The notion that greater learning outcomes will be achieved if the cognitive work is distributed amongst a group of individuals working together versus working alone has received mixed support when explored empirically (e.g., Daiute & Dalton 1993; Johnson & Johnson, 1991). This study examined the relationship between small-group collaborative learning structures and the potential predictors of groups' overall academic performance. We sought to identify specific factors that distinguished high-performing groups from low-performing groups in the classroom. Class attendance and individual-level academic performance were positively related to group-level academic performance. Further, it was predicted that groups consisting of an exceptionally high-performing member, or superstar, would achieve greater group-level academic performance than groups consisting of members who performed similarly. However, the greater the distance between the highest-performing member's score and the average of the other group members' scores on individual-level tasks, the lower the score on group-level tasks. This difference between the highest scoring group member and the rest of the members is referred to as the Superstar Difference Score. Qualitative and quantitative analyses indicated that the Superstar Difference Score is a reliable, negative predictor of group-level academic performance. Practical implications for classroom instructors and future directions for education research resultant from this study's superstar effect are discussed. (Contains 7 tables.)