Recent legal challenges to race-conscious college admissions processes have called into question what constitutes a sufficient level of diversity on college campuses. Previous research on the educational benefits of diversity has examined the linear relationship between diversity interactions and student outcomes, but multiple theoretical frameworks suggest that this relationship may be curvilinear. The present study investigated this possibility using a longitudinal sample of 8,615 first-year undergraduates at 49 colleges and universities. The results indicate that rare or moderate diversity interactions are associated with virtually no growth (and sometimes even slight declines) in leadership skills, psychological well-being, and intellectual engagement, whereas very frequent diversity interactions are associated with considerable growth. The results are similar regardless of students' race, institutional characteristics, and whether the interactions are interracial or across multiple forms of difference. Implications for institutional practice and future research are discussed.