may be capable of thinking in this way. But much of the time, when relying
upon our native resources, we may well depend upon mechanisms for
cognitive processing which answer to other priorities. We shall return to
develop these points in section 5 below. For the moment, our conclusion is
that philosophers have failed to provide any good reason for rejecting the
results of the psychological reasoning experiments.
4 Psychological explanations of performance
An idea shared at one time by several investigators was that what made
variants of the selection task – such as that given in section 2 above –
diYcult for so many subjects, was that they were abstract and schematic,
being about numbers and letters rather than more tangible or everyday
materials. So perhaps the explanation for divergent performance on the
selection task was that theorists classed variants of the task together
because of theirformalsimilarity, and yet these were diVerent processing
tasks for the subjects involved. According to the logician, the validity of an
inference is a matter of the truth-preserving quality of a general pattern or
schema which it exhibits, such as:All Fs are Gs; all Gs are H; so, all Fs are
H. Instances of that pattern will be valid, no matter what ‘F’, ‘G’ and ‘H’
may be – whether the particular argument is about cabbages or kings.
But perhaps it matters to the ordinary reasoner’s performance what a
particular piece of reasoning is about. If she is familiar with cabbages and
has had considerable practice in thinking about cabbages, perhaps she will
perform much better on a test of reasoning about cabbages than on a
similar test about kings (or triangles, or numbers, or sets). Several theorists
(Manktelow and Evans, 1979; Griggs and Cox, 1982; Johnson-Laird,
1982; Pollard, 1982; Wason, 1983) have explored this basic idea by sug-
gesting hypotheses concerning the way associational links might facilitate
performance on reasoning tasks.
In fact it turned out thatmerefamiliarity of subject matter has little
eVect upon task-performance. Although subjects do indeed perform better
on some familiar versions of the selection task, it is always possible toWnd
a version of the task which is equally familiar, but on which they perform
poorly. Yet we need not give up on the idea that performance on such tasks
is content-sensitive. Indeed, if we apply the general modularist research
programme to results obtained on the selection task, we might expect that
variation in performance on diVerent versions of the task would be at-
tributable to the operation of central modules specialising in diVerent sorts
of reasoning in diVerent domains. This is an hypothesis which a number of
investigators – most notably Cosmides and Tooby – have begun to ex-
plore.
Psychological explanations of performance 119