Learning New Feedforward Motor Commands Based on Feedback Responses

Abstract

Highlights d Long-latency stretch reflex responses can learn altered arm dynamics d This learning occurs with minimal engagement of voluntary motor responses d What reflex responses learn transfers to voluntary motor commands

Explicit feedback and instruction do not change shoulder muscle activity reduction after shoulder fixation

“…Although no restrictions were placed on movement trajectories, participants did generate their movements with an arc-like trajectory, as a result of moving predominantly the elbow joint Figure 2 A,C,E (Maeda et al 2017(Maeda et al , 2018(Maeda et al , 2020a(Maeda et al , 2020b . We found substantial shoulder agonist muscle activation prior to movement onset in all experiments as required to compensate for the torques that arise at the shoulder when the forearm rotates about the elbow joint Figure 2 B,D,F (Gribble and Ostry 1999;Maeda et al 2017Maeda et al , 2018Maeda et al , 2020aMaeda et al , 2020b . We then mechanically locked the shoulder joint of the robotic manipulandum, which cancelled the torques that arise at the shoulder with forearm rotation and removes the need to activate the shoulder muscles.…”

“…), explicit strategies in these tasks quickly improved learning speeds (for review, see Krakauer et al 2019) . We previously reported that people learn to reduce shoulder activity after shoulder fixation, which unlike many other motor learning paradigms, unfolded slowly over hundreds of trials and it was still incomplete by the end of the testing session (Maeda et al 2018(Maeda et al , 2020a(Maeda et al , 2020b . In addition, this learning took place without people knowing about this manipulation and in the absence of kinematic errors, as locking the shoulder joint restricts kinematics into an arc about the elbow joint.…”

“…To compare the changes in amplitude of muscle activity over time and across different phases of the protocols, we calculated the mean amplitude of phasic muscle activity across a fixed time-window, -100 ms to +100 ms, relative to movement onset. These windows were chosen to capture the agonist burst of EMG activity in each of the experiments (See Debicki and Gribble 2005;Maeda et al 2017Maeda et al , 2018Maeda et al , 2020aMaeda et al , 2020b .…”

“…First, although we found no reliable differences in the degree or rate of shoulder muscle activity adaptation across experiments, it is hard to prove the absence of an effect due to many intrinsic experimental factors, including sample size and number of repetitions. Note, however, that we tried to keep these experimental factors as close as possible to our previous studies (Maeda et al 2018(Maeda et al , 2020a(Maeda et al , 2020b , and in particular to a study in which we did observe reliable differences across experiments (Maeda et al 2020b) . Second, the metric used as feedback to the participant (muscle activity amplitude) is a relatively noisy measurement.…”

Shared internal models for feedforward and feedback control of arm dynamics in non-human primates

“…Many previous studies have demonstrated that the nervous system develops an internal representation of such arm dynamics that can be used to generate predictive (i.e., feedforward) muscle activity during self-initiated reaching (Gribble & Ostry, 1999;Debicki & Gribble, 2005;Gritsenko et al , 2011;Maeda et al , 2017Maeda et al , , 2018Maeda, Gribble, et al , 2020;Maeda, Zdybal, et al , 2020a, 2020b and also reflex (i.e., feedback) responses to mechanical perturbations, in a way that accounts for the arm dynamics (Lacquaniti & Soechting, 1984, 1986a, 1986bSoechting & Lacquaniti, 1988;Kurtzer, Pruszynski, et al , 2006;Kurtzer et al , 2008Kurtzer et al , , 2009Kurtzer et al , , 2014Kurtzer et al , , 2016Pruszynski et al , 2011;Crevecoeur et al , 2012;Maeda et al , 2017Maeda et al , , 2018Kurtzer, 2019;Maeda, Gribble, et al , 2020) . For instance, when generating single-joint elbow movements and when responding to mechanical perturbations that create pure elbow motion, the nervous system generates predictive shoulder muscle activity and robust shoulder feedback responses to counter the underlying torques that arise at the shoulder joint because of forearm rotation about the elbow joint (Gribble & Ostry, 1999;Kurtzer et al , 2008;Maeda et al , 2017Maeda et al , , 2018Maeda, Gribble, et al , 2020) . This similarity between feedforward and feedback control reflects, in part, the shared neural circuits between feedforward and feedback control at the level of the spinal cord, brainstem and cerebral cortex (Pruszynski & Scott, 2012;Kurtzer, 2014;Scott, 2016) .…”