Abstract: | A sample of 162 children aged from 5 to 9 was pretested to discover each child's "rule" for predicting the movement of a mathematical balance beam. Children then worked alone, with a partner who used the same rule, with a partner who was more competent, or with a partner who was less competent. If partners' predictions differed, the dyad members were asked to discuss and reach agreement, but were not given feedback. All children were subsequently given 2 individual posttests. The results revealed that regression in thinking was as likely a consequence as improvement, both proving stable. Benefits accrued primarily to those whose partner was more competent, but understanding of the outcomes of collaboration required attending both to the nature of the rules (whether they allowed consistent or inconsistent prediction) and the shared understanding attained during the paired session. |