1uwe Panther K Thornburg L Linda Metonymy and Pragmatic Infer
1uwe Panther K Thornburg L Linda Metonymy and Pragmatic Infer
Associate Editors
Jacob L. Mey
University of Southern Denmark
Herman Parret
Belgian National Science Foundation, Universities of Louvain and Antwerp
Jef Verschueren
Belgian National Science Foundation, University of Antwerp
Editorial Board
Shoshana Blum-Kulka Catherine Kerbrat-Orecchioni
Hebrew University of Jerusalem University of Lyon 2
Jean Caron Claudia de Lemos
Université de Poitiers University of Campinas, Brazil
Robyn Carston Marina Sbisà
University College London University of Trieste
Bruce Fraser Emanuel Schegloff
Boston University University of California at Los Angeles
Thorstein Fretheim Deborah Schiffrin
University of Trondheim Georgetown University
John Heritage Paul O. Takahara
University of California at Los Angeles Kansai Gaidai University
Susan Herring Sandra Thompson
University of Texas at Arlington University of California at Santa Barbara
Masako K. Hiraga Teun A. Van Dijk
St.Paul’s (Rikkyo) University Pompeu Fabra, Barcelona
David Holdcroft Richard J. Watts
University of Leeds University of Berne
Sachiko Ide
Japan Women’s University
Volume 113
Metonymy and Pragmatic Inferencing
Edited by Klaus-Uwe Panther and Linda L. Thornburg
Metonymy and
Pragmatic Inferencing
Edited by
Klaus-Uwe Panther
Linda L. Thornburg
University of Hamburg
List of contributors ix
Acknowledgments xi
List of contributors
Acknowledgments
K.U.P.
L.L.T.
Hamburg, April 2003
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.1 (1)
Introduction
On the nature of conceptual metonymy
tributors to this book. Nevertheless, throughout the book the ‘x for y’ notation
will be maintained because it has become an established convention in cogni-
tive linguistics. The use of small capitals is meant to reflect the assumption
that metonymy is a relation between concepts, rather than between real-world
denotata or referents.
On the basis of George Lakoff ’s (1987) and Ronald Langacker’s (1993) work,
which emphasizes the conceptual nature of metonymy, Günter Radden and
Zoltán Kövecses (1999: 21) have proposed a widely accepted characterization of
metonymy: “Metonymy is a cognitive process in which one conceptual entity,
the vehicle [also often called the ‘source’, KUP/LLT], provides mental access to
another conceptual entity, the target, within the same cognitive model.”
In what follows, we adopt this definition as a convenient starting point for
our discussion of metonymy. However, we see a need to constrain the scope of
this definition somewhat because it covers some linguistic phenomena that are
arguably very different from clear cases of metonymy. Consider, for example,
the italicized referential noun phrases in (1) and (2):
(1) The piano is in a bad mood.
(2) The loss of my wallet put me in a bad mood.
In sentence (1) the subject noun phrase has the standard metonymic interpre-
tation ‘the musician playing the piano’, with the meaning of piano providing
mental access to the concept of piano player. In sentence (2), the sense of the
loss of my wallet provides access to the concept of ‘non-possession (of the wal-
let)’. Are we therefore entitled to conclude that the relation between the concept
of loss and that of non-possession is a metonymic relationship, just as the re-
lation between the concept of piano and that of piano player is metonymic?
Intuitively, the answer seems ‘no’; and in fact, there is an important differ-
ence between the two cases. In sentence (2) the relationship between ‘loss’ and
‘non-possession’ is conceptually necessary, i.e., the proposition presupposed by
the referring expression in (1), ‘I lost my wallet at time t’, entails ‘I did not
have my wallet for some time span beginning at time t’. In sentence (1), the
relationship between the piano and the piano player is contingent; the presup-
position ‘There is a piano’ does not entail ‘There is a piano player’. In other
words, there is no metonymy loss for non-possession, but there is an often
exploited metonymy musical instrument for musician.
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.4 (4)
where the company name is used to refer to the automobile workers who walk
out of the work place. Pragmatically, metonymies are also found on the pred-
icational, propositional (where referential and predicational metonymies oc-
cur in combination), and illocutionary levels, respectively. An example of a
predicational metonymy is
(4) General Motors had to stop production.
where the necessity or obligation to stop production stands for the actually
occurring event of stopping production (obligation to act for action). The
metonymy involved is an instance of a high-level metonymic principle that is
very common in English and other languages: A potential event (e.g. the ability,
possibility, permission, obligation to undertake an action) is metonymically
linked to its actual occurrence. Events are conceptualized here as ICMs that
contain as subcomponents the modalities of their realization. Sentence (4) is
also a propositional metonymy because General Motors metonymically refers
(in this case) to the executive officers of the company.
Finally, an example of an illocutionary metonymy is given by the well-
known phenomenon of conventionalized indirect speech acts as in (5a), in
contrast to (5b):
(5) a. I would like you to close that window.
b. Close that window.
where the expression of the wish with regard to the action to be carried out
by the addressee (signaled by would like you to) metonymically evokes the re-
quest (5b) itself (see Gibbs 1994, 1999; Thornburg & Panther 1997; Panther
& Thornburg 1998). The basic idea is that an attribute of a speech act can
stand for the speech act itself in the same way that an attribute of a person
can stand for the person (see also Panther & Thornburg’s and Radden & Seto’s
contributions in this volume).
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.5 (5)
Our contention that the relation between the piano and ‘the piano player’,
on the one hand, and that between General Motors had to stop production
and ‘General Motors stopped production’, on the other hand, is of the same
kind, viz. metonymic, may look surprising at first sight. One might object that
the target meaning of (4) is really an implicature that comes about through
pragmatic strengthening of the proposition expressed in (4).2
However, note first that a metonymic analysis does not preclude a prag-
matic analysis in terms of conversational implicature. On the contrary, we
assume that conversational implicatures, or more generally, pragmatic infer-
ences, are often guided by preexisting metonymic principles.3
Second, and more importantly, referential, predicational, and illocutionary
metonymies share the property of highlighting or foregrounding their respective
target meanings. The source of a metonymy serves as a “reference-point” (see
e.g. Langacker 1993) whose sole purpose is to provide access to a target mean-
ing. That metonymy involves highlighting is a common assumption among
cognitive linguists (see e.g. Croft 1993). To illustrate, consider the following
larger context for sentence (4):
(6) General Motors had to stop production on Monday but they resumed it
on Thursday.
The but-clause in (6) only makes pragmatic sense if the clause General Motors
had to stop production on Monday has the foregrounded target meaning ‘Gen-
eral Motors stopped production on Monday’. Obviously, the backgrounded
source meaning of the first clause in (6) (the ‘obligation’ sense) is still ac-
tivatable, but this holds for standard cases of referential metonymy as well,
i.e., it is a general property of metonymy that source meanings are “active”
to some degree.
Third, the same metonymy can be triggered predicationally and referen-
tially. The obligation to act for action metonymy triggered in (4) and (6)
can also be derived from the nominalized (referential) counterpart of (4):
(7) General Motor’s obligation to stop production had a devastating effect on
the economy.
is often understood as an offer of the speaker to lend her car to the hearer (for a
metonymic treatment of speech acts, see Panther & Thornburg, this volume).
In utterance (9) a proposition analogous to the one in (8) is nominalized and
used referentially.
(9) My willingness to lend you my car surprised everybody.
The referential noun phrase in (9) lends itself quite readily to the (fore-
grounded) target meaning ‘My offer to lend you my car’. We see no reason to
treat the pragmatic implication of the noun phrase in (9) differently from the
target meaning of uncontroversial metonymies as in Table Four wants another
Chardonnay where Table Four stands for ‘the customer sitting at Table Four’.
The link between a metonymic source and its target may vary in strength. The
strength of a metonymic link depends on how conceptually close source and
target are to each other (cf. Panther & Thornburg 1998). The relevance of the
strength factor becomes obvious when an utterance like (5a) – where the con-
ceptual link between the mental attitude literally conveyed by the utterance, i.e.
the speaker’s wish that the addressee perform the action, and the targeted ac-
tual request itself is very strong – is compared to the relatively weak metonymic
connection between the contents of (10a) and (5b) (repeated below as (10b)):
(10) a. There’s a draft in here.
b. Close that window.
tent of the assertion (10a) and q stands for the propositional content of the
request (10b).
. Summary
is used in linguistic communication, the metonymic link between the two con-
cepts can be explicitly canceled without contradiction: She is a mother of two
daughters but she is not a housewife is semantically well-formed. A meaning
that, in cognitive linguistic terms, is stereotypically evoked via metonymy (see
Radden & Kövecses 1999: 27) or, in neo-Gricean parlance, via a generalized
conversational implicature, is generally not expressed through a separate lexi-
cal item; e.g., there are no simple lexemes for the concepts housewife mother
or alcoholic beverage in English, and, in fact, it would be redundant to have
such lexemes because their senses are easily accessible via metonymically based
conversational implicatures.
Regrettably, so far, there has been relatively little dialogue between schol-
ars working in a neo-Gricean framework and cognitive linguists, although the
objects of inquiry and even the analyses proposed do not seem incompatible to
us but point to possible avenues of convergence.
The contributions to this volume have been organized into four parts. Part I
is concerned with defining the role of metonymy in inferential utterance inter-
pretation (Ruiz de Mendoza & Pérez Hernández, Barcelona) and conceptual
blends (Coulson & Oakley). Part II focuses on the metonymic motivation of
grammatical structure (Stefanowitsch, Panther, & Thornburg, Köpcke & Zu-
bin). Part III explores the role of metonymic inferencing in linguistic change
(Ziegeler, Okamoto). Part IV closes the book comparing the exploitation
of metonymies from a cross-linguistic perspective (Radden & Seto, Brdar &
Brdar-Szabó).
The first three chapters of the volume demonstrate the significance of metonymy
as a conceptual tool for guiding inferencing in language and other cognitive
domains. Ruiz and Pérez’s paper sets the stage relating work on metonymy in
cognitive linguistics to relevance theory; Coulson and Oakley’s and Barcelona’s
papers are case studies that show the power of metonymic principles in con-
ceptual integration and the interpretation of humorous discourse, respectively.
In the first chapter “Cognitive Operations and Pragmatic Implications,”
Francisco Ruiz de Mendoza and Lorena Pérez Hernández link work on con-
ceptual metonymy in cognitive linguistics to Gricean pragmatics and rele-
vance theory. The authors start with the assumption generally accepted in cog-
nitive linguistics that metonymy and metaphor are tools for understanding
and reasoning about the world. They reduce metonymy to two basic types:
metonymies where the target concept is part of the source concept (target-
in-source metonymies) and metonymies where the source is part of the tar-
get (source-in-target metonymies). Relying on recent work by Papafragou
and Carston, Ruiz and Pérez argue that metaphor and metonymy are part
of what is said, rather than what is implicated – in contrast to previous
relevance-theoretic and Gricean analyses. However, the authors strongly object
to Carston’s idea that metonymy and metaphor are “loose” ways of speaking
with the principle of relevance as sufficient to account for their interpretation.
Rather, Ruiz and Pérez propose that the principle of relevance must be sup-
plemented by metaphoric and metonymic mappings, i.e. cognitive operations
available to speakers and hearers that are part and parcel of the their seman-
tic and conceptual knowledge. The authors also argue for a view of metaphor
and metonymy as a continuum. They present an interesting new analysis of
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.11 (11)
Barcelona, presents four case studies of the humorous mechanisms of jokes and
anecdotes. The interpretation of these genres requires – as has been pointed out
by a number of humor theorists like Attardo (1990) and Raskin (1985) – com-
plex inferential work on the part of the hearer. Barcelona wonders how it is
possible that listeners often arrive at the intended humorous reading of a joke
or anecdote at “lightning speed.” For him, this feat cannot be explained on
the basis of Gricean maxims alone (or their variants in the discourse world of
humor). Adopting a conception of metonymy that is inspired by Radden and
Kövecses (1999) (cf. Section 2.1 above) Barcelona proposes that in many if not
all cases the inferential work is facilitated by pre-existing metonymic connec-
tions in a cognitive frame or domain or by pre-existing metaphorical connec-
tions across frames. Metonymies thus help achieve the frame adjustment (cf.
Attardo 1990; Raskin 1985) that is necessary in order to grasp the punch line
of a joke or anecdote. Barcelona shows that the value of metonymy for prag-
matic inferencing can be appreciated only if one discards the traditional view of
conceptual metonymy as a purely referential phenomenon. As to the question
whether pragmatic inferencing can be reduced entirely to metonymic reason-
ing, Barcelona is not willing to commit himself to a wholly affirmative answer,
but he certainly thinks that metonymically based inferencing plays an essential
role in utterance interpretation.
The chapters in Part II of the volume are concerned with the interaction be-
tween metonymy and grammatical structure (see also the contributions in Part
III and Brdar & Brdar-Szabó in Part IV). In the three studies summarized be-
low metonymic origins of the linguistic phenomena under investigation are
still visible, but the metonymic relationship in many cases has become such an
integrative part of grammatical meaning that it is no longer defeasible.
In Chapter 4 “A Construction-Based Approach to Indirect Speech Acts”
Anatol Stefanowitsch offers an account of conventionalized indirect speech
acts (ISAs), specifically, requests such as Will/can you close the door? in terms
of Construction Grammar (see e.g. Goldberg 1995). Using some of Sadock’s
(1974) collocational criteria for conventionalized indirect requests (e.g. the
possibility to insert politeness markers like please, kindly, the conditional
would/could, and preposed reason clauses), Stefanowitsch shows that certain
aspects of conventionalized indirect requests are not predictable from their
form and meaning components and that they therefore qualify as construc-
tions. Stefanowitsch calls them ISA constructions and contrasts them with
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.13 (13)
utterances such as Are you able to close the door?, which can in certain con-
texts be used as an indirect request, but does not qualify as an ISA construc-
tion because the above-mentioned test criteria fail to apply to it. Stefanow-
itsch argues that ISA constructions are completely conventional, but that, de-
spite their partially unpredictable properties, they are motivated metonymi-
cally in the sense of Panther and Thornburg (1998). However, there is no
need for the speaker/hearer to process them metonymically because their prag-
matic function is part of their meaning. The metonymic motivation is cap-
tured in the construction grammar framework by positing metonymic inheri-
tance links between direct questions and conventionally indirect requests. Ste-
fanowitsch’s analysis amounts to postulating that there are two constructions
of the form Can you do A? One of them signifies a question and the other a
conventional request. In the last part of his paper Stefanowitsch tests the pre-
dictions of the construction grammar analysis against the neurolinguistic liter-
ature on indirect requests. Although the psycholinguistic evidence is not con-
clusive in all respects, there are some interesting indications that individuals
with right-hemisphere damage, who generally have trouble recovering non-
literal meaning, have no problems interpreting indirect request constructions
as requests, but they do have trouble interpreting other non-conventionalized
indirect requests as requests.
In Chapter 5 “Metonymies as Natural Inference and Activation Schemas:
The Case of Dependent Clauses as Independent Speech Acts” Klaus-Uwe Pan-
ther and Linda L. Thornburg investigate some if-clauses that look “incom-
plete,” i.e. lack a syntactically realized consequent clause. Many such “trun-
cated” conditional clauses qualify however as constructions in the sense of
Goldberg (1995) because they have a non-compositional conventional sense
associated with them. Panther and Thornburg identify three kinds of conven-
tionalized pragmatic functions of such if-clauses: deontic (involving speaker
commitment or hearer obligation) as in If you would like a cookie (offer) or If
you will come to order (request); expressive (e.g. surprise, shock, etc.) as in Why,
if it isn’t Susan! (expression of surprise); and epistemic (reasoning, expression
of belief) as in If it was a warning (challenge of a prior assumption). Using the
concept of mental space from conceptual blending theory and an approach to
indirect speech acts as conceptual scenarios, Panther and Thornburg make ex-
tensive use of the cancelability test known from Grice’s work on implicature to
determine the degree of conventionalization of the if-clause types they investi-
gate. They argue, quite in line with Stefanowitsch’s analysis of indirect requests
(this volume), that many truncated conditionals do not require any inferential
work on the part of the hearer even though the metonymic motivation of their
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.14 (14)
The contributions of Part III of the volume are concerned with the role of
metonymy and implicature in linguistic change, i.e. with the implicatures and
metonymic inferences associated with modals and their periphrastic counter-
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.15 (15)
ment is the most important part of the message, thereby “bring[ing] about a
certain expressiveness, that is, to perform a given speech act with particular
stylistic nuances [. . . ].” More formally, Okamoto proposes a metonymic infer-
ence that the addressee draws “on the basis of his/her knowledge of certain
frames and understanding of the specific context.” There is thus a metonymic
shift from ‘[S koto]’ to ‘[[P koto] Modality]’ where ‘Modality’ is supposed to
stand for the pragmatic function of the proposition P. For example, the use of
[S koto] as a directive speech act as in Mainichi ha o migaku koto ‘You brush
your teeth every day-koto’ is more indirect; and an exclamation such as Maa
oishii koto ‘Oh, it is delicious-koto’ is “less imposing and ‘feminine’.” The origi-
nal subordinate clause construction [S koto] develops thus into an independent
construction analogous to the if-clause constructions analyzed by Panther and
Thornburg (this volume). The illocutionary function of koto-clauses in these
constructions seems comparable to that of the German clauses introduced by
the complementizer dass in directives such as Dass du das nicht noch einmal
machst! (‘Don’t ever do that again’) or exclamations such as Dass ich das noch
erleben durfte! (‘That I would live to see this’).
The last two contributions to this volume demonstrate that the use of
metonymic principles may vary cross-linguistically and that metonymy in-
teracts with and is constrained by grammatical structure (see also Stefanow-
itsch, this volume; Panther & Thornburg, this volume). Radden and Seto’s
paper is more pragmatically oriented comparing commercial events in English-
speaking and Japanese-speaking cultural contexts; Brdar and Brdar-Szabó ana-
lyze the role of metonymy in coding linguistic action in English, Croatian, and
Hungarian from a typological perspective.
In Chapter 9, Günter Radden and Ken-ichi Seto investigate “Metonymic
Construals of Shopping Requests in have- and be-Languages.” The classifica-
tion into have- and be- languages derives from how the concept of possession
is encoded. have languages include English, German, Lithuanian, and Croa-
tian; be languages are Japanese, Chinese, Korean, Finnish, Hungarian, Polish,
and Hausa. The authors focus especially on the wording of shopping requests
in English and Japanese. An English sentence such as John has two children
would have to be rendered in Japanese as ‘At/To John are two children’. This
structural difference has consequences for how the two languages linguistically
code certain stages in the shopping scenario. Radden and Seto distinguish two
main phases in the shopping scenario: (i) the precondition, i.e. the availability
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.17 (17)
of the requested articles and (ii) the transaction, which is further subdivided
into (a) the transfer of the article from the salesperson to the customer, (b)
the reception of the article by the customer, and (c) the result, i.e. the cus-
tomer’s possession of the article. They then show that the metonymic cod-
ing of the speech acts that characteristically occur during these stages is par-
tially dependent on the structural resources of the language in question (see
also Brdar & Brdar-Szabó for grammatical constraints on metonymy). For ex-
ample, stage (i) of the shopping scenario is typically referred to in both lan-
guages by means of a metonymy, which, in a have-language like English, is
possession for availability (e.g., Do you have 40-watt light bulbs?) and, in a
be-language like Japanese, existence for availability (40 watto no denkyuu
(wa) ari-masu ka ‘Are there 40-watt light bulbs?’). Radden and Seto also point
out that in English a question about the possession and thus (metonymically
induced) availability of an article can stand for the requested transaction itself,
i.e. stage (ii) of the shopping scenario, whereas in Japanese the same pragmatic
function is achieved by means of a question about the existence of the arti-
cle. The authors demonstrate that politeness factors may actually cut across
the typological properties of languages. Thus an English speaker would avoid
a direct expression of stage (iia) (#Give me “The Times”!) whereas in other
have-languages, such as Lithuanian and Croatian, this wording would not be
considered inappropriate; analogously in a be-language like Hungarian the lit-
eral translation of the above would be infelicitous whereas in Japanese a direct
reference to the requested transfer would not sound offensive if it is used in
combination with deference markers.
Chapter 10, Mario Bdrar and Rita Brdar-Szabó’s contribution “Metonymic
Coding of Linguistic Action in English, Croatian and Hungarian,” is a de-
tailed study of English sentences like The President was clear on the matter and
their equivalents in Croatian and Hungarian. What is peculiar about the ver-
bal locution to be clear on some matter is that it is conventionally used to refer
to a speech act (‘to speak clearly on some matter’) where the speech act it-
self is not explicitly coded in the expression but conventionally evoked via a
metonymy manner (of linguistic action) for linguistic action. The de-
gree of conventionalization of this metonymy varies from a strongly implicated
but still cancelable target meaning to complete lexicalization that defies defea-
sibility. Brdar and Brdar-Szabó demonstrate that this predicational metonymy
is much more constrained in Croatian and Hungarian than in English. Thus
the above sentence would have to be rendered in Hungarian as Az elnök vilá-
gosan nyilatkozott ezzel az üggyel kapcsolatban (‘The president spoke clearly on
that matter’). Both Croatian and Hungarian are more likely to explicitly code
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.18 (18)
the linguistic action itself. The authors see a more general typological ten-
dency for these languages to avoid predicational metonymies, whereas refer-
ential metonymies of the type Beijing’s difficulties in Tibet are also systemati-
cally exploitable in Croatian and Hungarian. Bdrar and Brdar-Szabó suggest an
implicational relationship between referential and predicational metonymies:
Languages that systematically exploit predicational metonymies will also make
extensive use of referential metonymies; some languages will be largely re-
stricted to referential metonymies. In fact, Brdar and Brdar-Szabó argue that
cases such as I’ll be brief (without a complement like about NP), which actu-
ally have literal counterparts in Croatian and Hungarian, are really reducible
to referential metonymies of the type speaker for utterance, a subtype of
the more general metonymy agent for action. These cases would thus not
constitute counterexamples to the generalization proposed by the authors.
The authors of this volume share the belief that the study of conceptual
metonymy provides important insights into language use and language struc-
ture. Metonymy appears to be on a par with metaphor as far as its concep-
tual import is concerned. Promising projects for further research on the role of
metonymy in natural language would include a more systematic comparison
of the exploitation of metonymies from a typological perspective (cf. Radden
& Seto, Brdar & Brdar-Szabó, this volume), the role of metonymic thinking in
language acquisition, discourse-pragmatic conditions of metonymic uses, con-
straints on the creation of metonymic links, and, last not least, a hierarchically
organized taxonomy of conceptual metonymies found in human language.
Notes
References
Attardo, Salvatore (1990). The violation of Grice’s maxims in jokes. Berkeley Linguistics
Society, 16, 355–362.
Barcelona, Antonio (Ed.). (2000). Metaphor and Metonymy at the Crossroads: A Cognitive
Perspective [Topics in English Linguistics 30]. Berlin and New York: Mouton de Gruyter.
Croft, William (1993). The role of domains in the interpretation of metaphors and
metonymies. Cognitive Linguistics, 4, 335–370.
Dirven, René & Pörings, Ralf (Eds.). (2002). Metaphor and Metonymy in Comparison and
Contrast [Cognitive Linguistics Research 20]. Berlin and New York: Mouton de Gruyter.
Fauconnier, Gilles & Turner, Mark (1999). Metonymy and conceptual integration. In K.-U.
Panther & G. Radden (Eds.), Metonymy in Language and Thought [Human Cognitive
Processing 4] (pp. 77–90). Amsterdam and Philadelphia: Benjamins.
Gibbs, Raymond W., Jr. (1994). The Poetics of Mind: Figurative Thought, Language, and
Understanding. Cambridge: Cambridge University Press.
Gibbs, Raymond W., Jr. (1999). Speaking and thinking with metonymy. In K.-U. Panther &
G. Radden (Eds.), Metonymy in Language and Thought [Human Cognitive Processing
4] (pp. 61–76). Amsterdam and Philadelphia: Benjamins.
Goldberg, Adele (1995). Constructions: A Construction Grammar Approach to Argument
Structure [Cognitive Theory of Language and Culture]. Chicago and London: The
University of Chicago Press.
Grice, H. Paul (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 41–58). New York: Academic Press.
Lakoff, George (1987). Women, Fire, and Dangerous Things: What Categories Reveal About
the Mind. Chicago and London: The University of Chicago Press.
Lakoff, George & Johnson, Mark (1980). Metaphors We Live By. Chicago and London: The
University of Chicago Press.
Langacker, Ronald W. (1993). Reference-point constructions. Cognitive Linguistics, 4, 1–38.
Levinson, Stephen C. (1995). Three levels of meaning. In F. Palmer (Ed.), Grammar and
Meaning (pp. 90–115). Cambridge: Cambridge University Press.
Levinson, Stephen C. (2000). Presumptive Meanings: The Theory of Generalized Conver-
sational Implicature. Cambridge, MA, and London: The MIT Press.
Marmaridou, Sophia S. A. (2000). Pragmatic Meaning and Cognition [Pragmatics and
Beyond New Series 72]. Amsterdam and Philadelphia: Benjamins.
Norrick, Neal R. (1981). Semiotic Principles in Semantic Theory. Amsterdam and
Philadelphia: Benjamins.
Panther, Klaus-Uwe & Radden, Günter (Eds.). (1999). Metonymy in Language and Thought
[Human Cognitive Processing 4]. Amsterdam and Philadelphia: Benjamins.
Panther, Klaus-Uwe & Thornburg, Linda L. (1998). A cognitive approach to inferencing in
conversation. Journal of Pragmatics, 30, 755–769.
Panther, Klaus-Uwe & Thornburg, Linda L. (forthcoming). Metonymy. In D. Geeraerts & H.
Cuyckens (Eds.), Handbook of Cognitive Linguistics. Oxford: Oxford University Press.
Papafragou, Anna (1996a). Figurative language and the semantics-pragmatics distinction.
Language and Literature, 5(1), 179–193.
Papafragou, Anna (1996b). On metonymy. Lingua, 99, 169–195.
TSL[v.20020404] Prn:13/06/2003; 10:24 F: PB11301.tex / p.20 (20)
Radden, Günter & Kövecses, Zoltán (1999). Towards a theory of metonymy. In K.-U.
Panther & G. Radden (Eds.), Metonymy in Language and Thought [Human Cognitive
Processing 4] (pp. 17–59). Amsterdam and Philadelphia: Benjamins.
Raskin, Victor (1985). Semantic Mechanisms of Humor. Dordrecht: Reidel.
Ruiz de Mendoza, Francisco J., & Otal Campo, José L. (2002). Metonymy, Grammar, and
Communication. Albolote, Granada: Editorial Comares.
Sadock, Jerrold M. (1974). Toward a Linguistic Theory of Speech Acts. New York: Academic
Press.
Song, Nam Sun (1997). Metaphor and metonymy. In R. Carston & S. Uchida (Eds.),
Relevance Theory: Applications and Implications (pp. 87–104). Amsterdam and
Philadelphia: Benjamins.
Sperber, Dan & Wilson, Deirdre (1995). Relevance: Communication and Cognition (2nd ed.).
Oxford: Basil Blackwell.
Thornburg, Linda & Panther, Klaus (1997). Speech act metonymies. In W.-A. Liebert,
G. Redeker, & L. Waugh (Eds.), Discourse and Perspectives in Cognitive Linguistics
[Current Issues in Linguistic Theory 151] (pp. 205–219). Amsterdam and Philadelphia:
Benjamins.
Turner, Mark & Fauconnier, Gilles (2000). Metaphor, metonymy, and binding. In
A. Barcelona (Ed.), Metaphor and Metonymy at the Crossroads: A Cognitive Perspective
[Topics in English Linguistics 30] (pp. 133–145). Berlin and New York: Mouton de
Gruyter.
Verschueren, Jef (1999). Understanding Pragmatics. London: Arnold.
TSL[v.20020404] Prn:18/03/2003; 14:40 F: PB113P1.tex / p.1 (21)
P I
Cognitive operations
and pragmatic implication
. Introduction
are not regarded as being part of what is said or explicated, but as being derived
via implicature.
In Cognitive Linguistics, metaphor and metonymy are not considered as
tropes of figurative language, but as cognitive mechanisms used for drawing in-
ferences, and to reason about and understand the world. These cognitive opera-
tions are accomplished by means of conceptual mappings of knowledge from a
source domain into a target domain and are crucial for concept formation and
concept understanding. We shall propose in this chapter that metaphoric and
metonymic mappings are to be rightfully included in the relevance-theoretic
list of mechanisms that are used to obtain explicated meaning. In so doing,
we argue in support of Sperber and Wilson (and against Levinson 2000: 193–
198) that the distinction between explicature and implicature is a valid one, but
that metaphoric and metonymic interpretations – rather than being implicated
meanings – should be viewed as part of the explicit meaning of utterances, i.e.
as enrichments that come about by explicature.
Sperber and Wilson (1995) have criticized those pragmaticists who, follow-
ing Gricean postulates, have adopted as a working principle the view that any
aspect of utterance interpretation that falls outside the domain of disambigua-
tion and reference assignment is an implicature. Instead, Sperber and Wilson
have made the interesting claim that some of the cases that have been regu-
larly treated as implicatures are in fact cases of explicit meaning, which they
call explicatures. For them, an assumption is explicit if it is a development of
the logical form encoded by an utterance (Sperber & Wilson 1995: 182). A log-
ical form, in turn, is “a well-formed formula, a structured set of constituents,
which undergoes formal logical operations determined by its structure” (ibid.
p. 72). When a logical form is semantically complete – and therefore capable
of being true or false – it becomes a proposition. Incomplete logical forms are
stored in conceptual memory as assumption schemas that may be completed on
the basis of contextual information. Since, for Sperber and Wilson, complet-
ing an assumption schema – which has a logical form – in order to obtain a
proposition – which also has a (more developed) logical form – is an inferen-
tial activity (i.e., it exceeds mere decoding), it follows that, for them, studying
the way the logical form of an utterance is developed into its explicature is a
matter of pragmatics.1
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.3 (25)
Sperber and Wilson, together with other relevance theorists (e.g. Carston
1988; Blakemore 1992), have defended the view that there are three processes
involved in getting from an assumption schema to a full proposition: disam-
biguation, fixation of reference, and enrichment. While disambiguation and
reference assignment are familiar linguistic phenomena, the notion of enrich-
ment is entirely new. Consider the following example by Carston (1988; in
Davies 1991: 39):
(1) The park is some distance from where I live.
For B’s response to be relevant, A needs to have access to the (implicit) as-
sumption that one’s comfort while on holiday may typically be affected by in-
sects (rather than hidden microphones) and an excess of people. As a conse-
quence, we reason that the speaker did not enjoy his holiday. This information
is an implicature since it has its own distinct propositional form that functions
independently of the explicated information as the conclusion of an argument.
Both Sperber and Wilson (1995) and Blakemore (1992) interpret the lack
of literalness of metaphor and other ‘tropes’ as a matter of producing implica-
tures. For example, according to Blakemore (1992: 163), the metaphor
(3) My neighbor is a dragon.
These are the more central implicatures. Other weaker ones would have to do
with the nature of the neighbor’s unfriendliness, together with her behavior
and appearance. It is these weaker implicatures that justify the speaker’s not
using a non-metaphorical utterance like My neighbor is fierce and unfriendly.
Metaphor is thus seen as a way of optimizing relevance, which Sperber and
Wilson understand as achieving the adequate balance between processing cost
and meaning effects.
The implicature-explicature distinction, as it stands in the traditional
relevance-theoretic literature, has been the subject of considerable revision and
criticism. Among its weaknesses stands the problem of finding solid criteria to
distinguish what is implicated from what is explicated. Carston (1988) and Re-
canati (1989) are classic attempts to remedy this shortcoming. Carston (1988)
puts forward what she calls the ‘functional independence’ criterion, according
to which there is no functional independence between what is said and the en-
riched version of what is said, since the latter entails the former. This means
that the enriched interpretation cannot be an implicature. For example, as we
have seen above, in the sentence The park is some distance from where I live the
expression some distance is usually interpreted as a ‘long distance’ or ‘further
away than you think.’ This interpretation is a development of the blueprint
provided by what is said and entails it. In genuine implicated meaning, what
is said and what is implicated do not stand in the same kind of relationship.
For example, the same sentence could be used to convey a warning that the
addressee may not be able to walk such a long distance and should, therefore,
take a bus. The functional independence criterion has been criticized by Re-
canati (1989), who argues that Carston makes the mistake of using a formal
property of propositions (i.e. entailment) to distinguish explicatures from im-
plicatures. Recanati’s insight is essentially correct since what is involved in ex-
plicature derivation is not necessarily a logical development of what is said,
but an adaptation of what is said to contextual requirements. Thus, Carston’s
functional independence test does not fare well in cases of what Recanati has
called ‘saturation’ as a form of deriving explicatures. For example, the sentence
John is not good enough is to be interpreted as John is not good enough for a
certain activity or purpose (e.g. John is not good enough for that job/for Mary,
etc.). This development of the initial expression does not entail what is said (cf.
Ruiz de Mendoza 1999). Additionally, as we shall see below, Carston’s criterion
is unable to handle other forms of deriving explicatures such as the ‘loose use’
of concepts (‘loosening’) (cf. Carston 1997: 106).
Levinson (2000: 195–196) points out a number of problems in the analy-
sis provided by both Sperber and Wilson (1995) and Carston (1988). Thus, he
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.5 (27)
notes that, contrary to what Sperber and Wilson assume, the representations
of explicatures do not necessarily contain the semantic representation associ-
ated with what is said. For example, as he aptly observes, any implicature can
be added as a conjunct to what is said. In John’s three children came to the party,
it is possible to phrase the corresponding scalar inference in terms of either ex-
plicated meaning (‘the totality of John’s children, of cardinality three, came
to the party’) or as an implicated separate proposition (‘John has no more
than three children’). This same observation affects the theoretical status of
Carston’s functional independence criterion, since adding an implicature to
what is said yields a complex proposition that may entail what is said. For ex-
ample, the sentence The beach was crowded may be used to implicate ‘I couldn’t
rest.’ The sentence The beach was crowded and I couldn’t rest makes explicit such
an implicature, which thus becomes functionally dependent on what is said.
More recently, Carston (1997, 2000) has attempted to refine the explicature-
implicature division by looking in greater depth into the concept of enrichment
and by specifying further mechanisms of explicature generation. As a result,
some cases of what was previously considered a matter of implicated mean-
ing have been transferred to the domain of explicature derivation. Initially,
Carston (1988), following Sperber and Wilson (1995), accepted the proposal
of three mechanisms to derive explicatures: fixation of reference, disambigua-
tion and enrichment. Later, it was realized that, just like enrichment, cases of
what Sperber and Wilson (1985–1986) had called ‘loose use’ of language, which
included all tropes, also involved a departure from literalness, although in an
opposite direction. Thus, while the non-literal expression some time required
strengthening into ‘a long time,’ the interpretation of non-literal raw in an ut-
terance like This steak is raw involves a loosening of the lexical concept ‘raw’
from ‘not cooked’ to ‘underdone’ (and, therefore, ‘difficult to eat’). Conse-
quently, Carston proposed loosening as a mechanism to derive explicatures.
This mechanism typically applies to metaphor, which, thus far, had been dealt
with in relevance theory as a matter of implicature derivation.
In Carston (2000) both strengthening and loosening of concepts are treated
as forms of what in Relevance Theory circles has come to be known as ad hoc
concept construction, a term that goes back to earlier work by Barsalou (1983).
This involves the creation of a concept – as an adjustment to contextual re-
quirements – on the basis of a linguistic cue. In This steak is raw the loosening
process of raw is only possible in a context in which it is evident that the steak
is not literally raw (e.g. a customer is complaining about his steak being too
underdone for his taste). Interestingly enough, the strengthening of scalar con-
cepts also requires some sort of contextual adjustment. The expression some
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.6 (28)
time requires strengthening in It will take some time to repair your watch, but
not in He returned some time later (meaning ‘just a little bit later’).
The construction of an ad hoc concept is regulated by the principle of rel-
evance, according to which the hearer is entitled to assume that the intended
interpretation of an utterance creates the intended contextual effects with no
unjustifiable processing effort. Of course, this principle is sufficient to con-
strain the number of possible meaning implications of an expression that needs
this kind of adjustment. However, consistency with the principle of relevance
is insufficient to explain how the implications come about or what conceptual
mechanisms are involved in their derivation. Thus, the principle of relevance
does not account for what regulates the connection between ‘raw’ and ‘under-
done,’ or between ‘raw’ and any other target of the explicature derivation task.
For example, the sentence My steak is raw is not a complaint in a context in
which the customer likes ‘rare’ meat. The cognitive mechanisms underlying
the derivation of a whole range of different explicatures for the lexical concept
‘raw’ have to do with the conventionalized scalar nature of the various targeted
values and their connotations. Deriving explicatures on the basis of a scale is
tantamount to going up or down the scale (this is the cognitive operation) until
the hearer finds a point on the scale that will yield the relevant meaning effects
in terms of the context.
The example above is a case of what Carston (1997, 2000) categorizes
as loosening. We face similar problems when confronted with examples of
strengthening. Consider again the expression some time meaning ‘a long time’.
Interpreting this expression requires going up a scale of time measurement un-
til a point is reached where adjustment to contextual parameters is possible.
The principle of relevance regulates the extent of the strengthening task, but it
does not control how the task itself is achieved. The foregoing discussion sug-
gests that loosening and strengthening are cognitive mechanisms operating on
scalar concepts. But, there are also other cognitive mechanisms that play a role
in explicature derivation.
In Carston’s (1997, 2000) more recent work, metaphor and metonymy are
treated as other forms of constructing ad hoc concepts, which involve loosen-
ing and strengthening respectively.2 We believe this approach evinces the same
kind of weakness that we have already identified when dealing with scalar con-
cepts. For example, let us consider the metaphor Bill is a bulldozer (Carston
1997: 113), meaning that Bill is self-confident and determined. For Carston,
this expression makes a loose use of ‘bulldozer’ thereby creating an ad hoc
concept. However, by itself this is not enough to explain how we obtain the
relevant interpretation. In Cognitive Linguistics (cf. Lakoff & Johnson 1980,
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.7 (29)
1999; Lakoff 1987, 1993; Lakoff & Turner 1989), it would be postulated that
what guarantees the interpretation is the existence of an underlying conceptual
mapping from ‘bulldozer’ to ‘human being’ whereby we understand Bill’s be-
havior in terms of the figurative behavior (i.e. the way the machine functions)
that we attribute to a bulldozer. In other words, as with scalar concepts, we have
an underlying mental operation (i.e. a conceptual mapping) at work to provide
the range of meaning implications that are adequate to the context.
As the example above illustrates, a metaphor involves a conceptual map-
ping across two domains. Metonymy also involves a conceptual mapping, but
within one domain. In the sentence The sax has the flu there is a metonymic
shift from ‘sax’ to ‘saxophone player,’ where the player and his instrument stand
in a domain-subdomain relationship. As with ‘bulldozer’ above, ‘sax’ is con-
structed ad hoc for the purpose of identifying another concept with which it
has some sort of connection. However, the kind of metaphoric or metonymic
connection is not a matter of the principle of relevance as such. What this
principle does is merely draw our attention as interpreters to the necessity of
making a connection. Understanding how this connection is made and finding
out its communicative consequences depends on our ability to determine the
cognitive operation to be carried out for the sentence to be relevant in context.
What our discussion above suggests is that the task of deriving explicated
meaning involves performing any of a number of cognitive operations on the
basis of the blueprint provided by the linguistic expression and in connection
with – or as constrained by – the principle of relevance. To the mechanisms of
fixation of reference, disambiguation, saturation or completion, strengthening,
and loosening, we need to add conceptual mappings as discussed in the cogni-
tive linguistics literature. It must be noted that we have restricted the scope of
application of the notion of loosening, which, in our view, only holds for some
scalar concepts (those that do not require strengthening). So-called tropes like
metaphor and metonymy need separate treatment. This decision is in keeping
with what language itself reveals about the nature of these mechanisms:
(5) Loosely speaking, this steak is raw.
(6) *Loosely speaking, Bill is a bulldozer.
(7) *Loosely speaking, the sax has the flu.
As is evident from the examples above, the hedge loosely speaking is only com-
patible with the hyperbole in (5), but not with the metaphor and metonymy
in (6) and (7) respectively. This suggests that these latter tropes are not cases
of loosening. In what follows we shall study in more detail how conceptual
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.8 (30)
somewhere in that region, and searching the region is searching for a solution.
For a complete understanding of the full range of meaning implications of the
sentence, it is necessary to activate all the relevant correspondences.
The pragmatic notion of explicature as used by relevance theorists can
be improved from a cognitive linguistic perspective, if we regard explicature-
derivation as the result of cognitive operations. Moreover, a classification of
these operations may elucidate the nature of the explicatures that are ob-
tained through metaphoric mappings. In principle, in the case of many-
correspondence mappings, we expect a larger number of potential explicatures
than in the case of one-correspondence metaphors. Compare the following
sentences:
(10) John is a lion.
(11) You’re going nowhere that way.
From sentence (10) we obtain the explicature that John is courageous in the
same way as a lion is thought to be courageous (i.e. in a fierce instinctual way).
Remember that in the one-correspondence metaphor people are animals,
behavior usually maps onto behavior. Sentence (11), on the other hand, may
generate a greater number of explicatures based on the hearer exploring dif-
ferent aspects of the structure and logic of the many-correspondence journey
metaphor that is based on the path schema. One of the explicatures will be
more central than the others. However, in our view, this does not take away
from the explicit nature of all the inferences that are developed on the basis of
the conceptual material associated with the expression. Thus, in a situation in
which sentence (11) is uttered by an angry father to a rebellious teenage son, it
is possible to derive at least the following potential explicatures, of which the
first one is more central:
(12) a. The addressee is not going to achieve his expected goals (if he persists
in his behavior).
b. The addressee is not making any progress in life.
c. The addressee may make progress if he changes his way of doing
things.
d. The addressee is acting in an erroneous way.
e. The addressee may not have clear goals.
f. The addressee has erroneous goals.
As Lakoff and Johnson (1999) have noted, the central correspondence in jour-
ney metaphors is goals are destinations. This correspondence allows us
to derive the central explicature, i.e. (12a), which focuses on reaching the
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.11 (33)
In (13) it was not Napoleon but the army under his command that was de-
feated. In (14) it is not the fictional character Superman, but the actor Christo-
pher Reeve, who broke his back. The metonymy in (13) is a way of avoiding
the use of a longer and perhaps rather vague definite description that would
be more difficult to process; it is also a means of emphasizing Napoleon’s more
prominent role in the defeat. All this information is part of the explicature de-
rived with the aid of the metonymic mapping. This is so because it is part of our
knowledge that Napoleon organized an army with which he invaded Belgium,
but was later defeated at Waterloo by the English.
The metonymy in (14) would normally be used to avoid a long paraphrase
such as Christopher Reeve, the actor who played Superman, in order to achieve
successful reference. Despite the similarities of metonymic reference in (13)
and (14), the two metonymies are different in two respects. First, (13) instanti-
ates a metonymy in which the target domain (the army) is a subdomain of the
source domain (Napoleon), while (14) instantiates a metonymy in which the
source (Superman) is a subdomain of the target (Christopher Reeve, the actor).
Examples like (13) are cases of what we may call target-in-source metonymy;
(14), on the other hand, illustrates a case of source-in-target metonymy (see
discussion of examples (16) and (17) below for details on the relevance of this
distinction). Note additionally that, since the source in (13) and the target in
(14) provide frameworks of reference for their corresponding subdomains, it
is appropriate to refer to them as matrix domains.4 Second, the difference be-
tween the two types of domain-subdomain relationship we have identified has
communicative consequences that have a bearing on the kind of explicatures
to be derived. Thus, metonymies like the one in (13) allow the speaker to focus
on the source domain as the most relevant one while avoiding an uneconomi-
cal description like the army commanded by Napoleon. This mechanism is often
used when the speaker is unable to find the expression that actually designates
the intended referent, as in:
(15) The White House is trying to avoid another scandal.
whereby using the expression the White House the speaker avoids the problem
of having to name the accurate White House official in question.
On the other hand, metonymies like the one in (14) allow a speaker to
bring to the fore a cognitively salient subdomain of the matrix domain, with
its accompanying inferential effects. Thus, part of our conventional knowledge
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.13 (35)
about Superman (i.e. his supernatural strength) makes (14) a rather shocking
sentence from the point of view of the information it conveys. It is not only sad
but also ironic that the “man of steel” has seriously damaged his back.
It must be noted that metonymies are cases of one-correspondence map-
ping. This is a necessary consequence of the fact that in metonymies there is a
domain-subdomain relationship. This relationship only allows for two kinds of
conceptual operation: one involves highlighting a subdomain of the source, as
in (13) and (15) above; the other requires the expansion of the source into
a wider conceptual structure, as the case of (14). In having just one corre-
spondence, metonymies allow us to derive only one explicature, as will be
illustrated below.
In this connection, Ruiz de Mendoza (1997, 2000) argues for the existence
of a metaphor-metonymy continuum based on the distinction between one-
correspondence and many-correspondence metaphors. Here we want to sug-
gest that understanding the nature of this continuum is crucial to determining
the communicative effects of the explicatures derived by means of metaphoric
and metonymic mappings. Ruiz de Mendoza notes that while metaphors are
typically non-referential, one-correspondence metaphors may occasionally be
referential in nature, as in My tender rose abandoned me (cf. She is a tender
rose). Many-correspondence metaphors, on the other hand, can only be pred-
icative, and metonymies – which are by definition one-correspondence map-
pings – are usually referential. It is true that it is possible to find instances
of non-referential metonymies, as in She is a real brain or She’s just a pretty
face, or as in expressions containing verb-based metonymies (cf. Goossens
1990) such as giggle in ‘Oh dear,’ she giggled, where this verb stands for ‘to
say something while giggling.’ However, verb-based metonymies are necessar-
ily predicative and could not possibly function referentially, while predicative
uses of noun-based metonymies, which are very rare, need some sort of pa-
rameterization (adjectival or otherwise) of the source domain in order to sin-
gle out a feature that will subsequently map onto the target, just like in one-
correspondence metaphors (cf. Section 3). Because of this, it is possible to re-
gard them as borderline cases of metonymy: on the one hand, they share with
all metonymies the fact that there is an evident domain-subdomain relation-
ship between source and target; on the other hand, the operation of singling out
a feature of the source, rather than using its whole conceptual structure, makes
them resemble metaphors. If these observations are correct, the metaphor-
metonymy continuum would have cases of many-correspondence mappings
(which are predicative) at one end and clear cases of referential metonymy at
the other end. Referential uses of metaphor, verb-based metonymies (which
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.14 (36)
Ruiz de Mendoza (2000) and Ruiz de Mendoza and Pérez (2001) have postu-
lated the existence of what they call the Domain Availability Principle (or DAP).
According to this principle, only the matrix domain of a metonymic mapping is
available for anaphoric reference. In (16), the matrix domain is the target of the
metonymic shift (i.e. the customer who has ordered the ham sandwich), while
in (17), the matrix domain is the source of the metonymy mapping Nixon onto
the air force that carries out the orders he has sanctioned. The selection of the
matrix domain for anaphoric reference, rather than a subdomain, is perhaps
due to the fact that the matrix domain is conceptually more salient, since it
provides a larger amount of easily retrievable conceptual structure. This last
observation particularly holds for target-in-source metonymies, since it is of-
ten the case that their targets are not clean-cut specifications. Thus, in (17) it is
not known what specific section of the air force is involved in the bombing, but
we do know that Nixon is ultimately responsible for that action. In both kinds
of metonymic mapping, the explicature includes the information provided by
the linguistic expression, although each in a different way. In source-in-target
metonymies, what is said is conceptually expanded to include as much concep-
tual material as is necessary for the target concept to be compatible with the rest
of the predication. How much conceptual material is called up is regulated by
consistency with the principle of relevance, that is, we do not want to activate
more material than needed, but enough for the interpretive demands of the
moment. In target-in-source metonymies, whatever subdomain is highlighted
is determined by its degree of compatibility with the information provided by
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.15 (37)
In (18), my tender rose roughly means ‘the girl or woman that arouses in me
the same feelings as a tender rose.’ This explicature provides enough concep-
tual material for successful anaphoric reference. So, it is preferred – for this
purpose – to what is literally said by the expression.
The fact that anaphoric reference can be made to one of the conceptual
domains involved in metaphor and metonymy further supports our view that
these cognitive mechanisms generate only explicatures, since implicated mean-
ings seem to be poor candidates for anaphoric reference. In this connection,
consider again the short exchange in (2), expanded here as (19):
(19) A: Did you enjoy your holiday?
B: The beaches were crowded and the hotel was full of bugs.
A: I am sorry about that.
As remarked in Section 2, the implicated conclusion is that B did not enjoy his
holiday. However, this proposition is not the best candidate as antecedent for
anaphoric that, but rather the explicated information given by B’s response.
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.16 (38)
So far, we have looked into the communicative import of what we can call single
metonymies. However, sometimes metonymic mappings evince a rather more
complex structure. Consider the following examples:
(20) I love Picasso.
(21) I have a Picasso in the living room.
(22) Can you imagine? He is using Picasso as a bookend.
At first sight, Picasso in (20) stands for ‘Picasso’s pictorial work,’ but as in the
Napoleon example in (13), the conceptual structure of the matrix domain ‘Pi-
casso’ is somehow present in the explicature generated by the metonymic map-
ping. Thus, (20) communicates not only that the speaker likes Picasso’s work
a lot, but also that there is some reason for this, which is to be found in our
knowledge about Picasso (e.g. the speaker may be impressed by Picasso’s style,
by his mastery of color, and so on). Whatever additional knowledge is imported
from the matrix domain to construct the relevant explicature must be in har-
mony with the context and regulated by the principle of relevance. The situ-
ation is slightly different in (21). Here Picasso stands for ‘a specific sample of
his work.’ So, I have a Picasso can be paraphrased as ‘I have a specific painting
by Picasso.’ In order to arrive at this meaning, we need a double metonymic
operation from Picasso to ‘Picasso’s work’ to ‘a (unique) sample of his work.’
This situation is represented in Figure 1 below.
‘Picasso’
‘Picasso’s
work’
‘a specific
painting
by Picasso’
‘Shakespeare’
‘Shakespeare’s
work’
‘a book containing
(a sample of)
Shakespeare’s
work’
ous operation. This second mapping, from ‘Shakespeare’s work’ to the format
in which it is presented (typically a book or a number of books), allows us
to identify the referent in an economical way, while at the same time preserv-
ing the conceptual relevance of its source domain in such a way that not only
the format in which Shakespeare’s work is presented, but also its content is
important. Observe in this connection that the metonymic expression in (23)
highlights the content of Shakespeare’s writing rather than its material form, in
contrast to a non-metonymic paraphrase such as The book by Shakespeare is on
the top shelf, where it is the material form (e.g. a book rather than a CD-ROM,
microfiches, etc.) that is more central.
SOURCE TARGET
People applaud
A person would
rather bite his SOURCE
tongue
SOURCE TARGET
An animal
rears up SOURCE
METONYMY
METAPHOR A person
An animal rears
energetically
up
stands up
TARGET
METAPHOR in order to
(out of fear) as if argue in
to attack public
Figure 5. Metonymic development of the metaphoric source in get up on one’s hind legs
The examples we have discussed so far illustrate but one kind of in-
teractional pattern in our analysis, namely, metonymic development of a
metaphoric source. However, we propose others. Consider first the metaphoric
expression in (27):
(27) My lips are sealed.
This utterance may be used by a speaker to make a binding promise that he will
not reveal a secret. The act of sealing his lips is a figurative indication that he
will in no way open his mouth to disclose any information. But this indication
is part of a more general situation in which the speaker voluntarily decides to
keep certain information confidential.
Consider now the expression in (28):
(28) to win someone’s heart
The source domain of this metaphor contains a winner and a prize. The target
domain contains a lover who has succeeded in figuratively obtaining someone’s
heart. The heart, as a container of feelings, is chosen to stand for the feeling of
love. Since ‘heart’ and ‘love’ stand in a domain-subdomain relationship, we
have a case of metonymic highlighting of (a relevant part of) the metaphoric
target. Winning requires effort and tactics, an implication that is carried over to
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.22 (44)
TARGET
SOURCE
SOURCE
METONYMY
TARGET
the target domain of the metaphor, thus suggesting that the action of obtaining
someone’s love has been a difficult one. Figure 7 illustrates this process.
One final pattern of metaphor-metonymy interaction is illustrated in (29):
(29) She is the life and soul of the party.
SOURCE TARGET
METAPHOR
winner lover
METONYMY
TARGET
love
person party
Metonymy
Figure 8. Metonymic highlighting of a metaphoric source in She is the life and soul of
the party.
Notes
discourse connectives) are different from those that impose constraints on explicatures (e.g.
pronouns) (see Wilson & Sperber 1993: 19–23).
. In canonical accounts of Relevance Theory (e.g. Sperber & Wilson 1995; Blakemore
1992), metaphor and metonymy are treated as cases of interpretive (as opposed to descrip-
tive) use of language bringing about the production of strong and weak implicatures. In this
respect, Carston’s proposal, where – as we shall see below – loosening and strengthening are
forms of deriving explicatures, involves a relatively important readjustment in the theory.
The idea that metonymic shifts generate explicatures is also found in Papafragou (1995).
. To Lakoff and Johnson’s (1980) classification, it is possible to add situational metaphors
like He left the room with his tail between his legs. This metaphor invokes a situation in which
a person who has been defeated and humiliated decides to forego the pursuit of a certain
goal. This is based on the stereotypical representation of a dog leaving with its tail between
its legs after being defeated or otherwise punished. Situational metaphors are a type of struc-
tural metaphor. It must additionally be noted that metaphors based on image-schemas are
also a type of structural metaphor, since they allow us to understand abstract concepts in
terms of others grounded in bodily experience. For example, in the metaphor love is a
journey (Lakoff 1993), different aspects of a love relationship are understood in terms of
different elements of a journey: lovers are travelers, the vehicle is a love relationship, diffi-
culties in the relationship are impediments to travel, and goals are destinations (e.g., This
relationship is going nowhere, We are spinning our wheels, etc.).
. Ruiz de Mendoza and Díez (forthcoming) discuss a number of criteria that may be used
to determine which of the two domains of a metonymy qualifies as a matrix domain. Gen-
erally, the matrix domain of a metonymy is the most encompassing and/or most clearly de-
lineated (and therefore more readily accessible) of the two domains involved in a mapping.
When two domains are equally accessible, the matrix domain is a matter of the application
of a number of principles of cognitive saliency, like those identified by Langacker (1993)
and Kövecses and Radden (1998): human over non-human; container over content; con-
trolling entity over controlled entity; whole over part, among others. In the Napoleon-army
relationship it is the controlling over controlled principle that holds.
. According to Goossens (1990: 171) there is an additional possibility of interaction type,
namely demetonymization within a metaphorical context, as illustrated in the expression
pay lip service to meaning ‘support in words, but not in fact.’ Goossens argues that at first
sight this expression seems to be an example of metonymy within metaphor, where lip(s)
stands for ‘speaking,’ but that the figurative expression will work only if, in a second stage,
lip service is interpreted as ‘service as if with the lips only,’ where ‘lips’ is somehow dissociated
from ‘speaking,’ However, the status of such a process of demetonymization is questionable.
Contrasted with ‘actual (or genuine) service,’ lip service suggests that the person who offers
the service has no intention of keeping his word. Thus, lip service means ‘a promise or an
offer of service,’ and this linguistic action is carried out by means of speech.
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.26 (48)
References
Barcelona, Antonio (Ed.). (2000). Metaphor and Metonymy at the Crossroads: A Cognitive
Perspective [Topics in English Linguistics 30]. Berlin and New York: Mouton de Gruyter.
Barsalou, Lawrence (1983). Ad hoc categories. Memory & Cognition, 11, 211–227.
Blakemore, Diane (1992). Understanding Utterances. Oxford: Basil Blackwell.
Carston, Robyn (1988). Implicature, explicature, and truth-theoretic semantics. In R. M.
Kempson (Ed.), Mental Representations: The Interface Between Language and Reality
(pp. 155–181). Cambridge: Cambridge University Press. [Reprinted 1991 in S. Davis
(Ed.), Pragmatics: A Reader (pp. 33–51). New York, Oxford: Oxford University Press.]
Carston, Robyn (1997). Enrichment and loosening: complementary processes in deriving
the proposition expressed? Linguistische Berichte, 8, 103–127.
Carston, Robyn (2000). Explicature and semantics. UCL Working Papers in Linguistics, 12,
1–44. [To appear in S. Davis & B. Gillon (Eds.), Semantics: A Reader. Oxford: Oxford
University Press.]
Goossens, Louis (1990). Metaphtonymy: the interaction of metaphor and metonymy in
expressions for linguistic action. Cognitive Linguistics, 1, 323–340.
Johnson, Mark (1987) The Body in the Mind: The Bodily Basis of Meaning, Reason and
Imagination. Chicago and London: The University of Chicago Press.
Kövecses, Zoltán & Radden, Günter (1998). Metonymy: developing a cognitive linguistic
view. Cognitive Linguistics, 9–1, 37–77.
Lakoff, George (1987). Women, Fire, and Dangerous Things: What Categories Reveal About
the Mind. Chicago and London: The University of Chicago Press.
Lakoff, George (1993). The contemporary theory of metaphor. In A. Ortony (Ed.), Metaphor
and Thought (pp. 202–251). Cambridge: Cambridge University Press.
Lakoff, George & Johnson, Mark (1980). Metaphors We Live By. Chicago and London: The
University of Chicago Press.
Lakoff, George & Johnson, Mark (1999). Philosophy in the Flesh: The Embodied Mind and its
Challenge to Western Thought. New York: Basic Books.
Lakoff, George & Turner, Mark (1989). More Than Cool Reason: A Field Guide to Poetic
Metaphor. Chicago and London: The University of Chicago Press.
Langacker, Ronald W. (1993). Reference-point constructions. Cognitive Linguistics, 4–1, 1–
38.
Levinson, Stephen C. (2000). Presumptive Meanings: The Theory of Generalized Conver-
sational Implicature. Cambridge, MA, and London: MIT Press.
Panther, Klaus-Uwe & Radden, Günter (Eds.). (1999). Metonymy in Language and Thought
[Human Cognitive Processing 4]. Amsterdam and Philadelphia: Benjamins.
Papafragou, Anna (1995). Metonymy and relevance. UCL Working Papers in Linguistics 7.
Department of Phonetics and Linguistics, University College London.
Peña, Sandra (1997). The role of the event structure metaphor and of image-schematic
structure in metaphors for happiness and sadness. Miscelanea. A Journal of English and
American Studies, 18, 253–266. Zaragoza: University of Zaragoza Press.
Recanati, François (1989). The pragmatics of what is said. Mind and Language, 4, 294–328.
[Reprinted 1991 in S. Davis (Ed.), Pragmatics: A Reader (pp. 97–120). Oxford: Oxford
University Press.]
TSL[v.20020404] Prn:13/06/2003; 10:25 F: PB11302.tex / p.27 (49)
. Introduction
prompts the reader to construct two mental spaces, one to represent Seana’s
opinion of the statue, and one to represent Todd’s:
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.3 (53)
Seana Todd
Thinks Thinks
====== ======
s s’
hideous(s) wonderful(s’)
One virtue of mental space theory is that it explains how the addressee might
encode information at the referential level by dividing it into concepts rel-
evant to different aspects of the scenario. However, by partitioning the in-
formation, this method also creates a need to keep track of the relationships
that exist between counterpart elements and relations represented in different
mental spaces.
Consequently, the notion of mappings between mental spaces is a central
component of both mental space theory and the theory of conceptual blend-
ing. A mapping, or mental space connection, is the understanding that an object
or element in one mental space corresponds to an object or element in another.
For example in (1), the sentence about Seana’s and Todd’s respective opinions
about the statue, there is an identity mapping between the element s that rep-
resents the statue in Seana’s opinion space, and element s’ that represents the
statue in Todd’s opinion space. The mental spaces framework thus allows one
to represent the fact that the very same statue is referred to in the statue is
hideous and it’s just wonderful, in spite of its disparate properties in the two
opinion spaces.
Besides identity, such mappings can be based on a number of relation-
ships, such as similarity, analogy, and other pragmatic functions. Once linked,
the access principle allows speakers to refer to an element in one space by nam-
ing, describing, or referring to its counterpart in another space. Interestingly,
part of Fauconnier’s (1994) justification for the access principle allowing refer-
ence across different spaces was the existence of similar connectors operating
within a single mental space. For example, Fauconnier suggests that a prag-
matic function linking hospital patients to their medical conditions licenses
the metonymic reference to the patient in (2):
(2) The gastric ulcer in room 12 would like some coffee.
Just as pragmatic functions connecting (say) patients and their illnesses can
allow speakers to access and refer to an associated element in the same mental
space (see Nunberg 1978 for extensive review of pragmatic functions), cross-
space mappings based on identity and analogy can allow speakers to access and
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.4 (54)
In conceptual blending theory, the way in which the meaning of Coke in (3)
appeals simultaneously to conceptual structure from multiple domains is cap-
tured in a conceptual integration network (CIN). A CIN is an array of mental
spaces in which the processes of conceptual blending unfold (Fauconnier &
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.5 (55)
Turner 1998). These networks consist of two or more input spaces structured
by information from discrete cognitive domains, an optional generic space that
contains structure common to all spaces in the network, and a blended space
that contains selected aspects of structure from each input space, and fre-
quently, emergent structure of its own. Blending involves the establishment of
partial mappings between cognitive models in different spaces in the network,
and the projection of conceptual structure from space to space.
The CIN representing (3) involves two input spaces, a soft drink space and
a corporation space. In the soft drink space, an element d is set up to repre-
sent Coke, and is structured by a frame for soft drinks (viz. the element d is
construed as having the properties of the sugary, carbonated beverage drunk
by millions of people every day). In the corporation space, c’ represents the
Coca Cola corporation that manufactures Coke. Though the beverage and the
company that makes the beverage have very different properties, the elements
d’ and c’ are linked by a conventional metonymy that allows corporations to be
identified by their products. The corporation space also includes an element p’
that represents the profit generated by Coca Cola, Inc. during the first quarter
of 2001, and f’ to represent the predicted profit for the same quarter (viz. the
economic forecast). Conceptual structure in the corporation space involves a
frame for corporate profit, and for evaluating corporate profits. Moreover, a
conventional metonymy between corporations and their profits links elements
c’ and p’.
Soft Drink Corporation
Input Input
========= ==========
d d’
c’
p’
f’
The blended space in this network contains element c* linked by identity to
d in the soft drink space, and by metonymy to p’ (in the corporation space).
While conceptual structure in the input spaces comes from the domains of soft
drinks and corporations, the blended space includes partial structure from each
of the inputs as well as emergent structure of its own. Consequently, element
c* has some of the properties of Coke (in that it is a liquid), and some of the
properties of Coca Cola, Inc.’s first quarter 2001 profit (in that it was greater
than the forecasted profit).
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.6 (56)
For example, (4) involves a blend of a static construal of the spatial extent
of the blackboard (spatial input) with an abstract, image schematic under-
standing of a trajector that moves relative to a reference point, or landmark.
There is a mapping between the blackboard and the trajector and these two
elements are fused in the blended space. Similarly, there is a mapping between
the wall and the landmark, and these two elements are fused in the blended
space. In the motion input, the trajector’s motion ends at the landmark. Sim-
ilarly, in the blended space, the motion of the blackboard/trajector ends at the
wall/landmark. The path of motion can then be mapped onto the spatial input
to be construed as the spatial extent of the blackboard.
Spatial Blended Motion
Input Space Input
====== ====== ======
blackboard blackboard/trajector trajector
wall wall/landmark landmark
(static) (motion) (motion)
The fictive motion in (3) can be analyzed similarly by including a third input
space to the CIN (see Figure 1). The third input is structured by a schematic
characterization of the fictive motion schema in which a trajector moves along
an abstract path with a reference point (as in Fauconnier 1997).
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.7 (57)
Corporation Input
Soft Drink Input Motion Input
Blend
Figure 1. Conceptual integration network for (3) Coke flows past forecasts: Soft drink
company posts gains
ing in the headlines Tennessee Tramples Kentucky and Overseas Ballots Boost
Bush involve conventional metonymies between states and universities, univer-
sities and their football teams, politicians and their votes, in combination with
metaphoric mappings between combat and sports, and between greater quan-
tities and greater heights. One advantage of the blending framework is that it
allows the treatment of examples like (3) that require many sorts of conceptual
mappings to be set up in parallel in the course of meaning construction.
. Optimality principles
blends are more effective. Fauconnier and Turner (1998) argue for six such
optimality principles: (i) the integration principle, that representations in the
blended space can be manipulated as a single unit; (ii) the topology principle,
that relations in the blend should match the relations of their counterparts in
other spaces; (iii) the web principle, that the representation in the blended space
should maintain mappings to the input spaces; (iv) the unpacking principle,
that given a blended model, the interpreter should be able to infer the structure
in other spaces in the network; (v) the good reason principle, that creates pres-
sure to attribute significance to elements in the blend; and (vi) metonymic tight-
ening, that when metonymically related elements are projected into the blended
space, there is pressure to compress the “distance” between them. By reducing
the space of possible blending analyses, these constraints make blending more
principled.
Despite their poetic names, most of these principles invoke standard pres-
sures that obtain in all mapping problems (see Hofstadter 1995 for review).
The topology principle, for example, exerts normative pressure to construct
and maintain mappings in such a way as to preserve relational structure. In
research on analogical reasoning, this pressure is referred to as the structure
mapping principle (see Gentner & Markman 1997 for review). In research on
metaphorical mapping, this pressure is referred to as the invariance hypothesis,
the observation that the underlying mappings in metaphoric expressions are
almost always based on shared image schematic structure (see Brugman 1990;
Lakoff 1990; and Turner 1990). The web principle, that the representation in
the blended space should maintain its mappings to the input spaces, amounts
to the extension of the access principle to conceptual content in blended mental
spaces. Satisfaction of the web principle is what allows one to access elements in
the blend with names and descriptions from the input spaces, as well as what
allows the projection of structure from the blended space to other spaces in
the network. Finally, the unpacking principle, the dictate that given a blended
model, the listener should be able to construct structure in the other spaces
in the network, can be thought of as pressure to use conventional mapping
schemas that facilitate comprehension. Thus construed, the unpacking princi-
ple applies pressure to use conceptual metaphors, such as knowing is seeing,
and conventional metonymic mappings, such as part for whole, producer
for product, or container for contents.
The integration principle, the good reason principle, and the metonymic
tightening principle all specifically refer to the blended space, and conse-
quently, are unique to Fauconnier and Turner’s theoretical framework. How-
ever, we note that the integration principle, pressure to conceptualize the event
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.10 (60)
. Metonymic shifts
In their book Mental Leaps, Holyoak and Thagard (1995) claim that the differ-
ence between metaphor and analogy is that metaphors, especially literary ones,
are subject to “loose” and “shifting” mappings, which are the side effects of
metonymy. Treating it as a somewhat suspect technique, Holyoak and Thagard
argue that metonymy’s intrusion into metaphoric language places metaphor
outside the explanatory bounds of a theory of analogy. An unbridled force at
large in the literary universe, metonymy leads at best to analogical inconsis-
tency, at worst to incoherence. In support of their position, Holyoak and Tha-
gard point to the following excerpt from the writings of Ernest Hemingway:
His talent was as natural as the pattern that was made by the dust on a butter-
fly’s wings. At one time he understood it no more than the butterfly did and
he did not know when it was brushed or marred. Later he became conscious
of his damaged wings and of their construction and he learned to think and
could not fly anymore because the love of flight was gone and he could only
remember when it had been effortless.
(Quoted in Holyoak & Thagard 1995: 224)
Analyzing the passage, Holyoak and Thagard point to the fact that the writer’s
talent is initially mapped to the pattern of dust on the butterfly’s wings, and
later to the wings themselves. Further, they point out that there is no causal
relationship between patterns on a butterfly’s wings and its ability to fly, and
no reason why consciousness of wings should affect the butterfly’s ability to
fly. In blending theory, analogical mismatches like this are frequently used to
motivate the need for a blended space analysis. Indeed, in their discussion of
this example, Holyoak and Thagard resort to the use of slashes to represent
the conceptual fusion of ideas: “A butterfly’s pattern is not causally related to
its flight, so if talent is mapped to the pattern, then there is no reason why
consciousness of the talent/pattern should interfere with the ability to exercise
it” (Holyoak & Thagard 1995: 224).
In fact, one might also note that real butterflies cannot properly be said
to be conscious of anything, let alone be the sort of intentional creature whose
consciousness of a particular ability impairs the exercise of that ability. There is
indeed an analogical mismatch between the domain of the man’s consciousness
of his own talent and the realistic domain of butterflies. But where Holyoak and
Thagard suggest the passage involves metaphor “extended by the associative
aura created by metonymy,” we suggest it prompts a blended conceptualization
of the writer and the butterfly that exploits both metaphoric and metonymic
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.13 (63)
mappings. The first sentence, an explicit comparison between the man’s talent
and the dust on a butterfly’s wings serves to set up the two input spaces for the
ensuing blend: a “human” space with the man and his talent, and a “butterfly”
space with the butterfly and the dust. The blend between the two spaces is
prompted by the second sentence, “At one time he understood it no more than
the butterfly did and he did not know when it was brushed or marred.”
Although the sentence begins by referring to elements in the human
space, the focus gradually shifts to a blended space that concerns a hybrid
man/butterfly, exploiting partial structure from each of the input spaces. For
example, the initial use of the pronoun “he” (“he understood”) refers to the
man, while the second refers to the hybrid man/butterfly (“he did not know
when it was brushed or marred.”) The transition to reference to the blended
space is mediated by an ambiguity in the elided phrase “the butterfly did” in
“he understood it no more than the butterfly did.” This phrase could be inter-
preted as a comparison between the man’s understanding of his artistic talent
and the butterfly’s understanding of the dust pattern on its wings; or between
the man’s understanding of his talent and the butterfly’s understanding of the
man’s talent; or even between the man’s understanding of his talent and the
butterfly’s understanding of its own talent. Similarly, the first use of the pro-
noun “it” (in “he understood it”) refers to the man’s talent, while the second
(in “he did not know when it was brushed or marred”) can be understood as
referring alternately to the pattern of dust on the butterfly’s wing, the man’s
talent, or an element in the blended space with the attributes of both the dust
pattern and the man’s talent.
The multiple interpretations for this sentence can be captured in the con-
ceptual integration network in Figure 2. In the human space, the man does
not understand his artistic talent; in the butterfly space, the butterfly does not
understand the pattern of dust on its wings. In the blended space, the “he” is
a butterfly with the intentional powers of a human, and the dust pattern is a
feature of its wings that it could potentially understand, but doesn’t. The initial
blend conforms well to the first three optimality principles: integration, topol-
ogy, and web. A cognitive model of a butterfly that is not conscious of the dust
pattern on his wings is an integrated representation that is easy to manipulate.
It conforms to the topology principle because the relational structure in the
blended space corresponds to relational structure in the inputs. Moreover, it
conforms to the web principle because the mappings between elements in the
blended space and their counterparts in the input spaces are consistent.
However, the third sentence (“Later he became conscious of his damaged
wings and of their construction and he learned to think and he could not fly
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.14 (64)
anymore because the love of flight was gone and he could only remember when
it had been effortless”) employs slightly different mappings between the el-
ements in the blend and the inputs, thereby violating the web principle. As
Holyoak and Thagard point out, the man’s talent is initially mapped to the
dust pattern on the butterfly’s wings, but shifts to the wings themselves. This
violation of the web principle is offset by metonymic tightening, the pressure
to compress metonymically related elements in a blended space. In this case,
the metonymic relationship of adjacency between the butterfly’s wings and the
dust pattern licenses a mapping between the man’s talent – formerly mapped
to the dust on the wings – and the wings themselves.
The need for a blended analysis of the third sentence is readily apparent as
it makes little sense to talk about a man “conscious of his damaged wings,” nor
of a butterfly that, having lost the love of flight, “could only remember when it
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.15 (65)
had been effortless.” Though the mapping of the man’s talent has shifted from
the dust pattern on the wings to the butterfly’s wings themselves, the mapping
between the man and the butterfly remains intact, and the blended space again
features a hybrid creature with the body of a butterfly and the cognitive and
emotional capacity of the man represented in the human space. With the new
mapping scheme established, the blend in the third sentence satisfies both the
web and the topology constraints as the butterfly’s wings continue to map to
the man’s talent in a systematic way. For example, damaged wings correspond
to fading talent, inability to fly (an action that requires intact wings) corre-
sponds to the man’s inability to employ his talent (viz. writer’s block), and the
love of flight (enjoyment in the ability to use the wings for their intended pur-
pose) corresponds to the man’s love of writing (enjoyment in the ability to
exercise his talent).
Though Hemingway’s passage does indeed employ a series of analogical
mappings between conceptual structure taken variously from the domains of
butterflies and humans, it does not set up an analogy between the two domains.
That is, Hemingway does not exploit the reader’s knowledge of butterflies to
explicate notions pertaining to artistic talent. Rather, he exploits the reader’s
ability to integrate conceptual structure from disparate domains that enable
her to understand and empathize with the aging artist. While the reader may
not understand the artist’s joy in practicing his art, she can imagine the thrill
of flight. Moreover, having imagined the thrill of flight, the reader is in a better
position to empathize with the loss of this ability, and consequently its coun-
terpart in the human space, the man’s ability to write. If anything, the analogy
is from the blended conceptualization of the human butterfly to the experience
of the aging artist.
Metonymy is used here to shift the mapping schema in a way that vio-
lates the topology constraint, but optimizes integration. The metonymically
licensed slip is rhetorically motivated because the first blend serves the original
motivation of explicating the utter mindlessness of the artist’s talent, and the
second best serves the rhetorical motivation of explicating the emotive signif-
icance of a whole series of events as the artist’s ability changes over the course
of time. This includes the euphoric nature of the artist exercising his talent,
the loss of this ability and the associated regret, as well as the causal sequence
of events that produced the loss of talent. In subsequent examples we explore
other ways that metonymy licenses different sorts of blends, and the way in
which metonymic mappings affect the interplay of the optimality principles. In
particular, metonymic language frequently involves conflict of the integration
principle with the web and topology principles.
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.16 (66)
As Coulson (2000) has argued, the idiom digging your own grave entails much
more than a straightforward mapping from the source domain of grave digging
and the target domain of trouble, primarily because default interpretations of
this idiom lead to the inference that the deeper one digs the closer one gets
to dying, thus positing a direct causal relationship between grave digging and
death where none typically exists. The idiom is also fantastic in that the grave
digger and the corpse in a typical grave digging scenario both map onto the
same element in the blend. This idiom can be applied to any representation
of an individual’s actions interpreted as having untoward consequences that
the speaker thinks the individual does not foresee. Applicable target situations
can include anything from romantic disaster to academic failure to financial
ruin, as in You’re digging your own financial grave by investing all your money in
start-up Internet stocks.
Interestingly, part of the reason this example has been discussed so fre-
quently with respect to blending theory (e.g. Fauconnier & Turner 1998) is that
it cannot be accounted for by approaches to figurative language that involve
a strong commitment to the existence of shared conceptual structure in the
source and target domains. Indeed, analysis of the “digging” example in terms
of conceptual blending is motivated by the violation of the topology principle
in the disanalogous mappings that are set up between the grave digger to the
wrongheaded agent in the trouble space, and between the act of digging and
the wrongheaded act (e.g. investing in Internet stocks).
In accordance with our observations here, though, this stock example is
yet another illustration of the tradeoff between the topology principle and the
integration and unpacking principles. Although the digging your own grave ex-
ample violates topology, it does fulfill the integration constraint, allowing the
hearer to conceptualize the scenario in an integrated scene. Moreover, it ful-
fills the unpacking principle by utilizing conventional metaphoric mappings
between death and failure (Lakoff & Turner 1989), holes and situations (Lakoff
1993), and a conventional metonymic mapping between graves and death
(Turner 1987). The digger causes the grave’s existence, which maps metonymi-
cally onto death, which in turn maps metaphorically onto the wrongheaded
agent’s failure.
Metaphoric interpretation of the representation in the blended space thus
rests crucially on the metonymic identification of the grave with death. Ruiz
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.17 (67)
and Pérez (this volume) provide a number of examples that show the import
of metonymic inference for highlighting and expanding an under-specified
source domain in a metaphor. The proposal here is that this expansion of-
ten occurs in the blended space in the network, rather than in the source in-
put, thus accounting for the inferential influence of the target domain in these
processes.
A close relative to digging your own grave is blowing your own horn, both of
which exploit the X-your-own-Y pattern found in non-metaphoric examples,
such as get your own drink, and play your own instrument. In contrast to these
conventional examples, and indeed to the digging your own grave idiom, the
own-Y in blowing your own horn is not used in contradistinction to someone
else’s horn, but rather playing a horn for oneself rather than for another. The
horn in this case refers metonymically to the object of the trumpeting. The
puzzling thing about blowing your own horn, of course, is how and why blowing
your own horn maps onto praising yourself.
The verb blow and the noun horn provide the verbal cues for opening a
mental space representing the act of playing a musical instrument. In this mu-
sician space, the focal element ‘horn’ provides the reference point for accessing
and filling other slots in the frame such as ‘musician’. The event represented
in this space is the effect associated with producing a certain kind of sound.
The musician blows into the horn, which produces a distinctive (and loud)
sound, which captures the attention of others (possibly an audience). In fact,
in western ceremonies, horn playing is often a scripted part of a procession, ad-
monishing the crowd to pay attention. Blowing a horn, in effect, announces the
arrival of a very important person such that the attention of the crowd becomes
fixated on that person. Horn-playing can evoke this scenario via metonymic
inference (Ruiz & Pérez, this volume).
Further, just as there is a close part-whole relation between a musician and
her instrument, there is an even closer part-whole relation between a speaker
and her voice. This common part-whole topology establishes a close relation-
ship between the horn blowing input and the praising input we call the En-
comium space, reminiscent of the formal genre of speech in which the writer
enumerates the achievements and deeds of a living person. Since native speak-
ers of English know that this idiom is about acts of praising, the mental space
activates the focal element ‘speaker’ and ‘voice’. Once activated, the speaker
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.18 (68)
role and her most relevant feature map onto the focal elements in the musician
space, namely musician and horn.
The two spaces are analogically linked by the generic causal and temporal
relation obtaining between distinctive sound produced by a human being and
the subsequent effect it has on other minds. In the musician space, the musi-
cian blows his horn to draw the audience’s attention to some notable occur-
rence. In the Encomium space, the same relation holds between the vocalized
act and the subsequent effect it has in getting others to pay attention to the
accomplishments, deeds, and character attributed to the individual. This map-
ping has been lexicalized in the verb trumpet meaning ‘to praise’. By employ-
ing a conventional mapping between encomium and trumpeting, the blend in
blowing your own horn conforms to the unpacking principle.
Conformity to this entrenched mapping is also evident in the following at-
tested use of the blowing your own horn blend from a story in the Metropolitan
section of the New York Times, September 22, 1998:
I firmly believe that if you’re doing something interesting, you ought to tell
people about it,” Dr. Olivia said. “And if you’re blowing your own horn, do it
loudly. There’s no sense giving it a little toot.”
In this example, Dr. Olivia elaborates the blowing your own horn blend with a
self-conscious distinction between “blowing” your own horn, and “giving it a
little toot.” Focusing on the loudness-softness gradient, Dr. Olivia suggests a
mapping between the manner of articulation in the blend and the efficacy of
the bragging. Soft horn blowing in the blend maps onto less, and less noticeable
praise. This elaboration suggests a mapping between the degree of praise and
the volume of the trumpet sound – the same mapping that underlies the mean-
ing of the metaphoric expression muted praise. This in turn implies a mapping
between the praise and the sound, the speaker and the trumpeter. The map-
pings between elements of a typical Encomium space and a ritualized space we
call Trumpeting Royals are listed below.
Typical Trumpeting
Encomium Royals
======== =========
speaker trumpeter
hero royal
praise sound
voice trumpet
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.19 (69)
But, as noted above, the idiom refers to an atypical Encomium space in which
the speaker is the object of his own praise.
Typical Trumpeting Atypical
Encomium Royals Encomium
======== ========= ========
speaker trumpeter speaker
hero royal speaker
praise sound praise
voice trumpet voice
A topology-preserving mapping would proceed as in the following:
Typical Trumpeting Atypical Atypical
Encomium Royals Encomium Trumpeting
======== ========= ======== ========
speaker trumpeter speaker trumpeter
hero royal speaker trumpeter
praise sound praise sound
voice horn voice horn
In contrast, the idiom employs a mapping scheme like this:
Typical Trumpeting Atypical Blended
Encomium Royals Encomium Space
======== ========= ======== =========
speaker trumpeter speaker trumpeter
hero royal speaker horn
praise sound praise sound
voice horn voice horn
The violation of the topology principle inherent in the mapping between the
horn in the blended space and the object of the praise (viz. the speaker) in
the atypical Encomium space is supported by a conventional metonymic map-
ping between musicians and their instruments, as in The trombone is at his AA
meeting tonight. Or, an agent talking to a record producer Well, I can get you
a drummer, two guitars, and a bass, but you’ll have to find your own horn. As
in the examples discussed in previous sections, the metonymic mapping be-
tween horn and trumpeter in blowing your own horn makes it possible to sus-
tain a metaphoric interpretation of the model in the blended space despite the
violation of the topology constraint.
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.20 (70)
. Sculpture
Our analysis begins by considering the two input spaces that prompt the
construction of a third blended space.
Presentation Input Apocalypse Input Blended Space
============= ============= ===========
4 anthropomorphic 4 horsemen 4 anthropomorphic
figures figures
1 equine figure 4 horses 1 equine figure
The first space is a Presentation space contributing knowledge of artistic
medium: clay and glaze. Specifically, this space includes four anthropomor-
phic figures astride a single equine figure. The Apocalypse space contributes
minimal information about the Four Horsemen from the Book of Revelations,
namely that there are four horses and four horsemen, and that their ride her-
alds the end of the world. Viewers more familiar with the biblical account will
open a version of this space that includes knowledge that the four horsemen
are, themselves, personifications of Conquest, War, Plague, and Famine, each
of which mounts a horse of a different color: Conquest rides a white horse; War
rides a red horse; Famine rides a black horse; Plague rides a pale horse. The
two input spaces map counterparts onto each other via a similarity connec-
tor, since accessing each space depends on a relation of resemblance between
elements in each space. The established similarity mapping, in turn, allows ref-
erential structure in one space to trigger referential structure in the other. For
instance, Todd can now remark to Seana, “That hideous mass of clay predicted
the end of the world in 1942,” since referring to the medium of representation
can provide indirect mental access to the entity represented.
More interesting metonymic issues come to light as we consider features of
the composed blended space. In the blend, four anthropomorphic clay figures
sit astride a single equine figure, pictorially representing the Four Horsemen of
the Apocalypse quite differently from the way they are represented in the Apoc-
alypse space. This is due, in part, to material constraints imposed by the Presen-
tation space, a mental space that determines final material shape of the statue.
That is to say, conceptual integration in this blend works optimally only if the
representations can be compressed into one tightly integrated form. Present-
ing four figures astride one horse satisfies this integration constraint because it
makes efficient use of the Presentation space to present an integrated scene.
Moreover, the viewer’s attempt to satisfy the good reason (relevance) con-
straint might result in the construction of a mapping between the integrated
horse-and-riders scene as a snapshot of a singular, coordinated activity.
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.22 (72)
Japan
Hirohito
Time: 1942
Goal: World Domination
Death Space
====
Figure of Death
(i.e. Grim Reaper)
Goal: Cause Death
Blended Space
====
Hitler
Mussolini
Hirohito
Japanese Flag
Figure of Death
Missile
Horse
Time: 1942
Goal: World Domination and End of Humanity
The Axis Powers space is structured gradually as the viewer identifies the fig-
ures in the statue as Adolf Hitler, Benito Mussolini, and Emperor Hirohito. In
mental spaces nomenclature, each clay figure from the Presentation space maps
onto each referent in the Axis Powers space by a similarity connector. Once
this iconic relationship is established, viewers produce a value-role mapping as
each is construed as the leader of his respective country: Hitler is Leader of Ger-
many, Mussolini is Leader of Italy, and Hirohito is Emperor of Japan. The Axis
Powers space represents the figures as intentional agents acting in coordina-
tion with one another. In this space, each leader stands metonymically for each
nation, which, in turn, is understood as part of a corporate entity: a political al-
liance. Interpreters familiar with modern European history will access relevant
background knowledge about the Axis nations, such as the fact that the first
1936 alliance between Germany and Italy (known as the Rome-Berlin Axis)
was followed by a second 1940 alliance with Japan and, tangentially with Hun-
gary, Finland, Bulgaria, and Romania. In the Axis Powers space, the individual
nations act as one group.
With the blended construal, the formal features of the sculpture take on
new significance. For example, the interpreter may understand the referent
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.24 (74)
scene as an alliance wherein the single horse stands for the Axis Powers, and
the singular, coordinated event stands for the intentions and actions of each
nation under the alliance. In the blend, but in none of the input spaces, rid-
ing the horse stands for the sustained, coordinated effort of the three principal
nations to conquer the world, an inference licensed by the leader for na-
tion metonymy. Consequently, the interpreter does not only see Hitler, Mus-
solini, and Hirohito riding the horse, she sees Germany, Italy, and Japan acting
together in a military alliance.
Although the statue involves integration of conceptual structure from the
Apocalypse space with that in the Axis Powers space, three of the four horses
from the Apocalypse Input are omitted from the blended space. This occurs
because of positive pressure to accommodate structure from the Axis Powers
space, as well as an absence of pressure to preserve the precise topology of the
Apocalypse space. In the blend, as in the Axis space it projects to, the corporate
actions of the Axis powers manifest all the evils of the apocalypse in one politi-
cal alliance. The image of the three axis leaders riding a single horse in the blend
can be mapped onto their coordinated actions in the military alliance. More-
over, while the knowledge that the four horsemen of the apocalypse herald the
end of the world is important for producing the inferential implications of the
blend, the establishment of a precise mapping between particular leaders and
particular horsemen of the apocalypse is not. Consequently, there is no need
to preserve the metonymic mapping between horse color and personified evil,
noted above in our discussion of conformity to the good reason (relevance)
constraint.
The blend represents world conquest in terms of horseback riding, thus
compressing the complex chain of events involving millions of people to a
much more human scale activity involving four people and a horse. More-
over, the completed blend presents a dynamic event whereby the Axis powers
ride the horse of the apocalypse. The completed blend also takes on a distinct
temporal dimension, wherein the activities of the horse and the horsemen are
playing out in 1942. At this point, the meaning of the sculpture’s base takes
on new significance. With respect to the Presentation space the base is purely
functional, allowing the sculptor to display his figures with proper perspective.
But once the interpreter recognizes that the base is the entire Northern hemi-
sphere, she completes the blend in which the leaders of the Axis nations are
currently conquering the entire Northern hemisphere, which, in effect, stands
for Western civilization itself.
We have chosen to deal lastly with the most salient figure in the sculpture:
Death. In the sculpture, Death wears a German uniform, and appears to be the
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.25 (75)
figure actually riding the horse (the others appear as passengers). The inter-
preter recognizes the figure as Death because its face appears as a skull, thus
prompting the well-established metonymic compression of cause and effect,
where the effect of corporal decomposition comes to stand for its own cause,
death. As Turner (1987, 1991) has noted, the figure of the Grim Reaper is an
example of the generic is specific mapping (i.e., Death heralds the death of
an individual). The mere presence of death among these figures heralds the
death of the West as we know it, a very salient and plausible scenario in 1942.
The introduction of Death as the fourth horsemen comes about by virtue of
metonymic attribution to Death of elements from other mental spaces in the
blend. In our account, the presence of Death in the Apocalypse space automat-
ically opens a mental space for representations of Death as the Grim Reaper,
the common representation of death in Judeo-Christian lore.
It would be odd, however, to represent Death with his traditional priestly
cowl, robe, and scythe. Instead, Schreckengost represents him in a German uni-
form, carrying a missile in his right hand. It seems that to bring in wholesale
the figure of Death means violating the unpacking constraint, insofar as typi-
cal personifications of Death space come “packaged” with the features just de-
scribed. In this instance, violating the unpacking constraint satisfies the good
reason constraint. A priestly cowl and scythe do not have the same degree of
geopolitical relevance in 1942 as a German uniform and bomb do. The fact that
Death has to be wearing something means that clothing and accessories can be
projected from any mental space in the network.
Further, Death’s appearance exploits metonymic relations established in
the Axis Powers space, such that military uniform and bomb evoke both instru-
ments of war and effects of the war. Considered alone, each of these elements has
the potential to metonymically evoke various aspects of war. The military uni-
form, for example, is a salient part of the soldier’s appearance; the destructive
effects of a bomb are a salient aspect of its intentional construction; and death
is a salient side effect of war. Moreover, when presented together in an inte-
grated scene, each potential metonymic interpretation serves to reinforce the
others so that Death, the German soldier, serves as a cause-effect compression
of the instruments of war with the fatal effects of war. In fact, besides satisfying
the relevance constraint, the depiction of Death, the German soldier, helps to
optimize the integration principle.
How? We already know that Schreckengost has to choose a fourth horse-
man to complete his allegorical allusion, but unless he is going to introduce,
for instance, the leader of Hungary or Romania or Bulgaria or Finland (none
of whom are particularly notorious), he must choose a figure that does not vi-
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.26 (76)
olate the topology of the Axis Powers space. Stalin, for instance, would have
been an appropriately menacing choice (even in 1942), but would have dis-
integrated the corporate image, since Stalin and Russia were enemies to the
Axis nations. Choosing the personification of Death as the fourth horseman
(i) is appropriately menacing, (ii) preserves specific topology of the Apoca-
lypse space, and (iii) does not violate the topological relations recruited from
the Axis Powers space.
This brings us to the central ambiguity of the piece. Who is responsible for
the apocalypse? Like Death, the four horsemen are carrying out a divine plan
for the end of the world. In the Judeo-Christian tradition, especially, Death
is personified as a herald of death, and its heralding is understood to be the
proximate cause of an individual’s death. In other words, Death is not act-
ing of his own volition (in fact, it is not clear that Death has any volition at
all), it is merely acting out a divine mandate. But, is the interpreter to sup-
pose that Hitler, Hirohito, and Mussolini, as agents of the apocalypse, are also
executing a divine plan? Schreckengost’s own commentary suggests as much,
when he writes, “In the Four Horsemen of the Apocalypse . . . I saw a strange
resemblance to the four beasts let loose on the world today” (Adams 2000: 61).
References to the four beasts, then, refer to the leaders, with Death repre-
senting the results of their actions; however, Schreckengost’s use of the quasi-
modal verb phrase “let loose” suggests a more powerful entity permitting them
to act, lifting the barrier that holds them from the rest of the world. That an ex-
ternal and more powerful entity is being referred to is not in question. What is
in question is what or who is the ultimate instigator of these events? Is it God?
Or, is it the sum total of human actions – including World War I, the Treaty of
Versailles, the collapse of the Weimar Republic in 1933, the United States terri-
torial control over Hawaii, Chamberlain’s policy of appeasement, and so on –
that brought forth these beasts? Or, is it some combination of human folly and
divine retribution? All of these are plausible interpretations for metaphoric and
metonymic mappings.
As complexity increases, trade-offs between optimality principles become
inevitable. Comprehending Schreckengost’s sculpture involves maintaining
certain topological relations from the Apocalypse and Axis Powers spaces such
that a total of four horsemen appear and that three of them represent the lead-
ers of the Axis Powers nations. The final blend integrates the biblical and the
historical by preserving these topological relations. However, other topolog-
ical relations, such as the precise analogical mappings between horses color,
personified evil as rider, and political figure, are not preserved in the blend.
The overriding constraints in this example are integration and relevance as the
TSL[v.20020404] Prn:13/06/2003; 10:27 F: PB11303.tex / p.27 (77)
. Conclusion
References
Hofstadter, Douglas (1995). Fluid Concepts and Creative Analogies: Computer Models of the
Fundamental Mechanisms of Thought. New York: Basic Books.
Holyoak, Keith & Thagard, Paul (1995). Mental Leaps: Analogy in Creative Thought.
Cambridge, MA, and London: The MIT Press
Lakoff, George (1990). The invariance hypothesis: Is abstract reason based on image
schemas? Cognitive Linguistics, 1, 39–74.
Lakoff, George (1993). The contemporary theory of metaphor. In A. Ortony (Ed.), Metaphor
and Thought (pp. 202–251). Cambridge: Cambridge University Press.
Lakoff, George & Turner, Mark (1989). More Than Cool Reason: A Field Guide to Poetic
Metaphor. Chicago and London: The University of Chicago Press.
Langacker, Ronald W. (1987). Foundations of Cognitive Grammar, Vol. 1: Theoretical
Prerequisites. Stanford, CA: Stanford University Press.
Langacker, Ronald W. (2000). Grammar and Conceptualization [Cognitive Linguistics
Research 14]. Berlin and New York: Mouton de Gruyter.
Miller, George (1957). The magical number seven, plus or minus two. The Psychological
Review, 63, 81–97.
Nunberg, Geoffrey (1978). The pragmatics of reference. Bloomington: Indiana University
Linguistics Club.
Nunberg, Geoffrey (1995). Transfer of meaning. Journal of Semantics, 12, 109–132.
Oakley, Todd (in preparation). A Grammar of Attention.
Panther, Klaus-Uwe & Radden, Günter (Eds.). (1999). Metonymy in Language and Thought
[Human Cognitive Processing 4]. Amsterdam and Philadelphia: Benjamins.
Pelyvás, Péter (2000). Metaphorical extension of may and must into the epistemic domain.
In A. Barcelona (Ed.), Metaphor and Metonymy at the Crossroads: A Cognitive Perspective
[Topics in English Linguistics 30] (pp. 233–250). Berlin and New York: Mouton de
Gruyter.
Radden, Günter & Kövecses, Zoltán (1999). Towards a theory of metonymy. In
K.-U. Panther & G. Radden (Eds.), Metonymy in Language and Thought (pp. 17–59).
Amsterdam and Philadelphia: Benjamins.
Sperber, Dan & Wilson, Deirdre (1995). Relevance: Communication and Cognition. Oxford:
Blackwell.
Talmy, Leonard (2000). Toward a Cognitive Semantics, Vols. 1 and 2. Cambridge, MA, and
London: The MIT Press.
Turner, Mark (1987). Death is the Mother of Beauty: Mind, Metaphor, Criticism. Chicago and
London: The University of Chicago Press.
Turner, Mark (1990). Aspects of the invariance hypothesis. Cognitive Linguistics, 1, 247–255.
Turner, Mark (1991). Reading Minds: The Study of English in the Age of Cognitive Science.
Princeton: Princeton University Press.
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.1 (81)
Antonio Barcelona
. Introduction
The study of pragmatic inference has attracted linguists and language philoso-
phers for a long time, as a way of overcoming the limitations of a strictly “log-
ical” approach to the investigation of linguistic meaning. Speech act theory
and the theory of cooperative maxims (Grice 1975) greatly advanced our un-
derstanding of the way in which linguistic communication operates. Relevance
Theory (Sperber & Wilson 1986/1995), which privileges – as a sort of “super-
maxim” – the principle of relevance, has taken this understanding to a very high
degree of sophistication.
Jokes and funny anecdotes, especially those that place heavy demands on
the listener’s inferencing work, are particularly interesting, both for the prag-
maticist and for the cognitive linguist. Indeed, the study of humor has attracted
the attention of pragmaticists. Victor Raskin (1985: 103) has applied semantic
script theory and Gricean maxims to the study of humor, proposing a specific
set of “humor maxims”:
(1) Quantity: Give exactly as much information as is necessary for the joke.
Quality: Say only what is compatible with the world of the joke.
Relation: Say only what is relevant to the joke.
Manner: Tell the joke efficiently.
Attardo (1990) says that what a speaker does by telling a joke is to violate,
not just flout or exploit, a normal conversational maxim. But, Attardo claims,
when he does so, he is also being cooperative, because he expects his listeners
to switch to a “humor mode” once the punch-line of the joke has been reached.
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.2 (82)
Antonio Barcelona
Grice noted that by violating one of his “standard” maxims the speaker “will be
liable to mislead” (1975: 49). And, in Attardo’s view, this is exactly what hap-
pens in a straightforward interpretation of the text of a joke right before the
punch-line; the punch-line defeats this interpretation and the hearer then has
to seek an alternative interpretation. To accomplish this, the speaker replaces
the normal version of the maxims by Raskin’s special set of humor maxims.
In this process, he is normally helped by the speaker’s cooperative attitude –
a faint smile at the right moment, a wink, etc. So joke production and under-
standing is a kind of cooperative behavior, although it necessarily violates stan-
dard conversational maxims. The jokes and anecdotes analyzed below consti-
tute good examples of this cooperative behavior. In one of the anecdotes (“The
ironic doctor”), the real punch line, however, consists in the fact that one of the
characters fails to get the ironic intent of the doctor’s reply.
Attardo claims that the fact that the Gricean maxims of relevance and
quantity are necessarily violated in every joke seems to be evidence of the cor-
rectness of Relevance Theory (the other three maxims, quality, quantity, and
manner being subsumed under relevance); this necessary violation of quan-
tity and relevance also seems to support Horn’s (1984) proposal to reduce the
maxims to two super-maxims ‘Q’ and ‘R’.
Marín Arrese (1998) locates the main factor motivating a joke in the incon-
gruity of scripts, the “clash of worlds,” that, once it is clear to the listener that
a “joke-frame” has been created, leads to the operation of a new cooperative
principle.
My position will be neutral as regards the number or the status of these
communicative principles in humor understanding in the analyses that I will
be presenting below, since the goal of this paper is not to discuss the correct-
ness of the positions maintained by pragmaticists on these issues. I will simply
list some of the inferences that appear to become available to the listener of a
joke or a funny anecdote. However, I agree with relevance theorists that the
list of inferences to be drawn from an utterance is often open-ended (Wilson
1997). I also agree with both Raskin’s and Attardo’s insightful claim that humor
comprehension involves some kind of adjustment or change from the cognitive
script or frame (in Fillmore’s (1985) sense) that supports the initial, straight-
forward interpretation of the text of a joke right before the punch-line, to a
new script or frame. Therefore, I will also indicate for each joke which type of
frame adjustment takes place in it.
The goal of this paper is to study the role of metonymy in pragmatic in-
ferencing. The comprehension of a joke and of similar types of discourse of-
ten involves complex inferential chains. The wonderful fact is that people of-
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.3 (83)
ten activate such chains at lightning speed. How is this possible? Commu-
nicative principles like those put forward by pragmaticists, doubtless reduce
the cognitive effort required to arrive at the intended interpretation. But in
many if not all cases, these inferential processes are facilitated by pre-existing
metonymic connections in a cognitive frame (Panther 1994; Panther & Thorn-
burg 1998; Thornburg & Panther 1997), or by pre-existing metaphorical con-
nections across frames. Metonymic connections, in particular, seem to lie at
the very heart of pragmatic inferencing, as I hope to show by means of the
present paper.
. Metonymy
Antonio Barcelona
Metonymies are “mappings” in the sense that the source domain is connected
to the target domain by imposing a perspective on it. That is, the target domain
is understood “from” the perspective imposed by the source. This is one of
the reasons why, e.g. personal pronouns are not necessarily metonymies, even
though their abstract meaning (e.g. ‘third person, singular’) might anaphori-
cally “activate” their antecedents: they are not mapped onto them in this sense.
Saying that metonymy is a type of mapping is more adequate than saying that
it is a “stand-for” relationship, since the source does not necessarily substitute
unambiguously for the target: it merely activates it from a given perspective.2
To illustrate the above points, let us examine the sentence Picasso is not easy
to appreciate. In this sentence, picasso’s artistic work is a metonymic target,
and the activation of this target is carried out from the source Picasso, in his
role as artist, with the result that the hearer/reader is invited to conceptualize
this artistic work primarily as the outcome of Picasso’s artistic genius – as an
extension of his personality –, other aspects of this work being backgrounded.
Another fundamental property of metonymy is that the source maps onto
and activates the target in virtue of the experiential (hence pragmatic) link
between the roles each of them performs in the same “functional domain”
(i.e. a frame in Fillmore’s terms, or an ICM in Lakoff ’s terms).3 Fauconnier
(1997: 11) regards metonymy as a “pragmatic function mapping”; and Lakoff
(personal communication) claims that in metonymy the activation of source
role X brings about the activation of target role Y, both in the same concep-
tual frame or icm. In the previous example, the role artist (which is a domain
in the artistic activity frame) is pragmatically linked to the role art work
(another domain in the same frame), so that activation of the former normally
leads to activation of the latter.
The schematic definition of conceptual metonymy in (2) is based on
Kövecses and Radden’s (1998: 39) general definition of conceptual metonymy,
but it is more constrained than theirs. In Barcelona (in press), a set of addi-
tional specific definitions is proposed for the other general kinds of metonymy
represented by each of the various different phenomena (“purely schematic,”
“typical,” “prototypical”), which are covered by the schematic definition. Some
examples of the range of phenomena covered by the definition in (2) are (3),
(4), (5), (6) and (7).
(3) This book weighs two kilograms.
(4) This book is highly instructive.
(5) Belgrade did not sign the Paris agreement.
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.5 (85)
In example (3), the verb form weighs acts as a trigger to activate the physical
object sub-domain within the source domain book. That is, the whole do-
main book is mapped onto its sub-domain physical object, which is thus
mentally activated. In (4), the whole domain book is mapped onto its sub-
domain semantic content, which is thus mentally activated. Example (3)
would be a “peripheral” or “purely schematic” instance of metonymy, and
(4) would be closer to typical metonymies.4 As we can see from these ex-
amples, metonymy is, under the definition in (2), a very common, in fact
omnipresent, phenomenon in most linguistic expressions. This broad con-
ception of metonymy is, in fact, not exceptional in Cognitive Linguistics. A
similar conception underlies, for instance, Langacker’s notion of “active zone”
metonymies (Langacker 1993, 1999).5
Example (5) is a “prototypical” instance of metonymy, as it is referential
and as it has an individual (the Yugoslavian government is a collective individ-
ual) as target. Examples (6) and (7) are simply “typical” metonymies as they
are not referential; furthermore, the target in (7) is not an individual, but a
property (an emotional state). What is conventionally believed to be a possi-
ble behavioral effect of sadness (walking with drooping shoulders) activates its
cause (the emotion itself), so that an automatic inference is that the person
exhibiting this bodily behavior is sad.
I will start with relatively simple examples and then I will present a case that
requires extensive inferential work on the part of the reader/listener.
Context
As part of what is known in medicine as anamnesis (recalling medically relevant
facts), a pediatrician asks a young mother a number of routine questions about
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.6 (86)
Antonio Barcelona
her two-month-old baby: pregnancy, birth, weight at birth, and vitality and
vigor of the baby. With regard to the last point (vitality and vigor), the doctor
asks a very simple question and gets a delightfully naive answer.
Text
Doctor: And does your baby normally hold on tight to your breasts?
Mother: Oh, yes, doctor, just as if he were an adult!
Frame adjustment
Frame overlap, frame blend and frame shift. The nurturance frame unex-
pectedly overlaps with, and shifts to, the sexuality frame via a blend of the
nurturance frame with the adulthood frame.
As Raskin (1979: 332) claims, “much of verbal humor depends on a partial
or complete overlap of two or more scripts all of which are compatible with
the joke-carrying text.” I thus will distinguish in the ensuing analyses between
“frame overlap” (partial overlap) and “frame blend” (complete overlap). Frame
overlap takes place when in the conceptual world created by (part of) an utter-
ance or a text two cognitive frames remain clearly distinct but are linked by
means of a shared conceptual substructure; in the case at hand, nurturance
and sexuality share a minor sub-frame or sub-domain: the grasping by a
male of a woman’s breasts. Frame blend occurs when two mutually inde-
pendent frames are fused into an imaginary mental scene, or mental space.
Under the most likely interpretation of the above anecdote, the mother, in or-
der to highlight her baby’s vigor and vitality, merely intends to set up a counter-
factual mental scene in which, when breast-feeding, her baby is simultaneously
a baby and an adult.7 In fact, both types of frame interaction can be handled as
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.7 (87)
Antonio Barcelona
(invoked by term
’ adult)
Metonymy → Inference 1 Metonymy → Inference 2
[]
’
()
Context
The patient, a woman, is the type that never trusts a doctor. She asks an
impolite question and receives an ironic reply, which she takes literally.
Text
Patient: Excuse me, but have you been to medical school to get your M.D. degree?
Doctor: No, madam, I just got it at a lottery.
(After this, the patient files a complaint writing, in all seriousness, that she
cannot understand how a government-supported health center, “which is paid
for with our tax money,” has hired a doctor who obtained his degree at a lottery.
As this case shows, the main thrust of an ironic remark may fail to be grasped
by an obtuse addressee.)
Frame adjustment
Frame overlap, frame blend and frame shift. The medical education frame
unexpectedly overlaps with the lottery frame, both in the reality frame or
mental space, to create a counterfactual blend of both frames, which, in the
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.10 (90)
Antonio Barcelona
frame of reality, invokes the absurdity frame. The main inference is, thus,
due to a shift to the latter frame. However, this shift, intended by the doctor, is
not carried out by his literal-minded patient.
The medical education and the lottery frames are separately included
in the reality supra-frame. But the blend of these frames that we find in this
joke is inevitably confronted with the reality supra-frame or mental space,
which shows that participating at a lottery is not equivalent to studying at a
medical school as a condition for obtaining a medical degree; therefore, infer-
ence 2 is a counterfactual proposition in reality space. And when one reasons
within the reality frame, this blend automatically invokes the notion of absur-
dity. One of the things that enters the absurdity frame is the belief in the truth
of counterfactual situations.11 Two instances of this belief, which metonymi-
cally activate the whole frame, are the belief that doctor’s degrees are given out
at lotteries, and the belief that a doctor’s degree can be awarded to a person
who did not go to medical school.
Inference 1 arises on the basis of the metonymy condition for result. This
metonymy operates within the medical education frame. A pre-condition
for obtaining a medical degree is attending medical school. If the fulfillment of
the condition is questioned, so is the fulfillment of the result.
Inference 2 arises on the basis of the metonymy result for condition. Win-
ning a doctor’s degree at a lottery (result) stands for its condition (the fact that
medical degrees can be earned just by buying a lottery ticket). This metonymy,
thus, operates within a counterfactual blend between the medical education
and the lottery frames. The blend is possible because there is an overlap (a
shared sub-schema) between both frames: both exhibit a connection between
a (pre)condition and a result.
Inference 3 arises on the basis of the metonymy entity for its conven-
tional property. The entity here is a propositional entity (the plainly coun-
terfactual belief that doctor’s degrees can be given out as lottery prizes); the
“defining property” here is a property (absurdity) that is definitionally pred-
icated of this entity in reality space; this property is mentally activated by
the mental activation of the entity. As a result the absurdity frame is also
invoked.12
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.11 (91)
Inference 4 arises, on the one hand, on the basis of the same metonymy, en-
tity for its conventional property. The entity here is a propositional en-
tity, i.e. the belief (a plainly counterfactual belief) that doctors can get their
degree without going to medical school. And the conventional property of this
belief is again absurdity.
REALITY SUPRA-FRAME
Inferences 2–5 are counterfactual beliefs
( )
Metonymy → inferences 3–5
( )
Antonio Barcelona
that the same metonymic connection between a result and a condition is pre-
served in the blend of the two frames.
The following joke attributed to W. C. Fields is taken from Attardo (1990: 355).
Context
No context is provided
Text
Speaker A: Do you believe in clubs for young men?
Speaker B: Only when kindness fails.
Inferences
a. Meant and perhaps conveyed by Speaker A:13
1. Speaker A wants to know whether Speaker B believes in the conve-
nience, usefulness, etc. of (social) clubs for young men.
Frame adjustment
Frame overlap due to homonymy and frame shift. The club frame (a sub-
frame of the social institutions frame) overlaps with the conflict frame
(a sub-frame of the human interaction frame), and the interpretation shifts
to the conflict frame. The overlap between club and conflict is due to the
homonymy of the form clubs, which can correspond to two different lexemes:
‘club-1’ (a social institution) and ‘club-2’ (a heavy stick).14
Context
The anecdote is usually reported as taking place in the mid-1930s in Spain, dur-
ing a parliamentary session. An opposition M.P. concludes his savagely aggres-
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.14 (94)
Antonio Barcelona
sive speech against the Prime Minister with a seriously offensive accusation.
The M.P. has no evidence for his accusation; he merely intends to unsettle the
Prime Minister. Yet the latter retorts very cleverly. His strategy consists of ac-
cepting, for the sake of irony, the truth of the literal interpretation of the M.P.’s
utterance, thereby triggering a very different set of inferences.
Text
Opposition M.P. (referring to the Prime Minister): But what can we expect, after
all, of a man who wears silk underpants?
The Prime Minister (rising calmly): Oh, I would never have thought the Right
Honorable’s wife could be so indiscreet!
Inferences:
a. Meant and conveyed by the opposition M.P.
1. The Prime Minister is a homosexual.
2. The Prime Minister is unfit for office.
Frame adjustment
Frame overlap and frame shift. The underwear frame overlaps with the ho-
mosexuality frame. This frame, in turn, overlaps with the discretion frame,
which finally shifts to the adultery, heterosexuality, and stupidity frames.
That is, the M.P. implies that the Prime Minister is a homosexual, by accusing
him of wearing silk underwear (homosexuality). And the Prime Minister at-
tributes the knowledge of this fact to an indiscretion of the M.P.’s wife, thereby
linking this information about his supposed homosexuality to the woman’s in-
discretion (discretion frame). But then the most likely explanation for the
M.P.’s wife’s knowing such intimate details of the Prime Minister’s life is that
she has had a sexual affair with him, i.e., that she has been an adulteress, that
her husband has been cuckolded (adultery), and that the Prime Minister
is a heterosexual (heterosexuality). A further possible major inference re-
sulting from the previous inferences is that the M.P. has acted very foolishly
(stupidity) by revealing these facts in Parliament.
Inference 2 arises from the metonymy linking an entity to one of its conven-
tional properties (women and homosexuals were then thought to be unfit
for performing important social functions).
Inference 4 arises on the basis of the metonymy cause (the wife’s indiscreet
behavior) for effect (the fact that the M.P. knows the secret). It also arises on
the basis of another metonymy of the type result (the fact that the M.P. has
publicly revealed the secret) for precondition (knowing or having been told
about the secret).16
Antonio Barcelona
The first part of inference 6 also arises on the basis of the metonymy fact (see-
ing someone undress) for one of its conventional explanations (having
a sexual affair with that person). One of the explanations of the fact that a
woman has seen a man undress is that she has had, or has been about to have,
a sexual encounter with that man. The second part of this inference is due to
the metonymy definition (a married woman having sex with a man other
than her husband) for defined (the behavioral category called adultery).
Spelling out the definitional properties of a category (of behavior, in this case)
can automatically invoke the category.
Inference 7 arises on the basis of the metonymy behavior (having sex with a
woman) for kinds of people conventionally associated with that be-
havior (heterosexuals). Note that this is the same metonymy that led to the
offensive inference intended by the M.P. and listed above as inference 1, except
with source and target specified by different domains. The Prime Minister thus
seems to have, very skillfully, used exactly the same metonymy as the one used
by the opposition M.P. to suggest exactly the opposite inference.
Inference 8 arises on the basis of the metonymy definition (one’s wife hav-
ing a sexual affair with another man) for defined (the category of cuckolded
husbands).
among other implications, that the Prime Minister is not a homosexual and
that the M.P. is a cuckolded husband).17
. Conclusions
Metonymy has been shown, at least as far as the above sample of jokes and
anecdotes is concerned, to be at the basis of all of the pragmatic inferences
that can be drawn from these humorous pieces. One may disagree with the
exact naming of the metonymies proposed above, but it is undeniable, in my
view, that in each of them, the domain, concept or frame presented as source is
normally a mental activator of the domain, concept or frame presented as tar-
get. Therefore, a general conclusion that emerges from this brief study is that
conceptual metonymies often provide “ready-made” pointers towards plausi-
ble inferential pathways in the interpretation of a joke or an anecdote, and, in
fact, in the interpretation of any other kind of discourse. These pointers, which
are normally automatic, contribute greatly to the ease and speed of interpreta-
tion. A reader or hearer of these stories might have drawn other inferences in
addition to the ones provided above. But my guess is that they would have been
drawn on the basis of some metonymy. A similar claim is made by Coulson and
Oakley (this volume), who stress the fundamental role of metonymy in main-
taining connections between distant mental spaces in reasoning. Metonymies
(and metaphors) also constrain the range of possible inferences to be drawn
from an explicit proposition, a point clearly made by Ruiz de Mendoza (1997),
and Ruiz de Mendoza and Pérez Hernández (this volume).
Metonymy thus seems to constitute the very skeleton of pragmatic infer-
encing. Now, does this mean that pragmatic inferencing can be reduced to
metonymic reasoning? There are other important aspects of pragmatic infer-
encing that cannot simply be accounted for in terms of metonymic connec-
tions, but rather in terms of pragmatic principles and rules necessary for a com-
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.18 (98)
Antonio Barcelona
Notes
* I am grateful to my anonymous reviewers for their useful suggestions and to the editors
of the volume for their careful work.
. I have discussed in detail most of these problems and properties, except for the “func-
tional domain” and the “pragmatic link” notions (see below) in Barcelona (in press).
. Most inference-prompting metonymies do not involve substitution of source by target,
but activation of the latter by the former, as in Panther and Thornburg’s example (this vol-
ume) General Motors had to stop production, which yields the implicature ‘General Motors
stopped production’. This implicature is prompted, according to Panther and Thornburg,
by the metonymy obligation for action for action. As an anonymous reviewer of the
present volume correctly points out, the target does not eliminate the source, but is rather
“added” to the proposition. However, I support Panther and Thornburg’s claim that this
addition is prompted by the metonymic connection between source and target.
. In Barcelona (n.d.) and Barcelona (2002) I have proposed that the cognitive domain men-
tioned in the definition should be a “functional cognitive domain” (i.e. a frame or ICM), and
not just a taxonomic domain. In both papers, and in Barcelona (in press), I have also pro-
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.19 (99)
posed that the mapping in metonymy is unidirectional and asymmetrical, whereas the one
in metaphor is unidirectional and symmetrical. By “symmetrical” I mean that each source
element has in its frame a structurally equivalent role to its counterpart in the target (e.g., in
the love is a journey metaphor, the lovers have a role in the “romantic love frame,” which
is structurally equivalent to the role of the travelers in the “journey frame”).
. The difference between “purely schematic” and “typical” metonymies lies in the fact that
the target in the former is a “primary” domain or subdomain of the source, whereas the
target in the latter is a comparatively more “secondary” domain or subdomain of the source,
or is outside it, as in part for whole metonymies. In (3) physical object is a primary
subdomain in book, whereas semantic content is comparatively a more secondary (or
less primary) subdomain. Purely schematic, typical and prototypical metonymies constitute
a continuum of metonymicity. Purely schematic metonymies are contextual semantic values
occurring in the “literal” use of expressions, which points to the artificiality of the strict
literal-figurative distinction. For details, see Barcelona (in press).
. These two metonymies are “active zone” metonymies in which the “active zone” of the
notion book is different in each case.
. The reader is reminded that the lists of inferences analyzed in the ensuing case studies
does not exhaust the full set of the inferences that might be drawn on the basis of some
metonymy. We should also distinguish, in principle, the inferences that seem to have been
actually made by the direct participants in the stories, from those made by indirect partici-
pants (i.e. their observers) or by the listeners or readers of these stories, as they are narrated.
I include in my analysis both types: the inferences intended or made by the actual direct
participants, and those made by other participants, or by readers or listeners. Another im-
portant caveat is that the order in which the inferences are presented does not mean that
they are necessarily arrived at sequentially. Nor by numbering do I commit myself to a spe-
cific real-time ordering of the inferences, even though this ordering appears to reflect the
conceptual ordering among them, i.e., inference 1 below appears to set up one of the con-
ceptual frames needed for inference 2, just as the latter sets up one of the frames needed for
inference 3.
. Under a slightly different interpretation, the mother intentionally blends the baby’s vigor,
vitality and eagerness to feed with the vigor, firm grasp of her breasts and desire of a sexually
aroused adult. This less likely interpretation would not necessarily diminish the woman’s
naiveté (she may just have thought this blend an effective way of forcefully picturing her
baby’s strength, without thereby intending to convey the third inference).
. See Turner and Fauconnier (1995) and Fauconnier (1997); see also Turner and Faucon-
nier (2000), and Coulson and Oakley (this volume) for the important role of metonymy in
blending, and Barcelona 2000, for its basic role in metaphor.
. This activation of a frame (a whole) by mentioning an item of its conventional vocabulary
(a representative part) is in turn a ubiquitous metonymy (see Lakoff & Turner 1989: 108).
. As stated above, what I call a “frame blend” in my analysis of this anecdote and in that
of the other jokes and anecdotes can be analyzed in terms of Turner and Fauconnier’s model
of conceptual integration. The mother’s reply in the dialogue with the pediatrician would be
analyzed as an instance of the integration of three “input spaces” (adulthood, nurturance,
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.20 (100)
sexuality) into a blended space. In my analysis, I have represented the sexuality frame outside
the blend to highlight the suggestion that the inference that the mother engages in frequent
sexual activity may be a later by-product of the blend between adulthood and nurturance.
In the conceptual integration model, the main inference would occur in the blended space.
I have shown that this inference is facilitated by metonymic connections imported from
the input spaces or frames, such as baby’s tight hold on mother’s breasts for baby’s
vitality, or adulthood for full vigor. For more examples of the important role of
metonymy in blending, see Coulson and Oakley (this volume).
. An anonymous reviewer challenges the absurdity frame. The absurdity frame is a very
general conceptual frame that includes people’s knowledge of what qualifies as absurd: un-
reasonable, foolish, or ridiculous situations, behaviors, ideas, beliefs, etc. For instance, pay-
ing a lot of money for an object one does not need at all, or welcoming the Queen of England
in one’s pajamas. Many absurd situations or beliefs are at the same time counterfactual, such
as the idea that people can fly simply by moving their arms up and down. This frame can be
invoked directly by such words as absurd, ridiculous, outlandish, etc. or indirectly, as in this
case, by mentioning a typically absurd situation.
. This metonymy often reflects personal stereotypes. For instance, in a recent interview
in a regional newspaper in Spain (La Verdad, Murcia, 16/6/2000), David Byrne complains
about being known only as the “ex-leader of Talking Heads,” the rest of his career being
overlooked. This example shows that an entity can automatically invoke its conventionally
salient property.
. Speaker B may have realized this straightforward interpretation yet he may have ignored
it on purpose, or (which is much less likely) he may simply have failed to notice it.
. These two lexemes, club (‘association’) and club (‘a heavy stick’) actually have a Scan-
dinavian origin: the Old Icelandic form klubba, klumba ‘club (stick), a mass of anything,’
which was introduced in Middle English, and which was a cognate to Old Swedish klubb
‘club, lump, log,’ which in dialectal variants could also mean (metaphorically) ‘a lump of
people, a knot of people.’ All of these forms are in turn cognate to clamp, clamber, clasp, and
clump, all of which retain the ideas of ‘grasping,’ ‘holding together,’ and ‘mass’. In the 17th
century the term club was re-introduced from this Swedish dialect with the metaphorical
meaning ‘a lump of people,’ i.e. an association. Source: Skeat (1993 [1884]). But the two
senses are now so different, and are felt to be so unrelated, that Skeat himself treats them
as homonymous lexemes, rather than as two senses of one and the same lexeme. I have fol-
lowed him here. However, present day standard dictionaries differ in their treatment of this
form, though most tend to follow the homonymy position.
. According to Langacker this metonymy (an “active zone” metonymy in which a partici-
pant stands for its active zone, consisting of a relationship in which it participates) motivates
“raising” constructions. It also motivates other constructions in which the subject NP is
covertly propositional: A car is a good idea today (i.e., ‘Having/using, etc. a car is a good idea
today’); This knife is convenient (‘Having/using, etc. this knife is convenient’) (Langacker
1999: 327–332).
. If I am to trust my sources, the M.P. was actually a married man at the time of this
exchange, which explains his speechlessness in the face of the retort.
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.21 (101)
. Another possible inference is that the M.P., after being told the secret by his wife, does
not realize all of its negative implications for himself. This would show him as doubly
foolish: first for not grasping these implications and second for unwittingly stating them
publicly.
. Ruiz de Mendoza and Pérez Hernández (this volume) claim that metaphor and
metonymy are two fundamental explicature-deriving mechanisms. Most of the inferences
I have presented in my paper are implicatures, whose derivation has been shown to be regu-
larly facilitated by metonymic connections. So metonymy is instrumental in obtaining both
types of “pragmatic implications”, to use Ruiz de Mendoza and Pérez Hernández’s terms.
References
Arana, José Ignacio de (2000). Diga treinta y tres: Anecdotario médico. Madrid: Espasa Calpe.
Attardo, Salvatore (1990). The violation of Grice’s maxims in jokes. Berkeley Linguistics
Society, 16, 355–362.
Barcelona, Antonio (2000). On the plausibility of claiming a metonymic motivation for
conceptual metaphor. In A. Barcelona (Ed.), Metaphor and Metonymy at the Crossroads:
A Cognitive Perspective [Topics in English Linguistics 30] (pp. 31–58). Berlin and New
York: Mouton de Gruyter.
Barcelona, Antonio (2002). Clarifying and applying the notions of metaphor and metonymy
within cognitive linguistics: an update. In R. Dirven & R. Pörings (Eds.), Metaphor
and Metonymy in Comparison and Contrast [Cognitive Linguistics Research 20] (pp.
207–277). Berlin and New York: Mouton de Gruyter.
Barcelona, Antonio (in press). Metonymy in cognitive linguistics. An analysis and a
few modest proposals. In H. Cuyckens, T. Berg, R. Dirven, & K.-U. Panther
(Eds.), Motivation in Language: Studies In Honor of Günter Radden. Amsterdam and
Philadelphia: John Benjamins.
Barcelona, Antonio (n.d.). The distinction between metaphor and metonymy: A question
of (a)symmetry? Paper presented at the Conference on Researching and Applying
Metaphor. University of Manouba, Tunis, April 5, 2001.
Fauconnier, Gilles (1997). Mappings in Thought and Language. Cambridge: Cambridge
University Press.
Fillmore, Charles (1985). Frames and the semantics of understanding. Quaderni di
Semantica, 6–2, 222–254.
Grice, H. Paul (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 41–58). New York: Academic Press.
Horn, Lawrence (1984). Toward a new taxonomy for pragmatic inference: Q-based and R-
based implicature. In D. Schiffrin (Ed.), Meaning, Form, and Use in Context: Linguistic
Applications (pp. 11–42). Washington, DC: Georgetown University Press.
Kövecses, Zoltán & Radden, Günter (1998). Metonymy: Developing a cognitive linguistic
view. Cognitive Linguistics, 9, 37–77.
Lakoff, George & Turner, Mark (1989). More than Cool Reason: A Field Guide to Poetic
Metaphor. Chicago and London: The University of Chicago Press.
TSL[v.20020404] Prn:13/06/2003; 10:34 F: PB11304.tex / p.22 (102)
P II
Metonymic inferencing
and grammatical structure
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.1 (105)
A construction-based approach
to indirect speech acts*
Anatol Stefanowitsch
. Introduction
The central issue concerning indirect speech acts (ISAs) has always been the
question of how the hearer arrives at the interpretation intended by the speaker.
Traditionally, scholars have held that in order to interpret an ISA, the hearer has
to do a certain amount of inferencing, i.e., that the propositional content and
the illocutionary force of the indirect speech act are arrived at via a stepwise
application of inferencing rules of some sort (Searle 1975; Grice 1975; Sperber
& Wilson 1986). More recently, it has been suggested that the ISA activates one
part of a cognitive model that then metonymically evokes the whole model or
some other part of it (Thornburg & Panther 1997; Panther & Thornburg 1998;
Panther & Thornburg this volume, cf. also Gibbs 1994: 351ff.). For example,
the cognitive model for a ‘giving’ event includes the core idea of an act of giv-
ing, but also various peripheral ideas, e.g. that the hearer is willing and able to
perform the act, that the speaker wants something the hearer has, etc. By refer-
ring to any part of the model (e.g. Will/can you give me that book?, I want that
book, etc.), the speaker can evoke the whole model (part for whole), or the
core of the model (part for part).
Researchers in both paradigms have recognized that indirect speech acts
can have different degrees of conventionality. For example, the utterances Can
you close the window? and It’s cold in here can both be used as requests to close
the window. Both are traditionally assumed to be indirect requests, but the first
type of expression is conventionally used as a request, while the latter is not.
This distinction between conventionalized and non-conventionalized indirect
speech acts is long-standing and has been discussed, among others, by Searle
(1975), Morgan (1978), Bach and Harnish (1979), and Clark (1979).
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.2 (106)
‘Billy threw the ball at Diane’ or ‘Billy threw the ball in the direction of Diane,’
but only ‘Billy threw the ball intentionally in such a way that Diane could catch
it’ (cf. Goldberg 1995: 34f.).
In Construction Grammar it is assumed that any given expression will
instantiate several constructions at once (Goldberg 1996: 68): the sentence
The doctor kicked his wife the ball instantiates the subject-predicate construc-
tion (i.e. [SUBJ PRED]), the ditransitive construction (i.e. [SUBJ V OBJ1
OBJ2 ]/‘SUBJ causes OBJ1 to receive OBJ2 ’), the past tense construction (i.e.
[V-ed]/‘past’), two types of noun phrase construction ([the N] and [POSS N]),
and the lexical constructions (i.e. words) ball, doctor, his, kick, the, and wife.
In Construction Grammar terms, an expression is a well-formed expression
of a language if it is an instantiation of the combination of existing construc-
tions (including morphemes and lexemes) of that language (sometimes de-
scribed as the unification of constructions). Construction Grammar, then, is
non-derivational, since constructions are seen as basic units rather than the
result of an interaction between lexical items and syntactic rules, and it is
non-modular, since constructions directly link form and meaning/use.
Crucially, Construction Grammar does not view the set of constructions
making up a particular language as an unstructured list of items. Instead, con-
structions are seen as forming a highly structured inventory; four types of
inheritance links between constructions are posited.
First, there are polysemy links, which show the relation between different
extensions from some central meaning (Goldberg 1995: 75). For example, the
ditransitive construction mentioned above is linked by a polysemy link to for-
mally identical constructions with the meanings ‘X intends to cause Y to receive
Z,’ as in Jim baked Mary a cake, ‘X enables Y to receive Z,’ as in Jim allowed Mary
some cake, ‘X causes Y not to receive Z,’ as in Jim denied Mary cake, etc.
Second, there are metaphorical extension links, which show relations be-
tween two senses of a construction that are based on conceptual metaphors
(Goldberg 1995: 81). For example, the ditransitive construction is linked by a
metaphorical extension to a formally identical construction meaning ‘X com-
municates Y to Z,’ as in Jim told Mary a story (based on the conduit metaphor,
cf. Lakoff & Johnson 1980).
Third, there are instance links, which show the relation between a gen-
eral construction and a more specific instance of this construction (Gold-
berg 1995: 79). For example, the ditransitive construction is an instance of
the subject-predicate construction; it has a subject and a predicate that share
the properties of subjects and predicates in general, but it adds its own spec-
ifications, namely the exact type of verb phrase instantiating the predicate
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.4 (108)
In this section I will argue that conventionalized ISAs such as those given in
(1) must be considered constructions in their own right, i.e. form-meaning
pairs whose properties cannot be strictly predicted from other constructions or
general principles in the grammar of English (I will refer to such constructions
as ISA constructions):
(1) a. Can you pass the salt?
b. Would you mind closing the door?
c. I’d like a cheeseburger with fries.
I will then argue that these ISA constructions are nevertheless motivated by
other constructions and general principles. Drawing on the theory of Speech
Act Metonymies (Thornburg & Panther 1997; Panther & Thornburg 1998),
I will suggest a new type of inheritance link to be added to the apparatus
of Construction Grammar: the metonymic link (IMy ). This link allows us
to capture the (partial) motivation behind ISA constructions, while at the
same time acknowledging their independent status. I will also argue that non-
conventionalized ISAs are different from ISA constructions. While the latter,
being constructions in their own right, do not require inferencing on the part
of H (although see further below), the former do require such inferencing.
In order to demonstrate the need for positing ISA constructions in the first
place, it must be shown that conventionalized ISAs fit the definition given in
Section 2 above, i.e. that they do in fact have formal or semantic properties that
are not strictly predictable from other constructions of English.
Can you pass the salt? is that it has the form of a question, but is not uttered to
mean a question, but instead is (typically) uttered to mean a request.
However, the fact that a particular type of ISA has a different meaning than
its direct counterpart is not enough to argue for its status as a construction; it
must also be shown that this meaning is not predictable. The question how in-
direct speech acts convey their intended meaning has been variously answered
in terms of Gricean implicatures (e.g. Searle 1975), inferencing on the basis of
‘mutual contextual beliefs’ (Bach & Harnish 1979), conversational postulates
(Gordon & Lakoff 1975), conventions of use (Searle 1975; Bach & Harnish
1979; Morgan 1978), or abstract performative verbs in their semantic struc-
ture (e.g. Sadock 1974). All but the last two theories claim that the meaning of
ISAs is completely predictable on the basis of the construction they are based
on and general principles of communication. If these theories are right, then
there is no semantic justification for positing ISAs, conventionalized or not, as
constructions in the sense of Construction Grammar.
In addition to differences in terms of illocutionary force, there is typically
a restriction on the semantic roles of requests that is not present in the corre-
sponding direct constructions. For example, the subject of Can you X? or Would
you mind X? must be an agent, and the object (if present), must be a patient.
Thus, (2a) could be a question or a request, but (2b) could only be a question:
(2) a. Can you close the window?
b. Can you see the window?
Can you X? have the same structure as questions, they have some additional
properties that questions do not have, or that they do not have some of the
additional properties of questions. Second, it must be shown that convention-
alized ISAs differ formally from other ISAs with the same illocutionary force.
For example, it must be shown that the formal properties that distinguish a
request like Can you X? from the corresponding question are not properties of
indirect requests in general.
Both types of evidence do, in fact, exist. Beginning with the first type, it
has been noted that while conventionalized indirect requests typically allow the
preverbal occurrence of request markers like please or kindly, the correspond-
ing direct constructions do not (cf. e.g. Sadock 1974: 104):3
(3) a. Can you open the door?
⇒ Open the door! / Are you capable of opening the door?
b. Can you please/kindly open the door.
⇒ Open the door! / *Are you capable of opening the door?
(4) a. Would you mind opening the door?
⇒ Open the door! / Do you believe it would have a negative psycho-
logical effect on you if you opened the door?
b. Would you mind please/kindly opening the door?
⇒ Open the door! / *Do you believe it would have a negative psycho-
logical effect on you if you opened the door?
Thus, there is clear evidence that conventionalized indirect requests differ for-
mally from the direct question whose form they appear to share. The existence
of such differences has been recognized by proponents of a purely pragmatic
analysis of ISAs (such as Searle 1975 or Bach & Harnish 1979), but has not been
satisfactorily accounted for (a point which we will return to presently). Be-
fore doing so, however, we must address a potential counterargument against
using facts about please and kindly as evidence for a special formal status of
conventionalized ISAs at all. Recall that in order to argue convincingly that
conventionalized ISAs are constructions, it is not sufficient to show that they
have formal properties that distinguish them from the direct speech act whose
general form they share. We must also show that they differ formally from
non-conventionalized ISAs, which have no claims to construction status. Af-
ter all, the formal differences between ISA constructions and their direct coun-
terparts could potentially be linked directly to their illocutionary force. More
precisely, since please is commonly considered to be a request marker, we might
be tempted to link its distribution to the use of an utterance as a request. Pre-
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.7 (111)
verbal please (and adverbs like kindly) can occur with a wide variety of formal
structures, as long as they are used as requests:
(5) a. You can please/kindly give that back now.
b. You will please/kindly close the window.
c. (For the hundredth time,) I please/? kindly want the salt.
However, note that there are many ISAs that can be unambiguous requests in a
particular situation, but that cannot occur with preverbal please (and in many
cases cannot occur with please at all). Example (5c) is already somewhat odd for
many speakers without the material in parentheses, and the following examples
are clearly out:
(6) a. *It’s please/kindly cold in here. ∼ *It’s cold in here, please.
b. *You please/kindly can’t keep that. ∼ *You can’t please/kindly keep
that. ∼ ??You can’t keep that, please.
c. *This sauce could please/kindly do with some salt. ∼ ?This sauce could
do with some salt, please.
I will briefly return to this criterion in connection with the concept of speech
act metonymy below.
In sum, conventionalized indirect requests have at least three formal prop-
erties that distinguish them both from the direct speech act on which they seem
to be based and from non-conventionalized indirect speech acts with the same
illocutionary force: they can occur with preverbal please or request adverbs;
conditional modals express politeness instead of conditionality; and they can
have a preposed subordinate clause stating the reason for the request.4 Note
that they share the first and the third property with direct requests:
(9) a. Please/kindly close the door!
b. Since I’ve got my hands full, close the door!
As mentioned above, at least one of these formal properties, the possibility for
preverbal please to occur, has been acknowledged by proponents of a purely
pragmatic analysis, e.g. Searle (1975) and Bach and Harnish (1975). Thus, be-
fore we are fully justified in assigning construction status to conventionalized
requests on the basis of these formal properties, we have to ensure that they
cannot be accounted for by such a purely pragmatic analysis.
Searle’s explanation consists of two parts; first, he argues that please “ex-
plicitly and literally marks the primary illocutionary point of the utterance as
directive, even though the literal meaning of the rest of the sentence is not di-
rective” (Searle 1975: 68). As was pointed out above, this observation does not
account for the fact that preverbal please can occur in conventionalized indi-
rect requests, but not in non-conventionalized ones. Searle also seems to be
aware of this fact (Searle 1975: 75). However, he does not offer an explanation;
instead, he simply observes that “certain forms will tend to become conven-
tionally established as the standard idiomatic forms for indirect speech acts”
(Searle 1975: 76). Thus, if there is an explanation implicit in Searle’s discussion
at all, it is the circular statement that conventionalized indirect requests accept
preverbal please because they are conventionalized.
Bach and Harnish (1979: 188f.) also criticize Searle’s account, but their
own suggestion is hardly more enlightening. They simply suggest that the oc-
currence of preverbal please in indirect requests is always ungrammatical and
that indirect requests with preverbal please are “examples of syntactic liberty
[. . . ], ungrammatical but usable sentences that are perfectly acceptable to flu-
ent speakers” (Bach & Harnish 1979: 199). Even if we ignore the problems in-
herent in the notion of “ungrammatical but usable and acceptable sentences,”
note that it does not account for the crucial fact that preverbal please can occur
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.9 (113)
In light of the vast literature on indirect speech acts, the above account may
seem somewhat lacking in explanatory value. After all, it simply pairs seman-
tic/pragmatic functions with certain syntactic patterns, without making any
attempt to motivate the latter in terms of the former or to account for the fact
that all the ISA constructions listed here have roughly homonymous syntactic
patterns that function as questions or statements. In other words, the analysis
sketched out here so far does not explain why the particular ISA constructions
under discussion share their basic formal properties with particular types of
questions/statements, rather than any other (arbitrarily chosen) pattern.
There are two issues that need to be addressed. First, what happens in the
present account to the long and impressive series of inferential steps that a
hearer must take under the traditional view in order to arrive at the correct
interpretation of Can you pass the salt? (cf. e.g. Searle 1975, who needs 10 steps
to arrive at the intended meaning)? Second, if those inferences are dispensed
with, is there any motivated link between, for example, Can you pass the salt?
in its interrogative and in its requestive meaning? I will dispose of the first is-
sue in a somewhat cavalier fashion, by simply pointing out that such inference
processes were never very plausible to begin with, and that there is ample psy-
cholinguistic evidence for their non-existence (e.g. Gibbs 1994; Clark 1979). I
will however, come back to this issue briefly in my conclusion.
Let me turn instead to the second issue in more detail. Obviously, no
account of ISA constructions would want to deny that there is some rela-
tion between the ‘indirect’ and the ‘direct’ meaning of a construction. I be-
lieve that Panther and Thornburg’s theory of speech act metonymies pro-
vides a way of capturing this relation that can fruitfully be adapted to a con-
structional account.
In a series of publications, Panther and Thornburg have developed an ac-
count of the interpretation of indirect speech acts in terms of what they call
speech act scenarios, essentially idealized cognitive models of certain culturally
entrenched activities, that include not only an event itself, but also knowl-
edge about preconditions, results and consequences of this event. For requests,
such a scenario would include the following kinds of knowledge (Panther &
Thornburg 1998: 759):
(11) Simplified scenario for requests
before: H can do A. S wants H to do A.
core: S puts H under a (more or less strong) obligation to do A.
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.11 (115)
conventionalized indirect speech acts. An utterance like I always eat my egg with
salt, for example, can be understood as a request to pass the salt even though
it does not directly evoke any aspect of the request scenario. The reason is that
it evokes a possible motivation for S’s wanting the salt, which will then evoke
the before component of the request scenario. Panther and Thornburg dis-
tinguish utterances that directly evoke a part of the speech act scenario (which
they refer to as metonymic functions) from those that indirectly evoke a part of
it (which they refer to as indexical functions).
Panther and Thornburg’s model provides the missing piece for the analy-
sis of ISA constructions: their motivation. As mentioned in the overview of its
basic tenets, Construction Grammar does not view the linguistic system as an
unstructured collection of constructions, but as a highly structured inventory
of constructions that inherit aspects of form and/or meaning from one an-
other in intricate ways. In the case of the ISA constructions discussed above, I
propose that metonymic links provide the motivation for the partial structural
identity with the direct construction. These metonymic links are themselves
structured by the request scenario as defined by Panther and Thornburg. The
relationship between the question Can you X? and the request Can you X? can
be represented as shown in Figure 1.
Sem - ‹ ›
IFor {}
IFor request
The ‘direct’ construction at the top does not specify anything beyond its
syntax (Syn), i.e. the fact that it needs a subject and a verb phrase, and part
of the semantics (Sem), namely that can means ‘be able.’5 The semantic roles
of the subject and (if present) the object(s) or obliques will be specified by
whichever verb heads the VP, and the illocutionary force (IFor) will be speci-
fied by the discourse context in which the construction is used. If it is unified
with the interrogative construction (or rather, the subject-auxiliary inversion
construction), it will most likely be a question; if it is unified with the declara-
tive construction, it will most likely be a statement. However, it could – under
the right circumstances – have almost any illocutionary force.
The ISA construction at the bottom inherits the formal specifications, but
it does not inherit the semantics of can (I will leave open the question whether
can is actually completely empty, or whether it retains a weak trace of its mean-
ing). The ISA construction also adds the specification that the subject must be
2nd person, and that it must have the semantic role of ‹agent›. Contrary to the
direct construction, the ISA construction inherently specifies the illocutionary
force. This inherent specification of the illocutionary force also accounts for
the formal properties that it shares with direct requests (preverbal please, etc.).
The metonymic link captures the motivation of the ISA construction in
terms of the request scenario: an aspect of the before component (that H can
do A) stands for an aspect of the after component (that H will do A). How-
ever, the fact that the ISA is shown as a separate unit captures the fact that it
has construction status despite the existence of a motivating link, and thus al-
lows us to state its unpredictable formal properties. In other words, the ISA
has construction status in spite of the theoretical possibility to derive its mean-
ing from the request scenario on-line (as must in fact be done in the case of
non-conventionalized ISAs).
Note that the notion of speech act metonymy also allows us to motivate
one of the formal properties of direct requests and conventionalized indi-
rect requests (both of which we can now characterize as request constructions,
i.e. constructions that have the illocutionary force request directly associated
with them). Recall examples (8a, b), which show that request constructions
can have a preposed subordinate clause stating the hearer’s reason for making
the request. This fact can now be restated more insightfully: any construction
that directly evokes the core of the speech act scenario (i.e. any request con-
struction) can occur with a subordinate clause referring to the periphery of
the scenario, while non-conventionalized indirect speech acts which refer to
the a more peripheral component cannot take such a clause. Take examples
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.14 (118)
(13a, b), where the subordinate clause encodes aspects of the before compo-
nent (asterisks represent unacceptability under a request reading):
(13) The before component
a. (H can do A)
Since you’re sitting right next to the window...
...would you mind closing it? ∼ ...can/could you close it? ∼ ...I want
you to close it.
*...it’s cold in here. ∼ *...when will you close it? ∼ *...I need it closed.
b. (S wants H to do A)
Since I can’t get to the window...
...would you mind closing it? ∼ ...can/could you close it? ∼ ...I want
you to close it.
*...it’s cold in here. ∼ *... when will you close it? ∼ *...I need it closed.
The motivation for the facts in (13–15) is presumably that it is strange to evoke
a request scenario weakly via a speech act metonymy in the preposed subordi-
nate clause, and then follow it with an utterance that evokes it just as weakly
or even more weakly. Under these circumstances, the contents of the since-
clause can no longer be construed as a reason for the illocutionary point of the
main clause.
Returning to the issue of how ISAs mean what they mean, note that on
the present account the construction as a whole carries the illocutionary force
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.15 (119)
‘request,’ while at the same time it does not convey the ‘literal’ meaning of
ability (i.e., the semantics of can as an individual item in this construction is
(almost) zero). Thus, the ISA construction loses the meaning of the ‘direct’
construction. This does not have to be the case for every ISA construction.
Consider the case of the Would you mind X? construction. In this case, the
direct construction’s meaning of asking about the hearers potential objections
to the activity denoted by the VP is retained in the ISA construction, which
simply adds the inherent specification of the illocutionary force. This state of
affairs is shown in Figure 2.
This type of analysis captures the fact that many ISAs seem to retain the
meaning of the direct construction in addition to their illocutionary function.
Recall the obstacle hypothesis mentioned above: speakers often select request
constructions that make reference to the most likely obstacle to fulfillment, in
this case the hearer’s willingness.
An alternative to the analysis presented here would be to assume an on-
line motivation of all indirect speech acts in terms of speech act metonymies.
However, such an analysis would leave the unpredictable formal properties of
the type of conventionalized ISA discussed above unaccounted for. Thus, the
linguistic evidence alone is sufficient to support the analysis presented here.
However, additional evidence for such an analysis comes from neurolinguisics.
IFor {}
IMy:
IFor request
. Neurolinguistic evidence
The research on indirect speech acts has focused on individuals with right-
hemisphere damage (RHD), since it has repeatedly been observed that while
for such individuals the core linguistic abilities are typically intact, they have
difficulties interpreting various types of non-literal meaning. The general ex-
pectation would thus be that they would have a tendency to interpret ISAs lit-
erally, i.e. give them a direct interpretation even where the context encourages
an indirect interpretation.
There are five studies in particular that are relevant to the present discus-
sion (Hirst et al. 1984; Foldi 1987; Weylman et al. 1989; Stemmer 1994; Stem-
mer et al. 1994; and Brownell & Stringfellow 1999), though they have slightly
different foci. In the following discussion, I will be concerned with the evidence
they provide for the following questions: (i) Do RHD individuals have difficul-
ties interpreting ISAs (I will touch on the issue of production at several points,
but the literature is sparser here)? (ii) If so, are there differences between con-
ventionalized and non-conventionalized ISAs? (iii) If RHD individuals have
difficulties, what is the nature of these difficulties? I will also briefly discuss
how individuals with left-hemisphere damage (LHD), i.e. aphasics, perform
on some of the same tasks.
Hirst et al. (1984) found that RHD individuals do not in fact tend to give
conventionalized ISAs (of the form Can you X?) a direct interpretation. When
asked to judge the appropriateness of videotaped role-plays, they were equally
likely to judge direct and indirect interpretations as appropriate in a situation
where the context strongly suggested a direct interpretation. For example, for
the utterance Can you play tennis?, uttered in a situation where S and H are sit-
ting in the living room reading, they accepted as appropriate a verbal response
(where H would say “Yes”) in 75% of the cases, and a physical response (where
H would get up and start playing tennis in the living room) in 70% of the cases
(this difference was not significant).
On the other hand, they were vastly more likely to accept an indirect inter-
pretation where the context suggested one. For example, for the utterance Can
you pass the salt? in a situation where S and H are sitting at the dinner table,
they judged a verbal response as appropriate in 5% of the cases, but a physical
response in 97% of the cases.
Clearly, then, RHD individuals have difficulties with conventionalized
ISAs, but not in the expected way of taking them literally. Instead, the results are
opposite to the expectation: they clearly know the pragmatic meaning of ISAs,
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.17 (121)
and they apply it even in a context that by its incongruity makes an indirect
interpretation highly implausible.
Of course, Hirst et al.’s results cannot be generalized to all ISAs since they
used one of the most conventionalized ISA constructions available. The ques-
tion remains how RHD individuals perform with non-conventionalized ISAs.
The answer is provided in a similar study by Foldi (1987), in which sequences
of line drawings were used as stimuli rather than role plays. Presented with a
mixed set of ISAs consisting mostly of non-conventional ISAs (e.g. Will your
tray hold all these dishes? or Do you have everything to wash the car?), RHD pa-
tients showed a significant preference for direct interpretations as opposed to
both a group of aphasic individuals (see below) and the control group (Foldi
1987: 96). From the examples given, it seems that the contexts in Foldi’s study
allow both interpretations without producing the kind of incongruity found
in Hirst et al.’s examples, although they make an indirect interpretation much
more plausible.
Finally, Weylman et al. (1989) showed that in a task with verbal stimuli
(i.e. with verbal rather than visual descriptions of situations) RHD individu-
als interpreted ISAs with different degrees of conventionalization (Can you X?,
Are you able to X?, Is it possible for you to X?) indirectly more often than the
control group in a context that encouraged a direct interpretation (ibid.: 585).
This study thus confirms Hirst et al.’s results. Weylman et al. found that RHD
individuals showed no sensitivity to different degrees of conventionalization in
one task, and very little sensitivity in a second task (ibid.: 588). Superficially,
this may seem to contradict Foldi’s results. However, note that Weylman et al.’s
examples are all fairly conventionalized expressions; at the very least they are
clearly more conventionalized than Foldi’s examples.
With regard to production, Brownell and Stringfellow (1999: 460f.) found
that RHD individuals are able to produce ISAs, but that they have difficulties
varying the directness of their requests, especially with regard to the degree of
imposition on the hearer. Similarly, Stemmer (1994) and Stemmer et al. (1994)
showed that RHD patients have a tendency to overuse non-conventionalized
ISAs (in the form of hints), again, disregarding what would be appropriate in a
given context.
In sum, RHD individuals clearly have difficulties with ISAs. In the case of
expressions typically associated with conventionalized ISAs, their difficulty lies
in determining when the indirect reading is not intended. In the case of non-
conventionalized ISAs, their difficulty lies in discovering the intended meaning.
It seems, then, that RHD individuals have problems generally with taking con-
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.18 (122)
The neurolinguistic evidence, while still sketchy in part, makes clear at least
one thing: ISAs are always produced and interpreted by on-line inferencing.
However, the inferencing process is not one of deriving the non-literal, indirect
meaning of a construction from its literal meaning.
If this were the case, RHD individuals should have a uniform tendency to
interpret ISAs literally. Instead, the process is simply one of interpreting what
a speaker is trying to communicate on the basis of the meaning of the sen-
tence uttered and the context in which it is uttered. The meaning of a sentence
is the combined meaning of all constructions (syntactic, argument-structure,
lexical, etc.) that it instantiates. Since expressions may instantiate alternative
constructions at the same time, interpreting a speaker’s communicative inten-
tion often involves disambiguating homonymous expressions in terms of the
constructions they instantiate. The problem a hearer is faced with when hear-
ing strings like (16a–c) is not to process its ‘literal’ meaning, determine that the
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.19 (123)
utterance does not lend itself to a literal interpretation, and then infer possible
‘non-literal’ meanings:
(16) a. Can you close the window?
b. Would you mind telling me the time?
c. I would like a cheeseburger.
Instead, the problem faced by the hearer is to realize that each of these expres-
sions is ambiguous in terms of the conventional meanings attached to the con-
structions they instantiate, and to use context in order to determine which of
their conventionalized meanings is the one intended by the speaker, in the same
way as a hearer faced with (17) must determine which of the conventionalized
meanings of bank is intended:
(17) Sam went to the bank to take out a loan.
The RHD individuals clearly know that expressions like (16a) have two mean-
ings: recall that they accept both verbal replies (yes or no) or physical activities
(closing the window) as appropriate responses. However, they are not able to
disambiguate between them on the basis of the non-linguistic context (they
do not find it odd to play tennis in the living room). In a similar fashion,
they are not able to produce requests that are appropriate to the context. In-
stead, they often tend to use ISA constructions regardless of the situation, like
Brownell and Stringfellow’s patient JM, who consistently produced conven-
tionalized ISAs, even in contexts where normal controls unanimously preferred
direct requests: “It is as if she generated a single request frame and applied it
consistently” (Brownell & Stringfellow 1999: 460).
Of course, the existence of ISA constructions, i.e. of constructions that in-
herently specify their illocutionary force, does not entail that all constructions
work in this way. The sentences in (18) are not ambiguous between a ‘direct’
and an ‘indirect’ interpretation:
(18) a. It’s cold in here.
b. I wonder if it is already past seven.
c. I love cheeseburgers.
Yet, given the right context, they can have the same interpretations as (16a–c)
respectively. In this case, there does need to be some inferencing, i.e. the hearer
does need to ask herself ‘Why is the speaker telling me this?’ How this infer-
encing is done is essentially an open question. It is clear, however, that RHD
individuals have great problems with it, and do tend to interpret utterances like
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.20 (124)
those in (18) literally, even where the context leaves no doubt that they are to
be understood as requests.
The idea of metonymic reasoning cannot be proved or disproved by the
neurological evidence as it currently stands. It is very useful, however, since it
draws attention to the fact that ‘inferencing’ consists of the activation of related
aspects of culturally entrenched models. It allows us to capture neatly the mo-
tivation behind conventionalized as well as certain non-conventionalized ISAs.
One piece of evidence at least for the special status of non-conventionalized
ISAs that directly invoke a part of the request scenario is that RHD individ-
uals seem to have less trouble with these (cf. Weylman et al. 1989) as op-
posed to non-conventionalized ISAs that do not directly invoke part of the
scenario (cf. Foldi 1987). We might argue in this case (although further evi-
dence is certainly needed), that the existence of conventionalized metonymic
links between independently existing constructions also facilitates inferencing
processes that make use of these metonymic links in interpreting utterances
that do not instantiate ISA constructions.
Notes
* I would like to thank the participants of the panel “Metonymy and Pragmatic Inferencing”
at the IPrA Conference 2000 in Budapest, in particular Klaus-Uwe Panther, Linda Thorn-
burg, and David Zubin, for fruitful comments on an earlier version of this paper, for pos-
ing tough questions, and for pointing out weaknesses in my argumentation. All remaining
weaknesses are, of course, mine alone.
. There is a set of fundamentally similar, closely related theories that share this view of
language and are referred to as ‘Construction Grammar’ (e.g. Lakoff 1987; Fillmore 1988;
Fillmore & Kay 1993; Kay & Fillmore 1999; Goldberg 1995). The discussion in this section
essentially follows the exposition in Goldberg (1995).
. Note that here and elsewhere, the notion of predictability is used in the sense of the
definition quoted at the beginning of Section 2. A property of a construction is referred
to as predictable if it is expected on the basis of the properties of its component parts and
general ‘rules’ (e.g. other constructions or pragmatic principles) of the language in question.
. Note that this is not true of sentence-initial or sentence-final please, which under cer-
tain conditions can occur with questions too, at least if followed/preceded by an intonation
break, as in Please, is it true that you are planning a new movie at the moment. It seems that
this is only possible where the hearer is under an extremely weak obligation to answer the
question; the please seems to function as a request to answer the question, thus the example
just given means ‘Please tell me, is it true....’
. There is a further formal property with respect to which conventionalized indirect re-
quests resemble direct requests but not questions or non-conventionalized indirect requests:
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.21 (125)
their subject is restricted to a particular person (second person in Can you X? and Will you
X?, first person in I want X, etc. This difference, unlike the others, is predictable from the
request function of these expressions.
. I will leave open the question whether the direct speech acts here and below are ac-
tually constructions themselves, or whether their form and meaning can be completely
predicted from the constructions that they consist of (the specific lexical items used, the
subject-predicate construction, the subject-auxiliary inversion construction, etc.).
References
Bach, Kent & Harnish, Robert M. (1979). Linguistic Communication & Speech Acts.
Cambridge, MA, and London: The MIT Press.
Brownell, Hiram & Stringfellow, Andrew (1999). Making requests: Illustrations of how
right-hemisphere brain damage can affect discourse production. Brain and Language,
68, 442–465.
Clark, Herbert H. (1979). Responding to indirect speech acts. Cognitive Psychology, 11, 430–
477.
Fillmore, Charles (1988). The mechanisms of ‘Construction Grammar’. Berkeley Linguistics
Society, 14, 35–55.
Fillmore, Charles & Kay, Paul (1993). Construction Grammar. Ms. University of California,
Berkeley.
Foldi, Nancy S. (1987). Appreciation of pragmatic interpretations of indirect commands:
Comparison of right and left hemisphere brain-damaged patients. Brain and Language,
31, 88–108.
Gibbs, Raymond W., Jr. (1994). The Poetics of Mind: Figurative Thought, Language, and
Understanding. Cambridge: Cambridge University Press.
Goldberg, Adele (1995). Constructions: A Construction Grammar Approach to Argument
Structure [Cognitive Theory of Language and Culture]. Chicago and London: The
University of Chicago Press.
Goldberg, Adele (1996). Construction Grammar. In K. Brown & J. Miller (Eds.), Concise
Encyclopedia of Syntactic Theories (pp. 68–71). Oxford: Pergamon.
Gordon, David & Lakoff, George (1975). Conversational postulates. In P. Cole &
J. L. Morgan (Eds.), Speech Acts [Syntax and Semantics 3] (pp. 83–106). New York:
Academic Press.
Grice, H. Paul (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 41–58). New York: Academic Press.
Hirst, William, Le Doux, Joseph, & Stein, Susanna (1984). Constraints on the processing of
indirect speech acts: Evidence from aphasiology. Brain and Language, 23, 23–33.
Kay, Paul & Fillmore, Charles (1999). Grammatical constructions and linguistic gener-
alizations: The What’s X doing Y? construction. Language, 75, 1–33.
Lakoff, George (1987). Women, Fire, and Dangerous things: What Categories Reveal About the
Mind. Chicago and London: The University of Chicago Press.
TSL[v.20020404] Prn:13/06/2003; 10:41 F: PB11305.tex / p.22 (126)
Lakoff, George & Johnson, Mark (1980). Metaphors We Live By. Chicago and London: The
University of Chicago Press.
Morgan, Jerry L. (1978). Two types of convention in indirect speech acts. In P. Cole (Ed.),
Pragmatics [Syntax and Semantics 9] (pp. 261–280). New York: Academic Press.
Panther, Klaus-Uwe & Thornburg, Linda (1998). A cognitive approach to inferencing in
conversation. Journal of Pragmatics, 30, 755–769.
Sadock, Jerrold M. (1974). Toward a Linguistic Theory of Speech Acts. New York: Academic
Press.
Searle, John R. (1975). Indirect speech acts. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 59–82). New York: Academic Press.
Sperber, Dan & Wilson, Deirdre (1986). Relevance: Communication and Cognition. Oxford
and Cambridge, MA: Blackwell.
Stemmer, Brigitte (1994). A pragmatic approach to neurolinguisics: Requests (re)consid-
ered. Brain and Language, 46, 565–591.
Stemmer, Brigitte, Giroux, Francine, & Joanette, Yves (1994). Production and evaluation of
requests by right hemisphere brain-damaged individuals. Brain and Language, 47, 1–31.
Thornburg, Linda & Panther, Klaus (1997). Speech act metonymies. In W.-A. Liebert,
G. Redeker, & L. Waugh (Eds.), Discourse and Perspective in Cognitive Linguistics
[Current Issues in Linguistic Theory 151] (pp. 205–219). Amsterdam and Philadelphia:
John Benjamins.
Weylman, Sally T., Brownell, Hiram H., Roman, Mary, & Gardner, Howard (1989). Appre-
ciation of indirect requests by left- and right-brain-damaged patients: The effect of
verbal context and conventionality of wording. Brain and Language, 36, 580–591.
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.1 (127)
. Introduction
The syntactic correlates of speech acts are typically independent sentences such
as I will submit this article on time (promise) or Give me more time to finish this
article (request). Many speech acts are however realized as non-sentential con-
stituents such as Happy birthday! (nominal expression) or Sorry! (adjective)
that deviate from the sentential prototype. An especially interesting class of
speech acts is exemplified by the expressions in (1)–(10) below – with data from
English, German and French – that look like dependent clauses introduced by
a syntactically subordinating conjunction. Moreover, in German, such expres-
sions exhibit dependent clause, i.e. verb-final, word order, as seen in (7)–(9).
Interestingly, these apparent dependent clauses can however “stand alone” and
function as independent speech acts.
English
(1) If you will come to order. [request]
(2) Why, if it isn’t Susan! (Quirk et al. 1985: 842) [expression of surprise]
(3) If you would like a cookie. [offer]
(4) That you should say such a thing! [expression of indignation]
(5) That you dare to show your face here! [reproachful indignation]
(6) For you to even think that! [indignation]
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.2 (128)
German
(7) Wenn Sie jetzt bitte zahlen wollen. [request]
‘If you will please pay the bill now.’
(8) Daß (mir) niemand den Saal verläßt! [prohibition]
‘Nobody should leave this room!’
(9) Ob er wohl kommen wird? [question]
‘I wonder whether he will come.’
French
(10) Que personne ne sorte. (Grevisse 1993: 624) [prohibition]
‘Nobody should leave.’
In prior research (Thornburg & Panther 1997; Panther & Thornburg 1998) we
have defined speech acts and their felicity conditions in terms of scenarios –
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.4 (130)
Background Motivation
: H is willing to do A.
H will do A.
Realization
. Analysis of data3
For each of the examples in the data set, we analyzed the pragmatic force with
respect to (i) the content of the proposition in the if-clause in terms of its men-
tal space structure, (ii) potential metonymic links to various speech act scenar-
ios, and (iii) the degree of conventionalization, i.e. whether or not the prag-
matic force of the if-clause is cancelable. To illustrate, consider example (1).
If you will come to order is conventionally understood as a request to come to
order.4 To account for that fact, we begin with an analysis of the hypothetical
space triggered by if along the lines we have proposed in (11). First, we note
that the proposition within the clause, given in (1a):
Hypothetical space for If you will come to order
(1) a. Proposition p: you will come to order
explicitly refers to the addressees’ action in terms of future time frame and will-
ingness to undertake the action.5 Secondly, we assume that there is a strong
metonymic link between the hearers’ willingness to perform the action and the
ability to do so. Thirdly, because the hypothetical if space allows for potential
consequent propositions in some context, in the context of a noisy classroom,
say, in which a teacher utters If you will come to order, it is possible to infer the
consequent given in (1b):
(1) b. Inferable q: ... then I will begin the lecture.
Background Motivation
:
: H can do A. S wants H to do A.
:
: H is willing to do A.
H will do A.
Realization
Having accounted for the request interpretation of If you will come to order,
we further note that this reading is difficult – if not impossible – to cancel, as
tested in (1c), which yields a pragmatic contradiction:
(1) c. Cancelability: #If you will come to order ... but I’m not asking
you to do that.
In this section of the chapter we present additional if-clause data that have pre-
dominantly deontic pragmatic force. That is to say, what the if-clause impli-
cates requires that the world should change in such a way so as to match what
is metonymically evoked. As Searle (1983: 7 et passim), among others, has put
it, the “direction of fit” in these cases is from the world to words, the so-called
“so be it” use of language. In using if-clauses to issue directives or commissives
or to express wishes, a speaker uses language to talk about the way the world
will or should change to fit some propositional content.
.. Directives
We present here additional examples of if-clauses with directive illocutionary
force:
(12) “This is awful,” Julia exclaimed in consternation. “Do please tell her to
stop crying, Don Felipe. I can’t bear it. If you could explain it isn’t that I
really want to go home. I just have to.” [LOB.P1]
For the if-clause in (12) our intuition is that the speaker is making a request of
the addressee.6 At issue is how we derive the pragmatic request force from what
looks like a truncated conditional sentence. Using the methodology outlined
and demonstrated above, we first note that the mental space triggered by if
contains the hypothetical proposition:
(12) a. Proposition p: you could explain it isn’t that I really want to go
home
to the ability of the hearer to perform the action and the other two implicitly
evoke the speaker’s anticipatory gratitude for and desirability of the action.
As subcomponents of the Request Scenario, they activate the remainder of the
scenario, as represented in the shaded box:
Background Motivation
: : :
H can do A. S wants H to do A.
:
: S is grateful for H doing A.
Realization
In example (13) we find a different type of directive. Here the speaker appears
to suggest to the addressee that the two of them go up to the addressee’s room.
(13) “I have made a discovery, sir. It may be of no account, but I think that
you will find it – interesting. If we could go up to your room, sir....” Nick
wondered if he was about to be touched by a blackmailer, but the young
man sounded genuine enough. [LOB.P1]
H will do A.
:
: H benefits from A.
Realization
.. Offers
Our LOB data yielded no examples of if-clauses with the force of an offer
comparable to our constructed example in (3). Yet this example strikes us as
plausible and we include it in our analysis, reproduced here as (14):
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.10 (136)
In (14a) the proposition I can give you one is interpretable as being a subcom-
ponent of the Background branch in the Offer Scenario (at the same time a
before subcomponent), whereas the proposition I will give you one in (14b) is
interpretable as being a subcomponent of the Realization branch in the Offer
Scenario (simultaneously an after component). Also in the conceptual space
of the if-clause is the possibility of elaborating the proposition you would like a
cookie into the proposition p’:
(14) c. Proposition p’: you would like me (speaker) to give you a cookie
Background Motivation
: X is desirable to H.
: S is willing to do A.
:
: S will do A.
Realization
.. Wishes
We turn now to examples of if-clauses that have the force of a wish expression.
(15) ‘That girl’s nothing but a load of trouble, I’m warning you.’ ‘Kitty’s all
right,’ Bone contradicted flatly. ‘It’s her boy-friend that’s the trouble. If
we could get rid of him...’ Harry nodded his grizzled head like an old
hound. [LOB.L1]
(16) “But meanwhile, I must find her. If only I had a clue where to look for
her.” “Has it occurred to you that when you told her about us it was such
a shock to her that she has run away.” [LOB.P1]
We think both these examples have the force of a wish expression.7 However,
the wish interpretation is weaker in (15) and cancelable; in contrast, if only in
(16) makes the wish interpretation conventional and thereby uncancelable.
In (15) the hypothetical proposition:
(15) a. Proposition p: we could get rid of him
explicitly denotes the possibility for the interlocutors to carry out an action in
future time, an action that is doubly hypothetical by virtue of both if and could.
This gives rise to the inference that the action has not yet occurred.
Given the context in (15), an inferable consequence might be:
(15) b. Inferable q: ...then our troubles would be over.
as desirable. In the hypothetical space of (15), then, we can identify four com-
ponents that are inputs to the Wish Expression Scenario depicted in Figure 6:
(i) the explicit reference to the possibility to undertake an action to get rid of
him; (ii) the inference that the action has not occurred; (iii) implicit reference
to the desirability of the action; and (iv) implicit reference to resulting feelings
of satisfaction.
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.12 (138)
Background Motivation
:
: :
p does not exist. p is desirable to S.
:
p is realizable.
:
S is satisfied from the realization of p.
Realization
The cancelability of (15) contrasts with examples like (16) containing if only.
The proposition in this example presented in (16a):
(16) a. Proposition p: I had a clue where to look for her
is counter-factual at the time of speaking. The focus particle only singles out
one proposition to the exclusion of others – it highlights the importance or
relevance of that proposition for the speaker; only, then, triggers an implicature
of emotional involvement and high desirability with respect to the proposition.
The wish interpretation is not cancelable in this case, as seen in (16b):
(16) b. Cancelability: #If only I had a clue where to look for her, but I don’t
wish I had a clue where to look for her.
We also note that in (16) the proposition is non-factual but also potentially ful-
fillable. This contrasts with examples (17) and (18) in which the propositions
refer to non-occurrent past events that have no possibility of future realization.
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.13 (139)
(17) Tom, she thought. If only I could have asked Tom’s advice. But now it’s
too late for that. [LOB.L1]
(18) Watching him go, unable to speak, she felt that part of her was leav-
ing with him. She couldn’t hate him... If only he would have confided
in her, given some explanation. Now there was nothing – not even
friendship. [LOB.P1]
The Wish Scenario for (17) and (18) – in contrast to that given in Figure 6
– would contain a background assumption: Some state-of-affairs is impossi-
ble. Furthermore, the unfulfillability of such wishes implicates the absence of
emotional satisfaction. The emotions that are likely to be associated with un-
fulfillable wishes are strongly negative, such as regret, bitterness, anger and so
on – which likely characterize examples (17) and (18).
We now briefly summarize Section 3.1. We discussed if-clauses whose pre-
dominant function is deontic – serving to metonymically convey requests, sug-
gestions, offers, and wishes. We saw that the conceptual distance created by the
space builder if is exploited in the cases of other-directed speech acts like direc-
tives and commissives to minimize negative face-threat. We also saw that the
metonymically evoked pragmatic forces of these clauses tend to be uncance-
lable suggesting that they are highly conventionalized; i.e., the scenarios asso-
ciated with them are automatically activated. In such cases the if -clause can be
said to have achieved the status of a construction.
As we saw with unfulfillable wishes like (17) and (18), it is difficult to know if
the speech act was predominantly deontic – the expression of a wish – or pri-
marily an expression of emotion. In what follows we will focus on if-clauses
whose primary function seems to be the expression of a strong emotional
attitude with regard to some state of affairs.
.. Negative p
We begin with example (2) from Quirk et al. (1985), reproduced here as (19a).
We also provide some examples from the OED given in (19b–f). All of them
contain exclamation marks and/or other devices indicating that the if-clauses
are uttered with attendant emotions.
(19) a. Why if it isn’t Susan! (Quirk et al. 1985: 842)
b. If he is not equipped for a housebreaker! [1702 Vanbrugh False
Friend iii. ii]
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.14 (140)
c. And, so help me never! if his nibs didn’t go and dossed with her the
same night. [1846 Swell’s Night Guide 49]
d. ‘If it ain’t Frisco Red!’ exclaimed one prone figure. [1914 Sat. Even.
Post 4 Apr. 10/1]
e. ‘Oh, Gee, well, ain’t that the limit?’.. ‘If you aren’t the grouch.’ [1925
T. Dreiser Amer. Trag. I. xvii. 145]
f. Well, by jing, if it ain’t Tom. [Ibid. II. iii. 184]
We first note that this type of if-clause has the highest degree of syntactic and
pragmatic independence. That is, they don’t have plausible implicit consequent
propositions, except perhaps for absurdities like that in (19f ’):
(19) f ’. Inferable q: ?Well, by jing, if it ain’t Tom, then I’ll eat my hat!
Secondly, we note that all the propositions in (19a–f) assert at the moment of
speaking a non-factual state of affairs, for example it ain’t Tom, when in fact it
is precisely Tom. In other words, what the speaker does not do is simply assert
what is empirically true in reality space, which for (19f ’) might be:
(19) f ”. Well, by jing, it is Tom!
Emotional Space
By using an if-clause in this case the speaker can achieve multiple effects: avoid
directness, metonymically implicate the contradictory in reality space, namely,
you are the grouch (implicating “it is so”), convey a normative evaluation – one
shouldn’t be a grouch – and express a negative emotion such as dissatisfaction.
.. Positive p
Unlike the examples discussed thus far in Section 3.2, example (20) does not
contain a negative proposition in the if-clause:
(20) ‘You must think I like the military sticking its nose in.’ I said bitterly: ‘We
spend our lives running things the quiet way. Then the army arrives – a
blow, a false word – bang – suddenly there are shots. All right. If that’s the
way they want it. But don’t ask me to clean up the mess.’ [LOB.K1]
Here the speaker uses the hypothetical space of an if-clause to convey the
positive proposition:
(20) a. Proposition p: that’s the way they want it
which expresses a grudging concession to the way they want it and provides
additional grounds that the speaker evaluates the proposition as undesirable,
out of his control and most likely irreversible, and – like unfulfillable wishes –
gives rise to negative feelings like bitterness and regret. This dissatisfaction with
the state of affairs does not seem to be cancelable, as shown in (20c):
(20) c. Cancelability: #If that’s the way they want it, fine, it’s also my pro-
foundest wish.
The examples in (19) and (20) have been presented with the claim that their
primary function is the expression of an emotional attitude with regard to a
state of affairs – i.e. some “it is so” description. In the cases in (19), the if-
clause provides a mental space for conveying an expected state of affairs that is
contradicted in reality space. Because of the discrepancy between what is hypo-
thetically denied but empirically true, the construction is a potential vehicle for
the expression of attitudes like surprise or amazement, a use which is conven-
tional. In contrast, in cases like (20), the construction is used to convention-
ally signal disapproval of what is conceded in hypothetical space, an emotional
attitude that results from the speaker’s opposition to what he/she concedes.
(22) Judging from the spot where it lay it had been planted between the under-
side of the mattress and one of the cross-supports. If I hadn’t re-made the
bed... if Sonia and I hadn’t made love... Sonia. Nothing else accounted for
the presence of that hellish box. I’d left her alone in the bedroom when we
awoke from the brief sleep of exhaustion. [LOB.L1]
We note first that the if-clauses in (21) and (22) – unlike those in (19) – have
a low degree of pragmatic independence; rather, they give the impression of
being highly elliptical if-clauses. Secondly, they are classical cases of counter-
factuals. In uttering a counterfactual premise in hypothetical space – as repre-
sented in Figure 8 – the speaker pragmatically activates the shared background
knowledge in reality space that the proposition is false and at the same time
invites the hearer to consider the counterfactual proposition as a premise from
which to reason to unstated consequences.
Invited counterfactual
conclusions: q
Example (23) is similar to (21) and (22) in inviting the reasoner to com-
plete the conditional by drawing conclusions from the premise it expresses.
(23) Farland summed up. Quite fair to hold out on Winter. It seems he’s keep-
ing things back. If he knows about the knife... And if he knows that Wally
did attack the girl... There were voices in the hall and Winter entered with
the visitor.
In this example, however, the premise is not counterfactual but merely not
known to be true. Nevertheless, despite the lack of certainty about the truth
of p, the reasoner seems to believe that it is rational to assume that p. What
we see in (23) is a kind of hedged assertion or reasonable supposition that p.
Note, however, that the (weak) assertive force can be canceled very easily, as in
(23a) and (b):
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.18 (144)
(23) a. If he knows about the knife... But I don’t believe he really knows about
it...
b. And if he knows that Wally did attack the girl... But I don’t think he
knows that Wally attacked the girl...
As for the examples in (19), for (24) there do not seem to be any plausible
consequent propositions that follow from the if-clause. Our proposal for the
conceptual structure of (24) is represented in Figure 9. That is, given a context
in reality space in which some proposition like ‘it was a warning’ is generally
assumed to be true, the speaker, in using an if-clause, conveys in hypotheti-
cal space that he/she does not know whether the proposition is true, thereby
strongly implicating a challenge to the assumed truth of p in reality space.10
Indeed, the implicature raised by if p seems difficult, if not impossible, to
cancel:
(24) a. Cancelability: “I’ve told you I have no idea who this warning could
have been for. If it was a warning. #But I think it was a warning.”
We briefly summarize the analysis of data in this last section regarding the use
of if-clauses in relation to reasoning. We saw that in posing within an indepen-
dent if-clause a premise whose truth value is not known – as in (23), a speaker
can implicate in hypothetical space a weak assertion, which is cancelable, as
S challenges p
Notes
Another characteristic of this type of if-clause is that it is used for “afterthoughts”; i.e., it
cannot be used to initiate a conversation but rather expresses a reaction to some assumed
state of affairs.
References
Brown, Penelope & Levinson, Stephen C. (1987). Politeness: Some Universals in Language Use
[Studies in Interactional Sociolinguistics 4]. Cambridge: Cambridge University Press.
TSL[v.20020404] Prn:13/06/2003; 10:44 F: PB11306.tex / p.21 (147)
At the end of his tale about the three pledges of love, Karl Waggerl finishes with
the rather peculiar sentence in (1):1
(1) . . . als er in die Stube kam, da lag sein Mädchen auf der Bahre. Da wußte
er, daß sie es war, die er dreimal geliebt und dreimal verraten hatte, und
nun steckte sein Messer mitten in ihrer weißen Brust.
[. . . ] when he entered the main room, there lay his girl on the bier. And
then he knew that she was the one 2 whom he had loved three times, and
three times betrayed, and now there was his knife, thrust in the middle of
her white breast.
As the underlining indicates, Waggerl switches back and forth between fem-
and neut-gender in apparent anaphoric reference to the noun Mädchen.3 Our
goal in this paper will be to provide a general context in which this apparent
anomaly is explained. In the process we will first briefly review the state of
research on the semantics of nominal classification in German, and then pro-
vide evidence for both the historical and current productivity of neut-gender
classification for human beings, and females in particular, before returning to
the Waggerl text. As the discussion moves through these three steps the func-
tional role that metonymy plays will become apparent, first as part of the di-
achronic lexical processes that result in neut-gender human nouns, and then in
the pragmatics of referential tracking. In developing the argument we will fol-
low Lakoff ’s analysis and commentary on metonymic ICMs (Lakoff & Johnson
1980: Ch. 8; Lakoff 1987: Ch. 5). Four of Lakoff ’s (1987: 84–85) points about
metonymy are particularly pertinent to our analysis:
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.2 (150)
To these four points we add (e), which follows from point (c) and examples in
Lakoff and Johnson (1980: Ch. 8):
e. Some property evoked by B is relevant to the speaker’s current discourse
intent, and often is carried into the resulting discourse representation.6
á
<>
direct disourse-world reference
â á’
<> < >
(sexual) innocence (sexually) experienced
social naiveté social maturity
dependence independent
village life city life
¨ ¨
¨ ¨
Mädchen Frau
neut-gender fem-gender
There has been a great deal of research in this area, reviewed in such places
as Wienold (1967), Zubin and Köpcke (1984, 1986), Claudi (1985), Corbett
(1991), and Köpcke and Zubin (1996), but such research has tended to be anec-
dotal rather than systematic, and limited to a biased sample of high-frequency
nouns. In our own previous work (Köpcke 1982; Köpcke & Zubin 1983, 1984,
1996; Zubin & Köpcke 1981, 1984, 1986, in preparation) we have been able to
demonstrate the existence of underlying patterns of semantic motivation for
gender assignment.
We have argued for the following general principles underlying gender as-
signment in folk-taxonomic domains in the German lexicon (cf. Zubin &
Köpcke 1986):
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.4 (152)
But just how inclusive are such gender-motivating principles in the lexicon?
In order to test the generality of semantic principles, and of corresponding
phonological principles, we have constructed an extensive cross-sectional ran-
dom sample of the nominal lexicon in German (Zubin & Köpcke, in prepara-
tion), based on the Duden Universalwörterbuch (1983). In the analysis of this
sample we created a measure of cognitive entropy on a scale ranging from 0 (no
semantic association at all) to 5 (the nominal receives its gender based on a
fully productive semantic principle); intermediate steps correspond to increas-
ing degrees of generality of the basis for gender assignment.10 This was coupled
with a corresponding measure of formal entropy, again ranging from 0–5. The
combined mean entropy for nominals is about 5.0 indicating that on the av-
erage, gender assignment is fully motivated through a combination of seman-
tic, phonological and morphological factors. Some nouns have an entropy as
high as 10, indicating the cooperation of fully productive semantic and formal
factors. And very few nouns in the sample have negative entropy (which indi-
cates an anomalous gender assignment in conflict with overall semantic and
phonological factors).
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.5 (153)
Several of our studies include diachronic data showing that there is a general
drift toward greater semantic motivation in the evolution of gender assignment
in the lexicon. For example, the neut-gender human nouns – the topic of this
paper – have their beginnings as a lexical field in the lexicon with nouns such
as those in Table 1.
Table 1. The neut-gender nominal cluster for humans in the period between Middle
High German and Early New High German
As the middle column suggests, such neut-gender nouns have a wide va-
riety of etymological sources, with only a minority of them originating with
female human reference. For example, das Luder referred to meat used as bait
for trapping animals. One can only speculate about the origins of the cluster.
But the following nouns form a plausible basis:
a. das Weib has had neut-gender since its Germanic origins and was the
basic-level term for ‘woman’ until the 16th century, when it was gradu-
ally replaced by die Frau and began to take on its pejorative meaning. The
parallel semantic shift of other nouns such as die Dirne (‘girl’ > ‘prosti-
tute’) points to a general culture-historical basis for the shift (cf. Grimm &
Grimm 1984).
b. der Mensch developed a neut-gender alternate in the 16th century. Initially
this alternate made sex-neutral generic reference to humans, evident in ci-
tations from Luther, but then became increasingly restricted in reference
to women, first as a completely neutral referring expression (parallel to das
Weib), but then developing pejorative and objectifying affect in the 17/18th
century (cf. Grimm & Grimm 1984).
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.6 (154)
c. das Mädchen gradually replaced die Magd as the basic-level term for ‘girl’
in the 17/18th century.11 There was no corresponding shift of der Junge or
der Knabe ‘boy’ to das Jüngchen or das Knäblein as basic-level terms.
In the transitional period leading up to Early New High German the small
group of nouns in Table 1, perhaps led by das Mensch and das Weib, devel-
oped form-meaning correspondences that brought them into a lexical cluster
through a variety of processes detailed in Table 3. This small cluster was charac-
terized by semantic/pragmatic downgrading on the one hand, and neut-gender
on the other. The 19th and 20th centuries have seen a dramatic increase in the
size of the cluster, leaving no doubt about its productive potential. In our sam-
ple of about 100 neut-gender human nominals from this period, many have
entered the German lexicon in the last 50 years, some of these quite recently. A
few recent additions are given in Table 2.
Nowhere is semantic motivation in the lexicon stronger than in the do-
main of nominals referring to human beings. Human sex is the basis for the
typological distinction between gender systems, as in Indo-European or Aus-
tralian languages such as Dyirbal (Dixon 1968, 1982) and other non-gender
noun class systems such as those in Swahili/Bantu (Contini-Morava 2001) or
Navajo/Athapaskan (Young & Morgan 1987) that do not function to distin-
guish sex. Indeed human sex is the basis for the linguistic term ‘gender’ itself.
In German masc- and fem-gender are highly productive for nominals refer-
ring to human males and females, respectively.12 Yet in the German lexicon
there are long-noted exceptions to sex-based assignment, some already noted
above in Tables 1 and 2. A systematic examination of such human-reference
nouns shows that neut-gender nominals are not randomly distributed and have
formed a cluster in the lexicon with some modest productivity, as illustrated
in Table 2. The nominals in this cluster have the diachronic sources depicted
in Table 3.
At this point we turn to the central role that metonymy plays as a source of
neut-gender for human-reference nouns discussed in Section 3.1. Many are
based on metonymically structured perspectival ICMs, as defined in Figure 2.
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.8 (156)
experiential relation
referent
experiencer
Whatever their metonymic source, the examples given in (a)-(i) above all de-
pend on the perspectival metonymic ICM structure depicted in Figures 1 and
2. For example, das Ekel ‘disgusting person’ evokes a perspectival ICM in which
the experiencer feels disgust in the presence of the object of experience. Flinten-
weib ‘gun-woman’ evokes a cultural stereotype of the hardy independent fron-
tier woman when used as a referring expression. In general the examples dis-
cussed in this section are characteristic of the approximately 100 nouns in our
database of neut-gender downgrading human nominals.
We now turn to the final issue: given that there is a nominal cluster in the
lexicon that continues to attract human neut-gender nominals, and given that
the processes underlying the growth of this cluster are supported by metonymic
ICMs, we may pose the question, to what extent do these lexical structures play
a role in discourse processes, in particular in lexical choice and in pronominal
anaphora (see also Ruiz de Mendoza & Pérez Hernández, this volume)?15
We will illustrate the impact that these lexical structures have on discourse
processes in the short story Legende der drei Pfänder der Liebe (‘The Legend of
the Three Pledges of Love’) by Karl Waggerl. The story tells of a young village
pot-maker who leaves his girl and goes off to sell his wares in the surrounding
towns. She gives him three pledges of love: a hair ribbon, a ring, and a knife.
On his travels he camps alone, as promised, and each night he is visited by a
mysterious woman who sleeps with him. Each morning he gives this woman
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.11 (159)
one of the love pledges from his girl at home, and on the third night the knife.
On the way home he finds the ribbon and the ring, and begins to wonder.
Upon entering his house he finds his girl there with the knife in her breast, and
realizes that his girl and the mysterious woman are one and the same.
A few pertinent highlights will help to make clear the role that gender plays
in structuring the story. First, the story evokes two contrasting archetypes (cf.
Figure 1): the sexually innocent naive female bound to village life and moral-
ity (β ICM), lexicalized with the neut-gender noun Mädchen, and the sexually
experienced, citified, independent, amoral female (α ICM), lexicalized with
the fem-gender noun Frau. Second, in section A of the story (see Appendix) in
which the girl is the object of perspective, anaphoric pronouns are consistently
neut-gender. In section B in which the woman is perspectivized, anaphoric
pronouns are consistently fem-gender.
Finally comes the somewhat unusual syntax of section C, with which the is-
sues of this paper were introduced: “Da wußte er, daß sie es war, die er dreimal
geliebt und dreimal verraten hatte.” On a syntactic level the es is an expletive
pronoun in a cleft construction; i.e. a purely syntactic unit presumably inca-
pable of referring. But on a pragmatic level, such expletive pronouns can be
at least quasi-referential.16 In the sentence at hand, es can be secondarily taken
to evoke the neut-archetype (β in Figure 1), which then metonymically refers
to the innocent village girl character.17 The sie, on the other hand, evokes the
marked fem-archetype (α in Figure 1), which in turn metonymically refers to
the mysterious woman character.18 A further complication is presented by the
following relative clause – die er dreimal geliebt und dreimal verraten hatte –
containing two verb phrases, coordinately conjoined to the relative clause head,
the fem-gender pronoun die. Up to this point in the story, the referent of
“dreimal geliebt” (‘loved three times’) is, from the young man’s perspective,
the mysterious woman; and the referent of “dreimal verraten” (‘betrayed three
times’) is his village girlfriend. In other words, from his psychological perspec-
tive, these are two separate individuals. So who is the referent of the relative
pronoun die? It cannot be either the girl or the woman, metonymically ref-
erenced through the two cultural archetypes, since each of the two predicates
applies to only one of them. It can only be a direct, non-metonymic reference
(α in Figure 1) to that female person in the story world who plays both the role
of the girl and of the woman. The stylistic peculiarity of this sentence thus re-
sides in the fact that with extreme syntactic compactness Waggerl evokes both
culturally opposed archetypes for females, and then refers to the “real” person
standing behind the archetypes.
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.12 (160)
. Conclusion
The Waggerl story of the three pledges provides an extensive, systematic, and
sophisticated exploitation of the group of ICMs for females presented in Figure
1, leaving open the question whether such exploitation is limited to specific
literary genres such as the archaic folk tale. In conclusion we turn to some
short passages from current journalistic cultural critique to show that the use
of neut-gender terms for women, and the metonymic ICM they project onto,
has current and wide cultural validity. We take up each of the four perspectival
choices set out in Table 4.
The neut-gender term Weib is used by the feminist writer to satirize the use
of this noun by conservatives to express their distaste for feminist actions and
issues. Thus the passage illustrates both the use of this term by one segment of
society to express scorn and rejection toward another, but also the use of the
term by the target of scorn themselves to make fun of their tormentors.
The overall context makes clear that the model is presented as a visually at-
tractive object. Even the expression “healthy self awareness” refers to her con-
fident appearance, not to other, non-visual aspects of her person. Thus the
context strongly supports the use of the neut-gender noun das Model and the
metonymic model it projects onto.
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.13 (161)
The neut-gender noun Inkarnat literally denotes the flesh-toned colors used to
depict the model’s body, and thus metonymically refers to the model herself.
The choice of noun and its context decompose her into the play of paint color
on the canvas. The use of the noun Inkarnat in this context thus evokes the
sense of extreme analytic distance from the described person as a human being.
The writer’s point is that these archetypes are culturally current: listening to
and reading stories with these characters during childhood has negative effects
on the psyches of modern women.
The perspectives expressed in the metonymic ICM taken together form a
linguistically marked category, marked by the neut-gender of their associated
nominals. The increasing productivity of the ICM – evidenced by the consis-
tent assignment of neut-gender to the influx of primarily English loans over
the last 50 years – points to what appears to be a cultural distinction of current
relevance in German society in which men are monovalent (they have one sta-
tus: male) while women are bivalent: they are female, and they have specially
marked perspectival values.19
Notes
. For a synopsis of the story, see the last section of the paper and the Appendix. Feminine
gender referring expressions are marked with single underlining, neuters with double. The
abbreviations masc-, fem-, and neut-gender are used to refer to the morphological and lexical
gender properties of nouns, pronouns, and referring NPs in order to distinguish them from
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.14 (162)
both sex and the so-called “natural gender” of referents. The reader is asked to note the
alternation between feminine and neuter gender in this passage.
. lit: “. . . that she was it whom he had loved. . . ”
. In a standard syntactic analysis the clause “daß sie es war” is analyzed as an expletive con-
struction, with the implication that the pronoun es is non-referential. In Section 4.2 we will
argue that this pronoun is simultaneously licensed by the expletive construction and poten-
tially referential, following functionalist analyses of such phenomena as impersonal and re-
flexive constructions. In functionalist approaches to syntax such as Construction Grammar,
syntactic dependency and semantic/pragmatic value do not preclude each other.
. Square brackets enclose expansion of Lakoff ’s points.
. Lakoff ’s (1987: 84) formulation: “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” [italics ours]. We make use here of the final alternative in the disjunction.
. Examples from Lakoff and Johnson (1980: 36–37) will illustrate:
a. We need a couple of strong bodies (B) for our team (A=strong people)
b. There are a lot of good heads (B) in the university (A=intelligent people)
c. She’s just a pretty face (B) (A=good-looking woman)
Note that not only is the content of each of the metonymic B’s appropriate to the context
of utterance, and cannot be substituted in the others (e.g. ??? We need a couple of pretty
faces for our team) but that this specific content surfaces in the interpretation of each of the
utterances, e.g. strong bodies help the team succeed.
. Unmarked referential forms (α) have fem-gender, and refer directly to discourse-world
individuals. Marked (neuter) forms (β) evoke an intermediate scenario. Context may induce
an intermediate scenario for fem-gender forms, labeled (α ).
. While many basic-level nouns exhibit field-dependent gender assignment, others seem
to have idiosyncratic gender. Basic-level terms for drinks exhibit all three genders. Masc-
gender: Wein ‘wine’, Schnaps ‘liquor’, Saft ‘juice’; fem-gender: Milch ‘milk’, Limonade ‘soda’;
neut-gender Wasser ‘water’, Bier ‘beer’. But note that each is a hub for the productive as-
signment of gender to subordinate nominals, as pointed out below. The myth of extensively
arbitrary gender assignment in German and other noun-class languages stems largely from
almost exclusive attention being paid to a small set of high-frequency basic-level nouns.
. An exacting analysis of gender in oral and written discourse reveals that soda types in
fact vary between fem- and neut-gender. This variation is recent, and seems to stem from the
competition of Brause and Limonade on the one hand and (Mineral-)wasser on the other for
the dominating node in the lexical network.
. In other words, the higher the numerical value on the cognitive entropy scale, the
more semantically and/or formally motivated is gender assignment; conversely, the lower
the numerical value, the higher the degree of arbitrariness of gender assignment.
. It has often been argued that Mädchen and some of the other neut-gender human-
referring nominals have neut-gender only because of their derivational suffix -chen, which
is categorically associated with neut-gender, and do not have any semantic basis, other than
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.15 (163)
the diminutive meaning of the suffix itself. Proponents of such an hypothesis would have
to explain (i) that such chen-suffixed nouns have become conventionalized for reference to
females, and not to males; (ii) that Mädchen in particular has become the basic-level term
for ‘girl’, while the corresponding Jünglein remains a semantic diminutive for Junge ‘boy’;
and that (iii) an increasing number of non-diminutive terms such as Playmate, Groupie,
and Video-chick are receiving neut-gender assignment as they enter this lexical field. The
form-obsessed analyst must face the possibility that the structuring of this field is driven by
pragmatic forces and that word morphology is just a pawn in the process.
. And masc-gender is highly productive for generic-reference nominals, one of the main
sources of difficulty in attempts to make the language more sexually egalitarian (cf. Buß-
mann 1995).
. This in contrast to Langacker’s “optimal viewing arrangement,” in which only the refer-
ent is in the scope of predication.
. Note that the nominal expressing the emotion of revulsion itself is masc-gender: der Ekel.
. Pronominal anaphora is of course important for the current study when it agrees in
gender with its neut-gender antecedent.
. In the so-called “weather” construction the es in es regnet, schneit (‘it is raining, snow-
ing’) does not seem to evoke any referent. But in the context of other verbs the es is associated
with a vague, indeterminate referent: es hat eben geklopft (‘someone just knocked’); and in
the context of an extraposed complement, the es is understood as cataphoric: es gefiel ihm,
daß sie da war (‘it pleased him that she was there’).
. That Waggerl would use a double-entendre at this point is characteristic of his mastery
of the short story.
. A further problem with this pronoun sie is the lack of gender agreement with its an-
tecedent sein Mädchen (see Appendix). This is problematic for a derivation-based syntactic
theory, which in characterizing agreement maps fully specified features of a head (noun)
onto a target (pronoun), but not for a constraint-based theory such as HPSG, in which
referential indices may be under pragmatic control. In commenting on:
(a) That dog is so stupid, every time I see it I want to kick it. He’s a damned good hunter,
though.
Pollard and Sag (1994: 73–74) note that “as illustrated in [a], a pronoun that refers to an
entity already referred to by some earlier expression may have a new index with agreement
features different from those of the earlier expression, in order to serve some specific dis-
course purpose (in the present case, to signal a change of attitude toward the referent). In such
cases, according to our definition, the earlier expression does not qualify as an antecedent;
to put it another way, instead of being ‘referentially dependent’ on the earlier expression,
such pronoun uses must be regarded as deictic” [italics ours].
. A new archetype we have become aware of from current journalism, especially teen lit-
erature, is the sassy independent, unconstrained, sexually active, aggressive young woman
who is free of the traditional constraints of womanhood, embodied e.g. in cartoon charac-
ters such as “das Tankgirl.” The two nouns most closely associated with this archetype are
das Mädchen and the borrowed noun das Girl, both neut-gender. It looks like youth culture
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.16 (164)
has “stolen” these nouns from the old archetype of dependence and sexual innocence, and
appropriated them for their own archetype embodying strikingly contrasting values.
References
Bußmann, Hadumod (1995). Das Genus, die Grammatik und – der Mensch: Geschlech-
terdifferenz in der Sprachwissenschaft. In H. Bußmann & R. Hof (Eds.), Genus – zur
Geschlechterdifferenz in den Kulturwissenschaften, (pp. 115–160). Stuttgart: Kröner.
Claudi, Ulrike (1985). Zur Entstehung von Genussystemen: Überlegungen zu einigen
theoretischen Aspekten, verbunden mit einer Fallstudie des Zande. Hamburg: Buske.
Contini-Morava, Ellen (2001). Noun classification in Swahili. University of Virginia, web-
posted Ms. URL: jefferson.village.virginia.edu/swahili/swahili.html.
Corbett, Greville (1991). Gender. Cambridge: Cambridge University Press.
Dixon, R. M. W. (1968). Noun classes. Lingua, 21, 104–125.
Dixon, R. M. W. (1982). Noun classifiers and noun classes. In R. M. W. Dixon (Ed.),
Where Have All the Adjectives Gone? and Other Essays on Syntax and Semantics [Janua
linguarum, Series maior 107] (pp. 211–233). The Hague: Mouton.
Duden (1983). Deutsches Universalwörterbuch. Mannheim: Bibliographisches Institut.
Grimm, Jacob & Grimm, Wilhelm (1984 [1885]) Deutsches Wörterbuch. München:
Deutscher Taschenbuchverlag.
Köpcke, Klaus-Michael (1982). Untersuchungen zum Genussystem der deutschen
Gegenwartssprache [Linguistische Arbeiten 122]. Tübingen: Niemeyer.
Köpcke, Klaus-Michael & Zubin, David A. (1983). Die kognitive Organisation der
Genuszuweisung zu den einsilbigen Nomen der deutschen Gegenwartssprache.
Zeitschrift für Germanistische Linguistik, 11(2), 166–182.
Köpcke, Klaus-Michael & Zubin, David A. (1984). Sechs Prinzipien für die Genuszuweisung
im Deutschen: Ein Beitrag zur natürlichen Klassifikation. Linguistische Berichte, 93, 26–
50.
Köpcke, Klaus-Michael & Zubin, David A. (1996). Prinzipien für die Genuszuweisung im
Deutschen. In E. Lang & G. Zifonun (Eds.), Deutsch Typologisch: Jahrbuch des Instituts
für Deutsche Sprache (pp. 473–491). Berlin: Mouton de Gruyter.
Lakoff, George (1987). Women, Fire and Dangerous Things: What Categories Reveal About
the Mind. Chicago and London: The University of Chicago Press.
Lakoff, George & Johnson, Mark (1980). Metaphors We Live By. Chicago: The University of
Chicago Press.
Langacker, Ronald W. (1990). Subjectification. Cognitive Linguistics, 1, 5–38.
Pollard, Carl & Sag, Ivan A. (1994). Head-Driven Phrase Structure Grammar. Chicago: The
University of Chicago Press.
Wienold, Götz (1967). Genus und Semantik. Meisenheim am Glan: Hain.
Young, Robert W. & Morgan, William, Sr. (1987). The Navajo Language: a Grammar and
Colloquial Dictionary. Albuquerque: University of New Mexico Press.
Zubin, David A. & Köpcke, Klaus-Michael (1981). Gender: A less than arbitrary gram-
matical category. Chicago Linguistic Society, 17, 439–449.
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.17 (165)
Zubin, David A. & Köpcke, Klaus-Michael (1984). Affect classification in the German gender
system. Lingua, 63, 41–96.
Zubin, David A. & Köpcke, Klaus-Michael (1986). Gender and folk taxonomy: The indexical
relation between grammatical and lexical categorization. In C. Craig (Ed.), Noun
Classification and Categorization (pp. 139–180). Amsterdam and Philadelphia: John
Benjamins.
Zubin, David A. & Köpcke, Klaus-Michael (in preparation). The Irrgarten: Natural Cat-
egories in Language: A Study of Nominal Classification Systems with Particular Reference
to Gender in German.
Appendix
Excerpts from: Waggerl, Karl Heinrich. Legende von den drei Pfändern der Liebe. In: Kalen-
dergeschichten. Insel-Verlag, no year.
Underlinings: = feminine gender, = neuter gender. Square brackets indicate
bridging summary not in the original.
Da war ein armer Mann, ein Kesselschmied in einem Dorf, der hatte ein Mädchen, mit
dem er bald Hochzeit halten wollte. Und das war gut, denn das Mädchen liebte ihn
A mehr als alles in der Welt. Weil es aber nun am Geld für die Heirat fehlte, [zog] ... der
Mann in die Fremde, um seine Kessel in den Dörfern zu verkaufen. ... Da weinte nun das
Mädchen und bat ihn, zu bleiben. “Du wirst nicht wiederkommen,” klagte es, “ach, du
wirst mir untreu werden...!” ... Das Mädchen schwieg und verbarg seinen Kummer vor
ihm. Aber als er auszog, gab es ihm drei Pfänder der Liebe mit auf den Weg [Haarband,
Ring, und Messer ... Als er nach der Arbeit auf dem Markt sein Nachtlager fand], geschah
es, daß sich in der Dunkelheit eine fremde Frau an sein Lager gesellte. “Du gefällst
mir,” flüsterte sie, “du junger Kesselschmied!” ...Er küßte die fremde Frau und vergaß
alles und zog sie an sich. “Hast du kein Mädchen,” fragte sie...? “Nein,” antwortete
der Mann ... Und vor Tag, als die Frau von ihm Abschied nahm, und als sie zu
weinen anfing, da schenkte er ihr ein Band für das Haar zum Angebinde. ... [In der
zweiten Nacht] kam abermals eine Frau aus der Stadt an sein Lager, die sagte ihm süße
Worte ins Ohr und schlief bei ihm. “Hast du kein Mädchen daheim,” fragte sie leise...?
B “Nein...” ...auch in der dritten Nacht schlief er nicht allein, und sie schien ihm die kost-
barste von allen zu sein, diese Frau in der dritten Nacht. Die Frau schlang plötzlich die
Arme um seinen Hals und küßte ihn und weinte bitterlich. ... “Ach,” sagte die Frau,
“ich bin todtrauerig. Sicher hast du ein Mädchen daheim, das dich so liebt wie ich und
das vor Kummer stirbt....” Da verlangte der Mann nur noch heißer nach dieser Frau und
schwor seine Liebe vor ihr ab... Und am Ende der Nacht bat ihn die Frau um ein Zeichen,
daß sie an ihn denken könnte. ...er fand nur sein Messer in der Tasche, und das gab
er ihr zuletzt, weil es blank und scharf war, ein hübches Ding. ...
TSL[v.20020404] Prn:13/06/2003; 10:45 F: PB11307.tex / p.18 (166)
There was once a poor man, a village pot maker, who had a girl whom he soon wanted
to marry. And that was fine, since the girl loved him more than anything in the world.
A But since there was not enough money for the wedding, the man went off from home in
order to sell his pots in other villages. Now the girl cried and begged him to stay. “You’ll
never come back,” she wailed, “you won’t be faithful to me!” Then the girl was quiet and
concealed her sorrow. But as he left, she gave him three pledges of (her) love to take with
him, a hair ribbon, a ring, and a knife. When he had finished a day’s work at the market
and found a camp for the night, it happened that in the dark a mysterious woman came
to be with him. “You’re nice,” she whispered, “you little pot maker, you.” He kissed the
mysterious woman and forgot everything and drew her to him. “Don’t you have a girl?”
she asked. “No,” answered the man. And before morning, as the woman took leave of
him, and as she began to cry, he gave her a ribbon to put in (her) hair. The second
night a woman came again from town to his camp. She whispered sweet things in his
ear and slept with him. “Don’t you have a girl at home” she asked softly. No. ... Neither
in the third night did he sleep alone, and she seemed to him the most precious of all,
B this woman in the third night. The woman suddenly threw (her) arms around his neck
and kissed him and cried bitterly. “Ah!” cried the woman, “I’m so sad I could die. Surely
you have a girl at home who loves you as I do and who will die of sorrow.” Then the
man desired this woman even more passionately and declared his love to her. And at the
end of the night the woman begged him for a token for her to remember him by. But he
could find only the knife in his pocket, and that he finally gave her, because it was shiny
and sharp, a pretty thing. And now his thoughts turned to the trip home. [On the way
he finds at his sleeping places the ring and the ribbon] The last night he finally arrived
at home. And when he entered the main room, there lay his girl on the bier. And then
C he knew, that she was the one that he had loved three times, and three times betrayed,
and now there was his knife, thrust in the middle of her white breast.
TSL[v.20020404] Prn:4/06/2003; 11:43 F: PB113P3.tex / p.1 (167)
P III
Debra Ziegeler
. Introduction*
The analysis of counterfactual statements has provided the source for a diver-
sity of studies in recent years, ranging from comparative cross-linguistic re-
search (e.g. Kuteva 1998) to pragmatic approaches (e.g. Horn 2000; Van der
Auwera 1997), but much of this work has focused almost exclusively on the
counterfactual constructions found in the form of conditionals (see especially
Athanasiadou & Dirven 1997; Fauconnier & Sweetser 1996). While conditional
constructions are probably the most frequently cited environments for the ex-
pression of counterfactuality, counterfactual meaning is a form of inferential
reasoning associated with a number of textual and contextual clues but not
necessarily related to any particular grammatical construction. In this chapter,
the objective is to examine such clues in order to determine if there are cogni-
tive factors influencing the formation of a counterfactual implicature in the use
of a modal expression of ability. The chapter further aims to assess whether the
counterfactual meanings derive essentially from factors of formal complexity
of expression or from the conceptually based pragmatic relations of contiguity,
metonymy, or relevance.
The term ‘counterfactual implicature’ simply refers to a particular kind of
pragmatic implicature that is counterfactual in nature. In earlier studies (e.g.
Ziegeler 2000a) the creation of counterfactual reasoning was seen to be intri-
cately bound up with the co-operative maxims of Grice (1975), in particular,
the maxim of Quantity. However, in Levinson (1995), the maxim of Manner,
underlying what he reconstructs as a heuristic label, M: ‘marked descriptions
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.2 (170)
and modal adjectives like possible/necessary. However, for pairs like those in (1)
below (ibid.: 105), denoting past ability, Levinson argues that it is not Q1 but
rather M-inferences (as in (1b)) that complement Q2 inferences (as in (1a)):
(1) a. John could solve the problem (Q2 + > ‘and he did’)
b. John had the ability to solve the problem (M + > ‘but he didn’t’).
Visser suggests ‘have you not been able to’ as a present-day substitute for could
in (2). However, the modal is used in a negative context, which enables the abil-
ity meanings to become more transparent even in Present-Day English (PDE)
(‘couldn’t you gather any tidings of her’ might also serve as a suitable substi-
tute). According to Palmer (1986: 14), modals such as could and would cannot
be used to refer to a single event in the past; e.g. *he ran fast and could catch
the bus (1986: 93). It will be seen later in this chapter that isolated sentences
such as (1a), with a meaning of unambiguous past ability referring to single
events, are rarely found in bare declarative main clauses, even in diachronic
texts. Sentence (1a), then, is more likely to yield an R-based inference of future
potentiality than of past ability, leaving the two sentences non-synonymous.
The contrast expressed between the two forms, a modal and a semi-modal, pe-
riphrastic equivalent, may be related only to formal perceptual features, and
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.5 (173)
Levinson carefully avoids the past ability expressions used by Horn (1984,
1989) to illustrate the operation of the R-principle. Horn also discusses the use
of past ability expressions to provide the sources for implicatures, but in Horn’s
account, there is another more prolix, marked alternative of could, namely
be able to, which may be seen to produce the same Q2 inferences claimed by
Levinson (1995) for could in (1a):
(3) She was able to solve the problem.
which R-implicates:
(4) She solved the problem.
If, as Levinson seems to assume, could still contains senses of past ability in such
sentences, then was able to ought to be substitutable in (1b) as a more lexical-
ized, periphrastic marked alternative as well (be able to was noted by Denison
(1998: 171) as a suppletion for the use of could in assertive contexts). But in
spite of its relative markedness, the inferences obtained are also Q2 or R-based
implicatures, and are not complementary to those of could. Since (3) is also
a relatively marked expression by comparison to (1a), it is questionable that
relative markedness of formal expression is the only cause of the contrasting
implicatures.
The counterfactual implicature derived from the use of such expressions
must therefore be accountable to a more direct means of inferencing. Horn
(1984: 21) also discusses the fact that R-based implicatures can be canceled in
certain circumstances, and that the cancellation is obtainable without nega-
tion, using contrastive stress on the element in the sentence that derives the
implicature, in this case, the expression of ability. Thus:
(5) She was able to solve the problem . . .
could imply to the hearer that she only had the ability to solve it, but did not
actually solve it.
An appended clause introduced by an adversative conjunction such as but
and followed by a negative can produce the same cancellation; e.g.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.6 (174)
(6) She was able to solve the problem but (she didn’t, as) she didn’t have
enough time.
prediction as s/he would if the time reference were past. In present-day English,
the preterite forms of modals retain very little of their past temporal meaning,
but are reinforced by the use of the perfect auxiliary have to provide the same
senses of past time reference that were once available in Old and Middle En-
glish. Hence She could have come, but she said she had a lot of work to do is more
likely to express the speaker’s knowledge that the prediction was not fulfilled
than if the modal were in non-past form. The relation of the modal expressions
with the propositions they quantify can be illustrated thus to show differences
in temporal reference, where superscript t = reference time:
(10) a. non-past: mt [p] > p (by means of R-based implicatures); e.g. She may
be coming (mt [p]) > she is coming (p)
b. past: mt-n [p] (but X) > ∼p (by means of Q-based implicatures, or
canceled R-based), e.g. She might/could have come (mt-n [p]) (but she
said she had a lot of work to do) > she didn’t come (∼p).
The use of past time reference, for all intents and purposes, increases the
hearer’s evaluations of the speaker’s capacity to provide more accurate infor-
mation of the truth value of an utterance, and results in a higher probability
of a counterfactual implicature as a result. In this way, the Quantity maxim is
seen to interact with the maxim of Quality (‘do not say that for which you lack
adequate evidence’), the hearer believing there are grounds for the speaker’s
contribution to be backed by more accurate evidence, as is usually available
when referring to past events. The most likely interpretation, then, with past
time reference, produces a Q-based implicature of negative prediction.
Even with the contrastive clause removed, the modal expression in (11) alone
may stand for the counterfactual inference that the subject did not go to the
church; e.g. Did he go to the church? He would have gone ... . In such cases, the
cancellation of an R-based implicature originally suggesting the actuality of the
proposition predicated of the modal becomes part of the meaning of the modal
itself, and the modal acquires negative inferences via a metonymic process.
Such processes are similarly described for the development of sentence-final
particles out of conjunctions in Japanese, by Okamoto (this volume), in which
ellipsis of a subordinate clause originally introduced by a conjunction (koto)
results in the modal meanings it conveyed becoming conventionalized as part
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.10 (178)
of the semantics of the new construction; also, in the case of independent if-
clauses in English, German and French (see Panther and Thornburg, this vol-
ume), the semantics of the ellipted main clauses are conventionalized as part
of the meaning of a resulting speech act function for the stranded if -clause. In
all such cases, the metonymy is ‘part’ for ‘whole’.
The example in (11) might illustrate a possible diachronic route for the
development of counterfactual implicatures as the result of metonymic infer-
encing, and indicate the role played by the Gricean maxim of Quantity (revised
by Horn (1984) in terms of R- and Q-based implicatures on Quantity scales)
in the development of metonymic inferences of this kind. The Quantity impli-
cature is brought about by the quantificational evaluation of information in an
utterance, and will of necessity involve the understanding of part-whole rela-
tionships, one of the types of metonymic relationships discussed by Kövecses
and Radden (1998). On the Horn-scale, as discussed above, part-whole rela-
tionships are represented by weak and strong counterparts of gradable qual-
ities, quantities, and entities. In proposing modality as a measurement on a
Horn-scale, the part-whole relationships are seen as corresponding to predi-
cational equivalents of modal versus non-modal propositions. The expression
of counterfactuality, marked by only a modal predication type, involves the
assertion of a part (the modality) for an implicit whole (the modality and the
negated proposition it quantifies factually). The metonymic extension that cre-
ates the counterfactual implicature is of necessity grounded in the operation of
the maxim of Quantity 1, or Q-based implicatures (canceled R-based implica-
tures), and must be considered linked to Gricean principles of conversational
inferencing.
Apart from consideration of formal prolixity, there are a number of other con-
siderations that must be accounted for in formulating a hypothesis based on
a metonymical analysis. The examples given by Levinson (1987) and Horn
(1984) illustrate only one possible function for the use of past ability expres-
sions, and that is in expressing the attainment of a result. Less than adequate
consideration has been given to the possibility that expressions of past abil-
ity can, and frequently do occur in describing, for example, the inherent at-
tributes of the subject, also discussed by Panther and Thornburg as a type of
potentiality for actuality metonym, the domain of ‘character disposition’
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.11 (179)
The first two forms are acceptable for the possible reason that they can both
be used to express potentiality or possibility; had the ability to + V has not yet
grammaticalized its modal meanings in this way. As such it may be predicted
to be a recent introduction into the paradigm.
As well as aspectual changes in the environment of the past ability expres-
sion, a number of other factors may come into play, which affect the type of
inferences obtainable. The first is the environment of factive clauses. A factive
clause, in the present definition, is not intended to be defined in exactly the
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.12 (180)
same way as Kiparsky and Kiparsky (1970) define such predicates; that is, as
determined by the presupposition of truth in the that-clause complement or
the gerund complement of a matrix clause containing a particular factive verb;
e.g. regret (I regret that she was ill). In some cases, though, there is overlap
with their definitions; e.g. a relative clause with a referring head noun can be
classed as factive as it usually contains a presupposition. There is also factivity
associated with complements of verbs of reporting or communication (that-
clauses), which depends on the nature of the matrix verb: e.g. prayed/wished
that introduces a non-factive or counter-factive environment, while said that,
promised that, realized that, and found that are superordinate to clauses that are
taken in the present study to be factive.1 Examples of such clauses will appear
in the diachronic texts (see Sections 5.2.3–4); however, as an illustration of the
effect of such clauses on the inferences derived from modal and semi-modal
expressions, the following examples may suffice:
(14) a. John found that he had the ability to solve the problem.
b. John found that he could solve the problem.
In both (14a–b) the inferences are more likely to point to the actuality of
the complemented event, rather than the non-actuality. The stronger senses
of positive prediction are the result of the embedding of the modal or semi-
modal in a subordinate clause governed by a verb introducing a proposition of
presupposed factuality.2 Relative clauses may also contain a presupposition of
factuality, as noted above:
(15) a. The students who could solve the problem were given credit points.
b. The students who had the ability to solve the problem were given
credit points.
Sentence (16b) seems to imply that, in all probability, John had the ability to
solve the problem, and he did. Why this should be so is unclear, but perhaps
reasons attributable to sentence focus may provide a clue, the use of a restrictor
shifting attention away from the modality to the subject referent instead.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.13 (181)
. Interim summary
In the previous sections we have examined the question of whether the coun-
terfactual implicatures derived from the use of a semi-modal expression had
the ability to in contrast to a modal verb could are the result of M-inferences
of the type described by Levinson (1995) or Q-based implicatures of the type
described by Horn (1984) (and are therefore accountable by metonymic pro-
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.14 (182)
In order to verify these points, and to investigate whether such a stage existed,
it is necessary to consider the function of marking past ability modality in its
historical context. It is for this reason that the following survey was undertaken.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.15 (183)
Be able to + past-environments
N: 23 = 100%
0 1 (4.3%) 22 (95.6%)
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.17 (185)
Middle English
Total = 1 (Subordinate clause, factive)
Table 2a. Distribution by function and environment of be able to + past in EME main
clauses.
N=6
Non-factive clauses 1 1
Characterizing 1 2 3
Other 1 1 2
TOTALS 3 3 6
could-environments
Table 3. Frequency distribution of could forms in the diachronic survey. Forms inves-
tigated included cu∂, coud, couþ, couth, cuþ, kowth, cuth, cowd, koud, kowd, cowth, ku∂,
kouth, cou’d, culd, kouth.
N: 215 = 100%
Old English
Total = 8
Main Clauses: N = 1 (Lexical)
Subordinate Clauses: N =7
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.18 (186)
Lexical uses* 1 1 2
Factive clauses 2 2 4
Non-factive clauses 1 1
TOTALS 3/7 3/7 1/7 7
*These include two examples of the use of the past participle cu∂ , ‘known,’ a non-finite form, which is discussed as a
lexical item distinct from the modal paradigm by Goossens (1992: 381). However, it is included in the present analysis
because a number of non-finite forms are included in the data for be able to.
Middle English
Total = 41
Main Clauses: N= 25/41 = 60.9%
Table 6a. Distribution by function and environment of could forms in EME main
clauses.
N = 44
Table 6b. Distribution by function and environment of could forms in EME subordi-
nate clauses.
N = 122
for characterizing uses in the Middle English could main clauses: 12 of the 17
tokens came from a single text: Chaucer’s The General Prologue to the Canter-
bury Tales. However, the tokens for the other features are all fairly evenly dis-
tributed throughout the texts. The forms will be examined in turn, and their
role in the overall grammaticalization of the function of expressing past ability
will be discussed.
The predicate of past ability is dislocated in (18) because of the metrical re-
strictions of the verse, without which the sentence might be read as ‘there was
a monk, a manly man, (who was) able to have been an abbot,’ thus providing
the earliest example of this form in the texts, if such a reconstructed context
might serve as an adequate example. The use of the form refers to the abilities
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.21 (189)
and general characteristics of the subject, the monk, and reflects no possible
senses of past actuality; in fact, the non-actuality (that he was not a monk) is
more apparent, though this is not the dominant meaning in (18). The use of
the form has a general characterizing function. The only ME example in Visser
also has a generic predicate:
(19) the hauene was not able to dwelle in winter
‘the harbour was not suitable for winter dwelling’
c1380. Wyclif. The Acts of the Apostles, 27. 12
Visser (1969: 1750)
In (19) the subject is inanimate (‘the harbour’) and the predicate is a stative
verb (‘dwell’), thus illustrating again a characterizing function in the semi-
modal, and describing general facts about the subject. (It should be noted that
the subject of was is not the same as the subject of dwelle, suggesting that the
modal sense had not by then evolved.) There are no other examples of the semi-
modal appearing in either Visser or the Helsinki Corpus in the Middle English
data surveyed.
The Early Modern English data showed a far higher frequency of occur-
rence of the form: 22 tokens overall, with 6 in main clause uses and 16 in subor-
dinate clauses. The uneven distribution between main clauses and subordinate
clauses reflected that of all the forms across all the data: with the exception
of ME could, there was consistently a greater number of subordinate clause
uses than main clause functions. The reason may be that subordinate clause
functions usually reflect subjunctive moods, as noted by Bybee et al. (1994),
and such environments are semantically modal in any case, making them more
readily receptive to the introduction of modal and semi-modal forms. Factive
clauses again ranked high in frequency (87.45% in total), and negatives were
relatively frequent overall (49.95%).
Thus, by far the largest number of occurrences in the EME data for
was/were able to +V came from factive subordinate clauses, examples of which
occur in temporal subordinate clauses; e.g.:
(20) But yet wee must say plainely, That it appeareth strange vnto vs, when wee
were able to shew it, that since the comming to our Crowne, it was never
denyed vs by any of his predecessors
1570–1640. Robert Cecil, The Edmondes Papers
(Ed. G. G. Butler). p. 402
The basic meaning of were able in (20) seems to express the generic past abilities
of the subject, but restricted to a specific time reference; it is not clear whether
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.22 (190)
By the late 17th century, then, examples had started to appear in which the
use of was/were able to coincided with a meaning of ability specific to a sin-
gle event; i.e. accomplishment. In (21) the single event is still not totally in-
dependent of other conditions, but the data suggests that this is the route by
which meanings of past accomplishment may have developed out of earlier,
more lexical functions expressing the skills and general abilities of the sub-
ject. In negative constructions, however, the path to the development of mean-
ings of non-actuality is much faster. By the early part of the EME period, non-
accomplishment of specific events was already part of the meaning of the form
in factive subordinate clauses:
(22) and some of those that came thyther with the duke, not able to dissemble
theyr sorow, were faine at his backe to turne their face to the wall
1500–1570. Thomas More, The History of King Richard III
(Ed. R. S. Sylvester). p. 77
In (22) there is again a deleted relative clause understood in the context (‘who
were not able to dissemble their sorrow’), and the clause is factive. The con-
text indicates that the inability of the subjects was related to a specific moment;
i.e. when they came with the duke, and was not an inherent characteristic or
generic feature of a time-stable nature. This suggests that time-specific uses
were already appearing by the early 16th century, at least in negative subordi-
nate clauses. No negative examples appear before the EME period in the data,
though, clearly, (19) appears in Visser’s data as an example of a negative char-
acterizing function. By the 17th century, examples are appearing in negative
main clauses, with the meaning of negative accomplishment (non-actuality) of
a specific event:
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.23 (191)
(23) Before he was twenty, he came into the house of commons, and was on
the king’s side, and undertook to get Wiltshire and Dorsetshire to declare
for him, but he was not able to effect it.
1640–1710. Gilbert Burnet, Burnet’s History of My Own Time I
(Ed. O Airy). p. 1, I, 172
.. Could-forms
The use of the modal of past ability was much more frequent in the survey,
indicating that it is an older layer in the paradigm and therefore more highly
grammaticalized than the other two forms. Only eight tokens appear in the
Old English data, one in main clauses and seven in subordinate clauses. Three
out of the eight OE functions are unambiguously lexical, with noun phrase or
nominal clauses as objects, while factive subordinate clauses also occupy more
than half of the tokens, either lexical or non-lexical. Four of the eight appear
in negative sentences. In (24), could has a lexical function also expressing nega-
tion, in which the object is a NP, illustrating the earliest uses of the modal as a
verb expressing knowledge or the mental ability of the subject:
(24) ∂a he þa geornlicor me frægn be his þingum, þa sæde ic ∂æt ic his ∂inga
feola ne cu∂ e.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.24 (192)
‘When he earnestly asked me about these things, I then said I did not know
much about them.’
850–950. Alexander’s Letter. (Ed. S. Rypins). p. 27
Lexical functions continue into Middle English uses (which number 41 in to-
tal), where they still occupy a proportion of 22% of all environments. The per-
centage of negatives, though, drops to around 12.1%, a much lower proportion
than in the Old English data. A significant proportion (26.8%) of all environ-
ments (apart from lexical) is found in non-factive subordinate clauses, which
do not carry presuppositions as do factive clauses, and therefore are likely
to render the modal meanings contained in them to be hypothetical in na-
ture. Some non-factive subordinate clause environments included comparative
clauses of indefinite degree, or comparative clauses with non-referring subjects:
(25) the mair, John Norhampton, reherced as euel as he koude of the eleccion
on the day to-forn
1384–1425. Thomas Usk, A Book of London English
(Ed. R. W. Chambers & M. Daunt). p. 28
The subordinate clause in (25) does not contain an assertion nor does it con-
tain a presupposition, indicating that the extent of the subject’s abilities is un-
known. However, at the same time, Chaucer’s work is still replete with ex-
amples illustrating the unambiguous characterizing use of could, which he
uses profusely in one text to describe the attributes of the characters in The
Canterbury Tales, e.g.:
(26) Therto he koude endite and make a thyng,
Ther koude no wight pynche at his writyng;
And every statut koude he pleyn by rote.
1350–1420. Geoffrey Chaucer, The General Prologue to
the Canterbury Tales. (Ed. L. D. Benson). p. 28
In (26) three different functions are represented simultaneously: in the first ex-
ample, the use of could describes the capabilities of (presumably) the lawyer, in
a characterizing function; in the second the function expresses inability, non-
actuality and since the context does not refer to a single event, thus impossibil-
ity for the subject to þynche at (‘find fault with’) his writing; in the third, the
earlier use of the modal is represented in a full lexical verb meaning ‘know.’ This
example provides good evidence to demonstrate the co-existence of overlap-
ping stages of grammaticalization at one synchronic time-point and is indica-
tive of the gradualness of the developmental route from meanings of knowl-
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.25 (193)
edge to physical ability or skills, and finally to senses of possibility (or impos-
sibility in the case of negatives). Most of the examples from this text, though,
illustrate a characterizing function for could, and 15 of the 17 characterizing
uses listed in Table 5a are found in this one text.
The EME data is marked by an absence of lexical functions (only one exam-
ple), an increase in negatives (44.6% overall) and an increase in the number of
factive subordinate clauses (61.4% of all subordinate clauses; 45% of the total
EME figures). Also noteworthy is the emergence of counterfactual/hypothetical
meanings, created by the cancellation of R-based implicatures in main clause
environments. One such example is the following:
(27) I could be content, but it will aske some time, and I am going to such a
place vpon speciall busines.
1593. George Gifford, A Handbook of Witches. P. B2R
The R-based implicature in the first clause, suggesting that the subject will po-
tentially be content, is canceled by a conjoining clause expressing contrast and
therefore implying that the subject will not be content after all as he or she is
asking for more time. It should also be noted that by this time in the grammat-
icalization of could, most of the temporal meanings have faded and the modal
now refers to the potentiality of the predicate to be realized after, not prior to
the moment of speaking. In order to restore the bleached temporal meanings,
the perfect auxiliary have is now used, as noted earlier:
(28) I could haue bin well content to haue chose seuen Yeres Imprisonment, . . .
rather than I would this Day haue gyuen Euidence against Sir (Nicholas
Throckmorton) but sithence I must needes confesse my Knowledge, I
must confesse all that is there written is true.
1554. The Trial of Sir Nicholas Throckmorton
(Ed. F. Hargrave) P 1, 67.C2
When the modal clause alone (could have V-ed) may stand to represent the for-
mer modal clause + canceling clause (could have X but Y, therefore not-X), the
counterfactual potentiality for non-actuality metonym has developed.
What is important though, is that at the basis of its development is a scalar im-
plicature, an R-based implicature that has been canceled by an appended con-
trastive or adversative clause. That the basis is scalar is determined by the entail-
ment of impossibility in the negated modal (see e.g. (17)); such inferences are
the result of relationships between weak and strong elements on scales of the
type discussed in Horn (1972, 1984). There is no metonymic link between neg-
ative potentiality and negative actuality, as negative potentiality is simply in-
compatible with actuality. This development may be said to convert the impli-
cature into a Q-based one, thus creating a counterfactual metonym: the weak-
ened factuality contained in the modal expression now stands metonymically
for the counter-factuality of what it predicates.
It is clear from the data in the survey that the use of could as a verb meaning
past ability never generalized to attain meanings of past actuality, in the same
way as was/were able to. If it had, it would be hardly appropriate to suggest
grammaticalization as a motivation, since the generalization of generic mean-
ings to mark specific, time-stable events is an instance of entailment, and not
of the conventionalization of conversational implicatures typically associated
with the processes of grammaticalization. Furthermore, there is a notional in-
crease of agentivity that could be associated with the past actuality meanings in
able, an adjectival predicate, though could grammaticalized to undergo a loss
of lexical meaning instead of an increase, with its concurrent development as
an auxiliary verb. The loss of lexical uses that is clearly evident in the EME
period was also accompanied by a development of potentiality meanings in
main clauses, which could be modified by the addition of a coordinate clause
of contrast or adversity, or a subordinate clause of condition, concession or
qualification of the hypothetical meanings.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.27 (195)
Table 7a. Frequency and distribution of had the ability (to + V) in main clauses in
a small sample of the British National Corpus. Forms also included did not have the
ability (to + V), had no ability to, and had the ability to + V occurring within the scope
of a negative matrix clause or a negative subject NP.
Total: 48 = 100%
Main Clauses 27 = 58.3%
Environments counterfactual negative TOTALS
Table 7b. Frequency and distribution of had the ability (to + V) in subordinate clauses
in a small sample of the British National Corpus.
Subordinate
Clauses 21 = 41.6%
Environments characterizing counterfactual TOTALS
Factive 4 = 19% 15 = 71.4% 1 = 4.7% 20 = 95.1%
clauses
Non-factive 1 = 4.7% 1 = 4.7%
clauses
TOTALS 4/21 = 19% 16/21 = 76.2% 1/21 = 4.7% 21 = 99.9%
main clauses and 76.2% of subordinate clauses). The clearly low percentage of
counterfactual implicatures is obvious, and may suggest that Levinson’s (1995)
example (1b) is not typically representative of current patterns of use, and that
the analysis given to it suffers from the usual risks of employing constructed ex-
amples to make generalizations about language use. This can be seen in the fol-
lowing example, the only valid instance in the sample in which a counterfactual
inference might be said to develop:
(30) The penal jurisdiction of the county courts was to apply only where the
debt had been contracted under circumstances which implied an inten-
tion to defraud (which upset Wetherfield) or where the debtor had the
ability to pay but would not. (BPH 718)
The necessity in (30) is for a canceling clause, but would not, denying the pos-
sibility of an R-based implicature derived from the use of the ability expression
(‘had the ability to pay, and did pay’). The possibility is, of course, increased by
its environment, a relative clause that is factive in nature. However, even in a
main clause, cancellation is possible with a but-clause of adversity, suggesting
the likely non-actuality of the potential event:
(31) . . . I had the ability to communicate verbally but at a child’s level and
almost no ability to read. (CGU 719)
is still inferable from the information provided in the adversative clause alone.
For the same reason, a reinforcing clause is also still possible, introduced by
and and co-occurring with had the ability expressions without redundancy or
contradiction:
(32) She had the ability and now she had the break. (BP7 2026)
(33) All through his career he’s had the ability to lead and inspire and he’s really
got through to our players and conveyed that. (CBG 13033)
In (32) the predicate is not supplied, but may be implied in the previous con-
text. The conjoining clause (and now she had the break) suggests that the sub-
ject had the opportunity to accomplish the realization of the predicate, and
would be inappropriate if the first clause carried a default Q-based implica-
ture of the non-actuality of the focus of ability. It may also be redundant if
the R-based implicature was conventionalized (consider She was able to win the
race, and now she had the break, and therefore she did win in which the con-
joining clause is not required). Sentence (33) is a similar example, in which
the reinforcing clause (and he’s really got through to our players and conveyed
that) provides later confirmation of the subject’s characteristic skills or abili-
ties; thus ‘he’s had the ability to lead and inspire, and he has done that’ rein-
forces the R-based implicature of the actuality of the subject’s abilities. In the
data there were three instances of reinforcement in main clauses and one in
subordinate clauses.
The majority of the examples, though, as noted, described the generic skills
and abilities of the subject, with no modification, except perhaps the presence
of a frequency adverb such as always (in five of the main clauses and in two of
the subordinate clauses), e.g.:
(34) While Enzo Ferrari always had the ability to attract the best drivers in
the world, . . . the one man he never managed to get was Britain’s Stirling
Moss. (EXI 438)
The modal verb could, then, which at one stage of its development expresses
meanings of past ability, does not follow the same path of periphrastic modal
expressions of past ability such as was/were able to and go on to express past
actuality. The reason for this may be that as an older form of modality, it was
reduced, less periphrastic, and had a more frequent use historically, which led
in turn to semantic erosion and bleaching of its lexical senses of temporal-
ity and ability. In the case of could, there is a possibility that characterizing
uses created the evidential basis for the generation of hypothetical statements
about the potentiality of the subject, in both main clauses as well as subordinate
clauses. The weakening of temporal meanings of pastness in the modal would
contribute thus to the development of such meanings. Therefore, past actu-
ality in main clauses was never a stage in the grammaticalization path, as the
loss of temporal meanings ensured that it would never occur in such clauses.
Generic meanings of ability as a past characteristic may have served to jus-
tify the predictive nature of the new developing potentiality meanings: hence
the chameleon-like appearance of examples in the later part of the survey with
interpretations ambiguous between past ability or potentiality and future po-
tentiality (e.g. (25)). As the modal progressed on its way to become a marker
of potential modality, the past ability meanings were then renewed by was/were
able to, confirming the fact that the temporal meanings had been weakened.
When at a later stage, the meanings of potentiality were re-extended to past
contexts, the auxiliary have was introduced as a reinforcement, marking what
appears to be back-formation in could have + V-ed types.
Thus, what appear to be predominantly generic uses of was/were able to in
the earlier texts become extended in EME to meanings of ability relative to a
specific time point. Once the function has generalized in this way, R-based im-
plicatures producing metonymic extensions of past actuality familiar to main
assertive clauses will start to develop, as illustrated in (21). Had the ability to,
though, appears to be still at the stage at which the most frequent use is in the
expression of time-stable characteristics attributable to the subject, and exam-
ples similar to Levinson’s (1995) example (1b) that refer to single events are
found only with modifying clauses of adversity or contrast (see e.g. (30) in the
data surveyed). The presence of a modifying clause suggests that there is an un-
derlying R-implicature of potentiality associated with such expressions, and the
modifying clause acts to cancel or suspend it. No likely examples of Q-based
implicatures occur in had the ability to + V forms in bare or unmodified main
declarative clauses; the future conventionalization of this implicature to either
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.31 (199)
. Conclusions
Thus the results of the changes taking place in the development of the
modals and semi-modal equivalents may be described within the terms of a
metonymic analysis, but underlying the creation of the metonymic inferences
are certain principles explaining the direction of the metonym: either to ex-
press the actuality of the predicated event or its non-actuality, and such prin-
ciples appear to be founded in pragmatics and the notion of scalar relation-
ships between items. With further research in the field the correlations be-
tween the pragmatic principles of Quantity scales and metonymic inferences
may contribute an important cognitive dimension to the study of the Gricean
co-operative principle. The prospects remain an exciting possibility for future
research to reveal.
Notes
* I would like to thank Klaus Panther and Linda Thornburg particularly for their assistance
in the presentation of this paper at the “Metonymy and Pragmatic Inferencing” workshop
during the 7th International Pragmatics Conference, Budapest, July 7–14, 2000. A longer,
more detailed version of the paper may be found under the title “Past ability modality and
the derivation of complementary inferences” in Journal of Historical Pragmatics 2: 273–316
(2001).
. As pointed out by an anonymous referee, there is reason to doubt the factivity of clauses
following said that and promised that. However, in the present study, they will be classed as
such (with all good faith in the word of the subject referent) on the basis of the slightly higher
degree of expected realization of their complement propositions than can be attributed to
the members of the non-factive category.
. The strength of the inferences may vary according to the verb in the matrix clause, as
suggested by the same referee.
. The forms investigated were spelling variants taken from the OED. The only variants
appearing in the Helsinki Corpus (apart from able) were abil and hable, the earliest (and
only) citation for abil dated at 1350–1420 (the ME period), in The Cloud of Unknowing and
the Book of Privy Counselling, and the earliest citation for hable being in Roger Ascham’s The
Scholemaster (1563–1568), in the EME period.
. Acknowledgment is due to Matti Rissanen for permitting me access to the diachronic part
of the Helsinki Corpus of English Texts. Reference should be made to Kytö (1996) for details
on word count totals and for additional information regarding sources. A small portion of
the Middle English corpus (ME IV, as listed in the manual) was unavailable.
. This corpus can be found online at: http//sara.natcorp.ox.ac.uk. The sample is taken from
the 100 million-word corpus, but restricted to a maximum of 50 tokens.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.33 (201)
References
Athanasiadou, Angeliki & Dirven, René (Eds.). (1997). On Conditionals Again. Amsterdam
and Philadelphia: Benjamins.
Atlas, Jay D. & Levinson, Stephen C. (1981). It-clefts, informativeness, and logical form:
Radical pragmatics. In P. Cole (Ed.), Radical Pragmatics (pp. 1–61). New York:
Academic Press.
Bybee, Joan L. (1988). Semantic substance vs. contrast in the development of grammatical
meaning. Berkeley Linguistics Society, 14, 247–264.
Bybee, Joan L. (2003). The role of repetition in grammaticization: CAN in English. In R.
Janda & B. Joseph (Eds.), Handbook of Historical Linguistics. Oxford: Blackwell.
Bybee, Joan L., Perkins, Revere D., & Pagliuca, William (1994). The Evolution of Grammar:
Tense, Aspect and Modality in the Languages of the World. Chicago: The University of
Chicago Press.
Denison, David (1998). Syntax. In S. Romaine (Ed.), The Cambridge History of the English
Language, Vol. IV: 1776–1997 (pp. 92–329). Cambridge: Cambridge University Press.
Fauconnier, Giles & Sweetser, Eve (Eds.). (1996). Spaces, Worlds, and Grammar [Cognitive
Theory of Language and Culture 2]. Chicago and London: The University of Chicago
Press.
Grice, H. Paul (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 41–58). New York: Academic Press.
Horn, Laurence R. (1972). On the Semantic Properties of the Logical Operators in English. Los
Angeles: UCLA dissertation.
Horn, Laurence R. (1984). Towards a new taxonomy for pragmatic inference: Q-based and
R-based implicature. In D. Schiffrin (Ed.), Meaning, Form and Use in Context: Linguistic
Applications [GURT ‘84] (pp. 11–42). Washington, DC: Georgetown University Press.
Horn, Laurence R. (1989). A Natural History of Negation. Chicago: The University of
Chicago Press.
Horn, Laurence R. (1991). Given as new: when redundant information isn’t. Journal of
Pragmatics, 15, 313–336.
Horn, Laurence R. (2000). From if to iff : Conditional perfection as pragmatic
strengthening. Journal of Pragmatics, 32, 289–326.
Kiparsky, Paul & Kiparsky, Carol (1970). Fact. In M. Bierwisch & K. E. Heidolph (Eds.),
Progress in Linguistics: A Collection of Papers (pp. 143–173). The Hague: Mouton de
Gruyter.
Kövecses, Zoltán & Radden, Günter (1998). Metonymy: developing a cognitive linguistic
view. Cognitive Linguistics, 9, 37–77.
Kuteva, Tania (1998). On identifying an evasive gram: Action narrowly averted. Studies in
Language, 22(1), 113–160.
Kytö, Merja (compiler). (1996). Manual to the Diachronic Part of the Helsinki Corpus of
English Texts: Coding Conventions and Lists of Source Texts (3rd ed.). Helsinki: University
of Helsinki, Department of English.
Levinson, Stephen C. (1987). Pragmatics and the grammar of anaphora: a partial pragmatic
reduction of binding and control phenomena. Journal of Linguistics, 23, 379–434.
TSL[v.20020404] Prn:13/06/2003; 10:47 F: PB11308.tex / p.34 (202)
Primary sources
Old English
Rypins, S. (Ed.). (1971). (1924). Alexander’s Letter (Early English Text Society 161). New
York: Kraus Reprint Co.
Middle English
Benson, L. D. (Ed.). (1987). The Riverside Chaucer. Boston: Houghton Mifflin.
Chambers, R. W. & Daunt, M. (Eds.). (1967, 1931). Thomas Usk Appeals: A Book of London
English 1324–1425. Oxford: Clarendon.
Hargrave, F. (Ed.) (1730). The Trial of Sir Nicholas Throckmorton, Vols. I and IV, 2nd ed.
London: J. Walthoe Sen.
Sylvester, R. S. (Ed.). (1963). Thomas More, The History of King Richard III [The Complete
Works of St. Thomas More, Vol. II]. New Haven and London: Yale University Press.
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.1 (205)
Shigeko Okamoto
. Introduction
It has been recognized that metaphor and metonymy are two major processes
at work in grammaticalization (e.g. Hopper & Traugott 1993). In this chapter I
consider the role of metonymy in instances of grammaticalization in Japanese
that involve reanalyses of complementizers as sentence-final particles – reanal-
yses of grammatical morphemes that are quite common in Japanese and that
provide interesting examples of functional shifts in linguistic forms.
In Japanese, a verb-final language, there are many grammatical mor-
phemes used as markers of subordinate clauses – i.e. conjunctive particles (e.g.
ba ‘if ’, noni ‘although’, node ‘because’) and complementizers (see below for ex-
amples). These morphemes may also be used as sentence/utterance-final par-
ticles, expressing various pragmatic meanings. As discussed in a number of
recent studies (e.g. Ohori 1995; Okamoto 1995; Suzuki 1999), the latter usage
(i.e. sentence-final particles) can be considered as resulting from the functional
reanalyses of the former (i.e. markers of subordinate clauses) that involve a
shift of subordinate clauses to main clauses, constituting independent gram-
matical constructions in the sense of Fillmore et al. (1988), Kay and Fillmore
(1999), etc. The question is how and why these reanalyses may take place. The
present study addresses this question, focusing on the reanalysis of comple-
mentizers (comps) as sentence-final particles (sfps). It considers what moti-
vates and enables such a reanalysis, paying special attention to the nature of
pragmatic inference involved in the process of reanalysis, in particular, the role
of metonymy.
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.2 (206)
There are a number of morphemes in Japanese that can be used either as com-
plementizers or as sentence-final particles – e.g. no, koto, to, tte, and ka. As
comps, these morphemes have two functions: to mark a clause boundary by
nominalizing a complement clause and to indicate the epistemic status of the
proposition in the complement clause. For example, no is typically used for a
concrete or directly perceived event/state (e.g. (1));1 koto for an abstract or in-
directly perceived event/state (e.g. (2)); and to/tte for a quotative remark (e.g.
(3)) (cf. Kuno 1973; Josephs 1976).2,3
(1) Boku wa Yoshio ga Masao o butsu no o (*koto o/*to)
I tm sm om hit comp om
mi-ta.
see-pst
‘I saw Yoshio hitting Masao.’
(2) John wa nihongo ga muzukashii koto o (*no o/*to) manan-da.
tm Japanese sm difficult comp om learn-pst
‘John learned that Japanese is difficult.’ (Kuno 1973)
(3) Yoshio wa Masao ga shin-da to/tte (*no o/*koto o) it-ta.
tm sm die-pst comp say-pst
‘Yoshio said that Masao died.’
(7)) or for self-affirmation (e.g. (8)); and tte for reporting a proposition ex-
pressed by someone else (e.g. (9)) or for expressing the speaker’s insistence on
the proposition (e.g. (10)).4
(4) <from natural conversation>
A: Nee, kono zukini wa tada chiizu o mabush-ita dake na
hey this zucchini tm only cheese om coat-pst only cop
no.
sfp
‘Hey, did you coat this zucchini only with cheese?’
B: Aa, are, panko to hanhan na no.
oh that breadcrumbs with half-half cop sfp
‘Oh, that is half bread crumbs and half cheese.’
(5) Ii o- tenki da koto.
nice ppx weather cop sfp
‘What nice weather!’
(6) <from Asahi Shinbun, a daily newspaper>
Toku ni ta no habatsu ni tsuite wa issai hure-nai koto.
especially other factions about tm at all discuss-neg sfp
Miyazawa-ha no jimusyo ni konna ohuregaki ga haridas-are-ta.
faction-gen office in such order notice sm post-pass-pst
‘In particular, you must not discuss other factions at any time. Such a
notice of orders was posted in the office of the Miyazawa faction.’
(7) Nan da to. Tsumaranai da to.
What cop sfp boring cop sfp
‘What?! Boring?!’
(8) Moo kippu wa kat-ta to.
already ticket tm buy-pst sfp
‘I’ve already bought the ticket.’
(9) <from a TV drama>
Yama tte abunai n desu tte.
mountain tm dangerous aux sfp
‘Mountains are dangerous, I hear.’
(10) <from a TV drama>
Hontoo da tte. Ore puropoozu nan ka shite-nai tte.
true cop sfp I proposal such a thing do-neg sfp
‘It’s true, really. I didn’t propose marriage to her, really.’
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.4 (208)
One might argue that utterances such as (4)–(10) above are elliptical expres-
sions, or the result of main clause deletion, and hence that these morphemes in
question are not truly sfps at all, but rather are comps. For example, the mor-
pheme koto for exclamation (e.g. (11a)) might be analyzed as a comp, that is,
as an abbreviation of an expression such as S koto ni odoroita ‘(I) was amazed
that S’ (e.g. (11b)) or S koto to ittara nai ‘it is indescribable that S’ (e.g. (12b)):
(11) a. O- niwa ga kiree da koto.
ppx garden sm pretty cop sfp
‘How pretty the garden is!’
b. O- niwa ga kiree na/*da koto ni odoro-ita.
ppx garden sm pretty cop comp be amazed-pst
‘I was amazed that the garden is so pretty.’
Furthermore, the verb cannot be in the so-called potential form (-(rar)e). For
example, (13a) does not have a directive force:
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.6 (210)
Similarly, (15), in which the verb is followed by the auxiliary nakereba naranai
‘must,’ is unacceptable as a directive utterance:
(15) *Oya o daiji ni shinakereba naranai koto.
parent om take good care of must sfp
‘You must take good care of your parents koto.’
These examples indicate that a sense of duty is associated with koto, and that
koto imparts to the utterance a directive sense.
One might argue that (13), for example, could be paraphrased as (13b):
(13) b. Saku no naka ni hairanai koto o meejiru.
comp om order
‘I order you not to go inside the fence.’
The conclusion that the exclamatory or directive koto in [S koto] is not a comp
raises the question of how koto has come to be used as an sfp, or a modality
marker. Is this usage simply arbitrary or is it motivated?
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.7 (211)
two kinds of knowledge may be used in establishing the link between [S koto]
and [[Proposition koto] Modality].
Suppose, for example, (16) is used as an abbreviated expression:
(16) Mainichi ha o migaku koto.
every day teeth om brush
‘You brush your teeth every day-koto.’
To interpret (16) as a directive, one must among others know the frame for
directives. Modifying the scenario for requests given by Panther and Thornburg
(1998), I present a simplified frame for directives in (17):
(17) Simplified frame for directives
The before: H can do A (the action in question).
Doing A is desirable.
Sp wants H to do A.
Sp has the authority to put H under obligation.
The core: Sp puts H under strong obligation to do A.
The result: H must/should do A.
The after: H will do A.
As we can see, S in [S koto] refers to the after component in this frame.
In addition, the role played by koto in [S koto] cannot be ignored. As men-
tioned in Section 2, koto as a comp is typically used to indicate the epistemic
status of the proposition in the complement as being abstract or indirectly per-
ceived state of affairs. Further, it indicates that [S koto] is a constituent of the
main clause that may express the result of some kind of mental or perceptual
activity, including one’s attitude toward the proposition expressed by S. This
attitude, or stance, may be epistemic (e.g., Yamada-kun ga Amerika ni itta koto
wa tashika da ‘It’s certain that Yamada went to Amerika’), evaluative or emo-
tional (e.g., Ryooshin ga rikon-suru koto ni sansei da ‘I agree that my parents
will get divorced’), or deontic (e.g., Saku no naka ni hairanai koto o meejiru ‘I
order you not to go inside the fence’). This is a simplified frame for koto as a
comp, which highlights the part relevant to the interpretation of [S koto] as an
abbreviated expression. This is summarized in (18):
(18) Simplified frame for koto as a comp in [S1 koto S2]
– S1 represents an abstract or indirectly perceived state of affairs (P),
– S2, the main clause, may express one’s stance toward P in S1, a stance
that may be epistemic, evaluative/emotional, or deontic.
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.9 (213)
An utterance in the form of [S koto] (e.g. (16)) then serves as a kind of hint
for the implied meaning. On hearing such an utterance, the frame for koto as
a comp may be activated, which suggests that some kind of speaker’s stance
toward the proposition expressed in S is not explicitly mentioned (cf. Ohori
1997).6 Considering the proposition in S and the extralinguistic context, one
may then interpret the implied speaker’s stance as being deontic, in particu-
lar, directive. That is, the frame for koto ((18)) is coupled with the frame for
directives ((17)) by regarding the proposition in S as the after component
of the frame for directives and metonymically linking it to the core in the
same frame (see also Radden and Seto, this volume, for discussion of speech
act metonymy). At the initial stage of reanalysis, one’s knowledge of the spe-
cific context seems indispensable in that the implied meaning (i.e. the core in
the frame for directives in (17)) is inferable, only if there are sufficient contex-
tual clues for relating the utterance to the frame for an appropriate speech act.
For example, in the case of (16), the proposition in S can be considered the
after component of the frame for directives, only if the context fits the frame,
for example, if the context satisfies (some of) the conditions stipulated in the
before components in the frame for directives: if the speaker is the addressee’s
parent or teacher and is considered to have the authority of giving orders to
the addressee, if the action in question, brushing teeth every day, is considered
desirable, and if the addressee is assumed to be able to perform this action.
The points in (19) summarize the process of interpretation of [S koto] as an
abbreviated expression, conversationally implying the speaker’s directive stance:
(19) Summary of interpretive process:
– [S koto] as an abbreviated expression may be interpreted as a directive
on the basis of (a) knowledge of the frame for koto as a comp, (b)
knowledge of the frame for directives, and (c) knowledge of the specific
context.
– [S koto] activates the frame for koto as a comp, suggesting that the
speaker’s stance for the proposition expressed by S is implicit. Given
the context, the frame for directives may then be activated. That is, the
implicit speaker’s stance may be regarded as deontic, in particular, as
directive, because S in [S koto] corresponds to the after component of
the frame for directives, and because its context satisfies the conditions
in the before component of the same frame.
As we can see in (19), knowledge of frames is indispensable in the interpre-
tation of [S koto] as an abbreviated expression. It prompts the search for the
implicit meaning so that one can make sense of a given “incomplete” utter-
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.10 (214)
The reanalysis of the morpheme koto in Japanese discussed in this study illus-
trates the important role metonymy may play in grammaticalization at both
linguistic/rhetorical and cognitive levels. I have suggested that the reanalysis
of koto from a comp to an sfp is based on the use of [S koto] as a kind of
metonymic expression in that the whole is represented by its part. The mo-
tivations for the reanalysis, or the use of such a metonymic expression, seem
to be rhetorical and social concerns: (1) to foreground the information in the
“original” complement as the most important part of the message, (2) to bring
about certain expressiveness, that is, to perform a given speech act with partic-
ular stylistic nuances – i.e. spontaneity for the exclamatory koto rather than a
descriptive expression and conciseness for the directive koto, and (3) to use a
socially appropriate expression in a given situation – i.e. an indirect, less im-
posing expression of directive and a less imposing and “feminine” expression
of exclamation. Thus, the reanalysis of koto utilizes metonymy as an effective
linguistic device, or trope, that can satisfy these rhetorical and social needs (see
also Radden & Kövecses 1999).
Further, the process of reanalysis discussed above indicates that metonymy
is not simply a way of speaking, or a matter of trope. Rather, it illustrates an-
other important aspect of metonymy, namely, its role in thought processes
(Croft 1993; Panther & Thornburg 1998; Gibbs 1999; Radden & Kövecses
1999). I have argued that in the reanalysis of koto, conversational implicatures
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.12 (216)
Notes
* I would like to thank Klaus-Uwe Panther and Linda Thornburg for their valuable com-
ments and discussions.
. Note that the characterizations given here for no and koto have many “exceptions” the
discussion of which is beyond the scope of this study.
. Tte is a colloquial form of to.
. The following abbreviations are used in this paper: a = action, aux = auxiliary verb; comp
= complementizer; cop = copula; gen = genitive case marker; h = hearer; neg = negative
auxiliary; om = object marker; p = proposition; pass = passive voice; ppx = polite prefix; pst
= past tense marker; s = sentence; sfp = sentence-final particle; sm = subject marker; sp =
speaker; and tm = topic marker.
. Although not discussed here, the form tto is another variant of to and is used as an sfp
for making a declaration in a casual manner.
. See Okamoto (1995) for discussion of the morphemes no and to/tte.
. Ohori (1997) discusses what he calls suspended clause constructions in Japanese that
involve conjunctive particles (e.g. kara ‘because,’ noni ‘although’). Treating them as gram-
matical constructions, he accounts for their interpretation in terms of “framing effects,” in
which “the possible range of interpretation of the clause-linking form is constrained by the
constructional frame” (p. 476).
. Okamoto (1986) discusses the inferential process in the interpretation of “regular” ellip-
tical utterances in Japanese, which makes use of one’s knowledge of frames and that of the
specific context.
. The degree of reanalysis seems to differ depending on the morpheme; that is, some (pre-
viously) subordinate clauses (e.g. those with koto), but not others (e.g. those with node ‘be-
cause’), seem to have been fully established as main clause grammatical constructions with
“genuine” sfps. Further study on this issue will certainly enhance the understanding of the
process of reanlaysis concerning these grammatical morphemes.
References
Cook, Haruko M. (1990). An indexical account of the Japanese sentence-final particle no.
Discourse Processes, 13, 401–439.
Croft, William (1993). The role of domains in the interpretation of metaphors and
metonymies. Cognitive Linguistics, 4, 335–370.
Fillmore, Charles J. (1982). Frame semantics. In Linguistics Society of Korea (Ed.),
Linguistics in the Morning Calm (pp. 1–138). Seoul: Hanshin.
Fillmore, Charles J., Kay, Paul, & O’Connor, Mary C. (1988). Regularity and idiomaticity in
grammatical constructions: the case of let alone. Language, 64, 501–538.
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.15 (219)
Gibbs, Raymond W., Jr. (1999). Speaking and thinking with metonymy. In K.-U. Panther &
G. Radden (Eds.), Metonymy in Language and Thought (pp. 61–75). Amsterdam and
Philadelphia: John Benjamins.
Grice, H. Paul (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 41–58). New York: Academic Press.
Heine, Bernd, Claudi, Ulrike, & Hünnemeyer, Friederike (1991). Grammaticalization: A
Conceptual Framework. Chicago: The University of Chicago Press.
Hopper, Paul & Traugott, Elizabeth C. (1993). Grammaticalization. Cambridge: Cambridge
University Press.
Josephs, Lewis S. (1976). Complementation. In M. Shibatani (Ed.), Japanese Generative
Grammar [Syntax and Semantics 5] (pp. 307–369). New York: Academic Press.
Kay, Paul & Fillmore, Charles J. (1999). Grammatical constructions and linguistic gener-
alizations: The What’s X doing Y? construction. Language, 75, 1–33.
Kuno, Susumu (1973). The Structure of the Japanese Language. Cambridge, MA: The MIT
Press.
Lakoff, George (1987). Women, Fire, and Dangerous Things: What Categories Reveal About
the Mind. Chicago and London: The University of Chicago Press.
Martin, Samuel E. (1975). A Reference Grammar of Japanese. New Haven: Yale University
Press.
Maynard, Senko K. (1992). Cognitive and pragmatic messages of a syntactic choice: The case
of the Japanese commentary predicate n(o) da. Text, 12, 563–613.
Ohori, Toshio (1995). Remarks on suspended clauses: A contribution to Japanese
phraseology. In M. Shibatani & S. Thompson (Eds.), Essays in Semantics and
Pragmatics: In Honor of Charles J. Fillmore (pp. 201–218). Amsterdam and Philadelphia:
John Benjamins.
Ohori, Toshio (1997). Framing effects in Japanese non-final clauses: Toward an optimal
grammar-pragmatic interface. Berkeley Linguistics Society, 23, 471–480.
Okamoto, Shigeko (1985). Ellipsis in Japanese Discourse. Ph.D. dissertation. University of
California, Berkeley.
Okamoto, Shigeko (1995). Pragmaticization of meaning in some sentence-final particles in
Japanese. In M. Shibatani & S. Thompson (Eds.), Essays in Semantics and Pragmatics:
In Honor of Charles J. Fillmore (pp. 219–246). Amsterdam and Philadelphia: John
Benjamins.
Panther, Klaus-Uwe & Thornburg, Linda (1998). A cognitive approach to inferencing in
conversation. Journal of Pragmatics, 30, 755–769.
Panther, Klaus-Uwe & Thornburg, Linda (1999). The potentiality for actuality
metonymy in English and Hungarian. In K.-U. Panther & G. Radden (Eds.),
Metonymy in Language and Thought (pp. 333–357). Amsterdam and Philadelphia: John
Benjamins.
Radden, Günter & Kövecses, Zoltán (1999). Towards a theory of metonymy. In
K.-U. Panther & G. Radden (Eds.), Metonymy in Language and Thought (pp. 17–59).
Amsterdam and Philadelphia: John Benjamins.
Suzuki, Ryoko (1999). Grammaticalization in Japanese: A study of Pragmatic Particle-
ization. Ph.D. dissertation. University of California, Santa Barbara.
TSL[v.20020404] Prn:13/06/2003; 10:50 F: PB11309.tex / p.16 (220)
Teramura, Hideo (1984). Nihongo no Shintakusu to Imi II. (Japanese Syntax and Semantics
II). Tokyo: Kuroshio Shuppan.
Traugott, Elizabeth C. (1988). Pragmatic strengthening and grammaticalization. Berkeley
Linguistics Society, 14, 406–416.
Traugott, Elizabeth C. (1989). On the rise of epistemic meanings in English: An example of
subjectification in semantic change. Language, 65, 31–55.
Traugott, Elizabeth C. & König, Ekkehard (1991). The semantics-pragmatics of gram-
maticalization revisited. In E. Traugott & B. Heine (Eds.), Approaches to Grammat-
icalization (I) (pp. 189–218). Amsterdam and Philadelphia: John Benjamins.
TSL[v.20020404] Prn:4/06/2003; 15:33 F: PB113P4.tex / p.1 (221)
P IV
. Introduction
(see Gibbs 1986, 1994: 351–357; Thornburg & Panther 1997; Panther & Thorn-
burg 1999). Since have- and be-languages construe the notion of ‘possession’
differently, the metonymies linking the indirect wording to the intended speech
act meaning will also be different, i.e. different speech communities make use
of different metonymic reasoning in coding and understanding a request.
This paper investigates the successive stages of the shopping scenario in
which the notion of possession is relevant and compares the ways a speaker
of a have-language metonymically asks for goods in a shop as opposed to a
speaker of a be-language. We will first present contrastive data of have- and be-
languages on the metonymic construal of requests in a prototypical shopping
scenario and then discuss these findings.
Typically, only the precondition and one of the transaction stages are expressed
in communicating a shopping request. We will first look at the ways the pre-
condition of a shopping request, i.e. the article’s availability, is conceptualized
and will find that have- and be-languages make use of different metonymies
in construing the concept of ‘availability’ (Section 2.1). We will then examine
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.3 (225)
the ways in which one of the three successive subevents of a commercial trans-
action may be metonymically utilized in have- and be-languages to stand for
the commercial transaction as a whole (Sections 2.2-2.4)
Asking such a question only makes sense in a situation in which the customer
may reasonably suspect that the item is no longer available. In sentence (2),
such a situation might arise at the moment the fast-food joint is closing.
The article’s availability is an essential precondition for its purchase. It ei-
ther needs to be explicitly asked about by the customer or is taken for granted.
The customer cannot, as a rule, jump into the buying phase of a commer-
cial event unless s/he feels sure that the article is available. Thus, if 40-watt
light bulbs are usually only sold at electrical appliance stores, the customer
will not directly ask for one at a gas station without first inquiring whether
they are available. In this situation, a request such as Can I have a 40-watt light
bulb? or 40 watto no denkyuu o kudasai (lit.: ‘Give me a 40-watt light bulb’) is
pragmatically inappropriate.
The transfer of the article by the salesperson establishes the central subevent
of the commercial transaction requested by the customer. In terms of obstacle
theory, this phase represents the second potential obstacle to be overcome. The
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.5 (227)
In the East-Asian be-languages Japanese, Chinese and Korean, asking the sales-
person to “give” the customer an article is not considered offensive – it is, in
fact, the most neutral way of expressing a shopping request as illustrated in the
Japanese sentence (5):5
(5) Asahi shinbun (o) kudasai.
Asahi newspaper obj give.hon
Lit.: ‘Give the Asashi paper!’
‘Can I have the Asashi, please?’
The transaction of an article is only successful if the customer receives the ar-
ticle bought. This final stage of the transaction represents an achievement in
Vendler’s (1967) typology of situation types: it describes the non-volitional ter-
mination of an event. An achievement verb may often be used metonymically
to stand for an action leading to its achievement as in I am catching fish, where
the punctual achievement verb to catch is used in the dynamic sense of ‘trying
to catch.’ In the shopping scenario, the achievement of the buyer’s reception
of the article may be used to stand for its transaction by the salesperson, i.e.
the buyer expresses his wish to be given an article by means of the metonymy
reception for transaction. This metonymy is conventionally used in some
have-languages like German (7a) as well as in many be-languages like Japanese
(7b), Polish (7c), Chinese, Hungarian and Finnish and, in special situations,
also in Korean.7
(7) a. Ich bekomme zwei Kilo Tomaten.
I receive two kilo tomato-pl
‘Can I have two kilos of tomatoes?’
b. Tomato o ni kilo itadaki-masu/moraimasu.
tomato obj two kilo receive-hon
Lit.: ‘I’ll receive two kilos of tomatoes.’
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.7 (229)
The reception for transaction metonymy has the effect of mitigating the
face threat of the request, which may, amongst other things, be further alle-
viated by the use of a modal verb and the question form as in (7c). The in-
directness conveyed by this metonymy accounts for its widespread use in the
shopping situation. In Japanese, itadakimasu or moraimasu as in (7b) are the
conventional forms used by a customer to express his or her shopping request.
In other languages, the use of the reception phase for the article’s transaction
is pragmatically inappropriate. This applies to the have-languages English (cf.
#I’ll get/receive two kilos of tomatoes), Croatian and Lithuanian.
As a result of a commercial transaction, the article bought passes into the cus-
tomer’s possession. This future state of an article’s possession may, at least in
some languages, metonymically stand for its requested transaction. A have-
language that conventionally uses the metonymy possession for transac-
tion is English. It applies to situations that are mainly restricted to the order-
ing of food or drinks in a restaurant such as (8a), i.e. to non-permanent pos-
sessions. Of the be-languages considered, only Hungarian allows the speaker
to order food or drinks by metonymically referring to the resulting state as il-
lustrated in sentence (8b), which might be said in the situation in which each
member of a group places their order to a waiter. In accordance with its sta-
tus as a be-language, Hungarian construes such a request by means of the
metonymy existence for transaction.
(8) a. I’ll have a beer.
b. Nekem egy sör lesz.
‘me a beer become’
Most of the have- and be-languages studied do not permit either of these
metonymies. This may be because, in the chain of stages in the shopping sce-
nario, the resulting state is one step further removed from the central subevent
of transfer than the before-mentioned reception stage.
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.8 (230)
. Summary
The metonymic construals of shopping requests used in the eleven have- and
be-languages selected for this study are listed in Table 1.
Even if the number of languages analyzed is too small to claim any ty-
pological generalizations, the comparative results allow us to discern certain
cross-linguistic tendencies of metonymic construal. We can note the following
observations, which will be discussed in Section 3.
First, as should be expected from their typological status, have- and be-
languages use their own metonymic construals to express (i) availability and,
to a lesser extent, (ii.c) the result of a requested transaction as ‘possession’ and
‘existence,’ respectively.
Secondly, the metonymic construal of the transfer stage (ii.a), i.e. the possi-
bility of directly asking the salesperson to “give” the article wanted, is avoided in
most European languages but commonly used in the East-Asian be-languages
Japanese and Chinese and, to a lesser extent, in Korean.
Thirdly, the metonymic construal of the reception stage (ii.b) is rare in
most European languages but commonly found in the three East-Asian be-
languages as well as some European languages.
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.9 (231)
. Discussion
The ensuing discussion will attempt to find cognitive and cultural explanations
for the three phenomena observed. It is claimed that at least some of the struc-
tural differences discovered are not just arbitrary phenomena of language but
reflect conceptual and possibly cultural differences.
We will first look at the notions of ‘possession’ and ‘existence,’ which,
amongst other things, account for the different construals of availability. We
will next consider two forms of politeness, indirectness and deference, which
account for the absence and use of direct forms of request. Lastly, we will look
at the notions of action vs. process, which might account for the differences
found with respect to the metonymization of the reception stage.
have-languages like English pick out the possessor both as the theme and the
subject of the sentence and, thus, give prominence to the human. This is in con-
formity with many other areas in which have-languages, unlike be-languages,
focus on the human.8 be-languages such as Japanese may topicalize the posses-
sor as in (9b), but do not subjectize it. be-languages thus downplay the human
element and present the relationship between the two entities as a contiguity
relation, where the subject (the children) describes something that exists and
the complement (John) describes something in relation to which the subject’s
existence is predicated. The Location Schema prototypically applies to the spa-
tial location of things, but it also applies to the existence of things in the sense
of availability.
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.10 (232)
This poses the question of what counts as personal possession in Japanese. For
example, stamps do, so that I may use motsu in asking a friend to help me out
with stamps as in 50 yen kitte 10 mai motte-masu ka (‘Do you have ten 50-yen
stamps?’). However, due to their temporary nature, hamburgers are not con-
sidered personal possessions and, therefore, do not go well with motsu: *Big
Mac motte-masu ka (‘Do you have a Big Mac?’). Possessions may also be ab-
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.11 (233)
stract things such as interest or expenses, which may be ‘had’: kyoomi-o motsu
‘have interest’ and hiyo-o motsu ‘cover the expenses.’ Also money may be pos-
sessed but is more likely to be seen as existing: for example, ‘Do you have some
money?’ is rendered as ikuraka okane aru (Lit. ‘Some money be/exist?’). Things
that cannot be possessed are humans: thus, it is impossible to say ‘I have two
children’ instead of (9b), i.e. children are not regarded as personal possessions
in Japanese.
The notions of ‘possession’ and ‘existence’ are to be seen as forming a
conceptual continuum, which different languages may cut up differently. In
Japanese, only prototypical physical objects and abstract things can be pos-
sessed – these are things that can be controlled. Humans, transitory objects
including money, objects that are available but exist independently of us, and
objects in space cannot be possessed – they are only seen in a contiguity
relation to us.
Other languages may make different distinctions. The be-language Polish,
for example, uses the Existence Schema in questions about an article’s availabil-
ity but the Possession Schema in negated replies, i.e. something that is available
“exists” as in (11a), whereas something that is not available is “had” as in (11b):
(11) a. Czy jest duńskie masło?
q is Danish butter
Lit.: ‘Is (there) Danish butter?’
‘Do you have Danish butter?’
b. Nie ma.
not have
‘We don’t have any.’
The Possession Schema also takes over in shopping situations in which a cus-
tomer regularly buys a certain product at a certain shop. For example, the
customer may ask for his regular brand of beer such as EB by using a ‘have’-
question: Pani ma EB? (Lit.: Mrs. have EB?, ‘Do you have EB?’). The relation-
ship between the customer and the shop owner has become a personal one,
and the commercial event appears like an exchange of possessions.10
The reverse situation holds in have-languages, such as Lithuanian.11 A cus-
tomer asking a salesperson whether a specific item is available in the shop will
ask a ‘have’-question as in (12a), but a third person asking the shopper if the
item is available in the shop will use a ‘be’-question as in (12b):
(12) a. Ar turite 40 vatu˛ lempučiu˛?
q have-2p.pl 40 watt lamp-gen.dimin
‘Do you have 40-watt bulbs?’
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.12 (234)
In asking question (12b), the third person takes a distanced view of the shop-
ping scenario: the speaker’s attention is directed towards the existence of the
article in the shop, and the possessive relationship between the shop and the
article is out of focus.
A major difference between western and eastern cultures pertains to the ways
a person is attended to. Speakers of western languages tend to mitigate face-
threatening acts by using strategies of indirectness. A direct request such as
(3a) #Give me “The Times”! is felt to be rude in English and is therefore avoided
in polite interaction. The Japanese equivalent (5) Asahi shinbun (o) kudasai and
those of Chinese and Korean show, however, that direct requests are the nor-
mal forms used in speaking to salespersons. However, it would be misleading
to consider Japanese and English forms of request from a structural point of
view only.
As convincingly argued by Matsumoto (1988), the notion of ‘politeness’
and the linguistic strategies of politeness employed by a speaker are culture-
specific and fundamentally different in Western and Japanese cultures. The
Western notion of ‘politeness’ is based on the individual’s public self-image,
and impositions on an individual’s face are minimized by means of redressive
strategies such as conventional forms of indirectness. The Japanese notion of
‘politeness’ is based on a person’s position in society, and forms of politeness
are used to show the speaker’s deference to the supposedly higher-ranking ad-
dressee. The most important “relation-acknowledging devices” are honorifics,
i.e. conventional lexical or morphological forms by means of which the speaker
exalts the addressee and humbles himself or herself. Interestingly, salespersons,
who are of a much lower social rank than customers, are also spoken to in
Japanese, Chinese and Korean by using honorific forms.
The deferential aspect of honorifics can be seen in the etymology of some
Japanese honorific words. Kudasai in sentence (5) Asahi shinbun (o) kudasai
has as its bare form the honorific word kudasaru, which is etymologically re-
lated to kudaru ‘go down’ and suggests passing a favor down to an inferior
person. Thus, in using the exalting form kudasai, the Japanese speaker used
to express respect to a higher-ranking person. The same applies to the word
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.13 (235)
Section 2.3 showed that some languages, in particular the three East-Asian lan-
guages, allow the speaker to refer to the requested transaction by metonymi-
cally highlighting the reception stage of the shopping scenario. We will look
again at Japanese, where shopping requests are typically expressed as in (7b)
Tomato o ni kilo itadaki-masu (‘I’ll receive 2 kilos of tomatoes’). The metonymy
has the effect of focusing away from the agent’s action and viewing the event
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.14 (236)
. Conclusions
This study investigated the ways requests in an everyday situation, the shopping
scenario, are coded in different languages. Such requests are typically construed
metonymically, where the different stages of a commercial scenario may serve
as metonymic vehicles. These are, in particular, the availability of the article, the
transfer of the article by the salesperson, the reception of the article by the cus-
tomer, and the resulting possession of the article by the customer. The choice of
metonymies was shown to depend, amongst other things, on typological prop-
erties of the given language. The two types of languages distinguished for this
purpose are have- and be-languages, the former being typically represented
by English, the latter by Japanese. have-languages metonymically express the
notions of availability and, to a lesser extent, that of requested transaction, as
possession, be-languages construe these notions as existence. It has been ar-
gued that the notions of possession and existence form a conceptual contin-
uum, which is cut up differently by different languages and thus also accounts
for different metonymic usages.
have- and be-languages also tend to display different metonymic usages
with respect to the transfer and reception stage of the commercial transaction.
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.15 (237)
The use of a direct request is felt to be impolite in Western cultures but repre-
sents, in conjunction with honorifics as expressions of deference, the normal
form of shopping requests in East-Asian languages. The different cultural sys-
tems of politeness – indirectness vs. deference – account for the absence or
presence of the transfer for transaction metonymy. The metonymic use
of the reception stage for a requested transaction in East-Asian languages may
be relatable to culture-specific ways of viewing events: do-languages such as
English focus on actions and their results, become-languages such as Japanese
focus on processes as happening.
Notes
. In Heine’s typology, possessive have is subsumed under the Action Schema X takes Y,
since possessive verbs of ‘having’ derive from earlier meanings of ‘seize,’ ‘hold’ and the like.
. The terms ‘have-language’ and ‘be-language’ are used by Ikegami (1991) in his analysis
of representational differences between English and Japanese.
. A selection of be-languages that render sentence (1a) in a fashion similar to Japanese
in (1b) is listed below. We would like to express our thanks to Changhong Sui, Koo Izen,
Jae Jung Song, Jeong-Hwa Lee, Aila Radden, Karol Janicki, Elżbieta Tabakowska, Vitalija
Liutvinskiene, Joe McIntyre, Rita Brdar-Szabó and Mario Brdar for providing data on their
native languages.
Contrastive situations that might render the sentence acceptable are, for example, those of a
customer who wants twó kilos of tomatoes, not thrée, or two kilos of tomátoes, not potátoes,
or of a customer who, after resisting buying tomatoes, finally accepts. The buyer will then
introduce the sentence with the discourse response marker kulem ‘so, then,’ which is set off
by a pause, indicated here by a comma:
. Some of Ikegami’s pairs of examples in which English emphasizes the human where
Japanese presents the situation as thing-like or event-like include the following: English I
have a temperature corresponds to Japanese ‘temperature is,’ English John ran out of money
is rendered in Japanese as ‘(As for John), money became null,’ English I don’t understand you
is expressed in Japanese as ‘I don’t understand what you say,’ etc.
. Wordings such as (1a) in fact involve a further metonymy: it is not the salesperson who
possesses the items but the store, i.e. the person is used to stand for the institution. The
metonymy person for institution is motivated by a general principle of cognitive salience
(see Radden & Kövecses 1999): humans are in general more salient than institutions, and
entities we interact with, i.e. salespersons, are more salient than entities we do not inter-
act with, i.e. the shop. This does not, however, apply to the Japanese view of the world:
Japanese does not extend humans to institutions. Ikegami (1991: 301) nicely observed that
the notice We are closed today on the door of a shop would strike a Japanese speaker as odd.
Hence, the goods that are for sale at a store are neither seen as possessions of the store nor
metonymically as possessions of the salesperson but simply as existing in contiguity to the
store.
. We are indebted to Elżbieta Tabakowska for the Polish data. The situation is, in fact,
more complex. The Existence Schema is associated with the standardized shopping scenario
and expectations derived from it, while the Possession Schema tends to be associated with
negative expectations. Thus, a Polish customer may no longer expect to get rolls at a bakery
TSL[v.20020404] Prn:13/06/2003; 13:55 F: PB11310.tex / p.17 (239)
near closing time and ask ‘Do you still have bread?’ rather than ‘Is there still bread?,’ or
he may not expect to find a specific book in a bookstore and, therefore, form the Polish
question as ‘Do you have books by Shakespeare?’ and not ‘Are there books by Shakespeare?’
If the Possession Schema is used in situations that normally require the Location Schema as
in ‘Do you have beer?’ asked at a supermarket, the resulting meaning of counter-expectation
is that of the beer being sold illegally.
. We owe this interesting observation to Vitalija Liutvinskiene.
. We thank Yoshihiko Ikegami for providing this delightful example of Japanese polite-
ness.
References
Brown, Penelope & Levinson, Stephen (1987). Politeness: Some Universals in Language Usage.
Cambridge: Cambridge University Press.
Francik, Ellen P. & Clark, Herbert H. (1985). How to make requests that overcome obstacles
to compliance. Journal of Memory and Language, 24, 560–468.
Gibbs, Raymond W., Jr. (1986). What makes some indirect speech acts conventional? Journal
of Memory and Language, 25, 181–196.
Gibbs, Raymond W., Jr. (1994). The Poetics of Mind: Figurative Thought, Language and
Understanding. Cambridge: Cambridge University Press.
Heine, Bernd (1997). Cognitive Foundations of Grammar. New York, Oxford: Oxford
University Press.
Ikegami, Yoshihiko (1991). ‘do-language’ and ‘become-language’: Two contrasting types of
linguistic representation. In Y. Ikegami (Ed.), The Empire of Signs: Semiotic Essays on
Japanese Culture (pp. 285–326). Amsterdam and Philadelphia: Benjamins.
Matsumoto, Yoshiko (1988). Reexamination of the universality of face: Politeness
phenomena in Japanese. Journal of Pragmatics, 12, 403–426.
Panther, Klaus-Uwe & Thornburg, Linda (1999). The potentiality for actuality
metonymy in English and Hungarian. In K.-U. Panther & G. Radden (Eds.), Metonymy
in Language and Thought (pp. 333–357). Amsterdam and Philadelphia: Benjamins.
Radden, Günter & Kövecses, Zoltán (1999). Towards a theory of metonymy. In
K.-U. Panther & G. Radden (Eds.), Metonymy in Language and Thought (pp. 17–59).
Amsterdam and Philadelphia: Benjamins.
Thornburg, Linda & Panther, Klaus-Uwe (1997). Speech act metonymies. In W. A. Liebert et
al. (Eds.), Discourse and Perspectives in Cognitive Linguistics (pp. 205–219). Amsterdam
and Philadelphia: Benjamins.
Vendler, Zeno (1967). Linguistics in Philosophy. Ithaca, NY: Cornell University Press.
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.1 (241)
. Introduction
In (2) the noun phrase conference is used to refer to only one essential ingredi-
ent of a conference, viz. its participants.
In predicational metonymy one propositional content stands for another
propositional content. Assuming that the locus of metonymic mappings in
the examples listed in (1) above is indeed the predicatively used adjective, we
may classify them as predicational metonymies. It need not, however, be self-
evident that these are predicational metonymies, particularly in view of the
widely held traditional view that metonymy resides in the realm of nominals
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.4 (244)
(which is the reason why referential metonymies can perhaps also be called
nominal metonymies).
We shall reexamine our examples in light of the possibility that some can
perhaps be interpreted as referential or nominal metonymies, but let us now
review some evidence pointing in the direction of the conclusion that we are
dealing with predicational metonymies in the set of examples in (1) above.
The constructions we focus on here are not a spectacular type of metonymy,
and, just like many other types, are quite easy to overlook. However, a closer
look will reveal a number of relevant features. All the adjectives in the above
examples seem to specify one aspect of the linguistic action involved, viz. the
way in which it was performed, carrying more or less strong expectations as to
the effectiveness and ultimate result of the linguistic action, or the lack thereof.
That they denote the manner in which an implicit linguistic action is carried
out becomes obvious from paraphrases such as:
(3) a. I must speak openly with her, whatever the cost.
b. The President spoke clearly on the matter.
c. Livingstone, even when speaking humorously about a very distressing
period, could not disguise the discomfort.
Of course, it must be admitted that not all examples readily allow this type of
paraphrase. In some cases it is stylistically clumsy, while in other cases the con-
text also supports a paraphrase in which a mental predicate (verb or a complex
verbo-nominal expression, e.g. think, have an opinion, etc.) is used.
Notice that (1e), repeated here as (4a), actually has an explicit verb denot-
ing linguistic action in the prepositional complement following the predicative
adjective:
(4) a. The buyers were emphatic in declaring that they were right.
b. I need to call the garage (where my car is being serviced). They said
they’d have it ready by five o’clock.
Just like paraphrases, this also seems to indicate that the type of constructions
we are interested in here indeed involve predicational metonymies.
The set of predicative adjectives that appear in this construction includes
the following:
(7) accurate, articulate, baroque, bitter, blunt, boring, brief, bullish, clear,
coherent, cynical, definite, direct, dramatic, earnest, emphatic, explicit,
harsh, entertaining, factual, firm, forthright, frank, lukewarm, lyrical,
mum, poetic, open, pompous, precise, sarcastic, serious, short, silent,
specific, vague . . .
Croatian
(8) a. Moram biti otvoren s njom, po svaku cijenu.
must-1sg be open with her at any price
‘I must be open with her, whatever the cost’
a. Moram otvoreno porazgovarati s njom, po svaku cijenu
must-1sg openly speak with her at any price.
‘I must speak openly with her, whatever the cost’
b. Dragi kolege, bit ću kratak.
dear colleagues be will-1sg brief
‘Dear colleagues, I’ll be brief ’
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.7 (247)
Hungarian
(9) a. Nyíltan kell vele beszélnem, kerüljön, amibe kerül.
open must with-her speak. . .
‘I must speak openly with her, whatever the cost’
a. *Nyíltnak kell vele lennem.
open must with-her be
‘I must be open with her, whatever the cost’
b. Kedves kollégák, rövid leszek.
Dear colleagues, brief be-fut-1sg
‘Dear colleagues, I’ll be brief ’
b. . . . , ígérem, rövid leszek.
promise brief be-fut-1sg
‘I promise to be brief ’
c. Arthur visszaemlékezéseiben röviden említette a többi
Arthur recollections-poss-in briefly mentioned def other
tanárt.
teacher
‘Arthur mentioned briefly in his recollections other teachers’
c. *?Arthur rövid volt a többi tanárral kapcsolatban.
Arthur brief was def other teacher concerning
‘Concerning other teachers, Arthur was brief ’
c. *?Arthur rövid volt.
Arthur brief was
‘Arthur was brief ’
d. Az elnök világosan nyilatkozott / szólt ezzel az
def president clearly spoke spoke this-with def
üggyel kapcsolatban.
matter concerning
‘The President spoke clearly on that matter’
d. *Az elnök világos volt.
def President clear was
‘The President was clear’
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.9 (249)
A number of interesting contrastive facts emerge from these data. First of all,
we see that Croatian and Hungarian appear reluctant to make use of this type
of predication metonymy. This ties in nicely with the findings by Panther and
Thornburg, who report that another predication metonymy that seems to be
subject to restrictions of typological nature, potentiality for actuality, is
systematically blocked or only weakly exploited in Hungarian. We see that in
our Croatian and Hungarian examples, constructions with predicative adjec-
tives can be used very rarely, and they seem to be licensed only if the adjective
takes no further complement. Finally, Croatian and Hungarian favor the ex-
plicit mention of the linguistic action in the verbal part of the predicate, the
counterparts of English predicative adjectives are rendered in both languages
as adverbials of manner, phrasal or clausal. We also note the explicit mention
of the verb of linguistic action in temporal adverbial clauses in Croatian (cf.
(8d and f ). The fact that Croatian and Hungarian tend to explicitly mention
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.10 (250)
the linguistic actions and the tendency to render English adjectives as adverbs
of manner, lends further indirect support to the view that we are dealing here
with a predicational metonymy of the manner for (linguistic) action type.
It is significant that referential metonymy, however, does not seem to be
constrained in either Croatian or Hungarian in such a systematic way. Some ex-
amples, or concrete metonymies in Blank’s terms (1999: 183), may be culture-
specific and therefore lack metonymic counterparts in other languages, but,
generally, at the level of types of contiguity, i.e. in terms of image schemas,
there is a fairly close correspondence between English on the one hand, and
Croatian and Hungarian on the other. Compare some examples:
(10) a. Beijing’s difficulties in Tibet boil down to the Chinese leadership’s
relations with one man . . .
Croatian
b. teškoće Pekinga u Tibetu svode se na odnose
difficulties Beijing-gen in Tibet boil-down refl on relations
kineskog vodstva s jednom osobom . . .
Chinese-gen leadership-gen with one person
‘Beijing’s difficulties in Tibet boil down to the Chinese leadership’s
relations with one man . . . ’
Hungarian
c. . . . Peking Tibettel kapcsolatos nehézségei a kínai
Beijing Tibet-with concerning difficulties def Chinese
vezetésnek egy személlyel való viszonyára
leadership-poss one person-with concerning relations-to
vezethetők vissza . . .
relatable back
‘. . . Beijing’s difficulties in Tibet boil down to the Chinese leadership’s
relations with one man . . . ’
(11) a. Not even the great brains of Cambridge could solve his problem.
Croatian
b. Niti veliki mozgovi iz Cambridgea nisu mogli
Not-even great brains from Cambridge neg-aux could
riješiti njegov problem.
solve his problem
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.11 (251)
Hungarian
c. Még a nagy cambridgei koponyák sem tudták az ő
even def great Cambridge skulls neg could def his
problémáját megoldani.
problem-poss-acc solve
(12) a. Let’s have another glass.
Croatian
b. Popijmo još jednu čašu.
drink-imp yet one glass
Hungarian
c. Igyunk még egy pohárral.
Drink-imp yet one glass
(13) a. Dad used Scotch tape to piece together the torn-up photograph.
Croatian
b. Tata je koristio selotejp da sastavi poderanu sliku.
Daddy aux used tape-acc that fix torn-up photo
Hungarian
c. Papa cellux-szal ragasztotta meg az eltépett képet.
Daddy tape-with fixed pref def torn-up photo
These interesting crosslinguistic regularities seem to indicate that the distinc-
tion between referential, predicational and illocutionary metonymies may be
an important parameter in establishing a typology of metonymies. The dif-
ferences in the distribution of referential and predicational metonymies in the
three languages investigated here could lead us as far as to suppose that there
may perhaps obtain a sort of implicational relationship between the referential
and the predication type of metonymy, predicting that a language that makes
extensive use of the latter will also make heavy use of the former, while there
will be languages that will restrict themselves to referential metonymic mod-
els only. This suggestion is a far cry from postulating an implicational uni-
versal, it is rather to be understood as an invitation to a systematic study of
the relationship between various types of metonymies in as many languages
as possible. The evidence available at present seems to indicate (i) that the
pattern we present here is found repeatedly, across languages and across do-
mains, and (ii) that the kind of constraints on the two types of metonymies are
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.12 (252)
very different in nature (cf. Brdar-Szabó & Brdar 2001; Brdar-Szabó & Brdar
2002; Brdar-Szabó 2002; and Brdar, Brdar-Szabó, Gradečak-Erdeljić, & Buljan,
in press, where metonymies involving three cognitive domains are studied in
eight languages).
However, regardless of whether the observed crosslinguistic differences can
be shown to be more universal or not, the fact remains that Croatian and
Hungarian behave very differently from English, and that some explanation
has to be offered. In the remainder of our paper, we shall try, making use of
both synchronic and diachronic data, to show that these differences are the
result of an intricate interplay of cognitive and pragmatic, lexical and mor-
phosyntactic factors.
The few examples in Croatian and Hungarian in (1) where we apparently have
acceptable metonymic counterparts, can in fact be interpreted as referential
and not as predicational metonymies, i.e. they lend themselves to an analysis
in which the subject, which in our examples always denotes a person, i.e. the
speaker, stands for his/her utterance:
(14) I’ll be brief. (‘My speech/words, etc. will be brief ’)
1593 1797
blunt
1724 1848
coherent
Figure 1. Shifts from referential to predicational metonymy with blunt and coherent
numbers refer to the year in which a given use is first recorded in the OED;
the lighter shaded area denotes the time period in which adjectives collocated
with subject NPs that could be interpreted as referential metonymies, while the
darker shaded area denotes the period in which the adjective is used metonymi-
cally to code linguistic action. The two periods may overlap a great deal, so that
the referential metonymy continues in parallel with the predicational one, or
may perhaps sometimes be discrete, but as our main point is to demonstrate
that the referential metonymy temporally precedes the predicational one, the
issue of their overlap is of secondary importance here.
(15) a. 1593 Shakes. 3 Henry VI, v. i. 86 Trowest thou that Clarence is so
harsh, blunt, unnatural.
b. 1797 Mrs. Radcliffe Italian xiii. (1824) 606 Be pretty blunt with them
if they want to come in here.
c. Trollope Belton Est. iii. 27 He was blunt in his bearing, saying things
which her father would have called indelicate and heartless.
(16) a. 1724 Watts Logic iii. iv. 1 A coherent thinker, and a strict reasoner, is
not to be made at once by a set of rules.
b. 1848 Dickens Dombey 51 Be plain and coherent, if you please.
The data are sparse and not easy to come by, but it is significant that, while we
can, as yet, admittedly adduce only a few clear-cut cases, we have not been able
to find adjectives whose development would contradict our stipulations.
A similar proposal is put forward by Waltereit (1998: 63ff., 119ff.; 1999:
235f.), who, discussing verbal valency on French material, distinguishes be-
tween so-called insertional and role-level contiguities. He claims that the latter
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.14 (254)
the adjective kratak ‘short/brief ’, as well as its Hungarian counterpart rövid, are
semantically compatible with subjects denoting body parts, and in Croatian, to
a degree, compatible with subjects denoting humans, but the use of the future
tense (or the past tense) rules out the literal interpretation because being short
is not a quality that can be brought under the subject’s control and thus does
not obtain for a period of time at will.
On the other hand, in predicational metonymies of the manner for lin-
guistic action type we frequently find lexicalization and polysemy of adjec-
tival predicates. A selection of sense descriptions for a few adjectives in some
pedagogical monolingual dictionaries of English is given below:
(17) a. articulate
CCELD: ‘if you are articulate, you are able to express yourself easily
and well, especially when you are dealing with difficult ideas’
LDoCE: ‘expressing or able to express thoughts and feelings clearly,
esp. in words’
OALDoCE: ‘(of a person) able to put thoughts and feelings into clear
speech’
CIDE: ‘able to express, or expressing thoughts and feelings easily
and clearly’
b. blunt
CCELD: ‘when someone is being blunt, they are speaking directly and
simply without making any effort to be polite or to avoid
upsetting people’
LDoCE: ‘(of a person) speaking roughly and plainly, without trying
to be polite or kind’
OALDoCE: ‘(of a person, what he says) plain, not troubling to be polite’
CIDE: ‘saying what you think without trying to be polite or caring
for other people’s feelings’
c. brief
CCELD: ‘a piece of writing or speech that is brief, does not contain
to many words or details; used of persons’; ‘someone who
is brief when talking to another person does not say much
because they do not really want to speak to that person or
discuss that subject’
LDoCE: ‘to speak shortly’
OALDoCE: ‘lasting only for a short time or containing few words’
CIDE: ‘lasting only a short time or containing few words’
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.16 (256)
The formulations listed above allow us to safely conclude that the idea of
linguistic action is now conventionalized to such a degree that it is incorpo-
rated into one (or more) of their sense(s). Consequently, a productive ad hoc
metonymic reasoning based on an inferential model is now rendered superflu-
ous as the targeted meaning is now part of linguistic knowledge.
We do not want to claim that all the adjectives used in predicational
metonymies have to be polysemous in the way just exemplified above. Some
can still be open to metonymic inferencing, and some can be used in contexts
supporting both the interpretation of the sentence as a referential metonymy
and as a predicational one. In (6) above we have such an example where
the predicative adjectives explicit and coherent can be interpreted as predica-
tional metonymies, but simultaneously the NP Bloom (1993) is interpreted as
a referential metonymy.
Such co-existence of the two types of metonymies may be supposed to
have provided a bridge for the gradual development and spread of predica-
tional metonymies, and concomitant polysemy. We would like to hypothesize,
however, that this polysemy and the switch from referential to predicational
metonymy were facilitated by some structural (and semantic) facts about the
grammatical constructions in which they occurred. Consider the effect of the
presence viz. absence of the complement of the adjective in the following set
of examples:
(18) a. Our boss was vague.
b. Our boss was vague about when the pay rise was due.
(19) a. I was quite frank about it.
b. You’re not being frank with me, Mademoiselle.
c. I’ll be quite frank . . .
as well as Brdar-Szabó & Brdar 2002; Brdar-Szabó 2002; and Brdar, Brdar-
Szabó, Gradečak-Erdeljić, & Buljan, in press). While we do not have enough
diachronic data showing how the rise of specific types of adjective comple-
mentation ties in with the development of metonymic models and polysemy
in English, we can nevertheless synchronically compare English in this respect
with Croatian and Hungarian, which will be shown in the last section to be
different enough.
There are profound differences between English and the other two languages
concerning both the basic ascriptive construction and the extended one that
includes the complementation patterns of adjectives. It appears that the copu-
lar complementation pattern is less pervasive in Croatian and Hungarian than
in English, and particularly so in the case of the extended construction with a
pp as a complement of an adjective.
A number of contrastive studies (Ivir 1983; Brdar 1994) report that many
English predicative adjectives taking various complements (but also without
complements) do not find their Croatian and Hungarian counterparts in ad-
jectives but rather in verbal predicates (which is also in keeping with the less
analytic, i.e. more dynamic, typological preferences of the latter languages).
(20) a. Jack was silent.
b. Jack je šutio.
Jack aux silent-verb-past:3sg
‘Jack was silent’
c. Jack hallgatott.
Jack silent-verb-past:3sg
‘Jack was silent’
(21) a. . . . he was greatly afraid of Livia and at first wholly dependent on her
...
b. . . . jako se bojao Livije te je isprva potpuno
very refl afraid-verb . . . aux at-first wholly
ovisio o njoj/bio o njoj ovisan . . .
depended on her aux on her dependent
‘He was greatly afraid of Livia and at first wholly dependent on her’
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.18 (258)
Note that Croatian and Hungarian in such cases almost regularly fall back on
verbal predicates, or at least have them as alternative renderings of (22b).
Basically the same situation obtains in extended ascriptive constructions
with adjectives that are not derived from verbs. The extended ascriptive con-
struction itself coerces an event interpretation for which a suitable verb is
supplied (cf. Panther & Thornburg 1999a). Its gradual conventionalization
resulted in polysemy.
This more basic and general predication metonymy and the resultant
grammatical construction may have thus played a crucial role in the lexical-
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.19 (259)
We intuitively expect that linguistic actions are ontologically dependent on, i.e.
secondary, to mental ones, but with some effort, even the inference pointing
towards mental action could be perhaps cancelled.
Some adjectives in non-extended ascriptive constructions will also allow
an interpretation on which the expression was non-linguistic. Hence, the lin-
guistic action inference is again easily cancelable:
(24) He was emphatic but didn’t say anything.
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.20 (260)
On the other hand, we have cases of extended constructions in which the com-
plement simply coerces the linguistic action interpretation, and the inference
is non-defeasible.
(25) a. *Arthur was brief about his other teachers in his recollections but he
didn’t say anything.
b. *The buyers were emphatic in declaring that they were right, but they
didn’t say anything.
The above facts, we think, clearly show the role of structural factors, viz. gram-
matical constructions, in guiding this type of metonymic inferences in English,
from conversational implicatures towards conventional ones.
It is quite common for English predicative adjectives taking prepositional
phrases as complements to have Croatian and Hungarian counterparts that
are predicative adjectives but do not take pps as complements but rather nps
in various cases. In other words, pps are not extensively used as complements
of predicative adjectives in general in these two languages. More specifically,
English adjectival predicates taking pps as complements will regularly exhibit
verbal counterparts in Croatian and Hungarian, at least as one of the possibil-
ities, if not the only possibility. This of course also applies to our set of English
adjectives exhibiting the manner for linguistic action metonymy and their
counterparts in Croatian and Hungarian. We further find, as shown in (8) and
(9), that many of the pps that can perhaps follow adjectives in Croatian (occa-
sionally in Hungarian) are rather peripheral elements, adjunct-like (clauses or
adverbials) rather than complements (glede). It appears that there are signifi-
cant differences between English, on the one hand, and Croatian and Hungar-
ian, on the other, concerning the form that the specification of the active zone
of the metonymy (cf. Langacker 1995) assumes. The specification of the active
zone in English is far more schematic, e.g. a prepositional phrase as a comple-
ment of the adjective. If Croatian and Hungarian allow a predicative adjective,
then the referential type of metonymy (subject np for utterance) is more
likely, but even this usually requires more transparent coding of these events,
where active zones are made quite explicit (including even the explicit mention
of the verb of speaking) (cf. Brdar-Szabó & Brdar 1999).
The conclusion that we may draw from this is that Croatian and Hungar-
ian seem to lack the structural prerequisites that would bring about the switch
from referential to predicational metonymy, and lead eventually to the poly-
semy of adjectival predicates. Indeed, Croatian and Hungarian adjectives that
correspond to the English ones listed in (7) do not exhibit a comparable kind
and amount of polysemy. This may also explain why corresponding adjectives
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.21 (261)
tend to exhibit lower degree of communicative dynamism and assume the form
of analytic constructions. That the use or non-use i.e. choice of metonymies
may depend on typological properties of the given language is also shown by
Radden and Seto (this volume).
. Conclusions
Note
. Our corpus is part of a larger collection containing around 2,500 predicatively used ad-
jectives taking various types of complements, as attested in some 10,000 sentences excerpted
mostly from written sources.
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.23 (263)
References
Barcelona Sánchez, Antonio (1998). The state of the art in the cognitive theory of metaphor
and metonymy and its application to English studies. The European English Messenger,
7(2), 45–50.
Blank, Andreas (1999). Co-presence and succession: A cognitive typology of metonymy.
In K.-U. Panther & G. Radden (Eds.), Metonymy in Language and Thought [Human
Cognitive Processing 4] (pp. 169–191). Amsterdam and Philadelphia: Benjamins.
Brdar, Mario (1994). Infinitival Complement Clauses and Core Grammatical Relations
Subject and Object in English, German, Croatian and Hungarian. PhD dissertation,
University of Zagreb.
Brdar, Mario (1995). Kontrastivno-tipološki pristup proučavanju polisemije i sustava
komplementacije. [A contrastive-typological approach to the study of polysemy and
complementation systems]. Filologija, 24–25, 69–74.
Brdar, Mario, Brdar-Szabó, Rita, Gradečak-Erdeljić, Tanja, & Buljan, Gabrijela (in press).
Predicative adjectives in some Germanic and Slavic languages: On the role of metonymy
in extending grammatical constructions. Suvremena lingvistika (Zagreb).
Brdar-Szabó, Rita (2002). Referentielle Metonymie im Sprachvergleich. In M. Baróta,
P. Szatmári, J. Tóth, & A. Zsigmond (Eds.), Sprache(n) und Literatur(en) im Kontakt,
[Acta Germanistica Savariensia 7] (pp. 53–66). Szombathely: Maedinfo.
Brdar-Szabó, Rita & Brdar, Mario (1999). Predicative adjectives and grammatical-relational
polysemy: The role of metonymic processes in motivating crosslinguistic differences.
Paper presented at the International Workshop on Motivation in Grammar, Hamburg,
July 7–9.
Brdar-Szabó, Rita & Brdar, Mario (2001). Manner-for-activity metonymy across domains
and languages. Paper read at the 7th International Cognitive Linguistics Conference,
Santa Barbara, July 22–27.
Brdar-Szabó, Rita & Brdar, Mario (2002). Manner-for-activity metonymy in cross-linguistic
perspective. In B. Lewandowska-Tomaszczyk & K. Turewicz (Eds.), Cognitive Linguistics
Today (pp. 235–257). Frankfurt am Main: Peter Lang.
CCELDE (1987). Collins COBUILD English Language Dictionary. London and Glasgow:
Collins.
CIDE (1995). Cambridge International Dictionary of English. Cambridge: Cambridge
University Press.
Croft, William (1993). The role of domains in the interpretation of metaphors and
metonymies. Cognitive Linguistics, 4, 335–370.
Dirven, René (1990). Prototypical uses of grammatical resources in the expression of
linguistic action. In S. L. Tsohatzidis (Ed.), Meanings and Prototypes. Studies in
Linguistic Categorization. (pp. 267–284). London – New York: Routledge.
Dirven, René (1993). Metonymy and metaphor: Different mental strategies of conceptuali-
sation. Leuvense Bijdragen, 82(1), 1–28.
Dirven, René, Goossens, Louis, Putseys, Yvan, & Vorlat, Emma (1982). The Scene of
Linguistic Action and Its Perspectivization by SPEAK, TALK, SAY and TELL. Amsterdam
and Philadelphia: Benjamins.
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.24 (264)
Panther, Klaus-Uwe & Radden, Günter (1999). Introduction. In K.-U. Panther & G. Radden
(Eds.), Metonymy in Language and Thought [Human Cognitive Processing 4] (pp. 1–
14). Amsterdam and Philadelphia: Benjamins.
Panther, Klaus-Uwe & Radden, Günter (Eds.). (1999). Metonymy in Language and Thought
[Human Cognitive Processing 4]. Amsterdam and Philadelphia: Benjamins.
Panther, Klaus-Uwe & Thornburg, Linda (1999a). Coercion and metonymy: The interaction
of constructional and lexical meaning. In B. Lewandowska-Tomaszczyk (Ed.), Cognitive
Perspectives on Language [Polish Studies in English Language and Literature 1] (pp.
37–51). Frankfurt am Main et al.: Peter Lang.
Panther, Klaus-Uwe & Thornburg, Linda (1999b). The potentiality for actuality
metonymy in English and Hungarian. In K.-U. Panther & G. Radden (Eds.), Metonymy
in Language and Thought [Human Cognitive Processing 4] (pp. 333–357). Amsterdam
and Philadelphia: Benjamins.
Plank, Frans (1983). Transparent versus functional encoding of grammatical relations: A
parameter for syntactic change and typology. Linguistische Berichte, 86, 1–13.
Radden, Günter & Kövecses, Zoltán (1999). Towards a theory of metonymy. In K.-U.
Panther & G. Radden (Eds.), Metonymy in Language and Thought [Human Cognitive
Processing 4] (pp. 17–59). Amsterdam and Philadelphia: Benjamins.
Radden, Günter & Seto, Ken-ichi (this volume). Metonymic construals of shopping requests
in have- and be-languages.
Rudzka, Brygida (1982). The verb ASK and the scene of linguistic communication.
Unpublished paper. Catholic University of Leuven.
Rudzka-Ostyn, Brygida (1988). Semantic extensions into the domain of verbal
communication. In B. Rudzka-Ostyn (Ed.), Topics in Cognitive Linguistics (pp. 507–
553). Amsterdam and Philadelphia: Benjamins.
Rudzka-Ostyn, Brygida (1995). Metaphor, schema, invariance: The case of verbs of
answering. In L. Goossens, P. Pauwels, B. Rudzka-Ostyn, A.-M. Simon-Vanderbergen,
& J. Vanparys (Eds.), By Word of Mouth: Metaphor, Metonymy and Linguistic Action in
a Cognitive Perspective (pp. 205–243). Amsterdam and Philadelphia: Benjamins.
Sadock, Jerrold M. (1978). On testing for conversational implicature. In P. Cole (Ed.),
Pragmatics [Syntax and Semantics 9] (pp. 281–298). New York: Academic Press.
Searle, John R. (1975). Indirect speech acts. In P. Cole & J. L. Morgan (Eds.), Speech Acts
[Syntax and Semantics 3] (pp. 59–82). New York: Academic Press.
Seto, Ken-ichi (1999). Distinguishing metonymy from synecdoche. In K.-U. Panther &
G. Radden (Eds.), Metonymy in Language and Thought [Human Cognitive Processing
4] (pp. 91–120). Amsterdam and Philadelphia: Benjamins.
Taylor, John R. (1989). Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Ullmann, Stephen (1962). Semantics: An Introduction to the Science of Meaning. Oxford:
Blackwell.
Vachek, Josef (1961). Some less familiar aspects of the analytical trend in English. Brno
Studies in English, 3, 31–44.
Verschueren, Jef (1980). On Speech Act Verbs. Amsterdam: Benjamins.
Verschueren, Jef (1984). Basic Linguistic Action Verbs: A Questionnaire [Antwerp Papers in
Linguistics 37]. Antwerp: Universitaire Instelling Antwerpen.
TSL[v.20020404] Prn:4/06/2003; 16:20 F: PB11311.tex / p.26 (266)
Verschueren, Jef (1985). What People Say They Do with Words: Prolegomena to an Empirical-
Conceptual Approach to Linguistic Action. Norwood, NJ: Ablex.
Verschueren, Jef (Ed.). (1987). Linguistic Action: Some Empirical-Conceptual Studies.
Norwood, NJ: Ablex.
Waltereit, Richard (1998). Metonymie und Grammatik. Kontiguitätsphänomene in der
französischen Satzsemantik [Linguistische Arbeiten 385]. Tübingen: Max Niemeyer.
Waltereit, Richard (1999). Grammatical constraints on metonymy: On the role of the
direct object. In K.-U. Panther & G. Radden (Eds.), Metonymy in Language and
Thought [Human Cognitive Processing 4] (pp. 233–253). Amsterdam and Philadelphia:
Benjamins.
Name index
A D
Arana, José Ignacio de 85, 89 Díez, Olga 47
Athanasiadou, Angeliki 169 Denison, David 173
Atlas, Jay D. 171 Dirven, René 1, 169, 242, 254, 261
Dixon, R. M. W. 154
Attardo, Salvatore 12, 81–82, 92
F
B Fauconnier, Gilles 11, 84, 87, 99,
Bach, Kent 105, 109–110, 112–113 128, 169
Barcelona, Antonio 1, 7, 10, 12, 18, Fields, W. C. 92
23, 81, 83–84, 98–99 Fillmore, Charles J. 82, 84, 124, 146,
Barsalou, Lawrence 27 205, 211
Blakemore, Diane 25, 47 Foldi, Nancy S. 120–122, 124
Blank, Andreas 250 Francik, Ellen P. 225
Brdar, Mario 7, 10, 12, 16–18, 237,
241, 252, 257, 260–261
G
Brdar-Szabó, Rita 7, 10, 12, 16–18,
Gernsbacher, Morton Ann 244
237, 241, 252, 257, 260
Gibbs, Raymond W., Jr. 4, 105,
Brown, Penelope 129, 235
113–114, 215–216, 224–225
Brownell, Hiram 120–121, 123
Goldberg, Adele 12–13, 106–107,
Bußmann, Hadumod 163
124, 128
Buljan, Gabrijela 252, 257
Goossens, Louis 35, 40–41, 47, 186,
Bybee, Joan L. 188–189
242
Gradečak-Erdeljić 252, 257
C Grice, H. Paul 8, 13, 81–82, 105,
Carston, Robyn 10, 23, 25–28, 47 146, 169, 171, 211, 259
Chaucer, Geoffrey 188, 192 Grimm, Jacob 153
Clark, Herbert H. 105, 114, 225 Grimm, Wilhelm 153
Claudi, Ulrike 151
Contini-Morava, Ellen 154 H
Cook, Haruko M. 206 Harnish, Robert M. 105, 109–110,
Corbett, Greville 151 112–113
Coulson, Seana 7, 10–11, 97, Heine, Bernd 223, 237
99–100, 128 Hemingway, Ernest 11, 62, 64–65,
Croft, William 5, 215, 243 77
Name index
I N
Ikegami, Yoshihiko 231, 236–239 Norrick, Neal R. 9
J O
Johnson, Mark 2, 23, 28, 30, 32, 47, Oakley, Todd 7, 10–11, 97, 99–100,
107, 149–150, 162 128
Josephs, Lewis S. 206 Ohori, Toshio 205, 213, 217–218
Okamoto, Shigeko 7, 10, 15–16, 128,
K 132, 177, 205, 217–218
König, Ekkehard 211, 216 Otal Campo, José L. 1
Köpcke, Klaus-Michel 7, 10, 14, 149,
151–152 P
Kövecses, Zoltán 3, 9, 12, 47, 83–84, Pérez Hernández, Lorena 7, 9, 10,
178, 211, 214–215, 238, 245 18, 23, 97, 101, 146, 158
Kalisz, Roman 246 Pörings, Ralf 1
Kay, Paul 124, 205 Palmer, Frank R. 172
Kiparsky, Carol 180 Panther, Klaus-Uwe 1–2, 4, 6–8, 10,
Kiparsky, Paul 180 13–14, 16, 23, 83, 98, 105–106,
Kuno, Susumu 206 108, 114–116, 124, 127–129, 170,
Kuteva, Tania 169 176–179, 199–200, 211–212, 215,
Kytö, Merja 200 217–218, 224, 243, 246, 249, 258
Papafragou, Anna 9–10, 47, 98
L Peña, Sandra 31
Lakoff, George 2–4, 8, 23, 28–32, 47, Pollard, Carl 163
84, 99, 106–107, 109, 124,
149–150, 162, 211, 243, 246
R
Langacker, Ronald 3, 5, 47, 85, 100,
Radden, Günter 1, 3–4, 8–10, 12,
156, 163, 246, 260
16–18, 23, 47, 83–84, 178, 211,
Leech, Geoffrey 242
213–215, 223, 237–238, 245–246,
Levinson, Stephen C. 8, 15, 18,
262
23–24, 26, 129, 169–174,
Raskin, Victor 12, 81–82, 86
178–179, 181–182, 196, 198–199,
Recanati, François 23, 26
235
Rissanen, Matti 200
Rudzka, Brygida 242
M Rudzka-Ostyn, Brygida 242
Marín Arrese, Juana 82 Ruiz de Mendoza, Francisco J. 1,
Marmaridou, Sophia S. A. 1 8–10, 18, 23, 26, 31, 35–36, 47,
Martin, Samuel E. 217 97–98, 101, 146, 158
Name index
S U
Sadock, Jerrold M. 12, 109–111, 259 Ullmann, Stephen 242
Sag, Ivan A. 163
Schreckengost, Viktor 11, 70, 72,
75–77 V
Searle, John R. 105, 109–110, Van der Auwera, Johan 169
112–114, 133 Vendler, Zeno 179, 227–228
Seto, Ken-ichi 4, 8, 10, 16–18, 213, Verschueren, Jef 18, 242
223, 245–246, 262 Visser, F. Th. 172, 188–190
Skeat, Walter W. 100 Voßhagen, Christian 146
Song, Nam Sun 9, 237
Sperber, Dan 18, 23–27, 47, 81, 105
Stefanowitsch, Anatol 8, 10, 12–13, W
16, 105, 128, 146, 211, 216 Waggerl, Karl Heinrich 14, 149,
Stemmer, Brigitte 120–121 158–160, 163–164
Stringfellow, Andrew 120–121, 123 Weylman, Sally T. 120–122, 124
Suzuki, Ryoko 205, 217 Wienold, Götz 151
Sweetser, Eve 128, 146, 169 Wilson, Deirdre 18, 23–27, 47,
81–82, 105
T
Taylor, John R. 242 Y
Teramura, Hideo 206 Young, Robert 154
Thornburg, Linda L. 1–2, 4, 6–8, 10,
13–14, 16, 83, 98, 105–106, 108,
114–116, 124, 127–129, 170, Z
176–179, 199–200, 211–212, 215, Ziegeler, Debra 8, 10, 15, 169, 177,
217–218, 224, 243, 246, 249, 258 211, 217, 254, 256
Traugott, Elizabeth C. 205, 211, 216 Zubin, David A. 8, 10, 14, 124, 149,
Turner, Mark 11, 23, 29–30, 87, 99 151–152
Metonymy and metaphor index
Note: In this index we follow the widespread convention of notating metonymies as source
for target and metaphors as target is source. In those cases in which authors have not
followed this convention we have converted them into the conventional citation format.
Metonymies C
category member for category
A 87
ability for action 116 cause for effect 95
after component of directive for central subevent for whole
directive 115, 131–132, 213, event 228
215 characteristic behavior for
agent for action 18, 252, 254 female 157
archetypal figure for female characteristic living space for
157–158 female 158
argument for proposition 92 component of speech act scenario
author for work for for whole scenario 115, 130,
(non-unique) sample 39 145, 213–215
author for work for unique container for contents 59
sample 38
corporation for profits 55
country for army 73
B
before component of commissive
for commissive 135–137
before component of directive D
for directive 133–135 definition for defined 96
before component of exclamation
for exclamation 214–215
before component of wish for
wish 137–139 E
behavior for kinds of people effect for cause 60
conventionally associated emotional response to female for
with behavior 95–96 female 157
book for book as physical object entity for its conventional
85, 93, 99, 239 property 90–91
book for book as semantic entity for one of its
content 85, 99 conventional properties 95
Metonymy and metaphor index
M S
manner for linguistic action 17, salient body part for animate
242, 250, 255, 259–260 being 7
speaker for utterance 18, 252,
254
O
state for event 258
obligation to act for action 4–5
state for university 58
stating a fact for stating its
P implications 96
part for part 93, 105 status for female 157
part for whole 59, 93, 99, 105, 178 strategy (in a conflict) for
part of a frame for a whole conflict 93
frame 93 symptom for physical state 87
Metonymy and metaphor index
T Metaphors
transfer for transaction 228, a female is a (skeletal/scaffold)
230, 237 shape 155
type for token 93 a female is an animal 155, 157
argument is war 31
U conscious is up 30
university for football team 58 generic is specific 75
goals are destinations 32, 47
happy is up 31
V
people are animals 30–32
virtuality for actuality 177
politeness is distance 132
sad is down 31
W unconscious is down 30
willingness for ability 131
willingness to act for action
119
Subject index
D
F
default readings 8
face threat 223, 227, 229
defeasibility 8, 17, 259 see also factive clause(s) 179, 184–186,
cancelability 188–193, 199
deference 17, 231, 234, 237 factivity 180, 200
deontic function see communicative Finnish 16, 228, 230, 237
function, deontic fixation of reference 23, 25, 27, 29
dependent clause(s) 13, 127 foregrounding 5
directive(s) 16, 133, 139, 145, frame(s) 12, 82–84, 86–93, 95,
212–214 see also speech act 97–100, 106, 116, 123, 131,
scenario(s), Request Scenario and 211–217 see also idealized
Suggestion Scenario cognitive model(s) and
disambiguation 23–25, 27, 29 scenario(s) and speech act
do-language(s) 236–237 scenario(s)
Domain Availability Principle 11, 36 adjustment 12, 82, 86, 89, 92,
domain matrix 11, 243 95
double metonymy see metonymy, blend 86–87, 89, 99
double overlap 86–87, 89, 92, 95
shift 86, 89, 92, 95
French 127–128, 178, 253
functional cognitive domain 83–84,
E 98
elliptical expression(s) 208–209, 216 functional independence 26–27
enrichment 23, 25, 27, 30
entailment(s) 181 G
as opposed to metonymy 3 generic past abilities 189
Subject index