In order to maximize student access to microcomputers, teachers have assigned students to work at the computer in groups. There is little empirical evidence, however, to indicate the effect of grouping students on the social and instructional interactions that occur within those groups. Twenty learning-handicapped and twenty nonhandicapped boys from upper elementary grades participated for 10 minutes, three times per week for a 4-week period on a mathematics problem-solving task. Students worked in three different configurations: individually, with a handicapped partner, and with a nonhandicapped partner. All sessions were videotaped and interactions that occurred were assigned to categories using an observational instrument. Results indicated that nonhandicapped students behaved differently in heterogeneous dyads and homogeneous dyads. In heterogeneous dyads they made more management comments and used the computer rather than their partner as a resource. The task-related behavior of learning-handicapped students was similar in both configurations. As the task became more difficult, instructional behavior increased in all dyads. In all dyads, interactions were primarily task-related and positive in tone.