As a two-phase exercise in inter-district audit, with the emphasis on critical evaluation of routine clinical practice, three rheumatologists each examined the same 44 patients with shoulder pain, and recorded their diagnosis and the investigations and treatment they would carry out. In the first phase, 26 patients were seen by each rheumatologist separately; there was complete diagnostic agreement in only 46%, with wide variation in the frequency of requests for standard investigations, but all three rheumatologists recommended steroid injections for most patients. In the second phase, all three rheumatologists examined a further 18 patients together, discussed the symptoms and signs, and recorded their diagnoses separately. There was complete agreement in 78%. The presence of more than one lesion, and differences in the interpretation of certain physical signs, partly explain the lack of agreement in Phase 1. Treatment of specific shoulder lesions is highly concordant, with injection the major treatment modality, followed by physiotherapy. Perhaps the different diagnoses reached, and the fact that treatment might therefore be administered for the wrong diagnosis, may explain some treatment failures. Also, recruitment of patients for studies of the treatment of shoulder lesions requires care to avoid selection of a heterogeneous group.