This twin study estimates familial clustering of moderate and intense leisure-time physical activity and investigates quantitatively its genetic and environmental components. Study subjects are 3,344 male twin pairs aged 33-51 yr. Moderate activity levels were assessed with six questions about discretionary walking or stair climbing for exercise. Five questions assessed regular participation in specific, intense athletic activities (running, bicycling, swimming, racquet, and other sports). Familial aggregation is estimated by odds ratio of one twin engaging in an activity when his co-twin does. Monozygotic and dizygotic twin correlations were compared to estimate genetic and nongenetic sources of phenotypic variation. For each activity, the familial aggregation odds ratio was statistically significant with values between 2.9 to 4.6 for intense activities and between 1.4 to 1.9 for all moderate activities but one. Monozygotic twin correlations were higher than dizygotic, suggesting genes play a role in the observed phenotypic variation. For four questions, and a compromise scale of moderate activity, the difference between correlations was statistically significant (P < 0.05). In this cohort, much of the phenotypic variability for both moderate and intense activities is a result of familial effects. Genes may influence regular participation in specific intense exercises more than moderate activity, such as walking for exercise.