Gait and Posture 21 (2005) 4858

The assessment of body sway and the choice of the stability parameter(s)
J.A. Raymakers , M.M. Samson, H.J.J. Verhaar
Laboratory for Mobility Research, Department of Geriatric Medicine, Division of Internal Medicine, University Medical Center Utrecht,
P.O. Box 85500, 3608 GA Utrecht, The Netherlands
Received 27 October 2003; accepted 24 November 2003

Abstract
This methodological study aims at comparison of the practical usefulness of several parameters of body sway derived from recordings of
the center of pressure (CoP) with the aid of a static force platform as proposed in the literature. These included: mean displacement velocity,
maximal range of movement along x- and y-co-ordinates, movement area, planar deviation, phase plane parameter of Riley and the parameters
of the diffusion stabilogram according to Collins. They were compared in over 850 experiments in a group of young healthy subjects (n = 45,
age 2145 years), a group of elderly healthy (n = 38, age 6178 years) and two groups of elderly subjects (n = 10 and n = 21, age 6589
years) with stability problems under different conditions known to interfere with stability as compared to standing with open eyes fixing a
visual anchoring point: closing the eyes, standing on plastic foam in stead of a firm surface and performing a cognitive task: the modified
stroop test. A force platform (Kistler ) was used and co-ordinates of the bodys center of pressure were recorded during 60 s of quiet barefoot
standing with a sampling frequency of 10 Hz. In general, the results show important overlapping among groups and test conditions. Mean
displacement velocity shows the most consistent differences between test situations, health conditions and age ranges, but is not affected by
an extra cognitive task in healthy old people. Mean maximal sideways sway range is different among groups and test conditions except for
the cognitive task in young and elderly subjects. Standardised displacement parameters such as standard deviations of displacements and
planar deviation discriminate less well than the actual range of motion or the velocity. The critical time interval derived from the diffusion
stabilogram according to Collins et al. seems to add a specific type of information since it shows significant influence from addition of a
cognitive task in old subjects standing on a firm surface but not when standing on plastic foam. The critical time interval shows no consistent
relation to any other parameter. The influence of cognitive activity on balance merits further exploration. A new parameter, sum of maximal
deviation time (SDT) was proposed showing complete discrimination between frail elderly and other old subjects when obtained while visual
input was suppressed. It was concluded that mean displacement velocity seems to be the most informative parameter in most situations.
2003 Elsevier B.V. All rights reserved.
Keywords: Body sway; Center of pressure; Sum of maximal deviation time

1. Introduction
In the past 25 years, a fair number of parameters describing some aspect of body sway have been proposed. They
have never been compared in a critical study as to their usefulness in different conditions and for those wish to orient
themselves in this field it is virtually impossible to make a
valid choice. All these parameters are usually obtained in the
standing subject with devices that registrate the movements
of the body or its center of gravity, or mostly, the center of
pressure (CoP). The former are measured with mechanical

Corresponding author. Tel.: +31-30-250-8280;

fax: +31-30-254-4963.
E-mail address: raymakers@azu.nl (J.A. Raymakers).

0966-6362/$ see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.gaitpost.2003.11.006

or magnetic recording devices connected to the waist [1,2]

or the hip region [3], CoP co-ordinates are derived from
ground reaction forces registered with the aid of a force platform. All the proposed parameters describe some aspect of
the complicated movement pattern termed body sway and
thereby of stability. Besides the length of the CoP trajectory,
and its equivalent, the mean velocity of the CoP movements,
also the maximal range of the movement in perpendicular
horizontal dimensions and the total area of the horizontal
planar domain of the CoP and statistical derivations such as
the standard deviation of the displacement in both directions
and combinations thereof, have been used by several authors [26]. A dimensionless combined stability parameter
derived from phase plane analysis (combining variances of
two-dimensional velocities and of displacements) has also
been proposed [7].

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

A different approach was proposed by Collins and De

Luca [8,9]. These authors describe the data in the form of
a so called diffusion stabilogram by analogy with the random movements of liquid molecules. The squared distance
between randomly chosen pairs of CoPs is plotted against
the corresponding time interval. This leads the authors to the
definition of a critical time interval, a mean squared critical
displacement (MSCD) and a diffusion constant.
While the results of similar single parameters obtained
with different equipment have been compared in the literature, e.g. the length of the CoP trajectory and the area of
movement with force platform stabilometry and magnometry [10], an overall comparison of the different the ways of
elaborating crude CoP co-ordinates obtained with a static
force platform has not been published to our knowledge.
It has not yet been decided which of these parameters
provides the most useful information, i.e. is most effective
in discriminating between individuals with low and those
with an elevated risk of falling or morbid conditions, and
which one responds most to changes of the situation in which
the individual is maintaining his or her standing balance.
Many of the studies deal with a single way of evaluation
of CoP data in a limited number of subjects in a single
morbid condition and an overview of the applicability of the
approach in different situations is lacking.
We compared all these parameters based on CoP
co-ordinates obtained in over 850 experiments in 114 persons aged 2187 years, healthy or with conditions interfering with stability and tested under various conditions. We
propose a new parameter taking into account the duration
and magnitude of the greatest excursions of the CoP from
the center, which might indicate the risk of loosing balance.
The aim of this study was purely methodological. It is
strictly limited to the interpretation of CoP co-ordinates obtained with one single force plate. The use of two force
plates (one for each leg) and de possibility of calculating
movements of the center of gravity rather than those of the
CoP is outside the scope of this study. We hope to offer
guidelines for the use of force-plate data by comparing the
different ways of using crude CoP data in their ability to
discriminate between young healthy, elderly healthy people
and elderly patients at risk of falling, under different test
conditions that are known to influence stability and to obtain
information about the repeatability of these measurements.
We do not intend to investigate theories on the way stable
upright stance is achieved and maintained.
2. Material and methods
2.1. Participants
Young healthy participants (n = 45, age 2145 years)
were recruited from the hospital staff and persons aged over
60 years (n = 38, age 6178 years), by advertisement in the
local press and by asking healthy partners of patients with
Parkinsons disease attending the outpatient department.

49

The patient group consisted of 10 geriatric patients (age

7589 years) and 21 patients with Parkinsons disease (age
6587 years). The former were hospitalised at the department of geriatric medicine for various acute health problems
not primarily connected with mobility and had recovered so
as to be able to return home. They were considered to form
a sample of frail elderly subjects. They were able to stand
and to walk with a walking aid (rolator). The Parkinsons
patients were attending the outpatient clinic for Parkinsons
disease of the department of geriatric medicine and were in
stable condition and able to and stand and walk.
All young healthy participants gave informed consent to
participate after oral information. In the older group written
informed consent was obtained from all after oral and printed
information had been given. The tests were performed in
the laboratory for mobility research of the department of
geriatrics of the University Medical Center Utrecht between
September 1999 and September 2001.
2.2. Test procedure
The subjects were asked to stand quietly, barefoot, directly
on the surface of the force plate with the foot parallel on
both sides of a 4 cm broad T-shaped separator fixed to the
force-plate surface, since the position of the feet has been
reported to influence the results [11].
The basic instruction consisted of asking them to stand as
still as possible while visually fixing a black spot of 10 cm
diameter on the wall 150 cm before them at eye height. CoP
registration was performed for at least 60 s. Two protocols
were followed from there: the first included three consecutive modified stroop tests [12,13], where the subjects are
asked to name the colours of words (red, blue, green, yellow) printed on a table in front of them (at exactly the same
place as the black spot) in a colour different from what the
printed term says, without making mistakes for 120 s (only
60 s were recorded however). The word blue is for example printed in yellow, the word red in blue, etc. This intensive cognitive task is meant to distract the subjects attention
from the primary task of quiet standing. The subject then
steps off the force plate. A 10 cm thick layer of polyethylene
foam is placed on top of it and the subject is again asked to
stand on the plastic foam surface fixing the black spot. This
is meant to interfere with proprioceptive information from
the feet and ankles. The control experiment, the first stroop
test and the test standing on foam were used in the analysis.
The second protocol also started with a recording while
fixing the black spot, followed by a second recording with
eyes closed for 60 s, to evaluate the influence of visual input.
The protocol that was used initially in each individual was
repeated after 1 week under exactly the same conditions,
provided he or she was available, which was the case in
over 42 out of the 45 healthy subjects, in all but one elderly
healthy subjects, in 10 geriatric patients and in 18 of 21
patients with Parkinsons disease, who were dependent on
transport by benevolent relatives.

50

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

2.3. Equipment
A static force platform (Kistler ) connected to an eightchannel amplifier and A-D converter was used to record
ground reaction forces. The signals were processed on a PC
connected to the A-D converter using Bioware software to
obtain CoP co-ordinates, where x refers to sideways movements and y to fore-aft displacement.
A special programme was written to calculate additional parameters from the CoP data, not available with
the Bioware software. These were: the range of maximal
displacement in both horizontal directions, mean displacement velocity, the phase plane stability parameter [7], area
of stabilogram [2], planar deviation, mean squared critical
displacement, critical time and diffusion constant [8,9],
and the sum of the products of duration and displacement
of the maximal deviations from the average. The program
provides graphics of the mono-directional and planar displacements of the CoP and of the diffusion stabilogram. It
is available on request.
2.4. Calculation of parameters of sway
Recordings always lasted at least 60 s and sampling frequency was 10 Hz. The CoP stabilogram was inspected visually and it was decided that the first 10 s of the recording
would be ignored systematically to avoid disturbance from
delayed stabilisation of the recording equipment after the
person stepped onto the force plate.
Recordings were mostly extended to 120 s. However, in
elderly patients it was frequently seen that during the second minute varying inconsistencies occurred, e.g. a shifting
of the average CoP to a new location or exaggerated and unpredictable movements, which were interpreted as signs of
fatigue. Since recordings were much more consistent during
the first minute, only data from that interval were used for
analysis. So the interval between 10 and 60 s of the recording
was used and 500 pairs of CoP co-ordinates were obtained
in every subject. In a number of subjects data collection was
originally done with a sampling frequency of 50 Hz but in
the analysis co-ordinate pairs obtained at consecutive at 0.1 s
intervals were used, thus simulating a sampling frequency
of 10 Hz.
The following parameters were calculated: maximal range
of CoP displacement along the x- and y-co-ordinate, mean
velocity of CoP displacement (m/s), planar deviation, i.e.
the square root of the sum of the variances of displacements
in x (sideways) and y (fore-aft) direction, the phase plane
parameter [7] (square root of the sum of variances of velocity
and of displacement) and the area of the stabilogram. The
latter was established using an algorithm that constructs a
smooth closed curve (of irregular shape) that encloses all
recorded CoP points, the surface of which is calculated.
A new parameter was added: sum of maximal deviation
time (SDT). This is the sum of the maximal deviations of
the CoP from its center regardless of their direction multi-

plied by their duration. The calculation runs as follows: the

distances of the CoP to the center (represented by the average of all recorded co-ordinates) are arranged in a 16 16
bi-dimensional array according to their magnitude (1/16 of
the planar deviation of the individual test was used as classifying unit) and according to their duration expressed in units
of 0.1 s up to 1.6 s. The duration was counted from the moment the deviation reached the value corresponding to the
lower limit of the class in question until it decreased again
below this same limit. Even if more of these events characterised by this magnitude end duration occurred, only one
was used in the calculation. The deviation corresponding to
the cell with the maximal deviation in each duration class
was multiplied with the corresponding duration and these
products were summed. This provides the SDT. In other
terms its sums up the products of duration and magnitude
of uninterrupted intervals during which deviation is maximal, while all multiple short intervals where the CoP moves
rather quickly back and forth at short distance of the average
CoP are ignored. It is a combination of expressing the time
that subjects spend near the borders of stability and how
close to those borders this takes them. Hereby, we hoped to
detect eventually individuals with a greater risk of falling
(which will have to be the subject of a separate study).
The construction of the diffusion stabilogram [9] was performed using 100 sets of 50 data pairs randomly and repeatedly chosen at intervals increasing from 0.1 to 10 s from the
set of 500 CoP data. The squared mean distance between the
50 pairs of each set was plotted against the interval and regression analysis was performed in an iterative procedure so
as to provide two regression equations that fitted the initial
and the secondary phase of the plot optimally while the intersection of the two regression line fell not more than 0.1 s
from the interval that marked the transgression from the initial to the secondary regression domain. Typical examples
of standard and diffusion stabilograms in a single healthy
subject are shown in Fig. 1.
The time co-ordinate of the intersection point was termed
by the original authors critical time interval and the corresponding squared distance was taken as the mean squared
critical displacement [9]. The regression coefficient of
the secondary phase was taken to represent the diffusion
constant according to the same authors. The critical time
interval may be considered to represent the interval during
which displacement is governed by an open loop control
system to maintain stability. The mean squared critical distance corresponds in some way to the displacement that is
tolerated before a restraining action is taken by the same
system now using closed loop control, which action may
or may not be successful but tends to contain the random
walks of the COP within a limited space.
In the formulas, x and y indicate the values of the CoP
co-ordinates in the sideways and fore-aft directions. The
average co-ordinates being normalised at (0, 0). Thus x < 0
means left, x > 0 right, y < 0 means backward, y > 0
means forward. The parameters we investigated are defined

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

51

as follows:
Abbreviation (name)
PD (planar deviation)
Rfa (range fore-aft)
Rsw (range sideways)
Vd (displacement velocity)
Vm (mean velocity)
PP (phase plane parameter, Riley)
A (area)
SDT (sum of maximal deviation time)

DC (diffusion constant)
CI (critical time interval)

MSCD (mean squared critical displacement)

Formula
2
( x + 2 y); where 2 x and 2 y = squared
estimated standard deviation of x and y
|ymax ymin |
|xmax xmin |

((xi xi1 )2 + (yi yi1 )2 )

ti ti1


Vd
n2
( x + 2 y + 2 Vx + 2 Vy ; where Vx and Vy are
the velocities in x and y-dimension
Surface contained within the closed (irregular) curve
including all recorded CoPs.

16
i=1 tiDmax ; where t is time interval unit (0.1 s), i
the number of time-units and Dmax the maximal
deviation with a duration of t i summed for t = 0.1
to 1.6 s. The number of times that a deviation lasting
ti s occurs is not taken into account
Slope of regression line through points beyond the
critical time interval in the diffusion stabilogram [8,9]
Interval at which the slope of the regression of the
mean squared distance between random pairs on
their time interval in a diffusion stabilogram [9]
shows a significant break towards a more shallow
regression than the initial one
Mean squared distance between random samples of
CoP pairs in the diffusion stabilogram with a time
difference corresponding to the critical time interval

2.5. Statistical methods

Differences between experimental outcomes in the same
individuals were tested by paired Students t-test. Differences between groups in the basal situation were examined
by one-way analysis of variance and multiple comparisons
post hoc were done by Dunnets t-test. Variance analysis with a simple linear model using group number, test
number (the first or second of duplicate tests) and test condition (one out of four: eyes open, eyes closed, standing
on foam or performing the stroop test) as independent variables was used to test the influence of the health status, the
test situation and repetition on each of the parameters of
sway. To assess relationships between parameters and the
possibilities of data reduction linear correlations between
parameters were calculated and factor analysis was applied. The software SPSS v10.0 was used for the statistical
calculations.
Intra-individual standardised coefficients of variation
(sCV) for all parameters were calculated from duplicate
experiments by the formula:


(xi1 xi2 )2 /2n
sCV = 100%
90% range

Dimension
cm
cm
cm
cm/s
cm/s
No dimension
cm2
cms

mm2 /s
s

mm2

where xi1 and xi2 are paired observations from subject i, and
n is the number of paired observations and the 90% range is
the difference between the 5 and 95 percentile of the data.

3. Results
In the presentation, the results the data for every parameter
of sway obtained at the first test in each individual (using
two or three test conditions) was used in the calculation of
statistics. The results of duplicate tests performed on the
same day of after 1 week, were only used for calculation of
reproducibility.
3.1. The influence of age and health status
The results obtained under basal circumstances (eyes
open standing on firm surface) showed significant differences between young healthy subjects and elderly healthy
subjects for all parameters except the diffusion constant and
fore-aft range. Sideways range, area, velocity and MSCD
were significantly different in healthy elderly versus elderly
Parkinsons patients (Table 1). There were no significant
differences between Parkinsons patients and the group of

52

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

Fig. 1. Examples of stabilograms of a single healthy young person. The left panels represent the trajectory of the CoP recorded with eyes open (top)
and closed (bottom). The right panels show the diffusion stabilograms according to Collins et al. under the same conditions, with the indication of the
critical time interval and the mean squared critical displacement (MSCD). The slope of the regression line of the points beyond the critical time interval
constitutes the diffusion constant (angle a, right lower panel).

Table 1
Parameters of body sway in basal test conditions
Group

Range
fore-aft
(cm)

Range
sideways
(cm)

Area
(cm2 )

Mean
velocity
(cm/s)

Phase
plane

SDT
(cm s)

Critical
time (s)

Diffusion
constant
(mm2 /s)

MSCD
(mm2 )

7.1
3.2

1.98
1.31

29
40

39
24

Young healthy

Mean
S.D.

2.3
0.7

1.8
0.5

2.3
1.2

0.94
0.19

8.38
1.92

Elderly healthy

Mean
S.D.

2.5
0.8

2.4y
0.7

3.5x
2.1

1.59y
1.15

12.0x
6.9

7.1
3.8

1.59
1.0

21
36

61x
42

Elderly Parkinsons

Mean
S.D.

3.3y,b
1.0

3.1y,b
1.0

6.1y,a
4.2

2.07y
0.83

15.4y
5.7

4.0y,c
1.3

1.16y,a
0.32

14
30

111y,a
81

Geriatric patients

Mean
S.D.

3.2y,b
0.9

3.1y,b
0.9

6.1y,b
3.1

2.09y
0.97

20.0y,b
6.8

59
125

262y
701

12.2y,b
3.3

2.02
2.9

The mean values of the most important parameters of body sway in different groups in basal test condition: first test with eyes open fixing a black spot
at eye height and standing barefoot on firm surface. Multiple comparison by Dunnets t-test (two-sided). x 0.01 > P > 0.001, y P < 0.001 as compared to
young healthy. The means that are not marked by a superscript number are not significantly different from those of the young healthy group (P > 0.10).
a 0.05 > P > 0.01, b 0.01 > P > 0.001, c P < 0.001 as compared to elderly healthy. The means that are not marked by a superscript character are not
significantly different from those of the elderly healthy group (P > 0.10).

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

53

Table 2
The influence of age, health status, cognitive activity, visual and proprioceptive input
Influence

Age

Health Parkinson/
frail

Visual input eyes

closed

Cognitive task modified

stroop test

Propriocepsis plastic
foam

Test condition (personal

condition)
Planar development
Range fore-aft
Range sideways
Velocity
SDT
Phase plane
Area
Diffusion constant
Critical time
MSCD

Eyes open
(healthy)

+++
+++
+++
++
+++

+++

Eyes open
(60+)
++
++
++
++
+++
++
+++
+

+++

Young
(healthy)
++
+++
+++
+++
+++
+++
+++

+
++

Young
(healthy)

++

+++
+
+++

+++

Young
(healthy)
+++
+++
+++
+++
+++
+++
+++
+++
+++
+++

>60
(healthy)
+
+++
+++
+++
++
+++

++
+++

>60
(healthy)

+
++
++

>60
(healthy)
+++
+++
+++
++
+++
+++
+++

+++

Significance of mean differences between paired results obtained from persons in different conditions: young and healthy, old (>60 years), healthy or sick
and with as intervention closing of the eyes, the performance of the modified stroop test and the interposition of a layer of plastic foam between the force
plate and the feet as compared to the test on firm surface with eyes open () P 0.05; (+) 0.05 > P 0.01; (++) 0.01 > P 0.001; (+++) P < 0.001.

geriatric patients. The diffusion constant does not seem to

offer any discriminative information in the basal situation.
3.2. Relations between parameters
When factor analysis was applied to the data including all
the parameters of whom many are related a priori, most of
them they were grouped in one single factor with high and
consistent loadings. However the critical time interval according to Collins et al. appears to explain a specific part of
the total variance since it is does not show a significant correlation to any of the displacement parameters. By entering
mean displacement velocity, stabilogram area and the critical interval into an exploratory factor analysis up to 95% of
total variance could be explained with a two-factor model
using all data of all persons in a single test situation: eyes
open where the first two matrix components had the greatest
loadings on the first factor (0.940 and 0.952) and the critical
time interval on the second factor (0.990). The latter parameter thus appears to provide a different kind of information
as compared to those derived form displacement. The diffusion constant of Collins et al. and the mean squared critical
displacement do not seem to provide any additional information.
3.3. Discriminative potential
An overview of the significance of the influence of different approaches on the average results at multiple comparison between subject groups and test situations is given in
Table 2.
3.4. Reproducibility
To test the influence of intra-individual variability the tests
were repeated in most of the subjects after 1 week and in
all the geriatric patients also in duplicate at both occasions.
Variance analysis showed that differences between the first

and second test did not contribute significantly to the total

variance of the results, when differences due to age, and test
circumstances were also taken into account.
All the testing situations were evaluated for intraindividual variability by applying the paired t-test to
duplicate test results and intra-individual standardised coefficients of variation were calculated. Variance analysis was
applied using age group (young or old), test situation (eyes
open, closed, stroop test, standing on foam) and duplicate
number (1 or 2) as fixed factors to see the influence of
intra-individual variability on total variance. The results are
shown in Table 3. The variability seems rather large but is
similar to that reported in the literature.
To illustrate the impact of age and health status on stability in various conditions (eyes open or closed, standing
on firm or soft surface, performing and extra cognitive task)
means of displacement velocity and critical time are shown
in Figs. 2 and 3.
3.5. The influence of sampling frequency
All calculations were based on a sampling frequency of
10 Hz. We disposed of registrations with an original sampling frequency of 50 Hz in part of the subjects. In those all
the parameters were calculated in such a way that only every
fifth CoP recording from the original list was used. Thus a
sampling frequency of 10 Hz was simulated. We compared
both sets of results in the group of 10 geriatric patients.
Significant differences (P < 0.001) between the calculations based on the 50 and 10 Hz-based results were seen
for most parameters except for those derived from the diffusion stabilogram: critical time interval, diffusion constant
and MSCD (P > 0.30). The length of the trajectory and
the mean displacement velocity obtained at 50 Hz sampling
frequency were both 26.1% greater than when 10 Hz sampling was used, stabilogram area was 13.0% greater, Rileys
phase plane parameter 21.3%, and STD even 75%. This is
self-evident since much information is necessarily lost at the

54

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

Table 3
Features of reproducibility in duplicate experiments
Range fore-aft
(mm)

Range sideways
(mm)

Velocity
(cm/s)

Phase
plane

Area
(cm2 )

SDT
(cm s)

Diffusion
constant
(mm2 /s)

Critical
time (s)

MSCD
(mm2 )

0.125
0.567
0.573

8.1
2.207
0.032

Variance analysis (285 duplicate experiments in healthy persons, all test conditions) variance from duplication vs. residual variance
F
3.973
2.221
1.520
3.290
3.296
1.360
1.153
1.275
Significance P
0.044
0.137
0.218
0.070
0.070
0.271
0.283
0.259

6.982
0.008

t-test in 51 healthy subjects (eyes open on firm surface)

Mean difference
0.47
0.22
t
0.408
0.252
Significance P
0.685
0.802

0.69
0.877
0.384

0.36
0.724
0.473

0.14
0.650
0.519

0.17
0.518
0.607

Within subject standardised coefficient of variation for 51 healthy subjects (eyes open on firm surface)
sCV (%)
28
19
14
18
26
25

3.49 108
0.558
0.579

30

63

64

Mean differences, Students t-test of parameters in the basal situation calculated in duplicate experiments at 7 days interval in healthy subjects (n = 51).
Variance analysis with age group, test situation and duplicate number as fixed factors in 285 duplicate experiments in healthy subjects. The value of F
is the ratio of the variance attributable to duplicate and the residual variance. The bottom line shows the standardised coefficients of variation (sCV) for
51 healthy persons with eyes open.

lower sampling frequency. As expected the averages of planar deviation and standard deviations of displacement in xand y-direction obtained at 50 Hz were no more than 0.2%
different from those at 10 Hz. They simply are based on
a greater sample from the same data. The maximal range
recorded at both frequencies differed no more than 3% for
both dimensions.

4. Discussion
The many parameters that have been derived from sets of
CoP co-ordinates seem to provide very similar information
with the possible exemption of those derived by the approach
of Collins et al. This might be anticipated since they are all
based on the same crude data. Nevertheless, it is very difficult to compare reports from the literature when different
approaches of elaboration of the basal CoP recordings are
used. The aim of this study was to find a way of representing the CoP movements in a summary manner that would
yield a maximal discrimination between age groups and situations known to influence body sway. This could then in
a subsequent study help to detect individuals at high risk of
falling. The great number of experiments carried out in different groups of healthy and non-healthy subjects and under different test conditions allowed us to perform statistical
analysis that might identify those parameters or combinations thereof that would enable us to discriminate effectively
between the groups and/or the test situations.
Young and elderly subjects show significant differences in
the average results for the parameters obtained under basal
conditions. Sideways sway range is greater in older people,
as are the sway area, the mean velocity, the phase plane
parameter [7], and the mean squared critical displacement
in the diffusion stabilogram. Parkinsons patients show a
greater difference from normal in almost all parameters as
compared to healthy older subjects.

Collins and De Luca [8,9] compare the displacements of

the CoP to Brownian movements of a molecule of fluid that
are restrained at some level by a regulatory mechanism. At
that level the open loop control of the movements is replaced
by a closed loop control system. They apply Einsteins formula for the description of displacement of randomly walking molecules (diffusion) to CoP data and calculate the
mean squared critical displacement and the critical time interval at which the characteristics of the random Brownian
movement appear to be replaced by restraining influences
that ensure stability. The control system seems to limit the
extent of open loop controlled random movements and this
limit seems to be related to a time interval, describing the
temporal tolerance of the system [8,9]. While the comparison to particle physics is an interesting one, its theoretical
basis needs further exploration and there are fundamental
differences. As yet it is no more than a descriptive analogy. Its summary description shows relatively little differences among groups and interventions, but the critical time
interval appears to bear information that is different from
the one contained in the simple displacement parameters
as shown by factor analysis. In itself it is less discriminative. The critical time interval is influenced by the cognitive
task or blocking visual input as described by Collins and De
Luca [14]. It is not significantly dependent on age and unequivocally decreased by disturbing influences in young but
less significantly in elderly people. The information offered
by the diffusion constant is uncertain. It is only significantly
influenced by manipulation of visual and proprioceptive input in young subjects. In old people, whether healthy or
with general or specific (Parkinsonian) mobility problems,
MCSD shows significant differences between test situations
and health conditions.
The summed displacementtime parameter (SDT) was designed to detect subjects who show relatively long episodes
of major deviation from the average CoP and might thereby
be at particular risk of falling. Fig. 4 shows that SDT is not

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

55

Fig. 2. The influence of age on velocity of CoP displacement (upper panel) and critical time interval (lower panel) in healthy subjects in different test
conditions (quiet standing with eyes open, standing on plastic foam and performing the modified stroop test). Mean values are shown 2S.E.M.

negatively influenced by blocking visual input but that the

value in geriatric patients is already much greater than in the
other groups even when visual control is present. Impairment of proprioceptive input, which was not tested in geriatric patients, brings the value of SDT in the other groups up
to the geriatric level. SDT could be a more precise indicator
of falling risk than any other parameter. This finding may be
an indication that proprioceptive input from the lower limbs
is particularly important for balance maintenance in frail elderly subjects (which our geriatric patients were), when visual input is blocked. The relationship to actual falling risk
will have to be tested in a prospective design. In the other
groups, who lack the characteristic of frailty, closing the

eyes has no significant influence on SDT, while impairment

of proprioceptive input makes it increase even in healthy
people (Fig. 4).
Some additional ways of analysing the CoP data in a more
cybernetic way have been proposed. One of them aims at the
detection of periodical and non-periodical phenomena that
could be interpreted partly as reaction to disturbing stimuli,
as adaptation or as feed-back [15]. Finally it was proposed
to use vertical ground reaction forces for balance analysis
[16]. These approaches still seem to be very theoretical and
are outside the scope of this study. A promising approach is
the one comparing the center of pressure the displacement to
that of the center of gravity. It necessitates the simultaneous

56

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

Fig. 3. The influence of age and health status and the presence or absence of visual control on velocity of CoP displacement (upper panel) and critical
time interval (lower panel). Mean values are shown with bars representing 2S.E.M.

registration of body movements and displacement of the CoP

[17,18].
When displacement parameters of the CoP for different
groups and situations are compared, their value appears to
increase in conditions where stability may be expected to
decrease. These are: increasing age, closing of they eyes,
disruption of proprioceptic information by the placement
of plastic foam under the feet and diverting the subjects
attention from the task of quiet standing by an intensive
cognitive activity. The importance of attention of the subject

to the task of standing and its dependence on age have been

demonstrated in the literature [19,20].
In the literature, sampling frequencies from 10 to 100 Hz
have been reported. Many parameters are necessarily influenced by the choice of the sampling frequency. The
trajectory of the CoP appears more complex when more
intermediate CoP points are recorded and the trajectory
measured will therefore appear to be longer (and indeed
more accurate) when calculated from 50 Hz registrations.
For the measurement of velocity the same principle goes.

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

57

Fig. 4. The means (2S.E.M.) of SDT for the four groups of subjects (1: young healthy, 2: elderly healthy, 3: geriatric patients, 4: Parkinsons patients)
when examined with eyes open on firm surface in different test conditions (upper panel). The individual values are shown for closed eyes in the lower panel.

This was illustrated with the example of the geriatric patients in Section 3. Neglecting information by choosing a
low sampling frequency may reduce accuracy and thereby
discriminating capability. Comparison of results of studies performed with different sampling frequencies is not
possible.
It is unclear at first sight how the sampling frequency will
influence the parameters of the diffusion stabilogram. Since
this concept is based on random use of pairs of CoP data
its results will show variation even when repeated on the
same data set. However, we were not able to demonstrate a
significant influence of sampling frequency.
In conclusion, several of the displacement parameters
seem to contain the same amount of information but there
are some differences. Fore-aft range is not influenced by

age while lateral sway range is. It has been reported that
lateral sway range is the best single predictor of falling
risk. Mean velocity of displacement as a single parameter
discriminates well between test situations except for the influence of the cognitive task (stroop test) in elderly subjects.
It also has the smallest standardised intra-individual coefficient of variation, i.e. the smallest reproducibility error. The
total length of the CoP trajectory is directly arithmetically
related to mean velocity and recording time (which was the
same in all our recordings) and offers the same information,
provided the recording time is standardised. Both velocity
and trajectory length depend significantly on sampling frequency. The standardised descriptives, such as the standard
deviations of displacement and planar deviation seem to be
less useful. The characterisation of the critical time interval

58

J.A. Raymakers et al. / Gait and Posture 21 (2005) 4858

that is derived from the diffusion stabilogram according to

Collins seems to contain a special type of information that
needs to be explored further. This parameter appears to be
independent of sampling frequency and is not correlated
to any of the displacement parameters. We were unable to
attribute a meaningful place to the diffusion constant.
The final objective of this kind of investigation is to understand why people fall and/or to establish their risk of falling
prospectively. Studies allowing conclusions in this respect
are still scarce, despite the fact that the equipment has been
available such a long time [2123].
Acknowledgements
We acknowledge the help of Ms. Evelien Jacobs and Laura
Kleemans, Edwin Slok, Dineke Koek, Noor van Oosten,
Mieke Dammer who collected the experimental data and of
Paul de Vreede and Anne Cornelissen who helped with the
examinations and the maintenance of the equipment.
References
[1] Dornan J, Fernie GR, Holliday PJ. Visual input: its importance in the
control of postural sway. Arch Phys Med Rehabil 1978;59:58691.
[2] Lord SR, Clark RD, Webster IW. Postural stability and associated
physiological factors in a population of aged persons. J Gerontol
(Med Sci) 1991;46:M6976.
[3] Dean EM, Griffiths CJ, Murray A. Stability of the human body investigated by sway magnetometry. J Med Eng Technol 1986;10:12630.
[4] Black FO, Wall Jr D, OLeaty DP. Computerized screening of the
human vestibular system. Ann Otol Rhinol Laryngol 1978;87:853
61.
[5] Black FO, Wall C, Rockette HE, Kitch R. Normal subject postural
sway during the Romberg test. Am J Otolanrynol 1982;3:30918.
[6] Goldie PA, Bach TM, Evans OM. Force platform measures for
evaluating postural control: reliability and validity. Arch Phys Med
Reabil 1989;70:5107.
[7] Riley PO, Benda BJ, Gill-Body KM, Krebs DE. Phase plane analysis
of stability in quiet standing. J Rehabil Res Dev 1995;32:22735.

[8] Collins JJ, De Luca CJ. Random walking during quiet standing. Phys
Rev Lett 1994;73:7647.
[9] Collins JJ, De Luca CJ. Upright, correlated random walks: a
statistical-biomechanics approach to the human postural control system. Chaos 1995;5:5763.
[10] FitzGerald JE, Murray A, Elliot C, Birchall JP. Comparison of
balance assessment by sway magnetometry and force platforms. Arch
Otolaryngol Head Neck Surg 1993;119:416.
[11] Uimonen S, Laikatari K, Sorri M, Bloigu R, Palva A. Effect of
positioning of the feet in posturography. J Vestib Res 1992;2:349
56.
[12] Melzer I, Benjuya N, Kaplanski J. Age-related changes of postural
control: effect of cognitive tasks. J Gerontol 2001;47:18494.
[13] Siegrist M. Reliability of the stroop test with single-stimulus presentation. Percept Mot Skills 1995;81:12958.
[14] Collins JJ, De Luca CJ. The effects of visual input on open-loop
and closed-loop postural control mechanisms. Exp Brain Res
1995;103:15163.
[15] Fransson P-A, Johansson R, Hafstrm A, Magnusson M. Methods for
evaluation of postural control adaptation. Gait Posture 2000;12:14
24.
[16] nell A. The vertical ground reaction force for analysis of balance?
Gait Posture 2000;12:713.
[17] Hasan SS, Robin DW, Szurkus DC, Ashmead DH, Peterson SW,
Shiavi RG. Simultaneous measurement of body Center of pressure
and Center of gravity during upright stance. Part I. Methods. Gait
Posture 1996;4:110.
[18] Hasan SS, Robin DW, Szurkus DC, Ashmead DH, Peterson SW,
Shiavi RG. Simultaneous measurement of body Center of pressure
and Center of gravity during upright stance. Part II. Amplitude and
frequency data. Gait Posture 1996;4:1120.
[19] Nishiwaki Y, Takebayashi T, Imai A, Yamamoto M, Omae K. Difference by instructional set in stabilometry. J Vestib Res 2000;10:157
61.
[20] Marsh AP, Geel SE. The effect of age on the attentional demands
of postural control. Gait Posture 2000;12:10513.
[21] Brauer SG, Burns YR, Galley P. A prospective study of laboratory and
clinical measures of postural stability to predict community-dwelling
fallers. J Geront (Med Sci) 2000;55A:M46976.
[22] Gunter KB, White KN, Hayes WC, Snow CM. Functional mobility
discriminates nonfallers from one-time and frequent fallers. J Gerontol (Med Sci) 2000;55A:M6726.
[23] Maki BE, Holliday PJ, Topper AK. A prospective study of postural
balance and risk of falling in an ambulatory and independent elderly
population. J Gerontol Med Sci 1994;49:M7284.