Motor imagery (MI) is widely used to study cognitive aspects of the neural control of action. Prior studies were mostly centred on hand and arm movements. Recently a few studies have used imagery tasks to explore the neurophysiology of human gait, but it remains unclear how to ascertain whether subjects actually perform imagery of gait as requested. Here we describe a new experimental protocol to quantify imagery of gait, by behaviourally distinguishing it from visual imagery (VI) processes and by showing its temporal correspondence with actual gait. Fourteen young healthy subjects performed two imagery tasks and an actual walking (AW) task. During both imagery tasks subjects were sitting on a chair and faced a computer screen that presented photographs of walking trajectories. During one task (MI), subjects had to imagine walking along the walking trajectory. During the other task (VI), subjects had to imagine seeing a disc moving along the walking trajectory. During the AW task, subjects had to physically walk along the same walking trajectory as presented on the photographs during the imagery tasks. We manipulated movement distance by changing the length of the walking trajectory, and movement difficulty by changing the width of the walking trajectory. Subjects reported onset and offset of both actual and imagined movements with a button press. The time between the two button presses was taken as the imagined or actual movement time (MT). MT increased with increasing path length and decreasing path width in all three tasks. Crucially, the effect of path width on MT was significantly stronger during MI and AW than during VI. The results demonstrate a high temporal correspondence between imagined and AW, suggesting that MI taps into similar cerebral resources as those used during actual gait. These results open the possibility of using this protocol for exploring neurophysiological correlates of gait control in humans.