Figure 5.1 The Wason card-selection taskJudging from this version of the selection task it looks as if people are
checking on the correctness of the conditional solely by checking for
positive instances, as if seeking to conWrm a rule ‘All Fs are Gs’ by
examining the Fs and the Gs. Bad news for Popperians, you might think!
Having accepted Hume’s sceptical argument about inductive inference,
Popper (1971) claimed that we do not rely on induction anyway. The
impression that we do is a sort of ‘optical illusion’. Instead, we start from
some prior hypothesis and use the method of conjectures and refutations.
The initial interpretation of the selection-task data would seem to be that
the majority of people are, on the contrary, the most naive conWr-
mationists, and that they are not at all alive to the crucial signiWcance of
the potentially refuting instance. Practically everyone chooses the (a) card,
but on versions of the test like the one above few seem to grasp the
signiWcance of the (d) card as a potential falsiWer of the proposed
generalisation.
However, the results to be extracted from the selection task are more
interesting and subtle than indicated so far. Changing the details of the
task can change performance. You might expect that devising less ab-
stract forms of the card-test, dealing with topics which are concrete and
familiar to people, would signiWcantly improve performance. But not so:
familiarity turns outnotto be one of the crucial variables. However,
subjects do perform much better if they are dealing with what is called a
deonticversion of the task; that is, a case in which they are considering a
normative rule which may be kept to or broken, rather than a factual
generalisation. Thus, if the rule isIf anyone is drinking alcohol, then they
must be over 18 years of age, and the information about individuals’ ages
and drinks printed on cards reads: (a)drinking a beer, (b)drinking cola, (c)
26 years of age, and (d)16 years of age, subjects see the importance of
selecting the (d) card much more easily (Johnson-Lairdet al., 1972;
Griggs and Cox, 1982).
So what needs to be explained is why people do badly at some versions
110 Reasoning and irrationality