In 55 patients with type I or type II impingement lesions we performed arthroscopic subacromial decompression. Fifty-two patients are followed up 1 year postoperatively. In all patients the condition of the affected shoulder before and after decompression was documented using a 100-point shoulder score (pain on activity, 15 points; pain without activity, 15 points; function, 20 points; weight lifting, 10 points; muscle strength, 15 points; range of motion, 25 points). At follow-up we also documented the extent of passive inferior shift of the humeral head by ultrasound. The mean score preoperatively was 60.9 (+/- 13.8). Postoperatively there was a significant increase to 84.7 (+/- 12.5). The average postoperative hospital stay was 8.8 days (+/- 2.1). In 12 patients (23%) the postoperative score was less than 85 points, and in these the treatment was considered to have failed. Comparison of these patients as a group with those in whom the treatment was successful revealed no difference in age, a small but not significant difference in the preoperative duration of shoulder complaints, and no difference in the postoperative length of stay in hospital. However, there was a significant difference in the extent of passive inferior shift of the humeral head: in the failure group the mean inferior shift was 4.6 +/- 1.9 mm, while in the other patients the shift was only 2.7 +/- 1.0 mm. This difference was statistically highly significant. There was a statistical highly significant negative Pearson correlation coefficient of -5.56 between postoperative score and inferior shift of the humeral head. We conclude that patients with subacromial pathology and hypermobile glenohumeral joints may not be good candidates for subacromial decompression.