As part of a school-wide course transformation project at the University of Hawaii to improve student learning and retention, multiple geology and oceanography instructors are introducing two-stage exams (Gilley and Clarkston, 2014) in their undergraduate courses. The first stage is the traditional exam, where students take the exam individually. The second stage (which directly follows the first) is collaborative, in which groups of students answer the same (or a subset of) questions posed during the first stage. The group turns in a single exam paper, forcing the students to reach consensus on each answer. Groups may be formed by the instructor or self-selected by students, and can vary in size. Instructors can weigh the two stages of the exams however they like.

We analyzed 289 student scores on 14 two-stage midterm and final exams given by six different instructors. For each exam, the mean group score (stage two) exceeded the mean individual score (stage one), and all gains were statistically significant at α=0.05. Students who scored in the bottom quartile of the individual exam experienced the greatest mean improvement from individual to group. Students who scored in the top quartile of the individual exam had a lower, but still statistically significant, mean increase. The vast majority of groups had a group score that exceeded the scores of all individuals in that group, which argues against the theory that the increased group score is due to group members simply copying answers from the top-performing individual in their group. A cohort analysis revealed that groups containing all combinations of high- and low- performing students during stage one experienced statistically significant mean gains in exam scores, and selecting groups to include a mix of high- and low- performing students can be a highly effective way to proactively reduce the achievement gap.