High peak bone mass in early adulthood is an important protective factor against osteoporotic fractures in later life, but little is known about the effects of exercise on growing bone. The purpose of this cross-sectional study was to determine at which state of maturity (Tanner stage), the areal bone mineral density (BMD) differences between the playing and nonplaying arms of junior tennis players become obvious, and to clarify in each developmental stage which training and background variables, if any, could explain the interindividual differences in bones' response to mechanical loading. Ninety-one 7- to 17-year-old female tennis players and 58 healthy female controls were measured. In each Tanner stage, differences in BMD in playing and nonplaying (dominant and nondominant) arms (proximal humerus, humeral shaft, and distal radius) and BMD of the lumbar spine and nondominant distal radius were compared between the controls and players. Within each Tanner stage of players, the associations between training and background variables and BMD differences were analyzed with Spearman rank correlation coefficients. In players, BMD differences between the playing and nonplaying arms were significant (P < 0.05- < 0.001) in all Tanner stages, with the mean difference ranging from 1.6 to 15.7%. In controls, these dominant-to-nondominant arm differences were clearly smaller (ranging from -0.2 to 4.6%), but significant at some measured sites. In comparison with the relative side-to-side arm differences between the players and controls (i.e., examination of the training effect), the mean difference was not obvious and significant until the adolescent growth spurt (i.e., the girls in Tanner stage III with a mean age of 12.6 years). In the lumbar spine, significant BMD differences between players and controls were not found until Tanner stage IV (mean age 13.5 years; 8.7%, P < 0.05) and V (mean age 15.5 years; 12.4%, P < 0.05). In a nonloaded site of the skeleton (nondominant distal radius), no significant BMD differences were found between the players and controls in any Tanner stage. In the correlation analysis, the Tanner I and II players (mean ages 9.4 and 10.8 years) showed no significant associations between any of the predictive variables and the side-to-side BMD differences, while in Tanner stages III, IV, and V, such associations could be found; the total amount of training hours during the playing career and the number of training sessions per week showed a significant and systematic correlation (rs ranging from 0.43 to 0.80) with the side-to-side BMD differences in several measured bone sites. In conclusion, this study suggests that in a majority of female junior tennis players, the benefit of unilateral activity on bone density does not become clearly evident until the adolescent growth spurt or Tanner stage III. The total amount of training during the player's career and the current training frequency (sessions per week) seem to best explain the training effect on bone tissue, leaving, however, room for speculation on the still unknown factors that modulate the loading response of a growing bone.