Researchers studied the effect of cognitive conflict between peers in a collaborative problem-solving task in which the relationship between participants could not be described as "expert-novice" and in which development was potentially possible for all. A total of 156 subjects between 5 and 9 years of age, inclusive, participated in the study. Subjects were required to predict the direction a beam would tip when different numbers of weights were placed at differing distances from the fulcrum. Seven increasingly sophisticated rules for prediction have been distinguished. On the basis of pretest rule use, children were assigned to one of three treatment conditions: individuals retested as controls, an "equal rule" group pairing subjects who used the same rule, and an "unequal rule" group pairing subjects who used different rules. Children in each pair were of the same age, sex, and class in school. During the treatment phase, disagreements in prediction were resolved by discussion. The first posttest was administered approximately 3 days later and the second, 1 month later. Pairs did only slightly better than control individuals. Equal rule pairs did not show signs of development. Among unequal rule pairs, "lower partners" improved and "higher partners" declined. Concluding discussion examines three possible explanatory factors: amount of arguments, quality of reasoning, and cognitive state of the child. (RH)