Figure 3. Strategy suppresses implicit learning across individual participants.

(A–C) In Experiment 2, participants in the No PT Limit (no preparation time limit) group adapted to a 30° rotation. The paradigm is shown in A. The learning curve is shown in B. Implicit learning was measured via exclusion trials (no aiming). Preparation time is shown in C (movement start minus target onset). (D–F) Same as in A–C, but in a limited preparation time condition (Limit PT). Participants in the Limit PT group had to execute movements with restricted preparation time (F). The task ended with a prolonged no visual feedback period where memory retention was measured (E, gray region). (G) Total implicit and explicit adaptation in each participant in the No PT Limit condition (points). Implicit learning measured during the terminal no aiming probe. Explicit learning represents difference between total adaptation (last 10 rotation cycles) and implicit probe. The black line shows a linear regression. The blue line shows the theoretical relationship predicted by the competition equation which assumes implicit system adapts to target error. The parameters for this model prediction (implicit error sensitivity and retention) were measured in the Limit PT group. (H–J) In Experiment 3, participants adapted to a 30° rotation using a personal computer in the No PT Limit condition. The paradigm is shown in H. The learning curve is shown in I. Implicit learning was measured at the end of adaptation over a 20-cycle period where participants were instructed to reach straight to the target without aiming and without feedback (no aiming seen in I). We measured explicit adaptation as difference between total adaptation and reach angle on first no aiming cycle. We measured ‘early’ implicit aftereffect as reach angle on first no aiming cycle. We measured ‘late’ implicit aftereffect as mean reach angle over last 15 no aiming cycles. (K–M) Same as in H–J, but for a Limit PT condition. (N) Explicit adaptation measured in the No PT Limit condition in Experiment 2 (E2), No PT Limit condition in Experiment (E3, black), and Limit PT condition in Experiment 3 (E3, red). (O) Late implicit learning in the Experiment 3 No PT Limit group (No Lim.) and Experiment 3 Limit PT group (PT Limit). (P) Correspondence between late implicit learning and explicit strategy in the Experiment 3 No PT Limit group. (Q) Same as in G but where model parameters are obtained from the Limit PT group in Experiment 3, and points represent subjects in the No PT Limit group in Experiment 3. Early implicit learning is used. Throughout all insets, error bars indicate mean ± SEM across participants. Statistics in N and O are two-sample t-tests: n.s. means p > 0.05, ***p < 0.001.

Figure 3—figure supplement 1. Implicit error sensitivity varies with error.

(A) We empirically estimated error sensitivity on each trial in the limited preparation time (Limit PT) group in Experiment 2. The dashed horizontal line indicates the steady-state error sensitivity used in our competition theory predictions in Figure 3G. (B) Here we show error on each trial in the Limit PT group in Experiment 3. The horizontal cyan line shows the terminal error over the last 10 cycles. (C) Error sensitivity curves reported in Kim et al., 2018 denoted E1 (Morehead et al., 2017 results) and E2 (Kim et al., 2018 results). These two studies used invariant error-clamp tasks to isolate implicit learning. We compared our implicit learning measure in the Limit PT condition in Exp. 2 to these values. The vertical blue line shows the terminal error in B. The red star shows the terminal error sensitivity measured in A. In panels D–F, we show the same data as in A–C, except for the Limit PT condition in Experiment three where participants were tested on a personal laptop. Shaded error bars denote mean ± SEM across participants.

Figure 3—figure supplement 2. Comparing implicit and explicit adaptation via reported strategies.

In Figure 3, when analyzing the No PT Limit group (no preparation time limit) in Experiment 2, we measured implicit learning using exclusion trials at the end of adaptation. Next, we estimated explicit strategies by subtracting this reach-based implicit learning measure from the total adaptation measured over the last 10 cycles of adaptation (reach-based explicit measure). In addition, we also asked participants to report their explicit strategies after the probe period. Participants were shown a ring of circles surrounding each target and asked to indicate which circle best represented their aiming direction at the end of the experiment. We averaged this report-based explicit measure across all four adaptation targets, taking the absolute value for any misreported strategies (25% of all reports in opposite direction). We estimated report-based implicit learning by subtracting the reported explicit strategy from the total adaptation measured over the last 10 rotation cycles. (A) Here, we compare report-based explicit strategy with reach-based explicit strategy. Each point represents an individual participant. The solid line is the unity line. The bars at right show the mean value for each explicit measure. (B) Similar to A except here we compare report-based implicit learning with reach-based implicit learning. (C) Here we compared report-based implicit and report-based explicit learning measures. We also show the relationship predicted by the competition theory in blue (same as in Figure 3G). Error bars show mean ± SEM across participants. Statistics in A and B show paired t-tests: *p < 0.05.

Figure 3—figure supplement 3. Movement paths in Experiment 3 were straight and brisk.

In Exp. 3, we tested participants remotely in a laptop-based rotation study. Here, we show movement paths recorded by the computer in two example subjects: one in the Limit PT group (top row) and one in the No PT Limit group (bottom row). The left column shows trajectories during the baseline period. Note the four different groupings reflect the four different targets used in the task. The middle column shows trajectories during the rotation period. The color indicates the rotation trial number (blue is early in the rotation period, red is late in the rotation period). The data show a clear rotation in participant movement angle. Reach trajectories remained straight. Finally, the right column shows movement paths during the terminal period where participants were instructed to move straight to each target, without any cursor feedback.