Thirty-two emotionally disturbed and 23 normal adolescents were given two test of cognitive style developed by Kagan (Matching Familiar Figures and the Haptic-Visual Tests). The results were analyzed to evaluate two major hypotheses: (a) that emotionally disturbed adolescents perform differently on the tests than normal adolescents; and (b) that members of both groups reveal a high degree of stability among various subtest scores, both within the same test and across different tests. In testing the first hypothesis, the authors discovered minimal differences in mean scores for the different samples on each of the five subtests involved. However, a more complex analysis was carried out on the basis of the assumption that extreme scores on the various items were a more meaningful indicator of performance than simple means. This analysis yielded the conclusion that these emotionally disturbed adolescents were more impulsive (i.e., answered more quickly and made more errors) than their counterparts in public schools. An analysis of the intercorrelations between different tests and within tests in testing the second hypothesis revealed 19 out of 40 significant correlations, with an additional 14 out of the remaining 21 in the predicted direction. Finally, nine out of the 10 correlations combining the subgroups were statistically significant.