Exp Brain Res (2002) 143:8999

DOI 10.1007/s00221-001-0960-1

R E S E A R C H A RT I C L E

Daniel Lambertz Francis Goubel Chantal Prot

A method to evaluate reflex excitability of the human ankle

plantarflexors despite changes in maximal activation capacities

Received: 8 February 2001 / Accepted: 24 October 2001 / Published online: 16 January 2002
Springer-Verlag 2002

Abstract Stretch reflexes were evoked in the submaximally activated ankle extensors during sinusoidal length
perturbations. A mean stretch reflex (SR) amplitude

(SRA), i.e., SR area/SR duration was quantified for the

soleus muscle and for the gastrocnemii muscles. SRA

was also expressed in relative values (SRArel), i.e., SRA

was related to the corresponding background electromyogram (EMG). Sinusoidal length perturbations were applied in two ways: (1) at a fixed frequency (16 Hz) and
at different levels of voluntary contraction [10% to 70%
maximal voluntary contraction (MVC)] and (2) at a constant activation level (50% MVC) and at frequencies
ranging from 6 to 16 Hz. Then, new parameters are
proposed to characterize muscle reflex excitability. The
first parameter, SRindex, consisting in the slope of the

SRAEMG relationship, was considered to be more

representative of central influences on the reflex pathway. Secondly, the frequency distribution of the stretch

reflex was analyzed and the area under the SRArelfre

quency curve gave the second parameter (FDSRArel).
This parameter should account for both central and peripheral mechanisms on the reflex pathway. These parameters were found to be higher for the highly excitable
soleus than for the less excitable gastrocnemii muscles.

SRindex and FDSRArel can be proposed as a tool to analyze changes in reflex excitability, which can accompany
a process of neuromuscular plasticity. In order to validate the procedure, the proposed parameters were quantified in a case study before and after a period of plyometric training.
Keywords Reflex excitability Sinusoidal length
perturbation Stretch reflex Voluntary activation
Triceps surae
D. Lambertz F. Goubel C. Prot ()
Universit de Technologie de Compigne,
Dpartement de Gnie Biologique, CNRS UMR - 6600,
60205 Compigne cedex, France
e-mail: chantal.perot@utc.fr
Tel.: +33-3-44234392, Fax: +33-3-44204813

Introduction
Typically, stretch reflexes are obtained by imposing
joint perturbations either in force or in length while a
subject maintained a voluntary tonic activation. Reflex
responses of the triceps surae to imposed sinusoidal
force perturbations on the ankle joint were first described by Agarwal and Gottlieb (1977). Considering the
mechanical properties of the ankle joint a priori as a linear input-output system, imposed force perturbations induce changes in the resulting displacement. Thus, using
force perturbations, both variations in frequency and
in length can influence the stretch reflex amplitude
(Agarwal and Gottlieb 1980). Stretch reflexes in response to imposed perturbations in length have also been
studied using different methods: random perturbations
(Kearney and Hunter 1983), large and sudden ramp displacements (Kearney and Chan 1982; Toft et al. 1991;
Kirsch and Kearney 1993), or sinusoidal perturbations at
different frequencies (Rack et al. 1978; Neilson and
McCaughey 1981; Evans et al. 1983). So, during imposed sinusoidal length perturbation at a given level of
preactivation and at different frequencies, changes in reflex activities only depend on the variations in frequency, providing that muscle stretching is carefully controlled.
The aim of the cited stretch reflex studies was often
focused on the eventual mechanical importance of the
stretch reflex. In expressing joint stiffness as a force resisting an angular displacement, a large joint stiffness
provides good position control, notably during postural
tasks, which are generally non-stable, and favors the application of additional forces. With regard to the neurophysiological properties, the generation of the additional
forces depends on the reflex response. When mechanically induced, this reflex depends, firstly, on the elastic
properties of the musculo-tendinous structures in series
with the muscle spindles (Rack et al. 1983) and, secondly, on the supraspinal control (Sinkjaer 1997; Mirbagheri
et al. 2000). In turn, reflex excitability contributes to regulate muscle stiffness. Furthermore, both reflex excit-

90

ability and mechanical properties are affected by neuromuscular plasticity. Some evidence has been given that
training and disuse induce: (1) neural adaptation, including reflex excitability (Sale 1988; Duchateau 1995;
Enoka 1997), and (2) mechanical adaptation, including
the elastic properties (Cornu et al. 1997; Lambertz et al.
2001) of the neuromuscular system.
In the present study we propose a new characterization of the stretch reflex obtained during sinusoidal
length perturbations at different levels of muscle preactivation and at different frequencies. Such a new characterization offers the possibility to compare the reflex excitability of different muscles and to follow changes in
the reflex excitability of a given muscle during a period
of training or disuse, independently of the maximal activation capacities of the muscle, which are also altered by
such changes in the functional demand. The new parameters proposed in the present study were firstly quantified for a slow muscle (soleus), known to be highly excitable, and for less excitable muscles (gastrocnemii).
Secondly, a case study illustrated the physiological interest of the proposed parameters, quantified for the soleus
muscle before and after so-called plyometric training.
The intent of this type of training is to increase the instantaneous explosive power output as required during
short-term efforts such as sprinting or jumping. Then, the
proposed parameters were used to assess changes in reflex excitability due to such a physical conditioning.

Materials and methods

Apparatus and subjects
The ankle ergometer used for this study has been described in detail by Tognella et al. (1997), but will be reviewed briefly here.
This ergometer consisted of two main units: (1) a power unit
which contained the actuator, its power supply unit, position and
torque transducers, and its associated electronics, and (2) a driving
unit composed of a 486 PC-type computer equipped with a specific 12-bits A/D converter and a timer board. Angular displacement
was measured with an optical digital sensor and angular velocity
was captured from a resolver bound to the rotor. Angular torque
was obtained using a strain-gauge torque transducer. Specific
menu-driven software controlled all procedures and recorded mechanical variables and electromyograms (EMGs; 1 kHz sampling
frequency) for later analysis. A dual-beam oscilloscope gave the
subject visual feedback about the procedure in progress.
The main experiment was done on nine college students (seven
men and two women: 203 years, 1764 cm, 625 kg). All of
them gave their informed consent. Each subject was familiarized
with the experiment during a preliminary session, 1 week before
starting the tests.
The subject was comfortably seated without back support in
order to limit the contribution of the trunk muscles to the requested effort. The left foot was attached rigidly to an adjustable foothold, so that the horizontal bi-malleolar axis coincided with the
axis of rotation of the actuator of the ergometer. The knee was extended to 110 and the ankle was placed to 90, i.e., reference position. The thigh was maintained by a restraint system to keep it
immobilized during the efforts.
Surface EMGs were detected on each part of the triceps surae,
i.e., the soleus (Sol), the gastrocnemius lateralis (GL), and medialis (GM) muscles, using standard Ag/AgCl surface electrodes. The
electrodes were placed over the belly of each gastrocnemius mus-

cle and, for the Sol, 2 cm below the insertion of the gastrocnemius
muscles on the Achilles tendon. EMGs were recorded differentially, amplified, and bandpass filtered (11,000 Hz).
Furthermore, an additional experiment was done on a college
student who volunteered for this part of the present study. The aim
of this case study was to validate the proposed parameters as a
tool to analyze changes in reflex excitability before and after
8 weeks of plyometric training.
Experimental protocol
The torque from a maximal voluntary contraction (MVC) was determined in plantarflexion under isometric conditions while the
subject was asked to develop maximal contraction against the
actuator. The true MVC level of the day, corresponding to the best
torque, was determined by choosing the greatest of three attempts
to generate the maximal voluntary effort. Each of these attempts
lasted 4 s and were 2 min apart in time. Then, two experimental
paradigms were applied.
Firstly, stretch reflex activities at different contraction levels
were studied at a single frequency. Therefore, sinusoidal length
perturbations were imposed on the ankle joint at a constant frequency of 16 Hz while the subject maintained a submaximal plantarflexion torque, varying from 10% to 70% of MVC in steps of
10%. The displacement amplitude was fixed at 3 peak-to-peak.
This non-physiological high frequency was chosen because at this
frequency the musculo-articular system works under inertial conditions, which means that the dynamic stiffness of the musculoarticular system is constant whatever the level of contraction and
so, does not interact with the stretch reflex response (Matthews
1994).
Secondly, stretch reflex activities were characterized over a
range of frequency and a single level of contraction. Therefore,
sinusoidal length perturbations of 3 peak-to-peak were applied at
different frequencies ranging from 6 to 16 Hz in steps of 1 Hz.
During this part of the experiment the subject maintained the plantarflexion torque at 50% of MVC.
To perform these tasks the subject was instructed to hold a
constant level of force production by matching an oscilloscope
trace despite the oscillations in the torque signal. To facilitate this
task, the torque traces used as visual feedback consisted in a lowpass filtering of the output torque. Thus, all subjects were able to
keep their level of force production at the target value without any
difficulty, whatever the imposed frequency. The use of a frequency
range from 6 to 16 Hz should also eliminate the possibility for the
subject to resist the stretch voluntarily or to do tracking movements, since the high frequency tracking range was seldom above
7 Hz (Bennett 1994; Cathers et al. 1996). Data sampling started
when the subject reached the target value and sinusoidal perturbations were applied 1 s later. The sinusoidal oscillations lasted 4 s.
The subject relaxed as soon as the perturbation stopped and resting periods of at least 1 min were observed between the different
perturbation sequences to prevent muscle fatigue.
As for the additional experiment, sinusoidal perturbations were
imposed according to the two experimental paradigms before and
after the period of training. In each experimental situation, the
subject maintained contraction levels evaluated by considering the
actual MVC of the day. Furthermore, supramaximal electrical
stimulations were achieved in order to get the maximal motor
direct response (Mmax) of the Sol. The mean amplitude of the Sol
), expressed by the ratio between duration and area,
Mmax (M
max
was used to normalize stretch reflex activities and, thus, to take into account the different conditions of skin and surface electrodes
impedance in signals recorded on different days.
Data processing and analysis
Background activity (BGR) was defined as mean amplitude of the
rectified EMG recorded during the first second of sampling, i.e.,
before the sinusoidal perturbations were imposed. Then, as pro-

91
Fig. 1 A Raw data set of electromyogram (EMG) responses
to sinusoidal oscillations (arrow
indicates the beginning of the
lengthening phase). B, C Effect
of data processing on the stretch
reflex evoked during sinusoidal
perturbations, for the soleus
(Sol, thin lines) and the gastrocnemii (Gas, thick lines). In
B, background activities (BGR)
are degraded and the stretch reflex consisted essentially in the
more time-locked short latency
component. In C, rectification
is done prior to averaging.
Consequently, BGR activity remains present and the stretch
reflex appears as the short and
medium latency components.
Data obtained at 9 Hz and for
50% of maximal voluntary contraction (MVC)

posed by Evarts and Vaughn (1978), averaging prior to rectification was chosen to favor the quantification of time-locked or periodic EMG signals mixed with non-correlated EMG background
activity, such as the early component of a stretch reflex. The efficiency of this data processing is illustrated in Fig. 1, showing that
averaging of unrectified data favored the quantification of the
short latency response, whereas the later, less synchronized component was degraded by such a procedure.
Therefore, stretch reflex responses collected during sinusoidal
perturbations were analyzed as follows. Data were first inspected
visually to keep a number of successive periods where the required voluntary torque remained almost constant despite the imposed perturbations. After this windowing, the EMGs of Sol, GL,
and GM muscles were averaged over the remaining periods and
then rectified. GM and GL were summed up to express the gastrocnemii (Gas) activity. Since the EMGs of Sol, GM, and GL
were first averaged and then rectified, no correction of BGR was
necessary at this stage (see Fig. 1). However, this method can only
be used after the experimenter has carefully controlled that BGR
does not change notably during the muscle stretch.
Stretch reflex (SR) areas were determined by integrating the
averaged EMG over the time interval corresponding to the expected burst of reflex activity, i.e., as reported for the plantarflexors
inside a time interval which started and ended around 35 and
60 ms, respectively. In the present study, onset and offset of the
EMG burst were pointed out by the experimenter to compute latency and duration of the stretch reflex, respectively. In doing so,
the prominent peak of the reflex burst corresponded to the SR amplitude. Latency was always referenced
at the onset of the length
ening phase. A mean SR amplitude (SRA value was also calculat-

ed as the ratio between SR area and SR duration in order to consider possible influences of reflex duration. This was justified
since the literature indicates differences in reflex duration according to stretch modalities (Gottlieb et al. 1995; Morita et al. 1998).
Firstly,
the accuracy of the SR parameters (amplitude, area,
and SRA) was attested studying
the relationships between SR amplitude and, SR area and SRA, respectively. Then, the comparison
between both correlation coefficients
r should reflect possible effects of SR duration on SR area and SRA.
Secondly, stretch reflex activities were studied with regard
to

the background
level,
BGR.
For
this
purpose
Sol
and
Gas
SRA

and SRArel (i.e., SRA expressed with respect to the corresponding

BGR) were adjusted to normalized
BGR (BGR/BGR
max) in order

to establish normalized SRABGR and SRArel BGR relationships. BGRmax was achieved from the best MVC measurement of
the day.

Thirdly, SRA and SRArel frequency relationships were established. In the case of a linear regression between these parameters, the slope of the relationships can be calculated. However, this
method required that the correlation coefficient r matched a critical value. Thus, in order to avoid a dependency on the correlation
coefficient r, another parameter was proposed to characterize the
distribution of the stretch reflex over the different applied
frequencies.
This
parameter
represents
the
area
under
the
SRA
or

SRArel frequency curve and is

defined as the
frequency distribution of the stretch
reflex (FDSRA
rel or FDSRArel , respectively). The FDSRA and FDSRArel were obtained by numerical integration using the trapezium method.
With regard to the additional experiment, data processing was
conducted on Sol. Firstly, SR duration was quantified using the

92
data set of the first experimental paradigm before and after the
training period. Secondly,
according
to the first experimental
paradigm, normalized SRA and SRArel values were calculated

and
related to BGR/BGRmax. Thirdly, normalized FDSRA and
FDSRArel were quantified according to the second experimental
paradigm.
Statistics
A Students one-sample t-test was used to verify that mean values
were significant for the population. Based on these findings, a multiple regression analysis was used to investigate
the relations
among SR amplitude, SR duration, SR area, and SRA. Therefore, a
partial correlation procedure was applied which consisted in computing partial correlation coefficients rp which describe the linear
relationship between two variables while controlling for the effects
of one or more additional variables. Differences in Sol and Gas reflex activities for the two experimental paradigms were tested by
using Students t-test for independent samples. Data are presented
as mean SEM. Statistical significance was accepted at P<0.05.

Results
Pattern of the stretch reflex
Influence of BGR
In response to a sinusoidal length perturbation of low
amplitude and high frequency (16 Hz) the Sol and the
Gas muscles exhibited a rhythmic burst of activity at
each lengthening phase of the sinusoidal perturbation.
The stretch reflex consisted of a single burst of activity.
Its mean latency was 35.40.5 and 33.10.6 ms for Sol
and Gas, respectively. Mean SR duration for Sol did not
change significantly with the target torque (16.81.6 and
15.91.4 ms from the lowest to the highest contraction
level, respectively). In contrast, mean SR duration for
Gas changed with the level of BGR from 18.62.6 ms at
the lowest contraction to 14.41.3 ms at the highest contraction. For each level of contraction (10% to 70%

MVC) mean values of SR amplitude, SR area, and SRA

were calculated. Mean values were found to be significant for the population (P<0.05) as attested by Students
one-sample t-test. SR amplitude and SR area, as well as

SR amplitude and SRA, were highly correlated as attested by the linear relationships established between these
parameters (Fig. 2A, B). A possible influence of SR duration was statistically analyzed using the partial correlation coefficients method. For this purpose, SR amplitude
was used as control variable. Then, the correlation matrix was calculated for each muscle. As for the Sol, no
significant correlations were found between SR duration

and SR area or SR duration and SRA, whereas SR area

and SRA were significantly correlated (see Table 1). In

regard to Gas, Table 1 shows that SR duration was

highly correlated to SR area while SR duration and SRA,

and SR area and SRA showed no significant correlation.

Thus, in considering SRA values, SR duration plays only

a minor role in changes in reflex activities with the level
of BGR.

Fig. 2A, B Relationships between

mean stretch reflex (SR) parameters (amplitude, area and SRA) calculated for the population
(n=9). Sinusoidal perturbations were applied at different levels of
MVC (10% to 70%)
and at a constant frequency (16 Hz). A Relationships between SRA (circles) and SR area (squares), and SR
amplitude, respectively, for the soleus (Sol). In B, the same relationships are presented for the gastrocnemii (Gas). No improve
ments in the correlation coefficients r are found when using SRA
or SR area for Sol (A; r=0.97 vs r=0.97) and for Gas (B; r=0.95 vs
r=0.96). For clarity, no SEMs are shown
Table 1 Partial correlation coefficients
matrix between stretch reflex (SR) area, SR duration, and SRA for the soleus (Sol) and the
gastrocnemii (Gas) at different levels of maximal voluntary contraction (MVC; 10% to 70%) and at a constant frequency (16 Hz).
This matrix reveals that SR duration of Gas influences reflex activities when SR area is used

SR area

SRA
a Denotes

Sol

SRA
rp=0.792
P<0.05a

SR duration

Gas

SRA

SR duration

rp=0.604
P>0.05
rp=0.103
P>0.05

rp=0.610
P>0.05

rp=0.797
P<0.05a
rp=0.044
P>0.05

that the correlation coefficient rp is highly significant

Influence of frequency
Stretch reflex activities, evoked when maintaining 50% of
MVC, were always observed at frequencies higher than
9 Hz. For some subjects, no reflex activities were detect-

93
Table 2 Partial correlation
coefficients matrix between SR area,

SR duration, and SRA for the soleus (Sol) and the gastrocnemii
(Gas) at different frequencies (616 Hz) and at a constant level of
MVC (50%). This matrix reveals that SR duration of Gas influences reflex activities when SR area is used

SR area

SRA

Sol

SRA
rp=0.229
P>0.05

SR duration

Gas

SRA

SR duration

rp=0.557
P>0.05
rp=0.481
P>0.05

rp=0.286
P>0.05

rp=0.670
P<0.05
rp=0.418
P>0.05

was used as the control variable. Then, the correlation matrix revealed no significant correlation between SR area

and SRA neither for Sol nor for Gas. On the other hand,
SR duration and SR area were correlated for Gas but not
for Sol (see Table 2), indicating that Sol reflex activities
seemed to be independent of duration when the perturba
tion frequency changed. Thus, SRA will allow the comparison between reflexes which are different in duration.
In the subsequent sections, the stretch reflex will be ana
lyzed considering only SRA values.

SRA related to BGR

Fig. 3A, B Relationships

between mean SR parameters (ampli
tude, area and SRA) calculated for the population (n=9). Sinusoidal perturbations were applied at different frequencies (616 Hz)
and
at a constant level of MVC (50%). A Relationships between
SRA (circles) and SR area (squares), and SR amplitude, respectively, for the soleus (Sol). In B, the same relationships are presented for the gastrocnemii
(Gas). Correlation coefficient r is im
proved when using SRA instead of SR area for Gas (B; r=0.96 vs
r=0.78), whereas Sol (A; r=0.99 vs r=0.97) seems to be unaffected. For clarity, no SEMs are shown

able at the lowest frequencies (69 Hz). The latency for

Sol and Gas did not change significantly with frequency
and a mean value for Sol was 35.40.6 ms and for Gas
was 33.60.6 ms. Mean SR duration varied with the frequency from 20.11.7 ms at 6 Hz to 18.31.4 ms at 16 Hz
for Sol and from 19.81.5 ms at 6 Hz to 17.12.1 ms at
16 Hz for Gas. For each frequency, mean values of SR

amplitude, SR area, and SRA were calculated. Mean values were found to be significant for the population
(P<0.05) as attested by Students one-sample t-test. Then,
the influence of SR duration was estimated by the analysis

of the SR amplitude and, SR area and SRA relationships,

respectively. Sol SR amplitude and, SR area and SRA

showed good relationships (Fig. 3A). In contrast, Gas SR
amplitude and SR area were less correlated than SR am
plitude and SRA, as illustrated in Fig. 3B. This influence
of SR duration was statistically attested by using the partial correlation method. For this purpose, SR amplitude

TorqueEMG relationships were established for Sol and

Gas. Both relationships were found to be linear (r>0.98;
n=7) and their mean slopes were 0.630.07 and
0.300.04 Nm V1 for Sol and Gas, respectively.

The SRABGR relationships slightly differed for Sol

and Gas muscles. Firstly, as shown in Fig. 4A, Sol activation at 70% of MVC was already at about 90% of its
BGRmax, whereas Gas were only about 80% of BGRmax.

Secondly, Sol SRA increased sharply up to about 70%

of Sol BGRmax and then maintained a plateau value

where-as Gas SRA increased over the full range of the

tested BGR.
Then, Sol and Gas SR were related to their respective

levels of BGR. Expressing SRA in relative values,

should lead to linear SRA rel BGR relationships for both

Sol and Gas (see Fig. 4B). The negative slope of the relationships was significantly steeper for Sol than for
Gas as attested by Students t-test for independent samples (1.390.11 vs 0.850.06 for Sol and Gas, respectively).

SRA related to frequency

Firstly, the frequency dependency of Sol and Gas SRA

was studied at a given level of torque (50% of MVC)
over a frequency range from 6 to 16 Hz. Fitting was
satisfactory when using a linear regression (r=0.97 for
Sol and r=0.97 for Gas, P<0.05) for each muscle
(Fig. 5A). However, analyses of individual data showed

94

Fig. 4 Relationships between SRA (A) and SRArel (B), and the
normalized background (BGR/BGRmax), respectively, for the soleus (Sol; circles) and the gastrocnemii (Gas; squares). BGR (10%
to 70% of MVC) was quantified during the first second preceding
the perturbations (16 Hz). In B, the slopes of the linear relationships defined SRindex. For Sol, SRindexis 1.39 and for Gas, SRindex
is 0.85. Data are represented as mean SEM, which are representative for the population at P<0.05

that some subjects exhibited SRAfrequency relationships with a correlation coefficient which did not
match the critical value for a linear regression analy
sis (P>0.05). In such a case, the link between SRA
and frequency was approached by considering the area

under the SRAfrequency curve, defined above as

FDSRA (see Methods). Mean FDSRA for Sol was

1
1.350.32 mV s and mean FDSRA for Gas was

1.240.23 mV s1. Obviously, the slope or the FDSRA

should lead to similar results when a good linear regres
sion exists between FDSRA and frequency. This can
be easily verified by using the parameters from the
linear regression analysis (see legend Fig. 5). Similarly,

FDSRArel values were calculated and were found to be

significantly higher for Sol (15.602.96 s1) than for Gas
(9.471.29 s1) (see Fig. 5B) as attested by Students
t-test for independent samples.

Fig. 5 Relationships between SRA (A) and SRArel (B), and frequency, respectively, for the soleus (Sol; at the back) and the gastrocnemii (Gas; in front). BGR (50% of MVC) is quantified during

the first second preceding

the perturbations (616 Hz). FDSRA

(A) and FDSRArel (B) are expressed by the filled

areas. In both cases (A, B) fitting was satisfied when using a linear regression. The slopes in A are 12.86 V s (Sol) and 7.52
V s
(Gas) and in B are 0.16 s (Sol) and 0.07 s (Gas). Thus, FDSRA
can also be calculated by integration over the frequency using the
slope values (note, that intercept point values should be added for
integration). Data are represented as mean SEM, which are representative for the population at P<0.05

Although this part of the study considered only one

level of torque, the present results gave further informa
tion about the influence of BGR on the SRA frequency
dependency, since BGR, useful to maintain 50% of
MVC, varied notably from one subject to another. There

fore, a relationship between FDSRA (or FDSRArel)

and BGR/BGRmax can be established for the population.

Figure 6 shows FDSRA (Fig. 6A) and FDSRArel

(Fig. 6B) values of the nine subjects, who were asked to

maintain 50% of MVC. As one can see, FDSRA

showed no correlation with BGR/BGRmax due to scat
tered individual data, whereas FDSRArel tended to decrease linearly when normalized BGR increased. Thus,

FDSRArel has to be considered when studying the frequency dependency of the stretch reflex at different levels of BGR.

95

Fig. 6 Individual FDSRA (A) and FDSRArel (B) values of the

nine subjects as a function of normalized BGR activity. In A, scattering of data is noticeable. In B, the fit is improved (r=0.78;
P<0.05) and the slope of this relationship defined FD-SRindex. Data are obtained from the gastrocnemii while the subjects maintained 50% of MVC

Additional experiment

Fig. 7 Relationships between normalized SRA (A) and SRArel

(B), and the normalized background (BGR/BGRmax), respectively,
before (open circles) and after (filled circles) plyometric training.
In A, scattering of data after plyometric training is noticeable and
a direct evaluation of the training effect seemed to be difficult,
since BGR/BGRmax changed with training. In B, significant linear
relationships were found before training (r=0.81;
P<0.05) and after training (r=0.86; P<0.05) when using SRArel data. Thus, the
slopes can be used to quantify changes in reflex excitability, independently of changes in BGR/BGRmax

Firstly, SR duration was quantified at 16 Hz and at the

different levels of activation (1070% MVC) before and
after the training period. As a common finding, SR duration did not vary markedly with the level of activation
before and after the training period. Thus, the different
values of SR duration were averaged over the different
levels of activation. By this procedure, a decrease in SR
duration of about 5 ms was observed for this subject
after the training (19.301.36 ms before training to
14.101.94 ms after training). It can be argued, that this
decrease in SR duration, even at this high frequency, influences the analysis of the SR, and thus, validates the
use of mean amplitude reflex values as done in the following section.
Secondly, according to the first experimental para
digm, Fig. 7 shows the effect of training on SRA and

SRArel. As one can see, SRA after training varied markedly over the range of activation (see Fig. 7A), whereas

SRArel values led to linear relationships before and after

the training period (see Fig. 7B). Then, changes in SR

excitability due to training were assessed using the negative slope which, in turn, is also independent of changes
in BGR/BGRmax. The negative slope was found to be
steeper after training than before training (12.17 after
training vs 4.04 before training).
Thirdly, according to the second experimental para

digm, FDSRA and FDSRArel were quantified using

SRA and SRArel frequency relationships, respectively.

Figure 8 depicts the results of the SRA and SRArel

frequency relationships before and after the training peri

od. A good linear increase in SRA as well as in SRArel

over the frequencies was observed before training. After

training, SRA and SRArel increased linearly only over

the lower frequency range (69 Hz) and then, no further
changes occurred at higher frequencies (1016 Hz). Nor
malized FDSRA (1.25 s1 before training vs 1.27 s1 af
ter training) as well as FDSRArel (30.0 s1 before training vs 29.5 s1 after training) were found to be unchanged after the period of training.

96

velocity (Kearney and Chan 1982; Toft et al. 1991;

Fellows et al. 1993) or sinusoidal torque perturbation at
relatively low frequency (Agarwal and Gottlieb 1977,
1980). Thus, it can be assumed, with the authors cited
above, that the reflex responses of the ankle joint to sinusoidal length perturbations involve the short latency
component of a stretch reflex mediated via the monosynaptic pathway.
Pattern of the stretch reflex

Fig. 8 Relationships between normalized SRA (A) and SRArel

(B), and frequency, respectively, before (open circles) and after
(filled circles) plyometric training. A loss in the linear frequency
behavior of the stretch reflex after the training period at the higher
range of frequencies (1016 Hz) is noticeable

Discussion
In the present study a detailed electromyographic analysis of the stretch reflex led to introduce new parameters,
which can be useful for studying changes in muscle reflex excitability when changes in functional demands
(for example, a period of training or disuse), and thus in
motor drive, occur. Stretch reflex activities can be
evoked in different ways, such as sinusoidal perturbations. Knowing that muscle spindles are also sensitive to
slow and slight perturbations, sinusoids were often chosen because they involve no sudden impulsive movements that might stimulate, in a synchronized manner, a
large number of sensory receptors in an artificial and
non-physiological way (Rack et al. 1978; Evans et al.
1983).
The reflex activities described in the current experiments were evoked by a controlled sinusoidal length perturbation of low amplitude and at different frequencies.
These activities can be considered as a stretch reflex,
mediated via a spinal reflex pathway, since their latencies were within the range of the response to an Achilles
tendon tap, i.e., somewhat shorter than those reported in
other studies when using large ramp displacement at low

The data processing chosen (averaging prior to rectification) favored the analysis of the short latency component
of the stretch reflex (Evarts and Vaughn 1978; see also
Fig. 1). If a reflex study should also include the quantification of longer latency components of the stretch reflex,
EMGs should be rectified prior to averaging, and thus,
further BGR activity correction should be taken into account.
The SR duration of Gas was influenced by both the
level of muscle activation and the frequency, whereas
Sol SR duration was never modified by one of these parameters. The lower Gas SR duration at high levels of
muscle activation or at high frequency, accompanied by

an increase in SRA values (see Fig. 2), could represent a

higher synchronization of the waves composing the
stretch reflex of this heterogeneous muscle. This modification in SR duration could also be due to the fact that at
low frequencies and at low levels of tonic activity, muscle spindles would be activated in a more dispersed order. This effect of duration was obviously more effective
when considering SR area values but could be mini
mized when SRA values were analyzed (see Table 2).

Thus, further analysis of SRA values partly overcomes

the influence of SR duration. From this mathemati
cal/statistical analysis, SRA values were found to characterize best the stretch reflex activity since this parameter takes into account SR amplitude over a given duration. From a physiological point of view, Evans et al.
(1983) also reported that the use of the amplitude of the
EMG burst gave only limited information about the intensity of reflex activity in the plantarflexor muscles.
This was explained by the complex structure of the muscles in which large numbers of motor units were activated by the same reflex mechanism, with variable synchrony.
The case study indicated a decrease in SR duration after the training period. It can be proposed that the Sol became a more heterogeneous muscle after a period of
plyometric training due to a relative increase in the fast
type muscle fiber content (Almeida-Silveira et al. 1996).
Then, SR duration should increase rather than decrease.
However, as suggested by Cornu et al. (1997) plyometric
training should increase the tendon stiffness. This could
lead to a more equal distribution of the external imposed
length perturbation up to the muscle spindles and, thus, to
a better synchronization of the waves composing the
stretch reflex. Furthermore, it can be assumed that plyo-

97

metric training could induce changes in the solicitation of

afferent fibers, which should lead to a facilitation of the
reflex response (see Sale 1988) and, thus, a decrease in
reflex duration. Whatever the case, it is not the purpose of
this methodological paper to discuss in detail the effects
of the training, but this case study has shown that SR duration could be altered due to changes in the appropriated
physiological parameters when changes in the functional
demand occur. Thus, in order to find a midway solution

to analyze stretch reflex activities, SRA values should facilitate the comparison of reflex excitability, which can
be different in duration or recorded in different muscles.

SRA related to BGR

The experiments conducted at 16 Hz and at different

levels of MVC have shown a relation between SRA and

the prerequisite level of voluntary contraction. This influence, clearly observable for each subject, can also be
attested for the group of subjects provided that normalized data are considered. To illustrate the effect of the
level of contraction, the stretch reflex was often related
to external torque (Kearney and Chan 1982; Toft et al.
1991). However, assuming that a linear relationship exists between torque and EMG, as verified in the present
study, it seemed to be more reasonable to relate reflex
activities to the corresponding BGR activity rather than
to an external torque developed by a whole muscle
group. Therefore, the stretch reflex was analyzed with
respect to the pre-existing EMG as in previous studies
(Bedingham and Tatton 1984; Bloem et al. 1993). Hence,
when using normalized BGR data (BGR/BGRmax), it became possible to compare the reflex excitability of different muscles. For example, in the current study the difference in reflex excitability of Sol and Gas was approached by considering their respective BGR, which
differ when the subject maintained a given external
torque (see Fig. 4). This representation also gave the opportunity to reinvestigate the validity of the automatic
gain principle at the level of the human ankle extensors.
This automatic gain control, first proposed by Mardsen
et al. (1976) and later detailed by Matthews (1986) in
terms of frequency modulation and recruitment of motor
units, means that the size of the stretch reflex is directly
related to the amount of pre-existing muscle contraction.
In his study, Matthews elaborated the automatic gain
control on the elbow flexor muscles using square-wave
modulated vibration (100 Hz modulated at 5 Hz) and
tendon jerks at contraction levels below half-maximum.
The extent of the automatic gain compensation to a variety of muscles over a wider range of background activation could be then useful, since similarities were found
even earlier by other workers in related situations. Previously, Toft et al. (1991) failed to report the automatic
gain principle for the Sol, by using ramp displacement,
and proposed this principle as non-valid for the human
ankle extensors. The failure for these authors to confirm
the automatic gain principle could come from the fact

that they analyzed non-normalized reflexes obtained in

different subjects and that they related the Sol reflexes to
the plantarflexion torque developed by the muscle group.
The differences between ankle torque and EMG as a
measure of muscle tonic activity were also stated by
Kearney and Hunter (1983). These authors showed, that
reflex responses were more closely correlated with the
level of EMG activity than with ankle torque. Bloem et
al. (1993) also found that gain control was absent for the
GM but in their study the range of muscle contraction
was limited to that necessary to compensate gravity.
The present study extended this question by analyzing
independently Sol and Gas reflex activities over a large
range of muscle contractions. When considering, firstly,

SRA, it appeared that the automatic gain principle re

mained valid for Sol SRA up to about 70% of BGRmax

whereas Gas SRA increased over the full range of BGR.

These results could reflect the muscle differences in
terms of motor unit types. As shown by Johnson et al.
(1973) the Sol is a more homogeneous muscle composed
of a large (more than 80% in humans) number of slow,
low threshold motor units whereas Gas is a more heterogeneous muscle (around 50% of slow motor units). Thus,
it could be proposed that in a homogeneous muscle like
the Sol, the automatic gain control is available up to a
given BGR value. On the other hand, in the more heterogeneous Gas the recruitment phenomenon could be suspected to be present on a larger range of muscle activation. In accordance with this proposition, for such a muscle group, the automatic gain control was found to be
present over almost the full range of contraction. So,
when considering muscle activation rather than torque
values, it became possible to validate the automatic gain
control, up to the level where a further increase in activation could be exclusively due to the frequency modulation of motor units. Evidence of neural adaptation due to
a period of training has been mentioned by Sale (1988),
indicating that the expression of voluntary strength may
be linked to a skilled act, in which the nervous system
allowed a trained individual to fully activate prime
movers in specific movements and to better coordinate
the activation of all relevant muscles. Then, without automatic gain compensation, regulatory reflexes, whether
excitatory or inhibitory, would tend to be too mechanically powerful when pre-existing muscle contraction was
weak, and too feeble when muscle contraction was
strong (Matthews 1986).

Interestingly, when considering SRArel values, i.e.,

SRA divided by BGR, a significant negative relation was

obtained over the range of BGR for both muscles. From

a mathematical point of view, the use of relative SRA

values led to a linear relationship, while the slope
value should be independent of BGRmax. As previously
claimed by Bloem et al. (1993), this correction of stretch
reflex values by the ratio method is very effective when
a large range of muscle activation is tested. The slope
value was higher for the Sol than for Gas. This mode of
representation leads us to propose the negative slope of

the SRArelBGR relationship as an index of reflex excit-

98

ability (SRindex) for a given muscle (Fig. 4B). Using

SRindex, it will become possible to attest changes in muscle reflex excitability even if the muscle ability to be fully activated differs from one test to another, due to eventual neural adaptations like motor drive.
The usefulness of the SRindex has been shown with regard to the additional experiment. In this case study, the
SRindex, used to quantify changes in Sol reflex excitability over a large range of activation, was found to increase
after the period of training. It has been mentioned earlier
that mechanically induced stretch reflex activities depend on the supraspinal control as well as the peripheral
components of the reflex pathway such as tendon stiffness and/or spindle sensitivity. However, since SR activities were elicited at a frequency where the musculo-articular system works under inertial conditions, its mechanical response (i.e., the dynamic stiffness) does not interact
with the reflex response. In these conditions, a possible
influence of the mechanical properties of the structures
in series with the muscle spindles should be restricted to
the transmission characteristics of the imposed length
perturbation. Thus, the increase in SRindex might reflect
essentially a process of central adaptation on the reflex
pathway. One possibility, also proposed by Voigt et al.
(1998) to explain an increase in electrically evoked H reflexes during hopping, could be a decrease in the normal
level of presynaptic inhibition, which should lead to a facilitation of the reflex response (see Sale 1988).

SRA related to frequency

The results obtained for the population at 50% of MVC
and at different frequencies often led to a linear relation
ship between SRA and frequency ranging from 6 to
16 Hz. The slope of this relationship can then be used to
characterize the frequency dependency of reflex activities. Such a linear sensitivity of the stretch reflex to the
dynamics of the perturbation was already established using ramp perturbations (Gottlieb and Agarwal 1979).

However, for some of our subjects the individual SRA

frequency relationship failed to be linear. It can be argued that, at high frequency, antagonist muscles, such as
the tibialis anterior, were activated. Thus, antagonist activation can contaminate the agonist EMG by a crosstalk
phenomenon, leading to non-linearity. However, data
from preliminary experiments (unpublished observations) showed no tibialis anterior EMG activity, and thus,
other factors should be responsible for non-linearity.
Whatever the case, when individual data have to be
analyzed, the link between the stretch reflex and fre
quency can be quantified using the area under the SRA

frequency curve. This area, termed the FDSRA, can be

used to follow the reflex excitability of a muscle during
a period of hyper- or hypoactivity, provided that stretch
reflex activities were previously normalized (for example, using the M-wave). A further advantage for using

the FDSRA parameter instead of the slope consists in

the ability to assess changes in reflex excitability when

no changes in the slope occurs (i.e., in the case of parallel slopes). If these changes in functional demand are accompanied, as is often the case, with changes in submaximal muscle activation, it will be recommended to express data in relative values. In the present study the

FDSRArel was found to be higher for the slow Sol than

for Gas. Thus, FDSRArel can also be proposed as a

characteristic parameter of muscle reflex excitability.
The additional experiment illustrated the usefulness
of the FD parameter to study changes in reflex excitability, since this subject showed a partial loss of the frequency dependency of the SR after the training period. In this

case study, we reported that normalized FDSRA as well

as FDSRArel were unchanged. This result is in line with

findings from other workers who showed that mechanically induced reflex activities did not change after a period of plyometric training (Hkkinen and Komi 1986;
Almeida-Silveira et al. 1996; Voigt et al. 1998). Interest
ingly, the absence of changes in FDSRArel with training
should be then attributed to the above-mentioned differences in the frequency behavior of the SR before and af
ter training (see Fig. 8B). When analyzing FDSRArel
one should also account for changes in the muscle mechanical properties. Some results in the literature report a
decrease in stiffness of the active musculo-articular
system as a result of plyometric training (Cornu et al.

1997). Consequently, the unchanged FDSRArel should

reflect a balance between: (1) decrease in muscle spindle
solicitation (due to a lower stiffness) and (2) increase in
central excitatory influences (due to the above-proposed
removal of presynaptic inhibition).
Finally, considering the differences in BGR developed by the different subjects when maintaining 50%

of MVC, it appears that FDSRArel, as previously

FDSRArel (see Fig. 4B), decreased when the level of

BGR increased (see Fig. 6B). Then, the slope of the

FDSRArelBGR/BGRmax relationship can also be proposed as a new index (FDSRindex) to describe changes
in reflex excitability due to changes in functional demand when further individual levels of BGR are used.
FDSRindex should emphasize the evaluation of reflex
excitability even when changes in maximal motor drive
occur and can be proposed to follow, for example, the
time-course of rehabilitation.
In conclusion, the present study showed a simple description of the human stretch reflex evoked by sinusoidal
length perturbations, which represented a good tool to
test muscle reflex excitability. Sinusoidal length perturbations applied at different levels of BGR and at different
frequencies led us to propose new parameters to characterize reflex excitability, notably the frequency distribu
tion of the stretch reflex (FDSRArel). These parameters
could be useful to follow the time-course of the reflex excitability of muscles as for example, during rehabilitation
or training. Additionally, sinusoidal perturbations can
also be used to study mechanical properties of muscle. In
this way, sinusoidal perturbations can be used to study
both mechanical and neurophysiological parameters. It
appears then that a combined study of these parameters

99

would provide a better understanding of the consequences of, for example, a specific training program or a microgravity environment (Lambertz et al. 2001) in terms of
postural and movement control (Mirbagheri et al. 2000).
Finally, two indexes were proposed in order to characterize reflex excitability independently of changes in maximal activation. The first one, SRindex, gives the reflex excitability at a high frequency where the mechanical response of the ankle joint system does not interact with the
reflex response. On the other hand, the second index, FDSRindex, takes into account the frequency response, and
thus, the intrinsic mechanical properties of the neuromuscular system including muscle spindles. These indexes
should then facilitate the comparison of reflex excitability
between different muscles and different subjects. Moreover, the proposed parameters can be used when other
types of controlled perturbations are used (for example,
ramp displacements at different angular velocities). However, in the view of an SRindex evaluation, the perturbation
velocity should be high enough to ensure that the ankle
joint system works under inertial conditions.
Acknowledgements This study was supported by grants from the
Centre National dEtudes Spatiales (CNES) and the Ple GBM
Prinatalit-Enfance de la Rgion Picardie.

References
Agarwal GC, Gottlieb GL (1977) Oscillation of the human ankle
joint in response to applied sinusoidal torque on the foot.
J Physiol 268:151176
Agarwal GC, Gottlieb GL (1980) Effect of vibration on the ankle
stretch reflex in man. Electroencephalogr Clin Neurophysiol
49:8192
Almeida-Silveira MI, Prot C, Goubel F (1996) Neuromuscular
adaptations in rats trained by muscle stretch-shortening. Eur J
Appl Physiol 72:261266
Bedingham W, Tatton WG (1984) Dependence of EMG responses
evoked by imposed wrist displacements on pre-existing activity in the stretched muscles. Can J Neurol Sci 11:272280
Bennett DJ (1994) Stretch reflex responses in the human elbow
joint during a voluntary movement. J Physiol 474:339351
Bloem BR, Dijk JG van, Beckley DJ, Zwinderman AH, Remler
MP, Roos AC (1993) Correction for the influence of
background muscle activity on the stretch reflex amplitudes.
J Neurosci Methods 46:167174
Cathers I, ODwyer N, Neilson P (1996) Tracking performance
with sinusoidal and irregular targets under different conditions
of peripheral feedback. Exp Brain Res 111:437446
Cornu C, Almeida-Silveira MI, Goubel F (1997) Influence of
plyometric training on the mechanical impedance of the human ankle joint. Eur J Appl Physiol 76:282288
Duchateau J (1995) Bed rest induces neural and contractile adaptations in triceps surae. Med Sci Sports Exerc 27:15811589
Enoka RM (1997) Neural adaptations with chronic physical activity. J Biomech 30:447455
Evans CM, Fellows SJ, Rack PMH, Ross HF, Walters DKW
(1983) Response of the normal human ankle joint to imposed
sinusoidal movements. J Physiol 344:483502
Evarts EV, Vaughn W (1978) Intended arm movements in response to externally produced arm displacements in man. In:
Desmedt JE (ed) Cerebral motor control in man: long loop
mechanisms. Progress in clinical neurophysiology, vol IV.
Karger, Basel, pp 178192

Fellows SJ, Dmges F, Trpper R, Thilmann AF, Noth J (1993)

Changes in the short- and long-latency stretch reflex components of the triceps surae muscle during ischaemia in man.
J Physiol 472:737748
Gottlieb GL, Agarwal GC (1979) Response to sudden torques
about ankle in man: myotatic reflex. J Neurophysiol 42:91
106
Gottlieb GL, Chen CH, Corcos DM (1995) Relations between
joint torque, motion, and electromyographic patterns at the human elbow. Exp Brain Res 103:164167
Hkkinen K, Komi PV (1986) Training induced changes in neuromuscular performance under voluntary and reflex conditions.
Eur J Appl Physiol 55:147155
Johnson MA, Polgar J, Weightman D, Appleton D (1973) Data on
the distribution of fiber types in thirty-six human muscles: an
autopsy study. J Neurol Sci 18:111129
Kearney RE, Chan CWY (1982) Contrast between the reflex responses of tibialis anterior and triceps surae to sudden ankle
rotation in normal subjects. Electroencephalogr Clin Neurophysiol 54:301310
Kearney RE, Hunter IW (1983) System identification of the human triceps surae stretch reflex dynamics. Exp Brain Res
51:117127
Kirsch RF, Kearney RE (1993) Identification of time-varying dynamics of the human triceps surae stretch reflex. II. Rapid imposed movement. Exp Brain Res 97:128138
Lambertz D, Prot C, Kaspranski R, Goubel F (2001) Effects of
long-term spaceflight on mechanical properties of muscles in
humans. J Appl Physiol 90:179188
Mardsen CD, Merton PA, Morton HB (1976) Servo action in the
human thumb. J Physiol 257:144
Matthews PBC (1986) Observation on the automatic compensation of reflex gain on varying the pre-existing level of motor
discharge in man. J Physiol 374:7390
Matthews PBC (1994) The simple frequency response of human
stretch reflexes in which either short- or long-latency components predominate. J Physiol 481:777798
Mirbagheri MM, Barbeau H, Kearney RE (2000) Intrinsic and
reflex contributions to human ankle stiffness: variation
with activation level and position. Exp Brain Res 135:423
436
Morita H, Petersen N, Christensen LOD, Sinkjaer T, Nielsen J
(1998) Sensitivity of H-reflexes and stretch reflexes to presynaptic inhibition. J Neurophysiol 80:610620
Neilson PD, McCaughey J (1981) Effect of contraction level and
magnitude of stretch on tonic stretch reflex transmission characteristics. J Neurol Neurosurg Psychiatry 44:10071012
Rack PHM, Ross HF, Brown TIH (1978) Reflex responses during
sinusoidal movement of human limbs. In: Desmedt JE (ed)
Cerebral motor control in man: long loop mechanisms. Progress in clinical neurophysiology, vol IV. Karger, Basel,
pp 216228
Rack PHM, Ross HF, Thilman AF, Walters DKW (1983) Reflex
responses at the human ankle: the importance of tendon compliance. J Physiol 344:503524
Sale DG (1988) Neural adaptation to resistance training. Med Sci
Sports Exerc 20:S135S145
Sinkjaer T (1997) Muscle, reflex and central components in the
control of the ankle joint in healthy and spastic man. Acta
Neurol Scand 170:S1S28
Toft E, Sinkjaer T, Andreassen S, Larsen K (1991) Mechanical
and electromyographic responses to stretch of the human ankle
extensors. J Neurophysiol 65:14021410
Tognella F, Mainar A, Vanhoutte C, Goubel F (1997) A mechanical device for studying mechanical properties of human muscles in vivo. J Biomech 30:10771080
Voigt M, Chelli F, Frigo C (1998) Changes in the excitability of
the soleus muscle short latency stretch reflex during human
hopping after 4 weeks of hopping training. Eur J Appl Physiol
78:522532