Behavioural learning performance for the decision-making and the reaching task.

<p><b>(A)</b> Representative participant data showing how a reaching trajectory is gradually updated to match the hidden target trajectory (red). The colours of the lines indicate the sequence of attempts (ranging from green to blue), with later attempts being closer to the target trajectory. <b>(B)</b> Learning curves (positive and negative feedback conditions) for the participants in the decision-making task (DM) and the motor learning/reaching task (MO). Points achieved (y-axis) are plotted against the number of attempts (1-25). The dark red-triangle line and dark green-circle line represent the negative conditions in the DM and MO task respectively, while the light red-triangle line and light green-circle line represent the positive conditions. Error bars indicate 95% confidence intervals (CI) across 24 participants. <b>(C-D)</b> Two representative participants in terms of their curvature exploration. The curvature parameter (x-axis) ranges from −1 to 1, where −1 = ‘curve to the left’, 1 = ‘curve to the right’, and 0 = ‘straight movement’. The participant in (C) evenly explored the curvature dimension, while the participant in (D) concentrated on straight movements with little curvature. <b>(E-F)</b> Two representative participants in terms of their error reduction in both the direction (open circle) and curvature (solid circle) dimensions plotted against the number of attempts. For the participant in (E), the error in both dimensions was reduced to a relatively low level, while for the participant in (F) the error in curvature remained high. The latter was due to the lack of exploration in the curvature dimension as shown in panel (D). <b>(G)</b> Each participant’s mean curvature across all movements during the reaching task (blue circles; the absolute values were used for the movements with negative curvature). Four participants (10,16,18,22) were identified as outliers (red crosses). The blue circles (mean curvature values) were counted as outliers if they were larger than <i>q</i>3 + 0.15(<i>q</i>3 − <i>q</i>1) or smaller than <i>q</i>1 − 0.15(<i>q</i>3 − <i>q</i>1), where <i>q</i>1 and <i>q</i>3 were the 25<i>^th</i> and 75<i>^th</i> percentiles respectively.