Exp Brain Res (1997) 113:475–483
R E S E A R C H A RT I C L E
&roles:John J. Jeka · Gregor Schöner · Tjeerd Dijkstra
Pedro Ribeiro · James R. Lackner
Coupling of fingertip somatosensory information to head and body sway
&misc:Received: 25 March 1996 / Accepted: 6 August 1996
&p.1:Abstract Light touch contact of a fingertip with a sta-
tionary surface can provide orientation information that
enhances control of upright stance. Slight changes in con-
tact force at the fingertip provide sensory cues about the
direction of body sway, allowing attenuation of sway. In
the present study, we asked to which extent somatosenso-
ry cues are part of the postural control system, that is,
which sensory signal supports this coupling? We investi-
gated postural control not only when the contact surface
was stationary, but also when it was moving rhythmically
(from 0.1 to 0.5 Hz). In doing so, we brought somatosen-
sory cues from the hand into conflict with other parts of
the postural control system. Our focus was the temporal
relationship between body sway and the contact surface.
Postural sway was highly coherent with contact surface
motion. Head and body sway assumed the frequency of
the moving contact surface at all test frequencies. To ac-
count for these results, a simple model was formulated by
approximating the postural control system as a second-or-
der linear dynamical system. The influence of the touch
stimulus was captured as the difference between the ve-
locity of the contact surface and the velocity of body
sway, multiplied by a coupling constant. Comparison of
empirical results (relative phase, coherence, and gain)
with model predictions supports the hypothesis of cou-
pling between body sway and touch cues through the ve-
locity of the somatosensory stimulus at the fingertip. One
J.J. Jeka (✉) · P. Ribeiro
Department of Kinesiology, University of Maryland,
College Park, MD 20742–2611, USA;
e-mail: jj96@umail.umd.edu
G. Schöner
Laboratoire de Neurosciences Cognitives,
Center National de la Recherche Scientifique, Marseille, France
J.R. Lackner
Ashton Graybiel Spatial Orientation Laboratory,
National Center for Complex Systems, Brandeis University,
Waltham, MA 02254, USA
T. Dijkstra
Department of Psychology, University of Pennsylvania,
Philadelphia, PA 19104, USA&/fn-block:
© Springer-Verlag 1997
subject, who perceived movement of the touch surface,
demonstrated weaker coupling than other subjects, sug-
gesting that cognitive mechanisms introduce flexibility
into the postural control scheme.
&kwd:Key words Posture · Somatosensation · Fingertip ·
Entrainment · Velocity · Human&bdy:
Introduction
Flexible control of upright posture during quiet stance
and locomotion requires information about the contact
forces and textural properties of support surfaces derived
from somatosensory receptors, namely, muscle proprio-
ceptors and cutaneous afferents (Martin 1967; Horak and
MacPherson 1995). The importance of somatosensory
information becomes apparent when we encounter a slip-
pery surface and immediately adopt a more rigid posture
(e.g., “walking on eggs”) to compensate for the de-
creased friction of the support surface. However, our un-
derstanding of somatosensory contributions to upright
stance and locomotion is primarily empirical (Dietz
1992), with relatively few theoretical models to concep-
tualize extensive experimental findings.
We have developed a paradigm that separates the me-
chanical support provided by contact of the hand with a
rigid surface and the somatosensory cues provided by
such contact. When a standing subject lightly touches a
stationary surface with a fingertip, body sway is reduced,
even when contact forces are inadequate for mechanical
support of the body (Holden et al. 1994; Jeka and Lack-
ner 1994, 1995). The relationship observed between
body sway and the pattern of forces at the fingertip indi-
cates that subjects use slight changes in contact force at
the fingertip to gain information about the direction of
body sway, which allows attenuation of sway through
appropriate postural muscle activation (Jeka and Lackner
1995). These results indicate that somatosensory infor-
mation is a powerful orientation reference in the control
of human upright stance.
476

It is not clear, however, how somatosensory informa-

tion acts on posture. In particular, one must ask whether
somatosensory information forms part of the control
system, so that it couples in “closed loop” into the pos-
tural system. In the present study, we address this issue
with a method similar to the “moving room” paradigm
that has been used to demonstrate the coupling of visual
information with whole-body posture (Lee and Lishman
1975; Berthoz et al. 1979; Soechting and Berthoz 1979;
van Asten et al. 1988; Dijkstra et al. 1994a, b; Peterka
and Benolken 1995). Humans initiate responses that re-
duce body sway relative to movement of the visual
world, and such compensation is thought to be driven by
velocity-sensitive visual mechanisms (Schöner 1991). It
is not known, however, how dynamic somatosensory in-
formation couples to body sway. Experimental results
have shown that passive displacements of the arm
(Brandt et al. 1977) and passive stimulation of the feet
while sitting (Lackner and DiZio 1984) can result in
compelling illusions of self-rotation. With a moving con-
tact surface at the feet or hands, changing contact forces
can be perceived as body movement, movement of the
contact surface, or both (Lackner and DiZio 1992). Is
there, however, a coherent relationship between body
sway and dynamic somatosensory cues? In preliminary Fig. 1 A schematic diagram of the experimental situation. A sub-
empirical work, we found that contact of the fingertip ject is pictured in the tandem Romberg posture on the force plat-
form contacting the touch bar with the right index finger. The
with an oscillating surface led the center of pressure to touch bar either was stationary or moved sinusoidally in the medi-
assume the frequency of the moving surface (Jeka et al. al-lateral plane at a constant frequency (0.1, 0.2, 0.3, 0.4, and
1994). Here, we report a detailed account of the temporal 0.5 Hz) and amplitude (≈ 4 mm). For illustration, the subject is
relationship between dynamic somatosensory cues at the shown exceeding the threshold force of 1 N and the alarm is
sounding. In the actual experiment, the threshold was never ex-
fingertip and head/body sway that demonstrates strong ceeded. A video camera mounted to the ceiling directly above the
entrainment of the head and the body with the somato- subject’s head was used to measure head movements&ig.c:/f
sensory cues. We compare our results with those of a
model that predicts coupling of body sway to the veloci-
ty of the somatosensory stimulus. and Fz) registered by piezo-electric crystals in the corners of the
force platform.

Materials and methods Head motion

Subjects Medial-lateral head displacement (Hx) was measured with an

Five individuals participated, two women and three men, ranging
in age from 20–30 years. All subjects were healthy and physically
active, with no known musculoskeletal injuries or neurological
disorders that might have affected their ability to maintain bal-
ance. All were right-handed. The procedures used in the experi-
ment were approved by the Institutional Review Board at the Uni-
versity of Maryland. Informed written consent was obtained from
all participants in the study.
Apparatus and measures
ISCAN video system. A rigid, hollow aluminum tube attached to
an adjustable headband protruded 5 cm outward from the subject’s
forehead in the sagittal plane. A light-emitting diode (LED) was
attached to the end of the tube, and an ISCAN camera mounted on
the ceiling tracked the movement of the LED to measure Hx dis-
placement. The ISCAN system measures two dimensional move-
ment in a field of view 512×256 pixels. The camera was mounted
to the ceiling above the subject’s head. Because of differing sub-
ject heights, we normalized the field of view across subjects by
measuring the distance between the camera and LED and comput-
ing a calibration factor for each subject. The mean resolution
across subjects was 0.48 mm (Hx).

Figure 1 depicts the test situation, with a subject standing in a tan-

dem Romberg position (heel-to-toe) on a force platform, touching
with their right index fingertip a device used to measure the forces
applied.
Center of foot pressure
The force platform (Kistler Model 9261A) measured the ground
reaction forces through the feet. Medial-lateral coordinates of foot
pressure (CPx) were computed from the force components (Fx, Fy,
Fingertip contact forces
The “touch device” that the subject contacted with his or her index
finger consisted of a horizontal metal bar (46 cm×1 cm×2 cm) sup-
ported by a metal stand. A piece of white tape marked the point of
fingertip contact at the midpoint of the bar. The bar was positioned
parallel to the sagittal plane of the subject, with the stand resting on
a rigid wooden platform (155 cm×70 cm) that overlay the force
plate and extended beyond its lateral edges. The touch-device appa-
ratus on one side of the platform was balanced by a comparable
 477
mass on the other side (see Fig. 1). If the touch bar rested on the
floor independent of the force platform, then center-of-pressure
movements could be due to either actual body sway or to forces ap-
plied to the touch bar. Placing the touch bar on the wooden plat-
form insured that movements of the center of pressure reflected on-
ly movements and accelerations of the center of mass.
Two dual-element, temperature-compensated strain gauges
[Kulite Semiconductor, Type M(12) DGP-350–500] mounted on
the metal bar transduced the lateral (FL) and vertical (FV) forces
applied by the finger. The strain-gauge signals were amplified and
calibrated in units of force (newtons) and a comparator could trig-
ger an auditory tone when a specified threshold force was reached.
A computer-controlled stepper motor (Compumotor SX-5751)
was attached to the touch bar to move the bar in the medial-lateral
plane with a resolution of 0.0005 mm. One end of the touch bar
slid back and forth across a lubricated, smooth metal surface,
while the other end was held in place and served as an axis of rota-
tion in the medial-lateral plane. The stepper motor was pro-
grammed to move the bar sinusoidally at a constant amplitude
(4 mm peak-to-peak at the point of fingertip contact) and at differ-
ent frequencies. A potentiometer attached to the touch bar generat-
ed an analog signal of its movement, which was digitized for anal-
ysis. Even though the same amplitude was programmed for all fre-
quencies, the recorded amplitude was 3.32 mm (0.1 Hz), 3.56 mm
(0.2 Hz), 3.4 mm (0.3 Hz), 3.14 mm (0.4 Hz), and 2.82 mm
(0.5 Hz).

Procedure

The subject stood with right foot behind left along the center of
the anterior-posterior axis of the force platform. Adhesive tape
was used to mark the position of the feet on the platform, so that
the same foot position could be repeated throughout the entire ex-
periment. The touch bar was then adjusted to a comfortable height
(approximately waist level) and distance laterally from the subject,
to make contact with the right fingertip.
The subject’s task was to maintain the tandem stance with eyes
closed while keeping the fingertip force on the touch bar below
1 N. The touch bar was either stationary or moving at one of five
frequencies: 0.1, 0.2, 0.3, 0.4, or 0.5 Hz (total of six conditions).
The subject was not told that the touch bar could move. All sub-
jects completed every trial without ever exceeding the 1-N thresh-
old.
Before each trial, the subject took as much time as desired to
assume a comfortable stance, with their fingertip on the stationary
touch bar. Once ready, the subject closed their eyes and said “Go!”
and the experimenter initiated data acquisition that simultaneously
initiated touch-bar movement (i.e., on the moving bar trials). A
computer-generated tone signaled the beginning and end of the tri-
al. The subject stepped off the platform and rested for at least
2 min in between trials. The experimental trials were run in three
blocks of six (one trial of each condition per block) for a total of
18 trials. Conditions were randomized within a block. Trial dura-
tion was 80 s and all signals were collected in real time at 60 Hz.
The experiment lasted approximately 1 h.
Analysis
The experimental posture, tandem Romberg (heel-to-toe), was cho-
sen to enhance medial-lateral body sway. In previous experiments,
we found that the horizontal and vertical forces at the fingertip are
uncorrelated with anterior-posterior body sway during heel-to-toe
stance (Jeka and Lackner 1994, 1995). Consequently, we report here
measures related only to medial-lateral CPx and Hx displacement.
Linear systems analysis
Spectral analysis was performed on all signals; the magnitude-
squared coherence (MSC), the gain, and the phase were calculat-
Fig. 2 Overlaid time series of center-of-pressure (CPx), head
(Hx), and touch-bar (TBx) displacement in: a a stationary-bar con-
dition; and b a 0.5-Hz moving-bar condition&ig.c:/f
ed. The MSC is a measure of the strength of locking of body
sway to movements of the touch bar. MSC increases to a maxi-
mum of 1 with an increase in coupling strength. All spectra were
evaluated only at the driving frequency, as no systematic peaks
were observed at other frequencies. All spectra were calculated
with a Welch procedure (Marple 1987) in order to obtain consis-
tent estimates. We used 7 segments and factor 3 zero-padding&1fn.1: .
Gain was defined as the ratio of the CPx amplitude spectrum to
the touch-bar amplitude spectrum at the driving frequency. This
is a measure of gain relative to the surround (gain in the
“world”), unlike in systems theory where gain refers to the ratio
of the output signal over the input signal (gain at the “sensory
level”).
Time series of relative phase
We calculated a discrete time series of relative phase between:
CPx displacement and touch-bar displacement (TBx); and Hx dis-
placement and touch-bar displacement. The relative phase be-
1 The Welch procedure reduces the variance of the spectral esti-
mate by breaking the signal into overlapping segments and averag-
ing the spectral estimates of each segment. Each segment had one-
quarter of the length of the trial and was shifted by one-eighth of
the trial length relative to the neighboring segment (i.e., overlap-
ping segments). Factor 3 zero-padding made each segment equal
in length to the entire trial so that the sampling resolution of the
spectral estimate of each segment was equal to that of the full trial.
The choice of 7 segments is a compromise between the maximum
number of segments to achieve a good estimate of coherence and
the maximum number of cycles of data within a segment to get a
good estimate of phase (i.e., more segments mean shorter seg-
ments of actual data).&/fn:
478
Fig. 3 Spectral plots of CPx,
Hx and TBx in each condition.
Note that CPx and Hx dis-
placement frequencies are
broadband in the stationary-
bar condition (0.0 Hz), while
prominent peaks at the driving
frequencies emerge in all
moving-bar conditions. CPx
and Hx displacement amplitude
increase dramatically from the
stationary bar to the moving-
bar condition&ig.c:/f

tween CPx or Hx and TBx was determined as follows&fn.2:2: Significant
extrema of position and velocity traces of each signal were
picked. Relative phase was calculated by taking the time differ-
ence between an extremum of the target signal (CPx or Hx) and an
extremum of the reference signal (TBx) and dividing this by the
time difference of two extrema in the reference signal. This value
was multiplied by 360° to convert relative phase to degrees. The
mean and angular deviation of relative phase were calculated for
each trial.
2 A more detailed description of the peak-picking method and the
calculation of relative phase and angular deviation of relative phase
can be found in Dijkstra et al. 1994b. Dr. Dijkstra has made a Mat-
lab toolbox for relative phase analysis available to download from
“cattell20.psych.upenn.edu” in the “/pub/tjeerd/RelPhase.box” di-
rectory. (His e-mail address is: tjeerd@cattell.psych.upenn.edu.)&/fn:
Peak-picking was also used to calculate individual cycle-to-cy-
cle periods of CPx and Hx displacement for each trial. Individual
cycle periods were averaged as a measure of mean CP x and Hx dis-
placement frequency within each trial. Peak-picking was not used
in the stationary-bar trials, because relative phase calculations are
not possible with a stationary bar (i.e., no driving signal). Conse-
quently, in the stationary-bar trials, the most prominent peak in the
power spectrum was chosen for CPx and Hx displacement frequen-
cy. Four of the fifteen stationary-bar trials (five subjects×three tri-
als per condition) were not included in the analysis because the
spectrum was not unimodal.
Unless otherwise noted, one-way ANOVAs were used to test
each subject’s data separately, with frequency as the independent
variable at a significance level of 0.05.

Fig. 4 a Mean CPx displacement frequency; b mean Hx displace-
ment frequency. CPx and Hx frequencies averaged between 0.2 and
0.4 Hz in the stationary-bar condition (0.0 Hz), but then decreased
and increased with the frequency of the moving touch-bar. Note
that individual subject’s data are artificially offset at each driving
frequency so that the error bars are not obscured. Error bars in all
figures are the SEM&ig.c:/f
Results
A clear entrainment of body sway to the touch bar was
observed in the moving-bar trials. The contrast between
stationary-bar and moving-bar trials can be seen in
Fig. 2, which shows overlaid time series of CPx, Hx and
TBx displacement in a stationary-bar trial (Fig. 2a) and a
trial in which the bar is moving at 0.5 Hz (Fig. 2b). In
the stationary-bar trial, CPx and Hx displacement have no
prominent characteristic frequencies. In the moving-bar
trial, CPx and Hx displacement assume the same frequen-
cy as the touch bar.
The strong influence of touch-bar movement on head
and body sway is further illustrated in Fig. 3, which
shows spectral plots of CPx and Hx displacement from
individual trials of one subject for each condition. The
pronounced spectral peaks of CPx and Hx displacement
occur at the same frequency as the touch bar in each con-
dition. At the higher touch-bar frequencies (e.g., 0.4 Hz
and 0.5 Hz), an increase in the low-frequency compo-
nents of CPx displacement was also observed, but with
no consistent peak at any particular low-frequency com-
479
Fig. 5 a Mean CPx rms (RMS); b mean Hx rms. All subjects dis-
played a steady increase in displacement amplitude with increas-
ing frequency. At 0.5 Hz, CPx and Hx displacement amplitudes
were approximately twice that of the stationary-bar condition&ig.c:/f
ponent. Despite the strong influence of touch-bar move-
ment on body sway, questioning after the experiment in-
dicated that only one of the five subjects ever perceived
the touch bar as moving during the experiment&fn.3:3.
Mean frequency
Figure 4 shows that the mean CPx and Hx frequency, col-
lapsed across trials for each subject, matched the touch-
bar frequency in each condition, with some drop-off at
0.5 Hz. With a stationary bar, the mean CPx displace-
ment frequency was approximately 0.2–0.4 Hz (broad-
band spectrum); Hx frequency averaged close to 0.2 Hz.
In the moving-bar conditions, both CPx and Hx shifted to
the frequency of the touch bar. CPx and Hx mean fre-
quency showed a significant main effect for Frequency
for all subjects (P<0.001).
3 It is worth noting that all subjects reported perception of in-
creased self-motion on certain trials, but only subject 3 attributed
this to the influence of the moving touch-bar. The other subjects
expressed that they felt less steady on certain trials, but for no spe-
cific reason. One subject reported that the floor felt “spongy” on
certain trials (like “standing on a surfboard”), which suggests that
somatosensory information from the fingertip is interpreted cen-
trally with regard to expectations that the bar is stationary.&/fn:
480
Fig. 6 a Mean CPx gain; b mean Hx gain. CPx gain showed a
steady increase from 0.1 to 0.5 Hz. Hx gain remained constant
across conditions. The slope of subject 3′s gain was much flatter
and remained closer to 1 at all frequencies&ig.c:/f
Contact forces
Subjects maintained comparable levels of contact force
at the fingertip whether the touch bar was moving or sta-
tionary. Vertical force averaged close to 0.5 N and
showed no differences across Frequency for all subjects
(P>0.2). Lateral force averaged slightly above zero at
each frequency and also showed no effect for Frequency
for all subjects (P>0.1). These results indicate that sub-
jects applied very small, lateral (rightward) and medi-
al(leftward) forces with a slight lateral bias. The lack of
differences in contact forces across different frequencies
indicates that subjects maintained comparable contact
forces with a stationary or moving contact surface, even
though four of the five subjects indicated no conscious
perception of touch-bar movement.
Root-mean-square values
CPx and Hx rms in the stationary-bar condition was simi-
lar to that found in previous studies (Jeka and Lackner
1994, 1995), averaging between 0.2 and 0.3 cm, but in-
creased significantly in the moving-bar conditions. RMS
values for CPx and Hx in the 0.5 Hz condition were al-
most double those found with a stationary bar (0.0 Hz).
Figure 5 illustrates that CPx and Hx rms increased with
Fig. 7 Mean relative phase for a CPx versus TBx and b Hx versus
TBx. Both CPx and Hx led touch-bar movement at 0.1 Hz (positive
phase) and lagged behind touch-bar movement at frequencies
higher than 0.2 Hz&ig.c:/f
increasing frequency of the touch bar. A significant ef-
fect for Frequency on rms was found for all subjects
(P<0.0001) except subject 2, who showed no Frequency
effect for Hx rms.
Gain
Mean CPx gain increased with increasing touch-bar fre-
quency, shown in Fig. 6a. A significant Frequency effect
on CPx gain was present for all subjects (P<0.01) except
subject 3, whose gain did not increase with frequency.
The relatively constant gain for subject 3 is particularly
interesting, because she was the only subject who per-
ceived the movement of the touch bar during the experi-
ment. This suggests that conscious perception of touch-
bar movement allows adaptation to different properties
of the stimulus (see Discussion).
Mean gain of Hx displacement, shown in Fig. 6b, did
not depend on touch-bar frequency and averaged approx-
imately 2 across all conditions. Statistical analysis re-
vealed no effect for Hx gain as a function of Frequency
for all subjects (P>0.1). The constant Hx gain together
with an increase in Hx rms is a different relationship than
that observed for CPx rms and gain, which increase to-
gether as a function of frequency. An analysis of the per-
centage of spectral amplitude at the driving frequency
relative to the total spectral amplitude (i.e., at all fre-
 481

bar. Statistical analysis revealed a significant Frequency

effect on CPx and Hx mean phase for all subjects
(P<0.01).
The angular deviation of relative phase (CPx and the
touch bar; Hx and the touch bar) decreased with increas-
ing driving frequency with four of the five subjects
(P<0.01), suggesting an increase in coupling strength be-
tween touch-bar movement and head/body sway. The rel-
ative phase angular deviation of subject 3 showed no sig-
nificant effect for frequency (P>0.1), but displayed an
upward trend as driving frequency increased, reflecting a
decrease in coupling strength.

Mean-squared coherence
The MSC between CPx and touch-bar displacement,
shown in Fig. 8a, and between Hx and touch-bar dis-
placement, shown in Fig. 8b, demonstrate that the cou-
pling was strong (MSC was greater than 0.8) at all fre-
quencies of touch-bar movement. Significant effects for
Frequency emerged for CPx MSC (P<0.01) in subjects 1,
4, and 5. A significant Frequency effect was observed for
Hx MSC (P<0.01) only in subject 1. These effects are
due primarily to the increase in MSC observed as driving
frequency increased. The increase in coupling strength
indicated by the MSC results is consistent with the ob-
served decrease in angular deviation of relative phase
with four of the five subjects.
Fig. 8 Mean-squared coherence (MSC) between a CPx displace- Subject 3 did not show a significant Frequency effect
ment and touch-bar displacement and b Hx displacement and touch- for CP or H MSC, but her MSC showed a downward
x x
bar displacement. MSC was greater than 0.8 at nearly every fre-
quency for four of the five subjects, indicating strong coupling be- trend, consistent with the upward trend of her relative-
tween CPx, Hx, and touch-bar displacement. The MSC of subject 3, phase angular deviation results. Her decrease in MSC is
who perceived movement of the touch bar, decreased beyond 0.2 Hz&ig.c:/f particularly interesting because she was the only subject
to perceive movement of the touch bar. This suggests
that conscious perception of contact surface movement
quencies) demonstrated a decreasing percentage of total can influence the coupling to postural sway (see Discus-
spectral amplitude for the head as frequency increased. sion).
This decrease was significant in three of five subjects
(P<0.01). This may explain, at least partially, why head
gain is constant while head rms increases with increasing Discussion
frequency of the stimulus, because rms reflects the spec- We have demonstrated that contact of the fingertip with a
tral amplitude at all frequencies. The same analysis of stationary surface provides a powerful orientation refer-
CPx revealed a constant percentage of total spectral am- ence for improved control of upright stance (Holden et al.
plitude at the drive with increasing frequency in four of 1994; Jeka and Lackner 1994, 1995). The present results
five subjects (P>0.1), consistent with the parallel in- indicate that contact of the fingertip with a moving bar
crease in CPx gain and rms. One subject showed an in- leads to entrainment of the entire body to the frequency
crease in percentage of total CPx spectral power at the of touch-bar movement. Such entrainment is evidenced
drive with increasing frequency (P<0.01). by head and center of pressure displacement adopting a
harmonic structure that matches the frequency of the
Relative phase touch bar at all stimulus frequencies (cf. Figs. 3, 4) and
being temporally locked (coherent) with the touch bar at
Mean phase steadily decreased as touch-bar frequency all stimulus frequencies (cf. Figs. 7, 8). The small applied
increased. Fig. 7a shows the mean relative phase be- forces at the fingertip were not adequate to physically dis-
tween CPx and the touch bar; Fig. 7b shows the mean place the body (see Holden et al. 1994) and were compa-
relative phase between Hx and the touch bar. At the low- rable across stimulus frequencies. Therefore, entrainment
est touch-bar frequency (0.1 Hz), mean phase is positive, of postural sway with the touch bar must be due to senso-
indicating that CPx and Hx displacement lead touch-bar ry rather than mechanical coupling.
movement. As the driving frequency increases, CPx and To understand how the coherence between body sway
Hx displacement begin to lag the movement of the touch and the moving touch bar may come about and, in partic-
482
ular, which sensory signal drives this coupling, it is use-
ful to employ a concrete model. Following Schöner
(1991), a simple model can be formulated for the lateral
position, x, of the center of mass, by approximating the
postural control system as a second-order linear dynami-
cal system in the vicinity of the postural state at x=0 and
.
x=0:
.. . . . ––
x + αx + ω2x = c(d–x) + √Qεt (1)
where α equals the damping coefficient and ω equals the
eigenfrequency. The influence of the touch-bar stimulus
is captured by the right side of Eq. 1 as. the difference be-
tween the velocity of the touch bar (d ) and the velocity
.
of the body (x), multiplied
–– by a coupling constant, c.
Noise is added (√Qεt) to capture the random influences
on the equilibrium state. If the finger is moved
. . with the
body, it is plausible that this difference (d–x) is the sen-
sory signal available at the sensory surface. In this form,
coupling to touch contributes to the stability of posture
by increasing the effective damping of the system:
~ =α+c. For a stationary touch-bar (d. =0) this enhanced
α Consequently, gain relative to a constant amplitude input
damping predicts attenuated body sway. This is consis-
tent with previous experimental results demonstrating a
decrease in body sway when subjects touch a stationary
bar as compared to standing without contact of the bar
(Jeka and Lackner 1994, 1995).
The experimental results of the present study provide
various lines of evidence supporting the hypothesis that
touch couples to posture through the velocity of the so-
matosensory stimulus at the fingertip:
1. Relative phase between sway and touch-bar position
varies with frequency from approximately +20° for low
frequencies (0.1 Hz) to –90° for the highest frequency of
0.5 Hz, with zero relative phase observed around 0.2 Hz
(cf. Fig. 7). Given that lateral sway eigenfrequencies in a
stationary environment range from approximately 0.1 to
0.3 Hz (Scott and Dzendolet 1971), this result is consis-
tent with velocity-dependent coupling: A near in-phase
relationship relative to touch-bar position at 0.2 Hz
translates to a near 90° phase lag relative to touch-bar
velocity, which is the typical value expected close to the
eigenfrequency of the postural control system. Moreover,
reaching a 90° phase delay at the upper end of the attain-
able frequencies of sway translates into a 180° phase lag
relative to velocity as predicted from Eq. 1 (cf. also
Schöner 1991).
2. MSC increased with increasing frequency (Fig. 8).
MSC can be interpreted as the linear contribution of the
input to the output signal (Bendat and Piersol 1986). In
our model, coherence is an increasing function of the
strength of coupling between somatosensory input and
sway. For constant-amplitude (d0) periodic modulation
(at frequency ωd) of touch-bar. position, d(t)=d0 sin(ωd t),
the velocity-coupling term cd =cd0ωd cos(ωd t) increases
with frequency, reflecting the increase in peak velocity.
Thus, velocity coupling predicts the observed increase in
coherence with frequency.
3. Finally, we observed an approximately constant head
gain across all stimulus frequencies (Fig. 6b). In the mod-
el, the amplitude of sway is predicted to decrease less
with frequency than typically observed for a driven linear
system near resonance, because the velocity-dependent
coupling term increases in amplitude. Thus, the relatively
constant head gain is consistent with velocity coupling as
well. The increase in CPx gain (Fig. 6a) can likewise be
consistently understood. CPx reflects the inverse lateral
acceleration and is in phase with head movement (Fig. 7),
because the 180° phase shift of acceleration against posi-
tion is compensated by inverting the sign of acceleration
to obtain CPx. The dependence of CPx on acceleration im-
plies, however, dependence on the square of the driving
..
frequency: CPx~–x ~cω2d d0 sin(ωd t), resulting in an in-
crease in CPx gain consistent with the observed one. A
mean gain of 2 at the head can be related to the position
of the driving stimulus at the waist, that is, approximately
at the midpoint of the body’s height. A gain of 1 is ex-
pected at a point closest to the vertical position of the
touch stimulus, that is, at the waist. Because the body is
swaying approximately like an inverted pendulum, sway
amplitude is smallest at the ankles and largest at the head.
should result in proportionally higher gain above the
waist and lower gain below the waist. From the geometry
of inverted pendulum sway, a gain of 1 at the waist (the
midpoint of the body) corresponds roughly to a gain of 2
at the head.
These results were consistent across four of the five
subjects tested. The results of subject 3, who reported
conscious perception of touch-bar movement, differed
meaningfully. As the touch-bar frequency increased
above 0.2 Hz, the MSC of subject 3 decreased approxi-
mately linearly, whereas the MSC of other subjects ap-
proached 1. The lower MSC of subject 3 at higher fre-
quencies indicates that she was not as strongly coupled
to the touch-bar movement as other subjects. In terms of
the model, decreasing MSC implies decreasing coupling
strength with frequency. This interpretation is corrobo-
rated by the mean head gain of subject 3 (Fig. 6b), which
is at a consistently lower level (≈1) than for the other
subjects (≈2). At the same time, the slope of CPx gain for
subject 3 is lower than for other subjects. Both effects
are attributable to lower coupling strength. Because sub-
ject 3 was also the only subject who consciously per-
ceived the touch bar to be moving, it is tempting to as-
cribe her lowering of the coupling strength to an active
suppression of input to the postural control system from
what was no longer perceived to be a resting somatosen-
sory world. Cognitive mechanisms might introduce flexi-
bility into the postural control scheme and allow for pa-
rameters that may be fixed at the reflexive level to be
adaptively changed.
The pattern of results supports the hypothesis that so-
matosensory information from the fingertip and arm cou-
ples into the postural control system through the relative
velocity of touch bar and body movement. The neuro-
physiological basis for velocity-dependent, somatosenso-
ry afferent activity is well documented at the peripheral
(Johansson et al. 1982; Matthews 1988) and central lev-
els (Esteky and Schwark 1994). However, these results
are the first to demonstrate rate-dependent coupling be-

tween somatosensory input and whole-body postural
control. Velocity-dependent somatosensory input is thus
analogous to the inputs from the visual (Werkhoven et al.
1992) and vestibular (Howard 1986) systems, which also
represent velocity information.
Previous studies of visually driven postural sway
found similar phase relationships to the present results,
with steadily decreasing gain and larger phase lags at
stimulus frequencies greater than 0.2 Hz (Berthoz et al.
1979). Such results suggested that the postural control
system is passively driven by sensory information. Re-
cent evidence has shown, however, adaptive increases in
gain to a moving visual stimulus. Dijkstra (1994b) ob-
served that postural sway closely matched the amplitude
of visual motion even as distance to the visual display
was varied. Quantitative modeling revealed that not only
coupling strength to visual input, but also the autono-
mous nonvisual component of the postural control
system changed (Giese et al. 1996). The differences in
gain response between earlier and more recent studies
may be due to the amplitude of the visual stimulus,
which was much smaller in Dijkstra’s study than used
previously and more closely matched to the typical sway
amplitudes observed with human stationary stance. Pre-
cise matching of sensory and sway amplitude may result
in stronger coupling than previously observed and allow
adaptive mechanisms to unfold.
In the present results, we observe constant or increas-
ing gain despite large phase lags at higher driving fre-
quencies. We also observe much stronger coupling above
the eigenfrequency of sway than observed with visually
driven sway, which may be consistent with the process-
ing delays that are known to be longer with visual than
somatosensory pathways (Nashner 1981). However, be-
cause the velocity of the somatosensory stimulus in-
creased with driving frequency, the corresponding in-
crease in sensory drive makes it unclear whether cou-
pling or adaptive mechanisms are playing a role in these
differences between somatosensorially and visually in-
duced postural sway. Future experiments will explore
more completely the influence of somatosensory drive
on postural control, which we have demonstrated to be
as dramatic as that of full-field visual stimulation.
&p.2:Acknowledgements We thank Joel Ventura and Art Larson III
for technical assistance with the experimental apparatus. Ely Ra-
bin, Kelvin Oie, and Ryan Cleaver provided assistance with data
analysis and graphics. John Jeka and James Lackner were support-
ed by NASA grants NAGW-4374, NAGW-4375, and NAGW-
4733, Naval Training Systems Center grant NAWC-TSDN61339-
96-C-0026, and AFOSR grant F49620-95–1. Pedro Ribeiro was
supported by CAPES doctoral program grant. Gregor Schöner was
supported by DFG Sch 336/3–1. Tjeerd Dijkstra was supported by
NIH grant EY 09383.
References
Asten NJC van, Gielen CCAM, Denier van der Gon JJ (1988) Pos-
tural adjustments induced by simulated motion of differently
structured environments. Exp Brain Res 73:371–383
483
Bendat JS, Piersol AG (1986) Random data analysis and measure-
ment procedures. Wiley, New York
Berthoz A, Lacour M, Soechting JF, Vidal PP (1979) The role of
vision in the control of posture during linear motion. Prog
Brain Res 50:97–209
Brandt T, Büchele W,Arnold F (1977) Arthrokinetic nystagmus
and egomotion sensation. Exp Brain Res 30:331–338
Dietz V (1992) Human neuronal control of automatic functional
movements. Physiol Rev 72:33–69
Dijkstra TMH, Schöner G, Giese MA, Gielen CCAM (1994a) Fre-
quency dependence of the action-perception cycle for postural
control in a moving visual environment: relative phase dynam-
ics. Biol Cybern 71:489–501
Dijkstra TMH, Schöner G, Gielen CCAM (1994b) Temporal sta-
bility of the action-perception cycle for postural control in a
moving visual environment. Exp Brain Res 97:477–486
Esteky H, Schwark HD (1994) Responses of rapidly adapting neu-
rons in cat primary somatosensory cortex to constant-velocity
mechanical stimulation. J Neurophysiol72(5):2269–2279
Giese MA, Dijkstra TMH, Schöner G, Gielen CCAM (1996) Identi-
fication of the state space dynamics of the action-perception cy-
cle for visually induced postural sway. Biol Cybern 74:427–441
Holden M, Ventura J, Lackner JR (1994) Stabilization of posture by
precision contact of the index finger. J Vestib Res 4:285–301
Horak FB, Macpherson JM (1995) Postural orientation and equi-
librium. In: Shepard J Rowell L (eds) Handbook of physiolo-
gy. Oxford University Press, New York
Howard IP (1986) The perception of posture, self motion and the vi-
sual vertical. In: Boff KR, Kaufmann L, Thomas JP (eds) Sen-
sory processes and perception. (Handbook of perception and hu-
man performance, vol 1) Wiley, New York, pp 18-1–18-62
Jeka JJ, Lackner JR (1994) Fingertip contact influences human
postural control. Exp Brain Res 100:495–502
Jeka JJ, Lackner JR (1995) The role of haptic cues from rough and
slippery surfaces in human postural control. Exp Brain Res
103:267–276
Jeka JJ, Schöner G, Lackner JR (1994) Entrainment of postural
sway to sinusoidal haptic cues. Soc Neurosci Abstr 20:336
Johansson RS, Landström U, Lundström R (1982) Responses of
mechanoreceptive units in the galbrous skin of the human
hand to sinusoidal skin displacements. Brain Res 244:17–25
Lackner JR, DiZio P (1984) Some efferent and somatosensory in-
fluences on body orientation and oculomotor control. In: Woo-
ten BR, Spillman L (eds) Sensory experience, adaptation and
perception. Erlbaum, New Jersey, pp 281–301
Lackner JR, DiZio PA (1992) Sensory-motor calibration processes
constraining the perception of force and motion during loco-
motion. Posture Gait 1:93–104
Lee DN, Lishman JR (1975) Visual proprioceptive control of
stance. J Hum Mov Studies 1:87–95
Marple SL (1987) Digital spectral analysis with applications.
Prentice-Hall, New Jersey
Martin JP (1967) The role of the vestibular system in the control
of posture in man. In: Reuck AVS de, Knight J (eds): Myotac-
tic, kinesthetic and vestibular mechanisms. (CIBA sympo-
sium) Little, Brown, Boston, pp 92–96
Matthews PBC (1988) Proprioceptors and their contribution to so-
matosensory mapping: complex messages require complex
processing. Can J Physiol Pharmacol 66:430–438
Nashner LM (1981) Analysis of stance posture in humans. In:
Towe A, Luschei E (eds) Motor coordination. (Handbook of
behavioral neurology, vol 5) Plenum, New York, pp 527–565
Peterka RJ, Benolken MS (1995) Role of somatosensory and ves-
tibular cues in attenuating visually induced human postural
sway. Exp Brain Res 105:101–110
Schöner G (1991) Dynamic theory of action-perception patterns:
the “moving room” paradigm. Biol Cybern 64:455–462
Scott DE, Dzendolet E (1972) Quantification of sway in standing
humans. Agressologie 13:35–40
Soechting J, Berthoz A (1979) Dynamic role of vision in the con-
trol of posture in man. Exp Brain Res 36:551–561
Werkhoven P, Snippe HP, Toet A (1992) Visual processing of optic
acceleration Vision Res 32:2313–2329