Dan Sperber and Deirdre Wilson. 1997. Remarks on relevance theory and the social sciences. In Multilingua 16 (1997): 145-51.

“Some commentators have described the relevance-theoretic approach to communication as psychological rather than sociological. Often, this is intended as a criticism. We would like to respond by reflecting in very general terms about possible interactions between relevance theory and research programmes in the social sciences…”


Dan Sperber and Deirdre Wilson

Some commentators have described the relevance-theoretic approach to communication as psychological rather than sociological. Often, this is intended as a criticism.1 We would like to respond by reflecting in very general terms about possible interactions between relevance theory and research programmes in the social sciences.

It is true that most relevance-theoretic work so far2 has largely ignored aspects of communication discussed in the sociological literature: role playing, turn taking, class, gender, race and power, for example. Instead, the focus has been on issues typically discussed in psychology: attention, memory, intention, inference. However, this is more a reflection of what seems to us a sound initial research strategy (which is likely to change as the field develops) than some silly anti-sociological bias.

It is not enough to define sociology as what sociologists usually do and psychology as what psychologists usually do: such definitions show too much respect for institutional boundaries. In fact, one of us is a social scientist whose interest in our joint project sprang from his desire to rethink the place of meaning and communication in the study of culture and society (see Sperber 1975). More generally, we would like to help bring about a redefinition of disciplinary boundaries, including those between the cognitive and social sciences, and we see our work as contributing to both domains.

Communication is a paradigm case of social interaction, and any theory of communication is a theory of the most ubiquitous social phenomenon. All sociological theories either explicitly or implicitly appeal to some view of communication: any sociological description involves an account of who communicates what, to whom, when, why, and how. What social scientists have had to say on the ‘how’ question has mainly consisted in distinguishing various media and forms of communication, and showing which are used to communicate what. The question also needs to be approached at a more fundamental level, where the issue is, what basic mechanisms make communication possible at all? To the extent that social scientists have been concerned with this fundamental issue, they have generally accepted that some version of the code model explains the very possibility of communication, whatever the medium.3 This is not surprising since, until recently, the code model was the only one available. Social scientists have often quite rightly stressed how little this model explains of the complexity of social communication. However, their reaction has been not to look for an alternative model at the same fundamental level, but to look for much richer explanations at higher levels.

A coding-decoding mechanism, left to operate unhindered and in a vacuum, would create a copy of the communicator’s meaning in the recipient’s mind. The sociologically crucial fact that contents get transformed, distorted, lost or suppressed in most social communication cannot be explained in terms of such a basic mechanism. Unless some other mechanism is envisaged, those effects of communication that go beyond, or against, mere decoding must be explained entirely in terms of who is communicating what to whom, and why.

Relevance theory takes a different approach. It characterises communication as a different type of social process than does the code model. From the point of view of the code model, communication can be described as social because it is a form of interaction, but the abilities it presupposes in communicators are signal-oriented rather than other-oriented. All an encoder has to do is produce a signal; all a decoder has to attend to is a signal. This can happen without either communicator having any notion that there are other beings like itself, with mental states and capacities, or even that it is itself such a being. Thus, bees are social animals and code-communicators, but there is no reason to credit them with any form of subjectivity, let alone intersubjectivity.

If human communication were of the code type, self-awareness and other-awareness would not be presupposed by the very act of communication, although they might be fostered, or even engendered, by human communicative practices, or by the contents that get communicated. By contrast, if human communication is of the inferential type, it presupposes and exploits an awareness of self and others. Inferential communication is intrinsically social, not just because it is a form of interaction, but also, less trivially, because it exploits and enlarges the scope of basic forms of social cognition. Right or wrong, this is a strong sociological claim.

Communication, as we characterise it, is definitely not a copying process. Hence, the transformations, distortions and losses typical of social transmission are to be explained, not just as effects of the social uses of communicative practices, but as normal effects of communication itself. Indeed, what calls for some non-trivial explanation is the fact that extended chains of communication can succeed in distributing throughout a human population those enduring representational contents that are readily identified as a people’s culture. All past theories of culture have relied on a code model of human communication (which was the only one available). A theory of culture based on an inferential theory of communication should be interestingly different (see Sperber 1996).

Verbal communication typically conveys much more than is linguistically encoded. Here we have in mind not just the enrichment of linguistic meaning and the derivation of standard implicatures, but also ostensive or non-ostensive uses of the act of communication itself to convey claims and attitudes about the social relationship between the interlocutors. Sociolinguists have been particularly concerned with these aspects of verbal behaviour, and have studied them with sophistication and insight. In Relevance, we largely ignored them. We did not mean by this to deny their importance, or to express a lack of interest in the issues or the work done; we merely felt that, at this stage, we could best contribute to the study of human communication by taking it at its most elementary level, and abstracting away from these more complex aspects.

The idea that one can focus on micro-processes, abstracting away from the more complex social dimensions of communication, is itself often criticised. Many social scientists are holists, who deny the legitimacy of what they see as ‘reductionist’ approaches. However, any holist may one day be accused of illegitimate abstraction by someone claiming to be ‘more holist than thou’. No approach is so holistic that it can really take in the full range of interconnected facts, without abstracting away from whole dimensions of the reality being studied.

For instance, we have often been criticised (along with Gricean pragmatists and formal linguists) for using artificial examples designed to illustrate theoretical points, rather than data drawn from proper recordings of real-life verbal exchanges.4 We would argue that artificial examples elicit genuine and systematic intuitions, in both our readers and ourselves, that do provide evidence for or against a variety of theoretical claims. We would also readily grant that real-life verbal interactions exhibit properties and raise problems that tend to be ignored, or only sketchily dealt with, by those who, like ourselves, use mainly artificial examples. But then those who make audio recordings and careful transcriptions of authentic conversations can in turn be criticised for abstracting away from the bodily movements and visual monitoring involved in typical verbal exchanges. Should video recordings be used? Certainly, for some purposes; but they still abstract away from much of the personal and common history of the participants, which surely contributes to shaping and giving meaning to their interaction. And so on.

The real issue is not whether it is acceptable to abstract away from some aspects of reality, even though they play some role in the phenomenon studied. Everybody has to do that. Nor is there any serious argument to show that less abstraction is always better. Good abstractions contribute to relevance, and that is what makes them good abstractions. They do this by focusing attention on data about which interesting generalisations can be made. In a single domain, there may be several different levels of abstraction, each giving rise to interesting generalisations. We assume that this is so in the study of communication.

So far, the contribution of relevance theory to the study of human communication has been at a fairly abstract level. However, it seems to us to have potential implications at a more concrete sociolinguistic level. How does verbal communication convey social claims and attitudes, and play a role in the ‘negotiation’ of mutual relationships? (Incidentally, ‘negotiation’ – a key term in sociolinguistics – is an apt metaphor, but one that needs unpacking if the insight underlying it is to be turned into an explanatory account.) Here again, to the extent that the ‘how’ question has been considered at all, some version of the code model is generally assumed to apply. There are exceptions: we agree with Levinson when he says that

‘… only the most restrictive definitions of pragmatics would draw anything like a clear boundary between sociolinguistics and pragmatics… Indeed, pragmatics and sociolinguistics share many areas of common interest, and sociolinguists have contributed much to certain areas of pragmatics… However, pragmatics in turn has much to contribute to sociolinguistics; for in trying to understand the social significance of patterns of language usage, it is essential to understand the underlying structural properties and processes that constrain verbal interaction’ (1983: 374).

The recent work of Herb Clark (1992) illustrates one way in which inferential pragmatics and sociolinguistics can contribute to one another.5

Some of the ideas and distinctions proposed within the relevance-theoretic framework might be of use in this perspective. In particular, we have argued that some of what is conveyed is communicated ostensively, and some is not communicated at all,6 but merely made manifest; and this in turn is sometimes intentional and sometimes unintentional (see Wilson & Sperber 1993). Even in these non-ostensive cases, considerations of cognitive relevance determine the degree of manifestness achieved, and the likelihood that the information will be picked up, and therefore play a role in further interaction.

More generally: the most basic claim of relevance theory (the First, or Cognitive Principle of Relevance in Sperber & Wilson 1995) is that the pursuit of relevance is a constant factor in human mental life. This psychological claim has immediate sociological consequences. The existence of this constant factor is what makes it possible for an individual to infer with a modicum of success what other individuals are paying attention to, and even what they are thinking. Predictably, people are paying attention to the newly accessible information that seems most relevant to them, and are having thoughts inferentially derived by combining this information with the most relevant contextual information available. Try to imagine what human social life would be like if humans were not predictable to each other in this reliable way.

Because humans have such predictive purchase, they can act on each other’s mental lives. They do this by manipulating each other’s expectations of relevance. A second basic claim of relevance theory (the Second, or Communicative Principle of Relevance in Sperber & Wilson 1995) is that ostensive communication creates uniquely precise expectations of relevance in others. Ostensive communication is the most important means by which the psychological tendency to maximise relevance is socially exploited. There are also non-ostensive methods; while their effects on the exact content of other people’s thoughts may be less precise, they may still be of major social significance. For instance, non-ostensively produced evidence of deference or submission, or of expectations of deference or submission from others, is highly relevant and effective in maintaining or challenging power relationships. Most existing forms of human interaction involve ostensive and non-ostensive aspects simultaneously. The actual psychological and micro-sociological mechanisms involved in non-ostensive forms of interaction have not been adequately described. We believe that, here too, relevance theory might be of some help.


Austin, J.P.M. 1989 The dark side of politeness: A pragmatic analysis of non-cooperative communication. University of Canterbury, New Zealand, PhD thesis.

Brown, P. & Levinson, S. 1987 Politeness: Some universals in language usage. Cambridge University Press, Cambridge.

Clark, H. 1992 Arenas of language use. University of Chicago Press, Chicago.

Downes, W. 1984 Language and society. Fontana, London.

Forceville, C. 1994a Pictorial metaphor in advertising. Vrije Universiteit, Amsterdam.

Forceville, C. 1994b Pictorial metaphor in billboards: Relevance theory perspectives. In Muller, J. (ed.) 1994 Towards a pragmatics of the audiovisual. NODUS, Munster.

Hodge, B. & Kress, G. 1988 Social semiotics. Polity Press, Cambridge.

Jucker, A. 1988 The relevance of politeness. Multilingua 7: 375-84.

Leach, E. 1976 Culture and communication. Cambridge University Press, Cambridge.

Levinson, S. 1983 Pragmatics. Cambridge University Press, Cambridge.

Levi-Strauss, C. 1966 The savage mind. University of Chicago Press, Chicago.

Mey, J. 1993 Pragmatics: An introduction. Blackwell, Oxford.

Mey, J. & Talbot, M. 1988 Computation and the soul. Journal of Pragmatics 12: 743-89.

Power, R. & Dal Martello, M. 1985 Methods of investigating conversation. Semiotica 53: 237-57.

Schiffrin, D. 1994 Approaches to discourse. Blackwell, Oxford.

Sperber, D. 1975 Rethinking symbolism. Cambridge University Press, Cambridge.

Sperber, D. 1996 Explaining culture: A naturalistic approach. Blackwell, Oxford.

Talbot, M. 1994 Relevance. In R. Asher (ed.) The encyclopedia of language and linguistics 8: 3524-7.

Tanaka, K. 1994 Advertising language: A pragmatic approach to advertisements in England and Japan. Routledge, London.

Taylor, T. & Cameron, D. 1987 Analyzing conversation. Pergamon Press, New York.

Watts, R. 1989 Relevance and relational work: Linguistic politeness as politic behavior. Multilingua 8: 131-66.

Wilson, D. & Sperber, D. 1993 Linguistic form and relevance. Lingua 90: 1-25.

Zegarac, V. & Clark, B. forthcoming A cognitive account of phatic communication. To appear in Journal of linguistics.


1. According to Mey (1993: 81-2), ‘Relevance theory … does not include, let alone focus on, the social dimensions of language’. Talbot (1994: 3525-6) calls relevance theory ‘an asocial model’, and comments: ‘A drawback of the model … is its lack of any social element… Within the framework given, there is no way of discussing any divergence of assumptions according to class, gender or ethnicity’. See also Mey and Talbot 1988: 746-7.

2. There are some exceptions: see for example Austin 1989; Zegarac and Clark forthcoming; Downes 1984; Forceville 1994a, 1994b; Jucker 1988; Power & Dal Martello 1985: 251-3; Tanaka 1994; Watts 1989.

3.  See, for example, Levi-Strauss 1966; Leach 1976; Hodge & Kress 1988. Brown and Levinson’s work on politeness (1987) is a rare exception in approaching an important aspect of culture from a Gricean point of view.

4.  For useful surveys of social and cognitive approaches to ‘discourse analysis’ and their perceived strengths and weaknesses, see Taylor & Cameron 1987; Schiffrin 1994.

5.  Clark seems to believe that his evidence and arguments go against relevance theory. We don’t see why.

6.  Except in a very loose sense of ‘communication’. Many sociolinguists treat the term as applying both to information intentionally ‘given’ and information unintentionally ‘given off’. From this perspective, we have been criticised for ‘looking at only half the picture of language’:

‘Communication cannot be restricted to what people intend to communicate. People communicate more than they intend. Sperber & Wilson’s model crucially rests on the exclusion of precisely this’ (Mey & Talbot 1988: 746).

The issue is not, of course, whether non-ostensive forms of information-transmission exist, but whether they should be treated as communication. In Relevance, we argue that, in general, unintentionally transmitted information is subject merely to general cognitive rather than specifically communicative constraints: in other words, it falls under the First, or Cognitive Principle of Relevance rather than the Second, or Communicative Principle.