dividual thinkers. Science is, of course, a collective and inter-subjective
enterprise. New empirical results and new theories are normally arrived at
through collaboration and discussion between a number of investigators;
and, once published, they are subject to criticism and debate within the
scientiWc community as a whole, often over extended periods of time. The
fact that theory-conWrmation under these circumstances is strongly holis-
tic, provides no particular reason to think that the same will be true of the
cognition of ordinary thinkers, who have to make up their minds in real
time – often seconds or minutes rather than years – and often without any
opportunity for public debate.
Another point against Horgan and Tienson’s argument is that it is hard
to get it to marry with the sort of modularism about central cognition
which we have defended in this book. Their picture seems deWnitely to be
of asinglecentral system – a seamless web of belief – within which forces
and inXuencesXow back and forth in dynamical fashion. Our view, in
contrast, is that central cognition probably contains a wide variety of
conceptual modules which are each dedicated to processing information
about particular domains.
Of course, it somehow has to be possible for individual thinkers, with
their modularised minds, to engage in science. But it may be that this
possibility depends crucially upon (and is ‘scaVolded’ by) the external –
public – resources provided by spoken and written natural language,
mathematics, libraries, computers, and such like, as well as the cognitive
resources of the individual (Clark, 1998, and section 3.2 below). It is also
true, surely, that the various central modules have to be able to pool
together their information in some fashion, and communicate with one
another in such a way as toWx belief. Here, too, it may be that natural
language has an important part to play, as the sort oflingua francaby
means of which modular central systems interact (Carruthers, 1998a, and
section 3.6 below). Or it may be that the central ‘pool’ where the outputs of
diVerent modular systems can interact is provided byconsciousthought
(see Mithen, 1996, and sections 3.4 and 3.5 below).
These are, however, very much matters for future inter-disciplinary
research – at the moment we can only speculate. And we certainly do not
want to close oVthe possibility that some form of dynamical systems
theory may be needed to explain theWxation of belief in ordinary cog-
nition, even given the truth of modularism. But we do think that it would
be premature to abandon the algorithmic paradigm at this stage. Inves-
tigators should explore what can be done to model processing within a
variety of modular conceptual systems, and also the ways in which such
systems can interact, in algorithmic (but almost certainly probabilistic)
terms.
Mentalese versus connectionism 207