Four hundred and six patients presenting with shoulder pain underwent bilateral shoulder ultrasonography. Rotator cuff lesions were diagnosed in 197 patients. In 155 of the 308 patients who had additional arthrographic examinations, the arthrogram documented rotator cuff tears. Sixty-eight patients underwent surgery. There was good correlation between the ultrasonographic and arthrographic findings as well as the surgical results, with a sensitivity of 91% for each examination technique. Thirty-six postoperative patients were studied ultrasonographically; a minor to marked increase in rotator cuff echogenicity was demonstrated in them all. Its high predictive value makes ultrasonography the method of choice in diagnosing rotator cuff tears. Based on the evaluation of particular criteria, rotator cuff tears (greater than 1 cm) are commonly documented by ultrasonography. Only in patients with indeterminate or negative ultrasonograms in whom there is a high index of clinical suspicion, should arthrography be performed as a complementary imaging test. In postoperative follow-up the degree of scar formation at the reinsertion of the tendon is adequately documented by ultrasonography.