Generalization of delayed identity matching in retarded children.

In an extension of prior research, four retarded children were trained under an identity matching-to-sample procedure containing features previously shown to produce controlled generalization to novel stimuli. They first were taught to relate a particular handsign to the sample shape, then to maintain the handsign over a delay interval, and then to select from an array the comparison shape that permitted the handsign to be maintained (i.e., the shape identical to the sample). An initial test revealed little generalization of matching to novel stimuli, but after handsigns were trained to these stimuli, accurate generalized matching appeared immediately. The results replicated prior findings and demonstrated particular features of stimulus control sufficient to enable generalized matching. A behavioral account of relational matching was supported. The technique used in this study was shown to be effective in teaching abstract relations to nonverbal retarded children.