF. the breakfast-ACC.
(34) Mary, you can peel the potatoes; John, you can set the table.
Mari, te megpucolhatnád a krumplit, Jani, te megteríthetnéd az
asztalt.
Mary you PRT.-peel-POSS.-2.SG.PRES.COND.DEF. the potato-ACC.;
John, you PRT.-set-POSS.-2.SG.PRES.COND.DEF. the table-ACC.
(35) You can forget it.
Elfelejtheted.
PRT.-forget-POSS.-2.SG.PRES.IND.DEF.
The same analysis holds true for Hungarian but with one distinction: the
grammaticalization of the modality marker -hat-/-het- is much stronger in
Hungarian than that of can in English. This means that no literal ‘ability’
interpretation is available in Hungarian; in other words, the request interpreta-
tion is not cancelable. As a result, in examples (31b) and (32b), neither

question intonation nor question mark punctuation may occur.11 In examples
of this kind, the possibility marker -hat-/-het- has acquired the grammatical
function of signaling a requestive reading, which can neither be suspended nor
canceled — a clear case of grammaticalization. For an ability interpretation to
be available, the verb tud ‘be able, know how’ must be used, as in examples
(31a) and (32a). This does not mean, however, that a request interpretation is
not pragmatically possible with tud — it is, and it is cancelable.
Imprecations. Imprecations with you can are a highly conventionalized
kind of indirect speech act. Their imperative force is not cancelable: you
cannot say You can go to hell and insist on a literal interpretation of ability at
the same time. In both English and Hungarian the imperative and modal
constructions are in free variation as (36a, b) and (37a, b) exemplify. Since the
hearer’s want that the propositional content be realized can definitely be ruled
out in these utterances, the link between potentiality and actuality is even less
available with imprecations than with indirect requests.
In conclusion, the can and -hat-/-het- in these constructions have become
completely grammaticalized as markers of imprecation, just as -hat-/-het- has
become completely grammaticalized in Hungarian requests.
(36) You can go to hell.
a. Menj a pokolba. (‘Go to hell’)
go-2.SG.IMP. the hell-INTO
b. Elmehetsz a pokolba. (‘You can go to hell’)
away-go-POSS.-2.SG.PRES.IND.INDEF. the hell-into
(37) You can take this job and shove it.
a. Edd meg az állásodat. (‘Eat up your job’)
eat-2.SG.IMP. the job-2.SG.POSSESS.-ACC.
b. Megeheted az állásodat. (‘You can eat up your job’)
PRT.-eat-POSS.-2.SG.PRES.IND.DEF. the job-2.SG.POSSESS.-ACC.
2.4. Action metonymies (ABILITY TO ACT FOR ACTION)
We now turn to the domain of non-linguistic actions. The examples in (38)

and (39) are clear instances of the POTENTIALITY FOR ACTUALITY metonymy.
Quite regularly, the expression of abilities to act are used to stand for the
corresponding actions themselves. For the English and Hungarian sentences
in (38) where ability occurs in a past time frame, actuality is strongly impli-
cated, though in some contexts it is cancelable.
For the English and Hungarian sentences in (39), where ability occurs in
a future time frame, the metonymic link between potentiality and actuality is
less strong. In Hungarian the verb tud in (39a) focuses on the ability reading in
contrast to (39b), which according to one informant, focuses on the commit-
ment of the speaker to a future action. In other words, the actuality effect in
(39a) is very weak and easily cancelable, whereas in (39b) it is fairly strong
and therefore difficult to cancel (cf. ‘commissives’ in Section 2.3.2.). In fact
the Hungarian sentences in (39a, b) can be conjoined without an effect of
redundancy. The pragmatic sense of the resulting conjunct is paraphraseable
as I can come to your party and I will.
(38) John was able to finish his paper before the deadline.
John a határid²
o el²
ott be tudta fejezni a dolgozatát.
John the deadline before PRT. able-3.SG.PAST.IND.DEF. finish-INF.
the work-ACC.
(39) I can come to your party on Friday.
a. El tudok jönni a pénteki bulidra.
PRT. able-1.SG.PRES .IND.INDEF. come-INF. the Friday-ADJ. party-
YOUR-TO
b. Eljöhetek…
PRT.-come-POSS.-1.SG.PRES.IND.INDEF.…
2.5. The character disposition metonymy (DISPOSITION FOR OCCASIONAL

BEHAVIOR)
The domain of ‘character disposition’ is different from other types of the

POTENTIALITY FOR ACTUALITY metonymy we have considered thus far in that
the relationship between vehicle and target is one between potential behavior
and occasional actualizations of this behavior. The sentences in (40)–(42) are
judgments about character traits that have an empirical basis. They can be
paraphrased by means of existential quantification over times or events, e.g.,
Dogs can be dangerous means that there are times when dogs display danger-
ous behavior. The metonymy is systematically exploited in both languages.
The alternate Hungarian forms in (40a, b) are reported to be synonymous, but
in more general use is the form in (40a), (41) and (42) with the modality verb
tud, the focus particle is, and the infinitival form of the verb lenni ‘to be.’ The
metonymic effect of occasional actuality is not cancelable, e.g., it is impos-
sible to assert at the same time that Dogs can be dangerous but have never
been observed to be dangerous.
(40) Dogs can be dangerous.
a. A kutyák veszélyesek is tudnak lenni.
the dog-PL. dangerous-PL. PRT. able-3.PL.PRES.COND.INDEF. be-
INF.
b. A kutyák veszélyesek lehetnek.
the dog-PL. dangerous-PL. will-be-POSS.-3.PL.PRES.COND.INDEF.
(41) He can be very unfriendly.
Néha nagyon barátságtalan is tud lenni.
sometimes very unfriendly PRT. able-3.SG.PRES.IND.INDEF. be-INF.
(42) She can be very generous.
Nagylelk²u is tud lenni.
very-generous PRT. able-3.SG.PRES.IND.INDEF. be-INF.
2.6. The acquired skills metonymy (SKILL FOR DISPLAY OF SKILL)
The last metonymy that we briefly explore in this paper and illustrate in (43)–
(45) concerns skills or acquired abilities. This metonymy is similar to the
character disposition metonymy in that the link between potentiality and
display of behavior is restricted to certain occasions. Since skills are acquired
and tend to be exercised, we expect to find in language use a strong met-
onymic link between the ability to exercise a skill and its actual use. For
example, if you can speak five languages then, by implicature, you will
probably actually speak them. In English this metonymy is systematically
exploited but seems to be only partially exploited in Hungarian.
(43) Can the baby walk now?
Jár már a baba?
walk-3.SG.PRES.IND.INDEF. the baby
(44) Mary can speak five languages.
Mari öt nyelven beszél.
Mary five language-ON speak-3.SG.PRES.IND.INDEF.
(45) I can swim fifty laps.

Tudok úszni 50 hosszt.
able-1.SG.PRES.IND.INDEF.…
3. Results
The results of our analysis are summarized in Tables 1 and 2. The tables
include the modality forms, information for each conceptual domain regard-
ing whether or not and to what extent the POTENTIALITY FOR ACTUALITY
metonymy is exploited, and how strong the metonymic link from potentiality
to actuality is. We describe the degree of exploitation as ranging from ‘sys-
tematic,’ ‘weakly exploited,’ ‘not exploited,’ to ‘not exploitable.’ The strength
of the metonymic link is also indicated in scalar terms: from ‘enforced,’ i.e.,
non-cancelable and grammaticalized; ‘strong,’ i.e., cancelable in some con-
texts; ‘fairly strong’; ‘weak’; to ‘zero.’ With regard to indirect speech acts, the
tables also present the degree to which the modal has been grammaticalized as
a marker of illocutionary force. Interesting contrasts in exploitation and de-
gree of metonymic link are highlighted in bold face.
Two major contrasts between English and Hungarian emerge from our
study. The most important difference can be seen in the conceptual domains of
sense perceptions and mental states/processes. First, the degree of exploitation
of the POTENTIALITY FOR ACTUALITY metonymy in these two domains is nearly
opposite: English exploits the metonymy systematically, whereas in Hungar-
ian it is either systematically blocked (not exploitable) or only weakly ex-
ploited. Second, the metonymic link between potentiality and actuality is very
strong in English, i.e., almost not cancelable, whereas in Hungarian, it is either
non-existent (zero) or only weakly present.
Another contrast, more subtle yet highly interesting, occurs in the domain
of indirect directives. In both languages statements or questions containing
modality markers for ability/potentiality are systematically used as requests.
However, an important difference between the two languages is that in Hun-
garian this practice has become grammaticalized, whereas in English there is
still the possibility of canceling the request reading (cf. Searle 1975). As we
have shown in Section 1.2. (also see Thornburg and Panther 1997), the
hearer’s ability to perform the requested action is a BEFORE condition in the
Request Scenario and can stand for the whole scenario or other parts of the
Table 1. Exploitation of the POTENTIALITY FOR ACTUALITY metonymy in English

CONCEPTUAL DOMAIN Linguistic Form Exploitation Metonymic Link
Sense Perception can systematic strong
Mental States/Processes can systematic strong
Hedged Performatives can systematic strong
Indirect Speech Acts
(i) Commissives I can / can I systematic actuality effect fairly
strong; commissive
reading strong
(ii) Directives you can / can you systematic actuality effect weak;
request reading
strong
(iii) Imprecations you can not exploited actuality effect non-
existent; imprecation
reading enforced
Actions able / can systematic strong
Character Dispositions can systematic enforced
Acquired Skills can systematic strong
Table 2. Exploitation of the POTENTIALITY FOR ACTUALITY metonymy in Hungarian

CONCEPTUAL DOMAIN Linguistic Form Exploitation Metonymic Link
Sense Perceptions -hat- / -het-, tud not zero
excluded exploitable
Mental States/Processes sometimes tud weakly weak
exploited
Hedged Performatives -hat- / -het- systematic strong
Indirect Speech Acts
(i) Commissives -hat- / -het- systematic actuality effect fairly
strong; commissive
reading strong
(ii) Directives -hat- / -het- systematic actuality effect weak;
request reading
enforced
(iii) Imprecations -hat- / -het- not exploited actuality effect non-
existent; imprecation
reading enforced
Actions tud (-hat- / -het-) systematic strength varies
Character Dispositions tud (-hat- / -het-) systematic enforced
Acquired Skills tud (-hat- / -het-) weakly strong
exploited
scenario such as the AFTER condition that the hearer will actually perform the
action. Notice, however, that the metonymic link between potentiality and
actuality for indirect directives is relatively weak. Unlike commissives, whose
satisfaction is guaranteed by the speaker, requested actions are hearer-depen-
dent, i.e., not under the speaker’s control, and therefore not guaranteed.
4. Conceptual metonymy and conversational maxims
In this section we will briefly touch upon the problem of how the POTENTIAL-
ITY FOR ACTUALITY metonymy relates to two well-known pragmatic principles,
namely Grice’s (1975) First and Second Maxims of Quantity: Make your
contribution as informative as is required (for the current purpose of the
exchange) and Do not make your contribution more informative than is
required. The POTENTIALITY FOR ACTUALITY metonymy seems to be in conflict
with the first maxim. The first maxim triggers the implicature that if a speaker
says that something is potential or possible then it is not actual. A speaker
saying for example
(46) Professor Smith may be in her office.
conveys that she cannot make the stronger statement
(47) Professor Smith is in her office.
In such cases the First Maxim of Quantity seems to take precedence over the
POTENTIALITY FOR ACTUALITY metonymy. In other words, the route from
potentiality to actuality is barred: may be definitely cannot stand for is because
the first quantity maxim triggers the implicature that the speaker is not certain
about the actuality of the proposition.
However, in a sentence like
(48) I could hear him rummaging around in the basement.
the POTENTIALITY FOR ACTUALITY metonymy licenses the interpretation ‘I heard
him rummaging around in the basement.’ This interpretation can also be
derived in a Gricean framework, namely, via the Second Maxim of Quantity
(cf. what Horn 1989: 194ff calls an R-based implicature). The problem of
which maxim takes precedence in cases of conflict is a complex issue that is
beyond the scope of this paper (cf. Horn 1989: Chs. 3ff for detailed discussion).
There are however cases where the First Maxim of Quantity and the
POTENTIALITY FOR ACTUALITY metonymy (or Grice’s Second Maxim of Quan-
tity) are not in conflict but seem to interact. Consider (49) (from Sweetser
1990: 70):
(49) He may be a university professor, but he sure is dumb.
Sweetser remarks that, despite the modality, the first clause states, or rather
concedes, that the subject is a professor.12 In our terminology, the POTENTIAL-
ITY FOR ACTUALITY metonymy is operative in example (49). Yet the first clause
of the utterance additionally conveys a reservation, which the non-modal
predicate is a university professor would not imply. That is, to make the
actuality statement
(50) He is a university professor,…
is to evoke stereotypical professorial attributes in the mind of the hearer, e.g.,
that professors are highly educated, intelligent, well-read, etc., which can then
be canceled by a subsequent but clause. However, in using may be, as in (49),
the speaker signals from the very start, within the initial clause, that some
stereotypical implicature(s) such as ‘he is intelligent’ or ‘he is well-read’ cannot
be anticipated. Thus, in contrast to the actuality predicate is a professor, the use
of may be a professor contravenes a possible conclusion that would be
warranted on the basis of the stereotypical assumptions held about professors.
In the first clause of (49) the POTENTIALITY FOR ACTUALITY metonymy sanctions
the conclusion that the subject actually is a professor, whereas the First Maxim
of Quantity licenses the inference that he is a professor to less than the fullest
extent, an interpretation which is confirmed by the but clause.
Likewise, in the conventionalized metaphorical use of see ‘understand’
the potentiality marker can in sentences like
(51) I can see your point.
seems to have the same dual function. On the one hand, the POTENTIALITY FOR
ACTUALITY metonymy yields the interpretation ‘I see your point,’ on the other
hand, since the stronger statement I see your point has not been made, the First
Maxim of Quantity licenses the implicature ‘…but I have reservations.’ In
uttering I can see your point the speaker gives to understand that she cannot
make the stronger statement I see your point because understanding to the
fullest possible extent has not occurred.
In examples like (49) and (51) the conflict between the First Maxim of
Quantity and the POTENTIALITY FOR ACTUALITY metonymy is resolved in a
compromise: actuality is conceded but not to the fullest extent. In general then,
utterances with a potentiality marker like can — even when they are used
metonymically to evoke actuality — seem to be weaker than their correspond-
ing non-modal counterparts.
To summarize, the relationship between conceptual metonymies and
Gricean implicatures is far from being clear at the present stage of research.
5. The pervasiveness of the POTENTIALITY FOR ACTUALITY metonymy
In recent work Kövecses and Radden (1998) and Radden and Kövecses (this
volume) have proposed that the choice of metonymic vehicles for accessing
various targets is not random. They postulate default metonymies that repre-
sent the normal, i.e., preferred, way of indicating an intended target. In their
framework the cognitive principle ACTUALITY OVER POTENTIALITY is a default
case whereby ‘actual’ vehicles are selected to access ‘potential’ targets rather
than the other way around. They interpret our data from indirect speech acts as
exceptional to their default principle since these data instantiate a metonymy
that links a ‘potential’ vehicle to an ‘actual’ target. They argue that these
‘exceptions’ result from a conventionalized flouting of their principle CLEAR
OVER LESS CLEAR, a flouting that they claim is motivated by social norms such
as politeness. In this way they preserve the primacy of their cognitive default
principle ACTUALITY OVER POTENTIALITY.
However, our research calls into question the validity of the proposed
default principle ACTUALITY OVER POTENTIALITY. As we have amply demon-
strated in this paper, in addition to indirect speech acts, at least six other
cognitive domains exist in which the POTENTIALITY FOR ACTUALITY metonymy
is systematically exploited in English and to varying degrees in Hungarian.
Unlike indirect speech acts, the tokens of POTENTIALITY FOR ACTUALITY in
these other domains cannot be accounted for in terms of social conventions, in
particular, politeness principles. The use of an utterance such as I can smell the
garlic over I smell the garlic is not motivated by rules of politeness. The
pervasiveness of the POTENTIALITY FOR ACTUALITY metonymy in both English
and Hungarian, its systematic exploitation, and its degree of conventionaliza-
tion below the level of conscious awareness pose a serious challenge to a
theory that posits such a default cognitive principle as ACTUALITY OVER

POTENTIALITY.
6. Conclusion
Our pilot study was motivated by a curiosity to see how conceptual metony-
mies are realized in two very different languages. By using a scenario ap-
proach to describe states of affairs and choosing one basic metonymic
principle, the POTENTIALITY FOR ACTUALITY metonymy, we were able to deter-
mine (i) that this particular conceptual principle is more systematically ex-
ploited in English than in Hungarian in various cognitive domains and (ii) that
it is pervasive in disparate cognitive domains, a finding that suggests it is a
default rather than a non-default cognitive principle.
We believe that future research should cover a larger sample of languages
in which additional conceptual metonymies and cognitive domains are ex-
plored. Moreover, such an extended research program should elucidate the
relationship between Gricean implicatures and the kind of metonymic infer-
ences discussed in this paper.
Notes
1. Note that in this case the propositional content also changes.

2. For reasons yet to be determined the POTENTIALITY FOR ACTUALITY seems to be excluded
from the referential domain.
3. One anonymous reviewer pointed out that our use of ‘part’ does not coincide with the
traditional sense of metonymy based on referential contiguity and that we confuse
conceptual contiguity (as a metalinguistic concept) with referential contiguity. We claim,
however, that even referential (spatial) contiguities, e.g., the set of body parts are
themselves conceptual constructs. As such they are also entities in conceptual space. Just
as a prototypical body is conceptualized as having a head, trunk, arms, legs, etc., we
propose that a prototypical directive speech act has the components listed in (7). In this
sense it seems perfectly legitimate to refer to the felicity conditions of a speech act as
‘parts’ of the speech act scenario.
4. In Thornburg and Panther (1997) these concepts are explained more fully, especially the
difference between the pragmatic result (effect) of a speech act and the result (effect) of a
non-linguistic action.
5. In collecting our data, we presented native speakers of Hungarian with English sentences
which instantiate the POTENTIALITY FOR ACTUALITY metonymy and asked them to give a
natural translation in Hungarian. With one exception — the sentence in (38) with be able
— the English data occur with the modal auxiliary can; the Hungarian translations
contain, where possible, either the affix -hat-/-het-, which denotes possibility in a general
sense, or the verb tud, which is roughly equivalent to ‘be able, know how.’ In particular,
we wanted to know whether the ‘possibility’ and ‘ability’ modality markers -hat-/-het-
and tud conveyed the same metonymic effect (i.e., implicating actuality) as in the English
examples. We would like to thank various friends, colleagues and students for their time
and assistance in providing most of the Hungarian data: Andrea Szirmai, Éva Szabó,
Balazs Lövenberg, Zoltán Kövecses, and especially Rita Brdar-Szabó, who supplied
many helpful grammatical points.
6. For the glosses of the Hungarian examples the following conventions will be used: words
are separated by a space; morpheme boundaries are indicated by a hyphen; grammatical
functions are indicated in small capitals. The abbreviations used are: ACC = accusative,
ADJ = adjectival suffix, COND = conditional, DEF = definite conjugation, IMP = imperative,
IND = indicative, INDEF = indefinite conjugation, PART = participle, PL = plural, POSS =
possibitity, POSSESS = possessive, PRES = present tense, PRT = particle, SG = singular.
7. The abbreviation LOB stands for Lancaster-Oslo-Bergen Corpus.
8. In examples (15) and (16) the -ható complex may be regarded as an adjective-forming
suffix deriving ‘possibility-passive’ adjectives from verbal bases. We are grateful to Rita
Brdar-Szabó for providing the arguments in support of this claim and regard this analysis
as additional support for ours.
9. The factors that determine whether a hedged performative can actually be used to
perform the illocutionary act denoted by the performative verb are complex and will not
be dealt with in any detail. Suffice it to say that I can ask you to leave does not constitute
an act of asking nor does I must promise to be there constitute a promise.
10. With indirect commissives (offers) like (30), our Hungarian language consultants often
preferred the more indirect conditional over the plainer indicative mood for reasons of
politeness. This preference also holds for indirect directives.
11. We are grateful to Rita Brdar-Szabó for making this point clear to us.
12. This is an example of what Sweetser calls ‘speech act modality.’ Unlike (46), utterance
(49) is not available as a topic-introducing or discourse-initiating device and seems to be
restricted to a reactive slot in a discourse.
References
Bach, Kent, Robert M. Harnish

1979 Linguistic Communication and Speech Acts. Cambridge, Massachusetts:
The MIT Press.
Croft, William
Fraser, Bruce
1975 Hedged performatives. In P. Cole, J. Morgan (eds.), Syntax and Semantics.
Grice, H. Paul
1975 Logic and conversation. In P. Cole, J. Morgan (eds.), Syntax and Semantics.
Horn, Laurence R.
1989 A Natural History of Negation. Chicago: The University of Chicago Press.
Keresztes, László
1992 A Practical Hungarian Grammar. Debrecen: Debreceni Nyári Egyetem.
Kövecses, Zoltán, Günter Radden
1998 Metonymy: developing a cognitive linguistic view. Cognitive Linguistics 9:
37–77.
Lakoff, George
Panther, Klaus-Uwe
1981 Indirect speech act markers or why some linguistic signs are non-arbitrary.
In R.A. Hendricks et al. (eds.), Papers from the Seventeenth Regional
Meeting. Chicago Linguistic Society, April 30 - May 1, 1981. Chicago:
Chicago Linguistic Society, 295–302.
Panther, Klaus-Uwe, Linda Thornburg
1998 A cognitive approach to inferencing in conversation. Journal of Pragmatics
30: 755–769.
Searle, John
1969 Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cam-
bridge University Press.
1975 Indirect speech acts. In P. Cole, J. Morgan (eds.), Syntax and Semantics.
Vol. 3: Speech Acts . New York: Academic Press, 59–82.
Sweetser, Eve
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of
Semantic Structure. Cambridge: Cambridge University Press.
Thornburg, Linda, Klaus-Uwe Panther
1997 Speech act metonymies. In W.A. Liebert et al. (eds.), Discourse and
Perspectives in Cognitive Linguistics. Amsterdam and Philadelphia:
Benjamins, 205–219.
Part IV
Applications of Metonymy
“Mummy, I like being a sandwich”
Metonymy in Language Acquisition
Brigitte Nerlich, David D. Clarke

University of Nottingham
Zazie Todd
University of Leicester
In order to recognise how meanings are ar-

ranged in our heads, the surest way is to see
how they got in there in the first place. (Bréal
1991 [1884]: 149)
1. Introduction
The first use of the term ‘metonymy’ can be found in ancient Greek philoso-
phy, and more specifically in the famous debate about the arbitrariness or
naturalness of signs. In his account of Plato’s contribution to linguistics, Fred
Householder (1995: 93) points out that
Democritus (as quoted in Proclus’ commentary on the Cratylus 16) offered
four arguments (with four specially coined names) in favour of arbitrariness:
(a) ‘homonymy’ or ‘polysemy,’ i.e., the same sequence of phonemes may be
associated with two or more unrelated meanings; (b) ‘polyonymy’ or ‘isorro-
phy,’ i.e., the existence of synonyms; (c) ‘metonymy,’ i.e., the fact that words
and meanings change; (d) ‘nonymy,’ i.e., the non-existence of single words
for simple or familiar ideas.
362 Brigitte Nerlich, David D. Clarke and Zazie Todd
In Chapter 21 of his Poetics Aristotle distinguished between four classes

of ‘metaphors,’ which included what was later to be called ‘metonymy’ and
‘synecdoche.’ As Peter Koch (this volume) and Andreas Blank (this volume)
have pointed out, the first real definition of metonymy was given in the
Rhetorica ad Herennium which is sometimes attributed to Cicero: “denomina-
tio est quae ab rebus propinquis et finitimis trahit orationem, qua possit
intellegi res quae non suo vocabulo sit appellata” (Anonymous [Cicero] 1954:
IV, 32, 43).
In classical rhetoric, metonymy became a figure of speech distinct from
metaphor. It normally excluded shifts of meaning based on PART-WHOLE
relations, which were attributed to synecdoche. Nowadays, shifts of meaning
based on PART-WHOLE relations are included in metonymy and are actually
regarded as the most basic metonymical relations. This was already stressed in
1925 by the French linguist Gaston Esnault. Esnault distinguished between
metaphor, metonymy and synecdoche on modern lines, now advocated for
example by Ken-ichi Seto (in press, and in this volume). However, in his
definition of metonymy Esnault (1925: 29) referred to Democritus. He wrote:
“The meaning of the word metaphor is transfer, that of metonymy change (of
name), and that of synecdoche annexation.” Thus, Esnault agreed with tradi-
tional rhetoric, but as a semanticist he pointed out:
Metaphor is not a transfer of a word from one sense to another, it is an
intuition that gets transported; metonymy is not a change of a name imposed
on a thing, it is an objective relation conceptualised in shortened form;
synecdoche is not a larger or smaller logical intension imposed on a word, it is
a change in its logical extension. (Esnault 1925: 29)
And he returned to this topic of ‘shortening’ when he remarked:

Metonymy doesn’t open up new paths to follow as metaphorical intuition
does; instead it hurries over the stages in paths that are too well known and
shortens the distances so as to facilitate the rapid intuition of things that we
already know. (ibid.: 31)
What is important here is that metonymy enables us to say things quicker, to

shorten conceptual distances. This function of metonymy as an ‘abbreviation
device’ has also been stressed by Warren (this volume). Metonymy is a
universal strategy of cost-effective communication. In young children we can
observe overextensions based on the metonymical stretching of the meaning
attached to the few word forms children have acquired (‘compelled met-
onymical overextensions’). Later on we can observe ‘creative metonymical
‘Mummy, I Like Being a Sandwich’ 363
shrinking’ despite the fact that the children could express the same meaning
with the words they know. Later still, in adults, we can observe the use of
metonymies in, literally speaking, cost-effective communication, as for ex-
ample in headlines like “Brains of British Museum crack Crystal Skull
Riddle” (The Observer 26/6/96). After England’s victory over Holland in the
Euro 96 football championship one could read: “Our Boys done Gouda” (The
Sun, June 1996). We understand these headlines instantly, and yet they require
a lot of linguistic, contextual and conceptual ‘unpacking.’
2. Typologies of metonymies
Gaston Esnault distinguished between 38 types of metonymy, some of which

are well-known from the work of Lakoff and Johnson (1980), some less so.
However, Esnault and Lakoff and Johnson were not the only ones who
established typologies of metonymies or classifications of types of metonymies.
In this century alone (and this means leaving out all that has been written in the
traditional treatises devoted to rhetoric), we have found at least nine taxonomies
which were developed in relative isolation: Nyrop (1913), Esnault (1925), Stern
(1931), Ullmann (1951), Norrick (1981), Sappan (1987), Yamanashi (1987),
Fass (1991), Radden and Kövecses (this volume) (additional typologies can be
found in Andreas Blank’s and Ken-ichi Seto’s contributions to this volume). It
was only recently, in 1991, that Dan Fass actually reviewed past analyses of
metonymy and compared the work done by Stern, Lakoff and Johnson, and
Yamanashi. However, he overlooked Norrick’s typology, which to us seems to
be the most complete and coherent classification of types of metonymies.
Norrick distinguished the following types of metonymy and correlated them
with 18 metonymic principles:
I. Cause - effect
1. Cause - effect
2. Producer - product
3. Natural source - natural product
4. Instrument - product
II. Acts and major participants
1. Object - act
2. Instrument - act
3. Agent - act
4. Agent - instrument
III. Part - whole

1. Part - whole
2. Act - complex act
3. Central factor - institution
IV. Container - content
1. Container - content
2. Locality - occupant
3. Costume - wearer
V. Experience - convention
1. Experience - convention
2. Manifestation - definition
VI. Possessor - possession
1. Possessor - possession
2. Office holder - office
To analyze and compare these various typologies would require a separate
paper. In what follows, we shall see how one can use these classifications for
the study of metonymy in language acquisition.
3. Compelled metonymical overextensions
Studies of how children produce and, especially, how they understand meta-
phor, proliferate in the literature on language acquisition in general and
semantic development in particular. However, no work has been done up to
now on metonymy in language acquisition, and the possibility of having
metonymic overextensions in child language has hardly ever been noted
before (cf., however, Melissa Bowerman’s work, e.g., Bowerman 1978;
Gibbs 1994).
Our first hypothesis was that metonymical relations might be exploited in
overextensions produced by children up to age 2;5. We called these ‘com-
pelled metonymical overextensions’ because they are based on the fact that at
this age a child’s vocabulary, category and conceptual systems are still rela-
tively small and unstructured. This scarcity compels them to extend already
known words to cope with increasing communicative needs, to comment on
what they see and to request what they want.
A good overview of the research done on over- and underextensions is
provided by Eve Clark in her book The Lexicon in Acquisition. She writes:
Children do not start out already knowing the meanings of the words they are
attempting to pronounce and use. They must first assign some meaning to
them, and that takes time. In fact, children’s earliest mappings of meanings
onto forms diverge from adult usage in a variety of ways. (Clark 1993: 32)
One of the most important ways of mapping the meaning a child wants to
express onto a form which he or she has already acquired is overextension,
that is, by applying a word to members of the adult category and to members
of other categories that are perceptually similar, conceptually contiguous or
spatio-temporally related. Consider these examples:
(1) a. ball for balls of all kinds, and also for round hanging lampshades,
doorknobs, and the moon [based on perceptual similarity]
b. door for corks, jar-lids, box-lids, and gates when wishing to
have the relevant object opened or closed [based on functional
similarity]
Eve Clark distinguishes between two types of overextensions: (i) ‘overin-
clusions,’ where children extend a term to other entities from the same
taxonomy, e.g.:
(2) a. dada used for both father and mother
b. baby used for self-reference and all children
c. apple for apples and for oranges
and (ii) ‘analogical extensions,’ where children use a term for objects from
other taxonomies on the basis of perceptual similarity, e.g.:
(3) a. cotty-bars for abacus on the wall and picture of a building with
a columned façade
b. comb for centipede
Whereas analogical overextensions are based on recognizing and construing
similarities (and can therefore be regarded as the training-ground for use of
metaphors), overinclusions are based on both perceptual similarity and con-
ceptual contiguity (we shall later call them synecdochical overextensions, and
they are the basis for the construction of taxonomies).
That is, the words being extended are applied to instances of other categories
within the same or an adjacent conceptual domain. For example, a child may
over-extend a term within the domain of animals, as when horse is applied to
goats, cows, and sheep or within the domain of vehicles, as when truck is
applied to buses, tractors, and vans; or within the domain of clothing, as when
hat is applied to a crown. (Clark 1993: 34)
However, it seems to us that Eve Clark should have discussed in more depth a
third type of overextension based not so much on conceptual contiguity, but
on spatio-temporal or functional contiguity. As we shall see, such contiguities
are based on visual perception and are therefore the precursors to actual
metonymies.
This type of overextension has been studied from Piaget and Vygotsky
onwards under the heading of ‘complexive associations.’ Those studying
complexive associations have claimed that object names initially encode
complexive groupings that incorporate actions or locations associated with
objects (cf. Huttenlocher and Smiley 1993: 240).
Even the words themselves are said to be initially just auditory features of
situations where they habitually occur [...]. In short, it has been suggested that
young children group their experiences in a fundamentally different way than
older children or adults — in terms of co-occurring aspects of situations —
and that object names, rather than standing for particular types of objects, are
just another type of associate. (ibid.: 222)
Huttenlocher and Smiley contest this claim and argue that the extension of
children’s early object words match those of adults to a large extent. However,
after distinguishing more clearly between a word’s extension or meaning and
its use, they found that in certain cases other than naming contexts, such as
comments or requests, such ‘overextensions’ did occur; as when a child said
toy when seeing a certain bag which habitually contained toys, or when a child
said shoe when she saw a foot (ibid.: 226f). This means that children use
certain ‘metonymically’ based pragmatic strategies, just as adults do, so as to
achieve certain communicative effects.
When a child comprehends a word correctly but overextends it in production,
Clark [1978] concludes the overextension is not the result of an immature or
nonadult word concept but is rather a deliberate attempt to communicate
concepts for which the appropriate word cannot be remembered. (Pease et al.
1989: 113)
By contrast, the children’s semantic system as well as their systems of con-

cepts and categories, although not altogether different from the adult ones, are
as yet incomplete and unstructured and will have to be adjusted and modified
in the years to come.
Overextensions are therefore pragmatic strategies which are not com-
pletely different from adult communicative strategies. They are based on three
types of relationships which also give rise to polysemy historically, structure
meanings synchronically and allow adult speakers to vary word meanings

contextually. These relationships are metaphorical ones, metonymical ones
and synecdochical ones, all three based on Seto’s cognitive triangle. It seems
that even very young children begin to see similarities, connections and class-
inclusions, and that with age they gradually adjust to the way the adults
categorize and form associations and restructure their linguistic and concep-
tual systems accordingly. After this initial training in seeing and understand-
ing similarities, connections and class-inclusions, they can later exploit these
connections more creatively. We can therefore distinguish between three
types of overextensions:
(i) synecdochical: overinclusions exploiting relations such as concep-
tual contiguity inside a taxonomy (e.g., GENUS-SPECIES, SPECIES-
SPECIES and so on),1
(ii) metonymical: overextensions based on perceptual, spatio-temporal
and functional contiguity,
(iii) metaphorical: analogical extensions based on perceptual similarity.
It should be stressed that in the case of compelled metonymical overex-
tensions words are not overextended to label an object for which the child has
already acquired a more appropriate name. For example, as soon as the right
word for ‘horse’ has been learned, the use of woof or dog for ‘horse’ stops. In
the process of language acquisition, the extensions of words are adjusted until
the lexical field (and the categorical domain) to which they belong has
subdivided in just the same way as that of the adult caretakers (cf. Barrett
1982). All this is not the case in creative metonymical shrinking where the
metonymically used words may replace an already known word and cut across
subdivisions within lexical fields and even across lexical fields. In short,
compelled metonymies are used by children to cover up gaps in their tiny
lexicons, whereas creative metonymies are used to express something new by
not using the already available words in their lexicons.
In order to determine which types of metonymy are used in compelled
metonymical overextensions, we looked at the examples accumulated in
Barrett (1982) and the examples we could find in the corpus assembled by
Braine (1976).
Barrett (1982: 322ff) reports on some of the overextensions produced by
Hildegard as reported by Leopold (1939, 1949) at the one-word stage. These
are of three types: synecdochical, metonymical and metaphorical overexten-
sions. We shall provide a selection of examples for each type:

(i) Synecdochical overextensions:
(4) a. Papa for father, grandfather, mother (1;0); any man (1;2); more
appropriate lexical items learned: Mama (1;3), Mann (1;5)
b. Ball for balls (1;0); balloon/ball of yarn (1;4); observatory
dome (1;8); balls of tin foil and paper (called paper-ball),
marbles/ovoid ball (called egg-ball)/a spherical bead (called
ball-beads) (1;11); more appropriate lexical items learned: bal-
loon (1;10), beads (1;9)
(ii) Metonymical overextensions:
(5) a. Wheel for a wheelbarrow wheel (1;8); a wheelbarrow (1;10);
toy wagon/a ring (1;11); more appropriate lexical item learned:
wheelbarrow (1;11) [based on part-whole relationship]
b. Choo-choo for trains (1;7); bradyscope (1;9); airplane/wheel-
barrow (1;10); streetcar/a trunk (;11); more appropriate lexical
item learned: airplane (1;11), streetcar (1;11), wheelbarrow
(1;11) [based on the relationship between the noise associated
with an object and the object itself — one could call this
‘onomatopoeic metonymy’]
(iii) Metaphorical overextensions:
(6) a. Boat for toy sailboat (1;10); airship (1;11)
b. Milkbottle for milkbottles (1;10); bottle containing white tooth-
powder (1;10)
We also looked at a corpus of two-word utterances collected by Braine
(1976). At this stage in acquisition, overinclusions seem to diminish, whereas
compelled metonymies seem to increase. In the examples in (7) we indicate
the metonymical relation that is exploited, as well as the illocutionary force of
the utterance.
(7) a. David (1;9): want pocket. [container-contained] {request}
b. David (1;10): here fix it. [action-object] {comment}
c. David (1;10): here hello. [words-object] {comment}
Situation: indicating or identifying toy telephone (cf. also example
8f)
d. David (1;10): here more book [instrument-action] {request}

e. David (1;10): that hello. [see above]
Situation: indicating the toy telephone
f. David (1;10): want more spoon. [instrument-action]{request}
g. David (1;10): gimme that blow. [instrument-action]{request}
Situation: wants to blow the match out
h. David (1;10): u [=? you] hello. [see above]
Situation: David is instructing me to talk on his toy telephone, there
is imperative intonation
i. David (1;10): more put in. [action for place] {request}
Situation: has been putting tinker toys in their box, apparently
wants to put somewhere the pieces the adults are using
j. Jonathan (1;11.15): more book. [instrument-action] {request}
k. Jonathan (1;11.15): all gone blow.[instrument-action] {comment}
Situation: the match box is empty, i.e., there are no more matches to
blow out.
l. Jonathan (2;0): daddy eat. [action for object] {comment}
Situation: occurred twice in the apparent sense of daddy’s food,
once it referred to a piece of bread he had taken from my plate
As one can see, children with limited lexicons focus on one object or
attribute of an object to achieve certain speech acts in a metonymical way.
They focus on one salient feature in a set framework or frame of repeated
interactions with the caregiver or parent. They say book if they want to read,
blow when referring to a match, they say hello for telephone, and so on.
Gradually, through interaction with the caregiver or parent who grants their
requests, the children elaborate on their comments. Finally, the metonymical
overextensions, which function as place-markers in the interactional and
conceptual framework, will be replaced by the ‘proper’ words allocated them
by the adults, such as read, match and telephone.
4. Creative metonymical shrinking
Compelled metonymical overextensions normally peter out by the age of 2;5,

although they can still occur as the vocabulary continues to develop. Between
age 4;0 and 5;0 one can begin to notice a more creative exploitation of the
natural pathways of meaning which are metaphor and metonymy. These new
metonymies are based on what we call ‘creative metonymical shrinking,’ as

they are usually produced in order to communicate new ideas with the least
verbal effort. In contradistinction to compelled metonymical overextension,
these metonymies are produced despite the fact that the message could also
have been conveyed using conventional linguistic signs and syntactic con-
structions.
Metonymical shrinking occurs in a wide variety of linguistic forms
(nouns, derivations, compounds), as described in Section 4.1., and conversion
of nouns into verbs as described in 4.2.
The following examples are based on Matthew (Brigitte’s son). For the
purpose of the collection of examples we have used the concept ‘metonymy’
in a broad sense, including not just strictly lexical metonymies, but also
‘morphological metonymies,’ that is, derivations and compounds based on
metonymic associations.
4.1. Creative metonymies
(8) a. Matthew (4;11) (The ‘Title story’)

Matthew started school in January. At first we thought he might eat
the school dinners. But he didn’t like them and insisted on bringing
his own lunch box like most of his friends do. So in the end we
relented and, walking to school in the morning, he brandished his
lunch box saying to everybody he met: “I love being a lunch box.”
Then he thought a bit and said: “I love being a sandwich, I really like
being a sandwich” — one could really see the metonymical chain
extend from his arm through the lunch box to the sandwich and back.
What he meant by this metonymical utterance was that he liked to be
part of the children who were allowed to bring a lunch-box (i.e., a
sandwich) to school and were not forced to have this horrible stuff like
potatoes and veg served at the school dinner!
b. Matthew (5;0)
Matthew was wearing a pullover that went down to his knees — we
laughed and said it was a mini-dress, whereupon he pretended to be a
girl and started to curtsey. As he was not very good at it, he practiced
all afternoon. At bed-time he pulled off the pullover and said “I have
enough of it, I have been wearing this curtsey all day.”
c. Matthew (5;1)
Matthew was playing with his Playmobil operating theater. He was
asked by a doctor-friend “Do you want to become a doctor in an

operating theater, that is a surgeon like me?” — “No,” he said, “I don’t
like the breathing-in-stuff (anesthesiologist), I would rather be a
listen-to-your-heart-doctor” (physician).
d. Matthew (5;1)
While eating cheese on biscuits, he said: “I am eating a bone-sand-
wich.” — “Why?” — “Because cheese is good for your bones.”
e. Matthew (5;3)
We were in Germany, Easter-Sunday, Matthew was trying to find his
chocolate eggs, rabbits, toys, etc. Having filled up his basket to
overflowing, he looked round and said: “Are there any more Easters
to find?” Later he repeats the same sentence in German: “Sind da noch
ein paar Ostern?”
f. Matthew (5;3)
We were at home, Matthew tried to push me out of the house and said
“Uff,” then “I uffed you really hard,” and a bit later “I am pushing
you.” [Uff is a German interjection expressing effort.]
g. Matthew (5;4)
We were playing postman. Matthew told me to go into the living room
and close the door so that he could post a letter to me underneath it. I
forgot to close the door. Matthew came in with his letter and shouted:
“Mummy, where’s your door?”
h. Matthew (5;4)
Despite all our efforts to prevent such ‘utterances’ — here it is:
Holding out his finger as a gun, Matthew said “Mummy, I want to
have a toy gun, I don’t want the baddies just to be fingered.” And, on
a similarly macabre line, playing with a police car and a fire engine, he
said: “I’ll put the hose there so that I can spray people with water to
see if they wake up when they are not dead” and “I’ll put this [a
different hose attachment] on the police car to gun people.”
i. Matthew (5;4)
Matthew watched Daddy build a climbing frame. First came the
normal nuts and bolts, then the dome nuts, whereupon Matthew said
“Oh dear, now you have to do another lot of nutting.”
j. Matthew (5;4)
After watching a program called “The Treacle People,” we concocted
a drink for him, consisting of orange, apple and passion-fruit juice,
plus orange squash, plus water which Matthew then baptized
“Treacle-people juice.” And he now refers to this kind of juice quite
commonly in sentences like “Mummy, not treacle-people juice
again!”…
k. Matthew (5;4)
Matthew was making ‘potions,’ that is, mixing various juices and
water and filling them into pots. The first was a potion against
earache, the second, he decided, was “another earacher.”
l. Matthew (5;4)
This spring was the coldest one on record, so when I dared to wear a
T-shirt one Saturday instead of a woolly jumper, Matthew looked at
me with big eyes and said: “Mummy why are you all summered?”
Later in the day, he put on a T-shirt himself and said: “Look, I am all
summered too now!”
m. Matthew (5;5)
Playing with Lego, building a film studio, using a fire engine exten-
sion ladder as a camera stand, he said: “They want to camera them.”
n. Matthew (5;5)
Matthew was listening to pop music on the radio, started jumping
around, waving hands and feet, and said: “Mummy, I am drumming
and guitaring at the same time.”
4.2. The metonymical ‘verbing’ of nouns
Whereas Matthew’s metonymies based on nouns, such as curtsey for pullover,

can be regarded as creative and novel, his metonymies based on verbs are
rather run of the mill for children — almost like saying He gived me that
instead of He gave me that (cf. Clark 1982; Gibbs 1994: 422f). This met-
onymical ‘verbing’ of nouns is usually based on the metonymical relation of
CAUSE-EFFECT.
However, although common in young children, Matthew’s verbal me-
tonymies are not ‘just’ overextensions. They are not based on the fact that he
has gaps in his lexicon which he has to fill somehow (as in I died that spider,
which he produced when he was about 1;11).

Let us look more closely at the example to gun. He knew the verb to kill.
However, he did not want to use it because he knew it is bad to kill somebody.
This is one reason why he used what one could call a metonymical euphe-
mism. Another reason is the fact that to kill is a bland and arbitrary everyday
word. To use to gun instead is at one and the same time avoiding the use of to
kill and using a word with a much more concrete and realistic motivation. And
finally, the most obvious reason for using to gun instead of to kill seems to be
that he is holding the object in his hand. The concrete relationship between
object and hand is the direct metonymical trigger. In contrast, Matthew would
quite happily say “I am going to kill you” when running around without such
an object in his hand!
Using metonymies is therefore, even in child language, a case of figurative
remotivation based on Clark’s ‘principle of transparency,’ which we also find
in adult language. As pointed out in Nerlich and Clarke (1988), there are two
fundamental movements in the evolution of language, one toward arbitrari-
ness, the other toward remotivation. Through metaphor the conventional or
symbolic aspect of a sign is remotivated in the direction of iconicity (based on
a relation of similarity), through metonymy it is remotivated in the direction of
indexicality (based on a relation of contiguity). Metaphor and metonymy can
therefore be interpreted as meta-icons and meta-indexes respectively, in adult
as well as in child language (ibid.: 80).
In the case of to finger and to nut, the verbal metonymies are somewhat
more creative and really based on what we have called metonymic shrinking.
Why say “I don’t just want to use my finger and pretend that’s a gun so as to
shoot the baddies” when you can say “finger the baddies,” or why say “Daddy
has to put another lot of nuts in” when you can say “a lot of nutting”?
Matthew’s noun-based metonymies are all creative and based on verbal
and conceptual shrinking. Instead of being based on one type of classical
metonymical relation, such as CAUSE-EFFECT, these metonymies exploit quite a
number of ad hoc metonymic relations, best exemplified in the nonce-creation
of Treacle-people juice, where a relation between watching a certain program
on TV and making a new sort of drink is exploited in a linguistic act of
baptism.
5. Understanding metonymies
As Gibbs (1994: 422) has pointed out,

[c]hildren clearly hear various kinds of metonymic expression in daily con-
versation. Parents, in speaking to children, might use metonymical utterances
like The buses are on strike [...] and children appear to comprehend these
expressions. No empirical research has specifically investigated children’s
ability to understand such conventional metonymic utterances [...]. Yet there
is evidence on children’s production of novel lexical items and on their
understanding of indirect speech acts that suggest they possess some ability to
think metonymically.
Although we would in no way dispute the fact that children have “some ability
to think metonymically,” there seems to be a gap between the production and
understanding of metonymies (a well-known asymmetry in language acquisi-
tion). As far as the observation of Matthew can be regarded as evidence, we
found that he was quite happy to produce metonymies, first in overextensions
(although that was before his utterances were recorded), then in a more
creative way, but that he did not readily understand conventional or dead
metonymies or accept novel metonymies uttered by adults. At an older age,
the following examples of his non-comprehension were noted:
(9) a. Matthew (4;5)
When Matthew was 4;5 the trains were on strike. On our way back
from the childminder we used to walk over a railway bridge and watch
the trains rush through the tunnel. However, in the autumn of 1994 we
were standing there and nothing came. When Brigitte realized what
was going on she said: “The trains are on strike.” Matthew then
asked “What are they doing, can you explain?”
b. Matthew (4;1)
At around the same time Brigitte read a story to him entitled “Jake and
the jumble panic.” This time Brigitte asked him: “Did you understand
this?” and he said that he did not, so she explained to him what “I am
Electrical” meant (i.e., I am the man that sells the electrical goods) —
and now this is one of his favorite stories, he really savors it.
c. Matthew (4;1)
Matthew came home from school and announced that there would be
a table-top sale on Saturday morning — and he would like to go there
because he really could do with another table top (the week before we
had just bought him a ‘table-top,’ that is, a very wide shelf that can
also serve as a drawing-table). In this case he had not learnt the
meaning of a dead or conventional metonymy yet. In the next case he
understood, but did not accept a novel metonymy.
d. Matthew (4;4)
Matthew came home from school. I said: “Wow, you have eaten
your whole lunch box!” Matthew burst out laughing and said: “Oh
Mummy, you got that wrong — it’s: You have eaten everything out of
your lunch box!”
Our hypothesis is that children produce compelled metonymical overex-

tensions until about age 2;5. We are not quite sure what happens then, but by
about age 4;0 they start to produce metonymies quite naturally to express
things more quickly (creative metonymical shrinking). They learn conven-
tional metonymies or what one could call dead metonymies in the same way as
they learn any other word, some more readily than others (e.g., “Let’s see how
many mouths there are to feed,” uttered by Mummy at a party).
However, they resist relatively novel or innovative metonymies. That is,
they resist metonymical shrinking when it comes from the grown-ups, here
following what Clark has called the ‘principle of conventionality.’ They seem
to assume implicitly that what grown-ups say has to be right, has to follow the
rules, because the grown-ups are the ones that set the rules and establish the
norm. But this also means that they should always stick to the rules, too. As
comprehension is centered around rule-governed input, children censor their
parents’ output by what they believe should be the parents’ input. And yet in
their own production, in unsupervised, uncensored output, children produce
metonymies which deviate from the adult norm.
To return to comprehension, when introduced gradually, children find
metonymies quite funny. This sign of early metalinguistic awareness may
foster true understanding at a later date. From age 7, we think, one could get
them to see what is actually going on in the use of metonymies, and later still,
at around age 10, when they are fluent readers, one could make them discover
the wonderful things metonymies achieve in newspaper headlines (e.g., “Italy
beats Germany in cup final” or “She drinks lager and eats McDonald’s” (about
an anorexic-looking model, The Times 31/5/96) (cf. David Weiss, New York,
e-mail 4 May 1996).
5.1. Experimental study
The examples given above tell us a lot about one child’s development of
metonymy comprehension, but it is critical also to collect experimental data
from a larger group of children. An embedded story framework was used to
test the comprehension of metonymy by children in two age groups, and to see
whether the presence of clues in the text would help to improve understanding.
5.1.1. Method
The participants in this experiment were sixteen children at the local univer-
sity play group. Ten children were in the two and three-year-old group, and six
in the four and five-year-old group. There were approximately equal numbers
of girls and boys, and only children whose first language was English took
part in the study.
Two short stories were used as stimuli. One was “Jake and the Jumble
Panic,” taken from a ‘Ladybird’ book by Joan Stimson (Stimson 1992). This
story was selected because it contained four metonymies. The second story
was about some pirates and a treasure hunt, and was written especially for the
experiment. Although this story contained a total of sixteen metonymies, only
six were used as test stimuli in the experiment. Both stories were deemed
suitable for the age group, and none of the children had heard them before.
The metonymies in the stories were independent, such that failure to under-
stand one would not in itself cause problems in understanding the others. A
copy of the Pirate Story can be found in the Appendix.
Children were tested in small groups of two or three. The experimenter
read the two stories, pausing each time one of the metonymies was reached.
Children were then asked to point to which of two pictures showed what was
happening in the story. One of the pictures was always a literal (incorrect)
interpretation of the metonymy, while the other picture correctly depicted the
story. The pictures were presented concurrently, with each taking up half of a
sheet of writing paper. The ordering of the pictures was random so that
children could not consistently point in the same direction to give the correct
answer. The pictures were presented in such a way that no child could see
which picture the other child had chosen, i.e., they had to make the choice for
themselves. After a picture had been chosen, the experimenter continued with
the story without informing the children whether they had been correct.
5.1.2. Results and discussion

The means show that the two and three-year-olds gave the correct answer 4.5
times out of ten, which is approximately equal to chance performance. The
means for the four and five-year-olds was 6.5 out of ten. The distribution is
slightly skewed by the fact that one two-year-old got seven metonymies
correct. No children got all answers correct, however one child in the older age
group got nine out of ten. A Mann-Whitney test showed that the four and five-
year-olds performed significantly better than the two and three-year-olds (U =
47, p< 0.05). These results are shown in Figure 1.
4 Comprehension
0
2-3 years 4-5 years
old old
Figure 1. Results for metonymy comprehension task
A second analysis was carried out based on the story construction. In

“Jake and the Jumble Panic,” a clue was always present in the text before the
metonymy itself appeared. For example:
(10) Jake found himself by a lady selling sweaters and cardigans.
“I’m looking for Wilfred,” he said.
“Over there!” boomed Cardigans.
In this case there was a direct hint in the text that he was talking to a lady
selling cardigans, which may have helped the children to understand the
metonymy. This was not the case for the Pirate Story, in which the metony-
mies were all ‘stand-alone,’ i.e., without any preceding clues. The data for all
children were therefore analyzed according to whether or not there were clues
present in the text. There were four metonymies with clues, and six without.
The results showed that on average 72% of the metonymies were under-
stood when clues were present, compared to 40.5% when there were no clues.
A Wilcoxon test (Wilcoxon T = 9, p < 0.01) showed that there was a
significant improvement in comprehension for the metonymies for which
clues were available in the text compared to ‘stand-alone’ metonymies. A
graph showing these results can be found in Figure 2.
80
70
60
50
40 Comprehension
30
20
10
0
No clues Clues
Figure 2. Comprehension by text type for both age groups
This experiment shows that the four and five-year-olds have a signifi-
cantly better understanding of metonymy than the two and three-year-olds,
which is what might be expected from what has been said elsewhere in this
paper and from experimental studies of metaphor comprehension (e.g.,
Vosniadou 1987). In addition this experiment shows that children are able to
use clues in the text of a story to help them to understand metonymies. More
evidence would be needed to establish whether clues such as this enable
children to develop some understanding of metonymy in the first place, or if
they are just used in some instances once children already have some basic
knowledge of how metonymy works. One observation that could be worth
following up is that for the metonymy Twenty sailor hats were marching down
the gangplank, most of the children who correctly selected the picture of
sailors pointed to the sailors hats rather than another part of the picture. The
children may have realized that hats cannot march on their own and used that
as the basis for selecting a picture (although clearly there are many children’s
cartoons and books in which it is possible for inanimate objects to march, walk
and even talk). It may be that children’s general cognitive development or
understanding of the real-pretend distinction is as important as any under-
standing of metonymy itself. Further experiments could also be carried out to
investigate other aspects of metonymy comprehension and in particular to see
if there are links between development in comprehension of metaphor and
metonymy.
6. Conclusion
Understanding the use and comprehension of metaphor and metonymy in

language acquisition has important implications for language teaching, espe-
cially for reading and comprehension and later on for writing and composi-
tion. Without the knowledge of how to understand and use words with
multiple meanings, how to exploit subtle shades in meaning to their advan-
tage, and how to multiply the meanings of words themselves, children’s
chances of acquiring literary skills would be severely reduced. It is necessary
to build on what Gibbs has called the ‘poetic minds’ of children to enhance
and develop their semantic competence.
Note
1. An example of a (rather rare) SPECIES FOR GENUS synecdoche can be found on a toy
recently bought at a “Sealife Centre.” It consists of a plastic dolphin, whale, shark or
alligator which you put into water where it expands 200 times in size. On the package it
says: “Put sealife into a pan, sink or pool…” “Sealife can grow to 22 inches, wet” (on
synecdoche in general, cf. Nerlich in print and Nerlich and Clarke in print).
References
Anonymous [Cicero]
1954 Ad C. Herennium, de ratione dicendi (Rhetorica ad Herennium). London:
Heinemann.
Barrett, Martyn D.
1982 Distinguishing between prototypes: the early acquisition of the meaning of
object names. In S.A. Kuczaj (ed.), Language Development. Vol. I: Syntax
and Semantics. Hillsdale, NJ: Lawrence Erlbaum, 313–334.
Bowerman, Melissa
1978 Systematizing semantic knowledge: changes over time in the child’s orga-
nization of meaning. Child Development 49: 977–987.
Braine, Martin D.S.
1976 Children’s First Word Combinations. Monographs of the Society of Re-
search in Child Development 41 (Serial No. 164).
Bréal, Michel
1991 [1884] How words are organized in the mind. In G. Wolf (ed.), The
Beginnings of Semantics: Essays, Lectures and Reviews. London:
Duckworth, 145–151.
Clark, Eve V.
1978 Strategies for communication. Child Development 49: 953–959.
1982 The young word maker: a case study of innovation in the child’s lexicon. In
E. Wanner, L.R. Gleitman (eds.), Language Acquisition: The State of the
Art. Cambridge, Massachusetts: The MIT Press, 390–428.
1993 The Lexicon in Acquisition. Cambridge: Cambridge University Press.
Esnault, Gaston
1925 L’imagination populaire: Métaphores occidentales. Paris: Presses Univer-
sitaires de France.
Fass, Dan
1991 Metonymy, case role substitution and sense ambiguity. In D. Fass et al.
(eds.), Proceedings of the IJCAI Workshop on Computational Approaches
to Non-Literal Language: Metaphor, Metonymy, Idiom, Speech Acts and
Implicature, held at the Twelfth International Joint Conference on Artificial
Intelligence (24 August 1991, Sydney, Australia), 42–51.
Gibbs, Raymond W.
Householder, Fred W.
1995 Plato and his predecessors. In E.F.K. Koerner, R.E. Asher (eds.), Concise
History of the Language Sciences. Oxford: Pergamon, 9–93.
Huttenlocher, Janellen, Patricia Smiley
1993 Early word meanings: the case of object names. In P. Bloom (ed.), Lan-
guage Acquisition. New York: Harvester and Wheatsheaf, 222–247. (First
published 1987)
Leopold, Werner F.
1939 Speech Development of a Bilingual Child: A Linguist’s Record. Vol. 1:
Vocabulary Growth in the First Two Years. Evanston, IL: Northwestern
University Press.
1949 Speech Development of a Bilingual Child: A Linguist’s Record. Vol. 3:

Grammar and General Problems in the First Two Years. Evanston, IL:
Northwestern University Press.
Nerlich, Brigitte
in print Synecdoche: a trope, a whole trope, and nothing but a trope? In N. Norrick,
A. Burkhardt (eds.), Tropical Truth. Amsterdam and Philadelphia: Benja-
mins.
Nerlich, Brigitte, David D. Clarke
1988 A Dynamic model of semantic change. Journal of Literary Semantics 17:
73–90.
in print Synecdoche as a cognitive and communicative strategy. In A. Blank,
P. Koch (eds.), Historical Semantics and Cognition. Berlin: Walter de
Gruyter.
Norrick, Neal R.
Benjamins.
Nyrop, Kristoffer
1913 Grammaire Historique de la Langue Française. Vol 4: Sémantique. Copen-
hagen: Gyldendalske Boghandel Nordisk Forlag.
Pease, Deborah Myers, Jean Berko Gleason, Barbara Alexander Pan
1989 Gaining meaning: semantic development. In J.B. Gleason (ed.), The Devel-
opment of Language. Second edition. Columbus, etc.: Merrill, 101–134.
Sappan, Raphael
1987 The Rhetorical-Logical Classification of Semantic Changes. Braunton:
Merlin Books.
Seto, Ken-ichi
in print Metonymy and the cognitive triangle.
Stern, Gustaf
1968 [1931] Meaning and Change of Meaning. Bloomington: Indiana Univer-
sity Press.
Stimson, Joan
1992 Jake and the Jumble Panic. In J. Stimson, Bedtime Stories for Under Fives.
Paignton: Ladybird.
Ullmann, Stephen
1957 [1951] The Principles of Semantics. Oxford: Basil Blackwell.
Vosniadou, Stella
1987 Children and metaphors. Child Development 58: 870–885.
Yamanashi, Masa-Aki
1987 Metonymic interpretation and associative processes in natural language. In
M. Naga (ed.), Language and Artificial Intelligence, International Sympo-
sium on Language and Artificial Intelligence (16–21 March 1986, Kyoto,
Japan). North Holland: Elsevier Science Publishers, 77–86.
Appendix
The Pirate Story used in the comprehension experiment: ‘The metonymic treasure hunt.’
Tom and Mary had been promised a special adventure for their birthday. For one day they
would be able to live and work on a pirate ship. When they came to the harbor, dressed up
as pirates, there she was, the Black Princess swaying gently on the waves. “What a
beauty,” said Tom. “But where are the pirates?” asked Mary. And there they came —
twenty sailor hats were marching down the gang plank, stood in a row before them and
saluted them. Mary and Tom wanted to greet them, but then they realized that they didn’t
know their names. “What are your names?” asked Tom. “Har-har, my dears, this is the first
mystery you have to solve,” said the captain of the ship, who had just stepped in front of the
pirates as if from nowhere. “My name is Captain Hatty, by the way,” announced the
Captain. Look at the sailors closely and try to find out what their names are.
Peggy (-leg)
(Eye) Patchy
Stripy
Dotty
Tatty
and
Fatty
Now the children knew the names of the pirates and everybody went aboard. “All hands on
deck,” cried the captain, “set sails!” And off they went. The Black Princess was heading for
adventure. When it was lunch time, the ship’s cook shouted “Twelve bells, Twelve bells,”
come and eat your grub. Everybody sat down in the tiny galley. The cook, Fatty, brought
the meal, and looking round, he said jokingly: “Oh dear, there are two more mouths to feed.
Let’s hope we have enough fish fingers!” To celebrate the arrival of Tom and Mary, the
captain got out a bottle of lemonade and everybody got a glass. But just as they were about
to speak a toast, the ship wobbled a bit and the lemonade tipped over, but what was worse,
the treasure-map, that the captain was just about to show to the children, slid on the floor
and got a bit dirty. Quickly, Tom jumped up and saved it before it got all dusty. The captain
took it and nailed it down on the table, to be on the safe side. After a few days they arrived
on a desert island. Everybody got off the ship and explored the beach. Tom had brought a
camera and said to Mary. “Stand there between the rocks and do a Captain Hook for me!”
Then they all went into the deep dark jungle. The sun shone down and everybody got
thirsty. Mary said: “I wish I could have a Ribena.” “No Ribena here my deary,” said the
captain, “you’ll have to do with this nice fresh spring water.” They all had a drink and went
on. Finally they arrived at the spot where there was a cross on the treasure map and started
digging. They dug and they dug. But they couldn’t find anything. Then Tom had a look at
the map that Dotty had pinned to a tree and said: “But it’s upside down!” The captain took
Dotty by his collar and shook him, saying “Don’t you have any brain! We have wasted half
a day now. Let’s have a rest, before we start the search again.” They all put their heads
down for an hour and then started looking for the treasure again — this time in the right
spot. And after a little digging, the whole crew shouted: “There it is!” They could all see a
huge treasure chest. They opened it and found gold coins, jewels and, what a surprise, toys
for Tom and Mary. Everybody was very pleased. They went back to the ship and sailed
home.
Recontextualization of Metonymy in
Narrative and the Case of Morrison’s
Song of Solomon
Anne Pankhurst
University of Edinburgh
1. Introduction
Metonymic referentiality in narrative fiction is associated with a number of

powerful effects, achieved by recurring references to a concept, experience or
object. The scope of a referent is extended from a single episode to the entire
text, assisting in the development of a communicative linguistic interface
between writer and reader. Poetic thought processes like metonymy are fun-
damental to our way of thinking, so the claim that we understand poetry by
means of our everyday experiences (Lakoff and Turner 1989; Gibbs 1994) is
also relevant to the consideration of poetic elements in narrative prose fiction.
The reader then activates cognitive strategies which facilitate access to the
new context or world of experience proposed in a text.
In the case of Toni Morrison’s novel Song of Solomon (1977 [1989]),
scene, plot and characters are opaque to readers unfamiliar with the social and
ethnic background to the fictional narrative. Among various means used by
Morrison to make this world credible are recurring references to an earring
made and worn by the central character, Pilate. The apparently simple object
described in the text is a means of identifying its wearer, enabling us to answer
the question: who is Pilate? The answer is available within the fiction and at
386 Anne Pankhurst
the same time evokes the real world of the reader’s experience. It is more than
a description. The earring has numerous features which refer the reader to
more than one concept. This metonymy also has a number of meta-narrative
functions serving to structure the overall development of thematic material in
the text. Its recurrence in new contexts is an important element of the narra-
tive. It reactivates the reader’s memory and highlights important episodes.
2. Theories of metonymy in narrative
2.1. Jakobson on metonymy
Theories of the functions of metonymy in narrative are of interest to the

analysis both of literature and of oral narrative, e.g., Propp’s (1928) analysis
of Russian folk-tales, Labov’s (1972) account of natural narrative. Here, I
shall limit the discussion to literary narrative in prose fiction, and to the well-
known theories of Jakobson and Riffaterre. Starting with the linguistic analy-
sis of text, both theories select metonymy as a regular semantic principle of
substitution which extends beyond a given linguistic domain and the
syntagmatic organization of a sentence. While Jakobson is concerned with
extending the semantic principle of contiguity to the syntactic structure of
discourse, Riffaterre emphasizes the conceptual contiguity created by repeti-
tion of a metonymic association once it has been recognized.
Jakobson (1956) proposed a theory in which the nature of discourse is
defined as either metaphoric or metonymic. To this bipolar theory of language
he added an important proposal, that the use of metonymy characterizes realist
prose fiction. In this type of discourse, metonymy rather than metaphor
structures the sequence of events, occurring as a structural element of dis-
course as well as a one-word substitution based on semantic principles. The
realist author “follows contiguous relationships, metonymically digressing
from the plot to the atmosphere and from the characters to the setting in space
and time” (Jakobson 1956: 82). Prose is said to be “forwarded essentially by
contiguity” in the sequence of events. But the realist writer is also “fond of
synecdochic details” (Jakobson 1956: 78) which lend verisimilitude to a scene
or to the depiction of personal characteristics. Metonymic elements in the text
are seen as belonging both to description and to narration of events, thus
working at two levels. The first function fills out the details of a given
moment, using short-cuts or metonymic reductions as well as synecdoche to

convey intended meaning, while the second moves the text forward by a
metonymic relation of contiguity between events.1
Clearly, different objects can serve in this way as both elements of
description and motivators of further narrative action. In Song of Solomon
Morrison highlights an earring. This is apt, since jewelry is understood in
many cultures as a metonymic means to identify a person, and as a marker of
personal status (Hoebel 1972). Because of its visual prominence, it becomes a
means of establishing social and personal identity and can therefore serve to
define its owner’s domain of influence. There is a particularly strong associa-
tion between jewelry and more abstract concepts such as power and social
status, whether that power is acquired by wealth, magic or religion. The
regular semantic principle which allows substitution of a costume for its
wearer is thus extended when an item of jewelry is an intrinsic part of a
person’s whole identity, or a means of identifying the whole by the outward
part (Norrick 1981). When used in a fictional narrative, however, the very
familiarity of the concept provides more than a simple means to jog the
reader’s memory about the characteristics of the person whom it represents.
Perception and interpretation of a character’s role in different episodes of the
narrative are facilitated by the references already established through general
or encyclopaedic knowledge.
2.2. Riffaterre’s theory of metonymy in narrative
Riffaterre’s (1990) analysis of the functions of metonymy expands on

Jakobson’s theory considerably and will form the framework for evaluating
the effects of this particular OBJECT FOR OWNER metonymy in Song of Solomon.
The theory, as I shall show, is primarily concerned with explicating textual
features as linguistic and structural elements, but moves towards recognition
of the role of the reader’s cognition. In this respect Riffaterre’s approach to the
inferences drawn from metonymy underpins the view that literature is under-
stood through real-life experience.
The creation of subtexts is, in Riffaterre’s theory, one of the major
functions of metonymy. Subtexts are recurring elements (not episodes) of a
narrative, serving to identify characters, scenes or themes at different points. If
this is true, metonymy clearly departs from being the poetic embellishment
expressed in the words of Gibbons (1969 [1767]: 70), “[metonymy] gives a
388 Anne Pankhurst
vast scope and liberty to the fancy: it both adorns and invigorates our style.”
But metonymy retains its rhetorical function of persuasion. In this case, it
persuades the reader of who Pilate is, and what the earring is, in the domain of
physical appearance. It also plays an important part in marking and extending
associated conceptual domains within her person.
This well-known literary device can be related to the reader’s understand-
ing of what is intended through an interface that has to be structured if new
elements are to be satisfactorily communicated. Many repetitions may be
needed before metonymies “develop into subtexts that mirror the whole of the
text in which they are embedded and facilitate the text’s interpretation”
(Riffaterre 1990: 21). There must be a prolonged sequence, dispersed through-
out the narrative and weaving in and out of it, forming part of the referential
frame of the text. In this case, the development of the earring as a subtext is a
partial explanation of its effect. The interpretation of what has been under-
stood also depends on the reader recognizing that the conceptualization moves
through different contexts of experience while preserving its unity of domain.
How is this achieved? The metonymic reference is extended from its immedi-
ate context to the whole textual structure, by means of recontextualizations.
For Riffaterre, repetition of a linguistic device is an indicator of its importance
in the narrative, but it is also more than that. It enables positive or negative
marking of the episode in which it occurs. The earring, for example, is
described as ‘foolish’ in Macon’s negative view; but also as ‘wonderful,’
when Pilate’s dominance is unquestioned. There are affective as well as
cognitive effects.
The attempt to create reality or at least a possible world in fiction requires
a more complex function for metonymic references, which are known to be
‘two-faceted.’ The inference that the true referent of an earring is not the
object but its wearer (or indeed further implied characteristics of either the
ornament or the person) is relevant to this discussion. The fictional world must
have its own consistency or truth, understood by the assumed reader in terms
of a real experienced world and a rich personal encyclopedia of knowledge
and beliefs (cf., e.g., Langacker 1987). The single word ‘earring’ contains a
potentially rich chain of associations. Through a series of links, initially
following the metonymic pathway which allows substitution of POSSESSOR
FOR POSSESSION, the earring represents the person wearing it. The earring
attracts attention by its unusual form (a brass box) and therefore its function as
a container, standing for the personal treasure contained, is made salient.
Riffaterre raises a further question in claiming that without metonymy,

metaphor and other tropes it would be impossible to differentiate between
verisimilitude (the imitation of truth in fictional discourse) and factual report-
ing of events. Distinction between real life and fiction makes the presence of
tropes a special, literary phenomenon. Metonymy can lead to the construction
of symbolic value, for example, the use of pen to refer to both the profession of
writing and an author’s power of description and composition:
Let other pens dwell on guilt and misery. (Jane Austen, Mansfield Park: 420)
Metonymy also may have allegorical functions when it is consistently used to

structure the events of a fictional narrative intended to represent a real-world
situation (cf. John Bunyan, Pilgrim’s Progress). But metaphor and metonymy
are pervasive in everyday discourse as well as literature and cannot be distin-
guishing markers of literary as opposed to non-literary style (Lakoff and
Turner 1989; Ortony 1993; Gibbs 1994; Steen 1994).
In addition to metonymic relations between events (Jakobson’s ‘forward-
ing by contiguity’), fictional narrative requires characters to be represented
with enough depth and credibility for a reader to understand them as authentic.
This can be achieved by means of ‘synecdochic detail,’ the purpose of which
is to foreground the characteristics of fictional personae. It is part of the ‘mock
reality’ (Leech and Short 1981) or ‘fictional truth’ (Riffaterre 1990) which
deals with credibility, authenticity, objectivity and vividness in creating veri-
similitude, and without which a reader does not understand the world or
domain in which the writer’s intended meaning is realized. In an earlier
discussion of the functions of metonymy, Riffaterre made a general point
which extended the sense of PART FOR WHOLE synecdoches and is relevant to
the consideration of the earring as a potentially real object:
In the novel, the trend has been to dissolve or disperse the image of a character
into surrounding objects or to suggest a state of mind or the significance of a
dramatic situation through physical details that invite certain deductions or
inferences on the part of the reader. (Riffaterre 1982: 273)
Thus, an inferential reading of a text can be triggered by just those details

which create an apparently real world within the fiction. If a physical feature
or an object closely associated with a person stands for a non-physical quality
or an abstract idea, it constitutes an extension of meaning, moving possibly
towards metaphor (Dirven 1993). This intention to extend meaning is as-
sumed, and justifies the reader’s inferential process. Just such complex exten-
390 Anne Pankhurst
sions of meaning are to be seen in the functions of Pilate’s earring in Song of

Solomon.
3. A case study: Song of Solomon
3.1. The earring and the identity of Pilate
The first point examined here is the actual description of the earring and the
question of its complexity. Metonymy requires that an object stand for some-
thing with which it has close associations, but this earring is more than a
simple decorative object. It is known to contain a special piece of paper on
which the name ‘Pilate’ is written. It was the only word her illiterate father
ever wrote and is the only family document known to exist. It is therefore a
container for an important legacy and may be read as a means of accessing the
past.
The earring is first mentioned in an account of how members of the
family acquired their names. Macon Dead II recalls the birth and naming of his
sister, Pilate, and his father’s insistence that the name randomly chosen and
copied from the family Bible should be kept.
And it did stay there, until the baby girl turned twelve and took it out, folded
it up into a tiny knot and put it in a little brass box, and strung the entire
contraption through her left earlobe. (Song of Solomon: 21)
The earring is the vehicle of Macon II’s ambivalent feelings about his sister,
Pilate, once “the dearest thing in the world to him” and now “odd, murky, a
regular source of embarrassment.” Her “foolish earring” (Song of Solomon:
22) derives from her ragged clothes and disreputable lifestyle, transferring the
foolishness of the woman to the object by the epithet. Her poverty contrasts
with her brother’s material success and aspirations. Macon tells his son,
Milkman, the secret of the earring’s contents in warning him to stay away
from Pilate who “can’t teach you a thing you can use in this world” (Song of
Solomon: 53). Since Pilate is Macon’s sister and Milkman’s aunt, our curiosity
focuses on trying to understand why both object and wearer are ‘foolish,’ and
we therefore infer that there is a special motivation for this attitude yet to be
discovered.
Wearing the earring is a metonymy for Pilate’s physical appearance, but
it also stands for her family’s origins. It encapsulates her respect for her father,
her power over her brother and nephew, and her status as matriarch. To
account for it as synecdochic detail, a part for the whole, is insufficient. The
known implications expand the temporal and conceptual domains to include
the past, and other people’s memories besides those of Pilate herself. Within
this frame, the earring acquires the properties of a special message-bearing
object, symbolically representing the history of the wearer’s family.
Riffaterre invokes psychological plausibility in claiming that the impor-
tance of an object as a poetic device is heightened by repetitions, which confer
authority on it. As it is recontextualized, the cumulative effect becomes
stronger. In this case, Macon II narrated briefly the making of an earring out of
a brass box. A second mention of the earring, in another episode, permits the
reader to access another character’s experience and point of view. Pilate’s
single earring enhances her role and status, in the eyes of outsiders as well as
to her own family, but it is of particular importance to her nephew. When the
boy is brought to see her by his friend, Guitar, she has an overwhelming effect
on him. She will exercise great power and become the instrument for
Milkman’s rebellion against social common sense and his father’s wishes. The
earring, although only one of the striking elements of Pilate’s appearance, is
the salient feature of an extraordinary woman.
As they came closer and saw the brass box dangling from her ear, Milkman
knew that what with the earring, the orange and the angled black cloth,
nothing — not the wisdom of his father or the caution of the world — could
keep him from her. (Song of Solomon: 36)
While it could be argued that the earring is only one of the ways in which
Pilate influences Milkman in that episode (others involve food, story-telling,
assumption of seniority and the right to direct him), it is the highlighted
element. The consequences of this visit motivate much of the subsequent
narrative, within the theme of family origins.
3.2. The earring as family history
At a later stage of the narrative the account of the making of the earring is
expanded with a larger number of details of past history, after the potentially
dangerous relations between Pilate and Milkman have been established. The
story-teller is again Pilate’s estranged brother, Macon II, who relates family
history to Milkman (Macon Dead III). The brother and sister had been made to
abandon their father’s farm when it was forcibly acquired by white farmers
392 Anne Pankhurst
who killed their father, Macon Dead I (Jake). Pilate created the earring herself
during a period of isolation when she and Macon II were hidden for their own
safety by a woman called Circe. The stay in the house, with its intertextual
allusion to the stay of Ulysses and his sailors with the enchantress Circe, is
exile and imprisonment for the two children, deprived of the Edenic environ-
ment of their home. The repetition recreates and expands Macon II’s memory
of the past, but the negative marking of the earring as foolish is absent.
Before they left the farm she’d taken the scrap of brown paper with her name
on it from the Bible, and after a long time trying to make up her mind between
a snuffbox and a sunbonnet with blue ribbons on it, she took the little brass
box that had belonged to her mother. Her miserable days in the mansion were
spent planning how to make an earring out of the box which would house her
name. She found a piece of wire but couldn’t get it through. Finally, after
much begging and whining, Circe got a Negro blacksmith to solder a bit of
gold wire to the box. Pilate rubbed her ear until it was numb, burned the end
of the wire, and punched it through her earlobe. (Song of Solomon:167)
The choice of this special container for the word ensures that by inferen-
tial association the reader will understand that it has several levels of impor-
tance both for the characters in the story, and for the reader. Because the name
Pilate was given by her father, unlike their surname ‘Dead,’ which was
allocated officially and meaninglessly, the name will remain in the box as a
record of her true family history. Like a phylactery or amulet, it is bound to its
wearer and contains a special reminder that justification for power and identity
can be contained in one single word. Pilate’s action of creating the earring
preserves her identity and is her self-affirmation against powerlessness. Para-
doxically, lexical values in the phrases — the scrap of brown paper and the
little brass box — indicate that before the flight, the paper and the box were of
no special significance. They became valuable in the state of exile to which
Pilate and her brother were condemned, and which is structured by using the
earring as a mnemonic device.
Up to this point in the narrative, the earring has been no more than an
object by which Pilate can be identified. As the relationship between the boy
and his aunt grows, the earring acquires still greater significance. The different
implications combine the person with her symbolic role. Pilate is identifiably
different from all other people by the fact that she wears this single earring. As
metonymy, in the relationship of physical and mental contiguity within the
domain of the person, it stands for her sense of family and ancestry, her love
for her dead father, her strangeness in society’s eyes and her personal identity
as a vivid, flamboyant character with a strong self-will. The earring has

therefore a number of potential referents, each made salient in turn in the
recontextualizations.
3.3. The earring as encapsulation of folk memory
Temporal sequencing of a fictional world is a problem for the reader’s

understanding of what is happening at any given point of time. The existence
of the earring serves as a point of reference, and a stable element, in the non-
sequential account of the lives of different characters. The history of the
earring matches its owner’s development as she changes in the course of the
narrative. In different episodes she is an orphaned girl, an abused itinerant
worker, a marginalized but powerful woman in her city neighborhood, then a
magnetic attraction for her nephew and a trigger for his own search for his
family origins. Running parallel to Pilate’s story, but without the need for
flashbacks to retell the past, is Milkman’s story. After his childhood in the
City, he decides to find the family’s roots, in the rural South. During this
journey of discovery the earring subtext is again woven into the narrative as a
point of reference. Milkman is near the place where Pilate’s father (his
grandfather) was killed and the house where the earring was made. He hears
another account of its creation from the Rev. Cooper, which confirms to him
and to the reader the truth of Macon II’s account.
[Milkman] had walked right by the place where Pilate’s earring had been
fashioned, the earring that had fascinated him when he was little, the fixing of
which informed the colored people here that the children of the murdered man
were alive. And this was the living room of the son of the man who made the
earring. (Song of Solomon: 231)
Through Milkman’s experience, which verifies the truth of the family stories,
the reader understands that the earring and its contents exist in the past, present
and future of the narrative time. Given the complex spatio-temporal digres-
sions in this story, it functions as a necessary unifying or connective device for
the various concepts it contains (cf. Fauconnier and Turner, this volume).
3.4. The earring as sign of matriarchal power
The power of the earring as a referential frame to the text develops more with
each episode, notably at critical moments in Pilate’s family history. The most
394 Anne Pankhurst
significant is at the funeral of Pilate’s granddaughter, Hagar, who died after

Milkman ended her relationship with him. Pilate bursts into the service
shouting “Mercy!”, asserting her matriarchal right to express her grief in a
traditional song. The dramatic effect of her entrance is highlighted through the
viewpoint of the mortician, who is overwhelmed and dazzled by the authority
contained in the earring.
She tilted her head and looked down. Her earring grazed her shoulder. Out of
the total blackness of her clothes it blazed like a star. The mortician tried to
approach her again, and moved closer, but when he saw her inky, berry-black
lips, her cloudy, rainy eyes, the wonderful brass box hanging from her ear, he
stepped back and looked at the floor. (Song of Solomon: 317)
At this point, the focalizing view of an anonymous mortician generalizes the

effect of the metonymy. The earring, still the means of identifying Pilate,
acquires a number of new attributes which enhance its special function. It
blazed like a star, it is the wonderful brass box against a grief-stricken
background of total blackness of her clothes and her inky, berry-black lips,
her cloudy, rainy eyes. The prominence of the earring recalls the figurative
significance already created both in the case of its influence over Milkman,
and its function as container for Pilate’s name, but it now has wider social and
personal meaning, standing for the power of the matriarch.
At this point in the narrative it might be argued that the earring is no
longer metonymy, but has moved into a new, abstract domain of personal
power, becoming a metaphor. In the reader’s understanding, the power of the
object to blaze like a star gives access to a new aspect of its wearer, an
enhancement of her status. The character called Pilate develops from being an
outsider with rather unusual personal characteristics into a model of empower-
ment. From the reader’s point of view, however, the expert identification of
this transition to metaphor could be less important than the function of the
earring as a complex figurative marker within the text. Immediate interpreta-
tion of its significance bypasses linguistic and rhetorical analysis.
3.5. The destruction of the earring
The theories of Jakobson and Riffaterre, which have so far provided a frame to
the consideration of metonymy in fictional narrative, do not attribute a special
role to closure. Yet it is plausible to assume that communication with the
reader must be influenced by the nature of the ending, whether it is open or
closed, happy or tragic, predictable or surprising. At this moment of Song of

Solomon, the subtext of Pilate’s earring is again woven into her fate, engaging
the reader’s attention and providing an ironic comment from which inferences
may be drawn.
As we have seen, the earring contains Pilate’s desire to preserve the past
by means of the word, the specially-given name, the sole legacy of happiness
and home origins. Ironically, the earring is stolen by birds at the close of the
narrative. After Milkman has successfully found the family’s roots, Pilate
buries the earring along with her other treasure, a bag of bones. It is stolen
from the shallow grave by a bird. At the same moment Pilate herself dies,
murdered by Guitar. The figurative value of the earring is lost as it becomes an
indefinite, devalued something shiny, a mere object in a bird’s beak. From
this, the reader may infer that the earring’s special functions as keeper of
Pilate’s name and mark of her identity in this narrative have ended.
‘Should we put a rock or a cross on it?’ Milkman asked.
Pilate shook her head. She reached up and yanked her earring from her ear,
splitting the lobe. Then she made a little hole with her fingers and placed in it
Sing’s snuffbox with the single word Jake ever wrote. [....] Two of the birds
circled round them. One dived into the new grave and scooped something
shiny in its beak before it flew away. (Song of Solomon: 335)
This event returns the earring to the realm of the literal, but does not
negate its metonymic value. Although depersonalized and demythologized, it
still has a function in the structure of the narrative, insofar as it now could
motivate either of two potential outcomes. Because the metonymy has carried
an important amount of reference in the narrative, the first inference drawn is
that the closure is clear and dramatic. The earring is no longer defined as a
treasure; it is a shiny, ownerless thing. The move to anonymity is meaningful
in the thematic context, demonstrating to the reader that the writer’s intention
has been to focus only on Pilate’s life by means of the metonymic referential-
ity of the earring. In this sense, the metonymy has a meta-narrative function as
a mark of fictionality. The other potential outcome is open-ended. We do not
know what has become of the earring, but we can speculate as to its new
functions in some other narrative yet to be constructed.
396 Anne Pankhurst
4. Discussion
In Song of Solomon the earring is part of the physical description of its wearer,
always present and salient in her appearance, therefore she can be identified
by mentioning it. As a part for the whole, within the domain of the person, it is
a synecdochic detail of self-presentation. Its function is not, however, limited
to local use. In the macro-structure of the narrative, the earring is highlighted
by different focalizing voices at dramatic moments. It has been created during
the children’s flight from persecution in their home, it creates the bond
between Milkman and Pilate, it is the focus of Pilate’s matriarchal power
during the funeral of Hagar. It confirms Milkman’s discovery of his family’s
origins, and disappears at the death of Pilate. It becomes a short cut to our
understanding of Pilate’s importance for her family and reactivates memory of
previous events both for characters within the narrative and for the assumed
reader. Thus, it is evidence of a rich conceptual domain extending beyond the
person to family, race and social disempowerment.
The complex use of metonymy in this fictional narrative cannot be
accounted for by a single theory. Jakobson proposed that metonymy, retaining
its character as a figure defined by contiguity, facilitates forward movement of
the narrative when seen in a macro-context. This is not enough to explain how
the reader comes to the understanding that the single word ‘earring’ has the
property of multiple referentiality. The other part of Jakobson’s theory, the
provision of synecdochic detail, is also oriented towards exegesis of the text
rather than the reader’s process of understanding and interpretation. A more
satisfactory analysis is provided if we follow Riffaterre’s view that metonymy
is instrumental in creating subtexts which by force of repetition are crucial to
the reader’s grasp of the fictional world and may provide the basis for an
allegorical reading of the narrative.
Riffaterre defines subtext as a means of creating an actual or potential
relationship between the fictional world and the real world, achieved through
multiple references. When recontextualized, subtexts have a meta-narrational
function. They hold together disparate elements within the macrotext by a
network of inference because the patterns set up match existing patterns of
cognition (cf. Black 1993 for a discussion of similar effects generated by
metaphor). The availability of complex yet coherent multiple reference is of
great importance for the systematically cohesive development of longer narra-
tive (Miller 1985; Schulz 1992). If the fictional world is to be perceived as
true, albeit possessing fictional truth, the reader requires a number of entry
points into this world, created by means of reactivating what is already known
or believed in the real world. Understanding is, after all, an interactive process
involving the reader. “Narrative truth is a linguistic phenomenon experienced
through reading, a performative event with participation on the reader’s part”
(Riffaterre 1990: xiv).
The cognitive theory of metonymy (Gibbs 1994) addresses the impor-
tance of metonymic models of thought. Metonymy, as well as having linguis-
tic realizations, is a universal cognitive process which enables understanding
to take place. So in a very general way, it can be used in any kind of
communication for specific purposes, to express meaning, and will be under-
stood automatically.
People’s knowledge in long-term memory of coherent, mundane series of
events can be metonymically referred to by the mere mention of one salient
subpart of these events. We see that the mention of the subpart metonymically
stands for the whole event. (Gibbs 1994: 330f)
If this is generally true, it applies as much to narrative fiction as to

everyday spoken communication. As we read this narrative, we assume that it
has been created by Morrison with the intention of capturing a real or poten-
tially real world. We have inferencing power, and we share a number of
experiences and beliefs with the author and the characters. But the world of
Pilate and her family cannot be interpreted unless we recognize them. Met-
onymic reference through everyday objects is crucial in creating new con-
cepts. The mention of the earring acts as a trigger for a complex series of
references because it is a salient subpart of the woman’s physical appearance,
and we understand also that it plays a vital role in the reconstruction of her
family history and can therefore stand for it. To sum up, in the case of Song of
Solomon we can use two conceptualizations typical of metonymic thinking.
The object serves to identify its wearer, and the narrative recontextualization
of the object stands for potential, more abstract, interpretations. Its function is
to help the reader’s management of long text, working as a poetic device to
help the reader’s memory.2 By metonymic extension of the original referent,
through conceptual links, transition from physical to abstract domains is
effected in a way similar to the processes used in accessing an unknown world
in real life. In understanding, appreciating and teaching literature this is
valuable.
398 Anne Pankhurst
Notes
1. Jakobson’s theory of metonymy has been developed by other literary critics, notably
Lodge (1977) who argued strongly for the co-presence of metonymy and metaphor in
post-modernist fiction.
2. This function is not exclusive to metonymy. In a longer narrative, both metaphor and
metonymy can function as underlying structural devices in a text (Lodge 1977; Werth
1994).
References
Austen, Jane
1990 [1814] Mansfield Park. London: Penguin.
Black, Elizabeth
1993 Metaphor, simile and cognition in Golding’s The Inheritors. Language and
Literature 2: 37–48.
Bunyan, John
1974 [1678] A Pilgrim’s Progress. London: Collins.
Dirven, René
Gibbons, Thomas
1969 [1767] Rhetoric. Menston: The Scolar Press.
Hoebel, Edgar
1972 Anthropology: The Study of Man. New York: McGraw Hill.
Jakobson, Roman
1956 Two aspects of language and two types of aphasic disturbances. In R.
Jakobson, M. Halle (eds.), Fundamentals of Language. The Hague: Mou-
ton, 55–82.
Labov, William
1972 Language in the Inner City. Philadelphia: University of Pennsylvania Press.
1989 More than Cool Reason. Chicago: The University of Chicago Press.
Leech, Geoffrey N., Michael Short
1981 Style in Fiction. London: Longman.
Lodge, David
1977 The Modes of Modern Writing: Metaphor, Metonymy and the Typology of
Modern Literature. London: Edward Arnold.
Miller, J. Hillis
1985 Introduction to Bleak House. London: Penguin.
Morrison, Toni
1977 [1989] Song of Solomon. London: Picador.
Norrick, Neil
Benjamins.
Ortony, Andrew
1993 Metaphor and Thought. Second edition. Cambridge: Cambridge University
Press.
Propp, Vladimir
1928 Morphology of the Folk Tale. Austin: University of Texas Press.
Riffaterre, Michael
1982 Trollope’s metonymies. Nineteenth Century Fiction 37/3: 272–292.
1990 Fictional Truth. Baltimore: Johns Hopkins University Press.
Schulz, Victor
1992 ‘Deep structure signals’ in fiction. In D. Stein (ed.), Co-operating with
Written Texts. Berlin: Mouton de Gruyter, 335–360.
Steen, Gerard
1994 Understanding Metaphor in Literature. London: Longman.
Werth, Paul
1994 Extended metaphor, a text-world account. Language and Literature 3/2:
79–103.
List of Contributors
(Present addresses)
Andreas Blank
Philipps-Universität Marburg, Institut für Romanische Philologie
Wilhelm-Roepke-Str. 6D, D-35032 Marburg, Germany
E-mail: andreas.blank@mailer.uni-marburg.de
David D. Clarke
University of Nottingham, Dept. of Psychology
University Park, Nottingham NG7 2RD, UK
E-mail: David.Clarke@nottingham.ac.uk
René Dirven
Gerhard-Mercator-Universität Duisburg, Fachbereich 3, Sprach- und Li-
teraturwissenschaften
Postfach 10 15 03, D-47057 Duisburg, Germany
E-mail: Rene.Dirven@orchis.be
Gilles Fauconnier
University of California, San Diego, Cognitive Science Center
La Jolla, CA 92093, USA
E-mail: faucon@cogsci.ucsd.edu
Kurt Feyaerts
KU Leuven, Dept. Linguistiek
Blijde-Inkomststraat 21, B-3000 Leuven, Belgium
E-mail: kurt.feyaerts@arts.kuleuven.ac.be
University of California, Santa Cruz, Dept. of Psychology
Santa Cruz, CA 95064, USA
E-mail: gibbs@cats.ucsc.edu
402 List of Contributors
Louis Goossens
Universitaire Instelling Antwerpen, Dept. Taal- en Letterkunde: Ger-
maanse talen
Universiteitsplein 1, B-2610 Antwerpen (Wilrijk), Belgium
E-mail: louis.goossens@uia.ua.ac.be
Olaf Jäkel
Matin-Luther-Universität Halle-Wittenberg, Fachbereich Sprach- und
Literaturwissenschaften, Institut für Anglistik und Amerikanistik
D-06099 Halle (Saale), Germany
E-mail: jaekel@anglistik.uni-halle.de
Peter Koch
Eberhard-Karls-Universität Tübingen, Romanisches Seminar
Wilhelmstraße 50, D-72074 Tübingen, Germany
E-mail: peter.koch@uni-tuebingen.de
Zoltán Kövecses
Eötvös Loránd University, Dept. of American Studies
Ajtósi Dürer sor 19-21, H-1146 Budapest, Hungary
E-mail: kovecses@isis.elte.hu
Brigitte Nerlich
University of Nottingham, Dept. of Psychology
University Park, Nottingham NG7 2RD, UK
E-mail: bn@psychology.nottingham.ac.uk
Anne Pankhurst
University of Edinburgh, Dept. of Applied Linguistics
14 Buccleuch Place, Edinburgh EH8 9LN, UK
E-mail: afp@holyrood.ed.ac.uk
Klaus-Uwe Panther
Universität Hamburg, Seminar für Englische Sprache und Kultur
Von-Melle-Park 6, D-20146 Hamburg, Germany
E-mail: panther@rrz.uni-hamburg.de
Paul Pauwels
Katholieke Vlaamse Hogeschool
St. Andriesstraat 2, B-2000 Antwerpen, Belgium
E-mail: jverscha@pophost.innet.be
List of Contributors 403
Günter Radden
E-mail: radden@rrz.uni-hamburg.de
Ken-ichi Seto
Osaka City University, Faculty of Literature
3-3-138 Sugimotocho, Sumiyoshi-ku, Osaka, Japan 558
E-mail: seto@lit.osaka-cu.ac.jp
Linda Thornburg
Eötvös Loránd University, Dept. of American Studies
Ajtósi Dürer sor 19-21, H-1146 Budapest, Hungary
E-mail: lthornburg@isis.elte.hu
Zazie Todd
University of Leicester, Dept. of Psychology
University Road, Leicester LE1 7RH, UK
E-mail: ztl@le.ac.uk
Mark Turner
University of Maryland, Dept. of English Language and Literature
1102 Francis Scott Key Hall, College Park, MD 20742
E-mail: mturn@wam.umd.edu
Christian Voßhagen
E-mail: fs6a023@rrz.uni-hamburg.de
Richard Waltereit
Eberhard-Karls-Universität Tübingen, Romanisches Seminar
Wilhelmstraße 50, D-72074 Tübingen, Germany
E-mail: richard.waltereit@uni-tuebingen.de
Beatrice Warren
Lund University, Dept. of English
Helgonabacken 14, S-22362 Lund, Sweden
E-mail: beatrice.warren@englund.lu.se
Subject index
A blend(ing) 77-80, 84, 88-89

Action ICM 36-37, 43 metaphorical 86
action schema 277-278, 280-281, 284- blended space 77, 82
285 body part 264-265, 270
active zone 31, 250 bridge, metonymic 193, 196, 199-200,
ad hoc metonymy 139-140, 155 265
AFTER condition 337 bridging 197-198, 202
agent 246-249, 263, 277-278, 280, 285- by-name 211, 214
286
alternation C
frame-based 237, 239, 250 C(ategory)-related transfer 92
locative 239 C-relation 92, 113, 117
role 237-238, 249 case frame 277
‘swarm’ 240-241 case grammar 279
thematic 236 case role 177 (see also role, semantic)
ambiguity 235, 239-240, 248, 250 Category-and-Member ICM 34
anaphoric pronoun 10-11, 41 Category-and-Property ICM 35
anthropocentrism 285-286 Causation ICM 36, 38
anthroponomy 211 change
antonomy 291-293, 295, 306 semantic 158, 140, 142, 144, 156,
aphasia 171-172 160, 174, 184, 187, 247
argumentation theory 318 metonymic 140, 142-143, 145, 147,
association 172, 187, 256, 292, 318, 150, 155-156, 158
366, 390 clarity 50
metonymic 370, 386 class
associationist psychology 141-142, 149, inclusion 367
159-160 membership 283, 285
associative relation(ship) 142-144 co-presence 160-161, 169, 178-180,
attitude 296-299, 343 182, 184
cognitive
B domain 355
backgrounding 153 frame 184
basic-level term 281 model 312
BEFORE condition 337-339, 350 colloquial tautology 35, 62, 72
beneficiary 285 Commissive Scenario 345
406 Subject index
commissive 344-346, 348, 351 core 337-338

indirect 356 corpus 212, 219, 221, 224, 260-261,
communicative 271-272
efficiency 174 counterfactual 79
principle 44, 50 cultural model 313
compelled metonymic overextension
362, 364, 367, 369, 375 D
complementarity 291-294, 306 dative 285
composition 158, 161 dead metonymy 375
compound 123-125, 127, 132-133, 214- default relation 126
216, 218, 220, 370 demetonymization 269
compound surname 219-220 denominal verb 123-125
comprehension 376 deontic (also see necessity) 194, 196-
Concept ICM 23, 26, 28 197, 201-203, 207
concept metonymy 23, 26, 28-29 derivation 158, 161, 187, 370
concurrent processing model 70 direct object 233, 236-237, 239, 241,
Constitution ICM 32, 186 243, 245-246, 248, 250
constraint, grammatical 233 Direct-Reference ICM 25-26
Containment ICM 37, 41, 186 Directive Scenario 337
context-induced interpretation 193 directive
contextual effect 13 indirect 356
contiguity 19, 92, 95, 117, 130, 139- speech act 336, 344-346, 351
146, 149-151, 153-154, 157-160, 170- discourse 155
174, 177-180, 182, 184-185, 187, distance, metonymic 85
234-236, 238-240, 244-245, 281, 291, domain 9, 160, 173, 234, 256, 258-263,
295, 316-317, 320, 322, 329, 367, 265, 270, 289, 295, 301, 305
387, 396 boundary 320
conceptual 144, 146, 171, 173-174, conceptual 305, 350-351, 388
178-179, 184, 289, 305-306, 338, highlighting 153, 302
366-367, 386 mapping 153
referential 355 matrix 153, 160, 257-258, 260-261,
relation 242, 244-245 317-320, 328-329, 334
schema 182-184 dysphemism 265-266
spatial 142-143, 146
spatio-temporal 95 E
temporal and causal 177 effect
contradiction 306 rhetorical 52
contrariness 306 social-communicative 53
contrast 172 E-relation 92-93, 98, 117
Control ICM 27, 37, 40 ellipsis 142, 144, 171-172
conventionality 375 encyclopaedic knowledge 238
conventionalization 156-157 entailment 267
of conversational implicature 156 epistemic 193-206
conversion 37, 275, 277-281, 283-286, eponymous verb 65
370 error recovery model 70
Subject index 407
essive H
role 280 hapax legomenon 174
schema 280, 283-285 head 220-221
euphemism 264-266, 269, 272, 289, hedged performative 342-344, 351, 356
295, 299-301, 306 hierarchy
metonymic 53, 373 conceptual 328
evaluation/evaluative 292, 294-296, metaphoric 309-310, 314, 324
301-302, 306 metonymic 309, 321, 323
Event ICM 32-33 of grammatical relations 236, 248
event schema 275, 278-280, 285 of metonymic vehicles 29, 52
experiencer 248-249, 285 highlighting 11, 153, 174, 184, 258-259,
experiencing schema 285 270, 280, 302, 305
expressive 140, 155-156 hyperbole 215, 321-323
extension(al) 25, 153-154, 270, 317, hyponymy 94
320-321, 325, 365, 367
analogical 365, 367 I
metaphorical 281 ICM (Idealized Cognitive Model) 9, 17,
metonymic 150, 153, 325 19-23, 26-29, 67, 186, 256, 334
iconicity 250, 277, 282, 373
F idiomatic expression 309-310, 334
face-threatening act 343 illocution 176
figurative language 309 illocutionary
figure/ground 151-152, 154-157, 159 act 335, 343-344, 356
focusing, metonymic 286 force 350, 368
folk model 212 image schema 319
force-dynamic structure 87 implication 154-156, 178, 187
Form ICM 36 pragmatic 13
frame 9, 139, 145-155, 157-160, 172- implicature 193, 195, 349, 353-355
175, 178-180, 185, 234, 238, 276, 369 conversational 62, 67, 156-157, 334
generic 79 imprecation 344
frame model/theory 149, 151, 159 inalienable possession 242, 244-245
function of metonymy, pragmatic 12 indexicality 373
indirect object 243
G indirect speech 72
gender change 157-158 indirect speech act 62, 72, 335-336, 344,
generalization 118, 270 347, 351
generic space 77 inference 155, 198-199, 203, 205-206,
genitive 275 345
genus 114 metonymic 62, 355
gestalt psychology 151, 159-160 inheritance hierarchy 315
goal 280, 282, 285 innovation, metonymic 156-157, 174,
goal-over-source principle 282 375
grammaticalization 126, 346-347, 350 input space 77
Grice’s maxims 268 input-topology 87
instrument 246-249, 267-268, 277-281,
408 Subject index
285-286 262, 268, 270, 320

integration (network) 77-78, 81-82, 84, metaphor-metonymy complex (MMC)
89 103-104
irony 289-290, 295, 297-301, 306 metonymy
as abbreviation 128-129, 132
K conceptual 17-18, 182
knowledge structure 317, 328 creative 370
default 22-23, 44, 354
L definition 9, 17, 62, 92, 128, 138, 236
landmark (LM) 216-223, 261, 265, 271 expressive 305
language acquisition 361, 364, 367 illocutionary 335
lexicalization 140, 157, 160, 174, 186, innovative 375
241, 261 insertional 235-236, 240-242, 247
linguistic action 319-320 lexicalized 11, 174, 177
literal language 12 morphological 370
Location ICM 37, 41 non-default 22-23
location/motion schema 285 predicational 335
locative verb 280, 283 propositional 130, 335
range 153
M referential 25, 122-123, 125, 127-133,
manner 277, 279-281 276, 335
manner verb 280-281 reversible 25-27, 38
mapping 78, 256-257, 259, 261-262, substitution view 9, 13
301, 314-315, 329 within metaphor 257, 262, 265
cross-space 77, 84 minimal processing effort 12
metaphorical 22, 316 modal shift 193, 196
metonymic 22, 289 Modification ICM 37, 43
markedness 291, 294-295, 302 modifier 216-217, 219-220
maxim morphology 216
of conversation 68 motion schema 280, 282-283
of informativeness 115 motivation 211-212, 214, 224, 260, 265,
of quantity 352-353 289, 302, 306
maximal contextual effect 12 competing 51
meaning extension 257, 266, 272
mental N
bridge 21 naming strategy 225
motion 283 necessity 197-198, 201-207, 343
mention theory 298 inferable 196, 202
meronomy 93 network, conceptual 326
metaphor 62, 80, 130-131, 133, 139, nominalization 37
143-144, 146, 153, 160, 170-172,
183-184, 187, 225, 255-257, 259, O
262, 269, 272, 301, 309, 314, 316- object 279, 281
317, 329, 334, 362, 365-369, 373, 379 obligation 196, 343
metaphor from metonymy 193-195, 257, oblique object 240
Subject index 409
obstacle potentiality 343

potential 72 predicate 235, 238, 246, 249, 259, 261,
projected 73 266
onomasiological 150, 153, 159 transfer 10, 128
onomastics 211 predication 263, 270-271
ontological realm 21-23 preferred
opposition 229, 289, 291, 294-296, 301, directionality 29
305 vehicle 44-52
optimality preposition 216-219, 239, 248
constraint 89 pretense theory 298
principle 84 process 204
overextension 362, 366-369, 372 Production ICM 36, 39
metaphorical 368 projection constraint 85
metonymic 368-369 proper name 187
synecdochical 365, 368 prototype 149-151, 153, 175, 186, 258,
overinclusion 367 261, 264, 266-268, 271
overstatement 264-265 effect 66, 150, 160
prototypical member 150
P psychological reality 318
part 92-93
of an ICM 30, 36 R
partonomy 92-95 recipient 285
path 282, 285 recontextualization 388
patient 277, 279-281, 285-286 Reduction ICM 36
perception 313 reference 234, 275
Perception ICM 36, 38 contiguity-based 235
perception theory 314 ICM 23-24, 28
perceptual selectivity 46 metonymy 23-24, 279
perfect (aspect) 201, 204 metonymic 233-234, 236, 242-243,
performative verb 343-344 250, 385, 388, 397
permission 343 reference point 9-11, 174, 275-276
person perception psychology 313 expression 10-11
politeness (principle) 295, 354, 356 phenomenon 9, 19
polysemy 140, 158, 235-241, 247-250, Reference-via-Meaning ICM 25
361, 366 referencing 276
contiguity-based 248-249 referentiality in narrative fiction,
metonymic 139-140, 142, 144, 157- metonymic 385
158, 186 reflexive (clitic) 241-244, 250
popular etymology 144 Relational Grammar 245
possession 285 relationship, conceptual 319-320
ICM 37, 40 relevance 50-51
schema 285 Relevance Theory 13
possessive 241, 245 remotivation 373
possessor 245, 285 Request Scenario 346, 350
ascension 245-246 Restaurant ICM 51
410 Subject index
result 277 speech act scenario 344, 355

rhetoric 179, 185, 362 state-of-affairs scenario 337
role, semantic 233, 239, 280, 285 stereotype, social 312, 325, 328
roles, contiguous 235, 241 stereotypical property 35
stimulus 285
S strategy
salience 150-151, 154, 159, 174, 176, metonymic 227
215, 220, 260, 263, 275, 278, 282, pragmatic 366
289, 291, 294-295, 300, 313 strengthening, pragmatic 156, 159
relative salience 44 stupidity 309-310, 312, 325, 328
scale 270, 272, 293, 301, 306 subject 233, 235-236, 241, 243-245,
Scale ICM 31 247-250, 279
scenario 9, 69, 172-173, 178-179, 336 subjectification 199, 207-208
scene 173, 234, 267 subordinate-level term 281
schema 9, 146 succession/successive relation 161, 169,
schematicity 328 178-179, 181-182, 184, 187
schematization 315 surname 211-216, 218, 220-221, 223-
script 9, 69, 234, 234 225
selection restriction 235, 241 synecdoche 63, 91-92, 113, 153-154,
semiotic triangle 23 179, 185, 215, 362, 379
Sense Perception Scenario 339 synecdochic detail 386, 389, 391, 396
sensualist philosophy 141
shift T
metonymic 9-10, 13, 242, 244 taboo 53, 267
referential 10-11, 258, 317 target 9-11, 19, 21-22, 29, 130-132, 258,
shortcut, referential 12 261, 282, 295, 303-304, 310, 320, 334
shrinking domain 269-270
creative metonymical 362, 369-370, metonymic 24
375 tautology 35, 73
metonymic 367, 373 taxonomy 92-94, 212, 224-225, 367
sign 21 tension 263-265, 271
Sign and Reference ICM 37, 42 thematic role 235, 238, 247-250 (also
sign and reference metonymy 28 see case role, role, semantic)
Sign ICM 23-24, 28 theme 233, 248-249
sign metonymy 23-24, 29 Thing-and-Part ICM 31
similarity 141-144, 153, 171-172, 214, topology 85, 87
317, 320, 365, 367 trajector 261
slang 290, 295-296, 300, 304-305 trajectory 282
source 261, 280, 282, 285, 310, 320 transfer schema 285
source domain 270, 316, 326, 328 trigger 130-132
specialization 119 metonymic 373
species 114 trope
specification of landmark 216 metonymic 154, 157
speech act metonymy (also see me- rhetorical 139-141, 156, 317
tonymy, illocutionary) two-valued orientation 299, 301, 306
Subject index 411
typology of metonymy 173, 176, 178, vehicle 19, 21-22, 29, 334, 354
182, 185, 363 vehicle-to-target route 52
violation of cognitive and communica-
U tive principles 52-54
understatement 303
unpacking 85 W
web 85
V Whole ICM and its part(s) 30
vagueness 266 Whole-Part ICM 36
valence semantics 177 word formation 139, 153, 157-159, 187
value judgment 313 world knowledge 173-174 (see also
variation 261 encyclopaedic knowledge)
Author index
A Broca 171
Abbott, Black, and Smith 69 Brown 160
Aitchison 292 Bruner 313
Albee 298, 303 Bühler 160
Amin 160 Bunyan, John 389
Anderson 239 Busse 236
Andor 173 Bybee 195
Arance 185 Bybee et al. 193
Aristotle 1, 141, 160, 179, 362
Auden, W.H. 63 C
Clark 70, 71, 364, 365, 372, 373, 375
B Clark and Clark 108, 125, 280
Bach 211, 215, 226 Clark and Gerrig 71, 298
Bach, Adolf 211, 225 Clark, Eve 366
Bach and Harnish 342 Coates 197
Bain 141 Cooper 117
Baker 239 Cooper and Ross 45, 57
Balzac, Honoré de 61, 63 Copestake and Briscoe 133
Barrett 367 Cordier 146, 160
Barsalou 146, 249 Coseriu 160, 170
Bartlett 146 Cottle 216, 225, 226
Belletti and Rizzi 243, 250 Coulson 77, 78
Bergsten 125 Croft 19, 56, 62, 66, 153, 154, 173, 174,
Black 160, 396 186, 233, 238, 257, 258, 259, 263, 302,
Blake 245, 246 317, 319, 329, 334
Blank 141, 144, 145, 150, 159, 160, 170, Cruse 93, 100, 101, 117, 291, 292, 294,
171, 172, 186, 187, 233, 236 295
Bombeck 70
Bonhomme 145, 178, 179, 186 D
Bower, Black, and Turner 69 Darmesteter 141
Bowerman 364 Davies 311
Braine 367, 368 De Bleser and Bayer 185
Bréal 141, 159, 235, 361 Debus 226
Bredin 62, 105, 139, 145, 146, 154, 160, Deese 291, 292
169, 171, 177, 187 Dik 150, 173, 208
414 Author index
Dirven 56, 117, 280, 317, 320, 328, 331, Gibbs and McCarrell 73
389 Gibbs and Tenney 69
Dorward 215, 216, 225, 226 Givón 247, 248, 294
Dubois 144 Goossens 63, 103, 122, 193, 195, 209,
Dubost 236 256, 257, 258, 259, 269, 272, 317, 319,
Duchácek 169, 177 320
Dumarsais 177, 185 Goossens et al. 193
Gottschald 211, 223
E Graesser, Woll, Kowalski, and Smith 69
Eco 145 Grice 50, 57, 67, 68, 115, 121, 157, 268,
Eemeren et al. 318 335, 352
Ekman et al. 304 Grimshaw 243
Eliot, T.S. 63 Groupe de Liège 1
Esnault 362, 363 Gurd and Marshall 185
Esper 292
H
F Halliday 121, 129
Fass 133, 177, 248, 363 Happ 142, 143, 145, 160, 172
Fauconnier 77, 133, 186 Haspelmath 155
Fauconnier and Turner 77, 84 Hawkins 250
Fauconnier, Gilles 187 Hayakawa 299
Fillmore 146, 150, 160, 173, 177, 186, Heider 313, 330
239, 249, 266 Heine 193
Fleischer 226 Hemenway 30
Fontanier 169, 170, 177, 185 Herennium, auctor ad 140-141, 170, 185,
Francik and Clark 72 362
Fraser 342 Herskovits 56, 104
Frawley 24, 56 Hjelmslev 160
Frei 245 Hobbes 141
Fromkin 292 Hoebel 387
Holenstein 142, 143, 160
G Horn 352
Garavelli 170, 185 Hornby 300
Garman 292 Householder 361
Gauger 158 Huber et al. 185
Geens 260 Hume 141
Geeraerts 118, 150, 160, 321 Huttenlocher and Smiley 366
Geeraerts and Grondelaers 330
Geiger 322 I
Genette 62 Ikegami 282
Gernsbacher 68
Gerrig 70 J
Gibbons 387 Jackendoff 133, 233
Gibbs 1, 8, 18, 33, 35, 56, 62, 72, 73, 121, Jäkel 225, 226, 227
122, 364, 372, 374, 379, 385, 389, 397 Jakobson 62, 117, 143, 144, 160, 171,
Author index 415
172, 186, 386, 387, 396, 398 Leech and Short 389
Johansson 260 Lehrer 292
Johnson 183, 186, 255, 256 Leisi 125
Jongen 150 Leitner 125, 280
Joyce, James 63 Lejeune 208
Leopold 367
K Levi 125
Kainz 300, 306 Levin 62
Kastovsky 280 Liebert 160
Keller 176 Lindner 104
Kleiber 10, 12, 150, 160 Lipka 125
Koch 141, 145, 150, 155, 158, 159, 160, Locke 141
161, 173, 175, 186, 238, 250 Lodge 62, 63, 398
Koch and Oesterreicher 161 Lyons 117, 226, 292, 300, 306, 323
Köhler 160
König 195 M
König and Haspelmath 245, 246 Major 290, 296, 297, 300, 305
König and Traugott 156, 159 Marchand 279
Kövecses 3, 48, 49, 80, 81, 83, 121, 330 Martin 144
Kövecses and Radden 334, 354 Matthews 213, 226, 227
Kövecses and Szabó 57 Mellenius 125
Kruszewski 142, 143 Metzger 160
Kytö 209 Mill, James and John St. 141
Miller 396
L Minsky 146, 173
Labov 386 Morgan 57
Lakoff 2, 3, 19, 20, 25, 27, 33, 35, 38, 39, Morrison, Toni 385, 397
49, 50, 56, 57, 66, 67, 80, 81, 83, 94, Müller-Gotama 250
95, 115, 150, 186, 226, 233, 255, 256, Murphy and Andrews 292
257, 259, 290, 291, 294, 295, 310, 312,
313, 314, 315, 316, 317, 330, 334 N
Lakoff and Johnson 1, 2, 18, 19, 26, 40, Naumann 211, 215
51, 56, 64, 94, 95, 105, 154, 160, 169, Nerlich 117, 118, 141, 160, 185, 379
175, 177, 183, 226, 255, 256, 259, 271, Nerlich and Clarke 373, 379
321, 363 Norrick 34, 35, 41, 227, 363, 387
Lakoff and Turner 24, 34, 42, 56, 95, 385, Norvig and Lakoff 39
389 Nunberg 10, 12, 70, 122, 128, 133, 233,
Landis, Herrmann and Chaffin 291 234
Langacker 9, 19, 30, 31, 44, 56, 57, 96, Nyrop 141, 185, 363
109, 126, 127, 160, 174, 176, 186, 207,
233, 238, 239, 250, 259, 260, 261, 275, O
295, 310, 317, 328, 330, 388 Ogden and Richards 23, 26, 110
Larkin, Philip 64 Olsen 239
Lausberg 141, 154, 170, 177, 185 Ortony 389
Le Guern 145, 154, 185, 186, 233 Orwell, George 63
416 Author index
P Steen 389
Panther and Thornburg 335 Stein, Gertrude 73
Paretsky 113, 118 Stempel 161
Paul 141 Stern 133, 177, 363
Pauwels 268, 269 Stimson 376
Pauwels and Simon-Vandenbergen 260 Sweetser 5, 193, 353, 356
Pease et al. 366
Piaget 366 T
Propp 386 Tagiuri 330
Pustejovsky 104 Talmy 37, 104
Pustejovsky and Bouillon 104 Tannen 146, 249
Taylor 27, 55, 104, 105, 115, 150, 160,
Q 183, 227, 256, 257, 317
Quintilian 177 Thornburg and Panther 33, 333, 336,
344, 350, 355
R Thorne 290
Raible 144, 160, 186 Traugott 193, 195
Rastier 160 Traugott and König 156, 195, 199, 207
Reaney 211, 213, 216, 223, 225 Turner 77
Reddy 104 Turner and Fauconnier 77, 83
Reisig 141 Tversky 30, 93
Rice 331 Tversky and Hemenway 46, 93
Riffaterre 8, 386, 387, 388, 389, 391, Tylor 26
396, 397
Rosch 66, 160 U
Roudet 142, 143, 144, 160, 171, 172 Ullmann 49, 144, 154, 169, 170, 171,
Ruwet 185 177, 363
Ungerer and Schmid 121
S
Sandra 331 V
Sappan 363 Vosniadou 378
Saussure 142, 143, 160 Vygotsky 366
Schank and Abelson 69, 249
Schifko 154, 160, 161, 169, 177 W
Schmid 317 Warren 110, 117, 122, 123, 125, 133,
Schofer and Rice 1 150, 154, 171, 177, 178, 187
Schulz 396 Weinrich 104, 146
Schwarze 161, 187 Wells 110
Searle 62, 335, 336, 344, 345, 350 Wentworth and Flexner 290, 292, 295,
Seto 367 296, 297, 300, 302, 305
Shakespeare 64 Wernicke 171
Smith 213, 223, 225, 227 Werth 159, 398
Sorensen 110 Wertheimer 160
Spanoghe 245 Wierzbicka 93
Sperber and Wilson 12, 50, 176, 297, 298 Wiktorin 226
Author index 417
Wimmer 227 Y
Winston 239 Yamanashi 363
Winston et al. 100
Witkowski 226 Z
Wittgenstein 151 Zandvoort 279
Wundt 142, 303-304
Metonymy and metaphor index
A artist for his work 39, 51

ability to act for action 347 aspect – attribute 178
ability to perceive for actual perception author – product 184
339
ability to process for actual mental B
process 341 basic for non-basic 49
abstract for concrete 112 basic over non-basic 49
act of forming a percept for percept 39 blessed – stupid 171
action for agent 37 bodily over actional 46
action for instrument 37 bodily over emotional 46
action for object 37 bodily over mental 46
action for result 37 bodily over perceptual 46
activity – affected object 177, 184 body – part 245
activity – aim 182 bounded over unbounded 48, 51, 57
activity – consecutive activity 182
activity – object 180 C
activity – place 180 category for a member of the category
activity – product 177 27, 34
activity – result 182 category for defining property 35
activity – time 180 category for salient property 36
activity – time when the activity occurs cattle – money 171
184 cause – aim 182
activity – actor 125 cause – instrument 177
actor – object 180 cause for effect 22, 38, 43
actor – typical aspect 180 cause – effect 317, 372, 373
actor (inventor, etc.) – product 187 causer – result 124
actual for potential 33, 47, 57 central over peripheral 49, 53
actuality over potentiality 354 change of state is movement of pos-
agent – activity 177 sessed object 330
agent for action 37 clear over less clear 354
agent for instrument 187 clear over obscure 50, 52, 53, 55, 57
agent to result 111 collective body – state 181
agent – action 234 common over less common 49
agent – instrument 247 communication – human beings 153
anger is heat 80, 81, 82 comparant – compared 124
antecedent-consequent 129, 130, 132 concept for form 24
420 Metonymy and metaphor index
concept for its opposite 290 frame – frame 182

concept for thing/event 25, 46 function – object 160
concrete over abstract 46, 51, 52, 53 functional over non-functional 46
container for contents 29, 41, 43, 46,
125, 176 G
container – contents 103, 104, 234, 239, garment for person 125
240, 241, 317 generic for specific 34
contents for container 29, 41, 44 generic is specific 35
controlled for controller 26, 40, 44 genus for species 22
controller for controlled 27, 40, 44, 45, genus to species 114, 115
65, 271, 321 genus – species 113, 367
controller – controlled 105 goal – instrument 125
counter – public house 154 good gestalt over poor gestalt 47, 51, 53
D H
defining property for category 35 hamburger for customer 277
destination for motion 37 human over non-human 45, 51, 53, 57
deviant behavior for stupidity 324 human to non-human 11
disposition for occasional behavior 348
dominant over less dominant 47 I
ideal over non-ideal 48
E immediate over non-immediate 47, 51,
effect for cause 22, 38, 39, 43, 321, 333 57
effect for core 338 important over less important 49
emotion for cause of emotion 39, 47 important utensil for person 215, 224
ensuing situation to preceding situation inhabitants for place 41
108, 111 initial or final over middle 49, 53
ensuing situation to process 108, 109 institution for people responsible 271,
evaluative concept for its opposite 299 322
event for place 42 institution for place 28, 41
event – subevent 106 instrument – actor 181
events are actions 314 instrument – product 182
instrument for action 37
F instrument for agent 37, 187
face for the person 18, 19 instrument for product 40
fire – fireplace 154 instrument for result 125
form for concept 24, 42, 46 instrument to result 110
form for content 333 instrument/organ of perception for the
form for thing/event 26 perception 38
form-concept for thing/event 24, 46 interactional over non-interactional 46, 51
form-conceptA for conceptB 27 inventor – product 180
formA-conceptA for formA-conceptB 27 inventor for the thing invented 39, 176
formA-conceptA for formB-conceptA
28, 36 K
formA-conceptA for formB-conceptB 26 knowing is seeing 319
L organization – member 99
life is a journey 314, 316, 330 origin – object 125
living things – human beings 153 outgroup origin for stupidity 49, 311
M P
manner for action 37, 43 pars pro toto 154
manner of perception for the perception part – part 181
38 part – whole 181, 184
material – product 182 part for part 36
material constituting an object for the part for whole 31, 36, 43, 46, 84, 86, 106,
object 32, 48 113, 125, 259, 261, 271, 321, 389
material to result 110 part of a form for the whole form 36
matter for artifact 125 part of a thing for the whole thing 31
means for action 37 part – part 185
member of a category for the category part – whole 124, 179, 185, 234, 239,
27, 34, 42 243, 244, 245, 250, 344, 362
mental/physical state for object/person passing of time is moving entity 315
causing it 39 people for their possessions 72
messenger – angel 171 perception for result of perception 319
modified form for original form 43 perception for thing perceived 38, 45
more over less 47 period – period 182
person for the face 19
N physical/behavioral effect for emotion
necessity for motivation 333 causing it 39
new over old 50 place – object 181
place – place 182
O place (of origin or residence) for person
object – collective body 181 216, 224
object – function 180 place for activity performed at that place
object – typical aspect 180 42
object – place 160 place for event 42, 65, 271, 322
object for action 37 place for inhabitants 41
object for material constituting the object place for institution 28, 41, 65, 72 125,
32, 48 271, 322
object for owner 387 place for an institution located at that
object for user of the object 40 place 65, 72
object used for user 65, 271, 321 place for product made there 40
object – component 99 place for the event 65, 322
object – material 99 place for the institution 322
object – origin 125 place to result 111
object – place 124 place – object 124
object – property 317 positive end of a scale stands for the
objects used for their users 72 whole scale 294
occurrent over non-occurrent 47 positive for negative 304
ordered dish – customer 183 possessed for possessor 40, 41
422 Metonymy and metaphor index
possessor for possessed 40, 45 that it is done 39

possessor for possession 388 situationally more relevant over situ-
potential for actual 34 ationally less relevant 51
potential to actual 13 skill for display of skill 349
potentiality for actuality 33, 333, 335, sound for event causing it 39, 43
336, 338, 339, 340, 341, 342, 343, source – result 124
344, 345, 347, 348, 350, 351, 352, species for genus 22, 379
353, 354, 355 species to genus 114
potentiality for motivation 333 species – species 367
preceding situation to ensuing situation specific case for general rule 35
108, 110 specific for generic 34, 35, 48
preceding situation to process 108 specific over generic 48, 53, 57
preceding situation – process – ensuing state – place 180
situation 107 state/event for thing/person/state causing
preceding – ensuing 107 it 39
precondition – activity 182 states are locations 316
present for future 33, 47 states/attributes are things 314
present for habitual 33, 47 stereotypical over non-stereotypical 48
previous state – consecutive state 184 subevent for whole event 32, 48, 106
process of production – product 184 subject – set 124
process to ensuing situation 108, 109 subjective over objective 45, 52
process to preceding situation 108, 109 substitute form for original form 43
process to result 109
producer for product 39, 40, 45, 55, 72, T
125, 271, 321 thing perceived for perception 38
product for instrument 40 thing/event for concept 25
properties are possessions 35 thing/event for form 26
property for object 112 thing/event for form-concept 25
time for action 37
R time passing is moving entity 330
rare over less rare 49 times are locations 314, 316
relevant over irrelevant 51, 53 times are moving objects 314
result 262 times are things 314
result – cause 182 tool – object 181
result for action 37, 38 totum pro parte 154
result to preceding situation 110 traditional over non-traditional 50
result to process 110 typical aspect – activity 160, 180
result – causer 124 typical aspect – frame 160, 180
typical aspect – object 160
S typical over non-typical 49
salient property for a less salient one 311
salient property for category 36 U
salient quality for person 215, 224 unexpected over expected 50
seeing something done for making sure upper end of a scale for whole scale 32
V
virtuality for actuality 343 whole scale for upper end of the scale 32
visible over invisible 46, 51 whole thing for a part of the thing 31, 48
whole – part 99, 103, 104, 113, 124,
W 179, 185
whole event for subevent 32 words for the concepts they express 24,
whole event – subevent 106, 107 42, 46
whole for part 31, 35, 50, 113, 125, 255, 271
In the series HUMAN COGNITIVE PROCESSING (HCP) the following titles have been
published thus far or are scheduled for publication:
1. NING YU: The Contemporary Theory of Metaphor. A perspective from Chinese. 1998.
2. COOPER, David L.: Linguistic Attractors. The cognitive dynamics of language acquisition
and change. 1999.
3. FUCHS, Catherine and Stéphane ROBERT (eds.): Language Diversity and Cognitive
Representations. 1999.
4. PANTHER, Klaus-Uwe and Günter RADDEN (eds.); Metonymy in Language and Thought.
1999.
5. NUYTS, Jan: Epistemic Modality, Language, and Conceptualization. A cognitive-
pragmatic perspective. 2001.
6. FORTESCUE, Michael: Pattern and Process. A Whiteheadian perspective on linguistics.
2001.
7. SCHLESINGER, Izchak, Tamar KEREN-PORTNOY and Tamar PARUSH: The
Structure of Arguments. n.y.p.
8. SANDERS, Ted, Joost SCHILPEROORD and Wilbert SPOOREN (eds.): Text
Representation. Linguistic and psycholinguistic aspects. n.y.p.

1panter Klaus Uwe Radden Gunter Editors Metonymy in Language PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

1panter Klaus Uwe Radden Gunter Editors Metonymy in Language PDF

Caricato da

Copyright:

Formati disponibili

Metonymy in Language and Thought

HUMAN COGNITIVE PROCESSING is a forum for interdisciplinary research on the

EDITORIAL ADVISORY BOARD

Klaus-Uwe Panther and Günter Radden (eds)

Metonymy in Language and Thought

JOHN BENJAMINS PUBLISHING COMPANY

American National Standard for Information Sciences — Permanence

Library of Congress Cataloging-in-Publication Data

The present volume evolved from a workshop on metonymy held at Hamburg

Part I: Theoretical Aspects of Metonymy

Towards a Theory of Metonymy 17

Speaking and Thinking with Metonymy 61

Metonymy and Conceptual Integration 77

Distinguishing Metonymy from Synecdoche 91

Aspects of Referential Metonymy 121

Part II: Historical Aspects of Metonymy

Frame and Contiguity: On the Cognitive Bases of Metonymy and

Co-presence and Succession: A Cognitive Typology of Metonymy 169

Metonymic Bridges in Modal Shifts 193

Metonymy in Onomastics 211

Part III: Case Studies of Metonymy

Grammatical Constraints on Metonymy: On the Role of the Direct

Putting Metonymy in its Place 255

Conversion as a Conceptual Metonymy of Event Schemata 275

Opposition as a Metonymic Principle 289

Metonymic Hierarchies: The Conceptualization of Stupidity in

The Potentiality for Actuality Metonymy in English and Hungarian 333

Part IV: Applications of Metonymy

“Mummy, I like being a sandwich”: Metonymy in Language

Recontextualization of Metonymy in Narrative and the Case of

List of Contributors 401

Subject index 405

Author index 417

Metonymy and metaphor index 423

Klaus-Uwe Panther and Günter Radden

The authors of the contributions to this volume have different theoretical

2. Contributions to the volume

2.1. Theoretical aspects of metonymy

may, however, be overridden by factors such as style, taboo or politeness and

temporal contiguity between entities in the physical world, are called E-

2.2. Historical aspects of metonymy

The outstanding role of metonymy in triggering linguistic change has long

typology of metonymy” critically examines the classifications of metonymy

2.3. Case studies of metonymy

A variety of case studies investigate the operation of metonymy on various

tonymies (Dirven; Pauwels; Voßhagen); one paper studies the metonymic

2.4. Applications of metonymy

Two contributions are devoted to the significance of metonymy in language

3. Perspectives for future research

The papers collected in this volume certainly contribute to a deeper under-

still poorly understood. Of the many remaining problems to be solved we will

3.1. The nature of metonymic shift

highlighting is reflected in grammar. There are, of course, other grammatical

3.2. The pragmatic function of metonymy

An issue which has received relatively little attention in the discussion of

Gibbs, Raymond W., Jr.

Theoretical Aspects of Metonymy

1. The cognitive view of metonymy

The aim of this paper is to propose a conceptual framework of metonymy as a

1.1. Metonymy is a conceptual phenomenon

Metonymy is claimed to be not just a matter of names of things, but essentially

1.2. Metonymy is a cognitive process

The traditional view defines metonymy as a relationship involving substitu-