The purpose of this study was to examine the effect of three cognitive tasks on gait at a preferred walking speed, and at a fast speed, using dual-task methodology. A total of 29 stroke patients participated in the study. All 29 subjects performed 2 motor tasks (10-meter walk task and timed up and go task each at a preferred and a fast speed) and three cognitive tasks [Stroop, word list generation (WLG), serial subtraction (SS)] under dual-task conditions [cognitive-motor interference (CMI)] in a randomized order. Gait speeds were measured in six different conditions. A repeated-measure analysis of variance was employed to compare the results of the Stroop training, WLG, and SS tasks during preferred and fast walking. A Bonferroni adjustment use for post hoc analysis. The level of statistical significance was set at α=.05. A CMI effect occurred for performance of a 10-meter walking task at two different speed and a cognitive task (p＜.05). Stroop had a significantly greater effect than SS and WLG (p＜.05). The timed up and go task was affected when performed with fast walking speed during Stroop cognitive task (p＜.05), but was not affected if performed with preferred walking speed during a cognitive task (p＞;.05). This study showed that CMI of Stroop can be used as a rehabilitation program for stroke patients.