A prospective study was performed to determine the accuracy of magnetic resonance imaging (MRI) compared with operative findings in the evaluation of patients associated with rotator cuff tears. Fifty-four of 60 shoulders (58 patients) examined by MRI were confirmed as full-thickness tears and 6 as partial-thickness tears at the time of surgery. The oblique coronal, oblique sagittal, and axial planes of T2-weighted images with the 0.5 tesla MRI system were obtained preoperatively and compared with operative findings. MRI correctly identified 46 of 54 full-thickness rotator cuff tears and 5 of 6 partial-thickness tears. A comparison of MRI and operative findings in full-thickness cuff tears showed a sensitivity of 85%, a specificity of 83%, and a positive prospective value (PPV) of 99%. A comparison of partial-thickness tears showed a sensitivity of 83%, a specificity of 85%, and PPV of 39%. Linear regression analysis showed an excellent correlation between the MRI assessment and measurement at the time of surgery (r = 0.90, P < 0.01). MRI was useful in evaluating large and medium-sized rotator cuff tears, but less useful in distinguishing small full-thickness tears from partial-thickness tears.