Large-scale assessment programs are beginning to design group assessment tasks in which small groups of students collaborate to solve problems or complete projects. Little is known, however, about the validity of data from group assessment for making inferences about the competence of individual students. This study compared performance in small-group and individual assessment contexts to determine how well achievement scores from the group setting represented the mathematics skills of individual students and to determine what additional information about students' skills was provided by data on their behavior in the group setting. All students performed at high levels in the group setting; the profile of group interaction was a better predictor of individual achievement than was individual performance in the group setting.