key: cord-0270899-15rntkvh
authors: Weightman, Matthew; Brittain, John-Stuart; Miall, R. Chris; Jenkinson, Ned
title: Residual errors in visuomotor adaptation persist despite extended motor preparation periods
date: 2021-12-11
journal: bioRxiv
DOI: 10.1101/2021.06.28.450124
sha: 8f61d676d5e3ea38785d4a51b6da7f527b16a7f5
doc_id: 270899
cord_uid: 15rntkvh

A consistent finding in sensorimotor adaptation is a persistent undershoot of full compensation, such that performance asymptotes with residual errors greater than seen at baseline. This behaviour has been attributed to limiting factors within the implicit adaptation system, which reaches a sub-optimal equilibrium between trial-by-trial learning and forgetting. However, recent research has suggested that allowing longer motor planning periods prior to movement eliminates these residual errors. The additional planning time allows required cognitive processes to be completed before movement onset, thus increasing accuracy. Here we looked to extend these findings by investigating the relationship between increased motor preparation time and the size of imposed visuomotor rotation (30°, 45° or 60°), with regards to the final asymptotic level of adaptation. We found that restricting preparation time to 0.35 seconds impaired adaptation for moderate and larger rotations, resulting in larger residual errors compared to groups with additional preparation time. However, we found that even extended preparation time failed to eliminate persistent errors, regardless of magnitude of cursor rotation. Thus, the asymptote of adaptation was significantly less than the degree of imposed rotation, for all experimental groups. Additionally, there was a positive relationship between asymptotic error and implicit retention. These data suggest that a prolonged motor preparation period is insufficient to reliably achieve complete adaptation and therefore our results provide support for the proposal that limitations within the implicit learning system contributes to asymptotic adaptation levels. New & Noteworthy Residual errors in sensorimotor adaptation are commonly attributed to an equilibrium between trial-by-trial learning and forgetting. Recent research suggested that allowing sufficient time for mental rotation eliminates these errors. In a number of experimental conditions, we show that while restricted motor preparation time does limit adaptation - consistent with mental rotation - extending preparation time fails to eliminate the residual errors in motor adaptation.

School of Sport, Exercise and Rehabilitation Sciences, 24

The Introduction the elimination of residual errors), which are usually suppressed during naturalistic feedback 96 conditions. 97

More recently, Langsdorf et al. (2021) presented an alternative theory to explain why 98 under normal, non-clamped environments, the central nervous system fails to eliminate 99 residual errors during motor adaptation. They suggested an intrinsic speed-accuracy trade-100 off: where time consuming planning processes are interrupted by the imperative onset of 101 movement, resulting in fast, but inaccurate movements. In their study, they showed that 102 when an obligatory 2.5 second wait period was introduced between target presentation and 103 movement initiation, participants were able to fully adapt to a 45° rotation, leaving no 104 persisting errors at the end of learning. However, if this wait period was not enforced or was 105 introduced at the end of the movement, when no planning was assumed to be taking place, 106 participants failed to fully counteract the rotation. The authors highlighted mental rotation 107 as a time-consuming cognitive process potentially involved in the planning of visuomotor 108 adaptation. 109

For some time now, visuomotor adaptation has been framed as a combination of distinct 110 learning processes (Mazzoni & Krakauer, 2006 & Georgopoulos, 1993) . 126

Neuronal population vectors recorded in the monkey motor cortex gradually rotate from a 127 stimulus direction to a cued movement direction during the planning of a reach 128 (Georgopoulos et al., 1989; Lurito et al., 1991) . Additionally, the completion of mental 129 rotation tasks require long reaction times with larger magnitudes of rotation ( Ethical Review Committee). Participants were self-reported as right-handed (181) or left-157 handed (19) and used their preferred hand to complete the task (all but 5 of the left-handed 158 participants completed the task using their right hand). All participants had normal, or 159 corrected to normal, vision and reported no history of neurological disease. Initially, 180 160 participants were pseudorandomised into one of nine experimental groups (each n = 20), 161 which differed in the amount of preparation time provided (0.35, 1 or 2.5 seconds) and the 162 magnitude of cursor rotation (30°, 45° or 60°). These participants formed the online arm of 163 the study (Table 1) . To avoid any possible confounds relating to online data collection and 164 only after COVID-19 restrictions relating to in-person human testing were lifted, we also 165 collected data from an additional experimental group who completed the task in a 166 laboratory setting. Participants in this group experienced a 45° cursor rotation, with 2.5 167 seconds of preparation time (Table 1) We developed a visuomotor adaptation task using the behavioural science experiment At the beginning of every trial the cursor (red filled circle) would be located in the centre of 202 the screen (grey background). The cursor was surrounded by two larger red annuli, which 203 formed a bullseye-like formation (Figure 1b) . At a set moment, the outer ring disappeared, 204 followed by the inner ring 0.5 seconds later, and the cursor then changed colour from red to 205 white after a further 0.5 seconds. Participants were told that this sequence of events should 206 be treated as a countdown to movement and that they should aim to time their movement 207 so that it was initiated in synchrony with the cursor changing colour from red to white 208 c.

the on-screen cursor moved in accordance with participants' movements. Adaptation trials 246 immediately followed, in which the cursor feedback was rotated 30°, 45° or 60° clockwise 247 (depending on the experimental group, see below) relative to participants' movement. This 248 block lasted 280 trials, selected to ensure learning approached its asymptotic limit, based on 249 pilot data collected prior to the present study. While there is some indication that full 250 asymptotic saturation of learning was not achieved in all conditions, regression analysis of 251 the last 40 trials of adaptation indicated that there was no significant trend remaining (all p 252 > 0.19, with all slopes between 0.041 and -0.039) and thus we use the term 'asymptote' to 253 refer to the final state reached at the end of the 280 adaptation trials. Following adaptation 254 trials, participants performed no-feedback trials (n = 40). During these trials, participants 255 were told to stop using any strategies they might have employed to achieve the task 256 objectives and try to aim directly for the target. The cursor was hidden at all times during 257 the trial, apart from when located in the central position for the countdown cue. End-point 258 error was not displayed during the no-feedback phase. Veridical cursor feedback and display 259 of end-point error were then restored for the final de-adaptation trials (n = 40). 260

There were ten participant groups (Table 1) with any significant main effects or interactions followed up with Bonferrroni-corrected 292 multiple comparisons. Asymptotic differences from the imposed rotation magnitude were 293 compared using a Wilcoxon Signed Rank test, after data from some groups failed normality 294 checks (Shapiro-Wilk test). Mean reach angles across all baseline trials and during the first 295 16 no-feedback trials for each participant were used for baseline and retention 296 comparisons. Data from the laboratory group were processed and analysed separately from 297 the online data as this group was not factored into the original study design. In-lab data was 298 compared to its online counterpart using unpaired t-tests (two-tailed) and from the degree 299 of imposed rotation using a Wilcoxon Signed Rank test. All statistical analyses were carried 300 out in MATLAB (R2018b) and SPSS (IBM, version 27). All ANOVAs were run as general linear 301 models. The threshold for statistical significance was set at p < 0.05 and we report W, F, T 302 and p-values, as well as effect sizes from ANOVAs (partial eta squared (ηp 2 )). In summary, our results support previous work suggesting that mental rotation 357 contributes an interval to the planning process which is proportional to rotation magnitude, 358 and that short periods of preparation interrupt planning, leading the greater residual errors. 359

However, for the smallest rotation, we found no relationship between preparation time and 360 error, suggesting that mental rotation was completed within the allowed time. 

We aimed to test the assumption that increased motor preparation periods may allow for 429 more complete adaptation during visuomotor rotation tasks. As such, we hypothesised that 430 shorter preparation periods would be sufficient to fully counteract a 30° rotation, as a small 431 rotation would require less (and therefore quicker) mental rotation before motor execution. 432

We then predicted that this effect would scale with an increased cursor rotation, so that 45° 433 and 60° would require greater mental rotation and thus more time to fully compensate. 434

Indeed, we did find that restricting planning time for moderate and larger rotations resulted An additional difference between the two studies, which may have had some bearing on 499 the results, is the nature of response cueing used. We used a visual countdown sequence 500 which enabled participants to accurately synchronise their movements with the go-signal 501 and ensured a tight coupling between response times and the preparation time groupings. 502

In contrast, Langsdorf et al. (2021) displayed the target and instructed participants to wait 503 until they heard a tone before responding. This difference in protocol resulted in response 504 times closer to 3 seconds. One may argue that more than 2.5 seconds are required to fully 505 prepare for a 45° rotated reach. This is unlikely however, as we show that even a 30° angle 506 was not fully compensated, and there was no evidence for a relationship with the 507 preparation interval for those groups. Previous research has also shown that mental 508 rotation and aiming towards angles up to 90° can be achieved in Despite participants failing to eliminate residual errors during late adaptation, our data do 517 include some hallmarks of a speed-accuracy trade-off. We found that motor adaptation was 518 impaired in the 45° and 60° rotational groups, during both early and later stages, when Additionally, impaired adaptation associated with restricted preparation times was coupled 524 with an increased retention during no-feedback trials, which is consistent with the idea that 525 learning involved more implicit, procedural processes (Fernandez-Ruiz et al., 2011 ; Haith et 526 al., 2015) . The greatest retention was seen in the 0.35 seconds / 60° group, which implies 527 that adaptation in this group may have been weighted more so towards implicit processes -528 also highlighted by the larger residual errors at the end of learning. This suggestion is 529 reinforced by the equal retention seen for the 2.5 second condition, across all three rotation 530 magnitudes. In this case, the residual errors are approximately equal, because the 531 preparation time exceeded that required for mental rotation. However, these assumptions 532 should be treated with some caution after recent commentary on how the dichotomy of 533 implicit and explicit components of motor adaptation are inferred (Hadjiosif & Krakauer, 534 2021) . 535

Although not statistically evident, a speed-accuracy trade-off may also be apparent in the 536 In summary, our data suggests that extending motor preparation and planning periods 548 alone is insufficient to eliminate residual errors during visuomotor adaptation, irrespective 549 of the size of imposed cursor rotation. While increased preparation time may help to 550 improve error reduction at larger rotation magnitudes, our results suggest there remains a 551 limit at which learning saturates at asymptote, perhaps only overcome with priming of 552 explicit strategies, further instruction or changes to experimental variables. We also provide 553 further support of the use of online data collection methods in the study of motor control 554 and learning. Understanding why the central nervous system fails to fully adapt movements 555 in response to environmental changes may be key when aiming to optimise rehabilitation 556 protocols following brain injury or disease. 557 558

An implicit memory of errors limits human sensorimotor adaptation. 561 Nature human behaviour

Contributions 563 of spatial working memory to visuomotor learning

Cognitive channels computing action distance and 566 direction

Flexible explicit but rigid implicit learning in a visuomotor 569 adaptation task

The timing mega-study: 572 comparing a range of experiment generators, both lab-based and online

Delayed feedback 575 during sensorimotor learning selectively disrupts adaptation but not strategy use

Aiming error under transformed spatial mappings suggests a 578 structure for visual-motor maps

Relation between 581 reaction time and reach errors during visuomotor adaptation

Secondary tasks impair adaptation to 584 step-and gradual-visual displacements

Mental 587 rotation of the neuronal population vector

Cognitive spatial-motor processes

The explicit/implicit distinction in studies of 592 visuomotor learning: Conceptual and methodological pitfalls

The influence of movement 595 preparation time on the expression of visuomotor learning and savings

Adaptation to visuomotor rotations in younger and older 598 adults

Formation of a long-term memory 600 for visuomotor adaptation following only a few trials of practice

The dynamics of memory as a 603 consequence of optimal adaptation to a changing body

Independent learning of internal models 606 for kinematic and dynamic control of reaching

Learning of visuomotor 609 transformations for vectorial planning of reaching trajectories

Prolonged 612 response time helps eliminate residual errors in visuomotor adaptation

Estimating the implicit 615 component of visuomotor rotation learning by constraining movement preparation time

Cerebellar anodal tdcs increases 618 implicit learning when strategic re-aiming is suppressed in sensorimotor adaptation

Cognitive spatial-motor 621 processes

An implicit plan overrides an explicit strategy during 624 visuomotor adaptation

Explicit and implicit processes constitute 627 the fast and slow processes of sensorimotor learning

Dissociable cognitive strategies for sensorimotor 630 learning

Savings upon re-aiming in 633 visuomotor adaptation

The influence of awareness on explicit and implicit 635 contributions to visuomotor adaptation over time

Psychopy2: Experiments in behavior made easy. Behavior Research 639 Methods

Common processing constraints for visuomotor 641 and visual mental rotations

Mental rotation of three-dimensional objects

Overcoming motor "forgetting" through reinforcement of learned actions

Interacting adaptive processes with 650 different timescales underlie short-term motor learning

Flexible cognitive strategies during motor learning

Explicit and implicit contributions to 655 learning in a sensorimotor adaptation task

Divided attention impairs human motor 658 adaptation but not feedback control

Moving outside the lab: The viability 661 of conducting sensorimotor learning studies online

Sensory 664 prediction errors drive cerebellum-dependent adaptation of reaching

Visuomotor adaptation: 667 how forgetting keeps us conservative

Persistent residual errors in motor adaptation tasks: reversion to 671 baseline and exploratory escape

An implicit plan still overrides an 674 explicit strategy during visuomotor adaptation following repetitive transcranial magnetic 675 stimulation of the cerebellum. Experimental Results

Causal role of motor preparation 678 during error-driven learning

Direct and indirect effects 681 of cathodal cerebellar tdcs on visuomotor adaptation of hand and arm movements

Targeted tdcs 684 selectively improves motor adaptation with the proximal and distal upper limb

Implicit visuomotor adaptation remains limited after 687 several days of training

The authors declare no competing interests. 700 701