Intense exercise during childhood and adolescence may result in primary amenorrhea and low peak bone mineral density (BMD). After puberty, exercise may result in secondary amenorrhea and bone loss. Higher BMD in amenorrheic athletes than amenorrheic sedentary persons suggests that exercise may partly offsets the effects of amenorrhea. To examine this possibility, we measured BMD (g/cm2) by dual x-ray absorptiometry in 32 ballet dancer and 23 healthy controls of comparable age with regular menstrual cycles, 34 pre-pubertal female gymnasts bone age 8.9 +/- 0.2 years and 37 girls matched by bone age. Dancers had normal BMD at the weight bearing sites, not low, despite having oligomenorrhea, not high despite 32 hours of week dancing. BMD was lower by 4-6 percent at the non-weight bearing sites. BMD diminished in the dancers at the weight bearing femoral neck (r = -0.29, P = 0.1) and trochanter (r = -0.31, P = 0.09), and at the non-weight bearing arms (r = -0.29, P = 0.09) with increasing duration of amenorrhea. Dancers with less than 40 months amenorrhea had 5 to 7% higher BMD at the weight bearing, but not non-weight bearing sites. Dancers with more than 40 months amenorrhea had normal, not higher BMD at weight bearing sites and deficits of about 5 percent at non-weight bearing sites. In gymnasts, BMD was 10-15 percent (or 1 SD) higher than the bone age-predicted mean. Exercise may not offset the effects of amenorrhea. Bone loss may continue but from a higher level, perhaps attained prior puberty.