In two experiments, participants learned to discriminate between a pair of simply related, but very similar colors, in a two-choice categorization task. They were then tested over a wider range of isoluminant hues. Over these test values, both experiments yielded a postdiscrimination gradient that was initially peak-shifted but became monotonic through the course of testing. In Experiment 2, the presence of this early peak shift and subsequent change in gradient form were related to participants' inability to verbally characterize the difference between the training stimuli. This suggests a transition from generalization based on simple physical similarity to generalization based on a verbalizeable rule, as a consequence of additional relevant information becoming available during test. An explanation appealing to both associative and strategically controlled verbal processes provides an accurate account of the results.

The Perruchet effect refers to a dissociation between trends in the conscious expectancy of an event and trends in the strength or the speed of responding to that event, which suggests that learned changes in the performance of a response may be automatic. Despite being consistently demonstrated in conditioning studies and simple reaction time (RT) tasks, mixed results have been found in the choice variant of the Perruchet effect, especially when expectancy and responding are measured concurrently (that is, on the same trial). The present experiments examined why the dissociation disappears when concurrent measurement is used by directly comparing trials on which expectancy is measured to trials on which expectancy is not measured. In Experiment 1, expectancy was measured on a randomly chosen 50% of trials, whereas expectancy was measured every fourth trial in Experiment 2. In both experiments, the Perruchet effect was weakened on trials that immediately followed an expectancy rating but was still clearly evident on other trials, suggesting that automatic facilitation of RT based on recent trial history is temporarily masked, rather than abolished, by a concurrent expectancy judgment. (PsycINFO Database Record