The hypothesis that a history of one or more weight reductions and regains (weight cycling) is associated with lower site-specific bone mineral density (BMD) was examined in 169 premenopausal women, aged 29-46 years. Data on the previous 10-years' weight cycling history, present weight-bearing physical exercise, number of deliveries, present use of contraceptive pills or hormone-releasing coils, age at menarche and present menstrual status were collected by a self-administered questionnaire. Dietary intake was calculated from food records. The areal BMD (g/cm2) was measured with dual-energy X-ray absorptiometry (Norland XR-26). The lumbar spine (L2-4) BMD, adjusted to weight and age at menarche (ANCOVA), was 0.062 g/cm2 (95% confidence interval: 0.015 to 0.011 g/cm2; p = 0.01) higher in the non-cyclers (n = 68) than in subjects with reported weight-cycling history (n = 101). The corresponding difference for femoral neck BMD was 0.019 g/cm2 (-0.018 to 0.056; p = 0.30), for trochanter BMD 0.013 g/cm2 (-0.025 to 0.05 g/cm2; p = 0.50) and for distal radius BMD 0.022 g/cm2 (0.006 to 0.397 g/cm2; p = 0.008). A pairwise comparison of 34 weight-matched subjects (non-cycler vs cycler) gave similar BMD differences as found in the above (ANCOVA) analyses. The results suggest that weight cycling might be associated with lower spine and distal radius BMD.