has been raised against any form of inferential-role semantics. Fodor has
repeatedly objected (1987, 1990; Fodor and Lepore, 1992) that in the light
of Quine’s attack on the analytic–synthetic distinction there is no princip-
led way of distinguishing those inferential links which are constitutive of
meaning from those which are not. (See Quine, 1951, who argues that there
is no such thing as an inference which is valid in virtue of the meaning or
content of its terms; rather, there is just a spectrum of inferences to which
we are more-or-lessWrmly committed.) Fodor urges that the only adequate
response to this point is to recognise the atomicity of representational
content: that each particular belief has the content it has in virtue of its
distinctive causal relationship with the world (see chapter 7). Otherwise we
will be forced into the holistic absurdity of supposing a particular content
to be constituted by thetotalityof its inferential links – in which case it
must be doubtful if two thinkers can ever both have instances of a shared
belief-type.
However, this line of argument involves an oversimpliWcation. The
question is: what inferential linksmustbe in place if a thinker is to have a
belief with a certain content? The answer, ‘Those which are constitutive of
content’, is not unprincipled, even if it is not very informative. If you then
go on to ask: ‘How do you tell the links which are constitutive of content
from those which are not?’, you ask a diYcult question. But one thing
which ought to be clear is that there is no reason to suppose that there is
going to be a single principle capable of delivering appropriate answers for
all concepts, for all contents.
The problem with analytic inferential links, as determiners of content, is
not that they are unprincipled, but that simple and transparent cases are in
relatively short supply. Where we do have simple and transparent cases of
this sort of inferential link, the source of the principle is usually easy
enough to discern: we are dealing with aderivativeconcept which was
linguistically introduced to thinkers via the inferential link in question.
This applies toassassinate, and of course it also applies to the favourite
philosophical examplebachelor(therefore:unmarried man). These content-
constitutive inferential links are readily accessible to us because they are
the conduits through which we learnt the concepts in question. It is
instructive to comparebachelorwithbrother. In terms of necessary and
suYcient conditions ‘male sibling’ oVers as good a deWnition of ‘brother’ as
‘unmarried (adult) male’ does of ‘bachelor’. But a little girl can think that
her brother has just come into the room without thinking that a sibling of
hers has just come in.
It is quite evident that not all concepts are derivative: they cannot all be
learnt through explicitly and prescriptively establishing an inferential link.
Developmental studies suggest that our repertoire of conceptual contents
Philosophical arguments in defence of rationality 115