Most research on small group learning has focused on achievement, but few studies have systematically investigated the effects of group processes on achievement, or the influences of individual aptitudes and group composition on group processes. To investigate the relationship between student aptitudes, group process, and achievement in cooperative small groups in junior high school mathematics classrooms, and the stability of the relationship over time, students (N=105) in four classrooms participated in two studies. Initially, all students learned a one-week unit on consumer mathematics. Three months later, half of the students learned a one-week unit on probability. Students worked in four-person homogeneous-ability or heterogeneous-ability groups; they also completed achievement tests and the Eysenck Personality Inventory. Analyses of the data indicated that group process was a potent predictor of achievement in all studies; "asking a question and receiving no answer," the best predictor of achievement, was detrimental to achievement. The effects of group composition and student aptitudes on achievement were mediated by the group process variable. The mediating effect of group process and the magnitude of coefficients were stable across studies; group process was stable over time, both in average frequency and in individual student levels of participation. The results suggest a need to determine whether the stability of group process is generalizable to longitudinal designs. (Author/NRB)