1. Age-related bone loss is well established but reported rates of bone loss in the spine and femur vary widely. The aim of the present study was to investigate changes in bone mineral density in the lumbar spine and proximal femur in healthy postmenopausal women. 2. One hundred and thirty-eight population-based women, aged 45-65 years, recruited from general practice registers in 1990, were assessed at baseline; 108 returned for repeat assessment 4 years later, of whom 31 had taken hormone replacement therapy for 12 months or more of the 4-year study period. Bone densitometry of the lumbar spine and proximal femur was performed by dual-energy X-ray absorptiometry. Serum 25-hydroxyvitamin D and oestradiol were measured by RIA and serum intact parathyroid hormone by radio-immunometric assay. 3. The mean age at follow-up was 62 years (mean of 13.6 years after menopause). Lumbar spine bone mineral density was significantly higher in women who had received hormone replacement therapy for more than 12 months during the study period than in those who had not (P < 0.01). There was no difference between these two groups in the femoral neck or trochanteric bone mineral density. In the lumbar spine, the annual change in bone mass in untreated women was -0.39% (95% confidence intervals -0.60 to -0.09; P < 0.02) whereas there was a small gain in women receiving hormone replacement therapy [+0.36% (-0.12 to 0.84; P not significant)]. The annual change in bone mass in the femoral neck and trochanter was -0.51 and -0.45 respectively in untreated women (P < 0.01 and P < 0.02), and -0.16 and -0.15 in those receiving hormone replacement therapy (P not significant). 4. Our results demonstrate relatively low rates of bone loss in the spine and proximal femur in these healthy, population-based peri- and postmenopausal women. Hormone replacement therapy appeared to be associated with a significant protective effect on spinal, but not femoral, bone mineral density.