realisability. The idea seems to be that although we cannot be required to
provide natural conditions which arenecessaryfor a given content (since
that content might be realised in a variety ofdiVerentconditions), still it
must at least be possible toWnd natural conditionssuYcientfor a given
content, which play the realising role. But he still seems to think that the
reduction has to be some sort of conceptual one, since he allows that it can
be refuted by appeal to purely imaginary examples. And he still seems to
think that the suYcient conditions in question can be discovered bya priori
reXection. But why on earth would one expect that such an enquiry could
be conducteda priori, or that the suYciency in question should be concep-
tual?
5.3 The future of naturalised semanticsDoes this then mean that there is no truth whatever in the various reduc-
tive accounts of intentionality proposed by philosophers which we have
considered? No, we do not insist on that. For one thing, there is a puzzle
about the origin of content. The world did not always contain represen-
tations. So representations must somehow have evolved in the natural
order. We are happy to agree that the question of how that can have
happened can be answered by appeal to the basic elements of teleo-
semantics – speciWcally, a system of reliably sensitive indicator states
coming to be selected for purposes of behavioural control. (Compare a
modest account of the origins of monetary exchange – as against the
unpromising attempt to provide a reductive analysis of every species and
quantity of monetary value.)
For another thing, it may be that what can, indeed, be plausibly pro-
vided from the philosopher’s armchair arenecessary conditionsfor inten-
tionality, or for entertaining any given intentional content. This is surely to
be expected if concepts are represented in the form of prototypes, in the
way discussed earlier. For notice that amongst the prototypical features of
dogs listed above were two –has dogs as parentsandis a mammal– which
are at least plausible necessary conditions for a creature to be a dog. So one
might predict that in the case of semantic concepts, too, there are necessary
conditions which can be articulated more-or-lessa priori. (See Williamson,
1995, who makes just this point in connection with our concept of knowl-
edge.)
What we propose, then, is that philosophers should start to explore
variants of the naturalistic programmes discussed in this chapter – perhaps
with special emphasis on functional-role accounts, either in their ‘long-
armed’ form, or supplemented by causal co-variance constraints – with
a view to providing, from the armchair, not areductive account of
Naturalisation versus reduction 189