Relationship between sensorimotor adaptation and cognitive functions in younger and older subjects

Abstract

We investigated whether deficits of adaptive improvement in seniors are related to an age-dependent decay of the brain’s executive functions. Younger and older subjects completed a battery of cognitive tests, and preformed aimed arm movements before and during exposure to rotated visual feedback. In accordance with previous work, we found that adaptive improvement during exposure was degraded in seniors, while the transfer of adaptation to a new motor task was not. This pattern of findings confirms that strategic control but not sensorimotor recalibration is affected by old age. Using multiple linear regression (MLR) to extract separate executive components from our test battery, we found that basic response speed and decision-making, but not the inhibition of prepotent responses or mental flexibility, were degraded in our older subjects. Again using MLR, we found that degraded adaptive improvement in our seniors was partly related to the decay of basic response speed and decision-making, and partly to age-dependent phenomena not addressed by our cognitive-test battery. Finally, we observed that interindividual variability of cognition and adaptive improvement was larger in old than in young subjects, which could explain why some previous studies found degraded adaptation in seniorswhile others did not.

Appendix

Partitioning Of Variances

Consider the linear equation:

$$ y{\text{ = }}\alpha {\text{ + }}\beta _{{\text{1}}} {\text{*}}x_{{\text{1}}} {\text{ + }}\beta _{{\text{2}}} {\text{*}}x_{{\text{2}}} {\text{ + }}\beta _{{\text{3}}} {\text{*}}x_{{\text{3}}} {\text{ + }}\gamma _{{\text{1}}} {\text{*}}z_{{\text{1}}} {\text{ + }}\gamma _{{\text{2}}} {\text{*}}z_{{\text{2}}} {\text{ + }}\gamma _{{\text{3}}} {\text{*}}z_{{\text{3}}} $$

(3)

where x _k and y _k are performance measures from two different behavioral tests, x and z. The total variance of y can be partitioned into several components. The variance which y shares with any of x _k can be quantified as the squared multiple correlation between y and all x _k , denoted as R ²(x ₁, x ₂, x ₃), or for short as R ²(x). If R ²(x) is significant, then test x makes a reliable contribution to y.

In Fig. 4, the total variance of x _k is represented by the circle x, which can be thought of as the envelope of three shapes representing x ₁, x ₂, and x ₃, respectively. Area a+b, therefore, reflects the variance shared between y and any of x _k , or R ²(x). Similarly, area a+b+c represents the variance shared between y and any of x _k and z _k , quantified as R ²(x ₁, x ₂, x ₃, z ₁, z ₂, z ₃), or for short as R ²(x, z).

Fig. 4

The concept of common and unique variance, as the basis of MLR analyses. The top circle represents the variance of an independent variable y, and the bottom circles the variance of two groups of independent variables x and z. The overlap reflects the variance of y shared with x only (area a), z only (area c), and with x and z jointly (area b)

Based on these considerations we can determine area c, which represents the variance shared between y and z but not x, as R ²(x, z) - R ²(x); if this term is significant, then test z makes a reliable contribution to y which cannot be explained by test x. We can also determine area b, which represents the variance shared between y and both tests jointly, as R ²(x) + R ²(z) - R ²(x, z). If this value is significant, then both tests make a joint contribution to y.

To consider a third test w, we can combine tests x and z and then re-apply the above reasoning. For example, the variance shared between y and w but not x nor z can be calculated as R ²(x, y, z) - R ²(x, y).