Fig. 3: Changes in behavior and estimated learning parameters with positive scores (Experiment 2). | Nature Communications

Fig. 3: Changes in behavior and estimated learning parameters with positive scores (Experiment 2).

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

Fig. 3: Changes in behavior and estimated learning parameters with positive scores (Experiment 2).The alternative text for this image may have been generated using AI.

A Memory profiles of the two groups in Blocks 1 and 4 of meta-learning training with opposite reward functions (Promote, magenta, and Suppress, cyan). B Changes during blocks and estimated linear slopes in the initial/accumulated memory updates in the first/last R trials. C Changes in memory size during blocks and estimated linear slopes in the average score per trial. D Memory (aftereffect) profiles of the two groups in Blocks 1 and 4 of Probe. E Changes during blocks and estimated linear slopes in the initial memory update in the second E trial with rotation and the accumulation and retention of memory, measured as the average of all error-clamp trials. F Changes over blocks and estimated linear slopes in estimated learning parameters. G Meta-learning rates (η) for each experiment and the difference between punishments and rewards (difference between Experiments 1 and 2). For all panels, lines/dots/circles and error bars/shading represent group means and SEMs, except the error bars in (F) and (G), where box and whisker represent the 25–75th and 2.5–97.5th percentiles for the posterior density estimated by the MCMC method. Faded lines represent individual participants’ data in (B), (C), and (E). Each mean and SEM are calculated for data from 20 human participants per group (in total N = 40 for A–F, N = 80 for G). NS indicates “Not Significant” (two-sided Wilcoxon rank sum test for Training Block 1: p = 0.70 and Block4: p = 0.53, Probe-Block 1: p = 0.49 and Block 4: p = 0.23).

Back to article page