Elemental associability changes in human discrimination learning.

Abstract

Participants initially completed a discrimination task (D1) involving categorization of patterns with multiple common features, each feature being partly predictive of the correct response. In a subsequent target discrimination task (D2), these features were redistributed across new discriminative stimuli. The relative predictiveness of the features in D1 was either maintained in D2 (i.e., features were equally informative in D1 and D2) or switched (i.e., more informative features in D1 were made less informative in D2, and vice versa). Differential performance on D2 suggested that features most predictive of the correct D1 responses became more highly associable than features that were less predictive in D1. This finding suggests that the associability of individual stimulus elements changes as a consequence of their role in discrimination learning.