PURPOSE: It is common to assert that restoration of normal knee kinematics is essential for the best functional result after knee arthroplasty. Previous studies using the progression of the geometric centre axis have suggested that kinematics after unicompartmental arthroplasty is markedly different from the normal. For this study, the transepicondylar axis was used because this axis is closer to the flexion axis and should be a better reference for motion. The following hypothesis was tested: the transepicondylar axis would again show that the postoperative kinematics does not restore normal motion and is closer to that before replacement. METHODS: Seventeen osteoarthritic knees were tested before and after unicompartmental arthroplasty using a three-dimensional to two-dimensional registration technique tracking the transepicondylar axis to calculate translation and rotation of this axis. Results were compared for the seventeen knees before and after arthroplasty and were compared to the normal knee as measured in our previous study. RESULTS: Similar motion patterns in the pre- and postoperative knees were shown but both the pre- and postoperative motion were markedly different from the normal knee. CONCLUSIONS: This result supported our hypothesis. The clinical relevance is that medial unicompartmental arthroplasty cannot restore the motion of the knee to normal in the living knee. Therefore, it would be expected that the patient for unicompartmental knee might not feel normal. It may not be possible depending on ligaments alone to restore the knee to normal, and the changes in the articular shapes and the surgical procedure may also be necessary.