Implicit Adaptation Processes Promoted by Immediate Offline Visual and Numeric Feedback

Abstract

In adaptation learning, visual feedback impacts how adaptation proceeds. With concurrent feedback, a more implicit/feedforward process is thought to be engaged, compared to feedback after movement, which promotes more explicit processes. Due to discrepancies across studies, related to timing and type of visual feedback, we isolated these conditions here. Four groups (N = 52) practiced aiming under rotated feedback conditions; feedback was provided concurrently, immediately after movement (visually or numerically), or visually after a 3 s delay. All groups adapted and only delayed feedback attenuated implicit adaptation as evidenced by post-practice after-effects. Contrary to some suggestions, immediately presented offline and numeric feedback resulted in implicit after-effects, potentially due to comparisons between feedforward information and seen or imagined feedback.

Keywords:

• adaptation learning
• internal models
• knowledge of results
• implicit processes

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Sample size was informed based on estimates of effect size (Cohen’s d) and sample sizes used in previous literature, where concurrent and delayed feedback were compared for after-effects in a visuomotor adaptation paradigm (d = 5.6, n = 8-9/group; Shabbott & Sainburg, 2010). Based on these large effects, in a 3-group design, a minimum sample size of 18 was estimated (n = 6/gp, 1-β = 0.80, α = .05). Because we also had a numeric feedback group, we initially constrained our sample size to a minimum of n = 8. This was increased to n = 13/group, based on a post-hoc decision to double the number of practice trials for an additional n = 5/gp. A post-hoc calculation in G*Power for a 4 Group x 2 Time-point mixed design, based on Cohen’s f = .25 (a more moderate effect size), yielded a min. sample size of N = 48 (12/gp).

2 After excluding trials where movement times exceeded 1000 ms, the remaining percentage for trials where movement times exceeded 350 ms for the ONLINE group was 2.69%. To confirm that participants in the ONLINE group were not adjusting movements online, we also ran Pearson’s correlations comparing mean constant error (binned into 5-trial averages) at peak velocity to error at movement end. There was a large significant correlation between the two measures, r = .96, p<.001. There was no difference (based on a dependent t-test), between error at movement endpoint and error at peak velocity, t(294) = .059, p = .95. There was also no evidence that individuals in this group were slowing down (i.e., longer MT) in order to achieve greater accuracy and hence use the feedback to correct (i.e., error at peak velocity), r = −.013, p = .63.

3 Model fit of adaptation data was evaluated using Akaike’s Information Criterion (AIC) by modeling trial as a continuous variable (CE and RT). The linear model produced the lowest AIC for measures of CE (AIC = 28639) and RT (AIC = 54525) during adaptation, and for assessment of after-effects (CE, AIC = 14754)

4 We ran secondary analyses on the extended practice duration participants during the adaptation phase (practice-blocks 6-40, n = 5/gp). For CE (see supplementary Table 2), significant block differences starting at practice-block 9 were shown on all subsequent practice-blocks (all ps <.04). There was only a significant Group X practice-Block interaction for the OFFLINE + 3s group at practice-block 10, but no interactions for later practice-blocks. For all extended practice groups, there was a significant reduction in error and for the most part, there were not significant differences in how this was achieved. For RT (supplementary Table 4), no practice-block or group effects were observed, but there were Group X practice-Block interactions for the OFFLINE + 3s group (blocks 12, 22, 25, 27, 28, 30, 34, 39, 40; ps <.03). Relative to practice-block 6, RT’s increased for OFFLINE + 3s at these later practice-blocks more so than the increase in RT that was observed for OFFLINE_VIS group at the same practice-blocks. There was also an interaction for the OFFLINE_# group at practice-block 9 and 15 (ps<.02; see Figure 3). For VE (supplementary Table 6), a significant group difference was shown when comparing the ONLINE_VIS group to the numeric feedback group (p=.001; OFFLINE + 3s, p = .06). A significant Group X practice-Block effect was again present at practice-block 9 (OFFLINE_#, p=.001) and 24 (OFFLINE + 3s, p = .04), but not in later practice-blocks.

Additional information

Funding

This work was supported by a Natural Sciences and Engineering Research Council of Canada Discovery Grant (RGPIN-2016-04269) awarded to Nicola J Hodges

.