Knee joints affected by rheumatoid arthritis (RA) were randomized to either arthroscopic or open synovectomy, and the postoperative results were compared retrospectively. The materials consisted of 58 patients with 75 surgically treated knees. Arthroscopic synovectomy was performed on 37 knees of 30 patients, who were followed for five years and four months on average. Open synovectomy was performed on 38 knees of 28 patients, who were followed for five years and five months on average. The results were evaluated for ; recurrence of synovitis ; duration of the symptom-free period ; progressive bone changes on radiography ; range of motion (ROM) ; duration of hospital stay ; duration of surgery. Due to the less invasive surgical procedure, arthroscopic synovectomy resulted in less post operative pain, and there was no restriction of ROM after surgery, no need for postoperative rehabilitation, shorter hospital stay, lower incidence of osteoarthritic changes, and earlier return to the daily life. However, as synovial membrane removal was unavoidably insufficient, the effect of surgery lasted for only 18 months on average and recurrence of synovitis was observed in 73 % of patients. Open synovectomy meant relatively greater invasion and postoperative pain. Due to the restricted ROM in some cases, postoperative treatment was required and the hospital stay tended to be prolonged. Nevertheless, the recurrence rate was only 37 % during follow-up and the effect of surgery persisted for an average of 75 months, much better than that of arthroscopic synovectomy. In the treatment of RA knees, the indications for arthroscopic and open synovectomies should be carefully studied in the light of the merits and for demerits of each procedure.