exhibit something red; and if I want to check on the truth of the former I
shall open my eyes, whereas if I want to check on the truth of the latter I
shall pick up my vibrating-machine.
4.2 Elaborating functional-role semanticsHow best should functional-role semantics be developed? One question is
whether or not the identity of a content-bearing state should be made to
turn onallkinds of characteristic causes and consequences of that state
with an explanatory role. Thus suppose that my psychology happens to be
such that the thought, ‘There is a snake nearby’ always causes me to have a
panic attack – where it does not have this eVect because Ibelievethat
snakes are dangerous, but rather because of some association set up in my
childhood. Does this eVect contribute to the identity of the thought which I
thereby entertain, in such a way that if the thought which you would
express in the same words doesnotcause panic attacks in you, then we do
not entertain one and the same type of thought? Or suppose that the
thought, ‘Madonna is to visit SheYeld’ causes my hands to shake, but not
yours. Does this mean that we do not entertain thoughts with the very
same content, because our thoughts diVer slightly in their functional roles?
If we answer ‘Yes’ to these questions then we are proposing that content
should be analysed in terms of crude, undiVerentiated, causal role, where
anycause andanyeVect – whether cognitive, aVective, or brute physical –
can count in the individuation of content. If we answer ‘No’, then we need
toWnd some principled way of drawing a ring around the set of causes and
eVects we are interested in.
It does seem implausible that you and I should count as entertaining
distinct thoughts (which we would nevertheless express in the very same
words), merely because the ‘emotional colouring’ or the physical eVects of
those thoughts are diVerent. But how, then, are we to characterise the
relevant subset of a thought’s causes and eVects, which goes to make up
its content-deWning functional role? One obvious way forward, is to say
that it is only those causes and eVects which areinferentialin character
which count towards identity of content. But this then raises another
problem: whatisan inferential process? How do we distinguish inferences
from other sorts of cognitive transition? We obviously cannot say that a
transition is inferential just in case it can be broken down into intermedi-
ate steps, since this would bring it out that the transition from ‘P & Q’ to
‘P’ isnotan inferential one! But then nor can we say that a process is
inferential just in case it isvalid, or (more plausibly, since not all inferen-
ces are deductive ones) just in case itreliably generates truth from truth.
For this would be to introduce semantic notions into our attempt to
Functional-role semantics 179