The purpose of this study was to document predictors of long-term retention among minority participants in the Well Elderly 2 Study, a randomized controlled trial of a lifestyle intervention for community-dwelling older adults. The primary sample included 149 African American and 92 Hispanic men and women aged 60 to 95 years, recruited at senior activity centers and senior residences. Chi-square and logistic regression procedures were undertaken to examine study-based, psychosocial and health-related predictors of retention at 18 months following study entry. For both African Americans and Hispanics, intervention adherence was the strongest predictor. Retention was also related to high active coping and average (vs. high or low) levels of activity participation among African Americans and high social network strength among Hispanics. The results suggest that improved knowledge of the predictors of retention among minority elders can spawn new retention strategies that can be applied at individual, subgroup, and sample-wide levels.