A cohort of 3,595 white women aged 40-77 years was followed for an average of 10 years during which 84 new cases of hip fracture were identified. Triceps skinfold thickness and arm muscle area measured at baseline were examined as possible risk factors for hip fracture controlling for physical activity, height, menopausal status, calcium consumption, and smoking. Of these variables only arm muscle area, triceps skinfold thickness, and activity in recreation were independent predictors of hip fracture incidence using the Cox proportional hazards model. After adjustment, the estimated relative risk of hip fracture was approximately two for an increment of each anthropometric indicator (adjusted for the other) equivalent to comparing those at the 25th percentile to those at the 75th percentile (maximum width of 95% confidence intervals, 1.2-2.9). Risk of hip fracture was approximately two-fold for persons who reported little recreational exercise compared to persons who reported much recreational exercise (95% confidence interval, 1.2-3.2). Our findings are the first evidence from a prospective study that anthropometric indicators besides body mass index may have an independent relationship to risk of hip fracture.