From 1979 to 1984, eighty patients (119 knees) were arbitrarily selected for treatment with knee arthroplasty in which a posterior cruciate-substituting replacement was used. The average age of the forty-nine women and thirty-one men was 66.9 years (range, twenty-two to eighty-four years). Sixty-one right and fifty-eight left knees were operated on, and bilateral replacement was performed in thirty-nine of the eighty patients. The diagnosis was osteoarthritis in fifty-eight patients (eighty-eight knees), rheumatoid arthritis in fourteen patients (twenty-two knees), osteonecrosis in three patients (four knees), and traumatic arthritis secondary to a fracture of the tibia or femur in five patients (five knees). The average preoperative score on The Hospital for Special Surgery knee-rating scale was 47.5 points, and the average range of motion preoperatively was 88 degrees (range, 30 to 140 degrees). Of the 119 knees, eighty-seven had a varus alignment (maximum, 35 degrees) before knee replacement. After follow-up of two to eight years, the average score on The Hospital for Special Surgery scale was 90 points, and the average range of motion was 107 degrees. Of the 119 knees, 83 per cent were rated as excellent; 15 per cent, as good; none, as fair; and 2 per cent, as poor. Radiolucencies of one millimeter were present in 76 per cent of the knees; of two millimeters, in 7 per cent; and of three millimeters, in 3 per cent. No statistically significant correlation between radiolucencies and the clinical result was found. The results in knees of patients who had rheumatoid arthritis were not as good as those in knees of patients who had other diagnoses (F = 11.44). Our experience suggested that the posterior cruciate-substituting design provides more motion than do the cruciate-sacrificing surface-replacement designs, with no deleterious effects. The rate of infection (1.6 per cent) after these procedures, which were carried out in a standard operating theater with vertical airflow, was equivalent to that in other published series in which rooms with laminar airflow were used. Patients who had a bilateral procedure did as well as those who had a unilateral replacement, but they required approximately 3.5 more units of blood.