JRRD
Volume 45, Number 1, 2008
Pages 175–186
Journal of Rehabilitation Research & Development
Synchronous stimulation and monitoring of soleus H reflex during robotic
body weight-supported ambulation in subjects with spinal cord injury
Ross G. Querry, PT, PhD;
1*
Fides Pacheco, MD;
2–3
Thiru Annaswamy, MD, MA;
3–4
Lance Goetz, MD;
2–3
Patricia K. Winchester, PT, PhD;
1
Keith E. Tansey, MD, PhD
1
1
Spinal Cord Injury Laboratory, Department of Physical Therapy, The University of Texas Southwestern Medical
Center, Dallas, TX;
2
Spinal Cord Injury Service, Department of Veterans Affairs (VA) North Texas Health Care System,
Dallas, TX;
3
Department of Physical Medicine and Rehabilitation (PM&R), The University of Texas Southwestern
Medical Center, Dallas, TX;
4
PM&R Service, VA North Texas Health Care System, Dallas, TX
Abstract—We evaluated the accuracy of a novel method for
recording the soleus H reflex at specific points in the gait cycle
during robotic locomotor training in subjects with spinal cord
injury (SCI). Hip goniometric information from the Lokomat
system defined midstance and midswing points within the gait
cycle. Soleus H reflex stimulation was synchronized to these
points during robotic-assisted ambulation at 1.8 and 2.5 km/h.
Motor stimulus intensity was monitored and adjusted in real
time. Analysis of 50 H reflex cycles during each speed and gait
phase showed that stimulation accuracy was within 0.5° of the
defined hip joint position and that >85% of the H reflex cycles
met the +/–10% M wave criterion that was established during
quiet standing. This method allows increased consistency of
afferent information into the segmental spinal and supraspinal
circuitry and, thus, evaluation of H reflex characteristics during
robotic ambulation in subjects with SCI.
Key words: body weight-supported treadmill training, gait
training, H reflex, locomotor training, motor control, muscle
afferents, reflex activity, rehabilitation, robotic-aided training,
spinal cord injury.
manner, such as standing versus walking [2–5]. A pro-
found phase-dependent modulation, including modula-
tion associated with ambulation speed, also occurs during
walking and running [2,5–8]. In addition to reflex modu-
lation coupled to muscle activation, isolated joint angle-
dependent modulation of H reflex activity has been dem-
onstrated at the ankle [9–10] and the hip [11] independent
of motor neuronal excitation. The differences in the
H reflex modulation in these tasks is evidence that the
changes seen during the gait cycle are not simply due to
the α-motor neuron excitation level, as indicated by elec-
tromyography (EMG), but also may be modulated by
supraspinal, homonymous, and heteronymous afferent
inputs and interneuronal activity, as well as by intrinsic
Abbreviations: ANOVA = analysis of variance, ASIA =
American Spinal Injury Association, BWSTT = body weight-
supported treadmill training, DGO = driven gait orthosis,
EMG = electromyography, H
max
= maximal H reflex ampli-175
INTRODUCTION
The soleus H reflex has been used as a tool for
assessing monosynaptic reflex excitability in humans
during both rest and voluntary activity [1]. The amplitude
of the soleus H reflex is modulated in a task-dependent
tude, M
Hmax
= M wave amplitude at H
max
, M
max
= maximal
M wave amplitude, SCI = spinal cord injury, SD = standard
deviation.
*
Address all correspondence to Ross G. Querry, PT, PhD;
Department of Physical Therapy, The University of Texas
Southwestern Medical Center, 5323 Harry Hines Boulevard,
Suite V6.100, Dallas, TX 75390-8876; 214-648-1509; fax: 214-
648-1511. Email: ross.querry@utsouthwestern.edu
DOI: 10.1682/JRRD.2007.02.0028

176
JRRD, Volume 45, Number 1, 2008
properties of the motor neuron. To evaluate H reflex
modulation during the gait cycle over different condi-
tions or time, a methodology that addresses lower-limb
joint position, loading, and phase-dependent muscle acti-
vation patterns is preferred.
The most common method described in previous
investigations of timing stimulation of the soleus H reflex
during gait has been the use of a foot switch marker that
identifies the initial contact of the stance phase or the step
cycle EMG patterns of the soleus and tibialis anterior
muscles. This method involves the conduction of pseudo-
random stimulations of the H reflex across the gait cycle,
with post hoc division of the gait cycle into 8 to 20 phases.
Custom computer algorithms assign the acquired reflex
cycles to the appropriate phase of gait on the basis
of timing latencies or EMG activity for signal-averaging
of 8 to 10 reflex cycles at each point in the gait cycle [1,4–
5,12]. Although this method has been effective, analysis
of H reflex modulation across varying speeds would
require recalculation of gait phases on the basis of chang-
ing latencies with changes in ambulation speed. Addition-
ally, dividing the gait cycle into a number of phases (8 to
10) that include a range of joint positions could cause
increased variance of the H reflex across the phase as a
result of variations in muscle activation and joint position
at the hip and ankle.
The H reflex is also a commonly used clinical tool for
assessing reflex excitability after spinal cord injury (SCI)
[13–17]. H reflex responses after SCI have been shown to
be different than normal spinal cord physiology at rest and
during stepping [1,18–20]. Body weight-supported tread-
mill training (BWSTT) has been increasingly applied as a
clinical tool for rehabilitation of standing and walking in
patients with SCI [21–25]. Recent developments in
robotic devices for BWSTT have provided researchers
and clinicians the unique ability to monitor joint position,
torques, and subject performance with increased accuracy,
precision, and subject tolerance. The high repeatability
and control of gait kinematics with robotic locomotor sys-
tems allow for improved control of many of the factors
that modulate H reflex excitability. Robotic locomotor
training systems should allow accurate synchronization of
H reflex stimulation to defined points in the gait cycle.
This methodology should provide more consistent and
controlled afferent information from muscle activation,
joint position, and loading than the routinely used random
stimulation cycles. Robotic BWSTT may also allow better
control of or reduced step-to-step variability than manu-
ally assisted BWSTT. Data collected in this manner may
provide new insights into afferent and central regulation
of human motor control during the natural motor task of
walking, especially in subjects with SCI who are unable to
step on their own and have been difficult to study during
ambulation.
We conducted this study to develop and evaluate a
methodology that would result in more precise and accu-
rate stimulation of the soleus H reflex synchronously
with specific points in the gait cycle during robotic
BWSTT in subjects with and without SCI.
METHODS
Subjects
The local committees for the protection of human
subjects at the Dallas Department of Veterans Affairs
Medical Center and The University of Texas Southwest-
ern Medical Center approved this investigation. A total of
26 subjects (17 male, 9 female) volunteered to partici-
pate, including 4 subjects without SCI and 22 subjects
with SCI who had varying degrees of injury complete-
ness and functional ability. Table 1 summarizes the sub-
ject demographics. All subjects provided written consent.
Table 1.
Demographics of participants in H reflex methodology study. American Spinal Injury Association (ASIA) spinal cord injury (SCI) classifications
of A and B are described as motor complete, ASIA C and D as motor incomplete.
Classification
Sex
No.
Age (yr)
(mean ± SD)
Time Since Injury (mo)
(mean ± SD)Male Female
ASIA A 7 1 8 34.4 ± 5.2 60.8 ± 40.2
ASIA B 2 2 4 35.5 ± 12.2 51.9 ± 36.6
ASIA C 2 4 6 28.2 ± 10.0 44.8 ± 39.6
ASIA D 3 1 4 44.0 ± 11.7 26.4 ± 15.2Non-SCI 3 1 4 35.3 ± 3.5 —
SD = standard deviation.

177
QUERRY et al. Soleus H reflex during gait training and SCI
H Reflex Instrumentation
To record the soleus H reflex EMG, we placed sub-
jects in a prone position and placed Ag-Cl surface elec-
trodes (Blue Sensor, Ambu; Ballerup, Denmark) over the
soleus muscle. We prepared the electrode sites by shav-
ing the skin and mildly abrading it with prep-paper
and alcohol to reduce skin impedance to less than 5 kΩ .
The recording electrode was placed over the distal third
of the soleus muscle just below the insertion of the gas-
trocnemius muscle onto the Achilles tendon in order to
selectively record from the soleus. The reference elec-
trode was placed over the Achilles tendon approximately
6 cm above the calcaneus. A reference ground electrode
was placed over the fibular head.
The recording electrode signal was amplified at a
fixed gain of 500 by a bioamplifier (Biopac EMG 100C,
Biopac Systems Inc; Santa Barbara, California) and was
bandpass-filtered between 1 Hz and 5 kHz. The signal
was sent to a 16 bit analog-to-digital converter (Biopac
MP 150, Biopac Systems Inc; Santa Barbara, California)
and sampled at 2 KHz on a Pentium personal computer.
The computer-based data acquisition system (AcqKnowl-
edge, Biopac Systems Inc; Santa Barbara, California)
collected, monitored, and stored the signal on hard disk
for post hoc analysis.
The soleus H reflex was elicited by stimulation of the
tibial nerve in the popliteal fossa. A 2 in.-diameter poly-
mer anode was placed anteriorly just above the patella. We
used a handheld electrode to locate the optimum site for
nerve stimulation distal to the popliteal fossa. The crite-
rion for the optimum site was the motor point that yielded
the largest M wave amplitude during low-intensity stimu-
lation. The handheld electrode was replaced with a 1 in.
polymer cathode (Empi; St. Paul, Minnesota) that was
placed on the skin at the optimum stimulation site. The
electrodes were secured with adhesive tape so that the
stimulating electrodes constantly contacted the underlying
skin during all locomotor tasks. The cathode placement
distal to the crease of the popliteal fossa helped avoid elec-
trode movement relative to the nerve during the experi-
ment and ambulation. The nerve stimulus was a 1 ms
monophasic square pulse delivered by a constant current
stimulator (Digitimer DS7A, Digitimer Limited; Hertford-
shire, United Kingdom).
Robotic Instrumentation
AG; Volketswil, Switzerland) robotic gait orthosis. The
Lokomat driven gait orthosis (DGO) and Lokolift (Hoc-
oma AG; Volketswil, Switzerland) dynamic unweighting
system assist subjects during standing and walking. Sub-
jects were fitted with a weight-supporting harness, and
the Lokolift body weight-support system helped them
stand on the treadmill. We set body-weight support at
40 percent of the subject’s weight to ensure that a consis-
tent protocol was used with the subjects from each Amer-
ican Spinal Injury Association (ASIA) classification and
with the non-SCI subjects. This protocol included the
provision of an adequately safe environment for the sub-
jects with SCI, as well as sufficient support for them to
ambulate with robotic assistance for a minimum of 30
minutes. The rigid-framed DGO was secured and aligned
to the subject with cloth cuffs that attached around the
thigh and shank of the lower leg. The foot and ankle were
controlled by attachment of the spring-loaded straps on
the lower arm of the DGO to the subject’s forefoot. Pel-
vic straps connected the DGO to the weight-supporting
harness. Although the Lokomat uses a rigid frame struc-
ture that is aligned with the subject’s hip and knee and
helps stabilize the pelvis, the limb is secured with cloth
straps that cannot guarantee that the actual joint position
will be accurately aligned with the Lokomat joint axis
during ambulation. During locomotion, the subject’s gait
pattern was assisted by direct-drive linear actuators
aligned bilaterally at the hip and knee and computer-con-
trolled to generate a symmetrical gait pattern synchro-
nized to the speed of the underlying treadmill (Figure 1).
The Lokomat computer interface allowed the investigator
to adjust parameters of step length, hip, and knee range of
motion to approximate normal kinematics for each sub-
ject. Colombo et al. published a more detailed description
of the Lokomat device [26–27].
The Lokomat DGO included a computer interface
card that allowed goniometric position and direct current
motor force information from the hip and knee joints to
be output in real time during locomotor tasks. This infor-
mation was integrated with the external data acquisition
equipment to collect and monitor subjects’ hip joint
information within the Lokomat DGO and dynamically
synchronize the H reflex stimulation and response during
the gait cycle according to the defined criterion.
H Reflex ProtocolOnce subjects were instrumented for H reflex acqui-
sition, they were transferred into the Lokomat
®
(Hocoma
We analyzed H reflex responses by calculating the
peak-to-peak amplitude of the evoked motor response

178
JRRD, Volume 45, Number 1, 2008
recorded from the soleus muscle. Data acquisition was
triggered by each stimulation onset. Software-controlled
graphical displays allowed for a 100 ms time-amplitude
window representing the current H reflex stimulus-
response, as well as a graphical time-amplitude display of
the serial stimulus-response curves. Before testing the
H reflex under synchronized robotic locomotor condi-
tions, we recorded the H reflex and M wave recruitment
characteristics with the subject in the prone position. We
was gradually increased from a level below the H reflex
or motor (M wave) threshold to an intensity eliciting the
maximal M wave amplitude (M
max
). Specific identified
variables were the maximal H reflex amplitude (H
max
),
the M wave amplitude at H
max
(M
Hmax
), and the M
max
.
After placing the subject in the Lokomat, we repeated the
stimulus ramping protocol with the subject in a quiet
standing position with 40 percent body-weight unload-
ing. The stimulus-response output obtained in quiet
standing was then used to standardize the stimulus inten-
sity for all locomotor tasks. The M
Hmax
during quiet
standing was identified as the desired independent vari-
able to control during locomotor tasks. The M wave
amplitude is the response of the α-motor fibers to direct
stimulation. Maintaining the same proportion of activated
α-motor fibers is widely assumed to consistently activate
Ia afferents, allowing valid evaluation of H-reflex charac-
teristics, particularly H
max
across different tasks [2,28].
Synchronized Ambulation Protocol
Once placed in the Lokomat system, subjects
remained at 40 percent body-weight support and ambu-
lated at both 1.8 and 2.5 km/h. Hip and knee joint posi-
tions were sampled at 500 Hz from the Lokomat
goniometric output (Biopac MP150). A hardware digital
output channel was set so that the Lokomat hip position
information triggered the external stimulator output at
specifically defined points in the gait cycle for soleus
H reflex acquisition (Figure 2).
Initially, the midstance position of the gait cycle was
defined and selected as the criterion for synchronized
stimulation of the H reflex. Midstance (20% into the gait
cycle from initial contact) was selected as a period of sin-
gle-limb support, and afferent input through the lower limb
was similar to standing with the ankle in approximately 5°
dorsiflexion. Midstance was defined in the software algo-
rithm as 0° hip position from the Lokomat goniometric
output. We added a midswing protocol (75% into the gait
cycle from initial contact) as a second trial after confirm-
ing the stability of the midstance-synchronized acquisition
protocol and collected data on 15 of the 22 subjects
(5 ASIA A, 2 ASIA B, 2 ASIA C, 3 ASIA D, and 3 non-
SCI). Midswing was selected as a point of limb unloading
and defined in the software algorithm as the point of maxi-
mal hip flexion (30° ± 2°) recorded from the Lokomat
goniometric output (Figure 3). The corresponding position
Figure 1.
H reflex instrumentation with subject walking in Lokomat
®
(Hocoma
AG; Volketswil, Switzerland). H reflex stimulation and recording
electrodes are positioned before subject is placed in Lokomat.used a manually triggered ramping protocol to define the
stimulus-response relationship. The stimulus intensity
of the knee during the defined midstance and midswing
phases of the Lokomat’s programmed kinematic path

179
QUERRY et al. Soleus H reflex during gait training and SCI
were 7° ± 1° and 43° ± 2° of knee flexion, respectively
(Figure 3). The software algorithm was set to control the
external stimulator at these defined points. Trial 1 was
H reflex stimulation synchronized to midstance at both the
1.8 and 2.5 km/h ambulation speeds. Stimulation trigger-
ing frequency was approximately 0.5 to 1 Hz, depending
on speed and cadence, but user control of the stimulator
output to the subject allowed for interruption of the syn-
chronized trigger to prevent postactivation depression of
the stimulus.
During the locomotor tasks at 1.8 and 2.5 km/h, the
M
Hmax
was monitored in real time through the software
time-amplitude graphical displays. Deviations greater
than ±10 percent of the M
Hmax
standardized in the quiet-
standing condition for more than two sequential steps
resulted in adjustment of the stimulus intensity to restore
the appropriate M
Hmax
value. The criterion of ±10 per-
cent of M
Hmax
was initially defined as the acceptable
M wave variability that would minimally affect H reflex
amplitude response variability.
The series of H reflex events were recorded for
H reflex cycles were collected at each walking speed to
evaluate M wave variability. Ambulation speed was
Figure 2.
Specific points of gait cycle are defined and used to control stimula-
tion of soleus H reflex. Motor response is monitored for necessary
adjustment of stimulus intensity.
Figure 3.
Output of Lokomat
®
(Hocoma AG; Volketswil, Switzerland) gonio-
metric position of hip (black line) and knee (gray line) during ambula-
tion. Software control of external stimulator was set to predefined
points in gait cycle: (a) midstance, defined as 0° of hip flexion, and
(b) midswing, defined as point of maximal hip flexion. Synchronized
stimulator control pulse is shown.
Figure 4.
Graphical output of sequential H reflex stimulation cycles. H reflex
data acquisition was triggered by stimulus artifact and collected for
100 ms, allowing viewing and analysis of serial H reflex cycles without
acquisition of interstimulus information. Post hoc analysis defined each
100 ms H reflex event. Events that met set criterion for M wave ampli-
tude were signal-averaged for measurement of H reflex amplitude.100 ms event windows triggered by the onset of the stimu-
lation artifact (Figure 4). A total of 50 synchronized
increased from 1.8 to 2.5 km/h without software control
adjustments because stimulation control was synchronized

180
JRRD, Volume 45, Number 1, 2008
to real-time hip joint position with each step cycle. In
addition to the steady state ambulation H reflex cycles,
M
max
data were also acquired at each ambulation speed to
confirm the stability of the maximum motor neuron stimu-
lation amplitude.
For Trial 2, the software stimulator control window
was adjusted so that the defined midswing parameters
became the stimulus triggers. The synchronized stimula-
tion protocol at the two walking speeds and at M
max
were
repeated. Stimulation intensity was modified as needed to
maintain the M
Hmax
measured in quiet standing and used
during the midstance trial. After the two walking trials,
we remeasured skin impedance to identify any changes
that may have affected electrical signal amplitudes. Post
hoc analysis included stringent selection of the H reflex
cycles meeting the ±10 percent of M
Hmax
criterion from
the quiet standing and ambulation trials. All acceptable
cycles were signal-averaged with software event selection
and signal-processing algorithms (DataPac, Run Technol-
ogies; Mission Viejo, California).
Data Analysis
Data are presented as mean ± standard deviation (SD)
for central tendency and variance. Statistical comparisons
of subjects within ASIA classifications and non-SCI sub-
jects for standing and walking or midstance and midswing
phases were analyzed with paired t-tests that evaluated
differences in the measured variables (Excel, Microsoft
Corporation; Redmond, Washington). Two-way analysis
of variance (ANOVA) with Bonferroni analysis compared
differences between subjects within ASIA classifications
and non-SCI subjects for midswing and midstance vari-
ables (SPSS, SPSS Inc; Chicago, Illinois). The level of
significance was set at p < 0.05 for all analyses.
RESULTS
An important construct of this methodology was to
maintain the same proportion of α-motor neuron activa-
tion during sequential step cycles by monitoring the
M
Hmax
during synchronized stimulations. The effect of
the predefined threshold criterion of ±10 percent of
M
Hmax
on H reflex amplitude variability was analyzed
with initial data collection in motor complete, motor
incomplete, and non-SCI subjects. The typical response
amplitude on the ascending portion of the H-M sensitiv-
ity curve, with an attenuated sensitivity to increasing
M wave amplitude at H
max
and across the descending
portion of the curve to M
max
(Figure 5). This finding was
Figure 5.
M wave to H reflex response sensitivity graphs. Data represent tibial-
nerve stimulation ramping protocol from no M wave to maximal
M wave amplitude (M
max
). Presented are typical responses of subjects
(a) without spinal cord injury (SCI), (b) with American Spinal Injury
Association (ASIA) A (motor complete) SCI, and (c) with ASIA C
(motor incomplete) SCI. Solid vertical line represents measured M waveto the increasing stimulus intensity ramp was a higher
sensitivity of H wave amplitude to increasing M wave
that elicited maximal H reflex (M
Hmax
). Dotted vertical line represents
±10% of M
Hmax
. M
Hmax
percentage of M
max
is calculated.

181
QUERRY et al. Soleus H reflex during gait training and SCI
consistent among the subjects. Identification of M
Hmax
and the ±10 percent criterion lines provided evidence that
maintaining this range of motor nerve activation mini-
mally affected H wave variability within this range. This
criterion was then used to adjust stimulus intensity as
needed during data acquisition and for post hoc accep-
tance for H reflex cycles for all subjects and trials.
The M
max
varied for each individual subject but was
stable for any subject across conditions. The percentage
of M
max
where M
Hmax
occurred also varied with each
subject (Figure 5). In some subjects, H
max
occurred
before the onset of an M wave (0% of M
max
), while in
other subjects, H
max
occurred at >20 percent of M
max
.
Grouped data by SCI classification for this percentage of
M
max
in standing and walking are presented in Table 2 .
Paired t-test statistical analysis showed no significant dif-
ference in the M
Hmax
percentage of M
max
between stand-
ing and walking for subjects within ASIA classifications
and for non-SCI subjects (p > 0.05).
The described methodology resulted in precise syn-
chronicity of the soleus H reflex stimulation with the
defined midstance and midswing phases of the gait cycle.
The internal latency of the system from hip joint position
trigger to stimulation output to the subject was measured
as 16 ± 4 ms. During BWSTT at varying speeds, the
stimulation onset was highly accurate and repeatable.
Direct measurement of 20 sequential step cycles at both
1.8 and 2.5 km/h in 10 subjects during midstance resulted
in a pooled stimulation output to subjects at 0.3° ± 0.2° of
Lokomat hip extension position. Once the stimulus inten-
sity was adjusted during a testing condition to obtain
M
Hmax
, then intratrial adjustment within ±3 mA of cur-
rent maintained the M
Hmax
criterion in each of the prone,
standing, and ambulation trials. During the midswing tri-
als, the position of the knee was flexed approximately
30° more than during midstance. This alteration in elec-
trode distance required an increase of 3 to 5 mA in stimu-
lation intensity compared with the midstance level. Once
established, the stimulation criterion in midswing was
also maintained within ±3 mA
M wave amplitudes during serial H reflex stimuli indi-
cated minimal variability during a specific walking task.
During post hoc analysis, 1,150 step cycles at each ambu-
lation speed were evaluated in midstance and 350 step
cycles in midswing. The M
Hmax
variability resulted in
acceptance of 87.5 ± 12.6 percent and 84.7 ± 12.7 percent
of midstance H reflex complexes of pooled data at 1.8 and
2.5 km/h, respectively. During midswing, 88.8 ± 1.4 per-
cent and 90.6 ± 1.8 percent of the acquired H reflex cycles
met the M
Hmax
criterion of pooled data at 1.8 and 2.5 km/h,
respectively. Paired t-test results indicated no difference
between H reflex acceptance rates between ambulation
speeds for a given gait phase (p > 0.05). Acceptance rates
for stance and swing phases by SCI classification are pre-
sented in Table 3 . The ANOVA between subjects within
the ASIA classifications and the non-SCI subjects indi-
cated no significant differences between H reflex cycle
acceptance rates during each gait phase.
The total ambulation time required for each trial of
50 steady state cycles at each speed was only 4 to 5 min-
utes, even in subjects with SCI. The swing trial was con-
ducted on a separate occasion for seven of the subjects.
For the remaining eight subjects, it was completed in
sequence after the stance phase trial on the same testing
day and was established as the standard protocol. The
Table 2 .
M
Hmax
percentage of M
max
. Standing and walking data (mean ± stan-
dard deviation) shown grouped by spinal cord injury (SCI) classifica-
tion. No significant difference was found between standing and
walking percentages (p > 0.05).
Classification
(No. of Subjects)
Standing Walking
ASIA A (8) 15.6 ± 7.7 15.6 ± 9.1
ASIA B (4) 8.5 ± 9.1 10.0 ± 10.1
ASIA C (6) 8.4 ± 4.9 8.4 ± 4.8
ASIA D (4) 11.5 ± 7.8 11.8 ± 9.4
Non-SCI (4) 14.3 ± 4.4 13.9 ± 7.2
Table 3 .
Acceptance rates (mean ± standard deviation %) of H reflex cycles
meeting M
Hmax
criterion. Data are pooled for 1.8 and 2.5 km/h ambu-
lation speeds (n = 22 for stance, 15 for swing). No significant differ-
ences were found between spinal cord injury (SCI) classification
groups for each gait phase (p > 0.05).
Classification Stance Swing
ASIA A 83.6 ± 14.2 86.1 ± 4.2
ASIA B 82.5 ± 9.9 88.7 ± 5.6
ASIA C 86.0 ± 10.6 84.9 ± 4.1
ASIA D 89.2 ± 14.6 89.5 ± 8.1
Non-SCI 88.9 ± 12.4 90.1 ± 4.2
All Subjects 86.1 ± 12.6 88.1 ± 4.9ASIA = American Spinal Injury Association, M
Hmax
= M wave amplitude at
maximal H reflex amplitude, Mmax = maximal M wave amplitude.
ASIA = American Spinal Injury Association, M
Hmax
= M wave amplitude at
maximal H reflex amplitude.

182
JRRD, Volume 45, Number 1, 2008
total ambulation time for collection of swing and stance
phase data was within 30 minutes for all subjects.
Ultimately, the M
Hmax
criterion for monitoring stimu-
lation intensity and post hoc H reflex cycle acceptance
resulted in signal-averaged data. Examples of signal-aver-
aged data during the midstance trial for a typical subject
in each ASIA classification and for non-SCI subjects are
shown in Figure 6. The M
Hmax
defined in the standing
condition was maintained across ambulation at 1.8 and
2.5 km/h during the midstance and midswing trials (only
midstance phase is shown). The signal-averaged graphs
indicated that variability of the M wave within the ±10
percent criterion resulted in low H wave variability dur-
ing standing (three to five stimulation cycles) and
increased variability in H wave amplitudes during walk-
ing. The M wave variance was controlled in all trials, thus
providing evidence that the variance in H reflex ampli-
tudes was attributable to the integration of spinal and
supraspinal inputs on H reflex modulation and not to
methodology limits. The variability of the differences in
M
Hmax
in relation to H
max
that occurs on an individual
subject basis is also evident in Figure 6.
DISCUSSION
This study developed and evaluated a methodology
that could integrate the recent advances in robotic tech-
nology for assisted ambulation in subjects with SCI with
the acquisition of the soleus H reflex. The goal was to
increase the precision and accuracy of acquiring the
soleus H reflex at specific points in the gait cycle during
robotic BWSTT in subjects with and without SCI. The
main criteria for evaluating the methodology were (1) its
accuracy in stimulating the H reflex at a predefined point
in the gait cycle, (2) its ability to change ambulation
speed while maintaining synchronized H reflex stimula-
tion, (3) its ability to monitor M wave amplitude in real
time at the defined M
Hmax
in order to maintain equal
stimulus intensity during ambulation and across the dif-
ferent protocol phases, and (4) the variability of the
M
Hmax
during steady state ambulation speed to evaluate
the number of step cycles needed to obtain sufficient
H reflex cycles meeting M wave criterion while minimiz-
ing the ambulation time for subjects with SCI.
With the accuracy of the goniometric output from the
Lokomat coupled to the control of the external stimula-
tor, the soleus H reflex was acquired with less than 1°
variability in the Lokomat hip joint position on sequential
gait cycles. Although hip joint position controlled stimu-
lation, knee joint position within the Lokomat was also
consistent and therefore reduced the variability of affer-
ent information from the knee on the H reflex. One must
consider that, along with the Lokomat reporting mechani-
cal joint positions with high accuracy, actual hip and knee
Figure 6.
Signal-averaged H reflex cycles. Presented are typical subjects
(a) with American Spinal Injury Association (ASIA) A (motor com-
plete) spinal cord injury (SCI), (b) with ASIA B (motor complete)
SCI, (c) with ASIA C (motor incomplete) SCI, (d) with ASIA D
(motor incomplete) SCI, and (e) without SCI during standing (3–5joint positions are likely to be slightly different because
of the motion freedom of the limb held in place by the
cycles) and ambulation (approximately 35–40 cycles) at 1.8 and
2.5 km/h during midstance-synchronized stimulation.

183
QUERRY et al. Soleus H reflex during gait training and SCI
cloth cuffs.
*
Although this difference may occur, the
repeatability of the gait pattern should allow better con-
trol of stimulation guidance than was previously possible
during free walking or with manual BWSTT.
The majority of the studies investigating the soleus
H reflex during ambulation used random or incremental
H reflex stimulation across the gait cycle and then used a
post hoc division of H reflex cycles into 8 to 20 bins that
fell within a specified range of the gait cycle. For calcula-
tion of the H reflex amplitude of each bin, 8 to 10 reflex
cycles were averaged [1,3–6,12]. Although this method
may be time efficient for acquiring H reflex cycles across
the entire gait cycle, it results in division of the normal-
ized gait cycle (100%) into bins in which the hip and knee
joints would be positioned at random points covering 5 to
12 percent of the full step cycle but signal-averaged
together for a single data point. Variation in the joint posi-
tion of the knee, ankle, and specifically the hip, may alter
afferent input from joint loading and muscle activation
across a single bin. Afferent input from step to step may
also vary within a given bin. Given that H reflex ampli-
tudes have been shown to be modulated by both phase-
dependent muscle activation levels during ambulation
[2,5–8] and by ankle [9–10] and hip joint positions
[11,29], precisely stimulating sequential H reflex cycles
during the gait cycle may reduce variability in modulation
factors that affect H reflex output. Robotic-controlled
BWSTT provides highly repeatable gait characteristics
that would be difficult to control during therapist-assisted
BWSTT.
In addition to the high precision of gait-synchronized
H reflex stimulation during robotic-controlled BWSTT,
ambulation speed with this method could be adjusted as
desired or needed without loss of stimulation precision or
the need for instrumentation adjustments. This reduced the
time required to complete the protocol of 50 H reflex
cycles at two different ambulation speeds with body-
weight support. Collection of only 10 to 12 H reflex cycles
for signal-averaging at a specific gait phase, as is com-
monly reported, could be completed within a short time
and at several points within the gait cycle within 5 minutes
of ambulation. During investigations of H reflex modula-
tion in subjects with SCI, completing the protocol in a
minimal amount of time is a significant benefit because
of subject tolerance and fatigue factors. However, even
with the assistance of robotics, fatigue would still remain a
factor. Therefore, rapid acquisition of the required H reflex
cycles would augment the ability to minimize subject
fatigue and changes in motor activation. In order to
approximate a synchronized stimulation to a specific point
in the gait cycle, methods that use latency from a set trig-
ger such as initial contact would require tedious measure-
ments to adjust latency parameters. These measurements
would include measurements and calculations for each
change in ambulation speed and for each individual sub-
ject’s cadence for any given speed. Although changes in
body-weight support were not measured in this investiga-
tion, future use of this methodology should allow adjust-
ments with confidence in simulation synchronization.
Maintaining the same proportion of activated α-motor
fibers, measured by the M-wave, is widely assumed to
demonstrate a constant level of Ia afferent activation and
allow valid evaluation of H reflex characteristics, particu-
larly H
max
, across tasks [2,28]. Considerable variability
exists in how previous investigations have selected the
H reflex stimulation intensity criteria. Methods include
selecting a constant percentage of the M
max
to be main-
tained across conditions. Investigators have used intensi-
ties ranging from 10 to 30 percent [7,30–33], multiple
stimulation intensities with post hoc analysis to match
M wave cycles [2,6], or have not reported. Acceptable
M wave variance for these studies was between ±3 to
5 percent of selected intensity. If a single stimulus is
selected based on M
max
intensity across subjects, data
from this investigation would suggest that the position of
the stimulus intensity on the M-H recruitment curve
would vary between subjects. This variability may com-
plicate the interpretation of group results because individ-
ual data may be collected at different portions of the
curve. If the selected stimulus intensity is on the ascend-
ing limb of the M-H recruitment curve, small changes in
M wave amplitude may alter H reflex responses. The
methodology presented here selected M wave intensity
that corresponded with the individual’s maximal H reflex
response in quiet standing and with the acceptance toler-
ance of the criterion ±10 percent. The H
max
portion of the
recruitment curve has been shown to be an area of attenu-
ated change in H wave amplitude with M wave change
[2]. Our data supported that the criterion of ±10 percent
M
Hmax
resulted in minimal H reflex variability. The cho-
sen stimulus intensity M
Hmax
in quiet standing was used
across all trials. The M
Hmax
was monitored in real timeand required minimal or no adjustment from the initial
quiet standing value across experimental conditions.
*
Joe Hidler, personal communication and unpublished data, June 2006.

184
JRRD, Volume 45, Number 1, 2008
Comparing H reflex acceptance rates is difficult
because previous investigations have not stated the total
number of H reflex cycles accepted compared with the
number of cycles collected or the M wave and H wave
variability. The synchronization methodology resulted in
85 percent or greater of the 50 H reflex cycles meeting
the rigid post hoc criterion of M
Hmax
±10 percent.
H reflex acceptance rates did not differ between ambula-
tion speeds or between subjects with or without SCI. The
minimal variability of the M
Hmax
may result partly from
the consistency of joint position and afferent input previ-
ously discussed with this methodology. This high level of
stability will allow us to reduce the number of H reflex
cycles collected and further reduce the necessary data
collection time.
We chose midstance and midswing to allow evalua-
tion of synchronized H reflex acquisition at two different
positions within the gait cycle, with very different joint
and muscle activation patterns. Midstance was defined to
maximally load the hip and knee and approximate a neu-
tral ankle position, thus maximizing afferent information
from the kinetic chain. This position would be similar to
the loading characteristics during standing, which has
been shown to have increased H reflex amplitudes [2,4],
and would allow a comparison of H reflex modulation in
a dynamic ambulation activity and static standing with
very similar positioning. Midswing was defined at the
point of maximal hip flexion. Previous studies have indi-
cated that passive hip flexion of 20° to 30° attenuates the
H reflex amplitude compared with hip extension [11,29].
Therefore, the midswing position would allow the inves-
tigation of dynamic modulation of similar hip position-
ing. The Lokomat uses a spring mechanism placed under
the forefoot to control the ankle position for foot clear-
ance. During the stance phase, the strap-spring mecha-
nism that controls the ankle minimally affects the loading
characteristics in the Lokomat. During swing, however,
the spring force of the ankle mechanism maintains the
ankle in neutral to slight dorsiflexion for safety and does
not require active use of the tibialis anterior, even in sub-
jects without SCI. For this investigation, the stability of
M
Hmax
and M
max
was consistent in both midstance and
midswing. Although actual limb position and the robotic-
limb joint axis at the hip, knee, and ankle during gait will
differ, the differences may not be clinically significant
and robotic support provides a level of repeatability that
would be difficult to impossible to control manually.
robotic device that offered dynamic goniometric output.
Robotic devices continue to develop, and the Lokomat
device is currently in 10 Department of Veterans Affairs
facilities and more than 100 are in service worldwide, with
the potential for increased clinical and research use in the
future. All the hardware and software used with the Loko-
mat are commercially available and require no proprietary
equipment or programming abilities. The crossover appli-
cations that would allow use of this methodology with
other research measures or electronic goniometers for
overground, synchronized H reflex stimulation are
expected to be developed.
CONCLUSIONS
Commercially available hardware and software can
obtain a highly repeatable soleus H reflex response syn-
chronously with any specified position in the gait cycle
during robotic BWSTT in subjects with all ASIA levels
of SCI. Synchronized, kinematic-controlled stimulation
increased the consistency of hip and knee positions and
may provide increased control of the muscle length and
muscle force afferent information presented to the seg-
mental spinal and supraspinal circuitry that generate and
modulate the H reflex. Visual and software analysis of
M wave amplitude during acquisition required minimal
intratrial stimulus-intensity adjustment and resulted in a
high percentage of acceptable reflex cycles with no dif-
ferences between subjects with different ASIA SCI clas-
sifications and non-SCI subjects. Although 50 H reflex
cycles were collected during ambulation, the stability of
the data suggests that a reduced number of H reflex
cycles could be used without affecting data quality. The
growing development of robotic technologies is creating
new tools for the clinical rehabilitation and scientific
investigation of mechanisms of neural injury and SCI
repair. This methodology may be used to investigate
various research interests and provide new insights into
spinal cord rehabilitation.
ACKNOWLEDGMENTS
We would like to thank Nathan Foreman and James
Mosby for their efforts and contributions to this study.
Additionally, we would like to thank all the subjects whoThe main limitation to this methodology is that it was
specifically developed with the specialized Lokomat
volunteered to be part of this study for their patience,
time, and efforts.

185
QUERRY et al. Soleus H reflex during gait training and SCI
This material was based on work supported by the
Department of Veterans Affairs Rehabilitation Research
and Development Service (grant B4026I) and the Mobil-
ity Foundation Center at The University of Texas South-
western Medical Center, Dallas, Texas.
The authors have declared that no competing interests
exist.
REFERENCES
1. Capaday C. Neurophysiological methods for studies of the
motor system in freely moving human subjects. J Neurosci
Methods. 1997;74(2):201–18. [PMID: 9219889]
2. Capaday C, Stein RB. Amplitude modulation of the soleus
H-reflex in the human during walking and standing. J Neu-
rosci. 1986;6(5):1308–13. [PMID: 3711981]
3. Lavoie BA, Devanne H, Capaday C. Differential control of
reciprocal inhibition during walking versus postural and
voluntary motor tasks in humans. J Neurophysiol. 1997;
78(1):429–38. [PMID: 9242291]
4. Schneider C, Lavoie BA, Capaday C. On the origin of the
soleus H-reflex modulation pattern during human walking
and its task-dependent differences. J Neurophysiol. 2000;
83(5):2881–90. [PMID: 10805685]
5. Simonsen EB, Dyhre-Poulsen P. Amplitude of the human
soleus H reflex during walking and running. J Physiol.
1999;515(Pt 3):929–39. [PMID: 10066917]
6. Capaday C, Stein RB. Difference in the amplitude of the
human soleus H reflex during walking and running. J Phys-
iol. 1987;392:513–22. [PMID: 3446790]
7. Ferris DP, Aagaard P, Simonsen EB, Farley CT, Dyhre-
Poulsen P. Soleus H-reflex gain in humans walking and
running under simulated reduced gravity. J Physiol. 2001;
530(Pt 1):167–80. [PMID: 11136869]
8. Sinkjaer T, Andersen JB, Larsen B. Soleus stretch reflex
modulation during gait in humans. J Neurophysiol. 1996;
76(2):1112–20. [PMID: 8871224]
9. Hwang IS. Assessment of soleus motoneuronal excitability
using the joint angle dependent H reflex in humans. J Elec-
tromyogr Kinesiol. 2002;12(5):361–66. [PMID: 12223168]
10. Pinniger GJ, Nordlund M, Steele JR, Cresswell AG. H-
reflex modulation during passive lengthening and shorten-
ing of the human triceps surae. J Physiol. 2001;534(Pt 3):
913–23. [PMID: 11483720]
11. Knikou M, Rymer WZ. Hip angle induced modulation of H
reflex amplitude, latency and duration in spinal cord injured
12. Ethier C, Imbeault MA, Ung V, Capaday C. On the soleus
H-reflex modulation pattern during walking. Exp Brain
Res. 2003;151(3):420–25. [PMID: 12827331]
13. Calancie B, Alexeeva N, Broton JG, Molano MR. Interlimb
reflex activity after spinal cord injury in man: Strengthen-
ing response patterns are consistent with ongoing synaptic
plasticity. Clin Neurophysiol. 2005;116(1):75–86.
[PMID: 15589186]
14. Calancie B, Broton JG, Klose KJ, Traad M, Difini J, Ayyar
DR. Evidence that alterations in presynaptic inhibition con-
tribute to segmental hypo- and hyperexcitability after spinal
cord injury in man. Electroencephalogr Clin Neurophysiol.
1993;89(3):177–86. [PMID: 7686850]
15. Little JW, Halar EM. H-reflex changes following spinal
cord injury. Arch Phys Med Rehabil. 1985;66(1):19–22.
[PMID: 3966862]
16. Muller R, Dietz V. Neuronal function in chronic spinal cord
injury: Divergence between locomotor and flexion- and H-
reflex activity. Clin Neurophysiol. 2006;117(7):1499–1507.
[PMID: 16690351]
17. Schindler-Ivens SM, Shields RK. Soleus H-reflex recruit-
ment is not altered in persons with chronic spinal cord
injury. Arch Phys Med Rehabil. 2004;85(5):840–47.
[PMID: 15129411]
18. Andersen JB, Sinkjaer T. The stretch reflex and H-reflex of
the human soleus muscle during walking. Motor Control.
1999;3(2):151–57. [PMID: 10198147]
19. Faist M, Mazevet D, Dietz V, Pierrot-Deseilligny E. A
quantitative assessment of presynaptic inhibition of Ia
afferents in spastics. Differences in hemiplegics and para-
plegics. Brain. 1994;117(Pt 6):1449–55. [PMID: 7820579]
20. Fung J, Barbeau H. Effects of conditioning cutaneomuscu-
lar stimulation on the soleus H-reflex in normal and spastic
paretic subjects during walking and standing. J Neurophys-
iol. 1994;72(5):2090–2104. [PMID: 7884446]
21. Behrman AL, Lawless-Dixon AR, Davis SB, Bowden MG,
Nair P, Phadke C, Hannold EM, Plummer P, Harkema SJ.
Locomotor training progression and outcomes after incom-
plete spinal cord injury. Phys Ther. 2005;85(12):1356–71.
[PMID: 16305274]
22. Dietz V, Harkema SJ. Locomotor activity in spinal cord-
injured persons. J Appl Physiol. 2004;96(5):1954–60.
[PMID: 15075315]
23. Protas EJ, Holmes SA, Qureshy H, Johnson A, Lee D,
Sherwood AM. Supported treadmill ambulation training
after spinal cord injury: A pilot study. Arch Phys Med
Rehabil. 2001;82(6):825–31. [PMID: 11387590]
24. Wernig A, Muller S, Nanassy A, Cagol E. Laufband therapy
based on “rules of spinal locomotion” is effective in spinalhumans. Clin Neurophysiol. 2002;113(11):1698–1708.
[PMID: 12417222]
cord injured persons. Eur J Neurosci. 1995;7(4):823–29.
[PMID: 7620630]

186
JRRD, Volume 45, Number 1, 2008
25. Winchester P, McColl R, Querry R, Foreman N, Mosby J,
Tansey K, Williamson J. Changes in supraspinal activation
patterns following robotic locomotor therapy in motor-
incomplete spinal cord injury. Neurorehabil Neural Repair.
2005;19(4):313–24. [PMID: 16263963]
26. Colombo G, Joerg M, Schreier R, Dietz V. Treadmill train-
ing of paraplegic patients using a robotic orthosis. J Reha-
bil Res Dev. 2000;37(6):693–700. [PMID: 11321005]
27. Colombo G, Wirz M, Dietz V. Driven gait orthosis for
improvement of locomotor training in paraplegic patients.
Spinal Cord. 2001;39(5):252–55. [PMID: 11438840]
28. Pierrot-Deseilligny E, Mazevet D. The monosynaptic
reflex: A tool to investigate motor control in humans. Inter-
est and limits. Neurophysiol Clin. 2000;30(2):67–80.
[PMID: 10812576]
29. Knikou M, Rymer Z. Effects of changes in hip joint angle
on H-reflex excitability in humans. Exp Brain Res. 2002;
143(2):149–59. [PMID: 11880891]
30. Knikou M, Conway BA. Reflex effects of induced muscle
contraction in normal and spinal cord injured subjects.
Muscle Nerve. 2002;26(3):374–82. [PMID: 12210367]
31. Miyoshi T, Hotta K, Yamamoto S, Nakazawa K, Akai M.
Somatosensory graviception inhibits soleus H-reflex gain in
humans during walking. Exp Brain Res. 2006;169(1):135–38.
[PMID: 16365752]
32. Phadke CP, Wu SS, Thompson FJ, Behrman AL. Soleus H-
reflex modulation in response to change in percentage of
leg loading in standing after incomplete spinal cord injury.
Neurosci Lett. 2006;403(1–2):6–10. [PMID: 16723187]
33. Field-Fote EC, Brown KM, Lindley SD. Influence of pos-
ture and stimulus parameters on post-activation depression
of the soleus H-reflex in individuals with chronic spinal
cord injury. Neurosci Lett. 2006;410(1):37–41.
[PMID: 17046161]
Submitted for publication February 5, 2007. Accepted in
revised form July 24, 2007.