Tears of the rotator cuff are accompanied by muscle atrophy, which has direct implications on patient outcome after rotator cuff surgery. However, no reliable method exists to determine muscle volumes in situ. The current authors evaluated the reliability of magnetic resonance imaging in assessing rotator cuff muscle volumes. Muscle volumes were determined in 10 shoulders from cadavers using two methods: (1). oblique sagittal shoulder scans were obtained, the contours of the supraspinatus, infraspinatus and teres minor, and subscapularis muscles were traced, and muscle volumes were calculated using image analysis software; and (2). rotator cuff muscles were dissected and muscle volumes were measured by water displacement. The average magnetic resonance imaging volume of the supraspinatus, infraspinatus and teres minor, and subscapularis muscles were 36 +/- 12, 96 +/- 41, and 99 +/- 33 mL, respectively. There was a significant correlation between magnetic resonance imaging and water displacement measurements. The intraobserver and interobserver variabilities were less than 4%. The results indicate the magnetic resonance imaging is a reliable method to determine rotator cuff muscle volumes with good intraobserver and interobserver variability. This method may prove valuable preoperative assessment of rotator cuff muscles and prediction of outcome after rotator cuff repair.