Results of three clinical tests for detecting shoulder impingement syndrome (Neer's, Hawkins', and Yocum's tests) and four tests for determining the location of the rotator cuff lesion (Jobe's test [supraspinatus], Patte's test [infraspinatus], lift-off test [subscapularis], and palm-up test [long head of the biceps brachii]) were compared to intraoperatively observed anatomic lesions in 55 consecutive patients who had surgery for Neer's syndrome. For Jobe's and Patte's tests, both pain (denoting tendinitis) and functional impairment (denoting tendon rupture) were evaluated. All clinical tests were done by the same examiner and all surgical procedures (acromioplasty with or without rotator cuff repair) by the same surgeon. The location and extent of the lesions (size of the tear in the 34 patients with rotator cuff defects) were determined intraoperatively. The sensitivity, specificity, and positive and negative predictive values of each test were calculated. Sensitivity was satisfactory but specificity was poor, in particular for determining the location and type of rotator cuff lesions. The severity of functional impairment during Jobe's and Patte's maneuvers was not correlated with the size of the tear.