The present study examined the relationship between the cognitive skill of role-or perspective-taking and naturally occurring behavior of behaviorally disordered children. Twenty-six boys, aged five years, nine months to twelve years, two months were tested and observed at their treatment facility. It was predicted that children who could take the perspective of others would prefer peer to adult interaction, would more likely give positive attention to their peers and would be more likely to use affective language than their non perspective-taking peers. These and related hypotheses were examined by observing each participant's interactive behavior for 36 minutes distributed over three different settings, lunch, freetime and organized activity on six or more 2 different days. To determine perspective-taking skill, two perspectivetaking instruments were administered in a separate room at the treatment site. One measure (the Chandler role-taking task) required a child to tell a story from a series of three cartoon pictures and then retell the story from the point of view of a late arriving bystander. The other task (the Friendship interview from the Selman Measure of Interpersonal Understanding) assessed role-taking on the basis of the child's responses to questions about a filmstrip story that depicted a conmon dilemma between close friends. The variety and frequency of affective words was assessed by counting the affective words used by the child when responding to the first role-taking task, the cartoon stories. A vocabulary test was administered at the same time as the other cognitive measures. Before data analysis began, such methodological concerns as reliability of the observational code, reliability of the judges' scoring of the role-taking tasks and internal consistency of the measures were addressed. Cognitive measures, use of affective language and behavioral categories were then correlated with each other. The vocabulary test was used to partial general verbal skill from the relationship of role-taking and affective language. In addition to examining relationships among the measures, the children were divided into nerspective-taking and non perspective-taking groups and compared on the various behavioral and language measures.It was found that perspective-takers engaged in more neutral interaction with their peers than did non perspective-takers. While there was no negative relationship between perspective-taking and neutral interaction with adults, it was found that perspective-takers received less positive and negative attention from adults. Although no relationship was found between perspective-taking and th~ general categories of distributing positive attention to peers and adults, there was a relationship between a specific sub-category of positive attention, sharing and the Selman perspective-taking measure. Helping behavior was marginally correlated with the Chandler perspective-taking task. No relationship was found between perspective-taking and receiving positive or negative attention from pee...