which are true of us in virtue of our possession of a folk psychology are
inapplicable. And most psychologists say, in consequence, that prior to the
age of four children lack the conceptbelief, and are incapable of belief-
thoughts. (Perner, 1991, says that three year olds possess the distinct
conceptprelief, which is a kind of pared-down amalgam ofpretenceand
belief.) Yet they also say thatafterthe age of four we all possess the same
concept of belief, despite the many minor diVerences which may exist
between individuals in their beliefs concerning beliefs. Plainly, then, their
idea may be that the conceptbeliefis to be individuated by just those
inferential connections (such as the one from ‘A believes that P’ to ‘A is in a
state which may be false’) which are necessary and suYcient for the
relevant psychological generalisations to be true of us.
A general moral can be extracted from this sort of case. Fodor objects to
functional-role semantics on the grounds that it has no ‘principled’ way of
marking out the inferences which are constitutive of meaning; so all
inferences would have to be counted as meaning-constitutive; so content
becomes holistic and idiosyncratic; and so, disastrously, there would be no
contentful psychological laws. We maintain that Fodor has got the dialec-
tics of content back-to-front. TheWrst thing to assert is that there are,
undeniably, contentful psychological causal tendencies of a lawlike char-
acter. (Indeed, note how we rely upon them all the time – in order to
communicate, pragmatically, far more than we need explicitly state, for
example.) But then, given that there are such lawlike connections, a notion
of functional-role content can always be constructed – namely, from
whatever inferences are necessary for the laws to apply. Whatever is
idiosyncratic isprecisely what is omittedfrom the functional-role that
constitutes content. And there we have the required principle.
5 Naturalisation versus reduction
All three of the main semantic theories considered here (informational
semantics, teleological semantics, and functional-role semantics) are
driven by a desire tonaturalisesemantic properties, such as meaning, truth,
and reference. And this desire, in turn, is underpinned by the need to avoid
eliminativismabout content-based psychology, as we noted in section 1.
But in order to see whether anaturalisedsemantics would have to be a
reductivesemantics, we need to get a little clearer about what, exactly, a
reductive account of some phenomenonis.
5.1 Reduction and realityReductions come in (at least) two forms –conceptualandmetaphysical.
Conceptual reductions are generally attempteda priori, and purport to
184 Content naturalised