Epidemiological studies of bone mineral determinants rely heavily on measurements made with absorptiometric techniques such as single-photon absorptiometry and dual-energy x-ray absorptiometry. In general, absorptiometric data are expressed as areal densities (bone mineral density, BMD), obtained by dividing bone mineral content (BMC) by bone area or width (BA, BW). This size correction assumes that BMC and BA (BW) are directly proportional to one another, such that a 1% change in BA (BW) is matched by a 1% change in BMC. This is rarely the case, and the exact relationship depends on the population group, skeletal site, body size, instrumentation, and scanning conditions. Size adjustment determined by using predefined indexes, such as BMD and body mass index (BMI, wt/ht2), may fail to correct BMC fully for bone and body size, and may lead to spurious associations with other size-related variables such as calcium intake, energy expenditure, and grip strength. A general approach to size adjustment is described, in which BA (BW), weight, and height are incorporated in all regression models of BMC. Although BMD plays a valuable role in fracture-risk assessment and clinical management, we advocate that its use in epidemiological research be discontinued.