The objectives of this study were to evaluate (1) the effect of spinal muscle strengthening by loading exercises on the bone mineral density (BMD) of the spine, and (2) the effect of upper extremity loading exercises on the BMD of the midradius and femur in healthy, premenopausal women. The study design was a randomized, controlled trial of 3 years' duration. Ninety-six healthy, premenopausal, white women aged 30-40 years participated; 67 completed the study. All subjects were in good health (normal menses) and were active, but not athletic (that is, not involved in a regular sport activity). Subjects were randomized to an exercise or control group. The exercise group performed a supervised, non-strenuous, weight-lifting exercise program. Exercise performance was supervised once a week at the medical facility. In addition, the subjects performed the exercises twice a week on their own. Dietary calcium intake was to be maintained at 1,500 mg/day in both groups. Bone density was measured at the lumbar spine and hip with dual-energy X-ray absorptiometry at 0, 1, and 3 years. BMD of the midradius was measured with single photon absorptiometry. Measurements of muscle strength were obtained at baseline and every 3 months for 3 years. Maximal oxygen uptake was measured, and the level of physical activity was recorded. Compliance with the exercise program was excellent during the first year of the study, but decreased thereafter. At the end of 3 years, subject withdrawal was about 34% from the exercise group and about 22% from the control group (total subject withdrawal was about 30%). Muscle strength in the exercise group increased significantly at all involved skeletal sites (p values all < 0.001). There was a modest positive correlation between the BMD of Ward's triangle with spinal flexor strength (r = 0.32, p = 0.008) and with grip strength (r = 0.38, p = 0.001). Comparing study groups, we found no significant effect of the loading and nonstrenuous strengthening exercises in the exercise group or free physical activity group (our control group) on BMD at the spine, hip, or midradius measurement sites. In active, but not athletic premenopausal women, additional moderate weight-lifting exercises showed no significant effect on BMD.