Background: The Fugl-Meyer Assessment of the Upper Extremity (FMA-UE) has been used in many clinical studies and in stroke rehabilitation. In studies evaluating psychometric properties, confirmatory factor analysis (CFA) indicated that the FMA-UE is a multidimensional tool. Item Response Theory One-Parameter Logistic (IRT1PL) supports that item-difficulty hierarchy can be used as a treatment index of upper extremity function for stroke recovery. However, studies on the psychometric properties of the FMA-UE in Asian populations are lacking.Objectives: To investigate the dimensionality and item-difficulty hierarchy of the FMA-UE for stroke rehabilitation in Japanese patients. Methods: This was a cross-sectional study. The participants comprised 268 individuals admitted for de novo stroke (median age, 70.0 years; median days since stroke onset, 78.5) in 22 hospitals in Japan. The dimensionality of the FMA-UE was evaluated using CFA of selected items. The item-difficulty hierarchy of the FMA-UE using the appropriately selected model was demonstrated using IRT1PL analysis after confirming dimensionality.Results:Two reflex items were removed by utilizing the floor and ceiling effects. The 31- and 30-item FMA-UE exhibited a good model fit of the unidimensionality in the CFA. The 30-item FMA-UE was found to be a good model by model comparison (the 31-item vs. the 30-item). The item-difficulty hierarchy of the 30-item FMA-UE was found not to be consistent with the expected item order.Conclusions:This study provides evidence that the FMA-UE has multidimensionality and the 30-item FMA-UE is a valid instrument for measuring upper-extremity impairment after stroke.