Purpose: The purpose of this study was to assess the efficacy of MR arthrography to detect partial thickness rotator cuff tears. Materials and Methods: One hundred and seventy seven patients with a high suspicion for rotator cuff disease were studied by performing MR-arthrography and subsequent arthroscopy. The ability of MR-arthrography to detect partial thickness tears was evaluated according to the location of the tears. We determined the correspondence between the measurements of the articular side partial tears on MR arthrography and those on the arthroscopic findings. Results: The arthroscopic diagnosis of partial thickness rotator cuff tears was divided into 3 groups according to their location. There were 63 cases on the articular side, 41 cases on the bursal side and 20 cases on both sides. The sensitivity of MR-arthrography was 82% for the articular side tears and 11% for the bursal tears. The specificity was 88% for the articular side tears and 100% for the bursal tears. MR-arthrographic measurement correctly predicted 72% in 28 repaired cases of 56 articular side partial thickness tears. Conclusions: MR-arthrography may be a reliable tool for diagnosing articular side partial thickness rotator cuff tears, but it has limitations for bursal side tears.