Butat leastthereis acertaindegreeof consensusinthephilosophyof science
that Duhem and Quine were right to maintain that scientiWc theories and
hypotheses form a sort of network in which there are reticular relations of
evidential support (Duhem, 1954; Quine, 1951). This is a picture which
appears toWt belief-Wxation in general fairly well. A candidate belief gets
accepted if it coheres suYciently well with other beliefs. Although that
seems right, epistemology has plenty of work to do in order to specify what
exactlytherelationofcoherenceconsistsin. For present purposesthemoral
to be extracted is that a candidate belief needs to cohere with other relevant
beliefs, but what other beliefs are relevant depends on background knowl-
edge and so cannot be given ana priori, system-external, speciWcation.
As Fodor remarks ‘in principle, our botany constrains our astronomy, if
only we could think of ways to make them connect’ (1983, p.105). This
point can be illustrated by a supposed connection between solar physics
and Darwin’s theory of natural selection. Shortly after the publication of
The Origin of Speciesa leading physicist, Sir William Thompson, pointed
out that Darwin just could not assume the long time-scale required for
gradual evolution from small diVerences between individual organisms,
because the rate of cooling of the sun meant that the Earth would have
been too hot for life to survive there at such early dates. Now we realise
that the Victorian physicists had too high a value for the rate at which the
sun is cooling down because they were unaware of radioactive eVects. But
at the time this was taken as a serious problem for Darwinian theory – and
rightly so, in the scientiWc context of the day.
According to this argument, then – we will call it theNetwork Argument
- what you come to believe depends upon your other relevant beliefs. But
what other beliefs are relevant depends in its turn upon what you believe.
Does this show that central cognitive processes, and in particular belief-
Wxation and non-demonstrative inference, cannot be encapsulated? Does
the Network Argument refute the view that cognition is modular through
and through? No. It cannot be the case that considerations drawn from
epistemology and scientiWc practice (in fact the Network Argument is
really founded on the case for a sort ofcoherentismin epistemology)
should directly establish such a conclusion about psychology and the kinds
of processing systems inside individual heads. There has to be a connection
between epistemology and scientiWc conWrmation on the one hand, and
individual cognitive psychology on the other, because science has to be
something individuals can do, and those individuals will be justiWed, and
know that they are justiWed, in some of the things which they believe. But
the transition from the epistemology of conWrmation to the nature of
central processing is by no means as smooth as the Network Argument
would make out.
Input systems versus central systems 71