Thirty-eight patients with suspected rotator cuff tears were examined at 1.5 T by using a loop-gap resonator surface coil. The MR findings were compared prospectively in a blinded fashion with the results from double-contrast arthrography in all 38 patients, high-resolution sonography in 23 patients, and surgery in 16 patients. In the total group of 38 patients, MR imaging detected 22 of 22 tears and 14 of 16 intact cuffs as determined by arthrography. In the 16 surgically proved cases, MR and arthrography showed identical results, with 92% sensitivity in the diagnosis of 12 tears and 100% specificity in the diagnosis of four intact cuffs. In a subgroup of 23 patients, sonography detected nine of 15 tears and seven of eight intact cuffs as determined by comparison with arthrography. In 10 surgically proved cases, sonography was 63% sensitive in the diagnosis of eight rotator cuff tears and 50% specific in the diagnosis of two intact cuffs. For the diagnosis of rotator cuff tears, MR imaging is comparable to arthrography in both sensitivity and specificity. In this study, sonography was not as accurate in the diagnosis of rotator cuff tears as were the other two techniques. These results suggest that MR imaging should be considered the noninvasive test of choice for patients with suspected rotator cuff disease.