Catch trials in force field learning influence adaptation and consolidation of human motor memory

Abstract

Force field studies are a common tool to investigate motor adaptation and consolidation. Thereby, subjects usually adapt their reaching movements to force field perturbations induced by a robotic device. In this context, so-called catch trials, in which the disturbing forces are randomly turned off, are commonly used to detect after-effects of motor adaptation. However, catch trials also produce sudden large motor errors that might influence the motor adaptation and the consolidation process. Yet, the detailed influence of catch trials is far from clear. Thus, the aim of this study was to investigate the influence of catch trials on motor adaptation and consolidation in force field experiments. Therefore, 105 subjects adapted their reaching movements to robot-generated force fields. The test groups adapted their reaching movements to a force field A followed by learning a second interfering force field B before retest of A (ABA). The control groups were not exposed to force field B (AA). To examine the influence of diverse catch trial ratios, subjects received catch trials during force field adaptation with a probability of either 0, 10, 20, 30, or 40%, depending on the group. First, the results on motor adaptation revealed significant differences between the diverse catch trial ratio groups. With increasing amount of catch trials, the subjects' motor performance decreased and subjects' ability to accurately predict the force field-and therefore internal model formation-was impaired. Second, our results revealed that adapting with catch trials can influence the following consolidation process as indicated by a partial reduction to interference. Here, the optimal catch trial ratio was 30%. However, detection of consolidation seems to be biased by the applied measure of performance.

Keywords: control strategy; dynamic perturbation; interference; intermittent practice; internal model formation; reaching movements; robotic manipulandum; variability.

Figure 1

(A) Robotic device BioMotionBot. (B) Subject performing the horizontal point-to-point reaching task. The cursor corresponding to the position of the handle and the targets were displayed on a screen facing the subject.

Figure 2

Representative mean hand trajectories and mean hand velocity profiles (outward movements only) of one group. (A) Straight-lined baseline trajectories. (B) Disturbed trajectories at the beginning of force field adaptation (first set). (C) Reshaped straight-lined trajectories at the end of force field adaptation (last set). (D) Smooth bell-shaped, single-peak baseline velocity profiles. (E) Disturbed velocity profiles at the beginning of force field adaptation. (F) Velocity profiles at the end of force field adaptation showing bell-shaped, single-peak profiles.

Figure 3

Mean time courses of velocity vector correlation coefficient for all three days. After learning force field A on day 1 (left), subjects of each catch trial ratio group were divided into control and test groups. Test groups adapted to an interfering force field B = −A on day 2 (mid). On day 3, all groups were retested in force field A (right). On all three days, subjects were able to adapt to the changed dynamic conditions indicated by increasing correlation coefficients. All data is presented as mean values ±95% confidence intervals.

Figure 4

Mean time courses of signed perpendicular displacement 300 ms after movement start in force field trials. Positive (negative) values indicate deviations in clockwise (counterclockwise) direction caused by disturbance of force field A (force field B). On all three days, subjects were able to adapt to the changed dynamic conditions leading to decreased errors. All data is presented as mean values ±95% confidence intervals.

Figure 5

Mean time courses of after-effects measured by signed perpendicular displacement during catch trials. Negative (positive) values indicate deviations in counterclockwise (clockwise) direction and therefore after-effects appropriate to force field A (force field B). The magnitude of after-effects increases with ongoing practice and is least for subjects receiving 30 and 40% catch trials. All data is presented as mean values ±95% confidence intervals.

Figure 6

Mean time courses of learning index which relates catch trials and force field trials. Learning of the clockwise-directed force field A is indicated with negative values, learning of the counterclockwise-directed force field B has positive sign. All data is presented as mean values ±95% confidence intervals.

Figure 7

Comparison of degree of adaptation between catch trial ratio groups using velocity vector correlation coefficient (A), perpendicular displacement (B), and learning index (C). All three performance measures indicate a significantly decreasing degree in force field adaptation with increasing catch trial ratio. All data is presented as mean values ±95% confidence intervals; (^*) indicates significant differences between catch trial ratio groups.

Figure 8

Comparison of development in initial performance from learning session (day 1) to retest (day 3) of force field A measured by velocity vector correlation (A), perpendicular displacement (B), and learning index (C). Positive values indicate a performance improvement, whereas negative values indicate a decreased initial retest performance compared to naive performance. In general, test groups show impaired consolidation compared to corresponding control groups indicated by a significant effect of interference (control, test). For velocity vector correlation (A) and perpendicular displacement (B), there is also a significant interaction of interference and catch trial ratio, indicating different consolidation depending on the catch trial ratio. Thereby, consolidation is least impaired for 30% catch trial test group. All data is presented as mean values ±95% confidence intervals; (^*) indicates significant differences between catch trial ratio groups.

Figure 9

Mean values of signed perpendicular displacement in catch trials of the first movement set at retest of force field A (day 3). All control groups show significant negative after-effect values indicating predictive force compensation appropriate to force field A. Test groups show significant positive after-effect values indicating predictive force compensation appropriate to force field B. All data is presented as mean ±95% confidence intervals.