On pragmatic competence

Noam Chomsky's celebration - December 1998

Jean-Louis Dessalles

Ecole Nationale Supérieure des Télécommunications
46 rue Barrault - 75013 Paris - France


In 1637, Descartes postulated the existence of `Good sense'. As he described it, it now appears as a genuine Chomskian competence: "Good sense is, of all things among men, the most equally distributed". Following Noam Chomsky's insight which consisted in defining linguistic performance as resulting from a much simpler underlying competence, we pose the problem of determining the rational competence assumed by Descartes. From the observation that rational thought is what allows us to conceive relevant replies during spontaneous linguistic activity, we draw a parallel between relevance in conversation and relevance in thought. Relevance, defined by its relation to problematicity, thus appears as the core of our pragmatic competence.

Keywords: pragmatics, competence, relevance, conversation, frame problem, thought.


Noam Chomsky brought us to the idea that language relies on an innate basis, that we are all endowed with the same linguistic competence, despite the contingent fact that we speak different languages. He had of course in mind principally our syntactic ability which allows us to arrange words in appropriate order. In the sixties, this statement was far from obvious, and it is not yet universally accepted.

I want to make here a similar, though far more obvious, claim, which has many interesting consequences. We do have a pragmatic competence, which can be formally described, which is universal and which is unlikely to be learned at all. This competence allows us to use language in concrete situations, to utter relevant arguments, to be considered a competent conversant.

Very few studies have considered pragmatics as relying on an underlying, specific competence. Kasher (1991) considered the existence of pragmatic competence, but mainly at the social level (speech acts, social acts) or at the behavioral level (turn-taking in conversation). The main issue of pragmatics, namely why we choose to say what we say, is for the major part left out of the scope of his study.

It is usual to consider that the content of our utterances is governed by the general laws of reasoning and that these laws are by definition not specific to language. In Fodorian terms, they belong to a central cognitive system which is so general that its studies is for the time being untractable (Fodor 1983). According to this view, any formal universality we would discover in the content of conversations would be the result of the universality of rational mind, what Descartes (1637) called `le Bon Sens'.

Another influential school presents the human mind as designed to extract relevance from the environment and to offer it to others through language. Relevance is broadly defined as a property of stimuli that allows to draw inferences for an acceptable cognitive cost (Sperber and Wilson 1986). According to this view, we would choose to utter any message that can trigger rich and costless inferences in the addressee's mind. Nothing is said, however, about more specific abilities that would allow us to reach this goal. In other words, if we have a pragmatic competence, it is very general indeed, and relies on the sole relevance principle. Again, all other necessary ingredients, like the ability to draw inferences, are supposed to lie out of the language competence and are ascribed to a more general rational ability.

Though I agree with the initial point made by Sperber and Wilson about the central importance of relevance in pragmatics, I would like however to go one step further by describing relevance as constituting the core of our linguistic competence. I will make the following claims:

In what follows, I will first sketch some of these pragmatic rules. Then I will indicate how we may use such rules to account for the content of utterances. Lastly, I will suggest that our reasoning ability (Descartes' Bon Sens) is derived from our pragmatic competence, not the reverse.

Constraints on spontaneous language

The range of what can be said at a given point of a conversation is narrow and sometimes dramatically restricted, compared to the number of conceivable utterances which are merely related to the subject of the conversation. Whenever these constraints of relevance are violated (for experimental purposes or after a misunderstanding) we observe reactions like "Why do you say this ?" or "So what ?" that may even be aggressive. This is the kind of answer you get if, in the middle of a lunch, you utter with no apparent reason "67 times 127 is 8509". Using Franken's (1998) distinction between success and satisfaction, I will say that in such cases, communication is trivially successful but that it is not satisfied.

This phenomenon of non-satisfaction is crucial for the understanding of pragmatic competence. It plays the same role as grammaticality judgments in the determination of syntactical competence. "67 times 127 is 8509" is irrelevant in most contexts, though it is syntactically and semantically acceptable. We `do not see the point' in such an utterance, `it does not make any sense' (which is semantically false but pragmatically correct). The ability to make relevant points is a fundamental competence that all normal human beings share. Individuals who systematically lack it are considered insane. We are endowed with a genuine pragmatic judgment. Contrary to the grammatical judgment, it requires a specified context (there are indeed contexts in which "67 times 127 is 8509" is relevant), but it is equally constraining.

It can be shown that Sperber and Wilson's criterion of relevance is not accurate enough to account for our pragmatic judgment. They define relevance as a gradual property depending on inferences and cognitive cost. Contrary to what we expect from their relevance principle, however, the possibility of making easy inferences is often totally disconnected from the question of relevance. From "67 times 127 is 8509", we can easily infer that 8509 is not prime, that 8509 divided by 67 gives 127. These easy inferences do not change, even in an infinitesimal manner, the lack of relevance of the initial sentence. Worse, the description by these authors of our inference ability let suppose that we do perform such easy inferences (8509 is not prime, etc.) before judging that they do not bring anything worth attention. This point is at least doubtful, and is not supported by any psychological evidence. Listeners `know' that "67 times 127 is 8509" is irrelevant, they do not need to try all possible deductions from it. According to my own theoretical account, relevance and irrelevance are decided on the basis of formal properties that are present or absent in the situation made manifest by the utterance (Dessalles 1993, 1996).

In order to characterize what makes the difference between relevant and irrelevant utterances in everyday conversation, we may first observe that the behavior of conversants is very stereotyped. When involved in casual conversation, we are not aware of the very peculiar way we behave when introducing a new topic. Any new topic must make manifest a situation which is either:

Examples are "In the elevator, I was with a 7 ft. tall guy" (improbable), "Why is this tramp standing in line in front of this cash distributor ?" (paradox), "There is no immunity for Pinochet" (desirable). The claim is literally that there is no other possibility (we do not consider humor here). This explains why "67 times 127 is 8509", as topic introduction, is irrelevant in most contexts (try to find a context in which it introduces a relevant topic while avoiding the three categories above !).

A complete definition of these constraints on relevance, which applies to all utterances, not only topic introduction, states that any utterance must either refer to a problematic situation, or attempt to reduce the problematicity of a situation (problematics means here paradox or improbability or (un)desirability).

These constraints on relevant topics are claimed to be part of our pragmatic competence. We are naturally able to comply with these constraints, and we spontaneously require others to comply with them.

How utterances are computed

Any good model of grammatical competence must explain, at least partially, why a given sentence is arranged as it is and not otherwise. Similarly, a good model of pragmatic competence should account, at least partially, for the content of utterances. We will illustrate how this can be possible. Consider the following classical example (for more realistic examples, see (Dessalles 1998)):

A: I forgot to take my jacket.
B: Do you want me to close the window ?

According to classical accounts, we must suppose that B is able to make inferences like no jacket --> cold and cold --> close window. The reasons why these inferences are made while others, which are equally plausible on the basis of semantic association, are not even attempted, remains totally mysterious. For instance, B does not infer that the jacket must be hanging somewhere, or that A's jacket must have pockets, though these inferences may be done in other contexts. The main difference between these accounts (including Sperber and Wilson's) and mine is that the former use inferences to define relevance, while I consider that inferences are guided by relevance.

Let us consider the preceding example from another perspective. When hearing A's utterance, B tries to find some relevance in it. According to the theory, he must seek how A's utterance reveals a problematic context. In which way having no jacket is problematic ? In the current situation, we suppose that the only available problematic context is that A may be cold, which is undesirable. Now B must find a relevant reply. According to the theory, he must find a way to reduce or cancel the problematicity. He looks for causes of A's feeling cold, and he tries to interfere with these causal links (Dessalles 1998). The open window causes A's feeling cold. Offering to close it is thus relevant, by canceling the cause of the current problematicity.

We see on this simple example how relevance guides inferences, rather than merely resulting from them. B imagines that A is cold because he is looking for a problematic situation that A's utterance makes manifest. He then knows that he got A's point (no need to compute a cognitive effect / cognitive cost ratio, whatever these concepts may mean). The next step for B is to compute a relevant reply. B must look for some way to reduce problematicity. He performs an abduction from A's feeling cold, and comes upon the open window which is the likely cause of her trouble.

We see that inferences are not performed randomly, merely following the laws of free association. On the contrary. Each inference is triggered by a very specific need: the need to establish a problematic point, or the need to reduce a problematic point. Technically, such inferences are performed backwards (e.g. abduction: find a likely cause for a given effect). We are far from a scenario in which the human mind would spend its time in a silly search for all consequences of given premises. If A had simply stated "It's cold here", B would not have concluded that she probably had gooseflesh, or that her peripheral temperature may be lower than her internal temperature. In other words, only relevant inferences are attempted, which suggests that the functioning of the rational part of our mind is governed by relevance.

Pragmatic competence and rational thought

In The modularity of Mind, Fodor (1983) makes a radical distinction between peripheral modules and the central system. The latter is the place where rational thinking takes place. According to Fodor, there is no hope to understand the way rational thought is computed, mainly because there is no a priori limitation on what can be involved in such computations. Any piece of knowledge may be included as premise or conclusion of an inference. If we consider that inferences define relevance, as Sperber and Wilson do, then the way inferences themselves are performed is hopelessly mysterious, as stated by Fodor. In particular, there is no hope to solve the `frame problem': when in a position to perform an inference, we cannot go systematically through all we know to determine what information is interesting as premise. The frame problem is then to explain how we manage to delimit the set of beliefs that are useful to perform a rational task (Chiappe and Kukla, 1996).

The difficulty of this problem is dramatically reduced if we consider that relevance is not a consequence of our inferencing ability, but rather that it governs inferences. In other words, according to the thesis proposed here, relevance is a formal property of belief states that gives a frame to subsequent inferences. When a person makes manifest to you that she is cold, your inferencing ability is entirely directed towards finding likely causes for this state. Then, it is entirely focused on finding a way to interfere with the causal link. Fodor's argument of the `holistic' nature of central processes, which is supposed to make the study of the central system intractable, is an extensional argument. There is indeed no extensional a priori limit to what may come next in our reasoning, as there is no extensional limit on what may come as next argument in a conversation. However, there are strong intensional constraints on what may come next in both cases. Any element in a reasoning or in a conversation must be relevant. This relevance constraint is not a posteriori, as hypothesized by Sperber and Wilson. It is an a priori constraint, based on a formal property (problematicity) of a state of knowledge, and it gives an intensional frame on what will be computed. This relevance constraint is part of our pragmatic competence.

What is suggested here is that the competence that allows us to utter relevant arguments in conversation is also what allows us to perform rational thinking. Two principles are at work both in what can be observed from spontaneous use of language and in the functioning of the central system: (1) detect some problematicity ; (2) try to get out of it. These two principles, which are claimed to be constitutive of our pragmatic competence, guide both our talk and our thinking.

Conclusion: Chomsky's lesson

What Noam Chomsky taught us, beyond all the fundamental discoveries he made about language structure, is a method: behind any behavioral performance, try to discover an underlying competence, i.e. a general, simple, procedure that accounts for the apparent complexity of the performance. Then question the possibility that this competence is learned or is derived from a more fundamental one. If this is not the case, then make the hypothesis that this competence has an innate basis.

If we transpose this methodology to the pragmatics of language, then we come upon a new question which is to know whether a simple competence governs the choice of what we think and say. The usual objection: "Thinking and discourse are much too complicated to be put into simple equations" is unacceptable. On the contrary, the apparent complexity of these processes should urge us to look for an underlying simplicity, i.e. a pragmatic competence, as the apparent complexity of language drove Chomsky to the search for a syntactic competence.

Problematicity (the sensitivity to improbable, paradoxical or (un)desirable stimuli) certainly lies at the core of our pragmatic competence. A stronger claim is that our talk and our thinking are heavily constrained by the need to alternatively detect and resolve problematic situations. This claim suggests that our mind is performing only relevant inferences, i.e. inferences which attempt to establish or to reduce some problematicity. A consequence of this claim is that pragmatic competence is what best defines the rational side of mind. Individuals lacking this competence would be unable to utter relevant arguments or even to form relevant thoughts. Nobody has ever reported the existence of a culture on earth in which normal individuals would be lacking the ability to be relevant. For Descartes, there was no doubt that `Good Sense' was universal:

"I know of no other qualities that contribute to the perfection of the mind; for as to the reason or sense, inasmuch as it is that alone which constitutes us men, and distinguishes us from the brutes, I am disposed to believe that it is to be found complete in each individual".
Cognitive Science is now engaged in the scientific characterization of this universal competence. I suggested in this paper that this competence mainly consists in the detection and solution of problematic states of knowledge.


Chiappe, D. L. & Kukla, A. (1996). "Context Selection and the Frame Problem - Commentary on Sperber and Wilson's Précis of Relevance". Behavioral and Brain Sciences, 19(3), 529-530.

Descartes, R. (1637). Le discours de la méthode. Bordas, ed. 1965.

Dessalles, J-L. (1993). Modèle cognitif de la communication spontanée, appliqué à l'apprentissage des concepts - PhD Dissertation. Paris : ENST - 93E022.

Dessalles, J-L. (1996). "Pourquoi est-on, ou n'est-on pas, pertinent ?". Communication et Langages (107), 69-80.

Dessalles, J-L. (1998). "The interplay of desire and necessity in dialogue". In J. Hulstijn & A. Nijholt (eds), Formal semantics and pragmatics of dialogue. Enschede : University of Twente, TWLT-13, 89-97.

Fodor, J. A. (1983). The modularity of mind. Cambridge: MIT Press.

Franken, N. (1998). "Communication in relevance theory". Revista Aalicontina de Estudios Ingleses, 11, 59-74.

Kasher, A. (1991). "On the pragmatic modules: a lecture". Journal of Pragmatics, 16, 381-397.

Sperber, D. & Wilson, D. (1986). Relevance: communication and cognition. Oxford: Blackwell.