Fig. 4: Replication of previous reports on changes in the speed of motor learning by simulation with the present Meta-learning model. | Nature Communications

Fig. 4: Replication of previous reports on changes in the speed of motor learning by simulation with the present Meta-learning model.

From: Reinforcement learning establishes a minimal metacognitive process to monitor and control motor learning performance

Fig. 4: Replication of previous reports on changes in the speed of motor learning by simulation with the present Meta-learning model.The alternative text for this image may have been generated using AI.

Assuming that task error is a punishment feedback, the present meta-learning model replicated previous reports. A, B Changes in error-sensitivity (speed of learning, i.e., β) in different probabilities of flip in the perturbation direction influencing the history of error, reported in the original study (A) and simulated by the model (B). Sensitivity increased in a stable environment (z = 0.9, red) and decreased in a rapidly changing environment (z = 0.1, blue). Sensitivity did not show apparent changes in medium stability (z = 0.5, green). The simulated memory profile at the end of the task was also plotted for each condition (B, left). C, D Effects of manipulating task error on motor learning, reported in the original study (C) and simulated by the model (D). Acceleration of learning (saving, blue) disappeared when a task error was randomly given (green) or removed (red). E, F Effects of reward/punishment on trajectory error, reported in the original study (E) and simulated by the model (F). The inset shows the used meta-learning rates for the simulation. Learning accelerated only in adaptation to punishment (black), compared to adaptation to reward (red) or random positive (blue). For all panels, lines/dots and error bars/shaded areas indicate the mean and SEM. a.u. = arbitrary unit. Panel (A) is from David J. Herzfeld et al., A memory of errors in sensorimotor learning. Science, 345,1349-1353 (2014). DOI:10.1126/science.1253138. Reprinted with permission from AAAS. Panel (C) is adapted with permission from Leow, L. A., Marinovic, W., de Rugy, A. & Carroll, T. J. Task errors drive memories that improve sensorimotor adaptation. J Neurosci. 40(15), 3075-3088 (2020). https://doi.org/10.1523/JNEUROSCI.1506-19.2020. https://creativecommons.org/licenses/by/4.0/. Panel (E) is reproduced with permission from Springer Nature. Galea, J., Mallia, E., Rothwell, J. et al. The dissociable effects of punishment and reward on motor learning. Nat. Neurosci. 18, 597–602 (2015). https://doi.org/10.1038/nn.3956, Springer Nature.

Back to article page