[Purpose] To estimate the minimal clinically important difference for the Fugl-Meyer assessment of the upper extremity by using anchor-based methods in stroke patients with moderate to severe hemiparesis. [Participants and Methods] Fourteen patients who were hospitalized in a convalescent phase rehabilitation ward were included in this study. Fugl-Meyer assessment of the upper extremity was used to assess the impairment prior to intervention and at follow-up (six weeks later). Participants were asked to evaluate the degree of improvement of paresis of the upper extremity using the global rating of change scale at follow-up. The mean change in Fugl-Meyer assessment scores in the group of patients who answered “a little better, meaningful in daily life” in the global rating of change scale was considered as the minimal clinically important difference. [Results] The mean post-onset period of participants for analysis was 49.4 days. The minimal clinically important difference of the Fugl-Meyer assessment scores were 12.4 (upper extremity), 5.6 (upper arm), and 4.9 (wrist/hand). [Conclusion] A score of 12.4 in the Fugl-Meyer assessment of the upper extremity is likely to be perceived as meaningful in stroke patients with moderate to severe hemiparesis.