185
Journal of Vestibular Research 15 (2005) 185–195
IOS Press
Vertical linear self-motion perception during
visual and inertial motion: More than
weighted summation of sensory inputs
W.G. Wrighta,b,∗ , P. DiZioa and J.R. Lackner a
a
b
Brandeis University, Ashton Graybiel Spatial Orientation Laboratory, Waltham MA 02454, USA
OHSU, Neurological Sciences Institute, Portland OR 97006, USA
Received 11 May 2004
Accepted 17 June 2005
Abstract. We evaluated visual and vestibular contributions to vertical self motion perception by exposing subjects to various
combinations of 0.2 Hz vertical linear oscillation and visual scene motion. The visual stimuli presented via a head-mounted
display consisted of video recordings of the test chamber from the perspective of the subject seated in the oscillator. In the
dark, subjects accurately reported the amplitude of vertical linear oscillation with only a slight tendency to underestimate it. In
the absence of inertial motion, even low amplitude oscillatory visual motion induced the perception of vertical self-oscillation.
When visual and vestibular stimulation were combined, self-motion perception persisted in the presence of large visual-vestibular
discordances. A dynamic visual input with magnitude discrepancies tended to dominate the resulting apparent self-motion, but
vestibular effects were also evident. With visual and vestibular stimulation either spatially or temporally out-of-phase with one
another, the input that dominated depended on their amplitudes. High amplitude visual scene motion was almost completely
dominant for the levels tested. These findings are inconsistent with self-motion perception being determined by simple weighted
summation of visual and vestibular inputs and constitute evidence against sensory conflict models. They indicate that when
the presented visual scene is an accurate representation of the physical test environment, it dominates over vestibular inputs in
determining apparent spatial position relative to external space.
Keywords: Otoliths, self-motion perception, sensory integration, vertical linear oscillation, virtual environment
1. Introduction
It has been known for over a century [10] that visual
stimulation can induce self-motion in stationary individuals. Actual self-motion usually involves a concordant combination of visual, vestibular, and somatosensory stimulation. Many investigators have examined
the contribution of each sensory input and of their combinations to self-motion perception (e.g. [2,3,7,9,12,
∗ Corresponding
author: W. Geoffrey Wright, Ph.D., Oregon
Health & Science University, Neurological Sciences Institute, OHSU
West Campus, Beaverton, OR 97006, USA. Tel.: +1 503 418 2602;
Fax: +1 503 418 2501; E-mail: wrightw@ohsu.edu.
14,16,17,25]). Most studies of perceived translation
have involved the horizontal plane, however, rectilinear
vertical acceleration (cf. [6,11,20]), an inertial stimulation that remains parallel to gravity and alters only
the magnitude of background force, has received little attention. In studying the visual influence on selfmotion perception, recent technological advances have
now made possible the use of rich, naturalistic visual
stimulation. These types of stimuli have unique characteristics, in terms of potential fidelity with real situations to make them interesting tools in the study of
visual and inertial contributions to space perception.
In the present study, we assessed the self-motion
perception of subjects exposed to vertical inertial body
motion combined with various types of scene motion.
ISSN 0957-4271/05/$17.00 © 2005 – IOS Press and the authors. All rights reserved
186
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Evidence from previous studies focusing on horizontal
plane, linear motion [1] found that subjects were always able to detect above vestibular threshold fore-aft
accelerations of a rolling cart when visual input was
concordant with the inertial motion. Detection of cart
accelerations dropped to 75–80% accuracy when subjects viewed a visual scene that was stationary in relation to the cart. During exposure to constant, unidirectional visual flow, detection of the direction of cart
motion reached 80% when visual flow was concordant
with the direction of cart motion, but was below 50%
with discordant flow. In related studies, Ohmi [14]
found that vestibular/somatosensory stimulation associated with inertial motion completely determined the
perceived direction of self-motion when visual input
was co-linear and 180 ◦ out-of-phase. However, when
visual motion was misaligned with inertial motion by
30◦ or 150◦ , visual input determined perceived selfmotion direction. Berthoz et al. [1] and Ohmi [14]
concluded that visual-vestibular/somatosensory interactions entail non-linearities and proposed explanatory
schemes in which visual and vestibular/somatosensory
signals are combined with their relative weights determined by the conflict between them. Earlier, a similar model was developed to explain visual and semicircular canal contributions to rotary self-motion perception. In this model, sensory channels are summated
but with gains dependent on stimulus frequency and
stimulus conflict [25]. Other findings from studies
involving sensory conflicts during linear motion concerning self-motion [16] and oculomotor responses [9]
are also incompatible with simple linear summation
models.
Our aim was to study sensory influences on vertical
self-motion perception using sinusoidal vertical linear
oscillation (VLO) and various combinations of visual
stimulation. During VLO the resultant gravitoinertial
force (GIF) direction remains constant. By contrast,
rectilinear acceleration in an earth-horizontal plane alters the magnitude and direction of the GIF and requires
the central nervous system to resolve the tilt/translation
ambiguity of the resulting vestibular afferent signals [1,
12,15,23]. The constant direction of GIF during VLO
ensured that the visual and the inertially dependent
stimuli were co-linear. This made for simpler interpretation of the results in terms of linear versus non-linear
interactions. We were also interested in maintaining a
high level of visual fidelity, which matches real world
visual stimulation as closely as possible. By using high
resolution, visual input of the actual test environment,
we introduced dynamic and pictorial depth cues, relative size, and polarizing orientation input.
Our approach was to establish baselines for vestibular and visual self-motion perception separately, and
subsequently to put visual and vestibular/somatosensory signals either in concordant or discordant relation to one another. The visual stimuli used in this
study were video recordings of the actual environment
in which the subject was present as seen from the test
situation rather than striped patterns as has been characteristic of many earlier studies. This allowed us to
determine whether visual stimuli representing the actual test situations have a more dominant influence on
self-motion perception than spatial stripe or dot patterns. Four amplitudes of inertial VLO and a fifth stationary condition were each tested against five different visual conditions including darkness, a stationary
visual scene, in-phase vertical visual oscillation at two
magnitudes, and counter-phase visual motion. During sinusoidal oscillation, peak acceleration occurs at
minimum velocity and vice versa. This is important
because visual motion detectors are predominantly velocity sensitive, and vestibular and somatosensory receptors are acceleration sensitive. We used an oscillation frequency of 0.2 Hz, in order to keep the visual,
vestibular, and somatosensory stimuli within the dynamic range of all three systems, thereby preventing
one of the inputs from dominating by default.
2. Materials and method
Subjects: Nine subjects, four males and five females
participated. The experimental protocol was approved
by the Brandeis Human Subjects Board, and all subjects signed an informed consent form. None of the
subjects had a history of vestibular or motor deficiencies as determined by a health survey and a motion sickness history questionnaire. Eight of the subjects were
undergraduate students between the ages of 18–23 who
were unaware of the apparatus’ capabilities. They were
paid for their participation. One subject (55 years old)
was a laboratory member who was familiar with the
apparatus but not with the current experimental goals.
Apparatus: The VLO device was a screw-driven
machine with a 15 horsepower motor controlled
by LabView computer software. A body-conforming
NASCAR driver’s seat with a five-point harness was
attached to the oscillator’s motion platform. The seat
aligned the subject’s z-axis parallel to gravity. The
device was programmed to oscillate at a frequency of
0.2 Hz in sinusoidal profiles with peak-to-trough amplitudes that could be varied from 0.2–1.6 m. The
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
corresponding peak inertial accelerations ranged from
0.16–1.26 m/s 2 .
A digital video camcorder was mounted on the VLO
device to create visual (V) stimuli. The camcorder was
used to record the surrounding room from the same
eye level perspective that a subject would view it from
while being oscillated. At the lowest point of travel the
camera was 1.5 m above the floor and at the highest
point 1.5 m below the ceiling (5 m high). The VLO
device is near the center of a 13 m × 18 m room, which
is filled with other equipment. The camcorder focal
length was set such that the recorded scene had no magnification relative to normal vision. Three different visual scenes were recorded. One scene depicted high
amplitude, 1.6 m peak-to-trough visual VLO at 0.2 Hz
with an accompanying sound track of the sound that
the VLO device makes during this motion. The visual
flow of this dynamic visual scene had a peak velocity
of 1.0 m/s. The second recorded scene depicted low
amplitude, 0.2 m peak-to-trough, 0.2 Hz visual VLO.
The 1.6 m/0.2 Hz machine noise was dubbed in synchrony with the visually depicted low amplitude VLO,
thus equalizing sound cues with that of the high amplitude visual scene. The peak velocity of this scene was
0.13 m/s. The final scene recorded was a stationary
view of the test room from the same perspective the
subject sees the room when seated in the VLO chair
at the stroke mid-point. The auditory signal associated
with chair oscillation at 0.2 Hz, 1.6 m amplitude was
dubbed onto the visual recording.
A Virtual Research VR4 head-mounted display
(HMD) was used to display the visual scenes. The VR4
is a light, counterweighted device, designed around
binocular 3.3 cm LCD displays. Its optics provide a
60◦ diagonal field-of-view (FOV) at full overlap (36 ◦
Height × 48◦ Width). The HMD optics allow for the
full range of depth focus from 25 cm to infinity. Binocular disparity offset was not used. The VR4 has built-in
stereo headphones.
Procedure: Subjects (except the one laboratory
member) were led into the test room blindfolded and
seated in the VLO chair before their eyes were uncovered. This prevented them from seeing the vertical rails
of the VLO device, which were behind them but when
the blindfold was removed let them view the room from
the location that the visual scenes were recorded.
The subjects were tested according to a quasirandomized, full-factorial, repeated-measures design
which included five amplitudes of inertial VLO and five
visual conditions (Table 1). All VLO and visual motion
stimuli were 0.2 Hz. The five inertial VLO amplitudes
187
Fig. 1. Depiction of average perceived self-motion for each visual condition using slopes and y-intercepts from linear regression.
The slope would equal one (diagonal dashed line) if the amplitude of self-motion perception were veridical. In darkness (Dark),
self-motion perception was slightly less than one (slope = 0.86). If
only visual input determined the amplitude of self-motion perception
then the slope would have been zero and the y-intercept would equal
1.6 (upper dotted line) for the visual condition depicting 1.6 m visual
motion (HiV) or 0.2 m (lower dotted line) for the visual condition
depicting 0.2 m (LoV) and 0 in SR (Stationary Room). This did
not occur. The slopes in the two dynamic visual conditions, LoV
and HiV, were significantly reduced relative to Dark, but not in the
stationary visual condition, SR. In the phase-shifted visual condition
(PSLo), the slope is equal to its un-shifted visual match condition
(LoV), but the y-intercept is greater. This cannot be explained by a
linear model, because the phase of perceived motion changed with
increasing inertial amplitude (not depicted here).
were 0, 0.2, 0.4, 0.8, and 1.6 m peak to trough displacement. The five visual conditions included 1) darkness,
2) a stationary visual scene, 3) naturally coupled 0.2 Hz
vertical oscillation at 1.6 m amplitude, and 4) naturally
coupled 0.2 Hz vertical oscillation at 0.2 m amplitude,
and 5) visual scene oscillation phase reversed relative
to inertial oscillation. Sessions were run on five different days with at least 48 hours separating test days.
Each session entailed exposing the subjects to one visual condition at five different amplitudes of VLO. The
order of presentation of visual conditions across sessions and of VLO amplitudes within sessions was randomized. Subjects were exposed to each amplitude of
inertial VLO for three minutes while viewing the scene.
The time between conditions was five minutes. Once
the subject had donned the HMD, we allowed them to
support their head and the HMD against the chair back
during testing with only minimal restraint. Because of
the counter-weighting of the HMD, this did not induce
188
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Table 1
Experimental design
Session
1
2
3
4
5
Visual condition
Dark
Stationary Room (SR)
LoV (0.2 Hz at 0.2 m peak-to-peak amplitude)
HiV (0.2 Hz at 1.6 m peak-to-peak amplitude)
180◦ Phase Shift of LoV (PSLo)
torque on the head during peak accelerations. This arrangement also reduced the possibility of movement of
the HMD on the head. Using the built-in video oculography device of the HMD, we confirmed that the
HMD did not move relative to the eye during peak accelerations. Eye-tracking also showed that there was
no measurable linear vestibular-ocular reflex at the inertial acceleration levels that were used. We minimized
transmission of vibration from the VLO apparatus to
the subject using dense foam between the back and buttocks and racecar seat. Additionally, we used a high
volume auditory recording from 1.6 m VLO to drown
out motor noise equally in all conditions.
During each three-minute test period, four dependent measures were reported: onset of apparent selfmotion, perceived amplitude of self-motion, the “compellingness” of self-motion, and path of self-motion.
Self-motion onset time was measured with a stopwatch.
Subjects reported their apparent peak to trough amplitude of self-motion in terms of feet or meters depending on which units they were most familiar with in every day life. If subjects were unsure of precise scalar
magnitudes, they were instructed to scale their judgments in each condition relative the one another. All
judgments in feet were later converted to meters for
analysis purposes. Subjects made judgments of amplitude within the first minute of a trial and again toward
the end of the three minute trial. In the few instances
where the judgments differed, they were averaged. The
reliability of our metric was verified by random repetitions of trials for a subject and comparing it to their
previous trial. One subject was also tested at all levels
in the same visual condition on separate days and no
significant changes in self-motion were reported.
A continuous relative rating scale was used to determine the level of compellingness of perceived selfmotion. This was included as an independent measure
of perceived amplitude of self-motion. It was intended
to provide a second measure of visual-vestibular interaction that was not dependent on the ability to judge
metric distances, which may include inherent perceptual biases. This scale was dependent on an individual’s ability to judge an internal state on a normalized
Inertial motion (0.2 Hz VLO)
Peak-to-peak amplitudes of 0, 0.2, 0.4, 0.8, 1.6 m
Peak-to-peak amplitudes of 0, 0.2, 0.4, 0.8, 1.6 m
Peak-to-peak amplitudes of 0, 0.2, 0.4, 0.8, 1.6 m
Peak-to-peak amplitudes of 0, 0.2, 0.4, 0.8, 1.6 m
Peak-to-peak amplitudes of 0.2, 0.4, 0.8, 1.6 m
scale. The subjects were instructed beforehand how
the scale was to be applied. The scale ranged from
0–5. Each subject was given examples of the type of
perception that could be related to varying levels of
“compellingness.” Zero equates to no compelling selfmotion. Compellingness greater than zero equates to
a sense of self-motion velocity and/or displacement.
Subjects were given examples to help them quantify the
subjective measure. The classic case of illusory motion
one experiences when a train/subway on an adjacent
track pulls away was described to the subject during
the instructions to provide an example of compelling
self-motion in the absence of real motion and how this
feeling can be very convincing but dissipate quickly.
They were instructed that high compellingness should
be equated to self-motion with both velocity and displacement that persists and is stable. They were informed that pressure cues, wind cues, and visceral cues
may be indicative of a heightened compellingness, but
are not necessary to experience self-motion. The highest levels of compellingness should include velocity
and displacement with a recognizable path and direction.
Path was a recorded as a separate variable describing
their perceived direction of self-translation, the shape
of their perceived motion path (linear, curvilinear, rotational), and their body orientation relative to perceived
vertical. They supplemented their verbal reports by indicating their apparent motion direction and body orientation with a small handheld cylindrical rod, which
they had been trained to use. The experimenter took
notes on the rod motion and verbal reports. Subjects
were also monitored periodically for symptoms of motion sickness. They were free to stop the experiment
at any time but none chose to do so and no significant
increase in the motion symptoms was found.
3. Data analysis
Repeated-measures ANOVAs (SPSS) were used to
test for effects of VLO amplitude, visual condition, and
interactions. Mauchley’s test of sphericity was used to
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Fig. 2. Subject-reported self-motion onset times (seconds) for each
visual condition at each level of inertial motion (VLO). All latencies
decrease with increasing levels of inertial motion. Visual motion
either high (HiV) or low (LoV and PSLo) induced faster self-motion
onset times relative to when no visual motion was viewed (Dark
and Stationary scene, SR). The pattern of results shows that increasing levels of stimulation for both visual and inertial inputs reduces
latencies.
make Greenhouse-Geisser epsilon adjustments to the
degrees of freedom where necessary. Non-parametric
statistics were used to analyze the data from compellingness reports and when compared to amplitude
data. The experienced subject showed no difference
in pattern of results from the naı̈ve subjects so all data
were analyzed together.
4. Results
The 25 experimental conditions involved different
combinations of visual and vestibular stimulation. The
five visual conditions were each paired with the five
amplitudes of VLO at 0.2 Hz (0, 0.2, 0.4, 0.8, 1.6 m).
For ease of presentation, the results will be described
in terms of the individual visual conditions. Figure 1
illustrates the overall pattern of results.
Dark. No motion was perceived in the absence of inertial motion. Perceived amplitude of motion paralleled
inertial motion amplitude in a nearly monotonic function. Subjects tended to underestimate the amplitude
of motion by about 15% (see Table 2a and Fig. 3a). A
regression analysis including all the data points showed
a linear fit with a slope of 0.86 and a 0 y-intercept (see
Table 3a).
The latency to reports of self-motion decreased as the
amplitude of inertial motion increased, and their means
189
ranged from 34.8 s at 0.2 m amplitude to 0.2 s at 1.6 m
amplitude (see Table 2b or Fig. 2). This is a nearly
exponential decrease of onset time with increasing inertial motion amplitude. These results represent the
contribution of vestibular and somatosensory signals to
apparent motion in the absence of visual stimulation.
Stationary Visual Scene of the Test Chamber (SR).
No motion was experienced in the 0m inertial amplitude
condition. As inertial motion amplitude increased from
0.2 to 1.6 m, the amplitude of reported self-motion
increased as well from 0.12 m to 1.04 m (Fig. 3b). As
in the Dark condition, there was always a tendency to
underestimate the amplitude of inertial motion. The
latency of reporting the onset of self-motion varied in
an exponential fashion from 27.8 s at 0.2 m amplitude
to 0.4 s at 1.6 m amplitude (Fig. 2). No subject reported
visual motion regardless of inertial motion amplitude.
A statistical comparison of the data for the Dark and
the Stationary Room visual scene conditions indicated
no difference for any of the comparisons. However, a
non-significant trend (p < 0.10, n.s.) was found for the
two highest amplitudes of inertial motion wherein selfmotion tended to be underestimated in the Stationary
Room scene condition relative to the Dark condition. A
binomial probably test showed it to be highly unlikely
(p < 0.01) that only three out of 18 Stationary scene
trials would have a greater amplitude of self-motion reported relative to the Dark condition at the two highest
inertial levels. Accordingly, if the discordance between
visual and inertial input were increased even more, the
stationary visual field might begin to significantly decrease perception of self-motion relative to when no
visual input is present.
Low Amplitude Vertical Visual Oscillation (LoV). On
average, subjects reported vertical oscillation of 0.17 m
amplitude in the 0 m inertial motion condition indicating that vertical visual oscillation can induce apparent
vertical oscillation in stationary subjects. As inertial
motion increased from 0.2 m to 1.6 m, the amplitude of
apparent self-motion increased as well but non-linearly
from 0.30 to 0.86 m. At the low amplitudes of inertial oscillation, 0.2 and 0.4 m, the perceived amplitudes of self-motion were 0.3 and 0.38 m, respectively.
At 0.8 and 1.6 m amplitude inertial motion, reported
self-motions were only half the actual displacements
(Fig. 3c). Statistical comparisons with the Dark condition indicate that at 0.8 and 1.6 m inertial motion,
the 0.2 m visual amplitude suppresses apparent displacement of the body. The latency to reports of selfmotion decreased linearly as inertial motion amplitude
increased.
190
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Table 2
(a) Perceived amplitude of self-motion (meters)
Visual condition
Dark
Stationary Room (SR)
Low Amp (LoV)
High Amp (HiV)
Phase-shifted (PSLo)
0m
0.00
0.00
0.17
1.14
n/a
=
=
<
=
0.2 m
0.08
0.12
0.30
1.43
0.34
Amplitude of VLO
0.4 m
0.8 m
<
0.34
<
0.69
0.23 0.49
=
0.38
=
0.47
=
1.47
=
1.53
<
0.70
=
0.73
<
<
<
<
<
1.6 m
1.30
1.04
0.86
1.75
1.04
(b) Self-motion onset time (seconds)
Visual condition
Dark
Stationary Room (SR)
Low Amp (LoV)
High Amp (HiV)
Phase-shifted (PSLo)
0m
n/a
n/a
23.3
10.4
n/a
=
=
0.2 m
34.8
27.8
12.3
4.7
14.7
Amplitude of VLO
0.4 m
0.8 m
>
13.1
=
2.8
>
9.3
3.1
=
7.2
=
2.7
=
3.3
=
0.7
5.3
=
2.9
=
=
=
1.6 m
0.2
0.4
1.8
0.3
2.6
(”>” and “<” symbols indicate statistical significance at the 0.05 level. No statistical
difference is indicated by an “=” sign. Marginal significance, 0.10 > p > 0.05, is indicated
by a “” or “” symbol.)
Table 3
(a) Linear regression of perceived self-motion amplitude for each visual condition relative to VLO amplitude
Visual condition
Dark
Stationary Room (SR)
Low Amp (LoV)
High Amp (HiV)
Phase-shifted (PSLo)
Slope
(95% Confidence interval)
0.7 < 0.86 < 1.0
0.5 < 0.67 < 0.8
0.2 < 0.41 < 0.6
−0.1 < 0.31 < 0.8
0.3 < 0.49 < 0.7
y-intercept
(95% Confidence interval)
−0.2 < −0.02 < 0.1
−0.2 < −0.02 < 0.15
0.04 < 0.19 < 0.3
0.9 < 1.28 < 1.7
0.1 < 0.31 < 0.5
Pearson
correlation
0.82
0.72
0.58
0.21
0.59
F
(df = 43)
90.7
46.0
21.6
1.88
23.0
p-value
< 0.001
< 0.001
< 0.001
> 0.10, n.s.
< 0.001
(b) Average visual and vestibular weights using linear weighting
Visual condition
Dark
Stationary Room (SR)
Low Amp (LoV)
High Amp (HiV)
Phase-shifted (PSLo)
0m
Wvis
Wvest
–
–
–
–
0.85
–
0.72
–
–
–
0.2 m
Wvis
Wvest
–
0.40
0.40
0.60
–
–
0.88
0.12
1.3
−0.3
High Amplitude Vertical Visual Oscillation (HiV).
Vertical self-motion was perceived at all inertial amplitudes and was of comparable magnitude, ∼ 1.4 m, at all
inertial amplitudes except 1.6 m, where it was significantly higher at 1.75 m (Fig. 3d). This visual condition
was also characterized by the shortest latency to reports
of self-motion for any of the other visual conditions
except at 1.6 m inertial amplitude where latencies were
short for all of the visual conditions (Fig. 2). A comparison of this 1.6 m Visual condition with the Dark
condition indicated that the amplitude of self-motion
reported was greater for all inertial amplitudes of motion.
Amplitude of VLO
0.4 m
Wvis
Wvest
–
0.85
0.42
0.58
0.10
0.90
0.89
0.11
1.2
−0.2
0.8 m
Wvis
Wvest
–
0.86
0.39
0.61
0.55
0.45
0.91
0.09
0.41
0.59
1.6 m
Wvis
Wvest
–
0.81
0.35
0.65
0.53
0.47
–
–
0.59
0.41
Visual Oscillation Counterphase to Inertial Motion
(PSLo). With visual and inertial motion of comparably low magnitudes but of opposite directions, the
direction of apparent self-motion was driven by the
visual input. However, at the highest levels of inertial amplitude (0.8 m and 1.6 m), the majority of selfmotion reports (10/18) were in phase with inertial motion. At 1.6 m inertial amplitude, the magnitude of apparent self-displacement was significantly less than in
the Dark condition involving only inertial motion, thus,
it is apparent that the visual input still had an effect, in
that, it reduced the amplitude of perceived self-motion
(Fig. 3e) despite the fact that for most subjects it did
not determine phase. Moreover, at 1.6 m inertial am-
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
191
Fig. 3. Average amplitudes of reported self-motion for each visual condition across increasing amplitudes of sinusoidal vertical oscillation (error
bars: ± 1 SE). The diagonal (dashed-line) indicates where veridical reports of self-motion would be, if only inertial input drove self-motion
perception, however visual effects are evident. a. Dark condition, b. The visual scene displayed in the HMD is a first-person perspective of the
room in which the experiment is conducted and remains stationary relative to the subject at all times (SR), c. Low amplitude (0.2 m) oscillation
(LoV) of room is visually displayed in HMD, d. High amplitude (1.6 m) of oscillation visually displayed (HiV), e. Low amplitude (0.2 m) visual
oscillation phased-shifted 180◦ relative to inertial VLO is displayed (PSLo).
plitude all of the subjects initially had some difficulty
“deciding” on their direction of motion. Direction of
motion was determined by verbal responses made at
the moment of perceived direction change. Subjects
could also indicate their direction with the joystick. A
number of subjects showed difficulty in coordinating
their perceived direction of self-motion with voluntary
movement of the joystick, in that they reported one
phase but continued to move the joystick opposite their
verbal report. Others recognized their error without
recognizing that visual and inertial motion were outof-phase. This type of confusion only occurred in the
PSLo visual condition. These results indicate visual
dominance at amplitudes up to 0.8 m but a diminished
non-uniform influence at 1.6 m. Latencies to reports of
self-motion were significantly lower than in the Dark
condition, at 0.2 m and 0.4 m amplitudes, but were not
different for 0.8 m and 1.6 m (Fig. 2).
Compellingness of Self-motion. The results from the
compellingness ratings (Fig. 4) were significantly correlated (Spearman Rank Order correlations) with inertial amplitude of oscillation and perceived amplitude
of self-motion in all visual conditions. The dark condition showed the highest correlation with inertial amplitude (rs = 0.89) relative to other visual conditions.
It also showed the highest correlation (r s = 0.88) with
subject-matched ratings of self-motion amplitude in the
dark condition, which suggests that the internal metric
for compellingness of self-motion is comparable to perceived amplitude of self-motion. All other visual conditions showed significant correlations (p < 0.05) with
subject-matched ratings of self-motion amplitude in the
respective visual condition (SR: r s = 0.85; HiV: rs =
0.38; LoV: rs = 0.56; PSLo: rs = 0.39). Using the
Kendall coefficient of concordance (W K ), inter-subject
reliability with the compellingness scale was found to
be very high in all visual conditions (DK: W K = 0.93;
SR: WK = 0.89; HiV: WK = 0.75; LoV: WK = 0.74;
PSLo: WK = 0.60). Together these correlations provide independent evidence that as inertial amplitude
of motion is increased, the intensity of perceived selfmotion increases, which correlates with greater amplitude of perceived self-motion. Additionally, when visual motion is temporally concordant with inertial motion self-motion is perceived as more compelling than
when temporally discordant (i.e. HiV and LoV relative
to SR and PSLo) (Wilcoxon Matched Pairs test, p <
0.05).
192
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Fig. 4. Average ratings of compelliness of perceived self-motion on a
scale from 0–5 for each visual condition across increasing amplitudes
of sinusoidal vertical oscillation. The Dark and SR visual conditions
show linear increase with inertial amplitude. Both HiV and LoV
were rated as significantly more compelling than SR and PSLo.
5. Discussion
Our experimental paradigm provided assessments of
the separate influences of visual and inertial cues on
vertical self-motion perception and of their interaction
when visual and inertial cues were concordant or were
mismatched in amplitude or direction.
In the condition which assessed perception of selfmotion during purely inertial motion, we found that not
all of our subjects reported self-motion at the 0.2 m,
0.2 Hz Dark condition (16 cm/s 2 ). In other vertical
oscillation studies, these values have generally been
above threshold for test subjects, for example the 2–
10 cm/s2 reported by Gurnee [6]. However, an important feature of our experimental paradigm was that the
subjects, except for one, had no foreknowledge of the
capability of the test device. They initially knew they
would be subjected to motion but not on which axis.
At above threshold exposures, subjects underestimated
the amplitude of their motion and eye-tracking measurements showed that there was no measurable linear
vestibular-ocular reflex at the inertial acceleration levels tested. A stationary visual scene slightly attenuated
the perception of motion.
Lower and higher amplitude visual motion in phase
with the inertial stimuli clearly affected the amplitude
of experienced self-motion, decreasing or increasing it
respectively, and in the absence of inertial motion in-
duced illusory vertical oscillatory self-motion. Reversal of visual motion relative to inertial motion generated a complex pattern of experienced motion with the
visual input determining the direction of experienced
self-motion at all but the highest inertial amplitudes.
Our study included two independent rating scales for
describing self-motion perception. One was based on
an externally referenced metric scale of distance traveled, the other based on an internally reference scale
of motion compellingness. Non-parametric statistics
substantiated that both measures were significantly correlated with changes in inertial amplitude, except for
the highly visually driven HiV condition. Because
our experimental design did not follow a standard psychophysical methodology, biases inherent in psychometric measures may have been present. However,
we consider the robustness of our results reliable because we employed dual independent measures, we
used a repeated measures design, which reduces the effects of variance across subjects, and we applied nonparametric statistics, which are well-suited for identifying rank changes in ratings.
Visual-vestibular models of self-motion perception.
We examined our results in relation to two models.
The first was a general linear regression model where
a least-squares line fit and y-intercept were determined
for each visual condition considered separately (see Table 3a). The results of this model show that inertial
motion accounted for a highly significant level of variability for each visual condition (p < 0.001), except for
the high amplitude visual condition (HiV) where the
slope and intercept show that visual input dominates.
A decrease in slope is seen in the low amplitude visual
condition (LoV), which reflects its suppressive effect
on perception of inertial motion. The larger slopes and
0-intercepts found in the dark and stationary visual field
conditions faithfully reflect the dominance of inertial
input. In the phase-reversed visual condition (PSLo), a
regression line could be fitted to the reports of perceived
amplitude, however, the shift from visual to vestibular
dominance as inertial amplitude increased rendered the
results less meaningful.
We also evaluated a model of the relative contribution of visual and inertial inputs to perception of vertical
self-motion that employed linear weightings and summation of visual and vestibular/somatosensory inputs
(see Table 3b).
Wvis ∗ VelVIS dt + Wvest ∗
aVLO dt2
(1)
= Amplitude of Perceived SM
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
Wvis ∗ (Amp. VE scene) + Wvest ∗(Amp. VLO)
= Amplitude of Perceived SM
Wvis + Wvest = 1
(2)
(3)
The simple linear model included two weighting
factors, visual weight (W vis ) and vestibular weight
(Wvest ), where the sum of the weights always equals
one (Eq. 3). For visual conditions where only one input
was present, the relevant weight can be interpreted as a
gain value (e.g. W vest = 0.85 at 0.4 m VLO in the Dark,
thus the gain of the vestibular system here is 0.85).
Wvis weights the effect of visual input on perceived
self-motion amplitude. This
can be interpreted as the
integral of visual velocity ( VelVIS ), and corresponds
to the experimenter controlled visual motion amplitude
(see Visual Condition in Table 1). Wvest weights the
effect of vestibular/ somatosensory input on perceived
self-motion amplitude. This
input is the double integral
of inertial acceleration ( aVLO dt2 ) and corresponds
to the amplitude of VLO (see Inertial Motion in Table 1). The total amplitude of perceived self-motion is
the sum of these two weighted inputs. This model gives
reasonable fits for the visual conditions in which there
was in-phase motion for different inertial motions, but
the weightings had to be changed for the low amplitude
versus high amplitude visual motion conditions. It simply failed for the conditions involving phase reversal of
visual and inertial motion.
Because both models inadequately account for the
phase-reversed visual condition, an alternative to
weighted linear models needs to be considered. Their
failure might follow because summation of acceleration and velocity inputs does not result in directional
determinacy. Visual motion detection is velocity sensitive and its direction is determined by optic flow,
whereas, graviceptive motion detection is acceleration
sensitive. However, the direction of acceleration is consistent with many velocities including oppositely directed ones. Therefore, it is possible that the representation of graviception is labile enough to be mapped to
many velocity profiles, with its primary function being
magnitude indication. The perceptual specification of
self-motion and orientation might not just be governed
by raw sensory thresholds and inputs, but perhaps attentional and higher level cognitive factors help disambiguate a dynamic state, which could otherwise be consistent with many velocities and acceleration profiles.
An alternative to a differential weighted linear model
is conflict modeling. Zacharias and Young [25] developed a conflict model of visual-vestibular interactions
193
and self-motion perception in which they proposed that
at low levels of stimulation and conflict vision tends to
dominate. Higher amplitudes of vestibular stimulation
shift the balance toward vestibular dominance. This
model successfully accounts for a number of experimental findings involving both rotary and linear visual
and vestibular stimulation [2,8,24,25] including the elevation of thresholds for detection of horizontal linear
acceleration by conflicting visual stimulation and their
lowering by conforming visual stimulation (cf. [1]).
Such a conflict model, however, cannot account for
the findings reported here. In our high amplitude (1.6 m)
visual oscillation condition, the level of conflict varied
from zero when inertial input was 1.6 m, to greater levels for inertial inputs of 0.8 m, 0.4 m, 0.2 m, and 0 m amplitude. Yet, the reported self-motion remained roughly
constant across conditions showing only a slight proportionality to inertial amplitude. For the low amplitude visual oscillation condition, as conflict increased
self-motion perception switched from vestibular to visual dominance. Furthermore, with phase reversed visual motion, a reduction or cancellation of self-motion
did not occur. These patterns are the opposite of what
the model would predict.
Visual-vestibular reciprocal innervation. Studies of
visual-vestibular interactions using brain imaging techniques have raised the possibility of reciprocal interactions between visual and vestibular projection areas of
the brain affecting self-motion perception [3,4,22]. Rotating visual scenes activate medial, parietal-occiptal
cortex (PO) and simultaneously decrease activity levels
in parieto-insular vestibular cortex (PIVC). Activation
of PIVC by caloric irrigation is accompanied by decreased activity levels in PO. These physiological findings cannot adequately account for our experimental
findings either. For example, with phase reversed visual and inertial stimulation (PSLo), one would expect
dual reciprocal inhibition and thus attenuation of experienced self-motion. However, the amplitude of selfmotion our subjects experienced in this circumstance
was greater or not statistically different from conditions
involving only inertial motion (Dark). Some evidence
exists indicating that linear acceleration does not evoke
inhibition in PO [13] as it appears to during rotational
stimulation. Imaging studies of visuo-vestibular interactions are still at an early stage and it may be unrealistic to expect them to explain very much at this
time.
The visual stimuli we used in our experimental
paradigm differ from those used in many previous studies of visuo-vestibular interactions. Earlier studies have
194
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
tended to use patterns of stripes, dot clusters, or geometric figures. We used video recordings of the actual
experimental chamber recorded from the location of a
subject’s head along the oscillator track, presented to
the subject by means of a head-mounted display. Although increasing the FOV of the HMD display would
likely have increased the effect of the visual scene on
self-motion induction, several studies have shown that
the level of realism depicted in visual scenes and the
presence of depth cues also can enhance the effectiveness of a display in inducing linear vection and changes
in apparent heading (e.g. [5,14,18,19,21]). The strong
influence of the visual display of the test chamber in
our experiment, across all inertial motion conditions,
indicates an important contribution of higher level cognitive factors to perceived self-motion. Knowledge of
physical constraints, such as awareness of where one’s
body is relative to a solid floor or ceiling, size of the
common objects within the room and their relative distances in the visual scene, as well as visual orientation
cues, all contributed to the direction and magnitude of
apparent motion. But more fundamentally, the visually
presented view of the actual test chamber, a chamber
which cannot physically move, served as a spatial anchor in a way that arbitrary bars, stripes, or spots do
not and dominated the subjects self motion perception
over a broad range of dynamic conditions.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
Acknowledgements
[17]
This work was supported by Air Force grant #
F49620110171 and NASA Grant NAG9-1263. Dr. Simone Bortolami designed the vertical linear oscillator
used in this experiment.
[18]
[19]
[20]
References
[21]
[1]
[2]
[3]
[4]
D.E. Angelaki, Three-dimensional organization of otolithocular reflexes in rhesus monkeys. III. Responses to translation, Journal of Neurophysiology 80 (1998), 680–695.
A. Berthoz, B. Pavard and L.R. Young, Perception of linear horizontal self-motion induced by peripheral vision (Linearvection): Basic characteristics and visual-vestibular interactions, Experimental Brain Research 23 (1975), 471–489.
T. Brandt, P. Bartenstein, A. Janek and M. Dieterich, Reciprocal inhibitory visual-vestibular interaction – Visual motion
stimulation deactivates the parieto-insular vestibular cortex,
Brain 121 (1998), 1749–1758.
T. Brandt and M. Dieterich, The vestibular cortex: Its locations, functions, and disorders, Annals of the New York
Academy of Sciences 871 (1999), 293–312.
[22]
[23]
[24]
J.-Y. Frigon and A. Delorme, Roll, pitch, longitudinal and yaw
vection visually induced by optical flow in flight simulation
conditions, Perceptual and Motor Skills 74 (1992), 935–955.
H. Gurnee, Thresholds of vertical movement of the body, Journal of Experimental Psychology 17 (1934), 270–285.
L.R. Harris, M. Jenkin and D.C. Zikovitz, Visual and nonvisual cues in the perception of linear self motion, Experimental Brain Research 135(1) (2000), 12–21.
J. Huang and L.R. Young, Sensation of rotation about a vertical
axis with a fixed visual field on different illumination and in
the dark, Experimental Brain Research 41 (1981), 172–183.
C.E. Lathan, C. Wall and L.R. Harris, Human eye movement
response to z-axis linear acceleration: the effect of varying the
phase relationships between visual and vestibular inputs, Exp
Brain Res 103 (1995), 256–266.
E. Mach, Grundlinien der Lehre von den Bewegungsempfindungen, Engelmann, Leipzig, 1875.
R. Malcolm and G. Melvill Jones, Erroneous perception of
vertical motion by humans seated in the upright position, Acta
Otolaryngologica 77 (1974), 274–283.
D.M. Merfeld, W. Teiwes, A.H. Clarke, H. Scherer and L.R.
Young, The dynamic contributions of otolith organs to human
ocular torsion, Exp Brain Res 110 (1996), 315–321.
S. Nishiike, S. Nakagawa, A. Nakagawa, A. Uno, M. Tonoike,
N. Takeda and T. Kubo, Magnetic cortical responses evoked by
visual linear forward acceleration, Neuroreport 13(14) (2002),
1805–1808.
M. Ohmi, Egocentric perception through interaction among
many sensory systems, Cognitive Brain Research 5 (1996),
87–96.
G.D. Paige and D.L. Tomko, Eye movement responses to linear
head motion in the squirrel monkey. I. Basic characteristics,
Journal of Neurophysiology 65 (1991), 1170–1182.
T. Probst, A. Straube and W. Bles, Differential effects of ambivalent visual-vestibular-somatosensory stimulation on the
perception of self-motion, Behavioral Brain Research 16
(1985), 71–79.
A. Straube and T. Brandt, Importance of the visual and vestibular cortex for self-motion perception in man (circularvection),
Human Neurobiol 6 (1987), 211–218.
L. Telford and B.J. Frost, Factors affecting the onset and
magnitude of linear vection, Perception & Psychophysics 53
(1993), 682–692.
L. Telford, J. Spratley and B.J. Frost, Linear vection in the
central visual field facilitated by kinetic depth cues, Perception
21 (1992), 337–349.
E.G. Walsh, The perception of rhythmically repeated linear
motion in the vertical plane, Quarterly Journal of Experimental Physiology 49 (1964), 58–65.
W.H. Warren and K.J. Kurtz, The role of central and peripheral
vision in perceiving the direction of self-motion, Perception
& Psychophysics 51 (1992), 443–454.
R. Wenzel, P. Bartenstein, M. Dieterich, A. Danek, A. Weindl,
S. Minoshima, S. Ziegler, M. Schwaiger and T. Brandt, Deactivation of human visual cortex during involuntary ocular
oscillations: A PET activation study, Brain 119 (1996), 101–
110.
A.H. Wertheim, B.S. Mesland and W. Bles, Cognitive suppression of tilt sensations during linear horizontal self-motion
in the dark, Perception 30 (2001), 733–741.
L.R. Young, J. Dichgans, R. Murphy and T. Brandt, Interaction
of optokinetic and vestibular stimuli in motion perception,
Acta Otolaryngologica 76 (1973), 24–31.
W.G. Wright et al. / Vertical linear self-motion perception during visual and inertial motion
[25]
G.L. Zacharias and L.R. Young, Influence of combined visual and vestibular cues on human perception and control of
195
rotation, Experimental Brain Research 41 (1981), 159–171.