Content uploaded by Michiteru Kitazaki
Author content
All content in this area was uploaded by Michiteru Kitazaki on Jul 03, 2014
Content may be subject to copyright.
Michiteru Kitazaki*
Department of Information Science
and Engineering
Toyohashi University of Technology
Toyohashi, Aichi 441-8580, Japan
Takuya Kimura
Department of Knowledge-Bas ed
Information Engineering
Toyohashi University of Technology
Japan
Effects of Long-Term Adaptation
to Sway-Yoked Visual Motion and
Galvanic Vestibular Stimulation on
Visual and Vestibular Control of
Posture
Abstract
Human postural control is a multimodal process involving visual and vestibular informa-
tion. The aim of the present study was to measure individual differences in the contri-
butions of vision and vestibular senses to postural control, and to investigate if the indi-
vidual weights could be modulated by long-term adaptation to visual motion or
galvanic vestibular stimulation (GVS). Since GVS is a less expensive technique than a
motion platform and can be wearable, it is a promising virtual reality (VR) technology.
We measured the postural sway of observers induc ed by a visual motion or GVS
before and after a 7-day adaptation task. We divided participants into four groups. In
visual adaptation groups, visual motions were presented to either enhance voluntary
body movement (enhancing vision group) or inhibit voluntary body movement (inhibi-
ting vision group). In GVS adaptation groups, GVS was applied to enhance voluntary
body movement (enhancing GVS group) or inhibit voluntary body movement (inhibi-
ting GVS group). The adaptation to enhancing body-movement-yoked visual motion
decreased the GVS-induced postural sway at a low motion frequency. The adaptation
to the enhancing GVS slightly increased the GVS-induced postural sway and decreased
the visually-induced sway at a low motion frequency. The adaptatio n to the inhibiting
GVS increased the GVS-induced postural sway and decreased the visually-induced
sway at a high motion frequency. These data suggest that long-term adaptation can
modify weights of vision and vestibular senses to control posture. These findings can
be applied to training or rehabilitation systems of postural control and also to ada ptive
virtual-reality systems.
1 Introduction
1.1 Human Postural Control from Vision and Vestibular
Senses
Human posture is a multimodal process controlled by visual, vestibular,
and proprioceptive information. Visually-induced postural sway has been widely
studied. Spontaneous postural sway decreases by 50–60% with a stable visual
environment (Travis, 1945; Edwards, 1946), while a moving room can affect
the posture of both adults and infants (Lee & Lishman, 1975; Lee & Aronson,
Presence, Vol. 19, No. 6, December 2010, 544–556
ª 2011 by the Massachusetts Institute of Technology *Correspondence to mich@cs.tut.ac.jp.
544
PRESENCE: VOLUME 19, NUMBER 6
1974). When a visual field contains a large visual motion,
body sway of observers occurs at an identical frequency
to the visual motion (Lestienne, Soechting, & Berthoz,
1977). Thus, when we use a cyclic visual motion, the
sway occurs at the same frequency of the visual motion.
Recently, galvanic vestibular stimulation (GVS) has
been investigated for vestibularly-induced postural sway.
When a small current is applied to the left and right mas-
toid processes, observers incline in the direction of the
anodal ear (Day, 1999). When GVS is applied to a walk-
ing observer, the walking trajectory is affected (Fitzpa-
trick, Wardman, & Taylor, 1999). GVS has also been
applied to some virtual reality systems. For instance, vir-
tual acceleration of self-motion was induced by GVS
with simultaneous presentation of visual motions
(Maeda et al., 2005; Maeda, Ando, & Sugimoto, 2005).
GVS was also applied to enhance music experience for
virtual-reality entertainment (Nagaya et al., 2006) and to
transfer body balance as a vestibular telepresence system
(Yoshida, Ando, Maeda, & Watanabe, 2008). Since the
GVS is a small system and is inexpensive, it is advanta-
geous in comparison with other larger systems such as a
motion platform, and is becoming increasing popular in
the virtual reality field.
1.2 Plasticity of the Human Perception-
Action Syst em
The human perception-action coordination system
is known to rapidly adapt to new situations. The most fa-
mous and traditional paradigm to investigate the adapta-
tion of the perception-action system is inverted vision
with a prism scope (Stratton, 1896). When wearing the
prism scope, the observer’s perceptual world is inverted,
causing staggering. After prolonged adaptation (1–4
weeks) the perceptual world returns to a proper orienta-
tion, and the observer is able to walk, run, and ride a
bicycle. A more moderate version of this system is dis-
torted optical stimulation (Welch, 1969), in which man-
ual pointing to a visual target fails and is shifted to the
distorted position when the observer observes the world
through the distortion prism. After adapting to the
prism, the observer can correctly point to the target.
These studies suggest that the human perception-
action system is plastic and adaptive to a new environ-
ment. When observers are moving, they sense both vis-
ual and vestibular sensations. In daily life, visual informa-
tion and vestibular information are consistent. For
example, when we walk 5 m, visual information and ves-
tibular information are equivalent and have a value of
5 m. In some studies, the gain of visual motion when an
observer actually moved was manipulated, and the pro-
longed adaptation to a new gain resulted in modification
of the perceptual stable gain in the direction of the
adapted gain (Wallach & Canal, 1976; Becklen, Wallach,
& Nitzberg, 1984; Wallach, 1987). Using VR techni-
ques, adaptation to a visual-vestibular conflict for 45 min
was reported to modify the vestibular sensation, but not
the vestibulo-ocular reflex (VOR; Ivanenko, Viaud-
Delmon, Siegler, Israel, & Berthoz, 1998; Viaud-
Delmon, Ivanenko, Grasso, & Israel, 1999). The degree
of this adaptation is stronger for males than females
(Viaud-Delmon, Ivanenko, Berthoz, & Jouvent, 1998).
In a previous study, we manipulated the gain of visual
motion when an observer rotates his head when wearing
a head-mounted display (HMD), and examined adaptive
changes in the most stable gain perceived by the observer
(Kitazaki & Shimizu, 2005). We found that the stable
gain changed after active adaptation for 2 min, and that
the adaptation for visual stability is related to relatively
higher information processing, at least after the fusion of
binocular sources, but is specific to or modulated by the
retinal location. In clinical studies, Jenkins and col-
leagues reported that the VOR gradually changes for
120 days after vestibular ablation (Jenkins, Cohen, &
Kimball, 2000; Jenkins, 1985).
1.3 Relevance to Virtual Reality
We propose that combining GVS and visual stimu-
lation with the plasticity of the perceptual-action system
of the brain will provide new virtual reality applications;
for instance, an adaptive telepresence system for mobile
observers who wear an HMD for vision and a GVS sys-
tem for vestibular sensation, and who receive visual and
vestibular sensation from other people at distance loca-
tions. The telepresence system needs to send multimodal
Kitazaki and Kimura 545
information such as vision and vestibular sensation.
However, they may not necessarily be sent with high re-
solution. If the quality of vision is lower than the vestib-
ular information, users of the system should utilize the
vestibular information more than the degraded visual in-
formation. If users rely more on the vestibular informa-
tion with a long-term usage/adaptation, users would
gradually adapt to the telepresence system appropriately.
The aim of the present study was to measure individ-
ual differences in the contributing weights of vision and
vestibular senses to postural control, and to investigate
whether the individual weights could be modulated by a
long-term active adaptation to visual motion and GVS.
2 General Methods
2.1 Participants
Twenty-four graduate and undergraduate students
participated with written informed consent. All partici-
pants were male and 20–23 years old. Since a previous
study reported that the degree of vestibular adaptation is
stronger for males than females (Viaud-Delmon et al.,
1998), we employed only male participants. This study
was approved by the Committee for Human-Subject
Studies of the Toyohashi University of Technology.
2.2 Apparatus
All experiments were conducted in a semi-dark
room, as shown in Figure 1(a). Visual stimuli were pre-
sented on a rear-projection screen (width height, 2.43
m 1.82 m) by a 3-CRT projector (Barco Cine7/II;
1024 768 pixels, 60 Hz refresh rate). GVS was gener-
ated by a D/A device (National Instruments PCI-6704),
and applied through the left and right mastoid processes
with a pair of disposable Ag/AgCI electrodes (Ambu
Blue Sensor P-00-S; Figure 1[b]).
The head movement and center of gravity for the
indexes of postural sway were measured for all partici-
pants. The head motion was measured by a magnetic
3D motion tracker (Polhemus FASTRAK) at 60 Hz.
The center of gravity was measured by a force plate
(NEC EB1101) and recorded through an A/D device
(National Instruments PCI-6024) to a computer. The
computer controlled visual stimuli and GVS measured
both the head motion and the center of gravity. Thus,
the timing of stimulus presentation and recording pos-
tural data were synchronized.
2.3 Experimental Design
We employed a typical experimental paradigm to
investigate adaptation: pretests and posttests with an ad-
aptation (Figure 2). First, all 24 participants performed
an experiment to measure postural sway induced by vis-
ual stimulus and GVS (pretest). Next, the participants
adapted to one of four perception-action combinations
for 7 days. For five participants, visual motion was pre-
sented to enhance voluntary body movement (enhancing
body-movement-yoked visual motion). For six partici-
Figure 1. Experimental apparatus. (a) Experimental scene in a dark
room. (b) Disposable electrode at left mastoid.
546 PRESENCE: VOLUME 19, NUMBER 6
pants, visual motion was presented to inhibit the move-
ment (inhibiting body-movement-yoked visual motion).
For six participants, GVS was applied to enhance volun-
tary body movement (enhancing body-movement-yoked
GVS). For seven participants, GVS was applied to inhibit
voluntary body movement (inhibiting body-movement-
yoked GVS). After the adaptation period, all participants
repeated an identical postural-sway experiment to test
the effect of adaptation (posttest). The changes of visu-
ally- and GVS-induced postural sways before and after
the adaptation reflect the effect of adaptation and the
plasticity of the multimodal postural control system.
The individual variance of the adaptation methods was
examined as between-subject conditions (four partici-
pant groups), while the effect of adaptation was investi-
gated with within-subject design by comparing the
results of the pretest with those of the posttest. The
numbers of participants in the four adaptation groups
were unequal as some participants canceled prior to per-
forming experiments.
3 Visually- and GVS-Induced Postural
Sways (Pre test)
3.1 Visual an d GVS Stimuli
The modality of the motion stimulus was the most
critical independent variable. Either visual stimulus or
GVS was presented to a participant for each trial with a
within-subject design. For the visual stimulus, we pre-
sented a lateral motion of random dots on the front-par-
allel screen (width height, 58.38 50.58 in visual
angle) at 1.5 m viewing distance. Five thousand random
dots moved laterally (leftward and rightward) back and
forth with a sinusoidal speed modulation (travel dis-
tance, 11.68 to 5.88 left and right from the center). Each
dot was a square (0.458 0.458) and red (0.76 cd/m
2
)
on a black background. The density of the random dots
on the background was 15.9% (see Figure 3).
For the GVS, we presented a weak current (minimum
0.1 mA, maximum 0.5 mA) through left and right mas-
toids; the current modulation was sinusoidal and similar
to the visual motion. Although the resolution of current
modulation was limited to 0.1 mA steps by the appara-
tus, the postural sway of participants was natural (see
Figure 4[a]). The motion frequencies of visual motion
and GVS were varied as an independent variable (0.1,
0.2, and 0.3 Hz) with within-subject design. Since the
Figure 2. Diagram of experimental design.
Figure 3. Visual stimuli. (a) Schematic stimulus of random dot visual
motion. (b) Visual motion varied with motion frequency. Motion of ran-
dom dots plotted against time for 0.1, 0.2, and 0.3 Hz conditions.
Figure 4. Sample data of postural sway and frequency analysis.
(a) Postural sway data of a trial of a subject and the GVS current were
plotted against time. (b) FFT was applied to the left data, and the
sway power was plotted against the sway frequency.
Kitazaki and Kimura 547
duration of a trial was always 90 s, the 9, 18, and 27
cycles of visual or GVS motions were presented for 0.1,
0.2, and 0.3 Hz conditions, respectively. A fixation point
(a cross, 4.68 4.68) was placed at the center of the dis-
play, to which participants were asked to fix their eyes. In
the visual motion condition, moving random dots were
presented around the fixation point and, in the GVS
condition, only the fixation point was presented on a
black background.
3.2 Procedure
A session contained a condition combination of
two stimulus modalities (Vision or GVS) and three
motion frequencies (0.1, 0.2, or 0.3 Hz), where these
six conditions/trials were presented in a random order.
Each participant conducted 10 sessions. Thus, every
condition was repeated 10 times.
3.3 Results
All participants showed postural sways induced by
visual and GVS motions. An example of postural sway
induced by GVS for a trial of a subject is shown in Figure
4(a). We applied FFT to each trial data, and extracted
postural-sway powers at the frequency of the visual or
GVS motion (see Figure 4[b]). The postural sway of an
observer typically occurs at the same frequency as the
stimulus motion or modulation. We obtained two types
of indexes for postural sway: center of gravity and head
motion. However, as the results of these two indexes
were similar, only the center of gravity data are pre-
sented.
3.3.1 Individual Differences. Postural-sway
powers were presented at the same frequency of the
stimulus motion for visual- and GVS-modality condi-
tions (see Figure 5), and individual data (n ¼ 24) were
plotted to observe large individual differences. The indi-
vidual differences were larger for visually-induced sways
than for GVS-induced sways. The absolute power of the
most sensitive participant was 1,000 times larger than
less sensitive participants for vision, and 100 times larger
for GVS.
To determine the ratios of contributing weights of
vision and vestibular for each subject’s postural control,
correlations of visually-induced sways and GVS-induced
sways were plotted (see Figure 6), where each dot repre-
sents a subject. Data were divided into three motion-fre-
quency conditions. The horizontal axis indicates power
of visually-induced postural sways at the stimulus fre-
quency, and the vertical axis indicates power of GVS-
induced sways. There were significant positive correla-
tions between visually- and GVS-induced sways
(r ¼ 0.58 and p ¼ .0030 for 0.1 Hz, r ¼ 0.52 and p ¼
.0089 for 0.2 Hz, and r ¼ 0.53 and p ¼ .0081 for 0.3
Hz). Thus, subjects who were sensitive to the visual
motion were also sensitive to the GVS for the induced
postural sway. Since many subjects were plotted in the
upper-left side of the center oblique line, the GVS-
induced sways were larger than the visually-induced sway
for many subjects.
3.3.2 Effects of Sensory Modality and Motion
Frequency. Data were normalized using the individual
averages as absolute values of the postural sways were
large. Normalized data (relative value to the individual
average) are shown in Figure 7. Data are averages of all
24 participants with error bars of the standard errors.
The GVS-induced postural sway was larger than the visu-
ally-induced postural sway, particularly at low frequen-
cies. The GVS-induced sway decreased with a high fre-
quency. A repeated measures ANOVA (two ways; two
modalities three frequencies) showed a significant
Figure 5. Individual results of visually- and GVS-induced postural sways
(pretests). Sway powers induced by visual and GVS stimuli were plotted
against stimulus frequency conditions for all subjects.
548 PRESENCE: VOLUME 19, NUMBER 6
main effect of the modality (p ¼ .0002) and the fre-
quency (p ¼ .0087), and an interaction between them
(p ¼ .0211). These data were identical to those pre-
sented in our previous study (Kitazaki & Kimura, 2008).
4 Long-Term Adaptation
4.1 Adaptation Groups
Participants were randomly divided into four
groups prior to pretests: the inhibiting body-movement-
yoked vision group (six participants), the enhancing
body-movement-yoked vision group (five participants),
the inhibiting body-movement-yoked GVS group (seven
participants), and the enhancing body-movement-yoked
GVS group (six participants).
4.2 Methods
The apparatus was identical to the pretests except
that the force plate was excluded. The motion of each
participant’s head was monitored by a motion tracker,
and visual motion or GVS were continuously manipu-
lated online depending on their motion at 60 Hz. In the
enhancing vision adaptation, when the participant
inclined rightward, the random dots moved rightward to
induce more rightward postural sway (and vice versa). In
the inhibiting vision adaptation, when the participant
Figure 6. Correlation between visually- and GVS-induced postural sways (pretests). Each dot repre-
sents a subject. The horizontal axis indicates powers of visually-induced postural sways at the stimulus
frequency, and the vertical axis indicates powers of GVS-induced sways.
Figure 7. Averaged results of visually- and GVS-induced postural sways
(pretests). Individually normalized data of sway power induced by visual
and GVS stimuli were averaged and plotted against the stimulus motion
frequency. Error bars indicate standard errors.
Kitazaki and Kimura 549
inclined rightward, the random dots moved leftward.
For these two conditions, the manipulation of visual
motion was the opposite. In the enhancing GVS adapta-
tion, when the participant inclined rightward, the GVS
induced a rightward postural sway. In the inhibiting
GVS adaptation, when the participant inclined leftward,
the GVS induced a rightward postural sway. For these
two conditions, the manipulation of GVS was the oppo-
site.
Participants were asked to sway their body laterally
back and forth with a 30 cm travel distance (15 cm left
and right from the center) at 0.2 Hz. To guide this body
motion, two vertical bars in red and green were pre-
sented on the screen (see Figure 8). The red bar moved
left and right with sinusoidal speed modulation at 0.2
Hz. The green bar represented the current (online) posi-
tion of the participant’s head. We asked participants to
move the green bar on the red bar as accurately as possi-
ble. Each trial continued for 60 s. Each subject per-
formed 10 trials a day, and continued for 7 days. Thus,
every participant performed 70 trials in total.
5 Effects of Long-Term Adaptation on
Induced Postural Sway (Posttest)
5.1 Methods
The methods were identical to the pretest. All par-
ticipants performed the same experiment.
5.2 Results
5.2.1 Amplitude of Postural Sways. Identical
analyses to those of the pretest were performed. Data
were plotted for inhibiting vision (see Figure 9), enhanc-
ing vision (see Figures 10 and 11), inhibiting GVS (see
Figure 12), and enhancing GVS (see Figures 13 and 14)
groups. In Figures 9, 10, 12 and 13, the left graph repre-
sents pretest data of the same group subjects, while the
center graph represents posttest data. To determine indi-
vidual changes before and after the adaptation, the dif-
ference of the induced postural sway before and after the
adaptation was calculated by subtracting the pretest data
Figure 8. Display for the adaptation experiment.
Figure 9. Effects of adaptation in the inhibiting vision group. (Left) Pre-
test data of the inhibiting vision group in the same format as in Figure 7.
(Center) Posttest data of the inhibiting vision group. (Right) The differ-
ence between the posttest data and the pretest data were plotted to
see adaptation effects.
Figure 10. Effects of the adaptation in the enhancing vision group.
The graph format matches that in Figure 9.
550 PRESENCE: VOLUME 19, NUMBER 6
from the posttest data; these data are plotted in the
right-side graph of Figures 9, 10, 12 and 13.
There were no effects of the adaptation for the inhibi-
ting vision adaptation condition (see Figure 9), while the
visually-induced sway slightly increased and the GVS-
induced sway decreased at low frequency (0.1 Hz) for
the enhancing vision adaptation condition (see Figure
10). Repeated measures ANOVA for the data from Fig-
ure 9 (right) and Figure 10 (right) were performed to
examine the effects of the adaptation (three ways: two
adaptation groups two modalities three frequen-
cies). There was a significant three-way interaction
between adaptation (inhibiting/enhancing) modality
frequency (p ¼ .0491), which supports the observa-
tion that the enhancing vision adaptation only decreased
GVS-induced sway and increased visually-induced sway
at the low frequency. Single-sample t-tests were per-
formed to test whether power changes of postural sways
were different from zero for each condition; there was a
near-significant effect suggesting that GVS-induced sway
was decreased at 0.1 Hz (p ¼ .079). Next, paired t-tests
were performed to test the difference between the GVS-
induced sways and the visually-induced sways for each
frequency condition; there was only a weak tendency for
a difference between GVS- and visually-induced sways
Figure 11. Effects of adaptation in the enhancing vision group at
0.1 Hz in correlation plots. Correlations of powers of GVS- and visually-
induced sways were plotted following the format of Figure 6. Both pre-
test and posttest data were plotted. Arrows indicate individual changes
before and after adaptation.
Figure 12. Effects of the adaptation in the inhibiting GVS group. The
graph format follows that of Figure 9.
Figure 13. Effects of the adaptation in the enhancing GVS group. The
graph format follows that of Figure 9.
Figure 14. Effects of the adaptation in the inhibiting GVS group at 0.3
Hz in correlation plots. The graph format follows that of Figure 11.
Kitazaki and Kimura 551
with the enhancing vision adaptation at 0.1 Hz (p ¼
.1441).
For the 0.1 Hz condition of the enhancing vision
group before and after the adaptation, each subject’s
sway power was plotted to determine the visually-
induced sways and GVS-induced sways (see Figure 11);
these data were not normalized by individual averages.
For four of the five subjects, contributing weights of
vision increased while those of GVS decreased.
For the inhibiting GVS adaptation condition, the
GVS-induced sway increased particularly at the high fre-
quency (0.3 Hz), while the visually-induced sway slightly
decreased (see Figure 12). For the enhancing GVS adap-
tation condition, the GVS-induced sway slightly
increased and the visually-induced sway decreased at the
low frequency (see Figure 13).
Repeated measures ANOVA for the data from Figure
12 (right) and 13 (right; three ways: two adaptation
groups two modalities three frequencies) was per-
formed to test the effects of the adaption; however, there
were no effects as the individual differences were too
large. Next, single-sample t-tests were performed to
determine whether the power changes of postural sway
were different from zero for each condition. There was a
near-significant effect of increased GVS-induced sway at
0.3 Hz by inhibiting GVS adaptation (p ¼ .054). Paired
t-tests were performed to test the difference between
GVS-induced sway and visually-induced sway for each
frequency condition. For the inhibiting GVS condition,
there was a significant effect of the adaptation (p ¼ .026)
at 0.3 Hz where the GVS-induced sway increased signifi-
cantly more than the visually-increased sway. For the
enhancing GVS condition, there was a near significant
effect of the adaptation (p ¼ .060) at 0.1 Hz, where the
GVS-induced sway increased more than the visually-
induced sway.
For the 0.3 Hz condition of the inhibiting GVS group
and the 0.1 Hz condition of the enhancing GVS group
before and after the adaptation, each subject’s sway
power was plotted to determine ratios of the visually-
induced sways and the GVS-induced sways (see Figures
14 and 15, respectively). For five of the seven subjects in
the inhibiting GVS group, the contributing weights of
GVS increased and those of vision decreased (see Figure
14). For three of the six subjects in the enhancing GVS
group, contributing weights of GVS increased and those
of vision decreased (see Figure 15).
5.2.2 Phase D elay of Postural Sways. The
phases of postural sway at the same frequency of visual
or GVS stimuli were calculated, and pretests, posttests,
and their difference were plotted in Figures 16, 17, 18,
and 19. Postural sway delays to stimulus motions, partic-
ularly GVS-induced sways, have a large delay for
30–1208 (Kitazaki & Kimura, 2008), which may affect
the delay of postural sways. Similar ANOVAs (three
Figure 15. Effects of the adaptation in the enhancing vision group at
0.1 Hz in correlation plots.
Figure 16. Effects of adaptation on phase delay in the inhibiting vision
group. The phase delays were plotted following the format of Figure 9.
(Left) Pretest data of the inhibiting vision group. (Center) Posttest data.
(Right) The difference between the posttest data and the pretest data.
552 PRESENCE: VOLUME 19, NUMBER 6
ways: two adaptation groups [inhibiting/enhancing]
two modalities three frequencies) were performed for
visual adaptations and GVS adaptations. Single-sample
t-tests and paired t-tests were also performed, similar to
the power/amplitude analysis. There was no effect of ad-
aptation in the inhibiting vision group (see Figure 16).
For the enhancing vision group, the delay of visually-
induced sway increased at the middle frequency (0.2 Hz;
see Figure 17; single-sample t-test, p ¼ .064; paired t-
test p ¼ .076). However, there was no significant effect
of the ANOVA.
For the inhibiting GVS group, the delay of visually-
induced sways increased at the high frequency (0.3 Hz;
see Figure 18; single-sample t-test, p ¼ .079). For the
enhancing GVS group, the delay of visually-induced
sway decreased at the middle frequency (0.2 Hz; see Fig-
ure 19; single-sample t-test, p ¼ .090). For the three-
way ANOVA, there was a significant interaction of
modalities and frequencies (p ¼ .035), suggesting that
both inhibiting and enhancing GVS adaptations
increased the delay of visually-induced sway increase at
the high frequency and decrease at the low frequency.
6 Discussion
We performed a long-term (7-day) adaptation
experiment using a body-movement-yoked visual
motion and GVS. Participants received visual motions or
GVS to either inhibit or enhance their voluntary sway.
The enhancing vision adaptation decreased the GVS-
induced sway and slightly increased the visually-induced
sway only at the low frequency (0.1 Hz). The adaptation
to the inhibiting GVS increased the GVS-induced pos-
tural sway and decreased the visually-induced sway, par-
ticularly at a high motion frequency (0.3 Hz). The adap-
tation to the enhancing GVS slightly increased the GVS-
induced postural sway and decreased the visually-
induced sway, particularly at a low motion frequency
(0.1 Hz). These results are summarized in Table 1.
For the delay of postural sway, the enhancing vision
adaptation increased the delay of visually-induced sway
at the frequency of adaptation (0.2 Hz). The adaptation
to the inhibiting GVS increased the delay of visually-
induced postural sway at a high motion frequency (0.3
Hz), while the adaptation to the enhancing GVS slightly
decreased the visually-induced postural sway at the fre-
quency of adaptation (0.2 Hz).
Figure 19. Effects of the adaptation on phase delay in the enhancing
GVS group. The graph format matches that of Figure 16.
Figure 17. Effects of adaptation on phase delay in the enhancing
vision group. The graph format follows that of Figure 16.
Figure 18. Effects of the adaptation on phase delay in the inhibiting
GVS group. The graph format follows that of Figure 16.
Kitazaki and Kimura 553
6.1 Modification of Sway Amplitudes
and Modality Contributi ons
The enhancing vision adaptation increased the
weight of the visual control of posture relative to that of
the vestibular control, which is reasonable as the enhanc-
ing vision adaptation seemed to make visual information
more reliable for the purpose of controlling posture. We
suggest that the effect was limited to the lower frequency
as the difference of the weights of vision and vestibular
senses are large at the low frequency (see Figure 7).
The inhibiting GVS adaptation increased the weight
of the vestibular control of posture relative to that of the
visual control at a high motion frequency. Although this
result is controversial, speculatively, our visual and ves-
tibular postural control may not be good at higher fre-
quencies (Stoffregen, 1986; van Asten, Gielen, &
van der Gon, 1988; Kitazaki & Kimura, 2008). By
exposure to inhibiting GVS, a negative aftereffect might
occur at the frequency vulnerable to posture controls,
then GVS-induced sway would increase after adaptation.
The enhancing GVS adaptation increased the weight of
the vestibular control of posture relative to that of the
visual control at 0.1 Hz, which is similar to our findings
of the enhancing vision adaptation.
6.2 Modification of Sway Delay
For all conditions and groups, the delay of the
GVS-induced sway was not affected by the adaptation.
Thus, the temporal system of the vestibular control of
posture seems robust. The delay of the visually-induced
sway was increased by the adaptations to enhancing
vision at 0.2 Hz, and decreased by the adaptation to
enhancing GVS at 0.2 Hz. If the enhancing vision adap-
tation makes visual information more reliable to control
postures, the delay should be decreased. However, our
results were the opposite. We speculate that the temporal
aspects of visual control of posture are slightly sensitive
to the adaptation, but cannot be accounted for by the
modification of modality contributions. Further studies
on this area are required.
6.3 Potential Contributions to VR
Application
Our data suggest that the long-term adaptation to
the body-movement-yoked visual motion and GVS can
modify weights on vision and vestibular senses to control
posture. This reflects great plasticity in the perception-
action system. Our findings can be applied to an adaptive
virtual-reality system or a training or rehabilitation sys-
tem of postural control in the future. For the adaptive
telepresence system for mobile observers, our results
suggest that enhancing visual motion and GVS can mod-
ify the weights of visual and vestibular sensation for pos-
tural control. If the quality of vision is lower than the
vestibular information, the amplitude of the vestibular
information should be increased with the GVS. For
long-term usage/adaptation, users would gradually rely
more on the vestibular information, and the degraded
visual information would not deteriorate the perform-
ance. However, inhibiting visual motion would have no
effect for the adaptive system. There is also a limitation
with respect to the effective frequency. The adaptation
Table 1. Summary of Adaptation Effects*
[Visual motion] [GVS]
[Enhanced] þVisual sway
– GVS sway at low frequency
– Visual sway
þGVS sway at low frequency
[Inhibited] No effect – Visual
þGVS sway at high frequency
*The first and second columns indicate modalities of adaptation, and the first and second rows indicate the direction
of the adaptation (enhanced or inhibited). If postural sway is increased after adaptation, þ is indicated. If postural
sway is decreased after adaptation, – is indicated.
554 PRESENCE: VOLUME 19, NUMBER 6
effects with enhancing visual motion and GVS mainly
appear at low motion frequency. Thus, the proposed sys-
tem may be effective for slow self-motion rather than fast
movements.
If vestibular sensitivity worsens more than vision due
to aging or accidental disorder, then visual information
should be used to control posture more than the vestib-
ular information. We propose the use of an AR applica-
tion to enhance optic flow to increase the visual weight
for postural control as a training or rehabilitation system.
Optic flow can be presented by an HMD and the vestib-
ular information can be presented by GVS. Since both
the HMD and GVS are small and wearable, the AR train-
ing system can be used at home, and will not impair the
quality of life of the user. After use of this system for a
period, users will rely more on visual information rather
than vestibular information. As our proposed applica-
tions work gradually and adaptively, the effects may by
optimized by quitting adaptation at an appropriate time
while evaluating adaptation effects continuously. How-
ever, our findings are still preliminary, and further
experiments are required with a larger number of partici-
pants.
7 Conclusions
We measured the postural sway of observers
induced by a visual motion or GVS before and after a
7-day adaptation task. Adaptations to enhanced visual
motion and GVS that were yoked to the observer’s vol-
untary self-motion caused an increase in visually-induced
postural sway and GVS-induced postural sway, respec-
tively. Thus, the long-term adaptation can modify
weights of vision and vestibular senses over postural con-
trol. These findings can be applied to a training or reha-
bilitation system of postural control, and also as an
adaptive virtual-reality system.
Acknowledgm ents
This research was supported by the Nissan Science Foundation,
a Grant-in-aid for Young Scientists (B), and The Global COE
Program (Frontiers of Intelligent Sensing) by MEXT Japan.
References
Becklen, R., Wallach, H., & Nitzberg, R. (1984). A limitation
of position constancy. Journal of Experimental Psychology:
Human Perception and Performance, 10, 713–723.
Day, B. L. (1999). Galvanic vestibular stimulation: New uses
for an old tool. Journal of Physiology, 517, 631.
Edwards, A. S. (1946). Body sway and vision. Journal of Exper-
imental Psychology, 36, 526–535.
Fitzpatrick R. C., Wardman D. L., & Taylor J. L. (1999).
Effects of galvanic vestibular stimulation during human
walking. Journal of Physiology, 517, 931–939.
Ivanenko, Y. P., Viaud-Delmon, I., Siegler, I., Israel, I., & Ber-
thoz, A. (1998). The vestibulo-ocular reflex and angular dis-
placement perception in darkness in humans: Adaptation to
a virtual environment. Neuroscience Letters, 241, 167–170.
Jenkins, H. A. (1985). Long-term adaptive changes of the ves-
tibulo-ocular reflex in patients following acoustic neuroma
surgery. Laryngoscope, 95(10), 1224–1234.
Jenkins, H. A., Cohen, H. S., & Kimball, K. T. (2000). Long-
term vestibulo-ocular reflex changes in patients with vestibu-
lar ablation. Acta Otolaryngology, 120(2), 187–191.
Kitazaki, M., & Kimura, T. (2008). Frequency-phase analysis
of postural sway induced by visual motion and galvanic ves-
tibular stimulation. Journal of Vision, 8(6), 1046.
Kitazaki, M., & Shimizu, A. (2005). Visual-motor adaptation
to stabilize perceptual world: Its generality and specificity.
Proceedings of 15th International Conference on Artificial
Reality and Telexistence, 85–90.
Lee, D. N., & Aronson, E. (1974). Visual proprioceptive con-
trol of standing in human infants. Perception & Psychophysics,
15, 529–532.
Lee, D. N., & Lishman, J. R. (1975). Visual proprioceptive con-
trol of stance. Journal of Human Movement Studies, 1, 87–95.
Lestienne, F., Soechting, J., & Berthoz, A. (1977). Postural
readjustments induced by linear motion of visual scenes.
Experimental Brain Research, 28, 363–384.
Maeda, T., Ando, H., Amemiya, T., Nagaya, N., Sugimoto,
M., & Inami, M. (2005). Shaking the world: Galvanic ves-
tibular stimulation as a novel sensation interface. ACM
SIGGRAPH 2005 Emerging Technologies. doi: 10.1145/
1187297.1187315
Maeda, T., Ando, H., & Sugimoto, M. (2005). Virtual acceler-
ation with galvanic vestibular stimulation in a virtual reality
environment. Proceedings of IEEE Virtual Reality 2005,
289–290.
Kitazaki and Kimura 555
Nagaya, N., Yoshidzumi, M., Sugimoto, M., Nii, H., Maeda,
T., Kitazaki, M., & Inami, M. (2006). Gravity Jockey: A
novel music experience with galvanic vestibular stimulation.
Proceedings of the 2006 ACM SIGCHI International Confer-
ence on Advances in Computer Entertainment Technology.
doi: 10.1145/1178823.1178881
Stoffregen, T. A. (1986). The role of optical velocity in the
control of stance. Perception & Psychophysics, 39(5), 355–
360.
Stratton, G. M. (1896). Some preliminary experiments on
vision without inversion of the retinal image. Psychological
Review, 3, 611–617.
Travis, R. C. (1945). An experimental analysis of dynamic and
static equilibrium. Journal of Experimental Psychology, 35,
216–234.
van Asten, W. N., Gielen, C. C., & van der Gon, J. J. (1988).
Postural movements induced by rotations of visual scenes.
Journal of the Optical Society of America, A5(10), 1781–
1789.
Viaud-Delmon, I., Ivanenko, Y. P., Berthoz, A., & Jouvent, R.
(1998). Sex, lies and virtual reality. Nature Neuroscience, 1,
15–16.
Viaud-Delmon, I., Ivanenko, Y. P., Grasso, R., & Israel, I.
(1999). Non-specific directional adaptation to asymmetrical
visual-vestibular stimulation. Cognitive Brain Research, 7,
507–510.
Wallach, H. (1987). Perceiving a stable environment when one
moves. Annual Review of Psychology, 38, 1–27.
Wallach, H., & Canal, T. (1976). Two kinds of adaptation in
the constancy of visual direction. Perception & Psychophysics,
19, 445–449.
Welch, R. B. (1969). Adaptation to prism-distorted vision: The
importance of target-pointing. Perception & Psychophysics, 5,
305–309.
Yoshida, T., Ando, H., Maeda, T., & Watanabe, J. (2008).
Externalized sense of balance using galvanic vestibular stimu-
lation. Proceedings of 12th Annual Meeting of the Association
for the Scientific Study of Consciousness, 249–250.
556 PRESENCE: VOLUME 19, NUMBER 6
