Two studies explored the experience and performance of students at Williams College in three-person groups that were homogeneous or heterogeneous in rated academic ability. In accord with hypotheses from Festinger's (1954) social comparison theory, students in academically homogeneous groups had more positive experiences and performed better on measures of written and video-taped performance. These results differ somewhat from recent studies of peer effects among roommates and from a line of recent social comparison research regarding the effect of exposure to superior others on one's own performance. In addition, students in single-sex groups had higher scores on several self-report and performance measures. Qualifying this finding were additional results showing that women did better in single-sex, while men did better in mixed-sex groups. The overall results were framed in terms of social comparison dynamics. [Support for this research was provided by the Andrew Mellon Foundation through grants to the Williams Project on the Economics of Higher Education.]