Knowledge of the effects of exercise on bone mass in postmenopausal women is limited and controversial. Animal studies have shown that the response of bone to bending strain is an alteration of bone geometry. We studied 250 postmenopausal women, aged 52-72 years, willing to participate in a 6-month exercise program. The first 125 started the program immediately and the remaining 125 served as controls. The training program included exercises designed to maximize the stress on the wrist. One hundred and eighteen of the active group and 116 of the control group completed the study and were reassessed 6 months later. Bone mineral density (BMD) of the femoral neck, lumbar spine, ultradistal and proximal radius was measured by dual-energy X-ray absorptiometry (DXA) both before and at the end of the exercise program. The forearm was also evaluated by peripheral quantitative computed tomography, which measures the area, bone mineral content (BMC), and volumetric density for both the cortical and the trabecular component. The results showed that the DXA measurements at the femoral neck, lumbar spine, ultradistal and proximal radius were similar between the two groups. No significant difference was detected after the exercise program at the proximal radius. At the ultradistal radius, the cross-sectional area of cortical bone rose by 2.8 +/- 15.0% (SD, p < 0.05), apparently for both periosteal apposition and corticalization of the trabecular tissue. The volumetric density of cortical bone rose by 2. 2 +/- 15.8% (p < 0.1), and that of trabecular bone decreased by 2.6 +/- 10.7% (p < 0.01). The combined changes in both bone volume and density in the exercise group were associated with marked increase in cortical BMC (3.1 +/- 10.7%, p < 0.01) and decrease in trabecular BMC (-3.4 +/- 14.2%, p < 0.05), which were statistically different from those observed in the control group (p < 0.05). In conclusion, these results confirm that site-specific moderate physical exercises have very little effect on bone mass. However, it appears that some exercises may reshape the bone segment under stress by increasing both the cross-sectional area and the density of the cortical component. These structural changes are theoretically associated with increases in the bending strength.