Thirty consecutive patients who underwent double con trast CT arthrography prior to shoulder arthroscopy were prospectively studied. Results from both studies were recorded with the radiologist blinded to the ar throscopic findings when making the final CT reading. The sensitivity, specificity, and accuracy, respectively, of CT arthrography findings compared to arthroscopy were 50%, 100%, 96% for rotator cuff; 66%, 100% 96% for bicipital labral complex; 100%, 100%, 100% for loose bodies; 50%, 100% 93% for Hill-Sachs le sions; 90%, 73%, 83% for anterior labral defects; and 100%, 100%, 100% for posterior labral defects. We conclude that shoulder arthroscopy accurately deline ates abnormalities of the anterior and posterior labrum, bicipital labral complex, rotator cuff, joint synovium, and humeral head. CT arthrography accurately delineates capsular redundancy, loose bodies, hardware around joints, and bony glenoid rim abnormalities.