[Purpose] The purpose of this retrospective study was to determine the minimal clinically important difference for comfortable gait speed for patients with stroke. [Subjects] Data were analyzed from 35 patients undergoing inpatient rehabilitation. [Methods] Two characteristics of gait were measured, assistance required and comfortable gait speed. Patients were grouped as either experiencing or not experiencing a decrease of 2 or more levels of assistance required over the course of rehabilitation. Receiver operating characteristic curve analysis was used to identify the change in gait speed that best differentiated between patients who did and did not experience the requisite decrease in assistance required for gait. [Results] Twenty-one patients decreased 2 or more levels of assistance whereas 14 did not. Walking speed increased significantly more in the group who experienced a decrease in assistance of at least 2 levels. The receiver operating characteristic curve analysis showed a change in walking speed of 0.13 m/s best distinguished between patients who did versus did not experience a reduction in assistance required. [Conclusion] An improvement in gait speed of 0.13 m/s or more is clinically important in patients with stroke.