The aims of this study of premenopausal parous women were to determine whether low bone mass could be accurately identified by clinical risk factors and to describe the effect of an information and bone mineral density (BMD) feedback program on lifestyle behavior. The subjects were a convenience sample of 271 women who took part in a cohort study of cot death in 1988, and in a population-based study of the determinants of bone mass in 1996. These subjects were provided with BMD feedback according to their T-score. Those with a score < -1.0 at either the femoral neck or lumbar spine were sent a letter indicating that their BMD was low (n = 72), whereas those scoring above -1.0 at both sites (n = 199) were told that their BMD was normal. Both groups were given a comprehensive osteoporosis information leaflet. In late 1997 we were able to contact 256 subjects (95%). In logistic regression, the presence or absence of low BMD was correctly classified in 79% of cases (p < 0.0001) by a model containing body mass index, fracture history, smoking, breastfeeding history, and sports participation. The model had poor sensitivity (38%) but high specificity (93%). At follow-up, those with low BMD had similar smoking cessation rates (16 vs 17%, p = 0.93) but higher rates of increased calcium intake (61 vs 9%), calcium supplement use (39 vs 4%), and increased physical activity (41 vs 17%) (all p < 0.001) in comparison with those with normal BMD. We conclude that a feedback program can alter self-reported behavior in young women for at least 12 mo and that the magnitude of effect is greatest in those with low BMD. Identification of these subjects by clinical risk factors is good but suboptimal, suggesting that measurement of BMD may be necessary to target most accurately and effectively those at highest risk.