that defenders of narrow content might be well advised to look to some
such form of functional-role semantics if they seek a naturalised account of
their target notion. But it is important to see that functional-role semantics
does nothaveto be narrow. For the functional-role of a mental state can be
characterised in such a way as to embrace the worldly causes and eVects of
that state – these are then ‘long-armed’, or world-involving, as opposed to
‘short-armed’, or in-the-head, functional roles (Block, 1986). Thus we
might say that it is part of the functional role of the belief, ‘Madonna will
visit SheYeld’ to be caused by events involving Madonna herself, and to
cause me (given my desires) to stand all night in the rain outside the
SheYeld Arena in order to see her.
Functional-role semantics might well appeal to some form of causal
co-variance, or informational, account in order to characterise the world-
involving element of functional role. So we might say that amousethought
is one which carries information about mice (which is caused,inter alia,by
the presence of mice), appealing to the ways in which that thought interacts
inferentially with others in order to solve the disjunction problem, and to
distinguish it from other thoughts which might concern the same worldly
property. Somousehas the contentmouse, and notshrew-or-mouse,in
virtue of the fact that it tends to be caused by the presence of a mouse, and
because I am apt to infer from itnot shrew, andcan interbreed with
other mice, and so on. (For a number of diVerent proposals concerning
the development of a two-factor semantics, see Field, 1977; Loar, 1982;
McGinn, 1982; Block, 1986.)
4.4 The charge of holismPlainly, there is a sense in which functional-role semantics is holistic. For
in specifying the functional-role of one state – say,the belief that P– we will
have to mention others with which it may causally interact; and many of
these, too, will be states with semantic contents (the belief that Q; the desire
that R; and so on). And it might seem that this then raises a diYculty for
the project ofnaturalisingcontent. For how can one successfully reduce
into purely natural, causal, terms the contentP, if the statements that
attempt to do so have to mention yet other contents? It can easily seem that
this is notreducingcontent, but merely passing the buck from one content
to another. In reality, however, there is no diYculty here, and this (weak)
sense in which functional-role semantics is holistic raises no problem for
the naturalisation project. In fact it is a ubiquitous feature of scientiWc (and
other) theories, that they should contain terms which only get their sig-
niWcance from their relationship to the other terms in the theory. But as
Lewis has shown (1970), onecanstill deWne each of these terms by
Functional-role semantics 181