This study examined student aptitudes and student behaviors during small group interactions as mediators of the effectiveness of small group learning. The hypotheses to be investigated were that the effects of small group learning on student achievement are produced by students' participation in group interaction, and in the providing and receiving of higher-order explanations. High and low ability students are more often involved in this process than other students. A treatment and control group of fifth grade students completed ability and attitude pretests. The treatment group was trained in small group interaction. All students received regular mathematics classroom instruction. They worked on assignments in mixed ability groups of four students. Achievement, retention and attitude toward mathematics were assessed. A Mann-Whitney comparison showed that trained students participated in more task related interaction than control students. The effects of small group interaction depend on the ability level of the students. Interaction during small group work was most beneficial for low ability students. The study showed they can help themselves by teaching others. A high quality of interaction must prevail if the small group method is to be of maximal effectiveness. (Author/DWH)