Abstract
<p>Four-, five- and six-year-old children, 24 with mild mental retardation and 24 with learning disabilities, were matched with a normally achieving child of the same age, gender, race/ethnicity and testing language. All 96 children were presented the second version of a new screening test, comprised of ten different tasks which required active cognitive processing. Tasks were evaluated using nonparametric and parametric methodology. A task was retained for further evaluation if a cutoff on a frequency distribution could be established below which there were at least three exceptional children for each normally achieving child. Next, frequency distributions for each of the tasks were systematically combined and improvements in the associated sensitivity (SN) and specificity (SP) were used as criteria for task retention. Seven tasks were selected: picture pointing, semantic information picture probe, standard oddity, picture sequencing, relative size, taxonomic generation, and semantic information verbal probe. Internal estimates of classification accuracy were, SN = 83% and SP = 92%. Predictive discriminant analysis was conducted to get external estimates of the classification accuracy (i.e., SN = 83% and SP = 81%). Exact interrater reliability was $>$70% and scoring concordance within one point was $\geq$95%. These data indicate the potential effectiveness of this new rest for detecting mild cognitive impairments in young children.</p>