The knees of 20 patients with a unicompartmental arthroplasty on one s ide and a total replacement on the other were compared at follow-ups of over two years. Length of hospitalization was slightly less with the unicompartmental knees and recovery was faster. By one year, however, there was essentially no difference between the unicompartmental knees and the total replacements in spite of the fact that the former had a better range of movement. This suggests that flexion beyond 105 degrees is not of great significance to elderly people living in the West. It also suggests that, as the posterior cruciate was resected or lengthened in the majority of the total knee replacements, the presence or absence of this ligament does not affect the activities of daily living. Complications both early and late were more frequent with a unicompartmental knee, possibly implicating inferior patient selection, instrumentation, and implant design.