0 1994Butterworth-Heinemann Ltd

10.50~6411/94/030131-12

Surface and Wire EMG Crosstalk in Neighbouring

Muscles

M. Solomonow, R. Baratta, M. Bernardi, B. Zhou, Y. Lu, M. Zhu and S. Acierno

Bioengineering Laboratory, Department of Orthopaedic Surgery, Louisiana State University Medical Center,
New Orleans, LA 70112, U.S.A.

Summary: Surface and wire myoelectric activity of the medial gastrocnemius

(MG), lateral gastrocnemius (LG) and tibialis anterior (TA) of the cat were
recorded during supramaximal stimulation applied via their nerves before and
after the muscle nerve to the LG and TA were cut in order to determine the
amount of EMG crosstalk amongst neighbouring muscles. It was shown that
the peak-to-peak (pp) amplitude and mean absolute value (MAV) of crosstalk
M waves recorded from the LG and TA after their nerve was cut did not
exceed 5% of their maximal value for surface electrodes and 2.5% of their
maximal value for wire electrodes. EMG crosstalk values were similar in
terms of peak to peak and MAV. Surface EMG crosstalk values were
significantly higher in preparations in which a substantial amount of subcutane-
ous fat covered the muscles, being 20 (5 16.6) % MAV and 16 (2 12) %
p-p. During increasing force contraction (accomplished by orderly recruitment
of motor units) from lO-100% of the maximal force of the MG the
corresponding crosstalk in the LG and TA increased linearly with the EMG
of the MG. It is concluded that the crosstalk problem in surface recording is
negligible for most biomechanical studies in which standard EMG recording
protocol is employed, yet a warning is issued against the indiscriminate
recording of surface EMG from muscles covered by adipose tissue.

Key Words: EMG-Muscle-Electrical stimulation-Evoked potentials-

Volume conduction-Fat.

J. Electromyogr. Kinesiol., Vol. 4, 131-142, September

INTRODUCTION in patients with anterior cruciate ligament deficien-

cies11,15.22. Denny-Brown4, working with animal
Recording of myoelectric signals from the skin
muscles, pointed out that caution should be
surfaces overlying muscles is a common procedure
employed when using EMG as he was able to
in various biomechanical studies such as gait analysis,
record activity from inactive and even denervated
calculations of torques and forces transmitted across
muscles located remotely from an active muscle.
a joint, and in assessment of coactivation patterns
Similarly, Gydikov et aL9 also showed that electrical
activity could be recorded from passive muscles at
Received December 7, 19913. various distances from an active muscle, probably
Revised February 18, 1994.
Accepted February 23, 1994. due to volume conduction in tissues of non-homo-
Address correspondence and reprint requests to Prof. M. geneous and non-isotropic nature. Fetz and Cheney’,
Solomonow, Bioeigineering Laboratory, Department of Ortho-
paedic Surgery, Louisiana State University Medical Center, 2025
however, used cross-correlation techniques to con-
Gravier Street, Suite 400, New Orleans, LA 70112, U.S.A. clude that EMG recorded from muscles other than

131
132
the signal source are negligible. English and Weeks5
further elaborated that EMG signals arising from
one motor unit could be recorded with wire
electrodes within the various compartments of
the same muscle. Recently Etnyre and Abraham6
demonstrated that during voluntary contraction of
the tibialis anterior (TA), wire electrodes located
in the soleus muscle did not record any electrical
activity whereas surface electrodes over the same
muscle detected a sizeable myoelectric signal. These
findings cast a shadow on the reliability of surface
recordings, suggesting that EMG crosstalk from an
active muscle may exist amongst neighbouring
muscles, which may cause scientific interpretation
of the results to be incorrect. Furthermore, it is
also questionable as to whether wire electrodes do
not suffer from the same problem.
Limited and inconclusive attempts3v1* were made
to quantify the relative amount of EMG signal that
may be recorded in muscles near an active muscle.
DeLuca and Merletti3 applied surface electrical
stimulation at 20 pulses per second (pps) to actuate
the TA muscle of human subjects while recording
surface myoelectric signals from the peroneus brevis
(PB) and soleus (SOL) muscles. They concluded
that as much as 16.6% of the EMG from the TA
can be recorded elsewhere in the leg. Similarly, Koh
and Grabiner12 used surface electrical stimulation of
the human quadriceps femoris via the femoral nerve
to assess EMG crosstalk from the medial and lateral
hamstring group. They measured average crosstalk
of 17% and 11.3% of the quadriceps EMG in the
lateral and medial hamstrings, respectively.
Unfortunately, both studies suffer from major
common limitations. Surface electrical stimulation,
at low or high intensities, is associated by various
reflex arcs that originate from numerous receptors
located in the cutaneous tissue under the electrodes
or in the mixed sensory-motor nerve of the muscle
and conveyed by complex oligosynaptic pathways
to different muscles of the limb under study’,13.
This by itself may have caused increased EMG
activity in muscles other than the stimulated muscle
and may be erroneously considered to be crosstalk.
Similarly, both studies failed to ascertain if wire
EMG confirms or contradicts their results in order
to establish if the recorded signals were indeed
crosstalk. Furthermore, electrical stimulation
applied to the surface of the skin overlaying a
muscle normally excites skin receptors before the
muscle is fully excited. On increasing the intensity
of the stimulus, the small nerve fibres of the
Journal of Electromyography & Kinesiology Vol. 4, No. 3, 1994
M. SOLOMONOW ET AL.
cutaneous pain receptors become excited while
simultaneous muscle excitation results in only
4040% of its maximal force value. Based on
our long-term personal experience with surface
stimulation19, we have yet to observe a normal
subject that can tolerate stimulus amplitudes in
excess of 6045% of that required to induce maximal
stimulation before an excruciating pain results. This
calls into question whether the muscles in the
DeLuca & Merletti and Koh & Grabiner studies
were indeed supramaximally stimulated. Without
that factor as a reference, the data provided in
those two studies could not be used in assessing the
maximal expected crosstalk in practical applications.
Finally, both studies have expressed the crosstalk
as a percentage of the maximal EMG from the
signal source muscle EMG that could be found
elsewhere in the leg. This method of presentation
does not provide the biomechanician with direct
and reliable means of assessing how much of
the EMG recorded from one muscle should be
subtracted as it may possibly be crosstalk from a
neighbouring muscle. For example, what percentage
of the EMG recorded from the gastrocnemius
muscle is crosstalk from the TA? With this type of
information, one may accurately correct the
recorded EMG to a high level of accuracy.
The objectives of this study are to determine,
using an animal model, if surface as well as a wire
EMG crosstalk indeed exists; and if so, what is the
extent of its influence on neighbouring muscles
during true supramaximal stimulation of one muscle,
while the others are denervated. The relationship
of EMG from the signal source muscle to its
crosstalk in its neighbours during the full range of
submaximal activity will also be investigated to
provide investigators with the means to correct their
data collected in various biomechanical studies.
METHODS
Preparation
Twelve adult cats were anaesthetized with chlora-
lose (60 mg kg-l). An incision was made above the
popliteal fossa to expose the sciatic nerve just above
the peroneal branch. The nerve was gently cleaned
proximally and distally and all branches were
denervated except that of TA, medial gastrocnemius
(MG) and lateral gastrocnemius (LG). A stainless
steel tripolar cuff electrode was placed on the
exposed nerve for later connection to a stimulator.
 EMG CROSSTALK IN MUSCLES
Special care was taken not to disturb the skin below
the knee region by an incision so normal connective
tissue, muscle and skin condition were available for
EMG recording. Both hindlimbs were shaved twice;
first with a coarse shaver and then with a fine razor
to avoid the interference of hair with surface
recordings. The smooth skin was then abraded with
an alcohol pad to ensure good conductivity with
the surface electrodes.
After each experiment was completed, the leg
was further dissected to confirm that electrodes
were placed correctly in or on the designated muscle
and to identify the possible presence of subcutaneous
fat which may cause significant crosstalk”. The
animal was then frozen, and the mid calf (at the
level of the electrodes) was transversely cut to
identify the geometry of the muscles/bones.
Instrumentation
Stimuli were delivered by a dual-channel com-
puter-controlled stimulation system described in
detail by Zhou et a1.23 and Solomonow et a1.n’.
Briefly, an IBM-XT computer delivered two control
voltages to two stimulators. The first stimulator, a
voltage-controlled oscillator, delivered 100 ps pulses
of suprathreshold amplitude and rate that could be
linearly increased from 2-200 pps. Upon application
of a given constant voltage, the pulse rate could be
fixed at a corresponding desired value. This stimulus
was, therefore, designated as the firing rate (FR)
stimulus and applied to the nerve via the two
proximal electrode poles.
The second stimulator multiplied a linearly
decreasing voltage with a train of 100 ps pulses
at 600 pulses s-l. Because such high-frequency
stimulation blocks the transmission of action poten-
tials elicited by the FR stimulus to the muscle,
contraction was inhibited14~20~21.As the pulse ampli-
tude of this stimulus was decreased, the smallest
axons, having the highest excitation threshold to
electrical stimulus, escaped the block first and
initiated a mild contraction commensurate with their
low innervation ratio and low FR from the FR
stimulus. Further decrease of the stimulus amplitude
allowed motor units of progressively larger size to
escape the block imposed by this stimulus and to
conduct action potentials generated by the FR
stimulus to the muscle*. Under such conditions, the
average motor unit recruitment pattern was executed
in an orderly manner according to sizelo, with each
newly recruited motor unit excited at the prevailing
133
firing rate. This stimulus was designated as the
recruitment (REC) stimulus, and was delivered by
the two distal poles of the electrode.
Surface EMG recordings were made with (Ag-
AgCl) bipolar electrodes, each of 3 mm diameter
and 7.5 mm centre to centre distance placed over
the midline of each muscle slightly distal to the
motor point. Wire EMG was recorded with, two
insulated fine stainless steel wires with a 1 mm
exposed tip as the active electrodes and a common
electrode placed in a remote location in the animal.
The active electrodes were inserted in the muscle
with the aid of hypodermic needles with an interelec-
trode distance of 3-4 mm. The insertion location
was selected about one-half the distance from the
motor point and the distal tendon to reduce
movement and stimulus artifacts**. The EMG was
differentially amplified with common mode rejection
ratio of 100 dB, gain up to 200000 and bandpass
filtered over 10-500 Hz16. Differential recording of
the EMG was selected to reduce the effect of noise
and other undesirable artifacts to an insignificant
level relative to the myoelectric signal. The EMG
was sampled at 3000 samples s-’ and stored on an
IBM-AT computer for later analysis. Force was
isometrically measured with a Grass ET-10 trans-
ducer attached with a linkage to a limb restraint
system hinged about the ankle joint.
Force, EMG, and the two control voltages were
also recorded on a Gould 260 polygraph. The EMG
was also monitored on an oscilloscope.
Protocol
Initial trials consisted of adjusting the stimulus
amplitude to supramaximal threshold by gradually
increasing its intensity until no further increase in
twitch force and M-wave amplitude was detected.
Stimulus frequency of 10 pps was then applied and
a set of four trials each of 6 s duration was applied.
Only 3 s of each trial were recorded for analysis.
Record of each trial consisted of a set of six EMG
traces (wire and surface EMG of each of the three
muscles) and a force trace. The nerve branches to
the LG and TA were then cut gently without
manipulating the leg, and another set of four trials
as described above was taken. Finally, with stimulus
frequency set at 40 pps, and recruitment stimulus
thresholds adjusted properly, orderly recruitment
of motor units was executed over a period of 2-3 s
to result in an increasing force contraction from
lO--100% of the maximal force of the muscle.
Journal of Electromyogrophy & Kinesiology Vol. 4. No. 3. 1994
134
Surface and wire EMG traces as well as force were
recorded in four recruitment trials.
Analysis
Data were processed in three ways. First, the
peak-to-peak of each M wave, and the mean
absolute value (MAV) of each M wave (mean value
of the full wave rectified M wave in a 25 ms
window) recorded from each muscle before and
after the nerves were cut were calculated and
normalized with respect to their corresponding
values before the nerves were cut. This was perfor-
med to assess how much of the EMG recorded
from a muscle exerting its maximal force is due to
crosstalk in terms of percentage peak-to-peak (p-p)
and MAV.
Second, the EMG (p-p and MAV) recorded from
the LG and TA after their nerves were cut were
normalized with respect to the corresponding values
of the MG (the only active muscle throughout the
experiment). This was done in order to assess how
much of the EMG from a maximally active muscle
could be expected to contaminate a neighbouring
muscle, and for purposes of comparison with
previous studies3T12.
The third test was performed in order to assess
the relationships of the EMG recorded from a
muscle which increases its force from lO-100% of
its maximal to the crosstalk signal which is recorded
from inactive neighbouring muscles. This infor-
mation is useful in many scientific applications
where muscles are active at various submaximal
levels during daily functions such as walking,
climbing stairs, etc., which are submaximal. In this
procedure, the EMG (p-p and MAV) from each
M wave of the LG and TA (nerves cut) were
normalized with respect to their maximal M wave
before the denervation.
RESULTS
Figure 1 shows a typical recording of wire and
surface M wave trains from the MG, LG and
TA during suprathreshold stimulation at 10 pps.
Figure 2 shows wire and surface recordings of the
same preparation after the nerves to the LG and
TA were cut.
Table 1 gives the p-p and MAV of the mean
(+ SD) EMG p-p and MAV recorded from all the
MG, LG and TA preparations after the nerves
were cut. In this table the EMG of each muscle
Journal of Elecnomyography & Kinesiology Vol. 4, No. 3, 1994
M. SOLOMONOW ET AL.
was normalized with respect to itself before the
nerves were cut. For wire recordings, the mean
(+ SD) EMG crosstalk recorded from the LG was
2.02 (? 0.53) % MAV and 1.94 (? 1.0) % pp
and the mean EMG crosstalk from the surface
electrodes was 4.65 (? 1.68) % MAV and 4.55
(* 2.39) % p-p. For the TA, the mean wire EMG
crosstalk was 2.67 (5 1.39) % MAV and 1.63
(2 1.09) % p-p and the mean surface crosstalk was
5.14 (2 4.15) % MAV and 3.99 (+- 2.99) % p-p.
The EMG data of the MG, being the signal source
of crosstalk in the LG and TA was not expected
to change significantly, demonstrating only minor
(2-3%) changes due to typical intertrial variability.
Table 2 provides the mean (? SD) EMG crosstalk
in the LG and TA when normalized with respect
to the MG mean EMG, in order to assess how
much of the EMG signal from the MG can be
expected to be present in its neighbouring muscles
and for comparison with previous studies. The mean
wire EMG crosstalk in the LG was 2.58 (’ 1.17)
% MAV and 2.0 (2 0.88) % p-p and the mean
surface EMG crosstalk was 5.75 (* 2.79) % MAV
and 4.45 (? 1.03) % p-p. For the TA, the mean
wire crosstalk was 1.57 (? 0.73) % MAV and 1.21
(? 0.65) % p-p and the mean surface crosstalk was
2.29 (2 0.92) % MAV and 1.99 (* 0.77) % p-p.
Figure 3 shows a typical recording of wire and
surface M wave trains from the three muscles during
an increasing force contraction (orderly recruitment
of motor units) in which the force of the MG
increased from lO-100% of its maximal after cutting
the TA and LG nerves. The force is shown at the
bottom.
Figure 4 shows the relationship between the EMG
of the MG to the EMG recorded from the LG and
TA (with nerves cut) for increasing force contraction
from lO-100% of the MG’s maximal force. The
pooled means from five preparations were fitted
with first order linear regression curves having
correlation coefficients larger than 0.91. The EMG
of the TA and LG was normalized with respect to
its own maximal EMG before the nerves were cut.
Figure 5 shows a typical surface and wire M wave
trace similar to that of Figure 2 taken from a
preparation in which an appreciable layer of subcut-
aneous fatty tissue was discovered in the post-
experimental dissection. Five animals exhibited a
similar amount of significant crosstalk which was
directly traced to the subcutaneous fatty tissue
extending from the dorsal fat pad behind the knee
joint. The data collected from these preparations
 EMG CROSSTALK IN MUSCLES 135

3 r-

200 300
Time (ms)

FIG. 1. Typical M wave trains recorded with surface and wire electrodes from the MG, LG and TA during supramaximal stimulation
of their nerves (w wire recordings and s surface recordings).

were not included in Tables 1 and 2 and are
presented separately in Table 3. In these animals
with significant fatty tissue the mean wire crosstalk
was 3.77 (& 1.69) % MAV and 3.63 (-’ 1.77) %
p-p, and mean surface crosstalk was 20 (2 16.6)
% MAV and 16.2 (* 12.1) % p-p for the LG. In
the TA, the mean wire crosstalk was 2.18 (2 0.94)
% MAV and 1.52 (* 0.58) p-p % and mean surface
crosstalk was 5 (* 2.22) % MAV and 5.04 (* 2.82)
% p-p. Fatty tissue was never observed in the
subcutaneous tissue overlaying the TA, and was
always limited to the area overlaying the MG and
LG.

Journal of Electromyography & Kinesiology Vol.4. No. 3, 1994

136 M. SOLOMONOW ET AL.

3
2

9
ZO k-
I.
z -1
-2
-3 I J
3

‘;;
2
1
c 1
A A A A A
ZO
r: -1
-2
-3 J
3
2-

? ‘-
To
vl cl -I-
ti -2 -

?
.# -;

z
z:, :_
10
z”
Ll -1 -
-2 -
-3
3
2-

7 ‘-
Y0
b -1 -
-2 -
-3
3
2-
- cn l-

To
b -1 -
-2 -
-3 I I I I
0 100 200 300 400 500
Time (ms)
FIG. 2. M wave trains recorded from the same preparation as Figure 1 after the LG and TA were made inactive by cutting their
nerves.

DISCUSSION poses in most biomechanical studies. Surface

Several important facts emerge from the results
of this paper. Crosstalk was found in denervated
muscle located near an active muscle.. The crosstalk
recorded during supramaximal activity of the
signal source muscle is always less than 5% for
surface recordings and less than 2.67% for wire
recordings; a negligible amount for practical pur-
recordings are, however, heavily contaminated
with crosstalk if a subcutaneous layer of fat .is
located under the electrodes. Finally, for the full
range of force, the crosstalk is linearly related to
the level of activity of the signal source muscle,
such that submaximal contractions in a given
muscle are manifested in even lower crosstalk in
neighbouring muscles.

Journalof Electromyography & Kinesiology Vol. 4, No. 3, 1994

 EMG CROSSTALK IN MUSCLES 137

TABLE 1. Mean -t SD of EMG activity after cutting the nerves to the LG and
TA muscles expressed as a percentage of the activity that was present before
their muscle nerves were cut

MAV Peak-to-peak
--
n=7 Wire Surface Wire Surface

MG 104.57 f 5.92 101.07 + 7.51 103.57 ‘- 4.70 97.96 f 4.76

LG 2.02 ” 0.53 4.65 ? 1.68 1.94 t 1.00 4.55 t 2.39
TA 2.67 2 1.39 5.14 ? 4.15 1.63 t 1.09 3.99 r 2.99

TABLE 2. Mean _’ SD of EMG crosstalk in the LG and TA muscles expressed

as a percentage of the activity in the MG

MAV Pea k-to-peak

_

n=7 Wire Surface Wire Surface

LG 2.58 ? 1.17 5.75 + 2.79 2.00 + 0.88 4.45 t 1.93

TA 1.57 2 0.73 2.29 ? 0.92 1.21 -c 0.65 1.99 -c 0.77

The fact that EMG crosstalk does exist confirms
the results of the various studies which first described
it 4*5,9.The relative amount of the crosstalk due to
maximally activated muscle is, however, extremely
small, ranging less than 5% for surface recordings.
Such a low value could be simply neglected, as it
will have little or no impact on the results of various
biomechanical studies where relatively large changes
in EMG are expected upon changes in force and
torques15,22. It is also questionable if this amount
of crosstalk is relevant in neurophysiological studies
where gross or integrated performance of a muscle
is investigated. An exception could be work in
which single unit recordings are the objective,
similar to the conditions described by English and
Week$.
The fact that the crosstalk is rather negligible for
most practical studies is further compounded by the
fact that the experimental conditions of this study
created the background for the ‘worst case’ con-
ditions. In normal voluntary contraction of a muscle,
the active motor units fire asynchronously relative
to each other. This fact is manifested by relatively
lower amplitude of the resulting EMG than that
recorded during synchronous stimulated contraction
such as in this study. Lower amplitude EMG, as
Figure 4 implies, yields much lower crosstalk in
neighbouring inactive or active muscles. The 5%
crosstalk recorded from surface electrodes placed
over an inactive muscle is, therefore, an over-
estimation of the crosstalk that can be expected in
voluntary contraction.
Furthermore, the cat’s hindlimb used as the
preparation in this investigation is substantially
smaller than that of the human leg. The distances
which any EMG signal has to travel in the volume
conductor are, therefore, much shorter, further
enhancing the over-estimation of the crosstalk
reported in this study. On the other hand, in
humans, the larger muscles may generate more
EMG interference. Combining the opposing effects
of larger distances and increased signal intensity,
the net effect is that the proportions are similar
and therefore the crosstalk values found in the cat
are expected to be valid for human studies as well.
Using single differential recording, DeLuca &
Merletti3 reported 7% p-p and 6.5% MAV crosstalk
at the PB and 5% p-p and 5.4% MAV crosstalk
at the SOL of stimulated human subjects. These
values are percentages of the myoelectric signal of
the stimulated TA recorded over the alleged ‘inac-
tive’ muscles. This data could be compared with
the data presented in Table 2. Table 2 indicates
that 5.75% MAV and 4.45% p-p crosstalk was
recorded from the LG. For the TA, 2.29% MAV
and 1.99% p-p crosstalk was recorded. Although
the values are similar or lower to those reported
by DeLuca & Merletti3, one should note that the

Journal of Electromyography & Kinesiobgy Vol. 4. No. 3, 1994

 138 M.SOLOMONOWETAL.

TO.025 1

~0.000 -
b
-0.025 -

, / I -I
1000 1500 2000 2500
Time(ms)

FIG. 3. M wave trains of the three muscles (after denervation of the LG and TA) during an increasing force contraction. Note the
amplified scales of the LG and TA recorded relative to the scales of the MG records.

experimental conditions were not identical. While
supramaximal stimulation of the MG (the EMG
signal source) was fully ascertained in this study,
DeLuca & Merletti were unable to ascertain that
condition, stating that “only part of the tibialis
anterior was activated”. With fully activated muscle
the crosstalk values may have been larger, probably
due to various reflex arcs which may tend to increase
the activity of the PB and SOL. The exact crosstalk
and the level of activity of the muscle are not fully
clear, nor is the method by which one may use the
data possibly to correct for the crosstalk.
Koh & Grabinerl* reported 17.1% MAV and
14% p-p crosstalk in the lateral hamstrings due to
lateral vastus contraction. Similarly, 11.1% MAV
and 9.2% p-p crosstalk from the lateral vastus was

Journal of Elecmomyography & Kinesiology Vol. 4, No. 3, 1994

 EMG CROSSTALK IN MUSCLES

Surface MAV Surface P-P

___ 1

I 20 40 60 80 100

NORMALIZED MG MAV (%';) NORMALIZED MG P-P (e,

Wire MAV Wire P-P

0 ?C 10 60 80 100 0 20 40 60 80 13il

NORMALIZED MG MAV (2) NORMALIZED MG P--P (5)

FIG.4. The relationship between the EMG (MAV and p-p) of the MG and the crosstalk in the LG and TA during increasing force
contraction. a, The relationship for surface recordings in terms of MAV, LG. R = 0.975, TA, R = 0.983, n = 5; b, the relationship
for surface recordings in terms of p-p, LG, R = 0.989, TA, R = 0.989, n = 5; c, the relationship for wire recordings in terms of
MAV, LG, R = 0.925, TA, R = 0.913, n = 5; d, the relationship for wire recordings in terms of p-p, LG, R = 0.915, TA, R = 0.979,
n = 5. Note that for clarity, only the positive SDwas plotted for the LG and only the negative SDwas plotted for the TA. The crosstalk
of the LG and TA is expressed as a percentage of their maximal EMG before their nerves were cut.

present in the medial hamstrings. Their values are
two to three times larger than those reported here,
and those of DeLuca & Merletti3. Again, one must
note their protocol in which the EMG crosstalk
recorded from the hamstrings during quadriceps
stimulation was normalized with respect to the
EMG recorded during 100% maximal voluntary
contraction of the quadriceps. They claim that the
quadriceps were stimulated to a level where the
mean frequency of the EMG power density spectrum
was identical to that obtained during maximal
voluntary effort. This assumption is rather question-
able as well as the value of normalizing EMG of
stimulated muscle with respect to that of a voluntary
contracting muscle. In any case, the values of
crosstalk are extremely high, representing the effects
of reflex arc and their normalization technique.
The crosstalk recorded from wire electrodes was
much lower than the corresponding surface crosstalk.
This, in essence, represents the commonly accepted
fact that wire electrodes have a smaller recording
territory, being less exposed to signals from distant
sources and more sensitive to nearby electrical
activity. The values of the wire crosstalk were about

Journal of EIecIromyography & Kinesiology Vol. 4. No. 3, 1994

140 M. SOLOMONOW ET AL.

100 -
I
- 50-
.E
0-J
rp
3 -50 -

-100 -

100 -

- 50-
VI
! c n I'
g0

-50 -

-100 -

100 -

50 -
Y
4"
-50 -

-100 -

100 -

A_ h
V V

-100 -

100 -

50 -
7
5 O
-50 -

_ 00 c

/-\
m

z
-100
r I I I

0 100 200 300 400 500

Time (ms)

FIG. 5. An M wave train of surface and wire EMG recordings from the MG, LG and TA taken from a preparation with appreciable
subcutaneous fat covering the proximal half of the MG and LG muscles. The nerves of the LG and TA were cut. Note the visibly
appreciable amplitude of crosstalk in the LG. The scales of the M wave amplitude were given normalized with respect to the maximal
p-p amplitude of the M wave of the respective muscle when supramaximally, stimulated before cutting its nerve.

half of their respective surface crosstalk values. preferably employ wire electrodes instead of surface
This also confirms the observations of Etnyre & ones.
Abraham6, which point out that surface crosstalk A significant amount of crosstalk was found when
may be detected while simultaneous wire recordings recording surface myoelectric activity from muscles
may be at the noise level. Investigations that require covered by adipose tissue. In the LG,
a high degree of accuracy of EMG recordings should 20.05 + 16.62% MAV and 16.25 + 12.15% p-p of

Journal of Electromyography & Kinesiology Vol. 4, No. 3, 1594

 EMG CROSSTALK IN MUSCLES
TABLE 3. Mean k SD of EMG activity after cutting the nerves to the LG and
TA muscles expressed as a percentage of the activity that was present before
their muscle nerves were cut, in those cats carrying excessive fat
MAV
n=5 Wire Surface
MG 99.80 f 2.05 100.00 2 8.66
LG 3.77 f 1.69 20.05 2 16.62
TA 2.18 + 0.94 5.00 f 2.22
EMG at supramaximal contraction was due to
crosstalk. It should be noted that the standard
deviation of the mean values was also extremely
large, indicating that in some conditions as much
as 37% MAV and 28% p-p crosstalk may be
expected when a layer of fat exists between the
muscle and the electrodes. In preparations that
displayed appreciable crosstalk, the location of the
surface electrodes was changed several times in an
attempt to reduce the crosstalk prior to dissection.
Such efforts were not productive indicating that
electrode location was not an important factor.
After dissection, and identification of the fatty
tissue over the muscles in question, it was clear
that the amount of fat in thickness and surface area
varied from preparation to preparation. This could
be the source of the large variability in the amount
of crosstalk found in these cases.
Although the experimental conditions of this
study delineate that the ‘worst case’ crosstalk that
one may expect is negligible, the values obtained
from preparations with a substantial padding of
subcutaneous fat is appreciable. In fact, using. a
qualitative method to identify possible crosstalk
during testing of a large number of human subjects,
it was found that subjects who had noticeable
adipose tissue also displayed significant crosstalk
amongst the flexor and extensors of the elbow17.
The data of this study confirm experimentally our
previous observations. Further confirmations that
the subcutaneous fat layer is associated with excess-
ive crosstalk are the simultaneous recordings from
the wire electrodes. In the LG the wire crosstalk
amounted to only 3.77% MAV and 3.63% p-p,
similar values to the crosstalk recorded from prep-
arations in which fatty tissue was absent.
The reason for the increased crosstalk manifested
by the presence of fatty tissue is unclear and
deserves a full scientific investigation. Yet, the
141
Peak-to-peak
Wire Surface
99.0 2 2.24 91.60 1?I 6.27
3.63 2 1.77 16.25 f 12.15
1.52 2 0.58 5.04 ? 2.82
message to investigators is clear; EMG surface
recording from muscles covered by subcutaneous fat
is not reliable for scientific or clinical interpretations!
Wire recordings should be employed in such cases.
Typical muscles from which surface recordings
should be avoided are the gluteus, abdominals and
most muscles of individuals with a noticeable overlay
of adipose tissue.
Although the LG displayed significant crosstalk
in the presence of subcutaneous fat, the TA did
not. Upon dissection, it was clear that the fatty
tissue was limited to the proximal region of the
calf, extending from the fat pad commonly found
behind the knee and covering only the dorsal aspects
of the leg. The anterior calf was free of fat, which
is reflected in the low crosstalk values of the surface
EMG from the TA as shown in Table 3. These
findings provide additional confirmation that subcut-
aneous fat is the source of the significant crosstalk
found in the LG.
The relationships between the surface and wire
EMG of the MG to the crosstalk found in the inactive
LG and TA during increasing force contraction was
estimated by a first order linear regression curve.
The curves shown in Figure 4 describe a linear
relationship between the EMG level of the active
MG to the crosstalk found in LG and TA. This
implies that during submaximal contraction, the
‘worst case’ crosstalk in the absence of subcutaneous
fat is much lower than 5% for surface recordings
and lower than 2% for wire recordings. For example,
when the MG is active up to 50% of its maximal
level, the manifested crosstalk in the LG and TA
is less than 2.1% MAV and less than 2.3% p-p
respectively. This is within the range of the noise
level of most instrumentation used by biomechanici-
ans, and could simply be ignored.
In biomechanical studies that require a high level
of accuracy, the curves of Figure 4 could be used
Journalof Elecrromyography & Kinesiology Vol. 4, No. 3, 1994
142 M. SOLOMONOW
to correct the data for the effect of EMG crosstalk.
One should note, however, that the curves represent
‘worst case’ correction, since they overestimate the
crosstalk due to the synchronous discharge of
the stimulated muscle. While being ‘worst case’
estimates of crosstalk, the curves of Figure 4 rep-
resent experimentally based data obtained in highly
controlled conditions with full confirmation of the
electrode location and the absence of subcutaneous
fat verified by post-trial dissection.
Finally, it could be concluded that for surface
recording of the EMG with the appropriate size of
electrodes, correct placement over the muscle and
inter-electrode distance8 one should disregard the
effect of crosstalk in most skeletal muscles of the
extremities, spine and upper trunk. The possibility
of crosstalk should be carefully considered in
abdominal and buttock muscles and in individuals
with adipose tissue. For biomechanical studies
considering submaximal contractions such as walk-
ing, lifting and general motion of the extremities,
the crosstalk is even less pronounced, being at the
noise level and of no concern for the investigator.
Acknowledgement: This work was supported with
National Science Foundation Grant BCS-9207007.
REFERENCES
1. Arsenault B, Belanger A, Durand M, DeSerres S, Fortin
L, Kemp F: Effect of TENS and Topical Anesthesia on
soleus H-reflex and the concomitant influence of skin muscle
temperature. Arch Phys Med Rehabil 74:48-53, 1993.
2. Blair E, Erlanger J: A comparison of the characteristics of
axons through their individual electrical responses. Am J
Physiol 106:524-564, 1933.
3. DeLuca C, Merletti, R: Surface myoelectric cross talk among
muscles of the leg. EEG & Chin Neurophysiol 69~568-575,
1988.
4. Denny-Brown D: Interpretation of the electromyogram.
Arch Neural Psych 61:99-128, 1949.
5. English A, Weeks 0: Electromyogram cross talk within a
compartmentalized muscle of the cat. J Physiol416:327-336,
1989.
Journal of Electromyography & Kinesiology Vol. 4, No. 3, 1994
ET AL.
6. Etnyre B, Abraham L: Comparison of wire and surface
electrode recordings between antagonist pairs of muscles.
Proc 6th ISEK Congress, Tokyo pp 85, 1985.
7. Fetz E, Cheney D: Postspike facilitation of forelimb muscle
activity by primate corticomotorneuronal cells. J Neurophy-
siol44:751-772, 1980.
8. Fugelvand A, Winter D, Patla A, Stashuk D: Detection of
motor unit action potentials with surface electrodes: Influence
of electrode size and spacing. Biol Cybernet 67:14>153,
1992.
9. Gydikov A, Gerilovsky L, Gatev P, Kostov K: Volume
conduction of motor units potentials from different human
muscles to long distances. EMG & Clin Neurophysiol
22:105-116, 1982.
10. Henneman E, Somjen G, Carpenter D: Functional signifi-
cance of cell size. I Neurophysiol28:560-580, 1965.
11. Hof A, Pronk C, Best J van: Comparison between EMG
to force processing and kinetic analysis for the calf muscle
moment in walking and stepping. J Biomech 20:167-178,
1987.
12. Koh T, Grabiner M: Cross talk in surface EMG of human
hamstring muscles. J Orthop Res 10:701-709, 1992.
13. Lee K, Johnson R: Electrically induced flexion reflex in gait
training of hemiplegic patients: Induction of the flexion
reflex. Arch Phys Med Rehabil 57:311-314, 1976.
14. Solomonow M, Eldred E, Lyman J, Foster J: Control of
muscle force by indirect high frequency stimulation. Am J
Phys Med 62:71-82, 1983.
15. Solomonow M, Baratta R, Zhou B, Shoji H, Bose W, Beck
C, D’Ambrosia R: The synergistic action of the ACL and
thigh muscles in maintaining joint stability. Am J Sports
Med 15:207-213, 1987.
16. Solomonow M, Baratta R, Zhou B, Shoji H, D’Ambrosia
R: The EMG-force model of electrically stimulated muscle.
IEEE Trans Biomed Eng 34:692-703, 1987.
17. Solomonow M, Baratta R, Zhou B, D’Ambrosia R: EMG
coactivation patterns of the elbow antagonist. Exp Neurol
100:47tX477, 1988. .
18. Solomonow M, Baten C, Smit J, Baratta R, Hermens H,
D’Ambrosia R, Shoji H: EMG power spectral frequencies
associated with motor units recruitment strategies. J Appl
Physiol 68:1177-1185, 1990.
19. Solomonow M: Bibmechanics and physiology of a practical
functional neuromuscular stimulation powered walking
orthoses for paraplegics. In: Neural Prostheses, ed by Stein
R. Peckham H. Pooovic D. Oxford Press. New York. 1992.
20. Tanner J: Reversible blocking of nerve conduction by
alternating current excitation. Nature 195:712-713, 1%2.
21. Wednesky N: Wie rasch ermudet der Nerv. Zentralbl Med
Wiss 65-68, 1884.
22. Winter D: Biomechanics and motor control of human
movement. Wiley, New York, 1990.
23. Zhou B, Baratta R, Solomonow M: Manipulation of muscle
force with various firing rates and recruitment control
strategies. IEEE Trans Biomed Eng 34:128-139, 1987.