The screening potential of a taxonomic information task for the detection of learning disabled and mildly retarded children

Abstract

Normally achieving (NA) students were compared to mildly mentally retarded (MMR) same-age peers (Experiment 1) and to learning disabled (LD) same-age peers (Experiment 2). In both studies the same task was presented. The students were asked to (a) describe similarities and differences among exemplars of 12 different categories, (b) identify the categories, and (c) name the exemplars. In Experiment 1, group comparisons yielded large, consistent performance differences between the mildly retarded and their normally achieving chronological age (CA) matches. High levels of sensitivity, specificity, and predictive accuracy were obtained when task performance was used to assign individual students to educational groups. These levels were maintained or improved when using a reduced set of 5 categories and the different descriptors measure. Using cut-off scores from this reduced set was also associated with accurate student identification. In experiment 2, there were small mean differences favoring the NA over the LD students on all 4 dependent measures but none was significant. None of the psychometric measures was associated with adequate levels of classification accuracy, although identification was improved when only a subset of categories and a single dependent measure was employed. However, a subset of LD students generated fewer different descriptors than any of their same age NA peers indicating that the task might be sensitive to a subset of LD children.