33% of the students were bused to school from a lower class neighborhood (of Middle- Eastern background) and distributed equally throughout all the classrooms.

The study involved 351 students in all nine 8^th grade classes in this school; five were assigned at random to the Group Investigation (G-I) method, and four to the whole-class (W-C) approach. Six history and geography teachers, out of a total of 11, employed the Group-Investigation method. The preceding year almost all teachers had participated in 10 workshop sessions. All classes were observed several times during the course of the second year by outside observers to ensure that the classes were conducted according to the principles of the G-I and W-C methods (Y. Sharan & S. Sharan, 1992). The dependent variables measured were: students' academic achievement on teacher-prepared examinations, students' verbal behavior and the cognitive strategies employed during group discussions (67 6-person groups for 30 minutes per group) that were videotaped and analyzed by judges.

In short, findings revealed that students from the G-I classes expressed themselves more frequently and used more words per turn of speech than did their peers from the W-C classes. Students' from the W- C classes whose parents came from Western countries took more turns of speech during the group discussion, compared to their peers whose families were from the Moslem countries of the Middle East (all of the children in this study were Jewish). Turn taking in the G-I groups was symmetrical among students from the two ethnic groups. Furthermore, students from the G-I classes, from both ethnic groups, addressed more cooperative statements to Middle Eastern students than did their peers from the W-C classes where the Middle Eastern students' lower status in the group discussions was particularly obvious. Finally, scores on tests of academic achievement were higher among students from the G-I than from the W-C classes. This finding was true for students from both ethnic groups.

The central concepts of this study are: specific kinds of interaction among students that support learning; structured interdependence by means of a learning task that motivates students to participate; tasks that stimulate dis-agreement and "conflict" between group members and promote the search for a solution; workgroup size; student productivity. Each of these concepts receives careful theoretical elucidation in the introduction. The study included 36 3^rd and 4^th grade classrooms; in 12 classes students worked individually, in another--12 students worked collaboratively in groups where interactions were unstructured, and in the remaining 12 classes students also worked un-collaboratively with interactions structured "via role, goal, resource and reward interdependence" (page 188). Classes were also randomly assigned to dyadic or small group composition. Following initial training "students completed four weekly classroom sessions under these conditions." During a fifth session one workgroup which included the lowest-and-highest achieving students from each classroom was videotaped. The relatively large