J Appl Physiol
91: 1307–1317, 2001.
Muscle activity in the leg is tuned in response
to ground reaction forces
JAMES M. WAKELING, VINZENZ VON TSCHARNER, BENNO M. NIGG, AND PRO STERGIOU
Human Performance Laboratory, Faculty of Kinesiology, University of Calgary,
Calgary, Alberta, Canada T2N 1N4
Received 31 December 2000; accepted in final form 10 May 2001
the human body reacts to
input from its external environment. One such input is
the ground reaction force, which occurs during the
ground contact phase of each stride. One possible reaction is the modification of the muscle activity patterns in response to that force. The lower extremity
muscles are used for many different tasks during each
stride, including the control of skeletal position, joint
stiffness, vibrations of the soft tissue packages, stability during ground contact, and propulsion for the movement task. It has been speculated that there is a
requirement for the muscles to control and, thus, minimize soft tissue vibrations during locomotion (25) and,
thus, that there will be a change in muscle activity
patterns in response to different vibration loadings on
the lower extremity.
Impact forces in heel-toe walking and running are
forces resulting from the collision of the heel with the
ground, reaching their maximum (the impact peak)
earlier than 50 ms after first contact (25). The rate at
which the impact peak is reached is termed the loading
rate and is a correlate of the major frequency of the
impact peak. Impact forces have frequency contents in
the range 10–20 Hz and should be expected to produce
vibrations of the soft tissues of the body. However,
observations suggest that impact-related vibrations
are minimal in the muscular soft tissues of the lower
extremities during walking and running. One possible
mechanism to reduce soft tissue vibrations is to ensure
that the resonant frequencies of the soft tissues are
distinct from the frequency of the impact force and that
vibrations within the soft tissues are strongly damped.
The frequency and damping coefficients of vibrations in
the soft tissues of the lower extremities are increased
by increases in muscle force production and muscle
shortening velocity (34). The natural vibration frequencies of the triceps surae, quadriceps, and tibialis
anterior tissues are in the range 10–50 Hz, depending
on the levels of muscle activity (34). Therefore, it
should be expected that if the frequency of the impact
force is close to the natural frequency of the soft tissues, then additional muscle activity will be used to
minimize possible vibrations. Muscle activity has been
shown to respond to frequencies of applied continuous
vibrations in the medial head of the triceps brachii in
the arm during isometric tests (7). However, changes
in the myoelectric patterns of the lower extremities
that may occur in response to the different impact
conditions experienced during walking and running
have not yet been demonstrated.
To test whether the body reacts to the loading rate of
the impact force by modifying muscle activity, an experiment must be performed that can alter the loading
rates of the impact force. The magnitude of the impact
force typically increases with the running velocity (8,
12, 26); however, the magnitude of the impact force
remains relatively insensitive to changes in cushioning
from the shoe or ground. However, the shoe cushioning
does change the loading rate (a correlate of frequency)
of the impact force (8, 16, 17, 19, 26). Thus, changing
the material properties of the shoe provides a tool by
Address for reprint requests and other correspondence: J. M.
Wakeling, Human Performance Laboratory, Faculty of Kinesiology,
University of Calgary, Calgary, AB, Canada T2N 1N4 (E-mail:
[email protected]).
The costs of publication of this article were defrayed in part by the
payment of page charges. The article must therefore be hereby
marked ‘‘advertisement’’ in accordance with 18 U.S.C. Section 1734
solely to indicate this fact.
electromyogram; wavelet; impact force; time-frequency analysis
DURING WALKING AND RUNNING,
http://www.jap.org
8750-7587/01 $5.00 Copyright © 2001 the American Physiological Society
1307
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
Wakeling, James M., Vinzenz von Tscharner, Benno
M. Nigg, and Pro Stergiou. Muscle activity in the leg is
tuned in response to ground reaction forces. J Appl Physiol
91: 1307–1317, 2001.—During walking and running, the
human body reacts to its external environment. One such
response is to the impact forces that occur at heel strike. This
study tested previous speculation that the levels of muscle
activity in the lower extremities are adjusted in response to
the loading rate of the impact forces. A pendulum apparatus
was used to deliver repetitive impacts to the heels of 20
subjects. Impact forces were of similar magnitude to those
experienced during running, but the loading rate was varied
by 13% using different materials in the subjects’ shoes. Myoelectric patterns were measured in the tibialis anterior, medial gastrocnemius, vastus medialis, and biceps femoris muscles. Wavelet analysis was used to resolve intensity of the
myoelectric patterns into time and frequency space. Substantial and significant differences in the myoelectric activity
occurred between the impact conditions for the 50 ms before
and the 50 ms after impact, reaching 3 ms in timing, 16% in
wavelet number, and 154% in the intensity of the muscle
activity.
1308
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
METHODS
Approach to the problem. A pendulum apparatus was used
to apply repeatable impacts to the heel. Subjects lay supine
on a bed, which was cyclically swung so that the right heel
impacted a wall-mounted force plate. The amplitude of each
swing was set so that the impact forces were similar to those
experienced during walking and running. A harness supported the leg so that muscle activity was not required for
joint movement of the lower extremity. The frequency component of the impact force was modified by changing the
midsole properties of the subjects’ shoes, with a soft shoe
giving a lower-frequency component than a hard shoe. Muscle activity patterns in the lower extremities were measured
by electromyography (EMG) from the tibialis anterior, medial gastrocnemius, vastus medialis, and biceps femoris muscles. The myoelectric signals were resolved into components
in intensity, time, and frequency to determine which features
of the signals changed between the shoe interventions. The
data were used to test the hypothesis that the muscle activity
required to stiffen the leg for impact differs between the two
impact conditions produced by the shoes.
Subjects. Twenty men (76.4 ⫾ 1.7 kg body mass, 33.3 ⫾ 2.6
yr old) who regularly ran or exercised participated in the
study. They gave their informed consent in accordance with
the University of Calgary Medical Bioethics policy on research using human subjects.
Pendulum apparatus. A human pendulum apparatus was
constructed (Fig. 1), adapted from that described by Lafortune and Lake (18). The pendulum was made from a canvas
bed strapped around a polyvinyl chloride frame, and it was
able to swing freely via polyvinyl chloride rods attached to
the corners. The subject lay supine on the bed, with leg
suspension straps around the ankle and knee maintaining
the right knee at a flexion angle of 20°. In this manner, the
lower leg was held parallel to the ground, and the hip flexion
angle was 20°.
Subjects tested two shoe conditions. The upper, outsole,
and insole materials were identical between the shoes, but
the midsole material varied between a medium elastic (Shore
J Appl Physiol • VOL
Fig. 1. Pendulum apparatus used to test impact forces. PVC, polyvinyl chloride.
C ⫽ 55) and a soft viscous heel (Shore C ⫽ 26). The elastic
shoe had a hardness similar to that of most commercially
available running shoes; the viscous shoe was much softer
and was used to reduce the loading rate of the impact force.
Subjects had size 9–10 feet, and the appropriate shoe size
was provided. The shoes had similar mass: 289.1 g for the
elastic shoe (size 9) and 287.5 g for the viscous shoe (size 9).
The order in which the shoes were tested was varied randomly between subjects. Subjects did not perform exercise
immediately before testing. Each subject performed only one
test session for both conditions.
The pendulum was pulled back to a reference stop and
allowed to swing so that the subject impacted the wall with
his right heel. The impacts were cyclically repeated with one
impact per second. The reference stops were adjusted for
each subject so that the velocity at impact was kept constant
at 1.0 m/s (to the nearest 0.1 m/s as measured using 2
photocells). This impact velocity was used, because it provided an impact force with shape and magnitude similar to
those of impact forces during overground running (17). The
subjects were required to dorsiflex the foot and resist the
motion of the pendulum at impact. A videocamera confirmed
that impacts occurred only with the heel and that the foot did
not “slap” onto the wall. In order for the subjects to adjust to
each new shoe, 25–40 initial impacts were administered
before data collection. Impact forces were recorded for a set of
30 impacts for each shoe condition using a vertically mounted
force plate (Kistler Instrumente, Winterthur, Switzerland),
which was sampled at 1,200 Hz.
Myoelectric signals were recorded from the tibialis anterior, medial gastrocnemius, vastus medialis, and biceps femoris muscles. Bipolar surface electrodes (Ag-AgCl) were
placed on the muscle bellies after removal of the hair and
cleaning with isopropyl wipes. Each electrode was 10 mm
diameter, and the intraelectrode distance was 30 mm. A
ground electrode was placed on the medial malleolus of the
ankle joint. The EMGs were preamplified at the source and
then recorded using a Biovision (Wehrheim, Germany) system at 1,200 Hz.
Wavelet analysis of EMG. Wavelet analysis (33) was used
to resolve simultaneously the intensities of the EMG signals
in time and frequency space. The analysis yields an intensity
that closely approximates the power of the EMG signal. The
procedure to obtain the intensity from the measured EMG
signal consisted of the following three steps: 1) compute the
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
which the loading rate of the impact force can be
experimentally manipulated.
To isolate the changes in muscle activity due to the
ground impact from the muscle activity that is required for joint movement, an experimental procedure
must be used that can minimize the myoelectric activity required for joint movement. One such approach is
to use a human pendulum apparatus (17, 18). Use of
this apparatus involves the subject lying supine on a
bed with the leg supported by a harness. The pendulum motion causes the heel to impact with a wall, and
thus impacts are generated without requiring the
joints to move. The cyclical swinging of the subject on
the pendulum results in a series of impact forces that
mimic those found during running.
In this study, we use such a pendulum technique to
investigate the response of muscle activity in the leg to
different impact forces. To vary the loading rate of the
impact forces experienced by each subject, the viscoelastic properties of their shoe midsoles were experimentally manipulated. This study tested the hypothesis that muscle activity patterns in the lower
extremities change in response to the different impact
loading rates that occur at ground impact.
1309
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
by the calculation of the intensities would otherwise contain
some oscillations that are a result of interference. The width
of the Gauss filter was kept shorter than the time resolution
of the wavelet (width is 3⁄8 of that of the time resolution).
Thus the time resolution was not significantly enlarged.
Definitions. A wavelet domain was defined as the time
series of intensity resolved for one wavelet only. The frequency bands for each wavelet domain are shown in Table 1.
Intensities at frequencies ⬍10 Hz typically contain movement artifacts, and so data from wavelet domain 0 were
ignored. An intensity ij,k was calculated for each sample point
j and wavelet domain k. Time tj is the time of sample j, and
wavelet wk is the wavelet number for wavelet domain k. The
global intensity (Ij) gives the power in the myoelectric signal
at each sampling point and was defined as the sum of the
intensities over wavelet domains 1–7 for each tj
7
Ij ⫽
Center
Frequency,
Hz
Low Band,
Hz
High Band,
Hz
Bandwidth,
Hz
Time
Resolution,
ms
0
1
2
3
4
5
6
7
6.9
19.3
37.7
60.1
92.4
128.5
170.4
218.1
0.3
10.8
25.2
45.7
72.1
104.3
142.4
186.6
14.7
31.4
53.6
82.0
116.3
156.2
202.1
251.7
14.4
20.5
28.4
36.3
44.2
51.9
59.8
65.0
63.3
45.0
36.7
30.0
25.0
21.7
20.0
18.3
J Appl Physiol • VOL
j,k
The mean global intensity (Mi) over a 50-ms time window is
equivalent to twice the square of the root-mean-square
(RMS) value for that window and was given by
60
兺I
j
Mi ⫽
j⫽1
60
Mean time (Mt) for the time-intensity space is the time taken
for the mean myoelectric work to be done within the signal
and gives an index of when the myoelectric activity was
predominant within each 50-ms window. Mt was calculated
as
60
兺tI
j j
Mt ⫽
j⫽1
60
兺I
j
j⫽1
In a similar manner, the total intensity (I⬘k) for a 50-ms time
window for wavelet domain k is
60
I⬘k ⫽
兺i
j,k
j⫽1
The mean wavelet value (Mw) for the wavelet-intensity space
is a correlate of the mean frequency of the myoelectric signal
for a 50 ms time window and is given by
7
兺 w I⬘
k k
Mw ⫽
Table 1. Time and frequency characteristics for the
wavelet analysis
Wavelet
No.
兺i
k⫽1
k⫽1
7
兺 I⬘
k
k⫽1
Time 0 was defined as the time of impact, which was the
moment of heel strike. The subscripts “pre” and “imp” to the
variables Mt, Mw, and Mi denote data from the preactivation
(⫺50 ⱕ tj ⬍ 0 ms) and postimpact (0 ⱕ tj ⬍ 50 ms) windows,
respectively. For each shoe condition and for each subject, the
៮ i, M
៮ t, and
means of Mi, Mt, and Mw for the 30 impacts are M
៮ w, respectively. The intensity values for M
៮ i were normalM
៮ i,imp for the elastic shoe
ized so that the maximum of M
condition had a value of 1.
Statistics. The significance of the muscle responses was
tested with analyses of variance (ANOVAs). Separate bal-
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
wavelet-transformed EMG signal using a filter bank of wavelets that include intensity and damping factors, 2) compute
the intensity of the wavelet-transformed signal by adding its
square and the square of its time derivative divided by the
center frequency, and 3) apply a Gauss filter to the wavelettransformed signal.
In step 1, a filter bank of eight wavelets was used. Each
wavelet was well defined in time and frequency space, and its
integral over time was 0. The set of wavelets was nearly
biorthogonal and was produced by nonlinear scaling. The
scaling was adjusted to obtain a physiologically acceptable
time resolution at all frequencies considering the uncertainty
principle. Center frequency, bandwidth, and the time resolution are indicated in Table 1.
The convolutions between each wavelet and the EMG are
called the wavelet transform. The process is best visualized
in frequency space, where the convolution consists of multiplication of the Fourier-transformed signal by the Fouriertransformed wavelet. Such a multiplication is equivalent to a
filtering procedure. The filtered data (convoluted signal) were
submitted to an inverse Fourier transform to obtain the
wavelet transform of the signal in the time domain.
The wavelets, and thus the convoluted signal, were multiplied by a frequency-dependent intensity factor to take care
of the fact that the amplitude of the EMG signal, as seen at
one frequency point in the Fourier-transformed EMG signal,
contributed to the wavelet-transformed signal of three adjacent wavelets. Splitting the amplitude into three components
and forming the sum of the squares does not yield the
amplitude squared, because the cross terms are missing.
These missing cross terms were included in the analysis in
the following manner: the original wavelets, and thus the
convoluted signal, were divided by the square root of the sum
of the squared wavelet intensities at each frequency in Fourier space.
In step 2, the intensity of the EMG signal was computed.
The intensity of the EMG signal was defined in time space
from the wavelet-transformed signal by adding its square
and the square of its time derivative divided by the center
frequency. If this operation is applied to a pure sine wave,
then it yields the amplitude squared at any time. In the
present case, the computed intensity represents an approximation to the amplitude squared because of errors caused by
the finite sampling frequency. The approximation was further improved by using a damping factor to compensate for
these errors. The damping factors were optimized by matching the intensity obtained by the wavelet analysis of a pure
sine wave with its amplitude squared. The match between
the intensities and the amplitude squared was ⬎95% for all
frequencies in the range 10–250 Hz.
Step 3 consists of smoothing the intensity using a Gauss
filter. This step is required, because the approximation made
1310
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
anced-design ANOVAs were used for each muscle and each
time window (Minitab). The dependent variables used for
these tests were Mt, Mw, and Mi. Subject number, shoe
condition, and a “subject ⫻ shoe condition” interaction term
were used as factors in the ANOVA. Each ANOVA contained
data from both shoe conditions, all 20 subjects, and all 30
trials.
៮ i, M
៮ t, and M
៮ w were calculated between the
Differences in M
elastic and viscous shoe conditions for each subject. The
ranges of these differences are represented by their RMS
value.
Muscle fatigue during the experiment can result in a
reduction in the frequency content of the EMG (5). If the
frequency content of the EMG was systematically changed
with the fatigue state, then the order in which the interven៮ w. The effect of
tions were presented would be related to M
muscle fatigue and the order in which the interventions were
presented were tested in the following way. The differences
៮ w,pre and M
៮ w,imp between the shoe conditions were
in M
scored as ⫹1 or ⫺1 if the elastic condition was higher or lower
than the viscous condition, respectively. These scored values
were then ranked by whether the subject was presented with
the elastic or the viscous shoe condition first. If fatigue had a
significant effect on the results, then it would be expected
that the subjects who tested the elastic shoe condition first
would have a score of 1, and if they tested the viscous shoe
first, then the score would be ⫺1. The distribution of these
ordered scores was tested with a nonparametric runs test
(Minitab) to determine whether there was a significant effect
of muscle fatigue on the myoelectric response.
Values are means ⫾ SE. The range of differences in values
was indicated by their RMS value. Significance was tested at
a confidence level of ␣ ⫽ 0.05. Power analyses (29) showed
that in all cases the power of the ANOVA tests was ⬎99%.
RESULTS
Typical force and EMG for the vastus medialis muscle that resulted from one heel strike are shown in Fig.
2, A and B. The square of the EMG and the global EMG
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
Fig. 2. “Ground” reaction force (A), electromyogram (EMG), and intensity traces for a
single impact. B: EMG data recorded from the
vastus medialis muscle. Time 0, moment of
heel strike. C: rectified, squared signal; thick
line, global EMG intensity summed from
wavelet domains 1–7. EMG intensity was
further subdivided into low-frequency (11–54
Hz), medium-frequency (46–116 Hz), and
high-frequency (104–252 Hz) bands generated from the sum of wavelet domains 1 and
2 (D), domains 3 and 4 (E), and domains 5–7
(F), respectively.
J Appl Physiol • VOL
91 • SEPTEMBER 2001 •
www.jap.org
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
intensity are shown in Fig. 2C. The intensity became
visible 75 ms before impact. In the 50-ms preactivation
phase, the EMG intensity was contained almost exclusively in the summed intensity from wavelet domains 3
and 4 (Fig. 2E). During the 50-ms postimpact period,
additional activity could be observed in wavelet domains 1 and 2 (Fig. 2D).
Individual EMG responses. The EMG intensities for
the vastus medialis muscle for one subject are shown in
Fig. 3. In wavelet domains 1–4 the EMG intensity had
increased beyond the background levels by 50 ms before heel strike. In this example, the intensity was
generally the same between the two shoe conditions for
1311
the preactivation and postimpact periods for all wavelets. The main exception was at wavelet 2 for the 50-ms
preactivation period. Here the total intensity was 243%
greater for the viscous than for the elastic shoe condition, a difference that was significant. There were 160
different combinations of 20 subjects, 4 muscles, and 2
shoe conditions. The ratio of the SE to the mean intensity was calculated for the maximum intensity in the
postimpact period for each of these combinations.
These ratios had a range of 19%, as shown by their
RMS value.
EMG intensities were greater than background levels in the 50-ms preactivation period for the majority of
J Appl Physiol • VOL
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
Fig. 3. EMG intensities in wavelet domains
1–7 for the vastus medialis muscle for 1 subject. Differences in EMG intensity can be localized to a specific time (preimpact) and frequency band (wavelet domain 2). Time 0,
moment of heel strike. Bandwidths for each
wavelet domain are given in Table 1. Thick
dashed line, mean response of 30 impacts in
the elastic shoe; thick solid line, mean response of 30 impacts in the viscous shoe; thin
lines, ⫾1 SE at each time.
1312
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
Fig. 4. EMG intensities showed changes in the
preactivation period for tibialis anterior (A),
gastrocnemius medialis (B), and biceps femoris
(C) muscles for 3 subjects. See Fig. 3 legend for
explanation of lines. Wavelet domain 2 (bandwidth 25–54 Hz) is shown for A and B, and
wavelet domain 3 (bandwidth 46–82 Hz) is
shown for C.
femoris muscle in Fig. 6. This decrease was calculated
for 139 of the 160 subject-muscle-shoe combinations
tested, and ␹2 analysis showed this decrease to be
significantly different from random.
In some subject-muscle combinations, the timing of
the EMG intensity peaks was shifted between the shoe
conditions (Fig. 7). In the examples shown, the viscous
shoe condition compared with the elastic condition had
a 58% peak intensity, which occurred 28 ms later for
the medial gastrocnemius muscle (Fig. 7A), a 124%
peak intensity, which occurred 12 ms earlier for the
vastus medialis muscle (Fig. 7B), and a 229% initial
peak intensity with minimal time shift for the biceps
femoris muscle (Fig. 7C). In other subject-muscle combinations, the EMG intensity differed between shoe
conditions, although the underlying timing remained
Fig. 5. EMG intensities for subjects
1–20 for tibialis anterior (A), medial
gastrocnemius (B), vastus medialis (C),
and rectus femoris (D) muscles. For
each subject, 4 columns are shown for
each muscle in the following order
(from left to right): mean global intensity before impact and after impact
៮ i,pre and M
៮ i,imp) in the elastic condi(M
៮ i,pre and M
៮ i,imp in the viscous
tion and M
condition. Intensities were normalized
៮ i,imp in the elastic
to the maximum M
condition for each subject.
J Appl Physiol • VOL
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
subject-muscle-shoe combinations (Fig. 4). For the tibialis anterior muscle, such preactivation occurred for
every subject. For the medial gastrocnemius, vastus
medialis, and biceps femoris muscles, this preactivation was seen in 75% of the cases. The examples in
Figs. 3 and 4 illustrate the differences in preactivation
intensity that occurred for all four muscles tested. A
៮ i,pre and M
៮ i,imp is shown on a
detailed description of M
subject-by-subject basis for both shoe conditions in Fig.
5. These data showed differences in intensity between
the shoe conditions. The ratios of the mean preactivation intensities to the postimpact intensities across all
subjects were 96, 48, 44, and 73% for the tibialis anterior, medial gastrocnemius, vastus medialis, and biceps femoris muscles, respectively. There was a de៮ w,pre to M
៮ w,imp, as shown for the biceps
crease from M
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
1313
Fig. 8. Changes in EMG intensity for wavelet domain 3 (bandwidth
46–82 Hz) in 1 subject for the biceps femoris muscle between shoe
conditions. See Fig. 3 legend for explanation of lines.
the same. In Fig. 8, the two events occurred at 40 and
70 ms after impact. The intensity of the myoelectric
signal for the viscous shoe condition for the first event
was 135% of that for the elastic shoe but declined to
59% for the second event.
Range of individual EMG responses. Mean values of
៮ t, M
៮ i, and M
៮ w differed between the elastic and viscous
M
shoe conditions for the preactivation and postimpact
periods (Fig. 9). The direction of the response varied for
all measures of EMG activity examined, with some
subjects showing greater values for the elastic shoe
condition and others showing greater values for the
Fig. 7. Changes in EMG timing and intensity for wavelet domain 2
(bandwidth 25–54 Hz) in 1 subject for the medial gastrocnemius (A),
vastus medialis (B), and biceps femoris (C) muscles. See Fig. 3 legend
for explanation of lines.
J Appl Physiol • VOL
៮ i), wavelet number (M
៮ w),
Fig. 9. Difference in the mean intensity (M
៮ t) for each subject between viscous and elastic shoe
and time (M
conditions. { Show a consistency of response for the tibialis anterior
muscle in the preactivation phase; F show a larger range of responses
in the biceps femoris muscle during the postimpact phase.
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
៮ w) for the preactivation (M
៮ w,pre) and
Fig. 6. Mean wavelet number (M
៮ w,imp) periods for the biceps femoris muscle. Data
postimpact (M
points are means ⫾ SE for each subject. There was a significant
៮ w,pre and M
៮ w,imp. In all cases, M
៮ w,pre
positive regression between M
៮ w,imp.
was greater than M
viscous shoe condition. The smallest RMS range for the
differences in response occurred in the preactivation
phase for the tibialis anterior muscle, with RMS values
៮ w,pre for the shift in wavelet number, 26% of
of 5% of M
៮
៮ t,pre
Mi,pre for the shift in intensity, and 1.4 ms for M
(Fig. 9). The largest RMS ranges for the difference in
response occurred in the postimpact phase for the
medial gastrocnemius muscle, with RMS values of 16%
៮ w,imp for the shift in wavelet number and 154% of
of M
៮
Mi,imp for the shift in intensity. The largest RMS range
for a change in timing between the responses was 3.0
៮ t,imp for the biceps femoris muscle (Fig. 9).
ms for M
Significance of EMG responses. The significance of
the differences between Mi, Mw, and Mt between subjects, shoes, and subject ⫻ shoe condition was analyzed
using ANOVA, and the results are shown in Table 2.
Significant differences occurred between the subjects
for Mt, Mw, and Mi of the EMG response, and these
differences were ubiquitous for all four muscles tested.
The shoe condition produced significant differences in
the EMG response in half of the tests (Table 2), and
these differences occurred during the preactivation
and postimpact phases of the EMG signal for Mt, Mw,
and Mi. The magnitude of the changes in response is
1314
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
Table 2. Significance of the muscle response between shoe conditions and between subjects
50 ms Preactivation
Tibialis anterior
Subject
Shoe
Shoe ⫻ subject
Medial gastrocnemius
Subject
Shoe
Shoe ⫻ subject
Vastus medialis
Subject
Shoe
Shoe ⫻ subject
Biceps femoris
Subject
Shoe
Shoe ⫻ subject
50 ms Postimpact
Mt,pre
Mw,pre
Mi,pre
Mt,imp
Mw,imp
Mi,imp
⬍0.001*
0.114
0.176
⬍0.001*
0.030
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.275
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.003
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.251
0.234
⬍0.001*
0.045*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.877
⬍0.001*
⬍0.001*
0.667
0.072
⬍0.001*
0.857
⬍0.001*
⬍0.001*
0.851
⬍0.001*
⬍0.001*
0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.052
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
⬍0.001*
0.890
0.170
⬍0.001*
0.211
⬍0.001*
⬍0.001*
0.079
⬍0.001*
⬍0.001*
0.093
⬍0.001*
shown in Table 3. Significant effects of the subject ⫻
shoe condition interaction term were observed for Mw
for all 8 tests, and significant differences were observed
for 12 of the 16 tests of this interaction for Mt and Mi of
the EMG response. The results from this interaction
term show significant differences in the way in which
the subjects reacted to the two shoe conditions for Mt,
Mw, and Mi.
Runs tests were performed for preactivation and
postimpact phases for all four muscles, and in all cases
there was no significant effect of the order in which the
shoe conditions were presented to the subjects on the
change in their EMG frequency response.
Ground reaction forces. There were significant differences in the maximum impact force (Fmax), the time to
reach this maximum (⌬t), and the mean rate of force
increase (Fmax/⌬t) to this maximum between the two
shoe conditions (ANOVA for 1,200 subject-muscle-shoe
trial combinations). Changing from the elastic to the
viscous shoe condition resulted in a 4.1% decrease in
Fmax, a 10% increase in ⌬t, and a 12.6% decrease in
Fmax/⌬t (Table 4).
DISCUSSION
An earlier experiment using a pendulum device delivered impacts to a subject with a fully extended leg,
and the impact forces were shown to have magnitudes
and loading rates that matched those reported for
running (18). In this study, the knee was held at a
flexion angle of 20° to more closely resemble the leg
posture at heel strike during walking and running. A
20° knee flexion results in a reduction in the magnitude of the impact force. The magnitude and loading
rate of the impact force have been measured for running speeds between 3 and 5 m/s (23), and extrapolation of these data shows that the impact forces experienced during this present pendulum study (Table 4)
had magnitudes corresponding to a 2.5 m/s run and
loading rates corresponding to a 1.1 m/s walk. This
Table 3. Mean muscle responses to the two shoe conditions
50 ms Preactivation
Tibialis anterior
Elastic
Viscous
Medial gastrocnemius
Elastic
Viscous
Vastus medialis
Elastic
Viscous
Biceps femoris
Elastic
Viscous
50 ms Postimpact
៮ t,pre
M
៮ w,pre
M
៮ i,pre
M
៮ t,imp
M
៮ w,imp
M
៮ i,imp
M
25.5 ⫾ 0.2
25.2 ⫾ 0.2
4.39 ⫾ 0.03
4.44 ⫾ 0.02
0.94 ⫾ 0.02*
0.83 ⫾ 0.02*
25.5 ⫾ 0.2
25.3 ⫾ 0.2
3.93 ⫾ 0.04*
4.15 ⫾ 0.03*
1 ⫾ 0.02*
0.84 ⫾ 0.02*
29.1 ⫾ 0.2
28.2 ⫾ 0.2
3.19 ⫾ 0.03*
3.30 ⫾ 0.04*
0.52 ⫾ 0.05
0.44 ⫾ 0.06
30.0 ⫾ 0.2*
29.7 ⫾ 0.2*
2.32 ⫾ 0.04*
2.41 ⫾ 0.04*
1 ⫾ 0.03
0.99 ⫾ 0.07
28.9 ⫾ 0.2
28.6 ⫾ 0.2
3.00 ⫾ 0.02
2.99 ⫾ 0.02
0.42 ⫾ 0.02
0.42 ⫾ 0.02
30.8 ⫾ 0.2*
30.1 ⫾ 0.2*
2.57 ⫾ 0.02*
2.65 ⫾ 0.02*
1 ⫾ 0.02*
0.89 ⫾ 0.02*
25.9 ⫾ 0.2
26.6 ⫾ 0.2
3.19 ⫾ 0.03*
3.07 ⫾ 0.03*
0.75 ⫾ 0.06
0.77 ⫾ 0.07
29.7 ⫾ 0.2
29.9 ⫾ 0.2
2.42 ⫾ 0.03
2.46 ⫾ 0.03
1 ⫾ 0.04
1.12 ⫾ 0.06
Values are means ⫾ SE, pooled from all 20 subjects. * Significant interaction, P ⬍ 0.05, between shoe conditions as shown in Table 2.
J Appl Physiol • VOL
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
Values (ANOVA results, n ⫽ 1,200 for each test) are shown for the probability that no interaction occurred between individuals and shoe
type and mean time (Mt), mean wavelet value (Mw), and mean global intensity (Mi) before (pre) and after (imp) impact. * Significant
interaction, P ⬍ 0.05.
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
Table 4. Characteristics of “ground” reaction force
during heel strike
Shoe Midsole
Fmax, N
⌬t, ms
Fmax/⌬t,
N/ms
Elastic
Viscous
1,058 ⫾ 5.0
1,015 ⫾ 5.6
29.4 ⫾ 0.2
32.3 ⫾ 0.4
37.3 ⫾ 0.3
32.6 ⫾ 0.2
Values are means ⫾ SE, pooled from all subjects (n ⫽ 600). Fmax,
maximum impact force; ⌬t, time from heel strike; Fmax/⌬t, mean
loading rate.
J Appl Physiol • VOL
been demonstrated for graded muscle contractions in
the cat gastrocnemius (30). The timing, activation
level, and motor unit recruitment patterns leave characteristic features within the myoelectric signal. The
ability to resolve time, intensity, and frequency simultaneously gives insight into how the muscles are used
to accomplish a given task. Differences in the timing of
activation can be seen in Figs. 4, 7, and 8; differences in
the intensity of activation can be seen in Figs. 4, 5, 7,
and 8; differences in the frequency of the myoelectric
signal being confined to distinct wavelet domains can
be seen in Figs. 2 and 3.
In this study, the myoelectric signals were resolved
by wavelet analysis into their intensities in time and
frequency space. The intensity represents the power
within the EMG for any given time and frequency
band. EMG power spectra, which are traditionally
used to assess EMG frequency content, are calculated
from the square of the Fourier-transformed EMG signal (21) and are thus also a measure of the power
within the signal. Mw used in this study is comparable
to the mean power frequency used to measure EMG
contractions (32). By contrast, RMS analysis computes
the amplitude, and not a power, of the signal as a
function of time. The square of the RMS value is
comparable to half the intensity from this wavelet
analysis.
The wavelet analysis of EMG signals resulted in a
variety of intensity peaks and leads to the concept of
events in muscle activation. Events can be resolved
over a set of wavelets and thus have their own wavelet
spectrum. Events, which are separated by more than
the time resolution of the wavelet, represent distinct
events within the muscle activation. The example in
Fig. 2, D and E, shows a muscle activation event that
was localized to wavelet domains 3 and 4 and to the
preactivation period. Differences in the EMG signal
were thus characterized by the specific frequency band
and the specific time at which they occurred. Differences that occurred in the time and the frequency
content of the myoelectric signals were statistically
significant (Table 2), thus justifying the use of the
wavelet techniques to decompose the EMG into its
components in time and frequency space. Such separation of the myoelectric signal into time and frequency
components would not be possible using traditional
analysis techniques such as RMS or Fourier transforms. Time-frequency decomposition is possible using
windowed Fourier techniques; however, this approach
is not without its difficulties (15). The time resolutions
for each wavelet were chosen so that they covered
physiologically relevant durations on the order of 20–
40 ms. The higher-frequency wavelets progressively
resolved shorter times (Table 1); however, this resulted
in broader frequency bands for these wavelets because
of the uncertainty principle governing signal processing (15).
Differences in the intensity patterns that resulted
from the different loading rates were limited to a narrow range of wavelet domains (Figs. 2 and 3). The
differences thus occurred at specific frequency bands. A
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
experiment therefore demonstrated that muscle activity patterns were different when the loading rate of the
impact force was varied and when those impact forces
were within the physiological range experienced during walking and running. To isolate the muscle activity
patterns that occurred in response to the impact force
from those required for a complete walking or running
motion, the experiments were conducted on a pendulum with the lower extremity positions constrained.
The supine position of the subjects could affect perceptual and proprioceptive processes that occur during
impact, and this is the trade-off that must be made to
obtain the pendulum-generated impact results. The
impact forces with the elastic shoe had substantially
higher loading rates than those with the viscous shoe.
The different impact forces experienced with the two
shoe conditions resulted in clear differences in the
intensity patterns resolved by this functional EMG
analysis.
When a muscle is activated, it takes time before the
muscle force is fully developed. The time taken to reach
maximum force depends on factors such as the muscle
fiber type, the activation level, and the contraction
dynamics but, for isometric twitches in mammalian
muscle, is in the range 23–73 ms (3, 4, 10). The intensity of the myoelectric signal indicates the activation
level within the muscle. The force achieved by the
muscle depends on the activation level, the number
and fiber type of the muscle fibers activated, the contraction dynamics, and the history of previous contractions (31). Fast- and slow-twitch muscle fibers have
different intrinsic contraction properties (1), and it is
typically thought that slow-twitch fibers are recruited
for low-intensity activities, with a greater proportion of
fast-twitch fibers being recruited as increasing force is
required (6, 13). The frequency components of the myoelectric signal can indicate the muscle fiber type recruitment for any given activity, and a coefficient of
frequency is given by Mw in this study. The frequency
increases with the conduction velocity of the muscle
fiber action potentials (20), and action potentials travel
faster along larger-diameter cells (14). Fast-twitch fibers have diameters 1.2 times those of slow-twitch
fibers in the vastus lateralis (9), suggesting that they
should have higher-frequency components. In addition,
fast-twitch muscle fibers can have faster (1.5 times)
conduction velocities than slow-twitch fibers, even
when they have similar diameters (27). Muscle fiber
recruitment strategies can thus be determined from
the EMG frequency spectra, and this has previously
1315
1316
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
J Appl Physiol • VOL
3 and 20 showed a negligible intensity during the
preactivation period for the medial gastrocnemius, vastus medialis, and rectus femoris muscles, while subjects 1, 8, 9, 13, 16, 18, and 19 showed some periods of
preactivation that had substantially greater intensity
than the following postimpact period. It would seem as
if the individual responses to these impacts can be
grouped into distinct strategies, although the functional significance of the strategies is not yet understood.
Some locomotor studies are challenged by the fact
that muscles are used to move the limbs (to change the
joint angles) and, additionally, to provide the joint
stiffness required for the locomotor task. In such cases,
it can be difficult to segregate these two tasks from the
EMG signals. In this study, the leg motion was minimized, with the lower leg being supported in a harness.
By eliminating the muscle action required to move the
legs, the EMG activity recorded was mainly that necessary for minimizing soft tissue vibrations and for
generating the joint stiffness required to withstand the
“ground” impact. Significantly different myoelectric
patterns were seen in response to the different shoe
midsole conditions and, thus, the loading rate of the
impact force. These changes were present in the timing, intensity, and frequency content of the EMG and
occurred in the period 50 ms before impact and the
period 50 ms after heel strike. If “tuning” is considered
to be the alteration of the mechanical properties of the
leg due to changes in muscle activity, irrespective of
any motion that occurs in the joints, then the results
from this study have shown that the leg is tuned
differently in response to the changes in the loading
rate of the impact force.
We thank Darren Stefanyshyn for help during the study.
Financial support was provided by Adidas International, the Alberta Heritage Foundation for Medical Research, and the Swiss
National Foundation.
REFERENCES
1. Bottinelli R, Pellegrino MA, Canepari M, Rossi R, and
Reggiani C. Specific contributions of various muscle fibre types
to human muscle performance: an in vitro study. J Electromyogr
Kinesiol 9: 87–95, 1999.
2. Brody LR, Pollock MT, Roy SH, De Luca CJ, and Celli B.
pH-induced effects on median frequency and conduction velocity
of the myoelectric signal. J Appl Physiol 71: 1878–1885, 1991.
3. Burke RE, Levine DN, Tsairis P, and Zajac FE. Physiological types and histochemical profiles in motor units of the cat
gastrocnemius. J Physiol (Lond) 234: 723–748, 1973.
4. Burke RE, Levine DN, and Zajac FE. Mammalian motor
units: physiological-histochemical correlation in three types in
cat gastrocnemius. Science 174: 709–712, 1971.
5. De Luca CJ. The use of surface electromyography in biomechanics. J Appl Biomech 13: 135–163, 1997.
6. Denny-Brown D and Pennybacker JB. Fibrillation and fasciculation in voluntary muscle. Brain 61: 311–344, 1938.
7. Dupuis H and Jansen G. Immediate effects of vibration transmitted to the hand. In: Man Under Vibration—Suffering and
Protection, edited by Bianchi G, Frolvlov KV, and Oledzky A.
Amsterdam: Elsevier, 1981, p. 76–86.
8. Frederick EC, Clarke TE, and Hamill CL. The effect of
running shoe design on shock attenuation. In: Sport Shoes and
Playing Surfaces, edited by Frederick EC. Champaign, IL: Human Kinetic, 1984, p. 190–198.
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
net decrease in the frequency content of the EMG may
occur because of recruitment of a greater proportion of
slow- than fast-twitch fibers. However, a decrease in
the frequency content may also be caused by the decrease in action potential conduction velocity that occurs during fatigue because of a lowering of the interstitial pH (2) and is probably due to an accumulation of
metabolic by-products such as lactate within the muscle (22). Thus muscle fatigue may also appear as a shift
in the EMG frequency spectrum (11). However, we
have shown that fatigue was not a significant factor in
these results. Therefore, it can be speculated that the
change in myoelectric frequency patterns that occurs
with different loading rates of the impact force is due to
different patterns of muscle fiber type recruitment.
The preactivation period occurred before heel strike,
and muscle activity within this period was the result of
feedforward control mechanisms. Thus the different
muscle preactivation states that occurred between the
impact conditions (Table 2) must have been the consequence of the different muscle tuning requirements
that were anticipated for each heel strike. The postimpact period began at heel strike and would initially be
the result of feedforward mechanisms. However, toward the end of the 50-ms postimpact period, there
may additionally have been some stretch reflex responses. Short-latency reflex responses have been
shown to occur 41 ms after stretch in the triceps surae
muscles (24) and may occur with even shorter latencies
as low as 34 and 39 ms in the biceps femoris and rectus
femoris, respectively (28). The majority of the postimpact period occurred before the muscle activity could
have been influenced by stretch reflexes. Indeed, differences in postimpact muscle activity before the influence of stretch reflexes can be clearly seen in Figs. 2–4.
Toward the end of the postimpact period, however,
differences in short-latency monosynaptic reflexes may
have additionally contributed to the different muscle
activity patterns.
The intensity patterns were complex in time and
wavelet space. The results could be more easily submitted to statistical analysis by reducing the intensity
patterns to coefficients describing the mean time,
wavelet number, and intensity for the 50-ms time windows before and after impact. Such mean coefficients
have smaller magnitudes than the peak values in the
intensity patterns. Nonetheless, the RMS values for
the differences of the mean responses for all the subjects show that the response of the body to the different
shoe conditions can be of substantial magnitude. A
range of responses occurred across the individuals
៮ t,
tested, with some subjects showing an increase in M
៮
៮
Mw, or Mi between the impact conditions, whereas
other subjects showed decreases in these parameters
(Figs. 5 and 9). Furthermore, testing between elastic
and viscous midsole conditions resulted in large differ៮ t, M
៮ w, or M
៮ i for some subjects, whereas little
ences in M
change was observed in others. The way in which the
muscle activation patterns respond to different loading
rates of the impact force is thus subject specific. Pat៮ i can be seen in Fig. 5; for instance, subjects
terns of M
MUSCLE ACTIVITY TUNED TO GROUND REACTION FORCES
J Appl Physiol • VOL
22. Mortimer JT, Magnusson R, and Petersén I. Conduction
velocity in ischemic muscle: effect on EMG spectrum. Am J
Physiol 219: 1324–1329, 1970.
23. Munro CF, Miller DI, and Fuglevand AJ. Ground reaction
forces in running: a reexamination. J Biomech 20: 147–155,
1987.
24. Nardone A and Schieppati M. Medium-latency response to
muscle stretch in human lower limb: estimation of conduction
velocity of group II fibres and central delay. Neurosci Lett 249:
29–32, 1998.
25. Nigg BM. Impact forces in running. Curr Opin Orthop 8: 43–47,
1997.
26. Nigg BM, Bahlsen HA, Luethi SM, and Stokes S. The influence of running velocity and midsole hardness on external impact forces in heel-toe running. J Biomech 20: 951–959, 1987.
27. Sadoyama T, Masude T, Miyata H, and Katsuta S. Fibre
conduction velocity and fibre composition in human vastus lateralis. Eur J Appl Physiol 57: 767–771, 1988.
28. Schillings AM, van Wezel BMH, Mulder T, and Duysens J.
Widespread short-latency stretch reflexes and their modulation
during stumbling over obstacles. Brain Res 816: 480–486, 1999.
29. Sokal RR and Rohlf FJ. Biometry: The Principles and Practice
of Statistics in Biological Research. New York: Freeman, 2000.
30. Solomonow M, Baten C, Smit J, Baratta R, Hermens H,
D’Ambrosia R, and Shoji H. Electromyogram power spectra
frequencies associated with motor unit recruitment strategies.
J Appl Physiol 68: 1177–1185, 1990.
31. Van Leeuwen JL. Muscle function in locomotion. In: Mechanics
of Animal Locomotion, edited by Alexander RM. London:
Springer-Verlag, 1992, p. 191–250.
32. Viitasalo JHT and Komi PV. Signal characteristics of EMG
during fatigue. Eur J Appl Physiol 37: 111–121, 1977.
33. Von Tscharner V. Intensity analysis in time-frequency space of
surface myoelectric signals by wavelets of specified resolution. J
Electromyog Kinesiol 10: 433–445, 2000.
34. Wakeling JM and Nigg BM. Modification of soft tissue vibrations in the leg by muscular activity. J Appl Physiol 90: 412–420,
2001.
91 • SEPTEMBER 2001 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.32.246 on June 16, 2017
9. Gollnick PD, Armstrong RB, Saubert CW IV, Piehl K, and
Saltin B. Enzyme activity and fiber composition in skeletal
muscle of untrained and trained men. J Appl Physiol 33: 312–
319, 1972.
10. Gonyea WJ, Marushia AS, and Dixon JA. Morphological
organization and contractile properties of the wrist flexor muscles in the cat. Anat Rec 199: 321–339, 1981.
11. Hägg GM. Interpretation of EMG spectral alterations and alteration indexes at sustained contraction. J Appl Physiol 73:
1211–1217, 1992.
12. Hamill J, Bates BT, Knutzen KM, and Sawhill JA. Variations in ground reaction force parameters at different running
speeds. Hum Mov Sci 2: 47–56, 1983.
13. Henneman E, Somjen G, and Carpenter DO. Functional
significance of cell size in spinal motoneurons. J Neurophysiol
28: 560–582, 1965.
14. Hursh JB. Conduction velocity and diameter of nerve fibres.
Am J Physiol 127: 131–139, 1939.
15. Kaiser G. A Friendly Guide to Wavelets. Boston, MA: Birkhaeuser, 1994.
16. Lafortune MA and Hennig EM. Contribution of angular motion and gravity to tibial acceleration. Med Sci Sports Exerc 23:
360–363, 1991.
17. Lafortune MA, Hennig EM, and Lake MJ. Dominant role of
interface over knee angle for cushioning impact loading and
regulating initial leg stiffness. J Biomech 29: 1523–1529, 1996.
18. Lafortune MA and Lake MJ. Human pendulum approach to
simulate and quantify locomotor impact loading. J Biomech 28:
1111–1114, 1995.
19. Light LH, MacLellan GE, and Klenerman L. Skeletal transients on heel strike in normal walking with different footwear.
J Biomech 13: 477–488, 1979.
20. Lindström L, Kadefors R, and Petersén I. An electromyographic index for localized muscle fatigue. J Appl Physiol 43:
750–754, 1977.
21. Lindström L, Magnusson R, and Petersén I. Muscular fatigue and action potential conduction velocity changes studied
with frequency analysis of EMG signals. Electromyography 4:
341–353, 1970.
1317