Belief

1. Beliefs as sentences in mentalese: the language of thought

Fred’s belief in the Devil cannot literally be a relation between Fred and the Devil. Otherwise he could not have the belief, unless the Devil existed. One response is to treat belief as a relation to sentences. To believe is to ‘believe-true’ a sentence: Fred believes-true ‘The Devil exists’. But animals that lack a language have beliefs. My dog may believe that his master is home, or that it is time for a walk. Moreover, monolingual French speakers and monolingual English speakers may agree in what they believe, say, that it would be good if they knew more than one language, and yet they may not agree on which sentences they believe true. Finally, you might believe-true the sentence ‘The Devil exists’ and yet not believe that the Devil exists because you wrongly think that the word ‘Devil’ means ‘God’. In this case what you believe is that God exists while wrongly thinking that the sentence ‘The Devil exists,’ is a good sentence to use to express this belief. For these reasons, and others, belief is usually thought of as a relation to a proposition. A proposition is what is expressed by a sentence; it is what is in common between sentences in French and English that mean the same; the proposition expressed is what is grasped when you understand a sentence. Monolingual speakers believe alike by believing the same propositions; dogs have beliefs by virtue of believing propositions despite not having a language to express them; someone who believes that the sentence ‘The Devil exists’ is true while thinking that ‘Devil’ means ‘God’ does not thereby believe that the Devil exists because they are wrong about what proposition ‘The Devil exists’ expresses. These remarks slide over a lively controversy concerning the ontological status of propositions (see Propositions, Sentences And Statements). Our immediate concern will be with a popular view that gives sentences a prominent role in the account of belief, but in a way which avoids the problems just rehearsed.

According to the language of thought hypothesis (LOTH), not only do certain sentences serve to provide the propositional objects of beliefs (and thoughts in general) but, in addition, the beliefs are themselves sentence-like. A sentence may be viewed as made up of significant parts put together according to certain rules. In the same general way, according to LOTH, beliefs have parts put together in certain ways (see Language Of Thought).

How does LOTH mesh with the idea that beliefs are relations to propositions? The idea is that a belief’s propositional object is determined by how it is made up from parts which have representational or semantic properties – that is, the parts stand for things, properties and relations much as the parts of a natural-language sentence do (see Semantics). In English ‘biscuit’ represents certain things, and ‘crisp’ represents a certain property, and when we combine them together to form the sentence ‘Biscuits are crisp’ we get a sentence that makes a claim that is true or false according to whether or not the things have the property. This is how the sentence expresses the proposition that biscuits are crisp (see Compositionality). In the same way, there are brain structures that represent things and properties, and when these brain structures are put together in the right way we get, says LOTH, a more complex structure, a sentence in mentalese, that represents the things as having the properties – as it might be, the sentence of mentalese that says that biscuits are crisp, that expresses that proposition, and that thereby provides us with a token of the belief that biscuits are crisp.

This theory can allow that dogs have beliefs. Dogs might have a language of thought even though they do not have a public language. It can explain how monolingual speakers of different language can agree in belief – their sentences of mentalese may express the same propositions. It also provides an explanation of a number of phenomena associated with belief. First, it explains how what a person believes can be causally relevant to what else they believe and what they do. If you believe that Mary is at the party and then learn that Mary is always accompanied at parties by Tom, you will typically come to believe that Tom is at the party. What you believe combines with what you learn to produce a new belief. LOTH explains these causal transactions as transactions between the structures that are the various beliefs. Much as a computer processes information by manipulating electronically coded structures so we arrive at new beliefs by virtue of our brains manipulating the symbols of mentalese. Similarly, what we believe contributes to explanations of what we do. My belief that there is coffee over there together with suitable desires may lead me to move over there by virtue of its being a belief that there is coffee over there. LOTH accounts for this fact in terms of the causal influence of the sentences of mentalese on the causal path to bodily movement.

Second, LOTH explains the fact that typically one who can believe that Jill loves Mary can believe that Mary loves Jill, and in general if you can believe that aRb, then you can believe that bRa (the phenomenon known as systematicity). The fact that if you can believe that aRb, then you can believe that bRa is explained by the fact that the state that encodes the former is a re-arrangement of the parts of the state that encodes the latter. And, of course, this explanation generalizes to explain more complex cases.

Finally, LOTH can explain our ability to form quite new beliefs (the property known as productivity). Just as we can form new sentences by novel combinations of the relevant words of a public language, so the brain can form new beliefs by means of novel combinations of the relevant words of mentalese.

LOTH thus explains a lot of what needs to be explained. Nevertheless, there are two serious problems for this view as applied to belief.