Why Mental Representations Have To Have Contents,

100 J. A. FODOR
Why Mental Representations Have To Have Contents Quantification overtheoretical
commitment to—mental representations is what cognitive science
has in common with the Classical tradition in epistemology as it developed in
both Rationalist and Empiricist versions. In particular, according to both
Classical and current theories, behavior is typically the effect of mental
processes, mental processes are typically causal sequences of mental operations,
and mental operations have mental representations as their domains. This
general picture, which I have elsewhere (see [3]) called the Representational
Theory of Mind (RTM) is presumably well known and I shall spare the reader
further exposition.
However, current cognitive science4 differs in important respects from
earlier versions of RTM. In particular, the contemporary movement is explicit
in endorsing the claim that mental processes are computational: mental
operations apply to mental representations in virtue of "formal" or "syntactic"
(or, anyhow, nonsemantic) properties of the representations.5 So, to the
extent that one assumes that the content of a mental representation is some
sort of construct out of its semantic properties, it follows that mental operations
are defined without reference to the content of the representations they
apply to. (I am putting this very loosely; the question of the relation between
the semantic properties of a representation and its content will loom large later
on and I do not want to prejudice the issue at this stage.)
Now suppose (as above) that behavior is to be explained by exhibiting
its contingency upon the mental processes that cause it. And suppose that
every mental process is a sequence of mental operations and that mental
operations apply to mental representations in virtue of the form of the
representations. Then it looks as though it does not matter whether the notion
of the content of a mental representation is in jeopardy since it looks as though
that notion is never going to be required in order to give the explanations
that cognitive scientists want to give. The idea that mental operations are
formal can thus be taken to imply that the content of a mental representation
is a dispensible construct, at least for the purpose of cognitive science. Indeed,
some philosophers have read the moral in just this way: see, for example,
[12], [13], and some of the remarks in [10].
I chink, however, that this line of argument is not well-advised. Roughly,
the point is this: we want not just to be able to characterize the causal chains
upon which behavior is contingent, but also to state such true and counterfactual
supporting generalizations about the etiology of behavior as there are
to be stated. But, to put the point in a nutshell, it looks as though such
generalizations typically hold in virtue of intentional properties of the
behaviors that they subsume, and it looks as though we shall need to advert
to the content of mental representations as part of our account of the intentionality
of behavior.
Notice that just about all of the familiar, rough-and-ready examples
of generalizations about the way in which behavior is contingent on mental
states and processes appear to make essential reference to intentional properties
of the behaviors they apply to. Consider, for an instance, that paradigm of
spruced-up common-sense psychology, the practical syllogism. It says something
about how someone will act (or, if action is thwarted, what the agent
will try to do; or, at a minimum, about the properties that the agent would