Fifty patients with signs and symptoms of chronic impingement syndrome and/or rotator cuff tear were evaluated with shoulder arthrograms and ultrasonography. Ninety asymptomatic shoulders of a comparable age had ultrasonography to serve as a control group. All controls had normal ultrasonograms with no hypoechoic or sonolucent areas. The 50 symptomatic shoulder patients had the following: 28 had a normal arthrogram with either normal ultrasonograms or buckling of the supraspinatus tendon; eight had normal arthrograms, but ultrasonography indicated a thin (less than 4 mm) irregular supraspinatus tendon; 11 had complete rotator cuff tears visualized on both arthrography and ultrasonography (nine of these 11 patients had surgery confirming complete tears in all); two had a false-positive sonogram, and one had a false-negative sonogram. Thus, the ultrasonography's overall positive predictive value to detect a full-thickness rotator cuff tear was 85%, its negative predictive value was 97%, its sensitivity was 92%, and its specificity was 95%. Real-time ultrasonography is a diagnostically sensitive and specific noninvasive method to evaluate patients with shoulder impingement syndrome, leading to the recommendation that it be used as a primary imaging technique to obviate or supplement arthrography in evaluating rotator cuff disease. However, static ultrasonographic pictures, without real-time ultrasonography as a supplement, were not helpful.