We reviewed 124 consecutive kinematic condylar total knee replacements (in ninety-one patients) at two to four years postoperatively. One hundred and eleven (90 per cent) were rated as good or excellent. The average active postoperative flexion was 106 degrees (range, 94 to 120 degrees). Twenty-two knees (18 per cent) had incomplete, non-progressive radiolucent lines, less than one millimeter in width, at the tibial bone-cement interface; these were considered insignificant. Restriction of stair-climbing ability in this series appeared to be a function of involvement of multiple joints rather than of patellar replacement, as the rheumatoid patients with resurfaced patellae performed the worst. The osteoarthritic patients with involvement of a single joint performed the best, regardless of whether the patella was resurfaced or not. When compared with a similar series of total knee replacements in which the tibial component was made entirely of plastic, less reaction at the bone-cement interface was found with the metal-backed kinematic tibial component. We suggest that this finding is clinical confirmation of in vitro studies that demonstrated the advantage of metal-backed tibial components. We concluded that this procedure, if meticulously performed, will give predictably good to excellent results with a low complication rate, a good postoperative range of motion, and a favorable-appearing bone-cement interface at two to four years.