Correlation dimensions and average maximum finite-time Lyapunov (MFTL) exponents were used to examine the kinematics of continuous level walking in human subjects. Comparisons were made between overground (OG) and motorized treadmill (TM) walking in young healthy subjects and between diabetic neuropathic (NP) patients and healthy controls (CO) during overground walking. Surrogate data were used to examine the stochastic nature of the stride-to-stride variability seen in these walking patterns. There were three primary results. First, the motorized treadmill significantly constrained the neuromuscular control and normal movement kinematics of walking. Second, NP patients demonstrated greater correlation dimensions in their movement patterns and slowed their walking speeds to maintain maximum upper body stability during unrestricted walking over level ground. Third, stride-to-stride fluctuations in walking kinematics could be clearly distinguished from correlated Gaussian noise and demonstrated changes in their structure that were related to the loss of peripheral sensation.