Cognitive principle of relevance: Human cognition tends to be geared
towards the maximisation of relevance.
Communicative principle of relevance: Every utterance conveys a pre-
sumption of its own relevance.
In eVect, they are oVering an economics of cognition according to which
human processing operates in such a way that the beneWts of new cognitive
acquisitions are balanced against the costs of the processing involved in
their acquisition.
If we apply relevance theory to previously known results on indicative
versions of the selection task, it immediately secures a promising success,
by yielding an explanation as to why performance is improved if the
conditional under consideration is of the form ‘If F, thennotG’. In such a
case a majority of the subjects make the correct selection (Evans, 1972).
For example, in a slightly altered version of the test presented in section 2,
the conditional ‘If there is an A on one side of the card, then there isnota5
on the other’ would make the selection of cards (a) – containing an A – and
(c) – containing a 5 – correct; and subjects do indeed tend to make this
choice. Why?
Suppose that subjects check on the conditional by checking on what they
take to be its consequences, relying upon the inferences they intuitively and
unreXectively draw from it. Since any proposition has inWnitely many
consequences (fromPwe may inferPorQandP or Q or Rand so on),
subjects must plainly stop when they judge that they have generatedenough
consequences to secure relevance. If the conditional is tacitly of the form
‘For all x, if Fx then Gx’ (‘Every card is such that, if it has an A on one side
then it has a 5 on the other’), then inferringthat an instanceFashould, if the
conditional is true, also beG, is the inference which costs subjects least in
the way of processing eVort. Hence the almost universal choice of card (a),
which hasFinstanced (that is, having an A on it).
To make the other selection, subjects need to infer from the conditional
a negative existential: that there is no instance such that it isFandnot-G.
But this inference involving negation is more costly for subjects in terms of
processing eVort (it is well known that inferences involving negation are
more diYcult). Instead, they tend to infer (non-deductively but reliably)
from ‘All Fs are Gs’ to ‘There are Fs and Gs’, and so select theF-card and
theG-card, rather than thenot-G-card. However, with a negative version
of the conditional, of the form ‘For all x, if Fx then not Gx’ (‘Every card is
such that, if there is an A on one side, then there isnota 5 on the other’), the
burden of representing a negation within the scope of the negative existen-
tial has been lifted from subjects, and so the inference to ‘There is no
instance which is both an A and a 5’ is more readily available to them. So
on that version of the task they spot the signiWcance of selecting the
potentially falsifyingG-card.
124 Reasoning and irrationality