The results of clinical shoulder examination including 20 special tests were compared with subsequent arthroscopic findings in 45 patients. The sensitivity of the clinical diagnosis was 73%. Impingement syndrome was correctly diagnosed in 19 of 22 cases (86%), rotator cuff tears in 7 of 9 cases (78%). The highest sensitivity for stage II impingement was found for the supraspinatus test (85%) and the lift-up test (92%); the sensitivity of these tests for rotator cuff tears was 100% and 89% respectively. Differentiation between impingement syndrome with and without rotator cuff tear by one of these tests alone was not possible because of their low positive predictive values (26% and 56%). In contrast, in 90% of patients with negative rotator tests the rotator cuff was complete, while the negative predictive value of the supraspinatus test was 100%. Instability was confirmed in only 53% of cases; the Leffert test had the highest positive predictive value (73%). In conclusion, the clinical diagnosis of a shoulder lesion cannot reliably be achieved by single tests; rather overall evaluation by an experienced clinician is necessary.