Can the Structure of Motor Variability Predict Learning Rate?

Recent studies show that motor variability is actively regulated as an exploration tool to promote learning in reward-based tasks. However, its role in learning processes during error-based tasks, when a reduction of the motor variability is required to achieve good performance, is still unclear. In this study, we hypothesized that error-based learning not only depends on exploration but also on the individuals' ability to measure and predict the motor error. Previous studies identified a less auto-correlated motor variability as a higher ability to perform motion adjustments. Two experiments investigated the relationship between motor learning and variability, analyzing the long-range autocorrelation of the center of pressure fluctuations through the alpha score of a Detrended Fluctuation Analysis in balance tasks. In Experiment 1, we assessed the relationship between variability and learning rate using a standing balance task. Based on the results of this experiment, and to maximize learning, we performed a second experiment with a more difficult sitting balance task and increased practice. The learning rate of the 2 groups with similar balance performances but different alpha scores was compared. Individuals with a lower alpha score showed a higher learning rate. Because the alpha scores reveal how the motor output changes over time, instead of the magnitude of those changes, the higher learning rate is mainly linked to the higher error sensitivity rather than the exploration strategies. The results of this study highlight the relevance of the structure of output motor variability as a predictor of learning rate in error-based tasks.