Motor learning and transfer: from feedback to feedforward control

Abstract

Previous work has demonstrated that when learning a new motor task, the nervous system modifies feedforward (ie. voluntary) motor commands and that such learning transfers to fast feedback (ie. reflex) responses evoked by mechanical perturbations. Here we show the inverse, that learning new feedback responses transfers to feedforward motor commands. Sixty human participants (34 females) used a robotic exoskeleton and either 1) received short duration mechanical perturbations (20 ms) that created pure elbow rotation or 2) generated self-initiated pure elbow rotations . They did so with the shoulder joint free to rotate (normal arm dynamics) or locked (altered arm dynamics) by the robotic manipulandum. With the sho ulder unlocked, the perturbation evoked clear shoulder muscle activity in the long-latency stretch reflex epoch (50-100ms post-perturbation), as required for countering the imposed joint torques, but little muscle activity thereafter in the so-called voluntary response. After locking the shoulder joint, which alters the required joint torques to counter pure elbow rotation, we found a reliable reduction in the long-latency stretch reflex over many trials. This reduction transferred to feedforward control as we observed 1) a reduction in shoulder muscle activity during self-initiated pure elbow rotation trials and 2) kinematic errors (ie. aftereffects) in the direction predicted when failing to compensate for normal arm dynamics, even though participants never practiced self-initiated movements with the shoulder locked. Taken together, our work shows that transfer between feedforward and feedback control is bidirectional, furthering the notion that these processes share common neural circuits that underlie motor learning and transfer .

Vision Affects Gait Speed but not Patterns of Muscle Activation During Inclined Walking—A Virtual Reality Study

“…While we found that visual cues affect locomotion for a brief period after perturbations occur (i.e., virtual inclination transitions), the roles of proprioception and vestibular cues in controlling locomotion are continuous and likely to be associated with gait speed. Overall, locomotion relies on feedback (e.g., from joint mechanoreceptors that detect gravitational changes) and feedforward (e.g., anticipating center-of-mass change from prior experience) (Riemann and Lephart, 2002;Seidler et al, 2004;Maeda et al, 2019) mechanisms. For example, visual inputs are used in the process of planning locomotion to create a model of the environment in which the walking would occur, while proprioception is essential during the execution of the movement to update the feedforward commands derived from the visual inputs (Bard et al, 1995;Sainburg et al, 1995).…”

Vision affects gait speed but not patterns of muscle activation during inclined walking – a virtual reality study

“…While we found that visual cues affect locomotion for a brief period after perturbations occur (i.e., virtual inclination transitions), the role of proprioception and vestibular cues in controlling locomotion are continuous and likely to be dependent on gait speed. Overall, locomotion relies on feedback (e.g., from joint mechanoreceptors that detect gravitational changes) and feedforward (e.g., anticipating center-of-mass change from prior experience) (Riemann and Lephart, 2002;Seidler et al, 2004;Maeda et al, 2019) mechanisms. For example, visual inputs are used in the process of planning locomotion to create a model of the environment in which the walking would occur, while proprioception is essential during the execution of the movement to update the feedforwards commands derived from the visual inputs (Bard et al, 1995;Sainburg et al, 1995).…”