A prospective study was done of fifty total shoulder replacements in forty-four patients who were followed for an average of 3.5 years (range, 2.0 to 7.5 years). The preoperative diagnosis was osteoarthritis in thirty-three shoulders, rheumatoid arthritis in eleven, and a previous fracture of the humeral head in six shoulders. Nine of the shoulders had a tear of the rotator cuff. The Neer-II system of total shoulder replacement with a non-metal-backed component was used in all fifty shoulders. At follow-up, forty-four shoulders (88 per cent) had no significant pain. Of the six painful shoulders, four had loosening of the glenoid component and one had malposition of both components. Three of the six shoulders had no significant pain after revision. The average range of active forward elevation in all of the shoulders improved from 71 to 100 degrees, and both external and internal rotation improved as well. Five specific activities of daily living were evaluated, and the patients' ability to perform them improved from 14 to 78 per cent. Thirty-five shoulders (68 per cent) were rated by the patients as much better; thirteen shoulders (26 per cent), as better; and three (6 per cent), as no better. An incomplete lucent line was noted around five (10 per cent) of the cemented humeral stems and around thirty-seven (74 per cent) of the glenoid prostheses. Four arthroplasties (8 per cent) required revision: three for loosening of the glenoid component and one for malposition of both components.(ABSTRACT TRUNCATED AT 250 WORDS)