This study evaluates the most valid and reproducible method for directly measuring anteversion (torsion) in dried femora using a commercially available measuring machine. Each femur was placed horizontally on the surface of the machine and readings were obtained from the head, the shaft distal to the lesser trochanter, and the distal end. Using computer software, four different anteversion angles were calculated: the center head-neck line to the retrocondylar line (Method 1); the center head-neck line to the transcondylar line (Method 2); the anterior head-trochanter line to the retrocondylar line (Method 3); and the anterior head-trochanter line to the transcondylar line (Method 4). The methods were applied to 20 femora, which were measured twice by one observer. The most reproducible method of measuring femoral anteversion uses the bone surfaces on the anterior aspect of the head and greater trochanter and on the posterior aspect of the condyles (Method 3,95% confidence limits of t 0.4"). The other methods are shown to be reproducible to t 2.4", t 3.3" and t 1.7" (Methods 1,2, and 4, respectively, 95% confidence limits). Conversion factors between the different methods are: 12.5" + 0.8 x (anteversion angle) to change each of Method 2 to Method 1 and Method 4 to Method 3; and 8" + 0.7 x (anteversion angle) to change each of Method 1 to Method 3 and Method 2 to Method 4 (correct to within t 3"). o 1993 Wiley-Liss, Inc.