We estimated genetic effects on bone density in pre- and postmenopausal twins and critically considered the assumptions of the twin model. Bone mass in the radius, lumbar spine, and hip, anthropometric measurements, usual calcium and caffeine intake, tobacco and alcohol use, number of pregnancies and live births, menstrual history, usual physical activity, and medical history were measured in a volunteer sample of 171 twin pairs [124 monozygotic (MZ) and 47 dizygotic (DZ)], aged 25-80, free of diseases known to affect bone mass or mineral metabolism. At all skeletal sites, MZ intraclass correlations exceeded DZ correlations for both pre- and postmenopausal women, yielding highly significant estimates of heritability for bone mass. Adjustments for height, age, and environmental characteristics did not reduce the heritability estimates. However, many of these estimates were unrealistically high, suggesting some violation(s) of the assumptions of the twin model. Thus, the familial resemblance in bone mass is due primarily to genetic effects at all skeletal sites and at all ages, although the importance of genetic effects is diminished with aging, as evidenced by increasing within-MZ pair variability in older women. Because of failures in the assumptions of the twin model, however, particularly the greater MZ environmental similarity and the probability of gene interaction, heritability estimates are probably too high and require cautious interpretation.