To determine the most accurate roentgenographic technique for the measurement of limb-length discrepancy, the twenty lower extremities of ten cadavera were measured with use of both orthoroentgenograms and lateral scout computed-tomographic scanograms. The actual anatomical limb length also was measured, to serve as a control. The effects of flexion of the knee joint, use of an external fixator, cost and time of the examination, and exposure to radiation also were determined. No statistically significant difference in the measurements of the length of the femur was found between the two methods at neutral or at 15, 30, or 45 degrees of flexion of the knee. However, computed tomography was significantly more accurate than orthoroentgenography in the measurements of length of the tibia and of total length of the limb when the knee was flexed to 30 degrees or more (p less than 0.01). The placement of an Ilizarov fixator did not alter the results. The cost and time necessary to complete an examination were comparable for the two methods. However, computed tomography delivered only 20 per cent of the radiation needed for orthoroentgenography. Computed tomography is more accurate than orthoroentgenography for the measurement of limb-length discrepancy in patients who have a flexion deformity of the knee.