learning an environment–actor coordination skill: visuomotor transformation and coherency of...
TRANSCRIPT
RESEARCH ARTICLE
Learning an environment–actor coordination skill: visuomotortransformation and coherency of perceptual structure
Young U. Ryu Æ John J. Buchanan
Received: 13 April 2007 / Accepted: 6 May 2009 / Published online: 24 May 2009
� Springer-Verlag 2009
Abstract The coordination dynamics of learning were
examined in a visuomotor tracking task. Participants pro-
duced rhythmic elbow flexion–extension motions to learn a
visually defined 90� relative phase tracking pattern with an
external sinusoidal signal. There were two visuomotor
transformation groups, a correct feedback group and a
mirrored feedback group with feedback representing the
elbow’s motion transformed by 180�. In Experiment 1, the
to-be-tracked signal and the participant’s motion signal
were superimposed within a single window display. In
Experiment 2, the to-be-tracked signal and participant’s
signal were presented in separate windows. Before day 1
practice and 24 h after day 2 practice, participants
attempted visually defined 0�, 45�, 90�, 135�, and 180�relative phase tracking patterns either with or without
visual feedback of the arm’s motion. Before practice, only
the 0� and 180� tracking patterns were stable. Practice led
to a decrease in phase error toward the required 90� relative
phase pattern with a corresponding increase in stability in
both the experiments. No effect of visual transformation on
performance emerged during practice in the single window
task, but did emerge in the two window task. The one
window training facilitated transfer to the four unpracticed
relative phase patterns, whereas the two window training
display only facilitated transfer performance to a single
unpracticed relative phase pattern. These findings suggest
that the perceptual structure determined the degree of
learning and transfer and interacted with the visuomotor
transformation. The present findings are discussed with
reference to how the visual display constrains the coher-
ency of independent signals with regard to learning and
transfer and the role of perceptual discrimination processes
linked to transfer.
Keywords Coordination � Visuomotor tracking �Perception–action coupling � Feedback � Display �Coherency � Perceptual motor learning
Introduction
The ability to coordinate the movement of a limb or limbs
with an external event is essential for the performance of
many everyday activities. For instance, visuomotor tracking
is an essential skill for catching a ball, for driving a car, or for
dancing with a partner. For experienced individuals, the
actions just listed may seem trivial. Of course, experience
comes from practice with a task and practice is associated
with a change in behavior referred to as learning. The current
experiment attempted to identify changes in the coordination
dynamics of a practiced visuomotor tracking pattern in order
to identify the learning processes that underlie the develop-
ment of visuomotor tracking skills in general.
Extensive research has shown that in-phase (relative
phase / = 0�) coordination patterns in bimanual (Kelso
1984; Kelso et al. 1986), single limb (Kelso et al. 1991),
and visuomotor tracking (Wimmers et al. 1992; Peper
and Beek 1998) tasks are more stable than anti-phase
Y. U. Ryu � J. J. Buchanan
Human Performance Laboratories,
Department of Health and Kinesiology,
Texas A&M University, College Station, USA
Present Address:Y. U. Ryu (&)
Motor Control Development Laboratory,
Department of Biokinesiology and Physical Therapy,
University of Southern California,
Los Angeles, CA 90033, USA
e-mail: [email protected]
123
Exp Brain Res (2009) 196:279–293
DOI 10.1007/s00221-009-1847-9
(/ = 180�) coordination patterns. Thus, a series of
experiments employing quite different experimental tasks,
bimanual, single limb, and visuomotor tracking, reveals
shared dynamic principles that constrain coordination. The
differential stability between anti-phase and in-phase
coordination patterns has been modeled at the level of the
relative phase dynamics and at the level of the individual
component actions and their coupling (Haken et al. 1985;
Schoner et al. 1986). A key feature of the modeling was
that the in-phase and anti-phase patterns were interpreted
as stable fixed-point attractors of the collective variable
relative phase (/). Between these stable attractive states of
the model, a repeller at the relative phase value of 90�exists and will push the system towards the attractor at
either 0� or 180�. Both bimanual coordination research
(Zanone and Kelso 1992; Lee et al. 1995) and visuomotor
tracking research (Wilson et al. 2005) have shown that
when participants attempted a 90� relative phase pattern
there was a tendency to be drawn to either in-phase or anti-
phase. This implies that the ability to produce a stable 90�relative phase pattern in many skills must be learned
(Zanone and Kelso 1992; Lee et al. 1995; Buchanan 2004).
To examine the learning of a 90� relative phase pattern
in a bimanual task, Zanone and Kelso (1992) first per-
formed a pre-practice scan of the relative phase landscape
by requiring participants to produce 13 different relative
phases from 0� to 180� in 15� increments. The scanning
runs revealed that attempts at producing a 90� relative
phase pattern were attracted to either 0� or 180�. This
finding was consistent with the model of the relative phase
dynamics developed by Haken et al. (1985). After several
days of practice, post-practice scanning runs revealed that
the practiced relative phase pattern of 90� was produced
without attraction towards 0� or 180�. Thus, practice
resulted in the formation of a stable attractor at the relative
phase value of 90�. Further, the post-practice scanning
results demonstrated that practice of the 90� relative phase
pattern affected not only the required coordination pattern,
but also the entire coordination landscape (Schoner et al.
1992; Zanone and Kelso 1992, 1997). The ability to learn a
specific visuomotor tracking pattern, such as a 90� relative
phase between a limb and external signal, however, has not
been investigated from any theoretical viewpoint to our
knowledge. A primary goal of the current experiment was
to determine if learning a 90� visuomotor tracking pattern
shared similar dynamics with the learning of a 90� relative
phase bimanual pattern. Specifically, the experiment was
designed to determine (1) if a novel tracking pattern could
be learned and produced with and without visual feedback
of the arm’s motion, and (2) whether or not learning a
novel tracking pattern influenced performance (i.e., sta-
bility and accuracy) of the intrinsic visuomotor tracking
patterns of 0� and 180� (Wimmers et al. 1992).
Wilson et al. (2005) examined unimanual visuomotor
tracking when participants had to establish a 0�, 90� or
180� phase relationship between a moving cursor and arm
motion represented as a moving cursor in a visual display.
As expected, the 90� tracking pattern was not stable
whereas the 0� and 180� were stable, and Wilson et al.
concluded that this pattern of performance was consistent
with the Haken et al. (1985; Schoner et al. 1986) model of
bimanual coordination. This indicates that producing a
stable 90� relative phase visuomotor tracking pattern would
require practice (Liao and Jagacinski 2000). What might
influence the ability to learn a 90� visuomotor tracking
pattern? Research has demonstrated that transformed visual
feedback of the arm’s motion can influence the accuracy
and stability of visuomotor tracking. For example,
Roerdink et al. (2005) required participants to track an
external signal under three feedback conditions: no feed-
back, correct feedback (feedback representing the actual
wrist motion), and mirrored feedback (wrist motion trans-
formed by 180�). Results revealed that the accuracy and
stability of tracking was the worst in the no feedback
condition, suggesting the importance of feedback in
tracking performance. The transformed visual feedback did
not influence the in-phase pattern, but stabilized the less
stable anti-phase pattern when the visual transformation
produced a perceived in-phase pattern (Bogaerts et al.
2003; Roerdink et al. 2005; Wilson et al. 2005).
The transformation results in the two visual signals
moving in the same direction and a coherently grouped
visual motion structure is created that helps to stabilize the
anti-phase pattern. This suggests that coherency in the
visual structure can play an important role in detecting and
recognizing a particular relative phase pattern (Bogaerts
et al. 2003; Roerdink et al. 2005; Bingham et al. 1999; Zaal
et al. 2000), and therefore may play an important role in
learning a novel relative phase tracking pattern and transfer
of skill after learning. However, it must be noted that while
the presentation of 90� and 180� tracking patterns as a 0�pattern improves performance, the stability of the 90� and
180� patterns does not match the 0� pattern (Wilson et al.
2005). The coherency created by the visual display does
not completely override the role of limb proprioceptive
information. With regard to learning and transfer of visu-
omotor tracking skills, it is necessary to understand how
transformed feedback interacts with the coherency of the
visual display in order to understand the interaction
between visual and proprioceptive feedback. Since the 90�pattern is already an unstable pattern, transformation of the
arm’s motion by 180� will not result in a more stable
representation of the action as does the transformation of
the arm’s motion during anti-phase tracking. Thus, both
correct and mirrored displays will represent an unstable
relative phase pattern.
280 Exp Brain Res (2009) 196:279–293
123
Another factor that may influence learning and transfer
of visuomotor tracking skills is the distance between the
signals. Research has demonstrated that tracking perfor-
mance is most stable when the to-be-tracked signal in the
form of an open square and a cursor representing arm
motion are superimposed. When the two signals are not
superimposed, performance error increases and grows as
the gap between the to-be-tracked signal and arm’s signal
increases (Reed et al. 2003). This finding indicates that the
coherency created by transformed visual feedback in
tracking tasks could be influenced by signal separation
distance. In other words, the larger the separation in the
signals, the less of an impact on the performance
improvements of 180� and 90� patterns presented as 0�patterns. This possible interaction between transformed
visual feedback and signal separation distance and its
impact on learning and transfer of visuomotor tracking
skills was also examined in the current study.
Experiment 1 addressed the learning of a 90� relative
phase visuomotor tracking pattern under a correct and
mirrored (arm’s motion transformed by 180�) presentation
of the arm’s motion in the visual display. In this experi-
ment, the two signals were superimposed and presented in
the same window. It is predicted that both the correct and
mirrored representations will allow the 90� relative phase
pattern to be learned and reproduced with or without visual
feedback of the arm’s motion in the display. However,
when combining the bimanual work and visuomotor
transformation work, two possible outcomes may be
observed after training: First, since transformed feedback
can stabilize a 180� tracking pattern, practicing a 90� pat-
tern with transformed feedback may actually lead to an
increase in the stability of the 180� pattern when produced
with transformed feedback. This finding would be in con-
tradiction to the bimanual work of Zanone and Kelso
(1992). Second, practicing the 90� pattern with correct
feedback will destabilize the 180� tracking pattern when
produced with correct feedback. This finding would be
consistent with the bimanual work and the visuomotor
transformation work since the 180� tracking pattern with
correct feedback typically emerges as the least stable pat-
tern under correct and transformed conditions. Experiment
2 addressed the interaction between signal-distance in the
training display and the visual coherency created by the
transformed visual feedback. If the transformation effect on
tracking performance is independent of the distance
between signals, presenting the tracked signal and arm’s
signal in separate displays will not influence learning or
transfer. However, if the transformation effect is linked to
the superposition of the signals, visual coherency will be
reduced in the mirrored condition across separate display
windows, and result in decreased performance gains and
reduced transfer compared to the single window display.
Experiment 1
Materials and methods
Subjects
Twelve students (6 males and 6 females, mean age 22.2
years) volunteered to participate in the study. The IRB
board of Texas A&M University approved the experi-
mental protocol and consent form and all participants
voluntarily signed the consent form prior to the experiment.
All participants were self-reported right-handers and had
normal or corrected-to-normal vision.
Apparatus
Participants sat on a height-adjustable chair and faced a
computer monitor at a distance of 0.76 m. The partici-
pant’s right elbow was comfortably placed on a table in
a supine position, with the longitudinal axis of their
upper arm at about 45� to the surface of the table.
Participants wore a wrist orthosis to secure their wrist
joint in parallel with the longitudinal axis of the forearm.
In the visuomotor tracking literature, many experiments
are performed without vision of the moving limb (e.g.,
Bogaerts et al. 2003; Ceux et al. 2003; Roerdink et al.
2005; Wilson et al. 2005; Byblow et al. 1995) and with
vision of the moving limb (e.g., Buekers et al. 2000;
Carnahan et al. 1996; Foulkes and Miall 2000; Liao and
Jagacinski 2000). We chose to block the view of the
participant’s arm with a wall (0.84 m 9 0.48 m) moun-
ted on the table between the participants’ head and arm
(Fig. 1a). A computer monitor was used to display the
external tracking signal, the participant’s own tracking
motion, and an angle–angle plot of the tracking signal
versus the limb’s motion. The sinusoidal tracking signal
was produced with a Tektronix� AFG320 arbitrary
function generator. The oscillation frequency was set at
0.8 Hz with a total of 8 cycles in a trial. The oscillation
frequency of 0.8 Hz was chosen because previous
tracking studies have shown that multi-stability was
observed around 0.8 Hz (e.g., 0.67 Hz in Ceux et al.
2003; 1 Hz in Buekers et al. 2000). An OPTOTRAK�
3020 3D camera system recorded the position of infrared
light emitting diodes (IREDs). Three IREDs were
mounted as follows: (1) attached to a dowel held in the
hand, (2) lateral epicondyle of the elbow, and (3) on the
table corresponding to the dowel’s height when the
participant’s hand was extended to maximum extension
along the table (Fig. 1a). The participant’s elbow angular
motion was represented by the change in the angle hdefined by the dowel’s position with regard to the IRED
mounted on the table (Fig. 1a).
Exp Brain Res (2009) 196:279–293 281
123
Protocol
Each participant was randomly assigned to an experimental
group based on the presentation of the trace representing
the arm’s motion in the training display (Fig. 1b): (1) a
correct feedback group (COR, 3 males and 3 females) with
visual feedback of the elbow’s rotation representing the
movement as made, and (2) a mirrored feedback group
(MIR, 3 males and 3 females) with the elbow’s angle signal
was transformed by 180�. For the correct feedback group,
rotation of the forearm upward (flexion) produced an
upward motion of the elbow signal on the screen, and for
the mirrored feedback group, rotation of the forearm
upward (flexion) produced a downward motion of the
elbow signal on the screen. The angle h approached 0�when the elbow was fully extended. The angle increased
positively when flexing the elbow in the correct feedback
condition. In the mirrored feedback condition, the angle
decreased negatively when flexing the elbow since the
online visual feedback was multiplied by -1 during a trial.
The required joint amplitude corresponding to the external
sinusoidal wave was 61�. Participants were required to
learn to rhythmically coordinate the flexion–extension
movement of the elbow with the external signal so the
elbow’s motion lagged the tracked signal at a 90� relative
phase relationship (Fig. 1b). The external signal and the
online visual feedback of the elbow’s rotation were both
presented as sine waves that overlapped in the workspace.1
The workspace was 14 cm (length) 9 4.5 cm (height) and
presented two peak-to-peak cycles of the external signal.
Participants were given four tracking trials to familiarize
themselves with the production of the joint motion dis-
played on the computer screen. Participants were asked to
track the external signal with both a 0� relative phase
(2 trials) and a 180� relative phase (2 trials) visuomotor
pattern. For these trials, the oscillation frequency was set at
0.5 Hz with a total of 8 cycles in a trial. This oscillation
frequency was slower than the one used in the experimental
sessions (0.8 Hz) to reduce the practice effect on tracking
before the experimental scanning run. Participants were
informed of the visuomotor transformation relationship
during these familiarization trials.
Scanning sessions
A pre-practice scanning session was conducted prior to the
practice of the 90� relative phase pattern on day 1, and a
post-practice scanning session was conducted on day 3
without any preceding practice trials. The scanning ses-
sions consisted of 30 trials for 5 different relative phase
patterns: 0�, 45�, 90�, 135�, and 180� with each relative
phase pattern defined by the relationship between the
elbow signal and the external signal on the monitor
(Fig. 1c). The scanning sessions consisted of a visual and
proprioceptive feedback condition (VPF) and a proprio-
ceptive feedback condition only (PF). In the VPF condi-
tion, the elbow angle trajectory was plotted along with the
signal to be tracked, and in the PF condition, the elbow
angle trajectory was not plotted with the tracking signal.
The VPF condition (3 trials) was conducted first and the PF
condition (3 trials) was conducted second in every scan-
ning session.
Before performing each relative phase pattern, the
required relative phase pattern was shown to a participant.
1
2 3θ
Wall
0.76 m
Monitor
A
0 45 ° 90° 135° 180°C
BRequired pattern Visuomotor transformation
Arm
flex.
ext.
COR MIR
time
°
Fig. 1 a Experimental setup showing the participant relative to the
wall blocking vision of the arm and the monitor displaying the
signals. b The plot on the left portrays the representation of the to-be-
learned 90� pattern as viewed by the participants. The black curverepresents the external signal and the gray curve represents the signal
that a participant is supposed to produce. The middle plot and plot on
the right display correct (COR) and mirrored (MIR) arm motion
(dashed curves). c The five target patterns used in the pre- and post-
practice scans are shown
1 Typically, the signal to be tracked and the arm’s motion in
visuomotor tracking tasks are presented in the form of cursors or lines
continuously moving on a monitor. We chose to present the signals as
continuous sine waves. Our pre-practice scanning procedure revealed
intrinsic dynamics consistent with other tasks that have used
continuous cursors to examine tracking (Roerdink et al. 2005; Wilson
et al. 2005) or discrete light flashes to examine learning in bimanual
tasks (Amazeen 2002; Zanone and Kelso 1992, 1997). This suggests
that our display is not introducing oddities inconsistent with other
visual presentation techniques and conveys similar information to the
participant just in a different format.
282 Exp Brain Res (2009) 196:279–293
123
The external signal was represented as a black sine wave
while the signal that the participants were supposed to
reproduce with elbow motion was represented with as a red
sine wave (Fig. 1c). The assigned participant’s signal lag-
ged the external signal in the 45�, 90�, and 135� relative
phase patterns. Participants were instructed to watch the
pattern represented by the two sine waves and to try to
reproduce the pattern when they performed the scanning
trials. After watching one of the relative phase patterns,
participants immediately performed three consecutive trials
for the assigned relative phase pattern. The five relative
phase patterns were performed in a random order.
Practice sessions
The practice session was initiated after the scanning trials.
Practice sessions 1 (day 1) and 2 (day 2) consisted of 6
trials per block that yielded 36 practice trials per session (a
total of 72 practice trials). Participants were required to
coordinate rhythmic elbow flexion–extension movements
to a continuous sinusoidal wave at a 90� relative phase
relationship. The elbow angle trajectory was provided
concurrently with the external sinusoidal signal in every
trial. After an odd numbered trial, both the re-play of the
performed trial and its angle–angle plot (tracking signal vs.
elbow angle) were provided as terminal feedback. Partici-
pants were informed that a circle in the angle–angle plot
represented the required 90� relative phase relationship
between the tracking signal and the participant’s elbow
motion. A variety of tasks have shown that the angle–angle
display as a form of terminal feedback acts as a beneficial
form of KR (knowledge of result) to represent performance
changes associated with the learning of relative phase
patterns (Lee et al. 1995; Buchanan et al. 2007).2
Data analysis
Prior to any data analysis, the 3D IRED trajectories were
filtered with a dual-pass Butterworth Filter with a cutoff
frequency of 10 Hz. The filtered data were used to compute
the elbow joint angle. The first cycle of motion was
dropped and considered as an adaptation phase so that 7
cycles were analyzed from each trial.
A continuous relative phase was computed to charac-
terize the spatiotemporal tracking relationship between the
external signal and the elbow angle trajectory. For both the
tracking signal (hexternal) and elbow angle (helbow), indi-
vidual phase angles were computed for every sampled
point i in a trial, hi ¼ tan�1½ðdxi=dtÞ=xi�, with xi the nor-
malized position and dxi/dt the normalized instantaneous
velocity. For each sampled point i in a trial, the continuous
relative phase was computed as /i = hexternal - helbow.
Mean resultant vectors were computed from the observed
unit vectors (x, y) of the /i values for each scanning trial
and practice trial (Burgess-Limerick et al. 1991). The mean
resultant vectors were used to compute a mean phase angle
(/obs) and a magnitude of circular variance (uniformity). A
value of /obs = 0� represented in-phase tracking and a
value of /obs = 180� represented anti-phase tracking. In
the pre- and post-scanning procedures, there were five
required relative phase values, /req = 0�, -45�, -90�,
-135�, and 180�. The target relative phase in the practice
condition was /req = -90�. For each scanning trial and
practice trial, an absolute phase error (/AE = |/req - /obs|)
was computed and used to evaluate visuomotor tracking
accuracy as a function of practice. The circular variance
falls within the range of 0–1 with 1 representing perfect
uniformity. In order to submit the circular variance data to
inferential tests based on standard normal theory (Mardia
1972), the circular variance score was transformed to the
range 0 to ? as follows, s0 = -2 logn(1 - S0)0.5. S0
denotes the measure of circular variance on the interval
[0, 1] with s0 the transformed circular variance (TCV)
measure that was submitted to statistical analysis. This
TCV measure was treated as an estimate of the variability
in tracking performance. An increase in the uniformity
measure with practice will indicate an increase in the
stability of tracking performance.
Statistical analysis
The ANOVAs used to analyze the pre-practice, practice,
and post-practice data each contained different factors and
have been reported in each section separately. Duncan’s
multiple range tests (p \ 0.05) were used to analyze all
post hoc main effects, and simple main effect tests
(p \ 0.05) were performed to reveal differences when
significant interactions were present. Partial eta-squared
values (g2p) were computed and have been reported with
every significant F value.
Results
Pre-practice scanning
The individual trial relative phase means (/obs) from the
pre-practice scanning session have been displayed in
Fig. 2. With visual and proprioceptive feedback, the most
2 The bimanual learning work of Zanone and Kelso (1992) utilized
cycle-to-cycle relative phase plots with both a graphical and
numerical representation of the to-be-learned pattern as a means of
KR. We know of no studies that have directly compared angle–angle
plots with cycle-to-cycle relative phase plots to determine the most
relevant form of display with regard to learning such coordination
patterns. The literature clearly shows that both forms of KR are
beneficial.
Exp Brain Res (2009) 196:279–293 283
123
clustering occurred for the 0� relative phase pattern fol-
lowed by the 180� relative phase pattern for both the cor-
rect and mirrored feedback groups (Fig. 2a). With
proprioceptive feedback only, a dense clustering of
observed mean relative phases only emerged in the correct
feedback group for the 0� tracking pattern (Fig. 2b).
The absolute phase error values (/AE) and TCV values
from the pre-practice scanning sessions were analyzed in a
2 Visuomotor Transformation (COR, MIR) 9 2 Feedback
(VPF, PF) 9 5 Pattern (0�, 45�, 90�, 135�, 180�) ANOVA
with the last two variables repeated measures. A significant
main effect of Pattern was found in the /AE data
(F(4,40) = 4.49, p \ 0.01, g2p ¼ 0:31). Post hoc tests
revealed that the /AE values were significantly smaller
when performing the 0�, 45�, and 180� patterns compared
to the 90� and 135� tracking patterns, with the largest /AE
associated with the 135� pattern (Fig. 3a). The PF condi-
tion (/AE = 70.7�) was characterized by significantly lar-
ger phase error than the VPF condition (/AE = 29.6�)
(F(1,10) = 14.4, p \ 0.01, g2p ¼ 0:59). The analysis of the
/AE data also revealed a significant Visuomotor Trans-
formation 9 Feedback interaction (F(1,10) = 5.21,
p \ 0.05, g2p ¼ 0:32). Post hoc tests revealed two main
findings. First, the tracking performance of the mirrored
feedback group (/AE = 86.8�) was significantly less
accurate than the correct feedback group (/AE = 22.5�)
when tracking in the PF condition. Second, tracking per-
formance in the mirrored feedback group was significantly
less accurate in the PF condition (/AE = 86.8�) compared
to the VPF condition (/AE = 54.6�).
A significant main effect of Pattern was found in the
TCV data (F(4,40) = 12.61, p \ 0.01, g2p ¼ 0:56) (Fig. 3b).
Post hoc tests showed that the 0� pattern was more stable
(larger TCV values) than the other four patterns. The 180�
A
0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180CORMIR
B
RP pattern (deg)0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180
Fig. 2 The individual trial observed mean relative phase values
(/obs) from the pre-practice scanning trials are plotted for the VPF (a)
and PF (b) conditions as a function of required relative phase: 0�,
-45�, -90�, -135� and ±180�. The negative values represent that
the task required elbow motion to lag the external signal in both the
correct and mirrored conditions
A
0 45 90 135 180
Abs
. Err
or (
deg)
0
20
40
60
80
100
B
TC
V
0
1
2
3
4
5
C
RP pattern (deg)
TC
V
0
1
2
3
4
5
CORMIR
0 45 90 135 180
0 45 90 135 180
Fig. 3 Pre-practice scanning data. a Mean absolute phase error (/AE)
and b mean TCV values are plotted as a function of required target
pattern. c Mean TCV values are plotted as a function of visuomotor
transformation and required target pattern. The error bars represent
the standard error of the between subject variability
284 Exp Brain Res (2009) 196:279–293
123
pattern was more stable than the 90� and 135� patterns with
no difference in stability between the 45�, 90�, and 135�patterns. The analysis of the TCV data also revealed a
significant Visuomotor Transformation 9 Pattern interac-
tion (F(4,40) = 3.73, p \ 0.05, g2p ¼ 0:27) (Fig. 3c). Tests
of the interaction revealed that within the mirrored feed-
back group the 0� and 180� tracking patterns were more
stable than the other three patterns, and within the correct
feedback group the 0� pattern was more stable than the
other four patterns. The correct feedback group tracked at
45� and 90� with more stability than the mirrored feedback
group, while the mirrored feedback group tracked at 180�with more stability than the correct feedback group.
Overall, the 0� and 180� patterns were produced with the
most accuracy in the pre-practice trials, and the 90�tracking pattern had the second largest absolute phase error
and was not stably produced in the either the VPF or PF
conditions.
Practice
The absolute phase error values (/AE) and TCV values
from the practice trials were analyzed in 2 Visuomotor
Transformation 9 2 Day (day 1, day 2) 9 6 Block (1, 2, 3,
4, 5, 6) ANOVA with the last two factors repeated. Practice
produced a significant reduction in absolute phase error
from day 1 (/AE = 14�) to day 2 (/AE = 10�)
(F(1,10) = 14.14, p \ 0.01, g2p ¼ 0:59). A main effect of
Block was found in the /AE data (F(5,50) = 3.32, p \ 0.05,
g2p ¼ 0:25), and post hoc tests (p \ 0.05) revealed that the
/AE values from blocks 1 to 4 (B1, /AE = 14.3�; B2,
/AE = 12.9�; B3, /AE = 12.8�; B4, /AE = 12.1�) were
significantly larger than those from blocks 5 (/AE = 10.8�)
and 6 (/AE = 10.2�). The analysis of the TCV data
revealed a significant increase in the stability of visuomotor
tracking performance from day 1 (TCV = 3.67) to day 2
(TCV = 3.82) (F(1,10) = 61.87, p \ 0.01, g2p ¼ 0:86). A
significant main effect of Block was also found in the TCV
data (F(5,50) = 2.92, p \ 0.05, g2p ¼ 0:23), and post hoc
tests revealed that the TCV values from blocks 1–3 (B1,
TCV = 3.48; B2, TCV = 3.59; B3, TCV = 3.6) were
significantly smaller than the TCV values from blocks 4–6
(B4, TCV = 3.69; B5, TCV = 3.68; B6, TCV = 3.68).
Pre-practice scanning versus post-practice scanning
Figure 4 displays the individual trial relative phase means
(/obs) from the post-practice scanning session. The clus-
tering in the post-practice session was much tighter than in
the pre-practice session for both visuomotor transformation
groups in both the VPF and PF conditions (compare Fig. 4
to Fig. 2). To test for performance differences in the pre-
and post-practice scanning sessions, the /AE and TCV
values were analyzed in a 2 Visuomotor Transformation 9
2 Feedback 9 2 Session (pre-practice, post-practice) 9 5
Pattern ANOVA with the last three variables repeated
measures.
The analysis of the /AE data revealed a significant
increase in tracking accuracy from the pre-practice session
(/AE = 50.2�) to the post-practice session (/AE = 26.7�)
(F(1,10) = 122.94, p \ 0.01, g2p ¼ 0:92). Tracking perfor-
mance was more accurate in the VPF condition
(/AE = 21.8�) than in the PF condition (/AE = 55.3�)
(F(1,10) = 20.16, p \ 0.01, g2p ¼ 0:69). A main effect of
Pattern was found in the /AE data (F(4,40) = 5.21,
p \ 0.01, g2p ¼ 0:34). Post hoc tests (p \ 0.05) revealed
that the 0� (/AE = 27.6�), 180� (/AE = 34.8�), and 45�(/AE = 32.9�) patterns were produced with less error than
the 90� (/AE = 44.9�) and 135� (/AE = 52�) tracking
patterns. A Visuomotor Transformation 9 Pattern interac-
tion effect was found in the /AE data (F(4,40) = 3.33,
p \ 0.05, g2p ¼ 0:25) (Fig. 5a). Tests of the interaction
revealed that the 0� and 45� tracking patterns were more
accurate than the other three patterns in the correct feed-
back group. However, no significant differences were
found among the five tracking patterns in the mirrored
feedback group. The correct feedback group produced
A
0 45 9 0 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180CORMIR
B
RP pattern (deg)
0 45 90 135 180R
elat
ive
Pha
se (
deg)
-180
-135
-90
-45
0
45
90
135
180
Fig. 4 The individual trial observed mean relative phase values
(/obs) from the post-practice scanning trials are plotted for the VPF
(a) and PF (b) conditions as a function of required relative phase: 0�,
-45�, -90�, -135� and ±180�
Exp Brain Res (2009) 196:279–293 285
123
significantly less error for the 0� and 45� tracking patterns
than the mirrored feedback group.
Overall, tracking performance was more stable with lar-
ger TCV values in the post-practice session (TCV = 3.57)
compared to the pre-practice session (TCV = 2.84)
(F(1,10) = 107.61, p \ 0.01, g2p ¼ 0:92). Across both ses-
sions, stability of tracking performance was less in the PF
condition (TCV = 2.93) compared to VPF condition
(TCV = 3.48) (F(1,10) = 12.84, p \ 0.01, g2p ¼ 0:56). A
main effect of Pattern was found in the TCV data
(F(4,40) = 19.79, p \ 0.01, g2p ¼ 0:66). Post hoc tests of the
Pattern effect revealed that the 0� relative phase pattern
(TCV = 4.05) was the most stable tracking pattern, fol-
lowed by the 180� tracking pattern (TCV = 3.37). There
were no differences in the stability of the 45� (TCV = 2.94),
90� (TCV = 2.86), and 135� (TCV = 2.82) patterns, which
were all less stable than the 0� and 180� patterns. Simple
main effects tests of a significant Visuomotor Transforma-
tion 9 Pattern interaction (F(4,40) = 5.51, p \ 0.01, g2p ¼
0:36) revealed two main findings (Fig. 5b): First, within the
correct feedback group the 0� tracking pattern was the most
stable, and within the mirrored feedback group the 0� and
180� tracking patterns were the most stable. Second, the
mirrored feedback group was more variable when producing
the 45� relative phase pattern and more stable when pro-
ducing the 180� relative phase pattern compared to the cor-
rect feedback group.
The primary hypothesis put forth was that a significant
change in the accuracy and stability of the 90� tracking
pattern would emerge with practice and be consistent with
the bimanual findings reported by Zanone and Kelso (1992,
1997). The above analyses revealed significant Session and
Pattern effects, but not Session 9 Pattern interactions
localizing a large change in performance associated with
the 90� relative phase tracking pattern. The above findings
suggest two possible hypotheses regarding performance
change that may explain the lack of the interaction: (1)
practice of the 90� pattern did not alter performance, as
suggested by the Pattern effects which basically mirror the
pre-practice Pattern effects; or (2) practice of the 90� pat-
tern produced equivalent changes in performance in all
five tracking patterns, which the Session effects suggest
(compare Fig. 4 and Fig. 2). To test these two hypotheses,
pair-wise t tests (p \ 0.05) of the pre- and post-practice
absolute phase error data (/AE) and TCV data were con-
ducted separately for each pattern. The t tests of the /AE
data revealed that absolute phase error associated with the
45�, 90�, 135�, and 180� tracking patterns was significantly
smaller in the post-practice session compared to the pre-
practice session (ts(142) [ 2.87, ps \ 0.01) (Fig. 6a). The
t tests of the TCV data revealed that the post-practice
tracking performance for each tracking pattern was less
variable than the pre-practice performance (ts(142) \-2.97, ps \ 0.01) (Fig. 6b). Thus, the data support the
second hypothesis regarding the impact of practicing the
90� relative phase tracking pattern.
Discussion
Overall, the learning of the 90� relative phase tracking
pattern was not influenced by the mirrored condition. In
fact, practice with either correct or mirrored visual infor-
mation benefited learning, retention, and unexpectedly
transfer to all unpracticed patterns, with or without con-
current feedback. The positive transfer was found in both
the relative phase error and TCV data. Clearly, such gen-
eral transfer across unpracticed patterns is not consistent
with the general predications of the Haken et al. (1985)
model. To the best of our knowledge, tracking studies
usually display the to-be-tracked signal as a cursor moving
on the screen. Within the current task, we displayed the to-
be-tracked signal as two cycles of an actual sine wave.
Thus, the general transfer effect may have arisen from
either the use of sine waves as the display signals or the one
window display. To examine the impact of the single
window display on the general transfer effect, we ran a
second experiment wherein the to-be-tracked signal and
A
0 45 90 135 180
Abs
. Err
or (
deg)
0
10
20
30
40
50
60
70CORMIR
B
RP pattern (deg)
TC
V
0
1
2
3
4
5
0 45 90 135 180
Fig. 5 a Mean absolute error (/AE) and b mean TCV values are
plotted as a function of visuomotor transformation and required target
pattern. The error bars represent the standard error of the between
subject variability
286 Exp Brain Res (2009) 196:279–293
123
participants signal were presented in separate windows. If
the general transfer effect is the result of the display being
sine waves, then the same general transfer phenomenon
should emerge when the signals are displayed in separate
windows. However, if the general transfer effect is the result
of the coherency of the visual motion structure (Bogaerts
et al. 2003; Roerdink et al. 2005) and the superimposition of
the signals (Reed et al. 2003), then general transfer should
not occur, and transfer results more consistent with the
Haken et al. (1985) model should emerge.
Experiment 2
Materials and methods
Subjects
Twelve students (6 males, 6 females, mean age 23.4 years)
volunteered to participate in this study. The protocol and
consent forms were approved by the IRB board of Texas
A&M University and all participants signed a consent form
prior to the experiment. All participants were self-reported
right-handers and had normal or corrected-to-normal
vision.
Apparatus and data analysis
The same experimental apparatus, data collection proce-
dures, and data analyses used in Experiment 1 were
employed in this experiment.
Protocol
The task and procedures were identical to those used in
Experiment 1 except that the to-be-tracked signal and
participants signal were presented in separate windows
instead of superimposed in a single window. The display
windows on the monitor were 14 cm long 9 4.5 cm high.
The to-be-tracked signal was presented in the top window
and the participant’s motion signal was presented in the
bottom window. The bottom of window 1 and the top of
window 2 were 0.5 cm apart and this placed the peaks and
valleys of the two signals approximately 9.0 cm apart. A
9 cm distance was the largest employed by Reed et al.
(2003) when the to-be-tracked signal and the signal rep-
resenting arm motion were presented as cursors in the
visual display.
Statistical analysis
The ANOVAs used to analyze the pre-practice, practice,
and post-practice data sets each contained different factors
and have been reported in each section separately. Dun-
can’s multiple range tests (p \ 0.05) were used to analyze
all post hoc main effects, and simple main effect tests
(p \ 0.05) were performed to reveal differences when
significant interactions were present. Partial eta-squared
values (g2p) were computed and have been reported with
every significant F value.
Results
Pre-practice scanning
The clustering shown in Fig. 7 is consistent with that
shown in Fig. 2 and demonstrates a strong attraction
towards only the 0� and 180� relative phase patterns prior
to practice with the to-be-learned 90� relative phase. The
relative phase error data (/AE) and TCV data from the
pre-practice scanning sessions were analyzed in a 2
Visuomotor Transformation 9 2 Feedback 9 5 Pattern
ANOVA. A significant main effect of pattern was found in
the /AE data (F(4,40) = 7.64, p \ 0.0001, g2p ¼ 0:43) and
post hoc tests revealed that the 0� and 180� patterns were
produced with smaller error than the other patterns and that
the 90� target pattern was produced with significantly lar-
ger error than other patterns (Fig. 8a, pre-practice). Over-
all, /AE was larger in the PF condition (/AE = 63.4�)
A
0 45 90 135 180
Abs
. Err
or (
deg)
0
20
40
60
80
100 Pre-practicePost-practice
B
RP pattern (deg)
TC
V
0
1
2
3
4
5
* * * *
0 45 90 135 180* * * **
Fig. 6 Pre- and post-practice scanning data. a Mean absolute errors
(/AE) and b mean TCV values are plotted as a function of required
target pattern and scanning session. The error bars represent standard
error of between subject variability. Asterisks represent statistical
significance
Exp Brain Res (2009) 196:279–293 287
123
compared to the VPF (/AE = 37.7�) condition (F(1,10) =
16.16, p \ 0.01, g2p ¼ 0:62), and larger in the mirror visual
transformation (/AE = 61.8�) group compared to the cor-
rect (/AE = 39.2�) group (F(4,40) = 9.06, p \ 0.05,
g2p ¼ 0:47). The analysis of the TCV values found a
significant main effect of Pattern (F(1,10) = 13.42,
p \ 0.0001, g2p ¼ 0:57), and post hoc tests revealed that the
TCV values associated with the 0� phase pattern were
significantly larger (most stable) than the other patterns,
with the TCV values associated with the 90� and 135�phase patterns significantly smaller (least stable) than the
other phase patterns (Fig. 8b, pre-practice).
Practice
Overall, performance improved and stabilized with the two
window display used in this experiment. The /AE and TCV
values were analyzed in a 2 Visual Transformation 9 2
Day 9 6 Block ANOVA with repeated measures on block
and day. The analysis of /AE revealed a significant reduction
in tracking error from practice day 1 (/AE = 25.9�) to
practice day 2 (/AE = 15.4�) (F(1,10) = 23.8, p \ 0.001,
g2p ¼ 0:70), with more accurate tracking performance
emerging in the correct (/AE = 17.8�) versus mirrored
(/AE = 23.6�) visual transformation condition (F(1,10) =
5.04, p \ 0.05, g2p ¼ 0:36). Tracking performance also
became more stable with practice with larger TCV values
during day 2 (TCV = 3.10) compared to day 1 (TCV =
2.72) (F(1,10) = 23.03, p \ 0.001, g2p ¼ 0:7). A Day 9
Block interaction (F(5,50) = 3.81, p \ 0.01, g2p ¼ 0:28) was
also found in the TCV data. Post hoc tests found larger TCV
values (more stable tracking) in day 2 practice blocks 1–6
(TCV [ 3.03) compared to day 1 (TCV \ 2.97). Within
practice day 1, smaller TCV values were found in blocks 1–5
(TCV \ 2.72) compared to block 6 (TCV = 2.97). TCV
values were smaller in block 4 (TCV = 2.92) compared to
blocks 3 (TCV = 3.2) and 5 (TCV = 3.24) during practice
day 2.
Pre-practice scanning versus post-practice scanning
As in Experiment 1, the clustering was tighter around the
90� relative phase pattern following 2 days of practice
(Fig. 9). The issue is whether or not the clustering around
A
0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180COR
MIR
B
RP pattern (deg)
0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180
Fig. 7 The individual trial observed mean relative phase values
(/obs) from the pre-practice scanning trials are plotted for the VPF (a)
and PF (b) conditions as a function of required relative phase: 0�,
-45�, -90�, -135� and ±180�. The negative values represent that
the task required elbow motion to lag the external signal
A
0 45 90 135 180
Abs
. Err
or (
deg)
0
20
40
60
80
100Pre-practicePost-practice
B
RP pattern (deg)T
CV
0
1
2
3
4
5
* *
0 45 90 135 180
Fig. 8 Pre- and post-practice scanning data. a Mean absolute errors
(/AE) and b mean TCV values are plotted as a function of required
target pattern and scanning session. The error bars represent standard
error of between subject variability. Asterisks represent statistical
significance
288 Exp Brain Res (2009) 196:279–293
123
the non-practiced patterns changed significantly. The pre-
and post-practice /AE data and TCV data were analyzed in
a 2 Session 9 2 Visual Transformation 9 2 Feedback 9 5
Pattern repeated measures ANOVA.
Overall, /AE values decreased from the pre-practice
(/AE = 50.6�) to post-practice (/AE = 32.9�) session
(F(1,10) = 22.5, p \ 0.001, g2p ¼ 0:69) were larger in the
mirrored (/AE = 51.1�) compared to the correct visual
transformation (/AE = 32.4�) condition (F(1,10) = 7.14,
p \ 0.0001, g2p ¼ 0:41), and larger without visual feedback
in the PF (/AE = 53.3�) compared to the VPF
(/AE = 30.2�) condition (F(1,10) = 19.68, p \ 0.01,
g2p ¼ 0:66). Relative phase error varied significantly
between the five tracking patterns (F(4,40) = 8.53,
p \ 0.0001, g2p ¼ 0:46) and a significant Pattern 9 Session
interaction (F(4,40) = 3.78, p \ 0.01, g2p ¼ 0:27) was found
(Fig. 8a). Post hoc tests revealed a significant decrease in
/AE values from the pre- to post-practice sessions for the
target pattern of 90� and the unpracticed pattern of 45�(Fig. 8a). The TCV values overall increased significantly
(less variability) from the pre-practice (TCV = 3.01) to
post-practice session (TCV = 3.42) (F(1,10) = 10.95,
p \ 0.01, g2p ¼ 0:52), and varied significantly between the
five phase patterns (F(4,40) = 42.8, p \ 0.0001, g2p ¼ 0:81).
Post hoc tests of the Pattern effect found that the TCV
values for the 0� (TCV = 3.9), 180� (TCV = 3.5), and 45�(TCV = 3.1) patterns were larger than the TCV values of
the 90� (TCV = 2.7) and 135� (TCV = 2.7) patterns.
Retention of the practiced 90� relative phase as a function
of window display
In this section, an analysis of the /AE values and TCV
values from the day 3 retention trials for the 90� relative
phase pattern is undertaken with a 2 Window (Exp 1, Exp
2) 9 2 Visual Transformation 9 2 Feedback repeated
measures ANOVA. In the retention data, relative phase
error in the one window display (/AE = 32�) was not
significantly different from the relative phase error in the
two window display (/AE = 36�) for the required 90�relative phase pattern (p = 0.66). What did influence rel-
ative phase error was the nature of the feedback, with less
error when both visual and proprioceptive feedback (PVF)
were available (/AE = 21�) compared to when visual
feedback (PF) was not available (/AE = 47�)
(F(1,20) = 8.65, p \ 0.01, g2p ¼ 0:3). Variability in perfor-
mance based on the TCV values was influenced by the
number of windows (F(1,20) = 5.6, p \ 0.05, g2p ¼ 0:22)
and by the nature of the feedback (F(1,20) = 15.1,
p \ 0.001, g2p ¼ 0:43), with a significant interaction
occurring between these two variables (F(1,20) = 12.92,
p \ 0.01, g2p ¼ 0:39). Post hoc tests of the interaction
found that tracking performance at the required 90� relative
phase pattern was the most stable in the one window
condition when both proprioceptive and visual feedback
were present (1 window PVF TCV = 3.81; 1 window PF
TCV = 2.72; 2 window PVF TCV = 2.89; and 2 window
PF TCV = 2.85).
Discussion
Performance of the to-be-learned 90� pattern improved
with practice in both the correct and mirrored visual con-
ditions, with the improvements larger in the correct con-
dition as predicted. Even with the difference in conditions,
the separate window presentation did not remove the
learning opportunity associated with the feedback displays.
However, the two window display reduced the general
transfer effect to only one unpracticed pattern, the 45�pattern. The lack of a reduction in phase error across three
of the four unpracticed patterns strongly suggests that the
nature of the sine wave display in and of itself did not
produce the general transfer finding in Experiment 1. The
smaller performance improvements in the mirrored
A
0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180CORMIR
B
RP pattern (deg)
0 45 90 135 180
Rel
ativ
e P
hase
(de
g)
-180
-135
-90
-45
0
45
90
135
180
Fig. 9 The individual trial observed mean relative phase values
(/obs) from the post-practice scanning trials are plotted for the VPF
(a) and PF (b) conditions as a function of required relative phase: 0�,
-45�, -90�, -135� and ±180�. The negative values represent that
the task required elbow motion to lag the external signal
Exp Brain Res (2009) 196:279–293 289
123
condition together with the transfer results demonstrate that
the separation of the two signals decreased the visual
coherency of the display.
General discussion
The pre-practice results from the current experiments
produced three consistent findings with previous research
on the transformation of visual feedback. First, in the one
window display (Experiment 1), the mirrored feedback
group produced a more stable 180� tracking pattern than
the correct feedback group, and this supports the inter-
pretation that the transformed feedback produced a more
coherent visual motion structure defined by the 0� phase
relationship (Bogaerts et al. 2003; Roerdink et al. 2005).
However, in the two window display, there was no
advantage in the mirrored feedback group regarding the
180� tracking pattern. This indicates that the separation of
the signals reduced the strength of the visual coherence
created by the overlap in motion direction between the two
signals. Second, the transformed visual feedback did not
destabilize the production of the mirrored feedback group’s
0� tracking pattern in either experiment. This finding aligns
with other research demonstrating that participants ignore
visual feedback when the external signal and propriocep-
tive information coincide (Roerdink et al. 2005). Third, the
90� relative phase visuomotor tracking pattern was not
stable prior to practice, indicating that stabilizing a 90�visuomotor tracking pattern with a transformation that
provides a 0� relative phase representation between the
signal and arm’s motion, and learning a 90� visuomotor
tracking pattern are distinct process (Wilson et al. 2005).
Visuomotor transformation and visual coherency
in the enhancement of the 90� tracking pattern
The finding that the 90� tracking pattern was not stable
prior to practice is consistent with the bimanual studies
using a scanning procedure (Zanone and Kelso 1992, 1997)
and supports the conclusion that the 90� tracking pattern
did not correspond to an attractor of the visuomotor
tracking dynamics before practice. The level of perfor-
mance accuracy and the stability of the 90� tracking pattern
improved significantly across practice sessions and practice
blocks in both the one and two window display tasks. The
comparison of the retention data for the 90� relative phase
pattern from the two tasks also demonstrates the impor-
tance of the signal presentation. Although the relative
phase error was not significantly smaller in the one window
task compared to the two window task, the TCV values
were significantly smaller. Retention of the 90� pattern was
most stable in the single window display in the PVF (vision
and proprioception) condition, with no difference in the
stability of the 90� pattern between the one window PF
(proprioception only) condition and both the PVF and PF
conditions in the two window case. The advantage in
tracking stability emerged when the two signals were
superimposed visually and not separated. The advantage
does not emerge without visual feedback. This indicates
that signal presentation during training is important in
creating a global coherent visual structure that can influ-
ence the performance stability in dynamic perception–
actions tasks.
With the one window display, the correct feedback
group was less variable in performing the 90� relative
phase tracking pattern compared to the mirrored feedback
group before practice. However, no significant visuomotor
transformation effects were found in the practice sessions
as a function of day or block. The two window display,
however, resulted in poorer tracking accuracy in the mir-
rored group compared to the correct group during practice.
Why did the visuomotor transformation only affect learn-
ing in the two window display? The mirrored feedback in
this task did not convert the 90� relative phase pattern to a
more stable pattern such as 0�. Instead, the transformation
converted the 90� relative phase pattern to a 270� pattern,
which is a repeller just as the 90� pattern. In the one
window display, the superposition of the signals allowed
participants to more easily discriminate the mismatch
between the arm’s visual feedback representation and the
relative phase representation required by the task. The two
window display reduced visual coherency and weakened
the discrimination process making it harder for participants
to overcome the mismatch between the arm’s visual feed-
back representation and the required relative phase.
Modulation of the attractor landscape with practice
The dynamic pattern theory predicts that learning will take
the form of a phase transition, whereby a repeller will be
transformed into an attractor with practice (Schoner et al.
1992). This transition will occur when the to-be-learned
environmentally specified pattern does not coincide with a
preexisting stable attractor in the system’s coordination
landscape (Schoner et al. 1992; Zanone and Kelso 1992).
Characteristic features of a phase transition in learning will
be found in an increase in accuracy of the newly defined
pattern and an increase in the stability of the newly defined
pattern (Schoner et al. 1992; Zanone and Kelso 1992).
Overall, findings from both experiments regarding learning
of the 90� relative phase pattern are consistent with the
dynamical systems theoretical interpretation (Schoner et al.
1992) and with previous experimental results (Zanone and
Kelso 1992, 1997; Fontaine et al. 1997; Buchanan 2004;
Buchanan et al. 2007).
290 Exp Brain Res (2009) 196:279–293
123
Although good agreement between previous work and
the Haken et al. (1985) bimanual model was found with
regard to leaning the 90� pattern, the general transfer to
four unpracticed patterns in Experiment 1 raises some
issues of how far the analogy can be drawn between
visuomotor tracking and this model (Wimmers et al.
1992; Roerdink et al. 2005). For example, Russell and
Sternad (2001) argue that the bidirectional coupling in
the HKB model is inappropriate for unimanual visuo-
motor tracking skills, since only the participant adjusts
and the target oscillation is independent of these
adjustments. The display in the visual tracking paradigm
may allow the participant to overcome errors and phase
differences and override the entrainment strategies that
characterize interlimb coordination in many tasks (Rus-
sell and Sternad 2001). With the two window display,
the general transfer effect to the four unpracticed pat-
terns was eliminated, and only transfer to the unpracticed
pattern of 45� emerged. Byblow et al. (1995) found that
0� and 180� tracking patterns were more stable when
the curser to be tracked moved continuously across the
screen versus a discrete cursor only showing at the left-
most and the rightmost positions of the screen. The
differences in the discrete versus continuous cursor dis-
play may be analogous to our one and two window
displays. The sine wave presentation of the signals dur-
ing training provided a trail of visual information over
two complete cycles of motion. In the study by Byblow
et al. (1995), the continuous cursor may be viewed as
providing more information with which to regulate
tracking. In the current tasks, the sine wave display may
be viewed as possibly providing more information than a
continuously moving cursor. The extensive information
provided when training with the sine wave display in the
single window task may have increased the coherency of
the visual motion structure in the display more so than a
continuously moving dot does in comparison to a dis-
crete pulsed display. Moreover, 2 days of exposure to
such a rich display may have facilitated an improvement
in perceptual discrimination (or resolution) processes in
general (Gibson and Gibson 1955), and not just with the
to-be-learned pattern. Thus, the creation of the coherent
visual structure resulted in the development of a possible
perceptual discrimination capacity allowing for general
transfer with the integrated one window display in the
current study. The two window display, however,
reduced performance during practice, decreased stability
during retention, and significantly altered transfer per-
formance. The overlap in the single window provided a
more coherently grouped visual structure than the two
window display, just as the continuously moving cursor
provided a more coherently grouped visual structure than
the discrete cursor display.
Neural areas associated with visuomotor tracking
Neuroimaging studies have revealed a variety of neural
substrates involved in visuomotor skills (Alexander and
Crutcher 1990; Grafton et al. 2001). For example, the pre-
supplementary motor area has been shown to be associated
with encoding of the visuomotor relationship (Sakai et al.
1999), and the supplementary motor area has been shown
to be active in the initiation and guidance of visuomotor
tracking movements (Picard and Strick 2001). Imaging
studies have also demonstrated that basal ganglia lesions in
monkeys impair visuomotor tracking (Bowen 1969), and
have revealed a role for the basal ganglia in sensorimotor
adaptation processes early in training with a transformed
visuomotor condition (Seidler et al. 2006). In a tracking
task with compatible or incompatible visuomotor feedback,
Grafton et al. (2001) observed an increase in activity in the
pre-central gyrus, with maximal activity in the arm motor
area during compatible tracking and maximal activity in
the putative frontal eye fields during incompatible tracking.
Transfer to the compatible condition following training
with the incompatible condition resulted in a decrease in
activity in the putative frontal eye fields and an increase in
activity in the arm motor areas in the Grafton et al. (2001)
study. The current experiment also required learning of an
incompatible visuomotor display, and based on both per-
ceptual judgment tasks (Bingham et al. 1999; Zaal et al.
2000) and visuomotor tracking tasks (Bogaerts et al. 2003;
Roerdink et al. 2005) one that may have been more
incompatible than the task examined by Grafton et al.
(2001). A speculative link to the current experiments may
be that greater putative frontal eye field activity was
present when learning the 90� tracking pattern. This greater
activity may be associated with the transfer that was
observed to the other patterns, especially the 45� and 135�patterns. In other words, an increase in activity in the
putative frontal eye fields may be associated with percep-
tual discrimination processes in visuomotor tracking which
might account for the overall transfer performance in the
current task.
Conclusion
The present experiment revealed similarities and differ-
ences in the processes associated with learning a 90� rel-
ative phase pattern in bimanual and visuomotor tracking
tasks. The most striking difference was the degree of
transfer to other patterns following practice with a single
visuomotor tracking pattern that seemed to be dependent
on the presentation of the signals. This difference in
transfer performance between the one and two window
tasks raises the issue of the generality of the learning
Exp Brain Res (2009) 196:279–293 291
123
extension of the Haken et al. (1985) model, especially in
light of the strong similarities drawn between bimanual
coordination and visuomotor tracking with regard to the 0�and 180� relative phase patterns. The findings from the
current set of experiments join a growing body of research
demonstrating that visual displays can enhance motor
performance in a variety of tasks, e.g., bimanual coordi-
nation tasks (Mechsner et al. 2001; Tomatsu and Ohtsuki
2005; Hurley and Lee 2006; Kovacs et al. 2009) and rapid
aiming tasks (Fernandez and Bootsma 2008; Kovacs et al.
2008). What the current results suggest is that performance
advantages linked to visual displays arise from the global
structure of the display and the degree of visual coherency
created by this global structure.
Acknowledgments The authors would like to thank Dr. G.P.
Bingham and two other anonymous reviewers for their comments on
previous versions of this manuscript.
References
Alexander GE, Crutcher MD (1990) Preparation for movement:
neural representations of intended direction in three motor areas
of the monkey. J Neurophysiol 64:133–150
Amazeen PG (2002) Is dynamics the content of a generalized motor
program for rhythmic interlimb coordination? J Mot Behav
34:233–251
Bingham GP, Schmidt RC, Zaal FTJM (1999) Visual perception of
the relative phasing of human limb movements. Percept
Psychophys 61:246–258
Bogaerts H, Buekers MJ, Zaal FTJM, Swinnen SP (2003) When
visuomotor incongruence aids motor performance: the effect of
perceiving motion structures during transformed visual feedback
on bimanual coordination. Behav Brain Res 138:45–57
Bowen FP (1969) Visuomotor deficits produced by cryogenic lesions
of the caudate. Neuropsychologia 7:59–65
Buchanan JJ (2004) Learning a single limb multi-joint coordination
pattern: the impact of a mechanical constraint on the coordina-
tion dynamics of learning and transfer. Exp Brain Res 156:39–54
Buchanan JJ, Zihlman K, Ryu YU, Wright DL (2007) Interlimb
transfer of a relative phase and amplitude ratio pattern in a
multijoint task. J Motor Behav 39:49–67
Buekers MJ, Bogaerts HP, Swinnen SP, Felsen WF (2000) The
synchronization of human arm movements to external events.
Neurosci Lett 290:181–184
Burgess-Limerick R, Abernethy B, Neal RJ (1991) A statistical
problem in testing invariance of movement using the phase plane
model. J Motor Behav 23:301–303
Byblow WD, Chua R, Goodman D (1995) Asymmetries in coupling
dynamics of perception and action. J Motor Behav 27:123–137
Carnahan H, Hall C, Lee TD (1996) Delayed visual feedback while
learning to track a moving target. Res Q Exercise Sport 67:416–
423
Ceux T, Buekers MJ, Montagne G (2003) The effect of enhanced
visual feedback on human synchronization. Neurosci Lett
349:103–106
Fernandez L, Bootsma RJ (2008) Non-linear gaining in precision
aiming: making Fitts’ task a bit easier. Acta Psychol 129:217–
227
Fontaine RJ, Lee TD, Swinnen SP (1997) Learning a new bimanual
coordination pattern: reciprocal influences of intrinsic and to-be-
learned patterns. Can J Exp Psychol 51:1–9
Foulkes AJM, Miall RC (2000) Adaptation to visual feedback delays
in a human manual tracking task. Exp Brain Res 131:101–110
Gibson JJ, Gibson EJ (1955) Perceptual learning: differentiation or
enrichment? Psychol Rev 62:32–41
Grafton ST, Salidis J, Willingham DB (2001) Motor learning of
compatible and incompatible visuomotor maps. J Cognitive
Neurosci 13:217–231
Haken H, Kelso JAS, Bunz H (1985) A theoretical model of phase
transitions in bimanual coordination. Biol Cybern 51:347–356
Hurley SR, Lee TD (2006) The influence of augmented feedback and
prior learning on the acquisition of a new bimanual coordination
pattern. Hum Movement Sci 25:339–348
Kelso JAS (1984) Phase transitions and critical behavior in human
bimanual coordination. Am J Physicol 240:R1000–R1004
Kelso JAS, Scholz JAS, Schoner G (1986) Nonequilibrium phase
transitions in coordinated biological motion: critical fluctuations.
Phys Lett A 118:279–284
Kelso JAS, Buchanan JJ, Wallace SA (1991) Order parameters for the
neural organization of single, multijoint limb movement pat-
terns. Exp Brain Res 85:432–444
Kovacs AJ, Buchanan JJ, Shea CH (2008) Perceptual influences on
Fitts’ law. Exp Brain Res 190:99–103
Kovacs AJ, Buchanan JJ, Shea CH (2009) Bimanual 1:1 with 90�continuous relative phase. Exp Brain Res 193:126–129
Lee TD, Swinnen SP, Verschueren S (1995) Relative phase altera-
tions during bimanual skill acquisition. J Motor Behav 27:263–
274
Liao MJ, Jagacinski RJ (2000) A dynamical systems approach to
manual performance. J Motor Behav 32:361–378
Mardia KV (1972) Statistics in directional data. Academic Press,
London
Mechsner F, Kerzel D, Knoblich G, Prinz W (2001) Perceptual basis
of bimanual coordination. Nature 414:69–71
Peper CE, Beek PJ (1998) Are frequency-induced transitions in
rhythmic coordination mediated by a drop in amplitude? Biol
Cybern 79:271–300
Picard N, Strick PL (2001) Imaging the premotor areas. Curr Opin
Neurobiol 11:663–672
Reed DW, Liu XG, Miall RC (2003) On-line feedback control of
human visually guided slow ramp tracking: effects of spatial
separation of visual cues. Neurosci Lett 338:29–212
Roerdink M, Peper CE, Beek PJ (2005) Effects of correct and
transformed visual feedback on rhythmic visuo-motor tracking:
tracking performance and visual search behavior. Hum Move-
ment Sci 24:379–402
Russell DM, Sternad D (2001) Sinusoidal visuomotor tracking:
Intermittent servo-control or coupled oscillations? J Motor
Behav 33:329–349
Sakai K, Hikosaka O, Miyauchi S, Sasaki Y, Fujimaki N, Putz B
(1999) Presupplementary motor area activation during sequence
learning reflects visuo-motor association. J Neurosci 19:1–6
Schoner G, Haken K, Kelso JSA (1986) A stochastic theory of phase
transitions in human hand movement. Biol Cybern 53:247–257
Schoner G, Zanone PG, Kelso JSA (1992) Learning as a change of
coordination dynamics–theory and experiment. J Motor Behav
24:29–48
Seidler RD, Noll DC, Chintalapati P (2006) Bilateral basal ganglia
activated with sensorimotor adaptation. Exp Brain Res 175:544–
555
Tomatsu S, Ohtsuki T (2005) The effect of visual transformation on
bimanual circling movement. Exp Brain Res 166:277–286
292 Exp Brain Res (2009) 196:279–293
123
Wilson AD, Collins DR, Bingham GP (2005) Perceptual coupling in
rhythmic movement coordination: stable perception leads to
stable action. Exp Brain Res 164:517–528
Wimmers RH, Beek PJ, Van Wieringen PCW (1992) Phase-transi-
tions in rhythmic tracking movements–a case of unilateral
coupling. Hum Movement Sci 11:217–226
Zaal FTJM, Bingham GP, Schmidt RC (2000) Visual perception of
relative phase and phase variability. J Exp Psychol Human
26:1209–1220
Zanone PG, Kelso JAS (1992) The evolution of behavioral attractors
with learning: nonequilibrium phase transitions. J Exp Psychol
Human 18:403–421
Zanone PG, Kelso JAS (1997) Coordination dynamics of learning and
transfer: collective and component levels. J Exp Psychol Human
23:1454–1480
Exp Brain Res (2009) 196:279–293 293
123