To assess the accuracy of magnetic resonance (MR) imaging in the evaluation of the shoulder after surgery, MR examinations were performed in 31 patients before repeated surgery, and MR findings were correlated with the subsequent operative findings. In addition, the MR findings associated with prior surgery were reviewed, including altered structure of the acromion, soft-tissue metal artifacts, a surgical trough in the humeral head, nonvisualization of the subacromial-subdeltoid fat stripe, and intermediate signal intensity within the substance of the rotator cuff on images obtained with all pulse sequences. The MR criteria for full-thickness tears in the shoulder after surgery are the presence of fluidlike signal intensity on T2-weighted images that extends through an area of the rotator cuff or the nonvisualization of a portion of the rotator cuff. With use of these criteria, six of seven full-thickness rotator cuff tears were identified at MR imaging, with an accuracy of 90%. Partial cuff tears were indistinguishable from repaired tendons. Findings at MR imaging were 74% accurate concerning the presence of impingement.