which Horgan and Tienson oVer in support of non-algorithmic processing
(following Putnam, 1988) is unsound.
Horgan and Tienson premise their argument onepistemic holism, focus-
ing especially on science and scientiWc method. In science, anything can,
potentially, be relevant to anything else. As they put it:
Are whales and dolphins relevant to questions about human evolution? Not if the
task is to produce a Darwinian history ofHomo sapiens. But if the question is the
role of tool use in the evolution of human intelligence, then it could be relevant to
compare whale and dolphin cognition with human and (so far as possible) early
hominid cognition, given that whales and dolphins are said to be pretty intelligent
among nonhuman animals and that they are not tool users. (1996, p.38)
It is now a familiar and well-documented phenomenon in the history of
science, indeed, that what seem like completely independent questions in
domains of enquiry remote from one another, have a way of turning out to
be relevant to each other after all. This motivates the Quinean idea of a
‘web’ of scientiWc belief (Quine, 1951), in which relations of epistemic
support are distributed right across the system, and in which an alteration
in any one part of the web can, in principle, be accommodated by making
adjustments in remote regions.
Given the truth of epistemic holism, Horgan and Tienson then argue
that it is very unlikely that central cognition – and, in particular, belief-
Wxation – should be subserved by strict processing algorithms, whether
probabilistic or not. For it seems impossible to devise algorithms which
would allowanybelief or piece of evidence to be relevant to theWxation of
a given belief. Indeed, those working within classical approaches to cog-
nitive science face a dilemma, which has come to be known as the ‘frame
problem’.Eitherthey try to allow everything to be relevant to everything,
in which case it is hard to avoid combinatorial explosion – thus it is plainly
impossible to search through each of your beliefs, together with their
logical consequences, every time you make a decision.Orthey place a
‘frame’ around the information which is deemed to be relevant for a
particular processing task, in which case they can no longer do justice to
the holistic nature of belief-Wxation. The only way out of this dilemma,
Horgan and Tienson argue, is to abandon the search for processing
algorithms altogether, and to try to model central cognition in terms of
dynamical systems theory, which might then be instantiated in a connec-
tionist network.
One fallacy in this argument, however, is the same as that committed by
Fodor (1983), and discussed at length in chapter 3 above. It is that the
argument involves the move fromepistemicholism in science tocognitive
holism, or holism about belief-Wxation within the minds of ordinary in-
206 Forms of representation