Objective: to evaluate the effects of brisk walking on bone mineral density in women who had suffered an upper limb fracture. Design: randomized placebo-controlled trial. Assessments of bone mineral density were made before and at 1 and 2 years after intervention. Standardized and validated measures of physical capacity, self-rated health status and falls were used. Setting: district general hospital outpatient department. Subjects: 165 women drawn from local accident and emergency departments with a history of fracture of an upper limb in the previous 2 years. Women were randomly allocated to intervention (self-paced brisk walking) or placebo (upper limb exercises) groups. Intervention: both groups were seen at 3-monthIy intervals to assess progress, measure physical capacity and maintain enthusiasm. The brisk-walking group were instructed to progressively increase the amount and speed of walking in a manner that suited them. The upper limb exercise placebo group were asked to carry out a series of exercises designed to improve flexibility and fine hand movements, appropriate for a past history of upper limb fracture. Results: drop-outs from both intervention and placebo groups were substantial (41%), although there were no significant differences in bone mineral density, physical capacity or health status between drop-outs and participants. At 2 years, among those completing the trial, bone mineral density at the femoral neck had fallen in the placebo group to a greater extent than in the brisk-walking group [mean net difference between intervention and placebo groups 0.019 g/cm2, 95% confidence interval (CI) −0.0026 to +0.041 g/cm2, P= 0.056]. Lumbar spine bone mineral density had increased to a similar extent (+0.017 g/cm2) in both groups. The cumulative risk of falls was higher in the brisk-walking group (excess risk of 15 per 100 person-years, 95% CI 1.4–29 per 100 person-years, P < 0.05). There were no significant differences in clinical or spinal x-ray fracture risk or self-rated health status between intervention and placebo groups. Conclusion: the promotion of exercise through brisk-walking advice given by nursing staff may have a small, but clinically important, impact on bone mineral density but is associated with an increased risk of falls. Self-paced brisk walking is difficult to evaluate in randomized controlled trials because of drop-outs, placebo group exercise, limited compliance and lack of standardization of the duration and intensity of walking. Further work is needed to evaluate the best means of safely achieving increased activity levels in different groups, such as older women and those at high risk of fractures.