10. Language of Thought

10 Language of Thought

According to Jerry Fodor’s hypothesis, we can say that the structure of internal states that have content must be language-like.

Propositional attitudes

PA are attitudes that we may take to propositions.

Propositions are what constitute the content of synonymous indicative statements in different languages.

Intentional states and content

PAs are also called ‘intentional states’ because the attitudes are ‘about’ the proposition’s topic.

If you want it to rain soon, your desire is about its raining soon.

The PA’s are said to have content: the proposition is that content.

We see that sentences are used to give the content of thoughts. According to LoT , thoughts have a sentence-like structure. Thoughts have a syntax in the same way that sentences do – the whole is made up of a finite number of parts in regular combinations.

For example, in the thoughts that ‘snow is white’ and that ‘snow is cold’ there is a common element, and it is what makes the thought about ‘snow’.

The Language of Thought Hypothesis

Need to consider two questions:

1. How does a M state get to express a proposition?

2. What attitude is being taken to that proposition?

Wrt 1. A M state is built in a combinatorial fashion from atomic parts that have semantic properties.

Compare: a sentence is built from words which have meanings.

Atomic and molecular representations

Suppose we have the thought that biscuits are crisp. The idea is that there is a brain state X representing a biscuit, and brain state Y representing crispness, and a syntactic operation in the brain that combines these two states in a way analogous to the combination that produced the predication ‘biscuits are crisp.’

This syntactic system is known as mentalese.

The thought ‘biscuits are crisp’ is structurally molecular, and has a truth condition: it is true if what X represents has the property that Y represents.

What gives these states their representational power is a tough question – which is also unsolved in the philosophy of language.

Belief boxes and desire boxes

What makes the attitude to the proposition the attitude it is? He LoT has a causal/functional story that you believe Y(X) if the state representing the molecule is in the right relations to stimuli and behaviour. Rather than having to clarify that each time we simply say that Y(X) is a belief if the token ‘X is a Y’ is in the ‘Belief Box’. Other boxes are for Hope, Desire, Fear, etc…

The contrast with functionalism

For F-ism the role of the state determines the content. For LoT , the roles only determine which box the content goes into.

In both cases the content is related to the roles of the thought. For F-ism, the behaviours will explain what state it is, whereas for LoT , the state it is will explain the behaviours observed.

So the proposition is identified independently of role and the mentalese sentence then determines the role of the state instantiating it.

Why are we supposed to believe in the language of thought?

Because it can explain things that we need explained. In particular, it explains

a. Systematicity and productivity

b. Similarities in behaviour arising from different thoughts

c. How thought evolves causally.

Systematicity

If you can think aRb then you can say bRa. If thoughts are in a LoT that’s not mysterious since if you can say one you can say the other, but if thoughts have no internal structure of the relevant (language-like) kind, then it is not known how they could have that property.

Productivity

Both language and thought are productive – meaning that we can think and say and understand an infinite number of new sentences or thoughts. We know that the compositionality of language explains it for that system, and we can only assume that language-like compositionality does it for thought.

Causes and effects

The thought that Fred is happy and at the party has similar effects to the thought that Fred is happy. The similarities are those related to Fred being at the party and the differences are those related to Fred being at the party. The semantic behaviour of the relevant sentences mirrors these similarities and differences. Let’s suppose that similar effects have similar causes. Then the languagelikeness of thought is the obvious reason for the identical behaviours.

Evolution of thought

Thoughts evolve. One leads to another. If Jones believes that if it is snowing then it is cold, and if he then learns that it is snowing outside, then it is expected that he will believe that it is cold outside. This is explicable in language terms because the syntactic form of the setences expresses their logical form. For a LoT theorist it is evidence that the structure of the states instantiating the propositions are such that they are causally related in the right way. Computers work like this, so it’s not inconceivable. But computers work with a language-like syntax. Suppose thoughts do too.

The Map Alternative

This is an Inference to Best Explanation. Are there other possibilities? One that’s proposed is the Map story, where mental maps do what mentalese sentences do for the LoT . We note that any one who believes that thoughts are internal states representing reality as being one or another way, has to believe that these states are structured and that the representational powers of the states depend on their structure.

The similarities and differences in representations correspond to similarities and differences in the representing states.

Compare to Arabic or Roman numerals representing numbers.

Note that the brain’s representational capacities seem indefinitely large; so its capaciy must be the result of combinations of a finite number of elements. The combination process is such that the represented is a systematic function of the elements of the representation and the way they are combined.

Minimalism about the language of thought

There are many ways that things can represent in virtue of their structure. Language is one way, but maps are another.

It is arrogant for LoT theorists to call theirs the ‘Representational Theory of Mind’.

The way maps represent

Maps are informationally rich

They differ from language in at tleast two ways.

1. They are rich in information. They give more information than they need to. For ex. a map that shows that Tasmania is an island also shows that it is south of the mainland. Language doesn’t have that problem.

2. [They are semantically smooth.] There are no obvious basic semantic units for a map. Nothing like words, for example. Nor are there any obvious semantic divisions in a map.

Do maps explain the phenomena?

Are mental states maps in the head (instead of sentences?) I.e. are they structured internal states which have the properties 1 and 2 above [of semantic richness and smoothness]?

First, know that a map can represent all the information in the world. (The world itself, if nothing more convenient comes to hand, is such a representation.)

So, could mental maps explain the things that are supposed to justify LoT ?

Systematicity

If a map can show Tasmania south of Victoria , then it could show Victoria south of Tasmania . So, yes, some systematicity is possible.

Productivity

An indefinite number of easily understood maps (of novel situations) can be constructed according to rules.

The causal argument

The similarities and differences in maps cause effects that are similar or different in the required ways.

Evolution of thought

There are ways to combine maps so that the resultant map is a causal product of the predecessor maps. Cruise missiles use map updates to stay on course. The map at t_n is a causal result of maps at t_i for i < n, and received information.

The objection from messiness

The map theory might be plausible, but how likely is it. Also the map theory tells us nothing about how the mind actually works, whereas the LoT does. It says it’s like a computer.

The map theory could have a very messy implementation.

But given that cognition is a natural product, wouldn’t we expect messiness rather than the LoT ’s unnatural neatness?