This study examined the effects of instructional control and aptitude composition on performance, interaction, and attitudes during a computer-based cooperative science lesson. A sample of 152 fifth and sixth grade students completed the instruction either under the learner control (LC) or the program control (PC) treatment. In the LC treatment, students exercised control over the amount, review, and sequence of instruction. Students in the PC treatment followed a predetermined instructional path. At the end of the lesson, all students completed a posttest and an attitude questionnaire. They also indicated their confidence by predicting the score on the posttest. An identical test was administered two weeks later for measuring retention. Time for each group and interaction in representative groups were also recorded. Results suggested that both heterogeneous grouping and learner control had significant effect on learning, time on task, verbal interaction, and attitudes. Working in heterogeneous groups increased confidence more than working in homogeneous groups. High-ability students outperformed low-ability students on all dependent variables, except attitudes. The implications for designing microcomputer software and future research are discussed. Statistical tables showing means and standard deviations for scores of achievement, confidence, retention, attitude, instructional time, and verbal interaction are appended. (Contains 46 references.) (Author/KRN)