Relevance: Communication and Cognition 
by Dan Sperber and Deirdre Wilson.
Blackwell, 279 pp., £8.95, March 1986, 0 631 13756 4
Show More
Human Agency: Language, Duty and Value 
edited by Jonathan Dancy, J.M.E. Moravcsik and C.C.W. Taylor.
Stanford, 308 pp., $35, September 1988, 0 8047 1474 6
Show More
Show More

Collaborative writing is necessarily less common in the humanities than in science and medicine. And it seems rather less common now than in the Forties, when I was making a false start in medicine (the decade when Wimsatt and Beardsley collaborated on The Verbal Icon). Is it that we have become more competitive, more serious? ‘Not on your life, boy,’ my colleague John Hay in the anatomy lab used to say, but he would be helpfully turning a cadaver’s hand in response to my ‘Give me some palm, Olive’ and didn’t have to say he meant quite the opposite. Is it mutual supportiveness that makes collaborative writing work? Or the added range, the drawing on more and wider associations? In the case of Dan Sperber and Deirdre Wilson’s Relevance, collaboration has been able to draw on feminine and masculine resources – although I am far too afraid of my feminist colleagues to say which resources are which. Anyway, like Gilbert and Sullivan, or Somerville and Ross, Sperber and Wilson make a great team. Sperber’s On Anthropological Knowledge and Rethinking Symbolism, written solus, were brilliant essays: but Relevance is more systematic and technical, and Wilson may be guessed to sing its cantus firmus. The result is a decisive contribution to our thinking about language.

Relevance has a strategic importance, which makes its comparative neglect astonishing, and raises a question whether this may not be because of its inconvenience to some of our intellectual establishments. Perhaps, like Copernicus’s De Revolutionibus, it may take a considerable time to win recognition. For it offers nothing less than the makings of a radically new theory of communication, the first since Aristotle’s. From Aristotle to modern semiotics, all thinking about communication has been based on the code model, in which the communicative process is one of encoding and decoding. Recently, Paul Grice and others have turned to common-sense ideas, and to an inferential model based on the producing and interpreting of evidence. But that has not got very far. It is felt that Grice underestimates ambiguity, and that his maxims of conversation are altogether too question-beggingly convenient, too remote from real life. His principle of relevance remains so unparticularised that John Searle not surprisingly infers the impossibility of framing rules of conversation that correspond to actual experience. (Pragmatics, or the study of conditions affecting the use of language, is still often dominated by grammar-like ‘rules’.) Grice has not shown how unshared knowledge can come to be shared: his mutual expectation of truth, for example, hardly describes ordinary conversation. Meanwhile, Davidson’s anatomising of types of purity of intention, while interesting in itself, is similarly barren, so far as communication is concerned.

Most assume that a theory of communication must be based on either the code or the inferential model. But Sperber and Wilson hold that verbal utterances generally combine the two communicative modes, with coding-decoding subservient to inferential processes. They argue, however, that any purely inferential theory must ignore the actual diversity of psychological processes (to say nothing of the fact that decoding is often automatic). They aim to improve on Grice’s common-sense theory of 1957, while reversing his subsequent moves away from psychological plausibility.

On the other side, Sperber and Wilson concede that codes are very useful tools; that most human communication uses coding is no awkwardness to them. (A purist might fault them, in fact, for accepting too readily that codes – in the usual sense – operate at all.) They do not dispute that the code model explains how in principle communication could be achieved. But they show it to be inadequate descriptively: communication is not achieved by coding alone. Sometimes they make their case by persuasive Wittgensteinian exposition, rather than by close argument. But at crucial points they go far beyond Grice in working down into empirical psycholinguistic detail. Sometimes, indeed, they do not so much speculate cogently as actually explain. To an extent unusual in pragmatics, they win agreement by describing examples with such psychological subtlety that one comes to see the description as having explanatory force. (This in spite of examples that in themselves are rather artificial.)

An instance of the looser method is their example of communication achieved without coding (which I adapt freely, to show it more representative than Searle might say). Peter asks Mary, ‘How are you getting on with Lacan?’ and Mary responds by showing him a bottle of aspirin. Her behaviour is not coded: no convention says that displaying a bottle of aspirin means one is having problems with Lacan. Now, although Sperber and Wilson trace many steps taken in interpreting Mary’s action, they do not consider the objection that gestural codes may subsidiarily have contributed. But their argument could easily be supplemented. For, while we speak loosely of ‘body language’, gesture is not precisely coded, and Mary’s gesture (if it is one) would not be enough, qua code, to communicate anything like as much as the inferential process warrants.

An instance of Sperber and Wilson’s more closely argued mode would be their distinction between the fact that utterances take on the grammatical constraints of language codes and the supposition that thoughts are ‘encoded’ in utterances: ‘We see verbal communication as involving a speaker producing an utterance as a public interpretation of one of her thoughts ... Let us say that an utterance is an interpretive expression of a thought of the speaker’s, and that the hearer makes an interpretive assumption about the speaker’s informative intention.’ The distinction is often glossed over; yet it amounts to an entire phase of communication. Sometimes, particularly in writing, the search for words can be long and difficult. This crucial distinction is a fundamental reason why coding has to be subsidiary to inferential processes. What is encoded is never a thought, but only a representation of one.

The main contribution of Relevance, and the Ariadne’s thread of all its explorations, is the principle of ‘optimal relevance’. This is described in very specific psychological terms, and appears at last to satisfy Searle’s requirement for a non-circular account of relevance (the account he found lacking in Grice’s pragmatics). For Sperber and Wilson, communication depends on a shared cognitive environment and shared ‘domains of assumption’ about the world – assumptions that may be more or less strong, manifest, specific, universal, perceptible, or entailed by inference. The expectation shared by communicator and addressee, however, is not necessarily one of sincerity or truth, but simply one of easiest relevance: namely, that an immediate assumption (for example, about the previous utterance) will be modified. In ways minutely analysed in Relevance, the addressee’s mind continually selects meanings (consciously or unconsciously) by this criterion of optimal relevance – and is expected to do so by the communicator. Thus my classmate John Hay could infer that I was not merely feeding him a wretched catch-phrase, since it was too old to be relevant for its humour alone. The optimal relevance was to his immediate activity of positioning the cadaver’s hand: to its greasily oily awkwardness.

All this recalls Grice’s maxims of conversation (as well as the common-sense adage ‘take the obvious’): but it is worked out in a more penetrating and far-reaching way. Sperber and Wilson describe relevance generally, and yet precisely enough for it to have great explanatory power. It seems to me that everyday communication may, in fact, be achieved by such optimally relevant selection. It is a flaw in their exposition, however, that they do not dispose of alternatives to the ‘easiest path’ hypothesis. Their case would have been stronger, for example, if they had considered the possibility that particular circumstances might justify selecting less than optimally relevant meanings, or even justify random search processes, offset by multiple redundancies.

On the optimal relevance model, it is possible to infer very subtle implications indeed. Perhaps Sperber and Wilson’s greatest triumph is to have extended pragmatics to very weak implicatures – ambiguities, implications of implications, doubtful figures – of the sort that make up the richest part of conversation. Such oblique utterances are often more communicative than the imagined univocalities of philosophers and speech-act theorists. Here Relevance goes beyond the spurious simplicities of semantic and speech-act-theoretical approaches, entering the more complex pragmatic strategies of real life. Sperber and Wilson are at their best, in fact, where previous pragmatic approches have been weakest: in dealing with ‘vague’ utterances. Their account of irony, for example, is more rigorous than Grice’s, since it allows far less freedom of assumption.

If the theory of communication sketched in Relevance is as significant as I take it to be, this may be a suitable juncture to consider some of its implications for literary criticism. (Several critics, notably T.A. Shippey, have already applied modern pragmatics to good effect.) Sperber and Wilson do not spell out these implications, but some are fairly obvious. Relevance theory can have little joy for structuralists and deconstructionists, both of whom (in opposed ways) identify communication as coding-decoding. Structuralists stress the power of codes so extremely as to deny the communicator any role, while Derridadaists – and sometimes Jacques Derrida himself – stress the frailty of coding through regress of signification, and regard communication itself as enmeshed in undecidabilities. From the present point of view, Derrida’s asking by what authority a word indicates one meaning rather than another can be seen as an unwitting reductio ad absurdum of the old Aristotelian theory of meaning. Communication fortunately does not depend on meanings authorised by philosophers, nor even by the French Academy.

If inference is the more fundamental communicative process, contemporary methods of criticism all need to be thought through afresh. Not that the whole structure of structuralism and construction of deconstruction must fall like houses of cards. Both have led to insights, and lent useful support to demystification. But their doctrines (like Lacan’s) seem to have merely symbolic value: as explanations, they are comparatively superficial. (Even elementary histology is enough to inform one that the genetic code is not really ‘inscribed’ as a text.) New historicism and cultural materialism may be regarded as less inevitably compromised by the code model – it is communication itself that they reject. Sometimes, it is true, they neglect relevances more immediate than politics: but at other times they valuably restore neglected (political) domains of assumption.

The implications of Relevance for students of older literature are particularly significant. Meanings in the past are often said to be inaccessible, because of changes in the language code; and certainly linguistic changes often make old writing difficult, especially where abstract words are involved. One might deem the cipher of the past unbreakable, if the coding-decoding process were all. But once the fundamental role of inference in communication is grasped, many new interpretative resources become available. Interpreting past utterances need not entail decoding antique experiences inaccessible in their alterity: it may be enough to supply necessary associations and infer optimally relevant senses. True, old assumptions may be forgotten or unobvious. But while domains of assumption have changed in numerous ways, they have not changed beyond recognition. It is not in principle a hopeless task to learn enough to make sound inferences from the assumptions formerly taken to be optimally relevant. These are likely, in fact, to have been fairly obvious (which of course is not to say that the meanings of total utterances were obvious). Moreover, since domains of assumption have internal self-coherence, local progress in interpreting an old text enables further inferences about others.

But can we assume that literary works qualify as utterances like those of ordinary language communication? Sperber and Wilson argue that the principle of optimal relevance explains even cases of non-semantic utterance (as when a sentence is used merely to exemplify acoustic properties). And certainly there is no question of literature’s constituting a special case by virtue of its tropic nature. Indeed, they propose that ‘the notion of a trope, which covers metaphor and irony and radically distinguishes them from “non-figurative” utterances, should be abandoned altogether: it groups together phenomena which are not closely related and fails to group together phenomena which are’ (‘Mood and the Analysis of Non-Declarative Sentences’, Human Agency p.88; cf Relevance, p. 243). For them, non-literalness constitutes no departure from a communicative norm, since ‘the expectation crucial to communication is one not of [literal] truthfulness but of optimal relevance.’ It is not obvious, however, just how their approach should be applied to literary works, which may be utterances of sorts, but evidently lack pragmatic contexts in the ordinary sense. After all, as they write in their essay in Human Agency, an optimally relevant utterance is ‘one that communicates enough contextual implications to be worth the hearer’s attention, and puts the hearer to no unjustifiable effort in obtaining them’. Can such an ‘easiest path’ principle work in literature, with its exploitation of heightened states, and even of deliberate difficulty?

Sperber and Wilson argue for the importance of pragmatics over semantics: ‘the linguistic form of a non-declarative utterance’, for example, ‘vastly underdetermines the way it is understood. In this, as in every other aspect of interpretation, considerations of optimal relevance play a vital constraining and enriching role.’ As pragmatics contributes more to the understanding of such utterances, less need be attributed to semantics. Indeed, as they put it in the same essay, ‘The characteristic linguistic features of declarative, imperative or interrogative form merely encode a rather abstract property of the intended interpretation: the direction in which the relevance of the utterance is to be sought.’ But is not literature deficient in precisely the gestures, facial expressions and tones of voice, to say nothing of the relationship and shared situation, that allow hearers of real-life utterances to draw pragmatic inferences?

Or are we to suppose that, in compensation, literature establishes its own special context of literary conventions, fashions, allusion, formal structuring, and confirming resonances or redundancies? Genres, for example, are perhaps to be regarded as constituting, in Sperber and Wilson’s terms, shared domains of assumption. And all these traditions and forms may amount to a contextual situation that allows quasi-pragmatic inference. Readers attend in different ways to avant-garde and naive writers (to take only one obvious pragmatic adjustment). Heightened attention is given to literature, moreover; and there is an unwritten contract between writer and readers that narrows the search for relevant meanings. Indeed, literature may be not so very unlike Sperber and Wilson’s example of non-reciprocal communication, in which Mary leaves her broken hair-dryer for Peter to find (rather as Orlando leaves his poems on trees), thus avoiding the possible refusal that might have followed if she had modified their mutual cognitive environment by asking directly. Writers may leave various degrees of understanding open to different readers, or readers in various moods. But the communicative intention is often mutually manifest, and literature may not be so very remote from ordinary communication after all. Strikingly, the difficulties of interpreting literature seem mostly to arise from the richness of its communicative success. For poetic effects such as figurative language allow many weak implicatures, and call for complex inferences.

For some time, critical approaches to the problem of multiple weak implicatures have been dominated by permissive ideas based on semantics and the coding model. Structuralist and subsequent schools have taken for granted, not only (and rightly) that there are no ‘words on the page’ – that literary meanings have to be constructed by the reader – but also (and wrongly) that construction depends purely on semantic processing, free from all but grammatical constraints. Meanings have been arrived at by freeplay rather than inference. But once it is conceded that all communication involves inferential interpretation, the nature of literary criticism will be seen in a rather different light. If what constitutes a meaning is optimally relevant content, meaning can hardly be arrived at by freeplay alone (although that may be a useful local tactic for generating options). State-of-the-art interpretation is likely now to depend more on pragmatic inference. And the paradoxical result may well be that as pragmatics supplements and enriches semantics, ampler meaning will eventuate. There will be more initially ambiguous enrichments, but fewer eventual ambiguities. Indeed ‘every utterance has at most one interpretation consistent with the assumption that the speaker was aiming at optimal relevance: the first interpretation tested and found consistent with this assumption is the only such interpretation. In deciding which contextual implications the speaker intended to communicate, the hearer should thus select the minimal (because most easily recoverable) set of implications consistent with the assumption that the speaker was aiming at optimal relevance. In other words, he should take the utterance to be fully literal only if nothing less than full literalness will do.’ After a cycle of discordant or negative interpretation, hermeneutics may be entering another cycle of homoiotes.

However that may be, the repercussions of Relevance are likely in the long run to be great – felt first, perhaps, in the pragmatics of conversation, the philosophy of language, and reader-response criticism, but also in many other activities: construction of memory-models, pedagogy, machine learning and (doubtless) advertising and propaganda. I do not mean to give the impression that Sperber and Wilson amount to a composite Aristotle. But any new Aristotles who come along to attempt a modern theory of communication are likely to begin with positions a good deal closer to Sperber and Wilson’s than to Derrida’s or Searle’s. If, that is to say, they are still interested, then, in theories of meaning.

Send Letters To:

The Editor
London Review of Books,
28 Little Russell Street
London, WC1A 2HN

letters@lrb.co.uk

Please include name, address, and a telephone number.

Read anywhere with the London Review of Books app, available now from the App Store for Apple devices, Google Play for Android devices and Amazon for your Kindle Fire.

Sign up to our newsletter

For highlights from the latest issue, our archive and the blog, as well as news, events and exclusive promotions.

Newsletter Preferences