OBJECTIVE Partial thickness tears of the undersurface (articular surface) of the rotator cuff (RTC) have been recognized increasingly in recent years as a source of treatable shoulder pain in the athletic population. This study evaluated the efficacy of MR arthrography (MR-ARTH) in diagnosing these tears. METHODS The study design was a retrospective review of medical records of patients who had presented with refractory shoulder complaints and subsequently undergone MR arthrography with multiple signal MRI sequences followed by shoulder arthroscopy. Of particular interest were patients who had oblique T1 fat suppression (COT1FS), coronal oblique T2 (COT2), and coronal oblique T2 fat suppression (COT2FS) images taken. Seventy-six subjects met the study criteria. Investigators examined the MR-ARTH images from these patients' charts while blinded as to arthroscopic results, clinical signs and symptoms. RESULTS Based on COT1FS images, investigators identified nine subjects as having had full thickness tears, 28 as having had partial thickness tears of the undersurface of the rotator cuff (PRTC), and 39 as having had intact RTC. These results were compared to actual findings at arthroscopy: nine full thickness tears, 26 of 28 with PRTC and 34 of 39 intact. The sensitivity of MR-ARTH was 84%, with a positive predictive value of 93%. The overall accuracy was 91% (69/76). The specificity was 96%. That is, if a PRTC was not seen on the MR-ARTH images, it was very unlikely to exist. COT2 and COT2FS sequences failed to increase sensitivity and overall efficacy of MRI. CONCLUSIONS PRTC can be diagnosed accurately by MR-ARTH with gadopentatate contrast. A COT1FS sequence is recommended for evaluation when tears are suspected.