The positive effects of physical activity on human bone mass have been well documented in many cross-sectional studies comparing athletes with sedentary controls as well as in longitudinal follow-up. By applying peripheral quantitative computed tomography (pQCT), which has the advantage of measuring volumetric bone mineral density (BMD) and the ability to distinguish among trabecular and cortical components, it was demonstrated that cortical BMD of the dominant arm was not greater than that of the nondominant arm. Cortical drift toward the periosteal direction and an increase in cortical thickness resulted in an improvement of mechanical characteristics of the playing arm's midradius. An improvement in the mechanical properties of young adult bone in response to long-term exercise was therefore related to geometric adaptation, but not to an increase in BMD. The manner in which the recruitment and function of bone cells are coordinated differs between the growing and the nongrowing skeleton. In the former, modeling is the dominant mode, and in the latter it is remodeling. In the present study, the side-to-side difference of 92 middle-aged female tennis players who initiated training after bone had matured was analyzed by pQCT. The side-to-side difference detected suggested a paradoxical adaptation of the mature radius to unilateral use during tennis playing, and that tennis playing after bone had matured did not stimulate cortical drift in the periosteal direction, unlike that seen in young subjects. Unexpectedly, the cross-sectional areas (periosteal and endocortical area) of the radius were smaller in the dominant arm than in the nondominant arm in the middle-aged female players. The findings suggest that unilateral use of the arm after the third decade of life suppresses age-related changes in bone geometry.