This thesis includes four projects that examine motor learning and assess novel approaches for the training and evaluation of walking. In Project 1 we study motor adaptation in children aged 8-36 months using a split-belt treadmill. Splitbelt walking, in which one leg moves faster than the other, causes asymmetries in interlimb coordination. Adaptation is manifest as decreases in the asymmetries with practice, and an aftereffect (i.e., asymmetry in the opposite direction) upon the return to normal treadmill walking. Most children showed adaptation in double support time (temporal measure of interlimb coordination), but fewer showed adaptation in the spatial measures of step length and centre of oscillation. Hence, the mechanisms controlling adaptation of temporal coordination may mature before those controlling spatial coordination. In Project 2 we studied interlimb coordination and long-term training in infants aged 3-10 months. We found that most infants expressed the same type of coordination (i.e., alternate or synchronous) when kicking, a non-weight bearing movement, and when weight bearing on the treadmill. We also showed that daily practice of the non-preferred coordination in weight bearing for 1 month changed the preferred coordination for non-weight bearing movements. These findings suggest there is partial sharing of neural substrates for interlimb coordination of different leg movements in infants. In Project 3 we compared 2 methods of walking training – body weight-supported treadmill training (BWSTT) and an over-ground method involving challenging practice of relevant walking skills (called skill training). Four individuals with chronic incomplete spinal cord injury (ISCI) completed 3 months of BWSTT followed by 3 months of skill training, or vice versa. We found skill training to be as effective as BWSTT at improving walking skill, speed, endurance and confidence. In Project 4 we developed a new measure of walking for ISCI, called the Spinal Cord Injury Functional Ambulation Profile (SCI-FAP). It involves the timed performance of 7 common walking tasks. The SCI-FAP has high inter-rater and test-retest reliability, and discriminative and convergent validity. Collectively the findings of this thesis will contribute to the optimization of walking training programs for adults and children with damage to the central nervous system.