The aim of this paper is to provide evidence, both published and new, to support the notion that human infants are particularly good subjects for the study of the pattern generator for walking. We and others have shown that stepping can be initiated by sensory input from the legs or by general heightened excitability of the infant. New results are presented here to suggest that weight support through the feet and rapid extension of the legs are important proprioceptive inputs to initiate stepping. Our previous work has shown that infants can step at many different speeds when supported on a treadmill. The step cycle duration shortens as the speed increases, with the changes coming largely from the stance phase, just as in most other terrestrial animals. Moreover, we have shown that infants will step in all directions. Regardless of the direction of stepping, the step cycle changes in the same way with walking speed, suggesting the circuitry that controls different directions of walking share common elements. We have also shown that infant stepping is highly organized. Sensory inputs, whether proprioceptive or touch, are gated in a functional way so that only important sensory inputs generate a response. For example, touch to the lateral surface of the foot elicits a response only in sideways walking, and only in the leading limb. New data is presented here to show that the pattern generators from each limb can operate somewhat independently. On a split-belt treadmill with the 2 belts running at different speeds or in different directions, the legs showed considerable independence in behaviour. Yet, the pattern generators on each side interact to ensure that swing phase does not occur at the same time. These studies have provided insight into the organization of the pattern generator for walking in humans. It will be interesting in the future to study how maturation of the descending tracts changes walking behaviour to allow independent bipedal walking.