Explicit Action Switching Interferes with the Context-Specificity of Motor Memories in Older Adults

Abstract

Healthy aging impairs the ability to adapt movements to novel situations and to switch choices according to the context in cognitive tasks, indicating resistance to changes in motor and cognitive behaviors. Here we examined if this lack of "flexibility" in old subjects observed in motor and cognitive domains were related. To this end, we evaluated subjects' performance in a motor task that required switching walking patterns and its relation to performance in a cognitive switching task. Specifically, a group of old (>73 years old) and young subjects learned a new locomotor pattern on a split-belt treadmill, which drives the legs at different speeds. In both groups, we assessed the ability to disengage the walking pattern learned on the treadmill when walking overground. Then, we determined if this motor context-specificity was related to subjects' cognitive ability to switch actions in a set-shift task. Motor and cognitive behaviors were tested twice on separate visits to determine if age-related differences were maintained with exposure. Consistent with previous studies, we found that old adults adapted slower and had deficits in retention. Most importantly, we found that older subjects could not switch locomotor patterns when transitioning across walking contexts. Interestingly, cognitive switching performance was inversely related to subjects' ability to switch walking patterns. Thus, cognitive mediated switching interfered with locomotor switching. These findings were maintained across testing sessions. Our results suggest that distinct neural substrates mediate motor and cognitive action selection, and that these processes interfere with each other as we age.

Keywords: aging; generalization; human; locomotion; motor adaptation; motor learning; set-shift; split-belt treadmill.

Figures

Figure 1

Experimental Paradigms and Definition of Parameters. (A) Here the split-belt treadmill paradigm used for both age groups during their first and second visits is illustrated. Resting breaks, when subjects did not walk, are indicated by dashed lines. These were taken every 150 strides. (B) The left panel is a sample screen for the cognitive switching task that was used to assess cognitive switching ability. The right panel is a sample screen for the Symbol Digit Coding Task, which was used to assess processing speed. (C) This schematic adapted from Finley et al. (2015) illustrates Step Length Asymmetry (StepAsym), StepPosition, StepTime, and StepVelocity parameters.

Figure 2

Early Adaptation Behavior. (A) Stride-by-stride time courses during baseline and adaptation for StepAsym, StepPosition, and StepTime are shown. Shaded gray areas represent the adaptation period. Resting breaks, when subjects were not walking, are indicated by the white regions in between shaded areas. The last 50 strides of the second adaptation block (before subjects walk overground) are also shown. Black arrows indicate the decays in adapted state due to the passage of time during the resting breaks. Colored dots represent the average of 5 consecutive strides and colored shaded regions indicate the standard error for each group. (B) Bar plots indicate the mean time constants (i.e., τ) per group ± standard errors and statistical difference lines between groups illustrate significant ANOVA effects of age. Remember that a large τ indicates that subjects slowly adapted. Note that, on average, subjects' time constants are <150 strides indicating that they occurred before the first resting break. (C) Bar plots indicate the mean %Forgetting per group ± standard errors and statistical difference lines between groups illustrate significant ANOVA effects of age.

Figure 3

Scatter plots illustrate the relationship between %Forgetting and the adaptation time constant (τ). Multiple regression analysis indicate %Forgetting was a significant predictor of the adaptation rate of StepPosition, but not StepAsym or StepTime.

Figure 4

Late Adaptation Behavior and Learning. (A) Bar plots indicate the mean extent of adaptation (AdaptExtent) and adapted steady states per group ± standard errors. In general, all subjects reached the same adapted state. (B) Bar plots show the mean learning index per group ± standard errors. Recall that the learning index is quantified by the average after-effects on the treadmill during the catch trial, when both belts move at the same speed. We only found an age effect on the Learning Index for StepTime, which was driven by the smaller after-effects of young adults during their second visit compared to other groups (post-hoc p-values and statistical difference lines shown). We believe that this smaller after-effect indicates that young subjects can switch faster between the split and tied StepTime patterns during their second visit.

Figure 5

Overground Behavior. (A) Stride-by-stride time courses of StepAsym (left), StepPosition (middle), and StepTime (right) are shown for baseline and post-adaptation overground walking. Colored dots represent the average of 5 consecutive strides and colored shaded regions indicate the standard error for each group. (B) Bar plots indicate the mean Transfer values per group ± standard errors and statistical difference lines between groups illustrate significant ANOVA effects of age. These quantify the initial after-effects when walking overground right after split-belt walking. (C) Bar plots indicate the mean %Transfer values per group ± standard errors and statistical difference lines between groups illustrate significant ANOVA effects of age. %Transfer values indicate the amount of initial after-effects as a percent of AdaptExtent on the split-belt condition. In other words, %Transfer values takes into account how well subjects adapted their gait on the treadmill. While an age effect is found in all parameters, exposure effects are only found for StepAsymmetry, but not for StepPosition [Transfer: F(1, 35) = 2.68, p = 0.11 and %Transfer: F(1, 35) = 2.97, p = 0.094] or StepTime [Transfer: F(1, 35) = 0.01, p = 0.93 and %Transfer: F(1, 35) = 0.01, p = 0.92].

Figure 6

(A–D) Scatter plots of cognitive abilities vs. motor transfer. Scatter plots of cognitive ability vs. transfer. Each panel illustrates the scatter plots of cognitive abilities that we tested (i.e., cognitive switching and processing speed) vs. transfer and %Transfer of StepTime and StepPosition for younger and older subjects. A significant relation is only observed between old adults cognitive switching and transfer of StepTime after-effects when expressed as absolute values (Transfer) or as a percent of AdaptExtent on the treadmill (%Transfer). On the other hand, non-significant correlations were found between cognitive switching and Transfer of StepTime after-effects in young adults when quantified as Transfer [F(1, 8) = 2.32, p = 0.17] or %Transfer [F(1, 8) = 1.19, p = 0.20] of StepTime. Thus, motor and cognitive switching were only related in old, but not young subjects.

Figure 7

Washout of Split-belt After-effects Following Overground Walking. (A) Bar plots indicate the mean Washout values per group ± standard errors. These quantify the initial after-effects when returning to the treadmill after overground walking. (B) Bar plots indicate the mean %Washout values per group ± standard errors. %Washout values indicate the amount of remaining after-effects on the treadmill as a percent of AdaptExtent during the split-belt condition. In other words, %Washout values takes into account how well subjects adapted their gait on the treadmill. %Washout values of 100% indicate that the adapted movements on the treadmill remain intact after the overground walking experience.