An extension of previous attempts at modifying children's cognitive styles is discussed. Specifically, the present study employed sociometric peer models in order to ascertain whether: (1) impulsivity is modifiable through observation of salient models, and (2) sociometrically selected peer models are more effective than unselected models in the reduction of errors associated with increased response times. The 19-item Sutton-Smith and Rosenberg Impulsivity Scale 3 was administered to 98 fourth and sixth grade children (50 in the fourth grade and 48 in the sixth). Fifty children (28 males and 22 females) scored at or above the median of 10 on the IMP3 Scale; 25 of the children were assigned to the experimental condition and 25 to the control condition. The experimental design of the study was implemented in three stages: (1) pretest evaluation of cognitive style, using the Matching Familiar Figures test; (2) training in which the children in the experimental group were individually exposed to a trained reflective model, who was named by the experimental child, and who were not classified as impulsives; the children in the control group were not exposed to models; and (3) children in both groups were individually pretested on nine items of the Matching Familiar Figures test. The study data were subjected to statical analysis. Results indicated that impulsive children showed a change in response style after viewing a reflective model. Generally, males in the experimental condition showed a greater increase in mean response time than did females, as well as a stronger decrease in mean number of errors. (DB)