Stereoradiography with a base shift of the source of illumination was used to produce pairs of radiographs to be measured by stereophotogrammetric techniques. The direction of shift was parallel with the longitudinal axis of the body, so that each radiograph in the stereopair could be used for other clinical purposes. A base shift of 10 centimetres with a distance of 100 centimetres between the focus and the film gave acceptable value of stereoscopic parallax. The radiographs were measured using a Hilger and Watts medical stereometer. This method was checked with test specimens, namely an osteotomised pelvis in which one acetabulum could be rotated and an osteotomised femur in which the whole upper portion could be rotated against the shaft. Measurements made on the acetabulum and its radiographs showed a correlation coefficient of 0.9838 over the range 0 to 30 degrees of anteversion, with a mean error +2.54 degrees and a standard deviation of +/- 1.52 degrees (n = 21). For the femoral neck, over the range from 10 degrees of retroversion to 80 degrees of anteversion, the correlation coefficient was 0.9979, the mean error +2.46 and the standard deviation +/- 1.48 degrees (n = 30).