Combining advances from gait analysis and motor learning fields, this study aims to examine invariant characteristics and practice-related changes in the control of walking gait when learning a biomechanically constrained pattern, racewalking (RW). RW's regulation imposes a straightened knee at the stance phase which differentiates it qualitatively from normal walking. Using 3D motion analysis, we computed key kinematic variables from a whole-body model. Principal component analysis was then used as a tool to evaluate the evolution of normal walking synergies (S0) immediately at the first practice session (S1) and further with practice (S1-S4). Before the start of practice, normal walking was characterized by two predominant control dimensions explaining an upper-extremities/antero-posterior component (PC1) and a lower-extremities/vertical component (PC2). Compared to normal walking, the immediate increase at S1 in the number of PCs needed to explain a significant portion of movement variance could be suggestive of a recruitment of a task-specific component. With practice, the significant decrease in the variance accounted for by PC1 and in the correlations between many variables could indicate a destabilization of spontaneous tendencies to facilitate the adoption of more task-specific coordinative pattern. PC2 seemed to be reinforced with practice where a significant increase in its explained variance was observed. In sum, this study shows that common features in the gait control are preserved with practice, and the movement reorganization, however, seems rather defined by shifts in the relative contribution of some variables within each PC.
Keywords: Control dimension; Gait analysis; Motor learning; Movement reorganization; PCA; Racewalking.