The present study explores variations in the degree of automaticity and predictability of cyclical arm and leg movements. Twenty healthy adults were asked to walk on a treadmill at a lower-than-preferred speed, their preferred speed, and at a higher-than-preferred speed. In a separate, repetitive punching task, the three walking frequencies were used to cue the target pace of the cyclical arm movements. Movements of the arms, legs, and trunk were digitized with inertial sensors. Whereas absolute slope values (|β|) of the linear fit to the power spectrum of the digitized movements (p < .001, η2 = .676) were systematically smaller in treadmill walking than in repetitive punching, sample entropy measures (p < .001, η2 = .570) were larger reflecting the former task being more automated but also less predictable than the latter task. In both tasks, increased speeds enhanced automatized control (p < .001, η2 = .475) but reduced movement predictability (p = .008, η2 = .225). The latter findings are potentially relevant when evaluating effects of task demand changes in clinical contexts.
Keywords: cyclical movements; movement predictability; sample entropy; spectral slope values.