To determine the diagnostic performance of magnetic resonance (MR) imaging in the evaluation of suspected rotator cuff tears, eight asymptomatic volunteers and 32 patients with rotator cuff tendonopathy who underwent surgery were examined with MR imaging. Twenty-four of these patients also underwent contrast arthrography. The ability of MR imaging to depict the size of cuff tears and the quality of torn tendon edges was also evaluated. The MR imaging and arthrographic studies were reviewed without knowledge of surgical results or of the other studies. A scoring system was developed and a score assigned to each patient's MR study. The sensitivity of MR imaging for all tears (partial and full thickness) was 0.91, and the specificity was 0.88; whereas the sensitivity and specificity of arthrography were each 0.71. The scoring system improved the sensitivity to 1.0 and the specificity to 0.92. Linear regression analysis showed excellent correlation between preoperative assessment of the size of rotator cuff tears and measurement at surgery (r = .95).