Learning modes, feature correlations, and memory-based categorization.

Abstract

Several pairs of correlated features were embedded in descriptions that had a wealth of exemplar-specific (i.e., idiosyncratic) information, and sensitivity to these correlations was examined as a function of intentional and incidental encoding conditions. Participants in incidental conditions were able to access information about several embedded correlations, even when the correlations involved 3 rather than 2 dimensions, and when complex inferences were required to recover correlations. In intentional conditions, however, little access to correlations was observed. The richness of the stimuli made it difficult to detect correlations at encoding, and the representations that resulted from analysis appeared too impoverished to allow covariations to be recovered. The results are discussed in terms of the advantages of storing examples for addressing unanticipated needs and goals.