naturalise content, in a way which would be tantamount to abandoning
the project altogether.
There would appear to be two possible ways forward for functional-role
semantics, here. One would be to borrow a leaf from the teleo-semanticist’s
book, and say that a process is inferential just in case it occurs when our
cognition operates in the way that it issupposedto. Then the transition
from ‘P & Q’ to ‘P’ will come out as inferential, whereas the transition from
the thought, ‘Madonna will visit SheYeld’ to shaking hands will not,
just as intuition dictates. For theWrst transition is either innate (and
selected for) or at least maintained in cognition because of its success;
whereas the second is neither. But notice that this need not commit us to
saying that thoughts themselves have functions. Rather, we would be
saying that thoughts are individuated in terms of those of their normal
causes and eVects which happen in accordance with properly functioning
cognition.
The other way forward would be to characterise as inferential those
processes which happen in accordance with psychologicallaws, and/or
which are counterfactual supporting. This approach would probably cover
much the same ground as the previous proposal, but has the advantage
that content-involving psychology can then be characterised entirely a-
historically. So, provided it is alaw(or nomic tendency, at least) that
people who believe ‘P & Q’ will also come to believe ‘P’ceteris paribus, then
this transition can be used to characterise the functional role of the former.
But since it is presumablynotlawful that thoughts of Madonna visiting
SheYeld should cause shaking hands, this eVect will not form part of the
functional role of that thought.
4.3 Two-factor functionalismAnother point is that it can easily seem that functional-role semantics is
not really acompetitorto informational semantics or to teleological se-
mantics, in that it is not really addressing the same questions. For how can
the reference, or worldly truth-condition, of a thought be a matter of how
that thought functions intra-cranially? – how can it be a matter of the
network of inferences into which that thought can enter? It can seem
inevitable, in fact, that functional-role semantics is only a naturalised
account ofnarrowcontent, and that it cannot be extended to account for
widecontent. In which case someone might claim to be an informational or
teleological semanticist (about wide content)anda functional-role seman-
ticist (about narrow content), and there is no real competition.
It is certainly true that some versions of functional-role semantics have
been intended as accounts of narrow content only. And it is equally true
180 Content naturalised