Two hundred eighty-one patients underwent magnetic resonance (MR) imaging of the shoulder over a 2-year period. Eighty-eight patients underwent arthroscopic surgery, and their surgical results were correlated with the findings at MR imaging. MR imaging enabled accurate prediction of anterior labral tears, with a sensitivity of 95%, a specificity of 86%, and an accuracy of 92%. MR imaging was less effective in the prediction of tears of the superior labrum, with a sensitivity of 75%, a specificity of 99%, and an accuracy of 95%. These two categories accounted for the majority of the surgically correctable disease. MR imaging was found to be unreliable in the prediction of posterior (sensitivity, 7.7%) or inferior (sensitivity, 40%) labral tears, isolated cases of which occurred in only two (2%) of those undergoing surgery. MR imaging proved to be a highly accurate, noninvasive technique for the clinical evaluation of shoulder instability.