Figure 6. Competition predicts changes in implicit error sensitivity without changes in implicit learning rate.

(A) Haith et al., 2015 instructed participants to reach to Targets T1 and T2 (right). Participants were exposed to a 30° visuomotor rotation at Target T1 only. Participants reached to the target coincident with a tone. Four tones were played with a 500ms inter-tone-interval. On most trials (80%) the same target was displayed during all four tones (left, High preparation time or High PT). On some trials (20%) the target switched approximately 300ms prior to the fourth tone (middle, Low preparation time or Low PT). (B) On Day 1, participants adapted to a 30° visuomotor rotation (Day 1, black) followed by a washout period. On Day 2, participants again experienced a 30° rotation (Day 2, blue). At left, we show the reach angle expressed on High PT trials during Days 1 and 2. Dashed vertical line shows perturbation onset. At middle, we show the same but for Low PT trials. At right, we show learning rate on High and Low PT trials, during each block. (C) As an alternative to the rate measure shown at right in B, we calculated the difference between reach angle on Days 1 and 2. At left and middle, we show the learning curve differences for High and Low PT trials, respectively. At right, we show difference in learning curves before and after the rotation. ‘Pre-rotation’ shows the average of Day 2 – Day 1 prior to rotation onset. ‘Post-rotation’ shows the average of Day 2 – Day 1 after rotation onset. (D) We fit a state-space model to the learning curves in Days 1 and 2 assuming that target errors drove implicit adaptation. Low PT trials captured the implicit system (blue). High PT trials captured the sum of implicit and explicit systems (green). Explicit trace (magenta) is the difference between the High and Low PT predictions. At right, we show error sensitivities predicted by the model. (E) Same as in D, but for a state-space model where implicit learning is driven by SPE, not target error. Model-predicted error sensitivities are shown. Error bars across all insets show mean ± SEM, except for the learning rate in B which displays the median. Two-way repeated-measures ANOVA were used in B, C, D, and E. For B and C, exposure number and preparation time condition were main effects. For D and E exposure number and learning system (implicit vs explicit) were main effects. Significant interactions in B, C, and E prompted follow-up one-way repeated-measures ANOVA (to test simple main effects). Statistical bars where two sets of asterisks appear (at left and right) indicate interactions. Statistical bars with one centered set show main effects or simple main effects. Statistics: n.s. means p > 0.05, *p < 0.05, **p < 0.01.