Attentional responses during discrimination learning by retarded children

Abstract

In Experiment I, eight mentally retarded children were trained on a simultaneous two-choice discrimination problem and a series of discrimination-shift problems. Subjects performed overt observing responses to produce elements of the discriminative stimuli, making it possible to measure directly changes in attention to different aspects of stimuli during learning. The patterns of change in observing responses were generally in line with descriptions of attentional changes derived from two-process theories of discrimination learning; for example, the frequency of irrelevant observing responses was high during the pre-solution period during extradimensional shifts but was low during intradimensional shifts. Contrary to current theories, extradimensional shifts caused an immediate increase in irrelevant responses, and intradimensional shifts caused an increase in relevant observing responses. Subjects responded to later shift problems by initially increasing both relevant and irrelevant observing responses, then withholding irrelevant observing responses. Experiment II examined the effects of three variables, the provision and non-provision of a mechanical observing response, the stimulus dimensionality, and degrees of relevance of the irrelevant stimuli, on the discrimination learning and transfer performance of sixteen mentally retarded children. The subjects were trained on simultaneous two-choice discrimination problems using complex stimuli containing either dimensional stimuli or mixed- ‘dimensional stimuli. Subjects were then tested on either intradimensional or extradimensional shifts. When dimensional stimuli were used, intradimensional shifts were easier than extradimensional shifts but when mixed-dimensional stimuli were used, the relative difficulty of the intradimensional and extradimensional shifts depended on whether observing-response buttons were provided. When observing-response buttons were provided, intradimensional shifts were again easier than extradimensional shifts but when these buttons were not provided, intradimensional shifts were harder than extradimensional shifts. The relative ease of intradimensional over extradimensional shifts was found to be further affected by the degree of relevance of the less-relevant dimension. The superiority of the intradimensional over extradimensional shift performance was progressively reduced and then eliminated as the degree of relevance of the irrelevant dimension was gradually increased. Subproblem analysis showed that subjects typically treated subproblems independently, even though there was some degree of dimensional analysis of stimuli, and intradimensional shifts were usually easier than extradimensional shifts. This finding is inconsistent with the usual interpretation of the relative ease of intradimensional over extradimensional shifts as an indication of non independence of subproblems.