Transition between discrete and rhythmic dynamics in human walking

BACKGROUND AND AIM: Human movement is structured through a set of dynamic primitives, neuromechanical attractors that simplify control by the central nervous system. At the end-effector level, these primitives give rise to two main types of movements: discrete (point attractors or submovements) and rhythmic (limit cycles or oscillations). Evidence from upper-limb tasks (as wrist flexion-extension) suggests that these movement types rely on distinct neural substrates and exhibit asymmetrical transfer of motor learning. This framework helps explain why rhythmic arm movements are often less impaired than discrete ones following stroke. Previous studies have shown that under temporal constraints, fast discrete movements tend to merge into rhythmic oscillations, while slow rhythmic movements disrupt into discrete submovements. While this phenomenon has been extensively studied in the upper limb, the transition between discrete and rhythmic dynamics in lower limb movements, particularly during locomotion, remains poorly understood. Our study investigates whether such a transition occurs in locomotion, whether the threshold for this transition is consistent across individuals, and whether the process is affected by hysteresis. We also explore whether the coordinated muscle activations (muscle synergies), understood as functional groupings of muscles, reflect the underlying dynamic primitives that govern this transition. METHODS: 20 healthy adults walked on a treadmill across systematically varied speeds (0.5–5 km/h) under three conditions: incremental, decremental, and randomized progression. Kinematic data were recorded using a 10-camera Vicon motion capture system (100 Hz), and EMG signals were collected bilaterally from 12 lower-limb muscles. Movement classification into discrete or rhythmic dynamics was based on established smoothness metrics, including spectral arc length (SAL) and mean-squared jerk ratio (MSJR). RESULTS: Both MSJR and SAL showed significant changes around 2 km/h compared to comfortable walking speeds (4–5 km/h), suggesting the existence of a speed threshold marking the transition from discrete to rhythmic movement dynamics. The direction of speed progression slightly modulated the onset of the transition, indicating a mild hysteresis effect. These findings are consistent with hysteresis observed in walk–run transitions and suggest a nonlinear mechanism. Preliminary analysis of EMG-derived muscle synergies revealed distinct activation patterns corresponding to discrete and rhythmic states. CONCLUSIONS: This study extends the dynamic primitives framework to human locomotion, providing novel insights into how the central nervous system transitions between control strategies at different walking speed. A deeper understanding of these dynamics may have implications for motor learning, exercise strategies, and neurorehabilitation. Future research should investigate these mechanisms in clinical populations to evaluate their diagnostic and therapeutic potential.