Objective: Subluxation of the shoulder after stroke can be measured according to the method described by Van Langenberghe and Hogan. Methods: To evaluate the reliability of this method, the shoulder radiographs of 25 patients were available for this study. Two independent raters each assessed these radiographs twice. Results: The intrarater reliability was good: percentage of agreement was 88 and 84%, weighted κ, 0.69 [95% confidence interval (CI), 0.38-1 0] and 0.78 (95% CI, 0.60-0.95) for raters 1 and 2, respectively. The interrater reliability was poor: percentage of agree ment was 36 and 28%, κ, 0.11 (95% CI, 0.0-0.31) and 0.09 (95% CI, 0.0-0.23) in sessions 1 and 2, respectively. Subsequently the original method was adjusted by com bining two categories (no subluxation and beginning subluxation) into one (“no clin ically important subluxation”). Conclusions: After this adjustment of the categories, the interrater reliability improved [percentage of agreement, 72%, and κ, 0.49 (95% CI, 0.18-0.80)], but did not reach acceptable values.