Involvement of the autonomic nervous system in motor adaptation: acceleration or error reduction?

Abstract

In the last few decades motor adaptation was extensively studied observing the invariant features of reaching movements. In a parallel neurobehavioral line of research emotional learning was studied under the umbrella of the ‘two-factor theory of learning’. In this study we explore the relation between motor learning and the autonomic response (heart rate, HR) of subjects performing point to point reaching movements holding a computer mouse. We consider two alternative outcomes: one is that autonomic response correlates with the learning rate and the second is that the autonomic response correlates with the residual error at the steady state. Eighteen subjects performed reaching movements under perturbed visual feedback demonstrating learning and after effects of learning. The hand movement as well as an Electrocardiogram signal were recorded throughout the training and carefully analyzed offline to extract the trial by trial error as well as the heart period. The results show clear correlation between the change in HR and the residual error but no correlation between the change in HR and the learning rate supporting the second alternative that the sensitivity to errors but not the learning rate correlates with the autonomic response. A control group of another seven subjects underwent the same experiment without the perturbed visual feedback. This control group showed no change in the HR. Further studies are required to validate this hypothesis and unravel the mechanism by which the autonomic response correlates with the residual motor error.

Appendix

Linearization of cursor versus mouse velocity

The position of the cursor on the monitor served to define the position of the mouse/hand on the board. However, as occurs with standard computer mouse, the transformation between the velocity of the hand/mouse on-board $\left( {\overrightarrow {V}_{h} } \right)$ and the on-screen cursor velocity $\left( {\overrightarrow {V}_{c} } \right)$ is not linear. In addition, the on-screen curser speed had been set to slow speed (The new mouse speed was set to ‘1’ through ‘windows’ SPI_SETMOUSESPEED function) to force wide hand movements by the subjects. In an ad hoc calibration test of the practical hand movement range of V _h  = 0–120 cm/s we observed a fine fit to simple quadratic relationship between cursor velocity and hand velocity as presented by Eq. 14. Precise analysis of hand velocity according to the recorded original cursor velocity, should consider the inverse transformation assuming that there was no distortion in the angle of the velocity vector.

$$ V_{c} \approx 0.0063 \cdot V_{h}^{2} + 0.65 \cdot V_{h} $$

(14)

Considering only the positive solution one can solve for the inverse transformation as follows:

$$ V_{h} \approx - 51.95 + \sqrt {2661.2 + 158.73 \cdot V_{c} } $$

(15)

This corrective action was done off-line to approximate the original mouse/hand movements. No linearization correction was implemented on-line, thus keeping the nonlinearity of the velocity transformation, as in a standard mouse cursor behavior, during the game. Figure 8 presents the planar trajectories and velocities of the cursor and the hand/mouse motion, during two typical no-perturbation trials movements (R → L and L → R). The cursor velocity recordings served to calculate the hand/mouse velocity by applying the inverse transformation described by Eq. 15. In addition, assuming no distortion in the angle of the cursor velocity vector in respect to the angle of the hand velocity vector, the Cartesian vectors of $\overrightarrow {V}_{h}$ were calculated (i.e., V _hx and V _hy ). Since the sampling rate of the cursor position was constant, the Cartesian coordinates of the hand/mouse position (i.e., X _h & Y _h ) can be calculated according to Eq. 16.

$$ X_{h} (t) = X_{h} (t = 0) + \int\limits_{0}^{t} {V_{hx} \cdot {\text{d}}t} \quad Y_{h} (t) = Y_{h} (t = 0) + \int\limits_{0}^{t} {V_{hy} \cdot {\text{d}}t} $$

(16)

Fig. 8

a and c Typical single L → R, and b and d typical single R → L movements’ trajectories and velocities. The continuous lines refer to the cursor, and the dotted lines refer to the calculated hand/mouse motion. The hand path is fairly straight and the velocity is roughly bell shaped

Although there is a quite difference between the velocity of the cursor and the velocity of the hand/mouse motion, their path are rather similar (straight line) and their velocity profiles are bell shaped. Due to the similarity in the path we assume that the cursor location and deviation from straight line represents the hand’s deviation from a straight line.