When prelocomotor infants are supported on a motorized treadmill, they perform well-coordinated, alternating stepping movements that are kinematically similar to upright bipedal locomotion. This behavior appeared to be a component of independent walking that could not be recognized without the facilitating context of the treadmill. To understand the ontogenetic origins of treadmill stepping and its relation to later locomotion, we conducted a longitudinal study using an experimental strategy explicitly derived from dynamic systems theory. Dynamic systems theory postulates that new forms in behavior emerge from the cooperative interactions of multiple components within a task context. This approach focuses on the transitions, often nonlinear, where one preferred mode of behavior is replaced by a new form. Specific predictions about these transitions help uncover the processes by which development proceeds. Chapters II, III, and IV introduce dynamic principles of pattern formation and their application to development. In our application of these principles, we tested nine normal infants twice each month beginning from month 1 in a task where the treadmill speed was gradually scaled up and in an additional condition where each leg was driven by the treadmill at a different speed. Kinematic variables were derived from computerized movement analysis equipment and videotaped records. We also collected a number of anthropometric measurements, Bayley motor scores, and a behavioral mood scale for each month. Several infants stepped on the treadmill in their first month, but in all infants performance showed a rapidly rising slope from month 3 to month 6. Infants also showed corresponding improvement in adjustments to speed and relative coordination between the legs. In dynamic terminology, we found evidence that alternating stepping on the treadmill became an increasingly stable attractor during the middle months of the first year. Dynamic predictions that transitions would be characterized by increased variability and sensitivity to perturbation were borne out. Identifying the transitions enabled us to suggest a control parameter or variable moving the system into the stable response to the treadmill. This appeared to be the waning of flexor dominance in the legs during posture and movement that allowed the leg to be stretched back on the treadmill and so elicited the bilaterally alternating response. Further studies are needed to test this hypothesis. This dynamic analysis confirmed earlier suggestions that skill in general, and locomotion in particular, develops from the confluence of many participating elements and showed how emergent forms may result from changes in nonspecific components. A dynamic approach may be useful for understanding ontogenetic processes in other domains as well.