The nervous systems in various species do not change rapidly in evolution. Yet the idea that human gait would be controlled in much the same way as it is in other mammals, or indeed even in invertebrates (Clarac et al ., 2000) has met a lot of resistance (Capaday, 2001). Nevertheless, as evidenced by Dietz et al. in this issue of Brain (Dietz et al ., 2002), we should take this evolutionary idea seriously. From the work on cats it is known that the rhythmic muscle activities during gait are generated by specialized neural circuits located in the spinal cord (the so‐called central pattern generator, CPGs for locomotion). Each limb is controlled by such a CPG, either located in the rostral or in the caudal spinal cord, subserving forelimb and hind limbs, respectively (Duysens and Van de Crommert, 1998). These CPGs are coordinated by neurones, which interconnect both sides or which transmit information between the cervical and lumbar spine (Dietz, 2002 a , b ). Each CPG is guided by sensory feedback to facilitate the transitions between the phases of the step cycle. Two important sources of feedback to the CPGs have been identified. First, signals about the loading and unloading of the limb are important. During the stance phase …