Cementless surface replacement arthroplasty of the shoulder is designed to replace the damaged joint surfaces and restore normal anatomy with minimal resection of bone. We have used the Copeland shoulder arthroplasty for 14 years. Between 1986 and 2000, 285 surface replacement arthroplasties were implanted in our unit. The prosthesis has evolved during this time, but the principle of minimal bone resection has remained the same. Between 1990 and 1994, 103 Mark-2 prostheses were inserted into 94 patients (9 bilateral). The operations were carried out for the treatment of osteoarthritis, rheumatoid arthritis, avascular necrosis, instability arthropathy, post-traumatic arthropathy and cuff arthropathy. The mean follow-up was for 6.8 years (5 to 10). The best results were achieved in primary osteoarthritis, with Constant scores of 93.7% for total shoulder replacement and 73.5% for hemiarthroplasty. The poorest results were seen in patients with cuff arthropathy and post-traumatic arthropathy with adjusted Constant scores of 61.3% and 62.7%, respectively. Most patients (93.9%) considered their shoulder to be much better or better than before the operation. Of the 88 humeral implants available for radiological review, 61 (69.3%) showed no evidence of radiolucency, nor did 21 (35.6%) of the 59 glenoid prostheses. Three were definitely loose, and eight shoulders required revision (7.7%), two (1.9%) for primary loosening. The results of this series are comparable with those for stemmed prostheses with a similar follow-up and case mix. The cementless surface replacement arthroplasty diminishes the risk of complications involving the humeral shaft and periprosthetic fractures. Revision or arthrodesis can be undertaken easily since the bone stock has been maintained with no loss of length.