The purpose of this investigation was to determine the short-term reproducibility of measurements of whole-body and subregion bone mass and density, as well as body composition, made by dual energy x-ray absorptiometry. Bone mineral content, bone mineral density and body composition were measured on two occasions, 1 to 2 weeks apart, in 21 women (average age, 20.9 [standard deviation 1.6] years). The method errors of the duplicate measurements, expressed as a percentage of the combined mean values from the two sets of measurements (i.e., as a coefficient of variation), for whole-body bone mineral content, bone mineral density, lean tissue mass and fat mass were 1.6%, 1.1%, 1.4% and 1.8% respectively. The method errors for bone mineral density in the hip were 2.2%, 1.1% and 2.5% for the neck, trochanter and Ward's triangle respectively. On the basis of the method errors and the expected treatment effects, the sample sizes needed for intervention trials (e.g., exercise training) were calculated. All of the whole-body and most of the subregion bone mineral density and lean tissue mass measurements made by dual-energy x-ray absorptiometry were sufficient for detecting the small changes (about 2%) expected in trials with 20 subjects, whereas measurements of subregion bone mineral content and fat mass were less precise.