We attempted to assess the accuracy of magnetic resonance imaging (MRI) in determining the size of recurrent cuff tears in correlation with size measured at surgery. Thirty-seven shoulders in 33 patients who had reoperation for a presumed failed rotator cuff repair were retrospectively evaluated. All patients had preoperative MRI, the results of which were read by a musculoskeletal radiologist to determine the presence of a tear and to estimate its size. All tears were measured intraoperatively in sagittal and coronal planes. Thirty-three shoulders had recurrent tears of the rotator cuff, and MRI correctly identified the presence of 30 of them. MRI correctly identified only 1 of the 4 patients without a recurrent tear of the cuff. The correlation coefficient for MRI accurately defining the size of cuff tears was 0.46. The sensitivity of MRI for the diagnosis of retear was 91%, and the specificity was 25%. MRI is accurate in diagnosing a recurrent full-thickness cuff tear in postsurgical shoulders. However, it is relatively inaccurate in correctly defining the size of the tear. MRI also has a tendency to overdiagnose cuff tears in postsurgical patients with continued pain and symptoms.