Treadmill training may be used to improve gait rhythmicity in people with Parkinson's disease. Treadmills, however, alter dynamical stride time fluctuations in healthy adults in a manner that mimics pathologic states, indicating the stride-to-stride fluctuations that characterize healthy gait are constrained. It is unclear if treadmills similarly alter dynamic gait properties in Parkinson's disease. Do stride time fractal dynamics in individuals with Parkinson's disease differ between treadmill and overground walking? Fifteen participants with Parkinson's disease and 15 healthy age-similar adults walked for 6 min in a conventional overground condition and on a treadmill while wearing inertial measurement units. Gait speed, stride times and stride time variability were measured. Fractal exponents (α) were computed with adaptive fractal analysis. Inferential statistics were analyzed with mixed model analyses of variance and post hoc simple effects tests. Mean gait speeds decreased and stride times increased on the treadmill but did not differ between the Parkinson's and control groups. Stride time variability was greater in the Parkinson's than control group in both conditions. Most relevant to our research question, stride time fractal exponents were greater on the treadmill (mean α = .910) than overground (mean α = .797) in individuals with Parkinson's disease, but not in healthy controls. The fractal scaling exponent α emanating from stride time fluctuations during treadmill walking increased toward a 1/f signal of α = 1.0 that has been interpreted as an optimal structural variability for gait. The clinical implication is that treadmill training may promote more efficient walking dynamics in people with Parkinson's disease than conventional overground training.