To clarify their usefulness in screening massive rotator cuff tear, routine radiographs of the shoulder were compared in three groups: (1) shoulders with a massive rotator cuff tear (MRCT), (2) shoulders with a small full-thickness rotator cuff tear (SRCT), and (3) normal control shoulders. Two different statistical methods were employed to assess the gradation (order) of correlation in 83 shoulders (22 were MRCTs, 31 with SRCTs, and 30 controls). Superior migration of the humerus and deformity of the greater tuberosity showed by far the more significant contribution to MRCT than did other abnormalities. When the shoulders were screened for these two abnormalities, the sensitivity was 78% (2127) and specificity was 98% (5556).