To avoid giving the impression that this is an easy,ad hoc, explanation
of an already known result, Sperber and his collaborators have shown in
detail how use of relevance theory can enable experimenters to manipulate
subjects’ performance on indicative selection tasks. In order to secure high
levels of correct selections for conditionals of the form, ‘For all x, if Fx
then Gx’ – making the selection task easy – one needs to increase the
number ofF-and-not-Gchoices, by reducing the required processing eVort
and/or by increasing the cognitive eVect. To do the former, an experimen-
ter could exploit a lexicalised concept which covers theF-and-not-Gcases
(for example, ‘bachelor’; withF: ‘male’ andG: ‘married’). To do the latter,
oneWnds some way of makingF-and-Gcases less interesting to subjects
thanF-and-not-Gcases. For full details of the experimental conWrmation
of these predictions of relevance theory, see Sperberet al., 1995a.
The ability to control performance on a number of selection trials
(although still requiring further replication) appears to be striking con-
Wrmatory evidence in favour of relevance theory. But a question arises
concerning the relationship between relevance theory and the modularist
research programme. For the relevance-theory explanation of results on
indicative selection tasks is of a quite diVerent kind from the postulation of
a cheater-detection module dealing with deontic tasks which have a cost-
beneWt structure. Although diVerent, it does not appear to be incompatible
with modularity, however. For although the principles of relevance theory
are intended to apply to cognitive processing in general, Sperber and
Wilson are in no way committed to postulating a domain-general cognitive
system or ability. It is entirely possible that relevance theory describes the
constraints on cognitive eYciency which shape the functioning of a large
number of distinct processing modules – although we would agree that
how this might work out in a system of interacting central modules still
remains fairly mysterious.
5 Practical rationality
Naive subjects attempting a selection task may not be following an infal-
lible procedure for determining the truth-value of a conditional, but they
may still (if relevance theory is right about this) be processing information
in accordance with principles which secure a sort of eYciency worth
having. We should expect ordinary human reasoning to be shaped by
practical constraints which may not always elevate production of exactly
the right answer above other considerations.
Consider once again the Wason selection task. What is generally des-
cribed as ‘conWrmation bias’ in the execution of this task – looking only at
theFandGcases when testing a conditional of the form ‘For all x, if Fx
then Gx’ – makes a good deal of sense when seen as a heuristic appropriate
Practical rationality 125