We studied 336 elderly white women, of whom 22 had previously suffered a hip fracture and 22 had previously suffered a vertebral fracture. All subjects were 60 years old or older with a mean age of 73.7 years. Measurements of ultrasonic transmission velocity (UTV), broad-band ultrasonic attenuation (BUA) and stiffness (STF) were made at the os calcis using a Lunar Achilles ultrasound device. Measurements of lumbar spine bone mineral density (L2–4 BMD) and femoral neck BMD were made using dual-energy X-ray absorptiometry. The fracture groups were significantly older and had more years since menopause than the control groups. Logistic regression showed that measurements of UTV, STF and BUA discriminated between fracture and non-fracture subjects for both the hip (p<0.001) and spine (p<0.05). Femoral neck BMD discriminated both hip and vertebral fractures from controls (p<0.001 andp<0.01, respectively). Spinal BMD discriminated between subjects with vertebral fractures and those without (p<0.01), but not hip fractures (p=0.64). For hip fracture, areas under receiver-operating characteristic (ROC) curves were 0.85 for UTV, 0.83 for STF, 0.79 for BUA, 0.78 for femoral neck BMD and 0.53 for spinal BMD. For vertebral fracture, areas under the ROC curve were 0.68 for UTV, 0.70 for STF, 0.66 for BUA, 0.66 for femoral neck BMD and 0.67 for spinal BMD. To determine whether calcaneal ultrasonic measurements discriminated, independently of BMD, fracture from control groups, UTV, BUA and STF were adjusted for BMD, age and years since menopause using multiple regression analysis and the residuals from the regressions were incorporated into a logistic regression analysis. Adjusted ultrasonic measurements discriminated hip fracture from control groups (p<0.005 for UTV;p<0.05 for BUA;p<0.01 for STF) but not vertebral fracture (p=0.37 for UTV;p=0.53 for BUA;p=0.25 for STF). These results show that, when ultrasonic measurements were adjusted for BMD and age, they still discriminated between hip fracture and control groups. This finding supports the hypothesis that ultrasonic measurements contain information about bone strength not contained in bone density measurements that may be useful in predicting hip fractures. Therefore, calcaneal ultrasound measurements may provide a safe, low-cost addition to bone densitometry.