PhiN 12/2000: 1

Patrícia Matos Amaral (Coimbra)

Metaphor and Relevance

Sperber and Wilson have developed a cognitive theory of human communication that sees the interpretation process as an inferential strategy using linguistic information and other kinds of evidence to identify the speaker's informative intention. The pragmatic principles are therefore not language-specific and can be better defined as "problem-solving" reasoning schemes or "controlled guesswork". Although metaphor and figurative language are studied within the framework of Relevance Theory, the authors don't take full advantage of the paradigmatic role of such utterances to the Inferential Model. The aim of this paper is to show how the striking similarities between metaphoric utterances interpretation and the Inferential Model of communication can be used to support this theoretical proposal and to reveal some of the achievements of non-demonstrative inference. Interaction Theory, as well as some recent cognitive-based studies on metaphor, will be discussed and compared.

1 Relevance Theory and the Inferential Model of communication

Relevance Theory, as presented in Sperber and Wilson (1986)1, is an attempt to explain the general principles of human communication. Following Grice's William James Lectures, the authors' main thesis is that the expectations created by communication itself play a major role in utterance interpretation and are therefore constitutive of this process. Unlike Grice, however, Sperber and Wilson don't postulate a kind of conversational "ethics" "observance of which is regarded as providing standards of rational discourse" (Grice 1989: 368). The authors maintain that interpretation is primarily a cognitive phenomenon which depends on our ways of processing information. The co-operative or contract-like dimension of human communication is thus external to interpretation proper and is already a consequence of the nature of human cognition.

For the Inferential Model of communication (which is opposed to the Code Model), the semantic content of sentences, obtained by linguistic decoding, is only a part of the evidence provided by the communicator and is subordinated to the inferential process of identifying its communicative intention,2 which is the core of utterance interpretation. Unlike the decoding mechanisms, which are highly specialized – domain-specific – and therefore modular in Fodor's sense (Fodor 1983), inferential abilities involved in interpretation are not language-specific and share the properties of central or global thought processes: they are non-demonstrative (the addressee constructs a hypothetical assumption which can be confirmed or not) and informationally non-encapsulated because they can use as premise any information available to the system. In this sense Pragmatics should not be conceived of as a separate module, but rather as "the domain in which grammar, logic and memory interact." (Wilson and Sperber 1991b: 583).

PhiN 12/2000: 2

Within this model utterance interpretation is thought of as a product or a resultant of an inferential calculation that includes both deductive and non-demonstrative reasoning schemes. Although the process of "guessing" the communicator's intended assumption takes place in two stages (hypothesis formation and hypothesis confirmation), deductive reasoning plays a crucial role in the whole process. Sperber and Wilson underline the fact that such a system of rules is very useful in that it allows maximal utilisation of the available information because it assures the accuracy of the conclusions drawn by the system and therefore restricts the processing of new assumptions. In fact, most of the chapter dedicated to inference is about deductive rules.3

One of the main features of non-demonstrative inference – that importantly approaches utterance interpretation and scientific theorizing, and makes it difficult to study them both – is the free access to any information held by the system:

So, for example, in creating a scientific hypothesis to account for a certain range of data it is legitimate to rely on analogies with other domains of knowledge, seemingly random association of ideas, and any other source of inspiration that comes to hand. Once a hypothesis has been formed, the extent to which it is regarded as confirmed will depend on how well it fits not only with neighbouring domains of knowledge but with one's whole overall conception of the world. (Wilson and Sperber 1991a: 380)

According to Relevance Theory, the context used to process an assumption is not given prior to the interpretation process but, rather, constructed and modified by the addressee as necessary; therefore, the construction of the context is an essential part of the interpretation process. The context may include any information, whether stored in the memory or directly obtained from the encyclopaedic entries of the concepts which are the output of the linguistic device. As the authors note, this information, which is to be combined with verbally communicated information, does not need to be true or valid (in a logical sense) but rather efficient for the current purposes.

For Sperber and Wilson, humans process information in order to improve their representations of the world according to their own interests. In this sense the main goal of human cognition is relevance – the optimisation of one's cognitive resources, i.e., achieving cognitive effects without much processing effort. As far as human communication is concerned, each individual expects utterance interpretation to achieve contextual effects (modifications in one's assumptions as a consequence of interaction between new decoded information and old representations) without requiring too much processing effort, and will therefore interpret utterances accordingly. This brings us to the Principle of Relevance: "Every act of ostensive communication communicates the presumption of its own optimal relevance" (Sperber and Wilson 1986: 159).

PhiN 12/2000: 3

Because any utterance is expected to be relevant, the addressee's task will consist of looking for the contextual assumptions that will achieve the most productive utterance processing. Nevertheless, the construction of the context for utterance processing is not unlimited because every context expansion involves processing effort.

Since in recognizing an intended assumption individuals have free access to any stored information, the quantity and quality of the contextual assumptions depend mainly on the addressee's availability, as well as on the kind of utterance (Goalty 1997 relates Relevance Theory and Register Theory, attempting to adjust the notion of Relevance to the characters and specific purposes of different genres). Although Relevance Theory is a cognitive theory of human communication, not all of the contextual effects need to be seen as "cognitive" or "informational". In fact, an utterance may be optimally relevant by producing more or less determinate social and emotional changes in the context and thereby compensate the processing effort.

Since both explicit and implicit communication can be more or less determinate and there is a continuum between strongly conveyed thoughts (as when the intended assumption is very close to the logical form of the utterance) and a complex of vague thoughts, more or less freely inferred by the addressee (e.g. in metaphoric utterances), Grice's account of human communication is not adequate. According to Sperber and Wilson, it is necessary to make a distinction between strong implicatures and weak implicatures. While the former have been studied by pragmatists, the latter cannot be accounted for within a framework that sees communication as "strong communication".

In fact, in many cases relevance is achieved by the recovery not of a single fully determinate assumption, but by the identification of an open and wide complex of propositions whose contextual effects may only partially coincide with the ones intended by the communicator. While some of these assumptions will be salient, so that it would be impossible not to recover them, the recovery of others will depend mostly on the addressee's responsibility. These propositions are not recovered simultaneously, but according to a certain hierarchy, which partly explains the difficulty, or even impossibility, of finding a paraphrase for figures of speech that accounts for their cognitive "richness" and "condensation". In this sense, weak implicature could not be conceived of outside a framework where communication is seen as providing and interpreting evidence about the communicative intention and the degree of participation required (and granted) to recover it.

2 Metaphor

Max Black's account of metaphor, known as Interaction Theory because it is based on Richards' view of metaphor as an "interaction of two thoughts", represents a break in metaphor theorizing. Unlike the traditional views of metaphor, which explain its content by developing the metaphoric utterance into a comparison or describe its interpretation as the finding of a lacking word (tertium verbum) improperly substituted by the displaced word, the interaction view sees metaphor as an intellectual operation with a cognitive import. More importantly, the transfer or projection, which allows us to see one object in terms of another, therefore producing a redescription or a change of representations, is presented as "a creative response from a competent reader." (Black 1979: 29)

PhiN 12/2000: 4

Let us see how this model works. We'll take the following sentence as an example:


The sense of an utterance is a dinosaur.4

The metaphoric interaction takes place between the focus (the concept of dinosaur) and the frame ("the sense of an utterance", which is literal) in that the reader will apply the system of associated commonplaces evoked by dinosaur to the other concept. By system of associated commonplaces or implication complex Black means "a set of standard beliefs [...] (current platitudes) that are the common possession of the members of some speech community." (1968: 40) This set of propositions shall be conventionally and spontaneously evoked; they are culture dependent and may be created or altered ad hoc by the author. Black notices that the success of metaphor interpretation does not depend on the truth-value of these assertions. In fact, these chunks of information correspond to the stereotypical features which shape conventional representation of the objects rather than to scientific truths (cf. Metzing 1981 and Fillmore 1985 about the concept of frame).

So in the example sentence the system of commonplaces evoked by dinosaur would include some of the following propositions:



A dinosaur is a pre-historic animal.

b. A dinosaur is a reconstitution that relies on our knowledge about the evolution of life on Earth.

c. A dinosaur fed on plants.

d. Bone fragments and other remains are used as evidence for the reconstitution.

e. Nobody has ever seen a living dinosaur.

According to Black, the task of the hearer consists of constructing a corresponding set of assertions that fits the principal subject, i.e., the frame. In this sense, the way we conceive of "the sense of an utterance" is mediated by our representation of a dinosaur, or rather by the system of propositions it evokes, which is used as an instrument of description that organizes and transforms our view of the other subject. The metaphor is therefore like a filter which hides some aspects of the frame and reveals other ones (which couldn't be discovered otherwise). In this sense the analogy does not precede the metaphor but is rather created by it – a conceptual innovation is therefore produced by the interpretation process.

In constructing the new implications "some of the 'associated commonplaces' themselves suffer metaphorical change of meaning in the process of transfer from the subsidiary to the principal subject." (Black 1968: 42) For instance, in equating a dinosaur and the sense of an utterance it is necessary to transform the palaeontologist/scientist into an agent that is constructing an object so that we can identify him and the hearer of an utterance – in this analogy they have the same purpose or the same function, i.e., they are seen as two instantiations (in different domains) of the same abstract concept. Thus the complex of implications resulting from the metaphorical projection may include "a set of subordinate metaphors" (ibidem) through which the meaning of the metaphor is extended. In the example sentence, such metaphors would be

PhiN 12/2000: 5



The hearer is a palaeontologist.

b. The words and actions of the speaker are the evidence and remains used to reconstruct a dinosaur.

c. General knowledge about the world and communication is the scientific knowledge about the evolution of life on Earth.

According to Black, the connections between both subjects may be grounded on analogy or on other relationships that allow these "shifts" of meaning. In fact, Black says very little about the constraints on this projection.

Although Black mentions these subordinate or minor metaphors which are intended to be less important (his analogy with musical overtones is very significant) as an objection to the interaction theory – which would thereby become a circular explanation – it seems to me that they should not be seen as a problem, but rather as an achievement of this account.

In fact, as a description of the products5 of metaphor interpretation, they constitute an important insight about analogical reasoning. According to Holland et alii (1993), the mapping between an analogy's source- and target-domain (focus and frame in Black's terms) is the resultant of an inductive schema which is an ad hoc abstract category built to relate two distant domains. Correspondences between them are made possible by an identity of structure and roles created by the "role-like category" which is less specific than any of the domains and can continue to evolve as new examples require to be integrated in it. In the example sentence, interpreting the metaphor means that the hearer should identify two agents – hearer and palaeontologist – pursuing the same purpose – the reconstitution of an object – with the help of instruments or means, some of them directly connected with the object they want to reconstitute (the words and actions of the speaker, bones and other remains from the dinosaurs), others being supplied by the agent and helping to organize or give sense to the first ones (general knowledge about communication and inferring the speaker's intention, scientific knowledge to interpret archaeological data). The schema "someone has the task of trying to reconstitute an object by using and interpreting the available evidence" allows the hearer to integrate the two disparate concepts and give sense to the metaphorical utterance, since both "the sense of an utterance" and "dinosaur" can be seen as instantiations of this more general schema.

As shown in the example, the set of implications constructed for the primary subject, i.e., the metaphorical mapping, requires the hearer to transform or elaborate the information provided by the focal and the frame concepts in a new category. According to Gibbs (1994), one of the main limitations of Black's account of metaphor is that it identifies the meaning of the metaphor with the meaning of the frame (the resultant of the interpretation process would thus be its redescription), thus contradicting the sense of "interaction". In fact, since metaphorical projection is unidirectional and the focus is conceived of as a "filter", only the frame should be altered in this process. Moreover, that Black sees the production of subordinate metaphors as a problem of his account shows that the new implications are intended to be literal assertions, as if such a "translation" would be a condition for the metaphor's intelligibility. On the contrary, Gibbs argues, very often the new implications can only be understood within the very metaphor and are easily understood by the speakers, who speak and think with a poetic mind.

PhiN 12/2000: 6

Gibb's objections to Black's interaction view also concern other models of metaphor interpretation that describe it as a projection between two domains. Such theories find it difficult to explain how the mapping takes place, i.e., how new features that cannot be found in either the primary or the secondary subject result from metaphor interpretation and allow conceptual innovation. I think, though, that Black's remarks about the subordinate metaphors can lead us to some recent achievements in metaphor theorizing and even corroborate them. In fact, by presenting metaphor as a cognitive operation, whose import cannot be communicated otherwise, Black's theory underlies interactionist and cognitive approaches and in a certain sense already predicts these developments.

One of the most recent developments in metaphor theorizing has appeared within the framework of the theory of "blending" or "conceptual blending". This approach explains metaphor as a particular case of a general and pervasive cognitive process known as "blending" or "conceptual integration" (for a general presentation of both the framework and the linguistic and conceptual phenomena it applies to see Fauconnier and Turner 1994). According to this view, metaphorical projection takes place between mental spaces, which are defined as temporary representations constructed by the speakers when they conceive of an experience (imagined or not). These mental spaces largely depend on stable knowledge structures which are stored in memory (e.g. conceptual metaphors identified by Lakoff and Johnson 1980), but unlike them mental spaces are short-term representations intended to account for the conceptualisation needs of the speakers, some of which are novel and unique.

Unlike Black's account, this model posits the existence of four (or more) mental spaces, namely: two "input spaces" (which correspond to the focus and the frame), a "generic" space which comprises the abstract "skeletal" structure shared by both of them and a "blend" space, where the representations from the input spaces, and eventually from other domains, combine originating an innovative conceptualisation (the latter are called "middle spaces").6 This four-space model interestingly accounts for both processes and products of metaphor interpretation which could not be accounted for by two-space models.

In this sense, the generic space explains how the constructed system of implications about the frame gets to correspond the one evoked by the focus and how it implies a 'metaphorical change of meaning' of the commonplaces themselves. On the other hand, the blended space makes it possible to explain the novelty or "emergent structure" which could not be accounted for in Black's theory (although, as we saw, one of the main insights of his theory, explicitly stated in Black 1979, was the acknowledgment of metaphor's role in conceptual innovation). In fact, in constructing a blended space hearers recruit information from different sources (other than the two input spaces involved in the analogical projection) – in an operation known as "completion" (cf. Grady et alii 1999: 5) – and combine it according to a logic that is constitutive of the blend and cannot therefore be found in any of the input spaces (cf. e.g. the analysis of Bertrand de Born's character in Fauconnier and Turner 1994). This process, which accounts for different phenomena, from counterfactuals to fiction, means that by trying to make sense of a specific situation (which can also be triggered by the understanding of a particular utterance) humans derive inferences that could not be predicted or drawn from any of the information sources alone – knowledge is generated by the very process of integrating them.

PhiN 12/2000: 7

3 Metaphor and Relevance

In the beginning of this paper I argued that Relevance Theory does not take full advantage of metaphor's significance as a paradigm of the Inferential Model of communication. I now wish to argue the "strategic role"7 of metaphor in the understanding of the interpretation process as it is conceived of within the framework of Relevance Theory.

Sperber and Wilson define utterance interpretation as a construction of the addressee that depends on general reasoning strategies ("problem-solving" abilities) rather than on specialized cognitive processes (i.e. linguistic decoding mechanisms). The addressee identifies the communicator's intended assumption by formulating an interpretative hypothesis which may or may not be validated. In this sense, utterance interpretation depends mainly on a non-demonstrative inferential calculation. Similarly, there is a long tradition in looking at metaphor interpretation as a "risky" or a "discovery" procedure, in that the mappings allowing the hearer to make sense of the utterance are somehow "genial" and result from an inferential process which does not have a proof value.

In this sense, by identifying the "causally relevant" (Holland et alii 1993: 297) relations that allow the mapping of two disparate concepts, and therefore finding a solution for the "meaning problem" created by metaphor, the inferential calculation responsible for metaphor interpretation constitutes a hyperbolic manifestation of the reasoning strategies through which an addressee validates the relevance presumption. According to Holland et alii (1993), inductive reasoning mechanisms are cognitive processes that allow humans to cope with the complexity and indeterminacy of their everyday lives, when they have to act and make decisions in the absence of full information. Both analogical reasoning and the interpretative hypothesis through which the addressee reconstitutes the intended assumption can be comprised in the group of "all inferential processes that expand knowledge in the face of uncertainty" (Holland et alii 1993: 1).

Moreover, the very process of validating relevance presumption, by finding a context that allows the output of the linguistic device to be processed, is in a certain sense reproduced in metaphor interpretation8 . In fact, the whole point of interpreting a metaphor consists of combining or integrating information in a productive way. As we saw, metaphoric projection takes place from focus to frame, i.e. the direction of metaphor interpretation is the opposite from reading. This directionality or irreversibility of metaphor has been seen as the main ground for its heuristic potential and cognitive import, since we tend to model less known concepts upon concrete well described domains of our experience, therefore bringing about a conceptual innovation from available information. In this sense, the intellectual operation described by Black constitutes a good example of intelligent use of old information by the system to yield new information (cf. the importance of metaphor in scientific thought).

The construction of the context, which is presented in Relevance Theory as an essential and constitutive part of the interpretation process, can thus be seen as corresponding to the evocation of Black's system of associated commonplaces, the first step in metaphor interpretation. In both cases, the addressee is selecting and retrieving a set of stored general assumptions about the world which will feed the inferential calculation.

PhiN 12/2000: 8

We have seen that contextual effects result from the combination of contextual assumptions with the new information. We have also seen that metaphoric projection constitutes the application of the system of associated commonplaces from focus to frame. In this sense, the retrieval of contextual assumptions, as the identification of the relevant commonplaces of the source, is the essential inferential procedure through which the system takes advantage of its stored information, whose relevance relies on its potential multiplicative effect.

In the example sentence, some of the conclusions drawn from the projection of the system of associated commonplaces of "dinosaur" to "the sense of an utterance" could be:



The communicator's words and actions are the evidence left for the addressee to reconstruct her intention, as bones and other remains are used by the palaeontologist as a departure for his reconstitution;

b. The addressee is the palaeontologist;

c. General knowledge about the world and about communication is used by the addressee to identify the communicator's intention, as well as knowledge about the evolution of life on Earth is used by the palaeontologist to reconstruct a dinosaur;

d. The dinosaur is a palaeontologist's (science) hypothesis; so is the addressee's hypothesis about the communicator's intention.

This set of propositions not only shows the temporal development of metaphor interpretation as a process, but also reveals the specificity of its products, which can be conceived of as weak implicatures, as defined above. In fact, they are implicitly conveyed propositions (as they are triggered by the logical form of the utterance and inferred by the addressee), they cannot be rigidly determined in that they do not constitute a definitive closed set and they are hierarchically produced and functionally used by interpretation itself. These "vague effects" cannot therefore be summarized in a single fully determinate proposition (a strong implicature) since such a paraphrase precludes the very possibility of communicating gradually and progressively a great amount of information, which to a great extent is only limited by the interpretation availability of the addressee. In fact, through a wider expansion of context the analogy could be almost indefinitely enlarged.

In an important way, these features define the subordinate metaphors that Black identified as occurring during the metaphorical projection. As it was shown above, the identification/construction of the metaphor's grounds relies upon the transfer of the system of associated commonplaces from focus to frame: each of these displacements corresponds to a proposition which in a certain sense is "a little metaphor". The author interestingly does acknowledge that although the set of constructed implications can be explicitly formulated, the paraphrase will necessarily involve a loss of the metaphor's cognitive import:

For one thing, the implications, previously left for a suitable reader to educe for himself, with a nice feeling for their relative priorities and degrees of importance, are now presented explicitly as though having equal weight. (Black 1968: 46)

PhiN 12/2000: 9

Moreover, these propositions do not have a proof value, 9 since they cannot be demonstrated according to a deductive logical procedure: they constitute the hearer's response to the meaning problem created by the metaphorical utterance.

4 Concluding remarks

By comparing metaphoric utterances interpretation, as it is described by Interaction Theory, and the Inferential Model of communication proposed by Relevance Theory, I was pursuing two main objectives. Firstly, I wanted to reinforce the foundations of the Inferential Model, by showing how they are paradigmatically manifest in the condensed metaphoric formulation. Secondly, metaphor interpretation and, more generally, analogical reasoning can give us some insight into the inferential processes responsible for interpretation (as well as about other central processes, as proposed in Wilson and Sperber 1991a: 380).

In most of their texts about inference, Sperber and Wilson insist that the participation of non-demonstrative inference, which plays a central role in recognizing the intended assumption, does not preclude the action of the deductive device in this process. In the chapter "Inference" of Sperber and Wilson (1986), as well as in other papers on this subject, the authors almost exclusively describe the role of deduction in generating conclusions from the output of the linguistic device and premises supplied by the system, thereby assuring a logically strict control of the interpretation process and overcoming intuitive or arbitrary explanations (given the incipiency of our knowledge about non-demonstrative reasoning). After defining non-demonstrative reasoning as "a form of suitably constrained guesswork" (Sperber and Wilson 1986: 69), the authors intend to specify the nature of these constraints; in fact, their adequacy seems to be the very condition of their effectiveness, i.e., of their realism.

In this sense, the authors present the central system as a deductive device, therefore overseeing the role of non-demonstrative reasoning in the formation of hypotheses which constitute the premises of their examples (for a close analysis of examples in Wilson and Sperber 1981 and Wilson and Sperber 1991a, see Amaral 1999: 101–102) and they symptomatically hesitate between the expressions retrieval and construction of implicated premises.

In doing so, "Sperber and Wilson recognize the role of such creative hypothesis formation, but they somehow lose sight of its centrality" (Levinson 1989: 466), which corresponds to metaphor's lack of centrality within this theory.

As shown above, Interaction Theory also says little about the selection and projection of the implications from focus to frame, which brings us back to the main problem of inductive reasoning schemes. How can these probabilistic processes be constrained? How can we explain that absurd or irrelevant inferences are not produced by the system? And how can the implications evoked by the focus, as well as the contextual assumptions mobilized by the addressee, be simultaneously explored and transformed?10 Some constructivist (in Ortony's sense: see Ortony 1979: 2) approaches to metaphor have shed light on this subject (see e.g. Indurkhya 1994, who argues for an interactive view of cognition that accounts for its objectivity, or the "experientialist synthesis" proposed by Lakoff and Johnson 1980). According to Holland et alii (1993), induction should be studied in a pragmatic rather than in a syntactic perspective, therefore taking into account the goals to which cognition is oriented.

PhiN 12/2000: 10

According to Relevance Theory, humans pursue the optimisation of the processing cost-benefit relation, cognitive efficiency being a condition of their very survival. By allowing humans to conceive of and communicate a great amount of information and by being a major instrument of conceptual innovation, metaphor is relevant also by showing us not only the imaginative character of our rationality, but also the rationality of our imagination.


Amaral, Patrícia Matos (1999): Do paradigma ao modelo: a relevância da metáfora para a compreensão do processo interpretativo. Faculdade de Letras da Universidade de Coimbra (unpublished M.A. Dissertation)

Black, Max (1968): Models and Metaphors. Studies in Language and Philosophy. Ithaca, New York: Cornell University Press. (1st ed. 1962)

Black, Max (1979): "More about Metaphor", in: Ortony, Andrew (ed.) (1979): Metaphor and Thought. Cambridge: Cambridge University Press, 19–43.

Fauconnier, Gilles and Mark Turner (1994): "Conceptual Projection and Middle Spaces", in: UCSD: Department of Cognitive Science Technical Report 9401. (Available from "").

Fillmore, Charles (1985): "Frames and the Semantics of Understanding", in: Quaderni di Semantica 6/2, 222–254.

Fodor, Jerry (1983): The Modularity of Mind. Cambridge, Mass.: The MIT Press.

Gibbs, Raymond (1994): The Poetics of Mind. Figurative Thought, Language, and Understanding. New York: Cambridge University Press.

Goalty, Andrew (1997): The Language of Metaphors. London/New York: Routledge.

Grady, Joseph et alii (1999): "Blending and Metaphor", in: Steen, Gerard and Raymond Gibbs (eds.): Metaphor in Cognitive Linguistics. Philadelphia: John Benjamins (in press). (Available from "").

Grice, H. Paul (1989): Studies in the Way of Words. Cambridge, Mass./London, England: Harvard University Press.

Holland, John et alii (1993): Induction. Processes of Inference, Learning and Discovery. Cambridge, Mass.: The MIT Press.

Indurkhya, Bipin (1994): "Metaphor as Change of Representation: An Interaction Theory of Cognition and Metaphor", in: Hintikka, Jaakko (ed.) (1994): Aspects of Metaphor. Dordrecht: Kluwer Academic Publishers, 95–150.

PhiN 12/2000: 11

Lakoff, George and Mark Johnson (1980): Metaphors We Live By. Chicago/London: The University of Chicago Press.

Levinson, Stephen (1989): "A review of Relevance", in: Journal of Linguistics 25, 455–472.

Metzing, Dieter (1981): "Frame Representation and Lexical Semantics", in: Eikmeyer, Hans-Jürgen and Hannes Rieser (eds.): Words, Worlds, and Contexts. New Approaches in Word Semantics. Berlin/New York: Walter de Gruyter, 320–342.

Miller, George (1979): "Images and Models, Similes and Metaphors", in: Ortony, Andrew (ed.) (1979): Metaphor and Thought. Cambridge: Cambridge University Press, 202–250.

Ortony, Andrew (ed.) (1979): Metaphor and Thought. Cambridge: Cambridge University Press.

Paivio, Allan (1979): "Psychological Processes in the Comprehension of Metaphor", in: Ortony, Andrew (ed.) (1979): Metaphor and Thought. Cambridge: Cambridge University Press, 150–171.

Prandi, Michele (1992): Grammaire philosophique des tropes. Mise en forme et interptrétation discursive des conflits conceptuels. Paris: Les Editions de Minuit.

Rumelhart, David (1979): "Some Problems with the Notion of Literal Meanings", in: Ortony, Andrew (ed.) (1979): Metaphor and Thought. Cambridge: Cambridge University Press, 78–90.

Sperber, Dan and Deirdre Wilson (1986): Relevance. Communication and Cognition. Oxford: Basil Blackwell. (2ª ed. de 1995)

Wilson, Deirdre and Dan Sperber (1981): "On Grice's Theory of Conversation", in: Werth, Paul (ed.): Conversation and Discourse. London: Croom Helm London, 155–178.

Wilson, Deirdre and Dan Sperber (1991a): "Inference and Implicature", in: Davis, Steven (1991): Pragmatics. A Reader. New York/Oxford: Oxford University Press, 377–393.

Wilson, Deirdre and Dan Sperber (1991b): "Pragmatics and Modularity", in: Davis, Steven (1991): Pragmatics. A Reader. New York/Oxford: Oxford University Press, 583–595.

PhiN 12/2000: 12


1 I will refer to the first edition of Relevance. Communication and Cognition. Despite the revisions from the second edition (1995), the core of the theory that interests the aim of this paper hasn't been altered.

2 As in most texts from Sperber and Wilson, I will talk about the communicator as a she and the addressee as a he. I will talk about communicative intention (without making the distinction between informative and communicative intention proposed by the authors) because the latter concept includes the former and because in most cases identifying the communicative intention is enough to recover the informative intention. Thus the interpretive hypothesis will be a hypothetical inference about the communicator's communicative intention.

3 This point will be further discussed later in this paper.

4 I will only refer to nominal metaphors (according to Miller (1979: 233): "Nominal metaphors: BE (x, y) when an x is not a y."). My example is an adaptation from a comparison in Rumelhart (1979).

5 Gibbs (1994) distinguishes process and products of metaphor understanding. According to him, figurative language understanding does not require more effort than processing literal utterances, what can be shown by comparing processing times. Nevertheless this does not prove that mental processes involved are the same in both kinds of utterances; moreover, theory must account for the "phenomenological awareness of figurative meaning" (1994: 115). Gibbs suggests thus that linguistic understanding takes place through a succession of different moments that must be studied independently; speaker's feeling of figurative language may be a consequence not of early understanding processes, but of late products that result from figurative language understanding (like richness, indetermination and openendedness of meaning).

6 According to this theory, metaphorical blends may allow us to project inferences in different directions. However, metaphors characteristically present asymmetric topicality, i.e., "One of the inputs is topical and the other provides a means of re-framing the first for some conceptual or communicative purpose" (Grady et alii 1999: 13). This functional asymmetry of metaphor's terms is supported by empirical evidence (see Paivio 1979). Moreover, since other sources of information may be recruited to the blend (e.g. conceptual metaphors) the directionality of such mappings may also be inherited and structurally relevant.

7 This expression is used by Prandi, according to whom "l'instrument qui, comblant le décalage entre signifiés articulés et valeurs de message, adapte le contenu des énoncés à la cohérence du texte est le même qui confère une valeur communicative cohérente (coherent) aux énoncés contradictoires: le développement inférentiel." (Prandi 1992: 229).

8 "Intuitively, being relevant in a context is a matter of connecting up with the context in some way." (Sperber and Wilson 1986: 246)

9 In the postface of the 2nd edition of Relevance, the authors point out that Relevance Theory does not preview any distinction between true and false assumptions for processing purposes (according to their principles, both true and false assumptions can be relevant), which is seen as a point needing revision. That such epistemic conditions are not taken into account can be explained, I think, by the fact that Relevance Theory is not exclusively (nor essentially) concerned with descriptive uses of language (see their concept of interpretive resemblance). Interestingly, Sperber and Wilson underline that the truth of the conclusions drawn by the system is more important than the truth of the premises, and they significantly give fiction and analogical reasoning as examples. In fact, truth seems to be the adequacy to the system's goals (i.e. the outcome of the processing task, its effectiveness in the fulfilling of specific purposes) rather than the adequacy to a certain reality. In this sense, not only truth but cognition as well is envisaged in a pragmatic rather than in an epistemic way.

10 "The study of induction, then, is the study of how knowledge is modified through its use. [sic]" (Holland et alii 1993: 5).