success, and the other of which employs whatever standards are derivable
from normative systems of logic, probability theory and decision theory:
rationality 1 : thinking which is generally reliable for achieving one’s
goals;
rationality 2 : thinking which conforms with a (correct) normative the-
ory.
Given the immense success enjoyed by our species (so far!), it seems very
unlikely that ordinary human thinking should be lacking inrationality 1 ;
but it may fail to exhibitrationality 2. Evans and Over have followed up this
idea by proposing a dual process theory of thinking, according to which
much of our thinking is carried out in terms of tacit and implicit processes.
This can be accepted, with the modularist proviso that the level of implicit
processing is itself multi-modular.
Evans and Over build up a powerful case that most human reasoning is
governed by processes, such as those determining relevance and selective
attention, which are implicit and inaccessible to the subject. Indeed, they
report a particularly dramatic set of experiments in which subjects’ choices
in the selection task are manipulated by changing experimental variables,
and where subjects are encouraged to verbalise the reasons for their choice
(Evans, 1995). It turns out that eye-direction determines subsequent
choice, with subjects choosing just those options which they initially attend
to; and the explicit reasons subsequently oVered are mere rationalisations
of choices determined by a set of non-conscious heuristics. But Evans and
Over do not deny thatsometimesreasoning can be governed by processes
which are explicit and conscious. Indeed, their dual-process theory of
thinking maintains that both sets of processes can contribute to reasoning
to a greater or lesser extent, depending on task demands. One of the
distinguishing features of explicit (conscious) reasoning is said to be that it
is verbally mediated, and that it is much more inXuenceable by verbal
instruction. (We return to consider the possible connections between con-
scious reasoning and language in chapter 8.)
We therefore have two orthogonal distinctions in play: between two
diVerent notions of rationality, and between two diVerent levels of cog-
nition. Then one can (and should) ask of both implicit and explicit reason-
ing processes, to what extent they serve the subject’s goals; and one can
(and should) ask of both implicit and explicit processes, to what extent
they approximate to valid logical norms.
First, as regards implicit processes, it seems plain that they must largely
be rational 1 , given that most reasoning is implicit, and given the practical
success of our species. But it is also possible that some of these processes
should be rational 2 , involving computations which comply with valid
norms, but driven by pre-logical judgements of relevance. (This seems to
Practical rationality 127