Content uploaded by Jana Maresch
Author content
All content in this area was uploaded by Jana Maresch on Oct 22, 2020
Content may be subject to copyright.
Eur J Neurosci. 2020;00:1–15.
|
1
wileyonlinelibrary.com/journal/ejn
1
|
INTRODUCTION
Adaptation to visuomotor rotations is assumed to consist
of at least two main underlying processes: an implicit pro-
cess, termed here “implicit adaptation”—which is slow,
expressible at low reaction times, and rigid in different con-
ditions—and an explicit process, termed “explicit re-aim-
ing”—which develops rapidly, requires a long preparation
time, and is highly flexible (Bond & Taylor,2015; Haith,
Huberdeau, & Krakauer, 2015; Huberdeau, Krakauer,
& Haith, 2015; McDougle, Bond, & Taylor, 2015;
Taylor & Ivry, 2011; Taylor, Krakauer, & Ivry, 2014).
Many different methods are used in visuomotor rotation
tasks to assess these processes (Fernandez-Ruiz, Wong,
Received: 1 March 2020
|
Revised: 13 August 2020
|
Accepted: 14 August 2020
DOI: 10.1111/ejn.14945
RESEARCH REPORT
Methods matter: Your measures of explicit and implicit processes
in visuomotor adaptation affect your results
JanaMaresch1
|
SusenWerner2
|
OpherDonchin1,2,3
© 2020 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
Abbreviations: AI, awareness index; CR, consistent reporting; HDI, high
density interval; IR-E, intermittent reporting exclusion; IR-EI, intermittent
reporting exclusion & inclusion; IR-I, intermittent reporting inclusion;
ROPE, region of practical equivalence.
1Department of Brain and Cognitive
Sciences, Ben-Gurion University of the
Negev, Be’er Sheva, Israel
2Institute of Movement and Neurosciences,
German Sport University, Cologne,
Germany
3Department of Biomedical Engineering
and Zlotowski Center for Neuroscience,
Ben-Gurion University of the Negev, Be'er
Sheva, Israel
Correspondence
Jana Maresch, Motor Learning Lab, Ben
Gurion University of the Negev, Marcus
Family Campus, Ben Gurion Drive 1,
8410501 Be’er Sheva, Israel.
Email: jana.maresch@gmail.com
Funding information
H2020 Marie Skłodowska-Curie Actions,
Grant/Award Number: 642961
Abstract
Visuomotor rotations are frequently used to study the different processes underlying
motor adaptation. Explicit aiming strategies and implicit recalibration are two of
these processes. Various methods, which differ in their underlying assumptions, have
been used to dissociate the two processes. Direct methods, such as verbal reports,
assume explicit knowledge to be verbalizable, where indirect methods, such as the
exclusion, assume that explicit knowledge is controllable. The goal of this study was
thus to directly compare verbal reporting with exclusion in two different conditions:
during consistent reporting and during intermittent reporting. Our results show that
our two conditions lead to a dissociation between the measures. In the consistent
reporting group, all measures showed similar results. However, in the intermittent
reporting group, verbal reporting showed more explicit re-aiming and less implicit
adaptation than exclusion. Curiously, when exclusion was measured again, after the
end of learning, the differences were no longer apparent. We suspect this may re-
flect selective decay in implicit adaptation, as has been reported previously. All told,
our results clearly indicate that methods of measurement can affect the amount of
explicit re-aiming and implicit adaptation that is measured. Since it has been previ-
ously shown that both explicit re-aiming and implicit adaptation have multiple com-
ponents, discrepancies between these different methods may arise because different
measures reflect different components.
KEYWORDS
explicit, implicit, measures, methods, visuomotor rotation | adaptation | motor adaptation
2
|
MARESCH Et Al.
Armstrong, & Flanagan, 2011; Haith et al., 20152015;
Heuer & Hegele,2011; Huberdeau etal.,2015; McDougle
et al., 2015; Morehead, Taylor, Parvin, & Ivry, 2017;
Taylor et al., 2014; Werner et al., 2015). These methods
differ in their underlying assumptions: direct methods,
such as asking subjects where they are aiming (Hegele &
Heuer,2010; Taylor etal.,2014) and questionnaires at the
end of the experiment (Benson, Anguera, & Seidler,2011;
Hwang, Smith, & Shadmehr,2006) assume that knowledge
is verbalizable; the former is more directed at subjects’
strategic corrections and the latter taps into subjects’ gen-
eral understanding of the perturbation. Indirect methods,
such as manipulating subjects’ behaviour using task design
(Fernandez-Ruiz etal.,2011; Haith etal.,2015; Morehead
etal.,2017), assume knowledge to be controllable.
Both types of methods have shortcomings, which are
mainly addressed outside the motor field in consciousness
research. Direct methods seem straightforward, but subjects
may refrain from answering or be unable to report on certain
experiences (Timmermans & Cleeremans, 2015). Reports
may also be contaminated by the observer paradox: asking
subjects to produce subjective reports or to reflect on their
own performance may influence the very processes that are
being monitored (Newell & Shanks,2014; Timmermans &
Cleeremans, 2015). Specifically, subjects reporting seems
to improve adaptation and it may affect the amount of ex-
plicit re-aiming (Langsdorf, Maresch, Hegele, McDougle, &
Schween, 2020). Indirect methods may not confound mea-
surement and awareness as fully as a direct measure, but
there is still the danger that the experimental manipulation
influences awareness. Additionally, indirect methods must be
interpreted with care: since the subject has not declared ex-
plicitly what is in their head, we are reaching our conclusions
through inference (Cleeremans, Destrebecqz, & Boyer,1998;
Jacoby, 1991; Timmermans & Cleeremans, 2015). Thus,
there is no gold standard for measuring either implicit ad-
aptation or explicit re-aiming. All measures must be viewed
critically.
Visuomotor rotation experiments quite commonly rely
on a direct measure of explicit re-aiming, the verbal report,
and an indirect measure of implicit adaptation that we call
exclusion. In the verbal report, subjects are asked to report
their intended aiming direction, usually based on landmarks
presented visually around the target. Exclusion is measured
in trials where subjects are told that the perturbation has been
removed and asked to aim straight for the target (Hegele &
Heuer,2010; Taylor & Ivry,2013). When measured after the
adaptation, this is called aftereffect. The residual learned re-
sponse in an exclusion trial is a measure of implicit adaptation.
The literature contains reports both of consistencies
and inconsistencies between these measures (Bond &
Taylor, 2015; Bromberg, Donchin, & Haar, 2019; Leow,
Gunn, Marinovic, & Carroll,2017; Taylor etal.,2014). The
fact that they can be inconsistent dovetails with recent pro-
posals that explicit re-aiming and implicit adaptation may
both be composed of multiple components. For instance,
McDougle and Taylor (2019) propose that explicit re-aim-
ing may have different components that are computed and
cached (McDougle & Taylor,2019). Others suggest that im-
plicit adaptation has labile and stable components (Heuer &
Hegele,2015; Kim, Parvin, & Ivry, 2019; Miyamoto, Wang,
Brennan, & Smith, 2014). One possibility is that different
measures weigh the components differently.
Thus, it may be a surprise that comparison of the mea-
sures has not yet been the focus of a dedicated, controlled
study. Our goal was to compare explicit re-aiming and im-
plicit adaptation under different experimental conditions as
measured by verbal report and by exclusion. We hypothesize
that differences we see in different measures across condi-
tions may come from different expression of underlying com-
ponents of explicit and implicit adaptation.
2
|
MATERIALS AND METHODS
All data and scripts used in order to present the figures in
this article are available in the online repository (https://osf.
io/6yj3u/).
2.1
|
Participants
Our main comparison was between consistent reporting and
intermittent reporting, however, in order to control for pos-
sible effects of the intermittent measures themselves and to
obtain additional information about explicit and implicit pro-
cesses throughout adaptation, we also compared two addi-
tional intermittent reporting groups. Overall, 53 right-handed
participants completed the study and were randomly assigned
to one of four groups: (a) consistent reporting (CR; n= 12;
mean age: 26.5 [range 24–30]; 5 female), (b) intermittent re-
porting exclusion (IR-E; n = 17; mean age: 26.0 [22–34];
7 female), (c) intermittent reporting inclusion (IR-I; n=12;
mean age: 24.4 [22–28]; 7 female) and (d) intermittent re-
porting exclusion and inclusion (IR-EI; n=11; mean age:
25.3 [24–30]; 7 female). All subjects signed an informed
consent form, which included basic information about their
relevant medical status. Subjects were not included for par-
ticipation if they had previously participated in visuomotor
rotation research, if they had any neurological disorders or
if they suffered from vertigo. Subjects were contacted and
recruited through the department of Biomedical Engineering
and received monetary compensation for their participation.
The experimental protocol was approved by the Human
Subjects Research committee of the Ben Gurion University,
and followed the ethical guidelines of the university.
|
3
MARESCH Et Al.
2.2
|
Experimental apparatus and
general procedures
Participants made centre-out, horizontal reaching movements
while holding on to the handle of a robotic manipulandum
(Figure1a). Movement trajectories were sampled at 200Hz
and a resolution of 0.3×10−3 degrees on each joint of the
shoulder, which translates into a resolution in Cartesian coor-
dinates of less than 0.2mm. The stimuli were projected onto
a horizontal plane in front of the participants (BenQ MS527,
3300 ANSI lumens), which also occluded vision of the sub-
ject's hand. At the beginning of each trial, the manipulandum
guided the subject's hand towards a white circle in the centre
of the display (5mm radius), which was positioned approxi-
mately 45cm away from the eyes of the subject. A small, cir-
cular cursor (white, 3mm radius) provided continuous online
visual feedback during each reach (except for trials in which
no visual feedback was provided). After 50ms, subjects were
presented with a small target (red, 7mm radius) at a distance
of 10cm from the origin. Targets could appear on a circle in
one of eight possible, equally spaced locations (45° between
targets: 0°, 45°, 90°, 135°, 180°, −135°, −90°, and −45°).
Target locations were in a pseudorandom order such that all
targets were experienced once before being repeated in the
next cycle. Subjects were instructed to make fast and accu-
rate shooting movements to the target, “slicing” through the
designated target. After leaving the starting position, partici-
pants had between 400 and 600 ms to reach the target in order
for the target to turn green.
This movement time was long enough for subjects to
reach the target, however, it did not allow for corrections of
movements. The target turned blue and yellow when move-
ments were too slow or too fast, respectively. Additionally, a
happy ding indicated target hit and an annoying buzz sound
indicated target miss.
In some trials, subjects were asked to report their intended
movement direction before beginning the movement. To this
purpose, 42 landmarks, spaced 5.625° apart, were shown on
a circle around the target. Landmarks were positive in the
clockwise direction and negative in the counterclockwise di-
rection (Figure2a) and rotated with the target, such that the
same landmarks would always appear in the same location
relative to the target (Bond & Taylor,2015). Landmarks were
only present during reporting trials. Verbal reports were re-
corded both online by the experimenter and digitally for later
verification.
FIGURE 1 Experimental task. (a) Subjects were seated in front of a desk, which occluded sight of their arm and hand and served as projection
screen for the visual scene. With their right hand, subjects held the handle of a robotic manipulandum. (b) Procedures for the four experimental
groups. Note that presentation of the different epochs is schematic and not to scale, i.e., the x-axis does not reflect the actual length of the individual
epochs. Subjects in the CR group reported consistently throughout the entire adaptation (green boxes) while subjects in the IR-E, IR-I and the IR-EI
groups performed regular adaptation blocks without reporting (grey boxes). At the end of each adaptation block, subjects performed four exclusion
trials in the CR group (light blue boxes), report and exclusion trials in the IR-E group (green and light blue boxes), report and inclusion trials in the
IR-I group (green and purple boxes) or report, exclusion and inclusion trials in the IR-EI group (green, light blue, and purple boxes). Each group
started with a baseline block and ended with a Process dissociation procedure block
(a) (b)
4
|
MARESCH Et Al.
In each trial, the cursor and the hand movement were re-
lated in one of three different ways: veridical, no vision, or
rotated. Veridical feedback was provided in most baseline
trials. In some trials, the visual field was rotated so that the
cursor was not precisely on the hand. This included a block
of trials during the baseline designed to teach subjects how to
report their aiming direction. In these trials, the visual field
was rotated by between −30° and 30° on each trial by a dif-
ferent amount governed by a sum of sinusoids. We also used
rotated trials in the adaptation blocks, where subjects were
exposed to a consistent 60° rotation of the cursor in clock-
wise direction. Finally, during the “refresh” adaptation block
between final exclusion and final inclusion, the cursor was
rotated by 60°. There were two different types of no vision
trials: inclusion trials and exclusion trials. The inclusion and
exclusion tasks thus only differ in the instructions. Before in-
clusion, subjects were instructed to “use what was learned
during learning”, and, before exclusion, subjects were asked
to “refrain from using what was learned, and perform the
movement as during baseline.” These instructions were iden-
tical to the ones used by Werner et al. (2015) and Werner,
Strueder, and Donchin (2019). Instructions were given fol-
lowing a pre-determined instructions protocol, which was
strictly followed (the full protocol is available on https://
osf.io/6yj3u/). Prior to the experiment, each part of the ex-
periment was shortly explained (familiarization, baseline,
FIGURE 2 Intermittent and final measures. (a) The difference between reported aiming directions and the hand path during adaptation
provides us with the implicitreport measure (Taylor etal.,2014). (b) The exclusion measure gives an indication of the implicit knowledge a subject
has and thus cannot control. By subtracting the exclusion from the movement direction during the adaptation epoch (four trials), we obtain a
measure of the explicit knowledge a subject had and can possibly control (explicitexcl). Note that as opposed to the calculation in (a), we here use
the preceding adaptation epoch. (c) Process dissociation procedure measure as adapted from Werner et al. (2015). The Awareness Index (AI, red) is
derived from the difference between inclusion and exclusion whereas the Unawareness Index (UAI, yellow) is equal to the exclusion. AI and UAI
may not add up to behaviour (see Inclusion undershooting the hand path)
(a)
(b)
|
5
MARESCH Et Al.
adaptation with intermittent test blocks, posttest). At this
stage, subjects were told that it was important to remember
what the baseline and adaptation blocks were since we would
refer to these blocks in later stages of the experiment (during
both exclusion and inclusion trials in intermittent test blocks
and in the posttest). The names of the blocks were explicitly
repeated at the beginning of both baseline and adaptation.
Additionally, subjects were asked to restate the instructions
in their own words before each intermittent and final test
block so as to ascertain that they had understood the task. In
order to ensure that our instructions would not reveal the na-
ture of the perturbation, great care was taken in formulating
the protocol not to use words like “re-aiming” and “rotation”
or refer to dissociations between hand and cursor.
2.3
|
Behavioural task
Prior to the start of the experiment, participants were told to
expect a change in the task after the baseline block that may
increase task difficulty. They were instructed to continue try-
ing to slice through the target with the cursor. This was fol-
lowed by a short familiarization session of 16 trials, during
which subjects could ask questions and become familiar with
the apparatus and task.
Familiarization was followed by a baseline block of 80
trials. The first half of the baseline block was 34 trials with
veridical feedback interspersed with 6 trials with no visual
feedback. Then, subjects conducted 40 trials of training in
reporting their aiming directions (described above). The pur-
pose of these trials was purely to train subjects in reporting
aiming direction without exposing them to strong adaptation.
Because of the changing perturbation, subjects were not able
to either adapt or develop a working strategy.
After the baseline block, subjects did a series of five adap-
tation blocks of 80 movements each. In all adaptation blocks,
visual feedback of the cursor was rotated by 60°. Subjects
in the CR group reported their intended movement direction
during the adaptation blocks.
Between each adaptation block, subjects did an intermit-
tent measurement block. The composition of this block var-
ied between the groups (see Figure1b). The CR group did
four exclusion trials. The IR-E group did four reporting trials
and four exclusion trials. The IR-I group did four reporting
and four inclusion trials. The IR-EI group did four reporting,
four inclusion and four exclusion trials. Pilots using different
sequences of the intermittent test blocks showed no order-ef-
fect and block order was thus always the same. As described
above, instructions were shortly repeated before each inter-
mittent measurement block. This explanation lasted about
20–30 s, whereas the timeframe between each individual
measurement block was kept as short as possible, lasting ap-
proximately 5–10s.
At the end of adaptation, subjects in all groups performed
a posttest, comprised of the full process dissociation proce-
dure. During a short break between the end of adaptation
and the posttest, subjects were briefed on the last part of the
experiment, reminding them of the names of the blocks and
the inclusion and exclusion instructions. Depending on the
group, additional care was taken to explain the type of tri-
als this group had not experienced yet (e.g., inclusion for the
IR-E group). The process dissociation procedure consisted of
48 trials, 16 of which were exclusion trials, 16 were “refresh”
adaptation trials, and the last 16 were inclusion trials. The
order of exclusion and inclusion was counterbalanced be-
tween participants as suggested by Werner et al. (2015).
After completion of the experiment, subjects filled out an
online questionnaire about the experiment and the task. In
the first part subjects had to answer open questions about the
nature of the perturbation, and in the second part, the same
questions were posed as multiple choice.
2.4
|
Movement analysis
We characterized movement direction on each trial using the
direction of the hand movement on that trial. Movement onset
was defined as the time at which hand speed first exceeded
5cm/s. To assess movement direction, we tested three different
measures of movement direction: error at 8cm, first maximal
error (FME) and error at peak velocity (PVE). We found large
overlap between the different metrics, indicating that our re-
sults were not sensitive to the precise measure of movement
direction we used and that subjects followed our instructions to
produce fast shooting movements, which led to straight hand
trajectories. We decided to use PVE as our measure for move-
ment direction because it had the smallest within subject vari-
ability in our data. PVE was specifically defined as the angle
between a line connecting starting and target dot and a line be-
tween movement onset and movement position at peak veloc-
ity. Average time until peak velocity was comparable across
groups and the different types of trials (overall mean: 173ms).
All trajectories were rotated to a common axis with the target
location at 0°. Positive angles indicate a counterclockwise devi-
ation of the hand from the target, meaning a clockwise rotation
of the cursor. Reaction time for all groups was defined as the
time between target appearance and movement onset as used in
previous reports (Bond & Taylor,2015; McDougle etal.,2015;
Taylor etal.,2014). Note that in case of reporting trials, this
definition does not take into account any possible differences
between delays in reaction time caused by reporting and “pure”
reaction time as measured during non-reporting trials. In all
groups, a trial was omitted if the subject's movement fell short
of reaching the target. In the intermittent reporting groups, trials
were excluded if reaction times exceeded 1,000ms. No time re-
strictions were placed on the CR group; however, subjects were
6
|
MARESCH Et Al.
encouraged to report and move as quickly as possible. 1.5% of
all trials were thus excluded. Please note that movement times
were long enough for subjects to reach and “slice through” the
target on most trials. To visualize movement direction, we used
bins of four trials each, which was the size of the intermittent
measurement blocks.
In addition to movement direction, we had subject reports
for intermittent reporting trials in all groups and also for the
adaptation trials in the CR group.
We further summarized subject performance by calculat-
ing a number of additional measures from the movement di-
rections and reporting.
2.4.1
|
Implicitreport
The implicitreport was the difference between report explicit
and movement direction (Figure 2a). While the report ex-
plicit provides us with an estimate of the explicit re-aiming
subjects can verbalize, the implicitreport measures adaptation
outside explicit verbal access.
2.4.2
|
Explicitexcl
Explicitexcl is the difference between movement direction at
the end of adaptation immediately preceding the exclusion
epoch and the movement direction in exclusion. This is the
change in the movement direction caused by the subject's re-
aiming. As such, it is an additional measure of explicit re-
aiming, which measures the subject's explicit control over
their own behaviour as opposed to measuring verbalizable
explicit.
2.4.3
|
Awareness index
Awareness index (AI) was calculated for each subject using
average movement direction of the exclusion (
Mexclusion
) and
inclusion (
Minclusion
) epochs during the intermittent test (for
the IR-EI group) or the posttest (for all groups) according to
the following formula:
This calculation is based on the calculation used by
Werner et al. (2015, 2019) without the normalization they
used to allow comparison of different rotation sizes. Without
the normalization, the unawareness index reduces to the
mean of the final exclusion. Thus, we did not calculate the
unawareness index.
Note that we have two measures of explicit re-aim-
ing based on the exclusion: explicitexcl (adaptation minus
exclusion) and the awareness index (inclusion minus exclu-
sion). Generally, the awareness index will indicate less ex-
plicit re-aiming than the explicitexcl. The gap between them
is a result of inclusion being generally lower than adaptation,
i.e., closer to baseline (Werner et al., 2015, 2019). This may
be because of the lack of visual feedback in the inclusion
blocks or for other reasons.
2.5
|
Statistical analyses
We used a Bayesian statistical approach very similar to the
approach used in Werner et al. (2019) to fit a linear model
to our data. The Bayesian approach combines prior informa-
tion about population parameters with evidence contained in
the sample itself in order to create posterior probability dis-
tributions of the parameters (Kruschke,2014). In contrast to
frequentist approaches, Bayesian probabilities are thus state-
ments about probabilities of these parameters in general, and
not about the specific sample in the study. Full details of
the model are available in the online repository (https://osf.
io/6yj3u/) and here we describe only the key points. The de-
pendent variable for our linear model was the movement di-
rection of each subject on each trial. The model's independent
variables included the epoch type, the subject's group, and
the subject. The model also included the group by epoch type
interaction and the subject by epoch type interaction. There
were eight epoch types, see Table1 for the overview of these
epochs. Since not all groups did all epochs, we used imputa-
tion as described in Kruschke (2014), Ch. 20, to balance the
model. Bayesian imputation requires a hierarchical model so
that the statistics of the imputed data can be learned from the
existing data. Thus, we made the group by episode interaction
a hierarchical parameter. We gave it a t-distribution in order
to allow outliers and sampled both the variance and the de-
grees of freedom of this distribution, as well as sampling the
specific values for every episode and every group. The other
coefficients of the model were assumed to be normally distrib-
uted and were drawn from a broad, uninformative prior. The
AI
=
Minclusion
−
Mexclusion
TABLE 1 Overview of the different epochs. Shown are the
number of epochs as used in the Bayesian analysis and the number of
trials included in each epoch
Epoch types Number of epochs
Number of
trials
Baseline 1 80
Adaptation 100 400 (5×80)
Report 5 20 (5×4)
Exclusion 5 20 (5×4)
Inclusion 5 20 (5×4)
Final exclusion 1 16
Final inclusion 1 16
|
7
MARESCH Et Al.
standard deviation of movements around the linear model was
assumed to differ among subjects and it was also sampled for
each subject from a gamma distribution which was determined
through hierarchical sampling. MATLAB code and model are
available in the online repository (https://osf.io/6yj3u/).
The joint posterior distribution of the model's parameters
were sampled using JAGS (4.2.0, http://mcmc-jags.sourc
eforge.net/) called from MATLAB (2019b, the Mathworks,
Natick) using matjags (http://psiexp.ss.uci.edu/resea rch/
progr ams_data/jags/). We used four chains, 2,000 burn-in
samples and 10,000 samples per chain. Using standard di-
agnostics described by Kruschke (2010), we ensured that the
chains had converged to a unimodal distribution for all pa-
rameters and that the results were consistent across chains.
Again, full details of the sampling, the posterior samples and
code to produce all the diagnostic plots are available in the
online repository (https://osf.io/6yj3u/).
Using the sampled estimates of the linear coefficients, we
calculated the posterior distribution of the mean movement
direction for each epoch, subject, and group. We report our
results using the mean and 95% high density interval (HDI) of
these estimates of the mean. The HDI contains 95% of the dis-
tribution, in which every point has higher credibility than any
point outside this range. Furthermore, for group comparisons
we specified a region of practical equivalence (ROPE) of −3°
through 3° around a value of 0° difference between the groups.
The limits of this ROPE are an arbitrary choice, which is based
on what we consider a meaningful difference. We report all
actual differences in our results, so each reader can thus decide
for themselves whether differences are meaningful. We report
the percentage of the HDI that lies within the ROPE as a mea-
sure of the probability that the values are equivalent.
Unless otherwise indicated, brackets represent 95% HDI.
3
|
RESULTS
3.1
|
Adaptation and learning rates
The time course of movement directions showed a stereotyp-
ical learning curve for all four groups (Figure3a). However,
the learning curves differed between the CR and the inter-
mittent groups: adaptation proceeded slower and reached as-
ymptote later in the intermittent groups than in the CR group
(see difference between blue learning curve and the rest). We
compared differences in movement direction early and late
in adaptation by computing our estimate of the mean over
(a) the last eight trials of the first adaptation block and (b)
the last eight trials of the last adaptation block (asymptote
of learning). Figure3b depicts the mean movement direction
for early and late adaptation: the CR group showed greater
change in movement direction during early adaptation (50.7°
[48–53]) as compared to the IR-E (39.9° [38–42]), IR-I
(39.4° [37–42]) and IR-EI group (40.3° [38–43]). These dif-
ferences of 10°–11° are consistent across groups. Although
the magnitude of the difference in movement directions be-
tween groups decreased to about 9°, it persisted until the end
of adaptation (CR: 61.1° [59–34]; IR-E: 52.4° [50–54]; IR-I:
52.5° [50–55]; IR-EI 52.7° [50–55]).
3.2
|
Intermittent measures
In Figure4, we show the results of the intermittent test blocks.
Results for these measures were comparable within the IR
groups, we thus first present the IR-E group's results as rep-
resentative for all IR groups and we show an overview of all
groups in Figure 5. Figure 4a depicts the time course of the
intermittent measures for the CR and the IR-E group. Measures
of explicit re-aiming are in green (report and explicitexcl ) and
measures of implicit adaptation are in blue (implicitreport and
exclusion ). All measures in all groups changed over the first
two epochs and then stabilized. After examining the data, we
decided to use the average value of the last three epochs for any
comparisons between measures and groups, as behaviour had
stabilized and subjects showed little variation between epochs.
FIGURE 3 (a) Binned estimated mean movement direction per
group. Dotted lines show location of test blocks, note however, that the
actual intermittent tests are not presented here. The posttest (Process
dissociation procedure) is also not shown. (b) Early and late movement
direction. For the early adaptation, we used the last two epochs of the
first adaptation block (8 trials) and for the late adaptation, we used
the last two epochs of the last adaptation block (8 trials). Vertical
lines denote 95% HDI of mean of individual subjects. Error bars
show 95% HDI of mean of the group.
8
|
MARESCH Et Al.
Implicitreport showed slightly higher levels of implicit ad-
aptation than exclusion in the CR group (Figure4b, top left
column; 28.8° [26–31] and 23.8° [21–26], respectively). In con-
trast, implicitreport was much lower than exclusion in the IR-E
group (Figure4b, top right column; 29.1° [26–32] and 43.4°
[41–46], respectively). The fact that implicitreport and exclusion
are more different in the IR-E group arises because exclusion is
different between the groups: implicitreport shows similar values
between groups whereas exclusion values for subjects in the
IR-E group are ~20° higher than in the CR group.
Differences in measures of explicit re-aiming reflected our
findings for the implicit adaptation. Reported aiming direc-
tions in the last three blocks were lower than the explicitexcl
in the CR group (Figure4b, bottom left; 30.1° [29–31] and
37.0° [33–40]). In contrast, the IR-E group showed much
higher explicit re-aiming when using reported aiming direc-
tions than when using explicitexcl in (Figure4b, bottom right;
22.0° [20–24] and 10.0° [7–12]). While implicitreport was sim-
ilar between the groups, explicit re-aiming was greater in the
CR group than the IR-E group according to both measures of
explicit re-aiming. This difference was, however, much more
striking for the explicitexcl than for the report.
In Figure5, we present results also from the IR-EI group
and show that our results are consistent across subjects and
groups. The two measures produce different results, and
these results are different for subjects in CR and IR groups.
Exclusion is larger than implicitreport for almost all subjects in
the IR groups (orange and purple) while exclusion is smaller
than implicitreport for nearly all subjects in the CR group
(blue). Similarly, report is greater than explicitexcl in the IR
groups and the opposite is true in the CR group. Taken to-
gether, our results indicate that reporting and exclusion are
not measuring the same thing, although the difference may
be hidden in some experimental conditions.
3.3
|
Intermittent measures IR-I and IR-
EI group
Subjects in both the IR-I and IR-EI group performed inclusion
as well as intermittent report trials. The IR-EI group, moreo-
ver, completed the entire process dissociation procedure in
FIGURE 4 Intermittent explicit and implicit measures for the CR
and IR-E groups. Intermittent measures were comparable across the
IR groups, thus only measures for the IR-E group are presented here.
Note that individual subject HDIs are sorted and thus do not correspond
to subjects across sub-figures. (a) Time course of report and exclusion
measures per group. On the left we present the CR group, on the right the
IR-E group. (b) Intermittent measures per group. Bars show estimate of
the mean of the last three test blocks. Top row shows implicit measures;
bottom row shows explicit measures. Black lines depict 95% HDI for
means of individual subjects; error bars show 95% HDI for means per
group
CR group IR-E group
Explicit Implicit
(a)
(b)
FIGURE 5 Explicit and implicit measures for individual subjects
across all groups. (a) Exclusion versus implicitreport and (b) explicitexcl
versus report
|
9
MARESCH Et Al.
each intermittent measurement block. During inclusion, we
ask subjects to perform the same movement as during ad-
aptation, but without visual feedback. Figure6a shows that
subjects in both groups tended to undershoot during these
trials, meaning less adaptation during inclusion trials than
during the preceding adaptation trials (47.5° [45–50] and
43.4° [41–46] during inclusion for the IR-I ad IR-EI groups,
respectively). The difference between adaptation and inclu-
sion trials is relatively high, with the HDI of this difference
not including zero for either IR-I group (5.8° [3–9]) or the
IR-EI group (10.4° [7–14]). While the HDI and the ROPE
of the IR-EI group do overlap, this overlap reflects less than
0.01% of the posterior density. The percentage of the poste-
rior density in the overlap is slightly higher in the IR-I group
(0.04%). Thus the two values are certainly not equivalent and
may well be different for both groups. The fact that inclusion
may not be the same as movement direction at the end of ad-
aptation means that we need to be careful when interpreting
the explicitexcl.
We therefore calculated the AI using the intermittent mea-
surement blocks of the IR-EI group. This intermittent AI en-
abled us to compare the two explicit measures and the AI with
each other within the same group. As with the other intermit-
tent measures, we used the last three intermittent measure-
ment blocks for this comparison. Figure6b shows that the AI
estimates explicit re-aiming lower than both other measures,
while the report provides the highest estimate (report: 17.8°
[16–20]; explicitexcl: 12.0° [8–15]; AI: 1.2° [−5–2]). This
further emphasizes that report estimates explicit re-aiming
higher than exclusion based measures.
3.4
|
Posttest
Our posttest, the process dissociation procedure, consisted
of inclusion and exclusion blocks at the end of adapta-
tion. At this point, each group received a short instruction
about the posttest, which led to a short delay of approxi-
mately 2 min. Since subjects had experienced exclusion
and/or inclusion intermittently (depending on the group),
we expected these measures to produce similar results in
the final test. However, this was not the case. In Figure7,
we show the different final measures per group. Notably,
final exclusion is similar across groups (Figure7a,b; CR:
31.6° [29–34], IR-E: 32.6° [31–35], IR-I: 37.6° [35–40]
and IR-EI: 31.9° [29–34]), contrasting with our findings in
the intermittent measurement blocks. This difference be-
tween intermittent and final exclusion was positive for the
CR group (8.5° [5–11]) while it was negative for the IR-E
(−11.7° [−15 to −9]) and IR-EI group (−9.7° [−14 to −6]).
The consistency of this result across individuals is shown
in Figure 7c, left scatter. In the discussion, we consider
the possibility that this difference is the result of decay in
a labile component of implicit adaptation caused by the
time delay between the end of adaptation and the posttest
(Hadjiosif & Smith, 2013; Kim et al., 2019; Miyamoto
etal.,2014; Morehead,2018).
For this reason, we also considered the differences be-
tween the movement directions at the end of adaptation and
in the posttest refresh block. We show in Figure7 that move-
ment direction generally reverts back towards baseline for
most subjects between the end of adaptation and the refresh
phase (Figure7c, right scatter plot). This means that there is
a decay in the learned adaptation during the interval between
the end of adaptation and the beginning of the posttest. This
is larger in the IR groups and it is consistent with the earlier
hypothesis of a labile implicit component that is missing in
the CR group.
Unlike final exclusion, final inclusion measures are dif-
ferent between groups. Subjects in the IR-I group showed the
lowest values for the final inclusion (Figure7b; IR-I: 31.7°
[29–34]). Subjects in the IR-E and IR-EI group showed sim-
ilar and higher inclusion values (IR-E: 46.2° [44–48]; IR-EI:
45.7° [43–48]), whereas subjects in the CR group were clos-
est to full compensation during their inclusion (Figure 7b;
CR: 51.0° [49–53]).
Finally, our results for the AI reflect our findings de-
scribed in the previous paragraph. The AI differed between
groups and showed the same pattern as the final inclusion
FIGURE 6 Intermittent measures for the IR-I and IR-EI groups.
(a) Intermittent inclusion for the IR-I and IR-EI group. (b) Explicit
intermittent measures for the IR-EI group. Shown is the mean of the
reported aiming direction, the mean exclusion explicit and the AI
(calculated from the difference between inclusion and exclusion).
Black lines denote 95% HDIs for individual subjects, error bars show
95% HDIs for the respective group
10
|
MARESCH Et Al.
where subjects in the CR, in the IR-E and IR-EI groups
showed more awareness than subjects in the IR-I (Figure8;
CR: 19.0° [16–22], IR-E: 13.4° [11–16] and IR-EI: 13.5°
[10–17] as compared to IR-I: −5.0° [−8 to −1]).
3.5
|
Reaction times
Figure9 shows reaction times for the different groups during
early (empty circles) and late adaptation (full circles) against
our measures of explicit and implicit adaption. The CR-group
showed consistently higher reaction times compared to the
other groups. This is not surprising as subjects in this group
were required to pause and state their intended aiming di-
rection before moving towards the target (see also Bond &
Taylor,2015; McDougle etal., 2015; Taylor et al.,2014)).
Reaction times did not differ between early and late adapta-
tion and they were also not related to the different measures
within each group (data not shown). The figure indicates
that one possible explanation for the differences between the
groups may be the differences in reaction times.
FIGURE 8 Awareness index for all groups calculated as the
difference between final inclusion and exclusion. Vertical lines
represent 95% HDI for individual subjects, error bars show 95% HDI
per group
FIGURE 9 Reaction times for all groups. Shown are reaction
times per measure (left and right column) for implicit and explicit
processes (top and bottom row). Filled circles show reaction times in
late adaptation (last two adaptation bins), whereas open circles show
reaction times in early adaptation (first two adaptation bins). Note that
the x-axis is presented on a logarithmic scale
FIGURE 7 Posttest per group. Shown are the posttest measures
and the differences between intermittent and final epochs. (a)
Movement directions during posttest. Note that exclusion is presented
before inclusion, whereas the order was randomized across subjects.
Shaded area denotes HDI. (b) Final exclusion (left) and final Inclusion
(right). Shown is the mean per 16 trials. Vertical lines represent 95%
HDI for individual subjects, error bars show 95% HDI per group. (c)
Differences between intermittent and final movement directions for all
groups. Left: Scatter plot of intermittent versus final exclusion. Right:
Scatter plot of movement direction at the end of adaptation versus
during refresh
(a)
(b)
(c)
|
11
MARESCH Et Al.
3.6
|
Questionnaire
As described in the methods, each subject filled out an online
questionnaire starting with three open questions about their
understanding of the task. After this, the same questions were
framed as multiple-choice questions. Answers and corre-
sponding codes can be found in the online repository (https://
osf.io/6yj3u/). We coded the answers to the open questions in
categories ranging from 1 to 5 (1 means full understanding of
the task, 5 is no understanding). Three researchers separately
assigned the codes to the answers of each subject. We present
the median values and interquartile range of the codes across
the three researchers and subjects per group. Subjects in the
CR, IR-E, and IR-EI group scored 2 points (2±0, 2±0.5,
2±0.5 for CR, IR-E, and IR-EI, respectively) in the open
questions (subjects understood that a consistent rotation was
applied), whereas the IR-I group scored 3±0 points (sub-
jects understood that a rotation or perturbation was applied
but thought it was not consistent). This stands in line with our
final inclusion and AI results where the IR-I group showed
the lowest values.
In the multiple-choice questions, subjects were (a) asked
to provide an explanation for the mismatch between the cur-
sor and the hand and (b) asked in which direction the cursor
was rotated relative to their hand. Subjects in the CR group
were able to better characterize the mismatch than subjects
in the IR groups (percentage correct: 92%, 53%, 67%, and
75% for the CR, IR-E, IR-I and IR-EI group, respectively).
When asked to classify the direction of the rotation, subjects
in all groups were close to chance level (percentage correct:
58%, 41%, 58%, and 42% for the CR, IR-E, IR-I, and IR-EI
group, respectively). Here, the first multiple choice question
reflects a higher understanding of the task for subjects in the
CR group, however, no further parallels can be drawn with
our behavioural results.
4
|
DISCUSSION
Our goal was to compare different types of measures of ex-
plicit re-aiming and implicit adaptation in a visuomotor rota-
tion task to determine whether they measure the same thing.
Our central result is that they do not. We compared two
measures of explicit re-aiming (report and explicitexcl) and of
implicit adaptation (implicitreport and exclusion) in four con-
ditions. Both report measures (report and implicitreport) meas-
ure verbalizable knowledge whereas the exclusion measures
(exclusion and explicitexcl) give an indication of what subjects
can control. We compared a consistent reporting group (CR),
which also performed intermittent exclusion trials, with three
intermittent reporting groups who performed exclusion tri-
als (IR-E), inclusion trials (IR-I) or both exclusion and inclu-
sion trials (IR-EI). Performing both inclusion and exclusion
trials provides a full process dissociation procedure as used
by Werner etal.(2015, 2019), which provides an additional
measure based on exclusion. We found that consistent re-
porting affected our measures differently than intermittent
reporting, regardless of the intermittent measures used.
Our measures produced consistent results in the CR group
(Figure5 bottom, blue dots), but in the other groups we found
differences in both explicit re-aiming (report was greater than
exclusion) and in implicit adaptation (exclusion was greater
than report). This can be seen in Figure5 in the orange and
purple dots. The results of the process dissociation procedure
from the IR-EI group suggest that this discrepancy in the IR
groups may be even greater because considering exclusion
without also considering inclusion may overestimate explici-
texcl (Figure6b).
It is worth making clear what these differences actually
mean. A typical subject in the CR group will have fully
adapted to the 60° rotation. When asked, however, they will
report an aiming angle of only 30°. When asked to shift their
aim back to the target, they will shift their aim by more than
30°. Typically, they will shift their aim by 40°, suggesting
that they may have more control over their aiming direction
than expressed in the report.
In contrast, a typical subject in one of the IR groups will
have adapted partially. Rather than a full 60° adaptation, they
will move towards 50°. When asked, they will report an aim-
ing angle of 20°. However, when asked to shift their aim back
to the target they will shift their aim by only 10°, showing
that they have less control over their aiming direction than
they think they do.
The differences in the explicit re-aiming shown by the two
measures may reflect biased measurements. For instance, it is
possible that report underestimates the explicit re-aiming in
the CR group. The CR group has faster and more complete
adaptation than the IR groups (Figure3), and these have pre-
viously been associated with more explicit re-aiming (Benson
etal.,2011; Haith etal.,2015; Leow etal.,2017). This is in
line with previous work suggesting that verbal responses may
not reveal all of subjects’ knowledge, especially when knowl-
edge is held with low confidence or is retrieved in a differ-
ent context (Eriksen,1960; Nisbett & Wilson,1977; Shanks,
Rowland, Rowland, & Ranger,2005). Conversely, it has been
suggested that prediction tasks—such as reporting—during
sequence-learning, concept-learning and grammar-learning
are based on feelings of familiarity and, therefore, lead to an
overestimation of awareness (Cleeremans & Elman, 1993;
Shanks & John, 1994). Accordingly, explicit re-aiming as
measured by report may be overestimating the actual explicit
re-aiming as a result of feelings of familiarity. Since the CR
group was reporting continuously, feelings of familiarity—
and thus overestimation of explicit re-aiming—should be
greater in this group as compared to the IR groups. Although
the CR group shows less variability in their reports—an
12
|
MARESCH Et Al.
indication that subjects in this group were more accustomed
to reporting—our data shows the opposite: lower explicit
re-aiming as measured by report than by explicitexcl in the
CR group and larger explicit re-aiming as measured by report
than by explicitexcl in the IR groups. Explicitexcl could also
be biased. The change in exclusion between intermittent and
final in the IR groups might reflect bias in the intermittent
exclusion. Thus, none of our measures necessarily reflects
a privileged measure of explicit re-aiming and all may have
biases. However, across both measures and both levels of re-
porting, we do not see a theory of bias that would explain all
of the differences in our results.
An alternative explanation for the dissociation between
our two methods may be that explicit re-aiming is not a
single process. This is not a new idea (Redding, Rossetti,
& Wallace, 2005; Redding & Wallace, 1996; Witt &
Proffitt, 2007). Specifically regarding visuomotor rotations
and reaching movements, McDougle and Taylor (2019) the-
orized that explicit re-aiming is a combination of calculated
and memorized re-aiming. Our results might arise because
the CR and IR groups have different levels of each of these
components and, in addition, different methods of measuring
re-aiming may reflect different combinations of these com-
ponents. For instance, in the McDougle and Taylor model,
calculated re-aiming is re-computed for each movement, a
computation that takes time. Cached re-aiming is not re-com-
puted and does not require a longer reaction time, but it is
limited to a small number of targets. The reduced reaction
times in the explicitexcl suggest that it is the computed com-
ponent that expresses itself more strongly in the explicitexcl.
However, a recent paper suggested that transition from com-
puted to cached strategy leads to an increase in the exclusion
explicit (which can be derived from the difference between
aftereffect and adaptation trials; Huberdeau, Krakauer, &
Haith,2019). Under this assumption, we generated the graph
in Figure10 that shows one (of several) ways that a specific
sensitivity of report and exclusion to the different compo-
nents could lead to the results we see. Since further evidence
is needed in order to determine which component is cached
and which is computed, we present the model with two dif-
ferent components, called comp 1 and comp 2. The specific
model that generated this figure and associated calculations
are available in the online repository (https://osf.io/6yj3u/).
Multiple levels or components of re-aiming are one possible
interpretation that could explain our results. However, much
additional work is required to determine the actual underly-
ing components and the way they are reflected in the different
measures of re-aiming.
Discrepancies in estimates of implicit adaptation mea-
sured by report and exclusion have been reported previ-
ously (Bond & Taylor, 2015; Leow et al., 2017). Leow
etal.(2017) compared reporting and non-reporting groups.
Their results are congruent with ours: the aftereffect shows
more implicit knowledge in the non-reporting group than
in the reporting group. One consideration that should be
mentioned at this point are the longer delays due to report-
ing in the CR group, which may have led to larger decay
of labile implicit in this group, and, possibly to more ex-
plicit re-aiming. However, reporting times were so short
(2–4s) that substantial decay of implicit adaptation seems
unlikely. Bond and Taylor (2015) report some conditions
where aftereffect showed less implicit knowledge than re-
port. The direction of these effects is the opposite from our
own, but the task conditions are also different from those in
our experiments. Thus, different measures of implicit adap-
tation may produce different results, and these differences
may depend on task conditions.
Differences in adaptation between reporting and non-re-
porting groups have been reported previously (Bromberg
et al., 2019; Leow etal., 2017; Taylor etal., 2014). Actual
asymptotic values differ between studies, which may be the
result of different rotation sizes (we are using 60°, while most
other studies use 30° or 45°). Larger rotation sizes may lead
to more explicit re-aiming and, consequently, to a higher
FIGURE 10 Example of how computed and cached components
could combine in order to form explicit re-aiming as measured by
exclusion and report for both consistent and intermittent reporters.
Our model assumes that each component is weighted differently
for each measure. We present one possible solution for the different
components, which assumes the computed component expresses
itself more strongly in the explicitexcl (Comp2 in the Figure): in the
CR group, report reflects 37° cached and 28° computed re-aiming. In
the IR groups, report reflects 5° cached and 27° computed re-aiming.
Report would thus primarily reflect computed re-aiming (77%) and
exclusion would primarily reflect cached re-aiming (77%).
|
13
MARESCH Et Al.
asymptote. However, it has also been proposed that the dif-
ference in asymptote does not arise from larger contributions
of explicit re-aiming, but from the additional time subjects
have when reporting, which allows for improved movement
planning (Langsdorf et al., 2020).
Some concerns regarding our data need to be addressed.
Exclusion is large in the IR groups compared with previous
findings regarding implicit adaptation in visuomotor rotation
tasks (Benson etal., 2011; Bond & Taylor,2015; Hegele &
Heuer,2010; Heuer & Hegele,2011, 2015; Leow etal.,2017;
McDougle, Bond, Bond, & Taylor,2017; Morehead, Qasim,
Qasim, Crossley, & Ivry,2015). This could be a result of dif-
ferences in task design: the use of a robotic manipulandum
rather than a tablet, passive rather than active return of the
hand, or the use of intermittent reporting. Interestingly, the
large exclusion in the IR groups actually decreases over time,
so that when it is measured several minutes after adaptation,
as part of the process dissociation procedure, subjects in all
groups have the same amount of exclusion (Figure6b). Some
have reported that implicit adaptation is composed of stable
and labile components (Hadjiosif & Smith,2013; McDougle
etal.,2015; Miyamoto etal.,2014; Morehead,2018), and it
may be that intermittent reporting leads to more labile im-
plicit, which, in part, explains the decay of the exclusion in
the IR groups (Figure6c). Since temporally labile implicit has
been shown to decay at a time constant of 15–30s (Hadjiosif
& Smith, 2013; Morehead, 2018), one could postulate that
our intermittent measures may have been affected by this
decay. This may indeed be the case in the inclusion in the
IR-EI measure, since this group performed inclusion more
than 30 s after the reporting trials (also see Figure 6a: the
IR-EI group shows less inclusion than the IR-I group). In all
other groups, reporting was almost immediately (5–10s) fol-
lowed by exclusion or inclusion trials, rendering large decays
improbable. Another consideration that may have led to dif-
ferences between intermittent and final estimates is a change
in subjects’ understanding of the instructions. However, since
instructions were consistent across groups, we would expect
the same change in understanding in all groups, which was
not the case.
We also need to consider possibility that our exclusion
measure is biased because it generalizes around the intended
aiming direction rather than around the actual movement
(Day, Roemmich, Roemmich, Taylor, & Bastian, 2016;
McDougle et al., 2017). Our results show slightly less ex-
clusion than implicitreport in the CR group, which stands in
line with generalization around the intended aiming direc-
tion. However, our data show a difference of ~5° (Figure4b
top and Figure 4c) while generalization studies show ~10°
difference between exclusion and implicitreport. Moreover our
IR groups show exactly the opposite effect of what general-
ization around the intended aiming direction would predict:
exclusion is ~15° larger than implicitreport. It is also worth
noting that our task conditions were quite different than those
used in the generalization studies. Specifically, we used eight
targets rather than one and we used online instead of end-
point feedback (Day etal.,2016; Krakauer, Pine, Ghilardi,
& Ghez,2000; McDougle etal.,2017; Schween, Taylor, &
Hegele,2018). These differences could affect the width of
the generalization function.
Finally, we want to revisit the starting point of this paper:
the fact that all methods have their shortcomings. When we
focus on quantifying subjective experience, all measures will
face interpretational complications. As such, it has been pre-
viously suggested in memory research and sequence learning
that subjects may fail to perform exclusion and inclusion re-
liably (Dodson & Johnson,1996; Graf & Komatsu,1994)—
either as a result of not understanding the instructions or
because of the complexity of the instructions—or they may
use different strategies for exclusion and inclusion leading to
distorted estimates of explicit and implicit processes (Barth,
Stahl, & Haider,2019). Failure to exclude and/or include may
then lead to biased estimates of the AI. Although great care
was taken to make instructions as clear as possible (see meth-
ods: Experimental apparatus and general procedures) and
although our exclusion and inclusion results largely match
previous reports in visuomotor rotation studies (Werner et al.,
2015, 2019), it is possible that our methods are not measuring
exactly what we think they are. Our results (and results of
others) might thus be plagued by systematic errors. The field
needs more comparisons across measures, such as the ones
in this paper, to make progress in overcoming this method-
ological gap.
In summary, our central finding is that different ap-
proaches to measuring explicit re-aiming and implicit ad-
aptation lead to different results. Thus, they may reflect
different components of the explicit and implicit processes.
We speculate that explicit re-aiming is not a single pro-
cess but made up of different components that contribute
differently to measures of re-aiming depending on the task
conditions. Methodological and theoretical developments
are needed to fully understanding how these components
combine to produce reaching movements. We need a new,
updated model of the explicit and implicit processes during
adaptation.
ACKNOWLEDGEMENTS
The authors thank Sivan Hazan for technical support and help
with the data collection. Furthermore, the authors also thank
Dr. Shlomi Haar, Dr. Matthias Hegele, Dr. Liad Mudrik, and
Dr. Ronen Segev for comments and discussions about our
data.
This work is part of the PACE project (itn-pace.eu), which
received funding from the European Union's Horizon 2020 re-
search and innovation program under the Marie Skłodowska-
Curie grant agreement No 642961.
14
|
MARESCH Et Al.
CONFLICT OF INTEREST
No conflicts of interest, financial or otherwise, are declared
by the authors.
AUTHOR CONTRIBUTIONS
J.M., O.D., and S.W. contributed to the idea and design of the
experiment. J.M. carried out the data collection, processed
the experimental data, drafted the manuscript, and designed
the figures. O.D. performed the Bayesian statistical analysis.
All authors discussed the results and contributed to the final
manuscript.
PEER REVIEW
The peer review history for this article is available at https://
publo ns.com/publo n/10.1111/ejn.14945
DATA AVAILABILITY STATEMENT
All data and MATLAB scripts used for analysis and plotting
of the figures are available in the online repository (https://
osf.io/6yj3u/).
ORCID
Jana Maresch https://orcid.org/0000-0002-0758-5043
Susen Werner https://orcid.org/0000-0001-9664-4617
Opher Donchin https://orcid.org/0000-0003-0963-4467
REFERENCES
Barth, M., Stahl, C., & Haider, H. (2019). Assumptions of the pro-
cess-dissociation procedure are violated in implicit sequence
learning. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 45(4), 641. https://doi.org/10.1037/
xlm00 00614
Benson, B. L., Anguera, J. A., & Seidler, R. D. (2011). A spatial ex-
plicit strategy reduces error but interferes with sensorimotor adap-
tation. Journal of Neurophysiology, 105(6), 2843–2851. https://doi.
org/10.1152/jn.00002.2011
Bond, K. M., & Taylor, J. A. (2015). Flexible explicit but rigid im-
plicit learning in a visuomotor adaptation task. Journal of
Neurophysiology, 113(10), 3836–3849. https://doi.org/10.1152/
jn.00009.2015
Bromberg, Z., Donchin, O., & Haar, S. (2019). Eye movements during
visuomotor adaptation represent only part of the explicit learning.
bioRxiv, 724864. https://doi.org/10.1523/ENEURO.0308-19.2019
Cleeremans, A., Destrebecqz, A., & Boyer, M. (1998). Implicit learning:
News from the front. Trends in Cognitive Sciences, 2(10), 406–416.
https://doi.org/10.1016/S1364 -6613(98)01232 -7
Cleeremans, A., & Elman, J. (1993). Mechanisms of implicit learning:
Connectionist models of sequence processing, Cambridge, MA:
MIT Press.
Day, K. A., Roemmich, R. T., Taylor, J. A., & Bastian, A. J. (2016).
Visuomotor learning generalizes around the intended movement.
eneuro, 3(2). https://doi.org/10.1523/ENEURO.0005-16.2016
Dodson, C. S., & Johnson, M. K. (1996). Some problems with
the process-dissociation approach to memory. Journal of
Experimental Psychology: General, 125(2), 181. https://doi.
org/10.1037/0096-3445.125.2.181
Eriksen, C. W. (1960). Discrimination and learning without awareness:
A methodological survey and evaluation. Psychological Review,
67(5), 279. https://doi.org/10.1037/h0041622
Fernandez-Ruiz, J., Wong, W., Armstrong, I. T., & Flanagan, J. R.
(2011). Relation between reaction time and reach errors during vi-
suomotor adaptation. Behavioural Brain Research, 219(1), 8–14.
https://doi.org/10.1016/j.bbr.2010.11.060
Graf, P., & Komatsu, S.-I. (1994). Process dissociation procedure:
Handle with caution! European Journal of Cognitive Psychology,
6(2), 113–129. https://doi.org/10.1080/09541 44940 8520139
Hadjiosif, A., & Smith, M. (2013). Savings is restricted to the tem-
porally labile component of motor adaptation. Translational and
Computational Motor Control, 409–419. https://doi.org/10.1523/
JNEUR OSCI.1046-15.2015
Haith, A. M., Huberdeau, D. M., & Krakauer, J. W. (2015). The influ-
ence of movement preparation time on the expression of visuomotor
learning and savings. Journal of Neuroscience, 35(13), 5109–5117.
https://doi.org/10.1523/JNEUR OSCI.3869-14.2015
Hegele, M., & Heuer, H. (2010). Implicit and explicit components of dual
adaptation to visuomotor rotations. Consciousness and Cognition,
19(4), 906–917. https://doi.org/10.1016/j.concog.2010.05.005
Heuer, H., & Hegele, M. (2011). Generalization of implicit and ex-
plicit adjustments to visuomotor rotations across the workspace
in younger and older adults. Journal of Neurophysiology, 106(4),
2078–2085. https://doi.org/10.1152/jn.00043.2011
Heuer, H., & Hegele, M. (2015). Explicit and implicit components
of visuo-motor adaptation: An analysis of individual differ-
ences. Consciousness and Cognition, 33, 156–169. https://doi.
org/10.1016/j.concog.2014.12.013
Huberdeau, D. M., Krakauer, J. W., & Haith, A. M. (2015). Dual-
process decomposition in human sensorimotor adaptation. Current
Opinion in Neurobiology, 33, 71–77. https://doi.org/10.1016/j.
conb.2015.03.003
Huberdeau, D. M., Krakauer, J. W., & Haith, A. M. (2019). Practice
induces a qualitative change in the memory representation for visuo-
motor learning. Journal of Neurophysiology, 122(3), 1050–1059.
https://doi.org/10.1152/jn.00830.2018
Hwang, E. J., Smith, M. A., & Shadmehr, R. (2006). Dissociable effects
of the implicit and explicit memory systems on learning control of
reaching. Experimental Brain Research, 173(3), 425–437. https://
doi.org/10.1007/s0022 1-006-0391-0
Jacoby, L. L. (1991). A process dissociation framework: Separating automatic
from intentional uses of memory. Journal of Memory and Language,
30(5), 513–541. https://doi.org/10.1016/0749-596X(91)90025 -F
Kim, H. E., Parvin, D. E., & Ivry, R. B. (2019). The influence of task
outcome on implicit motor learning. eLife, 8, e39882. https://doi.
org/10.7554/eLife.39882
Krakauer, J. W., Pine, Z. M., Ghilardi, M. F., & Ghez, C. (2000).
Learning of visuomotor transformations for vectorial planning of
reaching trajectories. Journal of Neuroscience, 20(23), 8916–8924.
https://doi.org/10.1523/JNEUR OSCI.20-23-08916.2000
Kruschke, J. K. (2010). Bayesian data analysis. Wiley Interdisciplinary
Reviews: Cognitive Science, 1(5), 658–676.
Kruschke, J. (2014). Doing Bayesian data analysis: A tutorial with
R, JAGS, and Stan, London, UK: . Academic Press. https://doi.
org/10.1016/B978-0-12-405888-0.09999-2
Langsdorf, L., Maresch, J., Hegele, M., McDougle, S. D., &
Schween, R. (2020). Prolonged reaction times help to eliminate
residual errors in visuomotor adaptation. bioRxiv. https://doi.
org/10.1101/2019.12.26.888941
|
15
MARESCH Et Al.
Leow, L.-A., Gunn, R., Marinovic, W., & Carroll, T. J. (2017). Estimating
the implicit component of visuomotor rotation learning by con-
straining movement preparation time. Journal of Neurophysiology,
118(2), 666–676. https://doi.org/10.1152/jn.00834.2016
McDougle, S. D., Bond, K. M., & Taylor, J. A. (2015). Explicit and im-
plicit processes constitute the fast and slow processes of sensorim-
otor learning. Journal of Neuroscience, 35(26), 9568–9579. https://
doi.org/10.1523/JNEUR OSCI.5061-14.2015
McDougle, S. D., Bond, K. M., & Taylor, J. A. (2017). Implications
of plan-based generalization in sensorimotor adaptation. Journal
of Neurophysiology, 118(1), 383–393. https://doi.org/10.1152/
jn.00974.2016
McDougle, S. D., & Taylor, J. A. (2019). Dissociable cognitive strate-
gies for sensorimotor learning. Nature Communications, 10(1), 1–3.
https://doi.org/10.1038/s4146 7-018-07941 -0
Miyamoto, Y. R., Wang, S., Brennan, A. E., & Smith, M. A. (2014).
Distinct forms of implicit learning that respond differentially to
performance errors and sensory prediction errors. Translational and
Computational Motor Control (TCMC). https://doi.org/10.1038/
s4159 8-018-34598 -y
Morehead, J. R. (2018). Implicit visuomotor adaptation has temporally
stable and labile components. 28th Annual Meeting of the NCM
Society.
Morehead, J. R., Qasim, S. E., Crossley, M. J., & Ivry, R. (2015).
Savings upon re-aiming in visuomotor adaptation. Journal of
Neuroscience, 35(42), 14386–14396. https://doi.org/10.1523/
JNEUR OSCI.1046-15.2015
Morehead, J. R., Taylor, J. A., Parvin, D. E., & Ivry, R. B. (2017).
Characteristics of implicit sensorimotor adaptation revealed by
task-irrelevant clamped feedback. Journal of Cognitive Neuroscience,
29(6), 1061–1074. https://doi.org/10.1162/jocn_a_01108
Newell, B. R., & Shanks, D. R. (2014). Unconscious influences on de-
cision making: A critical review. Behavioral and Brain Sciences,
37(1), 1–19. https://doi.org/10.1017/S0140 525X1 2003214
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know:
Verbal reports on mental processes. Psychological Review, 84(3),
231. https://doi.org/10.1037/0033-295X.84.3.231
Redding, G. M., Rossetti, Y., & Wallace, B. (2005). Applications of
prism adaptation: A tutorial in theory and method. Neuroscience &
Biobehavioral Reviews, 29(3), 431–444. https://doi.org/10.1016/j.
neubi orev.2004.12.004
Redding, G. M., & Wallace, B. (1996). Adaptive spatial alignment
and strategic perceptual-motor control. Journal of Experimental
Psychology: Human Perception and Performance, 22(2), 379.
https://doi.org/10.1037/0096-1523.22.2.379
Schween, R., Taylor, J. A., & Hegele, M. (2018). Plan-based gener-
alization shapes local implicit adaptation to opposing visuomotor
transformations. Journal of Neurophysiology, 120(6), 2775–2787.
https://doi.org/10.1152/jn.00451.2018
Shanks, D. R., & John, M. F. S. (1994). Characteristics of dissociable
human learning systems. Behavioral and Brain Sciences, 17(3),
367–395. https://doi.org/10.1017/S0140 525X0 0035032
Shanks, D. R., Rowland, L. A., & Ranger, M. S. (2005). Attentional
load and implicit sequence learning. Psychological Research
Psychologische Forschung, 69(5–6), 369–382. https://doi.
org/10.3389/fpsyg.2014.00465
Taylor, J. A., & Ivry, R. B. (2011). Flexible cognitive strategies during
motor learning. PLoS Computational Biology, 7(3), e1001096.
https://doi.org/10.1371/journ al.pcbi.1001096
Taylor, J. A., & Ivry, R. B. (2013). Implicit and explicit processes in
motor learning (pp. 63–87). Cambridge, MA: MIT press. https://
doi.org/10.7551/mitpr ess/97802 62018 555.003.0003
Taylor, J. A., Krakauer, J. W., & Ivry, R. B. (2014). Explicit and implicit
contributions to learning in a sensorimotor adaptation task. Journal
of Neuroscience, 34(8), 3023–3032. https://doi.org/10.1523/JNEUR
OSCI.3619-13.2014
Timmermans, B., & Cleeremans, A. (2015). How can we measure
awareness? An overview of current methods. Behavioural Methods
in Consciousness Research, 21–46. https://doi.org/10.1093/acpro
f:oso/97801 99688 890.003.0003
Werner, S., Strueder, H. K., & Donchin, O. (2019). Intermanual transfer
of visuomotor adaptation is related to awareness. bioRxiv, 617407.
https://doi.org/10.1371/journ al.pone.0220748
Werner, S., van Aken, B. C., Hulst, T., Frens, M. A., van der Geest, J.
N., Struder, H. K., & Donchin, O. (2015). Awareness of sensorimo-
tor adaptation to visual rotations of different size. PLoS One, 10(4),
e0123321. https://doi.org/10.1371/journ al.pone.0123321
Witt, J. K., & Proffitt, D. R. (2007). Perceived slant: A dissociation be-
tween perception and action. Perception, 36(2), 249–257. https://
doi.org/10.1068/p5449
How to cite this article: Maresch J, Werner S,
Donchin O. Methods matter: Your measures of
explicit and implicit processes in visuomotor
adaptation affect your results. Eur J Neurosci.
2020;00:1–15. https://doi.org/10.1111/ejn.14945