College students, ranked as high, medium, or low ability, were assigned to homogeneous or mixed-ability groups of two through five members for a test. Performance proved proportional to the number of high ability group members. The underlying group process was inferred through social decision scheme analysis. (Author/SJL)