To determine whether the presence and extent of a tear of the rotator cuff could be predicted on the basis of a patient's history, physical examination, and radiographic findings, detailed data from the histories and physical examinations of 103 patients who were known to have a tear of the rotator cuff were correlated with the radiographic and operative findings on these patients. An age-matched control group of fifty-one patients who had similar symptoms, but whose arthrograms showed normal results, was used to establish a baseline incidence of ten specific radiographic findings in the shoulder. Two discrete groups of patients who had a tear of the rotator cuff were identified. Twenty-eight patients (27 per cent) had a tear of a single tendon; the histories and the physical and radiographic findings in this group were consistent with a symptomatic local mechanical-impingement process in the shoulder. Sixty (80 per cent) of the seventy-five patients in the other group had a history of acute trauma to a shoulder. The patients in this second group were older and were non-athletic, and had not previously had symptoms that were severe enough to need treatment. These patients were subsequently found to have a complete tear of more than one of the tendons of the rotator cuff. Multiple radiographic findings in the shoulder and other coexisting orthopaedic conditions also were more common in these patients. In this group, we believe that acute trauma in a shoulder that had chronic degenerative changes, rather than localized mechanical impingement, caused the tendons to rupture.