The purpose of this study was to investigate the validity of the supraspinatus test as a screening test for detecting torn rotator cuff and to determine what its valuable positive signs were. Both the empty-can test and full-can test were performed on 200 shoulders diagnosed by magnetic resonance imaging (MRI)-and in some cases, surgical findings-to have full-thickness or partial-thickness torn rotator cuff s, or no tear in the rotator cuff . During the maneuver, the presence of pain or weakness or both pain and weakness were recorded as positive signs, and the distribution of these signs were analyzed according to the degree of tear. The predictive values were calculated in 2 ways by considering (1) only full-thickness tears as tears and (2) both full- and partial-thickness tears as tears. The 2 tests and the 2 ways of considering partial-thickness tears were compared. Pain and weakness were severity-dependent, and the empty-can test had a higher incidence of pain. The sensitivities of the 2 supraspinatus tests in all positive signs were higher when including partial-thickness tears in the tear group ; however, their specificities were higher when excluding partial-thickness tears. Both pain and weakness were interpretive for the supraspinatus test, and both tests were sensitive to full- and partial- thickness tears and specific for full-thickness tears.