Social cognitive mapping (SCM) is a common approach to identifying peer groups in developmental research. However, this approach involves three stages that each implies a unique conception of peer group. This article aims to bring conceptual clarity to the identification of peer groups using SCM by demonstrating how the meaning of peer groups differs at each stage of SCM. First, in the data collecting stage, interaction groups identify sets of children that hang out together. Second, in the data aggregating stage, co‐membership groups identify sets of children who are members of many of the same interaction groups. Third, in the data analyzing stage, similarity groups identify sets of children with similar patterns of relationships with their peers. After reviewing these three conceptions of peer groups, we briefly discuss some potential problems with using SCM as a tool to measure children's social networks and peer groups. Finally, we conclude by arguing that despite these issues, SCM remains a valuable methodology, and indeed one with untapped potential. Thus, we offer suggestions for the appropriate application of these theoretically and empirically distinct conceptions of peer group, noting that developmental researchers using SCM must identify which conception of peer group is used and justify why this conception is the appropriate one.