setting up some sort of associational link. However, a diVerent inter-
pretation of why the ‘meaningful’ rule is easier for subjects is possible in
the light of the cheater-detection hypothesis: the conditional may be
understood in such a way as toprescribethat the beneWt of sealing the
envelope is only available if the (higher) 5d cost of postage is met. This
interpretation is conWrmed by further tests comparing the performance of
subjects who varied in their knowledge of these postal arrangements.
American subjects did not perform as well on the test as subjects from
Hong Kong, where such a postal rule was enforced. But when the rule was
provided with a rationale – it was stated that sealed mail was treated asWrst
class – the American subjects did just as well on the selection task as Hong
Kong students (Cheng and Holyoak, 1985). We suggest that the rationale
served to aid subjects’ reasoning by giving the task a cost-beneWt structure,
hence bringing the cheater-detection module into play.
4.2 Relevance theoryThe hypothesis of a cheater-detection module is strongly supported by the
fact that it is not only able to explain the divergence in performance
between deontic and indicative versions of the selection task, but also
yields a more precise speciWcation of where the divergence is to be found.
Yet it cannot account for all the variation in performance to be found in
various versions of the selection task, since there is divergent performance
even on problems which do not diVer in terms of possessing or lacking a
cost-beneWt structure. Furthermore, even if we do have a special modular
capacity for reasoning about costs and beneWts in social exchange, this
does not explain why we should be so surprisingly poor at many indicative
versions of the selection task, and in particular why we shouldWnd it so
diYcult to spot the signiWcance of the potentially falsifying (d) card. The
most impressive attempt to account for subjects’ performance on these
indicative versions of the selection task involves the application of rel-
evance theory, as advocated by Sperber and Wilson (1986/1995).
The thinking behind relevance theory is that new information presented
to an individual is always going to be processed in the context of previous
beliefs and earlier thought. The new information will be more or less
relevant to the individual depending upon the cognitive eVects it produces,
in terms of new beliefs or modiWcations of existing beliefs. Sothe greater
the cognitive eVect, the greater the degree of relevance. But a cognitive eVect
will require processing, and heavy processing eVort is likely to be too
expensive to indulge in. Sothe greater the processing eVort required to
extract a piece of information, the lesser its degree of relevance. Sperber and
Wilson sum up their theory in terms of two general principles:
Psychological explanations of performance 123