The results of twenty-four non-constrained total shoulder replacements that were done in twenty patients who had treatment of rheumatoid arthritis were retrospectively reviewed to determine how those results were affected by the severity of the disease. All of the patients had Class-IV functional capacity, and 92 per cent had Stage-III or IV rheumatoid progression. Nine (38 per cent) of the shoulders had a tear of the rotator cuff. The mean length of clinical follow-up was 4.5 years (range, two to ten years). Preoperatively, all of the patients had disabling pain and limited function. Postoperatively, twenty-two (92 per cent) of the patients had no appreciable pain, and eighteen (75 per cent) had no significant functional limitation (p less than 0.001). Active elevation improved by 88 per cent, and external and internal rotation also improved significantly. Motion, relief of pain, and functional improvement were not significantly greater in the patients who had an intact rotator cuff. Radiolucent lines developed around ten (42 per cent) of the glenoid prostheses, but only two of the prostheses were surrounded by a complete line and were thought to be loose. No revisions were done. We believe that a non-constrained total shoulder replacement affords excellent relief of pain, satisfactorily improves range of motion, and improves function in patients who have severe rheumatoid involvement of the shoulder. However, because motion and function are severely restricted preoperatively, the end-results are not comparable with those that have been reported for patients who have less severe rheumatoid disease.