Addresses of correspondence: origgi@euro-edu.com; gershompalma@email.com

In the postface to the second edition of Sperber & Wilson [1986] the authors revise and refine several aspects of Relevance in to reply to objections and developments arising after the book was published. What interests us directly here is the more philosophical claim Sperber&Wilson make in order to connect relevance and truth (see § 3.2.1. in the Postface of [Sperber & Wilson 1995: 263]).

Truth is troublesome. In shaping a theory of truth one major issue is ontology: Which entities are the legitimate bearers of truth? Four main classes of entities are the most plausible candidates:

1. Utterances

2. Sentences

3. Propositions expressed

4. Mental states (beliefs)

By "utterance" we mean any speech act produced by a speaker. By "sentence" we mean the smallest linguistic unit which can be assigned a truth-value. By "proposition expressed" we mean the content, that is, the information conveyed by the utterance. By "belief" we mean any mental state which is supposed to represent a state of affairs.

What a theory of truth is supposed to do is to select the appropriate level of truth-bearers and to derive from it the semantic properties of other levels. In setting up "truth conditions" for Relevance Theory, Sperber & Wilson choose the fourth option, that is, mental states. As they say in the postface "when we say that an utterance is true, we really mean that its interpretation is true, and this is the output of a cognitive process of comprehension" [Sperber & Wilson 1995: 264]. Thus, according to them, truth is a property of states of our minds, and more precisely, of the mental outputs of our cognitive processes of interpretation. According to Relevance Theory, a process of interpretation starts from an input, which can be either an ostensive stimulus or an utterance, then processes it in order to determine a set of assumptions. The assumptions serve as input to a further processing (interpretation), which involves saturation, enrichment, computation of relevance, etc. This leads to conclusions, that is, a set of new or revised beliefs. These are the cognitive effects of the information processing.

The aim of connecting relevance and truth is to find some way of epistemically constraining our assumptions according to what Sperber&Wilson call in the postface the First Principle of Relevance: "Human cognition tends to be geared towards the maximisation of relevance" [Sperber&Wilson 1995: 260]. They choose to impose the epistemic constraint on the outputs of the process of interpretation, that is the cognitive effects a set of new assumptions has produced.

To be relevant an assumption has to be such that it brings about cognitive gains. A cognitive gain is, virtually by definition, some modification in the belief stock of a cognitive system increasing the ratio of true beliefs to false ones.

Constraining the outputs instead of, for example, the entire set of assumptions activated by an ostensive stimulus has a lot of advantages. It allows us to treat as relevant inputs false information that turns out to have true cognitive effects. One example of this kind of information-processing in which the truth of assumptions is immaterial to cognitive gains is fiction. A second one is well exemplified by the employment of false antecedents or of antecedents about which we had better remain silent in conditional reasoning (for example, taking as hypothesis in mathematics unproven conjectures , or worse, conjectures known to be formally indecidable, or taking as the premisse of a mathematical proof the negation of the hypothesis we wish to prove, as in reductio ad absurdum).

On the basis of these intuitions Sperber&Wilson propose the following replacements of the definitions (42) and (43) of Relevance to an individual :

(42) Relevance to an individual (classificatory)

An assumption is relevant to an individual at a given time if and only if it is relevant in one or more of the contexts accessible to that individual at that time.

replaced by:

(6) Relevance to an individual (classificatory)

An assumption is relevant to an individual at a given time if and only if it has some positive cognitive effect in one or more of the contexts accessible to that individual at that time.

(43) Relevance to an individual (comparative)

Extent condition 1: An assumption is relevant to an individual to the extent that the contextual effects achieved when it is optimally processed are large.

Extent condition 2: An assumption is relevant to an individual to the extent that the effort required to process it optimally is small.

replaced by:

(7) Relevance to an individual (comparative)

Extent condition 1: An assumption is relevant to an individual to the extent that the positive cognitive effects achieved when it is optimally processed are large.

Extent condition 2: An assumption is relevant to an individual to the extent that the effort required to achieve these positive cognitive effects is small.

It is clear that this definition has the advantage of not collapsing together truth and relevance. Although much of the information that we may process is not concerned with truth, its relevance depends on its causal role in the modification of our set of true beliefs. On the other hand, true beliefs may not be relevant per se, they may still have a crucial role in determining the relevance of new assumptions. Definition (6) seems to suggest that truth is a property of beliefs worth having, relevance is a property of ostensive stimuli worth processing. As we already said, we are concerned here with the cognitive principle of relevance: Relevance in cognitive terms is a property of assumptions, in communicative terms it is a property of interpreted utterances.

Still, there seems to be a problem with the new definition of relevance to an individual. Both relevance and truth are defined relative to a set of contexts that are accessible to an individual at a given time. Nevertheless, truth intended as a cognitive feature of our beliefs seems to be an eternal property, while relevance of assumptions is a contextual property of some temporally contingent cognitive states.

Connecting relevance and truth to the same context, the revised principle seems to force a distinction between what seems to be relevant and turns out not to be and what is relevant simpliciter. This is a very strong way of linking relevance and truth. Truth is a monotonic property. If a belief is held true at time t1 and then turns out to be false at t2 it means that it was false at t1. Is relevance also a monotonic feature of assumptions? Sperber & Wilson suggest that it is. As they say in the postface some assumptions may seem relevant but in fact turn out not to be. Example (4) in the postface stresses this point:

(4) Peter is a jealous husband. He overhears Mary say on the phone to someone, "See you tomorrow at the usual place". Peter guesses rightly that she is speaking to a man and infers, quite wrongly, that she has a lover and does not love him any more.

In this context Peter is inferring (and these are the cognitive effects of his interpretation of "See you tomorrow at the same place") that:

1 Mary said "See you tomorrow at the usual place"

2 Mary is speaking to a man

3 Mary is speaking to her lover

4 Mary does not love Peter any longer

In Sperber & Wilson [1995] the fact that Peter discovers at a later time that 3-4 are false entails that his assumptions aren’t relevant or are less relevant than they seemed to be. According to Sperber&Wilson, the fact that (3) and (4) are discovered to be false at a later time entails that even (2), which is true and relevant because it leads to rich contextual effects, seemed to be relevant but in fact was not. We believe that it was.

The burden of the discussion is to show that there can be cases of relevant assumptions for an individual that turn out to lead to false beliefs and nevertheless remain relevant even in the context of the revised set of beliefs entertained by the cognitive system.

Consider a variant case involving jealous Peter. Suppose as in (4) that Peter overheard the sentence "See you tomorrow at the usual place" uttered by Mary on the phone. Assuming Mary is speaking to a man and that that very same man is her lover, jealous Peter infers that Mary does not love him anymore. In Sperber & Wilson that Peter discovers at a later time that 3 and 4 are false entails that his assumptions are not relevant or that they are less relevant than they seemed to be.

However assumptions are such that when combined with general principles, maxims, attitudes and emotional states they are motivationally involved in the explanation or comprehension of behaviour. We endorse an intentional/psychological view of the mind according to which the laws of psychology typically express causal relations among mental states that are specified under intentional descriptions (of the kind: X believes that P). Therefore, ceteris paribus if X thinks that P is the case and desires that P be the case, and knows that the best means available to obtain P is to do a, then X does a. In Peter's case, the set of cognitive effects inferred from Mary's utterance contains obviously false beliefs. They function however as motivational inputs.

Consider the following scenario. After overhearing Mary on the phone, jealous Peter is consumed by a need to see who is stealing his wife. The next day he follows Mary, and discovers that she is meeting her father at the usual bar. Now, it seems overwhelmingly plausible that Peter will have to revise his set of beliefs about the fidelity of Mary and what not. At the same time it seems to us that we do not have any principled ground to say that Peter's assumptions should now be seen as not relevant or less relevant than thought prior to the discovery. On the contrary it is exactly their relevance that prompted Peter's actions. It is crucial for jealous Peter to be certain of Mary's fidelity. It is the very relevance of the beliefs entertained by him when overhearing her on the phone that justifies the extra effort he puts in discovering the truth of the matter. The overall final gain in cognitive benefit makes the trade off positive: his beliefs do not turn out to be any less relevant because they are false.

The reason is that even if false, the role played by (3)-(4) in terms of motivational structure of Peter's beliefs makes them relevant for him at both times. (3)-(4) are premisses triggering Peter's cognitive processes that lead him to action. It is because of (3)-(4) that he spies on Mary the next day. They are then relevant in term of positive cognitive effects being instrumental in delivering true beliefs.

However this introduces a difficulty for the revised principle of relevance and the close link between relevance and truth suggested in the postface. If a false belief can remain relevant even when discovered to be false (as our microscenario suggests it can), then, it would follow that relevance and truth are linked in a sense weaker than that of Sperber & Wilson.

If cognitive systems are selected by evolution to be efficient in delivering more true beliefs than false ones, then, ex-officio a system guided by relevance principles will tend to be a truth-producer. It would remain true though that to maximize positive cognitive effects false information continues to be relevant even when new contexts reveal its falsity. (3) and (4) are then relevant in terms of positive cognitive effects in the sense of being in the long run such that they are instrumental in delivering knowledge (true beliefs).

One possible way of fixing the claim problem could be to modify principle (6) as follows:

An assumption is relevant to an individual at t1 if and only if:

(i) either it has some positive cognitive effects in one or more of the contexts accessible to that individual at t1

(ii) or its effects at t1 can be used as the premisses of an inferential process having positive cognitive effects for the individual in some context accessible at t2.

This disjunctive form weakens principle (6), allowing cases in which contextual and eternal properties of assumptions can be kept separated. Relevant information leads us toward truth, but the context in which positive cognitive effects will be obtained may not be one of the accessible contexts at a certain time, as principle (6) seems to suggest.

J. Fodor (1987) Psychosemantics, MIT Press, Cambridge, Mass.

D. Sperber, D. Wilson (1986) Relevance, Basil Blackwell

D. Sperber, D. Wilson (1995) "Postface to the Second Edition" in Relevance, second edition, Basil Blacwell.