Trial-by-trial adaptation of movements during mental practice under force field

Abstract

Human nervous system tries to minimize the effect of any external perturbing force by bringing modifications in the internal model. These modifications affect the subsequent motor commands generated by the nervous system. Adaptive compensation along with the appropriate modifications of internal model helps in reducing human movement errors. In the current study, we studied how motor imagery influences trial-to-trial learning in a robot-based adaptation task. Two groups of subjects performed reaching movements with or without motor imagery in a velocity-dependent force field. The results show that reaching movements performed with motor imagery have relatively a more focused generalization pattern and a higher learning rate in training direction.

Figures

Figure 1

Experimental setup and trial protocol.

Figure 2

Time series of actual movement errors and the corresponding model predictions are shown. The changing trend of model output conforms with the actual movement errors. For the sake of clarity, the values are plotted after averaging every 5 samples.

Figure 3

Generalization patterns in 4 directions are shown. Free parameters are reduced to 5 by averaging the parameters existing at same directional difference values. Shaded regions show the deviation in parameter values across all subjects.

Figure 4

Figure shows averaged ERD patterns in both MI (in black) and No-MI (in dotted gray) groups. The ERD was calculated for each direction and was averaged within the sets. MI group has shown more prominent ERDs during imagery time.

Figure 5

Motor learning rate in all directions is shown. The solid bars represent mean values while corresponding standard deviation values are represented by the limits put on bars.

Figure 6

Motor learning rate that is, transferred in adjacent direction is shown. The impact of motor learning on immediate next direction (with 45° difference) is averaged across all subjects.

Figure 7

Averaged generalization pattern in all possible directional differences for MI and No-MI groups. Solid bars show the mean values, while the deviation is represented by the limits put on bars.

Figure 8

Polar plots for the eigenvalues of all odd subjects are shown. These plots signify that the model built for each subject is stable with the eigenvalues lying inside the unit circle.