RESULTS. Sixty-seven (41%) of the 163 patients had a joint effusion, bursal fluid, or both. Joint effusion alone was seen In 35 patients. Fourteen of these had a normal rotator cuff at surgery, and 21 had a rotator cuff tear (sensitivity, 22%; specIficity, 79%; positive predictive value, 60%). Bursal fluid alone was seen In 10 patIents, seven of whom had a rotator cuff tear (sensitivity, 7%; specificity, 96%; posItive predictive value, 70%). In 22 patients, fluid was seen in both the bursa and the joint; 21 had curgically proved rotator cuff tears (sensitivity, 22%; specificity, 99%; posItive predictive value, 95%). Of the 232 asymptomatlc shoulders, 16 (6.9%) had Isolated joint effuslons, eight (3.4%) had isolated bursab effusions, and four (1 .7%) had both joint and bursal effusions.CONCLUSION. The sonographic finding of lntraarticular fluid alone (without bursal fluid) has both a low sensitivity and a low specificity for the diagnosis of rotator cuff tears. However, the finding of fluid In the subacromial/subdeltold bursa, especially when combined with a joint effusion, Is highly specific and has a high positive predictive value for associated rotator cuff tears. Sonographically detected fluid In both the joint and the bursa is an uncommon finding in asymptomatlc shoulders. The sonographic observation of fluid In the subacromial bursa, either Isolated or cornbined with a joint effusion, should prompt a careful evaluation of the supraspinatus tendon for tear.