In 100 knees undergoing a total replacement, the angles between the tangent line of the posterior condylar surfaces, the anteroposterior axis as described by Whiteside, the transepicondylar line, and the trochlear line were measured. Also measured were the sulcus angle, the transepicondylar width, the height of the condyles, and the thickness of the various cuts. Radiologic measurements made were: the mechanical angle, the hip center-femoral shaft angle, the transcondylar angle, and the tibial plateau-tibial shaft angle. The mean values of these measurements were calculated, and comparisons were made according to gender and the mechanical axis using the Student's t test. Correlations between the various measurements were calculated. The transepicondylar axis was found to be a reliable landmark to properly rotate the femoral component, and was easier to locate at surgery than the anteroposterior axis. In trochlear dysplasia and in some valgus knees, relying on the anteroposterior axis can induce an excessive external rotation of the femoral component, and the opposite can happen in some varus knees. The anterior extent of the condyles is highly variable in arthritic knees, and cannot be used to orient the prosthesis. The ratio between the transepicondylar width and the height of the condyles is constant, but some narrow femora could require narrower implants to avoid medial-lateral overhang of the femoral component.