2.1 The case for MentaleseThe classical solution to these three problems has been that beliefs are
relations to internal sentences, as Fodor has consistently argued (1975,
1978, 1987 Appendix; see also Field, 1978; Davies, 1991). For sentences
have contents which are systematically determined from the contents of
their component words, together with rules of combination. If you under-
stand the words, and know the rules of syntax, then you must be capable of
understanding new combinations of those words, never before encoun-
tered. And by the same token, of course, sentences are productive, in virtue
of the fact that rules of syntax are recursive. So the sententialist hypothesis
provides us with solutions to the problems of systematicity and produc-
tivity: thought is systematic and productive because there isa language of
thought(LoT).
Moreover (and providing us with a solution to the third problem also)
sentence tokens can have causal powers, by virtue of being physical
particulars. If beliefs and desires consist of sentences, or sentence-like
structures, encoded in some distinctive way in the brain, then there will be
no diYculty in explaining how beliefs and desires can be causes. (By way of
analogy, think of the manner in which sentences can be stored in magnetic
patterns on an audio-tape. These sentence tokens thencausethe sound-
waves which result when the tape is played.) And if we suppose, in
addition, that the mind is arranged so as to eVect computations on these
sentences in ways which respect their syntax, then the causal roles of the
sentences will respect their semantic properties. For semantics is, in part, a
reXection of syntax. And then we shall have explained successfully how
beliefs and desires can have causal roles which depend upon their semantic
contents.
For example, a logical concept likeandornotcan be carried by a lexical
item of some sort, distinguished by its capacity to enter into certain
characteristic patterns of inference. Roughly, ‘&’ meansandprovided that
the computational system within which it belongs ensures that it is go-
verned by the following forms of inference:(P & Q)]P; (P & Q)]Q;
andP,Q](P & Q). And a concept such asbus-stop, too, can be
constituted by some lexical item (bus-stop, as it might be) characterised
both by its causal connections with worldly objects (bus-stops), and by the
way in which itWgures in distinctive patterns of inference (such asbus-stop
]buses should stop) involving yet other lexical items from other parts of
the language of thought.
It is worth noting that the argument for Mentalese can be considerably
strengthened, by asking justwhypropositional attitudes should be sys-
tematic. Is it merely a brute fact about (some) cognisers, that if they are
Mentalese versus connectionism 195