Retrospective revaluation and inhibitory associations: does perceptual learning modulate our perception of the...

The reported experiments seek to reconcile the conflict between proposed explanations of retrospective revaluation of contingency judgements in experiments with humans (e.g. Dickinson & Burke, 1996) and those of a perceptual learning phenomenon discovered in rats (Espinet, Iraola, Bennett, & Mackintosh, 1995). The experiments employ a medical diagnosis scenario, in which subjects are asked to judge whether particular symptoms signal the presence or absence of a fictitious disease. In Experiment 1, following alternating exposure to pairs of symptoms, AX and BX, subjects were informed that symptom A was predictive of illness. As a result of this training, subjects rated symptom B as predictive of the absence of the illness. In Experiment 2, a possible mechanism behind this effect was considered. In particular, it was established that pre-exposure in which the AX pair reliably preceded the BX pair produced the same pattern of results as those shown in Experiment 1. In contrast, when the order of pairs was reversed, so that BX reliably preceded AX, the reduction in ratings for symptom B was not obtained. Several extant associative theories are discussed, but none are found to account adequately for this pattern of results. To overcome these inadequacies, a parsimonious theory is suggested that explains all the effects observed in these experiments.