Social networks and friendships were tracked over a 3-week period for 132 students enrolled in fourth- and seventh-grade classrooms. Individual interviews were employed to collect data on friendships. Social group membership was determined by the composite social-cognitive map (SCM) procedure and by self-reports. Considerable overlap was found among the methods for establishing relational patterns (i.e., friendships, self-reported groups, composite social-cognitive maps). When loose criteria for stability were employed, high stability was found in friendships (p = .56-.75) and social group membership (p = .90) over a 3-week period. But when stringent criteria were employed, only modest social relationship stability was observed in both methods, suggesting that there is much greater fluidity in peer bonds than has been generally recognized. The relationship between friendship and social network measures, the distinctive information yielded by social networks, and gender and age differences in group structure, fluidity, and friendships are discussed.