The purpose of this study was to evaluate an ultrasound method on dried femora to decide if it is sufficiently reliable to consider its application to living subjects in health and disease. It reports the application of a standard static image B-mode ultrasound scanner to measure femoral anteversion (strictly torsion) on dried bones. The method involves creating images of bony surface contours at both the upper and the lower ends of the femur placed in a water bath. The calculation of femoral torsion involves measuring the angular difference between a head-neck line at the upper end and a transcondylar line at the lower end of the femur. This approach, using the surface contour of the bone at the upper end of the femur, is similar to that used by Lange & Pitzen (1921) and differs from conventional ways of measuring anteversion in dried bones using a central femoral head-neck axis. It is shown that in ultrasound scanning of the upper end of the dried femur to measure femoral anteversion, the head-neck line must be at, or rostral to, the base of greater trochanter. The intra- and inter-observer reliability was assessed in 10 femora, each scanned ten times by two observers. It is shown that femoral torsion can be measured by the ultrasound method to an accuracy of +/- 2 degrees (95% confidence limits) without a significant difference between the observers. The ultrasound method is compared with a new mechanical method for creating surface contours at both ends of the femur and found not to be statistically different in 37 femora. Two real-time ultrasound methods reported by other workers for measuring femoral anteversion in dried bones are reviewed. It is suggested that in applying ultrasound to measure femoral anteversion in living subjects, the B-mode static image scanner is superior to the use of real-time scanners.