BACKGROUND: The involvement of attention in the control of the walking-related rhythmic stepping mechanism remains unclear in contrast to control of balance phases in human walking. Concurrently to the dual-task related decrease in stride velocity, an increased stride-to-stride variability has been observed and interpreted as interference caused by competing demands for attention resources. It has been also shown that stride time variability depends on stride velocity, whereas the relationship between stride length variability and stride velocity remains controversial. It remains unknown whether dual-task related changes in stride time and stride length variability result from an effect of stride velocity and/or of the attention-demanding task. OBJECTIVE: To determine the respective roles of stride velocity and attention-demanding task on dual-task related changes in stride time and stride length variability. METHODS: Mean stride velocity, mean, standard deviation and coefficient of variation of stride time and stride length were collected using the ambulatory device Physilog in 40 healthy young adults while walking at self-selected speed and while walking with enumerating animal names. RESULTS: Enumerating animal names while walking provoked significant changes in gait but not in cognitive task performance. Balanced analyses of covariance with a repeated measures design showed that stride time variability increased significantly under dual-task condition independently of dual-task related decrease in stride velocity, while stride length variability did not change under dual-task. CONCLUSION: These results show that the dual-task related increase in stride time variability depends on the attention-demanding task rather than stride velocity, suggesting that healthy young adults devote a part of their attention to the control of the rhythmic stepping mechanism at normal self-selected walking speed.