Body weight is one of the principal determinants of bone density and fracture frequency, but there is significant disagreement in the literature regarding the relative contributions of the lean and fat components of body weight to this relationship. As previous studies have not considered the possible role of exercise in soft tissue-bone density interrelationships, we measured areal bone mineral density (BMD), fat mass, and lean mass in eumenorrheic premenopausal women and determined whether the interrelationships of these variables are influenced by the subject's exercise status. Subjects with mean activity levels of more than 140 kilojoules/kg.day (equivalent to undertaking vigorous physical activity for > 1.5 h/week) were classified as exercisers. In the nonexercising subjects (n = 36; age, 36 +/- 8 yr), BMD was markedly weight dependent (0.45 < r < 0.62), and this was contributed to by both fat and lean tissue. Because this finding may have arisen from the mutual dependence of soft tissue mass and areal BMD on body size, fat and lean masses were reexpressed as a percentage of body weight. The percent fat tended to be positively related to areal BMD (0.23 < r < 0.35), whereas the percent lean was inversely related to this index. A second way of obviating the mutual dependence of soft tissue mass and areal BMD on body size is to derive BMD/height as an index of volumetric bone density. This parameter was only related to lean mass in the femur, whereas the correlations with fat mass were little changed. The percent fat was positively (0.29 < r < 0.43) and the percent lean was negatively (-0.43 < r < -0.29) related to BMD/height throughout the skeleton, including the femur. In the exercising subjects (n = 63; age, 33 +/- 8 yr), fat mass and lean mass were unrelated to BMD/height (r < 0.23). However, the percent lean was positively correlated with BMD and BMD/height in the femoral neck (r = 0.28 and r = 0.31, respectively). It is concluded that bone density is only associated with fat mass in sedentary women. In exercisers, femoral neck density is related to lean mass, possibly through the effects of weight-bearing exercise on both of these variables.