Although all agree that the results of total knee replacement (TKR) are primarily determined by surgical skill, there are few satisfactory alternatives to the 'apprenticeship' model of surgical training. A system capable of evaluating errors of instrument alignment in TKR has been developed and demonstrated. This system also makes it possible quantitatively to assess the source of errors in final component position and limb alignment. This study demonstrates the use of a computer-based system to analyse the surgical skills in TKR through detailed quantitative analysis of the technical accuracy of each step of the procedure. Twelve surgeons implanted a posterior-stabilized TKR in 12 fresh cadavers using the same set of surgical instruments. During each procedure, the position and orientation of the femur, tibia, each surgical instrument, and the trial components were measured with an infrared coordinate measurement system. Through analysis of these data, the sources and relative magnitudes of errors in position and alignment of each instrument were determined, as well as its contribution to the final limb alignment, component positioning and ligament balance. Perfect balancing of the flexion and extension gaps was uncommon (0/15). Under standardized loading, the opening of the joint laterally exceeded the opening medially by an average of approximately 4 mm in both extension (4.1 +/- 2.1 mm) and flexion (3.8 +/- 3.4 mm). In addition, the overall separation of the femur and the tibia was greater in flexion than extension by an average of 4.6 mm. The most significant errors occurred in locating the anterior/posterior position of the entry point in the distal femur (SD = 8.4 mm) and the correct rotational alignment of the tibial tray (SD = 13.2 degrees). On a case-by-case basis, the relative contributions of errors in individual instrument alignments to the final limb alignment and soft tissue balancing were identified. The results indicate that discrete steps in the surgical procedure make the largest contributions to the ultimate alignment and laxity of the prosthetic knee. Utilization of this method of analysis and feedback in orthopaedic training is expected rapidly to enhance surgical skills without the risks of patient exposure.