Book Review: Rationality in an uncertain world: Essays on the cognitive science of human reasoning

This book is partly a retrospective of the authors’ contributions to the ®eld of reasoning. Their views are well known. The basic presumption is that cognitive scientists have ignored the fact that human inference is uncertain and committed themselves instead to logicÐa calculus of certain inference. Everyday inferences, on the other hand, are defeasibleÐthat is, a particular inference can be cancelled in the light of changing circumstances. Probability theory, rather than logic, offers a better approach to understanding the mind and the nature of human inference because it has uncertain inference at its core. The eight chapters of part 1 elaborate the problems of ``logicism’’. The advantages of their preferred alternativeÐthe probabilistic approachÐare expounded in the seven chapters of part 2, six of which consider the Wason selection task. In the Wason selection task, individu als are presented with an indicative conditional claim (e.g. ``If there is a vowel then there is an even number’’) that applies to four cards, each of which has a letter on one side (either a vowel or a consonant) and number (either odd or even) on the other. Presented with cards whose upper faces reveal A, K, 4, and 7, individu als have to select those cards and only those cards that need to be turned over in order to prove the claim either de®nitely true or de®nitely false. Invariably, individuals choose either the antecedent of the claim (the card A), or the antecedent and the consequent of the claim (cards A and 4) with much greater frequency than the logically correct solution (cards A and 7)Ðonly the latter two cards could reveal a possible counter-example (a card with an A and a 7 on it). Performance deemed irrational by logicism is reconstrued as rational on a probabilistic analysis of the problem. According to Oaksford and Chater (O&C), individu als apply what works in the everyday world. In the everyday world, ``most properties about which people reason apply to only a small proportion of objects in the world’’ (p. 19). In such a context, where the assumption of rarity holds, it would be odd indeed to examine, for example, non-black things in order to assess the truth or falsity of the claim that ``All ravens are black.’’ Instead, inspecting ravens and black things is normatively correct. The equivalent for the selection task is selecting the true antecedent and the true consequent (A and 4 in the example above), which is what individu als tend to do. O&C’s central goal is to offer a computational level account of human inferenceÐthat is, they seek to characterize what job is done by the mind. And they adopt a particular position on this. Their supposition is not only that the mind is adapted to the environment but also that its adaptiveness conforms to a normative standard (here the calculus of probability). Others (e.g. Evans & Over, 1996; Gigerenzer & Goldstein, 1996) have sundered the link between normative and adaptive rationality. But clearly even if we suppose a connection, there is no claim that the mind/brain performs Bayesian (or other probabilistic) computations, and O&C leave open what the nature of these computations may actually be, though they do offer a measured view of the relevance of connectionist approaches to this question. The authors compile their case, sometimes wittily, and invariably with scholarshipÐevident in the citations of work in the text, in the footnotes, and in the technical detail supplied. Their THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2000, 53A (1), 281±283