When a course designed around cooperative, problem-centered instruction attracts learners with a wide range of experience in the topic, should learners be grouped heterogeneously or homogeneously in terms of their relative expertise? In this study, learners were randomly distributed between the two types of groups; learning gains, satisfaction, and problem-solving discourse were compared. Overall, no significant differences were found between heterogeneous and homogeneous groups. However, groups solving relatively ill-structured problems exchanged significantly more elaborated explanations than groups solving relatively well-structured problems.