During a two-year period, eighty-nine patients who were scheduled to have a total knee arthroplasty for the treatment of degenerative osteoarthrosis were randomly assigned to one of two groups: resurfacing of the patella or retention of the patella. All patients received the same posterior cruciate-sparing prosthesis, and all operations were performed by, or under the direct supervision of, one of us. Three patients died in the early postoperative period. The remaining eighty-six patients (118 knees; fifty-eight that had had resurfacing of the patella and sixty that had not) formed the study group. They were followed for a mean of thirty months (range, twenty-four to forty-four months). Evaluation was performed with use of the clinical scoring system of The Knee Society, a patient-satisfaction questionnaire, specific questions regarding patellofemoral symptoms and function, and radiographs. All clinical evaluations were performed by the same research nurse, without the involvement of a physician, in a blinded manner (neither the nurse nor the patient had knowledge of whether the patella had been resurfaced). Preoperatively, the mean Knee Society score, on a scale ranging from 0 to 200 points, was 89.7 points (range, 33 to 132 points); postoperatively, this score improved to a mean of 172.7 points (range, 98 to 200 points). With the numbers available for study, we could detect no significant difference between the knees that had had patellar resurfacing and those that had not with regard to the over-all score (p = 0.63), the subscore for pain (p = 0.56), or the subscore for function (p = 0.77). We also could detect no difference between the treatment groups, with the numbers available, with regard to patient satisfaction or the responses to questions involving the function of the patellofemoral joint, including the ability to exit from an automobile, to rise from a chair, and to climb stairs. Thirty-two patients had bilateral total knee replacement with resurfacing of the patella in one knee and retention of the patella in the other. These patients expressed no clear preference for either knee. Eight (13 per cent) of the sixty knees that had not had resurfacing were painful anteriorly compared with four (7 per cent) of the fifty-eight that had; this was not a significant difference (p = 0.38), with the numbers available. The anterior pain that was noted postoperatively was predominantly of new onset; it had not been observed preoperatively in three of the four knees that had had resurfacing or in four of the eight that had not. No additional treatment options were offered to the patients who had anterior pain in the knee after resurfacing. However, six (10 per cent) of the sixty knees that had not had resurfacing had it subsequently, because of anterior pain in the knee, after the twelfth postoperative month (range, fifteen to thirty-nine months). The pain decreased in four of these knees. Thus, total knee arthroplasty with retention of the patella yielded clinical results that were comparable with those after total knee arthroplasty with patellar resurfacing, but it was associated with a 10 per cent prevalence of the need for subsequent resurfacing. The prevalence of anterior pain after total knee arthroplasty was not influenced by whether or not the patella had been resurfaced. The postoperative clinical scores, the postoperative development of anterior pain, and the need for subsequent resurfacing were not predicted by the presence of preoperative anterior pain, obesity, or the grade of chondromalacia observed intraoperatively. Because of the short duration of follow-up, these results should be considered preliminary. Additional follow-up is planned.