The purpose of this study was to determine the intrarater and interrater reliability of a manual anterior humeral head translation test. Fifteen subjects were positioned lying in a supine position with their identity shielded from examiners. A standard manual anterior humeral head translation test was performed and repeated with the glenohumeral joint in 90 degrees of elevation in the scapular plane, with use of the grading method proposed by Altchek and Dines in 1993. Reliability was assessed with the coefficient of agreement and kappa statistic. Intrarater reliability was 81.4% comparing grade I and II translation. This decreased to 54% when examiners distinguished between grades I, I+, II, and II+. Interrater reliability for the same comparisons was 70.4%, decreasing to 37.3%. On the basis of these data, the technique of manually assessing anterior humeral head translation studied has poor overall interrater reliability and only fair intrarater reliability. The test-retest accuracy of humeral head translation is enhanced when examiners only determine the relationship of the humeral head relative to the glenoid rim.