Muscular Responses and Movement Strategies During Stumbling
A. M. SCHILLINGS,1,2B.M.H. VAN WEZEL,1TH. MULDER,2,3AND J. DUYSENS1
1Department of Medical Physics and Biophysics, University of Nijmegen, 6525 EZ Nijmegen;2Sint Maartenskliniek
Research, 6500 GM Nijmegen; and3Institute of Neurology, University Hospital Nijmegen, 6500 HB Nijmegen,
Schillings, A. M., B.M.H. Van Wezel, Th. Mulder, and J. Duysens.
Muscular responses and movement strategies during stumbling over
obstacles. J. Neurophysiol. 83: 2093–2102, 2000. Although many
studies have investigated reflexes after stimulation of either cutaneous or
proprioceptive afferents, much less is known about responses after more
natural perturbations, such as stumbling over an obstacle. In particular,
the phase dependency of these responses and their relation to the stum-
bling behavior has received little attention. Hence response strategies
during stumbling reactions after perturbations at different times in the
swing phase of gait were studied. While subjects walked on a treadmill,
a rigid obstacle unexpectedly obstructed the forward sway of the foot. All
subjects showed an “elevating strategy” after early swing perturbations
and a “lowering strategy” after late swing perturbations. During the
elevating strategy, the foot was directly lifted over the obstacle through
extra knee flexion assisted by ipsilateral biceps femoris (iBF) responses
and ankle dorsiflexion assisted by tibialis anterior (iTA) responses. Later,
large rectus femoris (iRF) activations induced knee extension to place the
foot on the treadmill. During the lowering strategy, the foot was quickly
placed on the treadmill and was lifted over the obstacle in the subsequent
swing. Foot placement was actively controlled by iRF and iBF responses
related to knee extension and deceleration of the forward sway. Activa-
tions of iTA mostly preceded the main ipsilateral soleus (iSO) responses.
For both strategies, four response peaks could be distinguished with
latencies of ?40 ms (RP1), ?75 ms (RP2), ?110 ms (RP3), and ?160
in the step cycle. The phase-dependent modulation of the responses could
not be accounted for by differences in stimulation or in background
activity and therefore is assumed to be premotoneuronal in origin. In mid
swing, both the elevating and lowering strategy could occur. For this
phase, the responses of the two strategies could be compared in the
absence of phase-dependent response modulation. Both strategies had the
same initial electromyographic responses till ?100 ms (RP1-RP2) after
perturbation. The earliest response (RP1) is assumed to be a short-latency
stretch reflex evoked by the considerable impact of the collision, whereas
the second (RP2) has features reminiscent of cutaneous and propriocep-
tive responses. Both these responses did not determine the behavioral
response strategy. The functionally important response strategies de-
pended on later responses (RP3-RP4). These data suggest that during
stumbling reactions, as a first line of defense, the CNS releases a rela-
tively aspecific response, which is followed by an appropriate behavioral
response to avoid the obstacle.
I N T R O D U C T I O N
The pattern and timing of motor output during human loco-
motion are determined by a mixture of influences, some arising
from neural circuits entirely within the CNS and others arising
from a variety of sensory afferents. The electromyographic
(EMG) responses in leg muscles occurring after stimulation of
cutaneous and proprioceptive afferents during locomotion have
been described in many studies (see reviews Dietz 1992;
Duysens et al. 2000). The amplitudes of the responses to such
stimuli were dependent on the phase (or time) of stimulation in
the step cycle. For instance, electrical stimulation of the human
sural nerve yields facilitation of the ankle flexor muscle tibialis
anterior during early swing, but leads to suppression when
delivered during late swing (reflex reversal) (Duysens et al.
1990, 1992, 1996; Tax et al. 1995; Van Wezel et al. 1997;
Yang and Stein 1990).
It has been assumed that the phase-dependent response mod-
ulation adapts the responses in a functional way to the circum-
stances at various times in the step cycle. Previous studies have
suggested that the phase-dependent responses and the corre-
sponding joint angle changes following selective cutaneous
stimulation might be functionally relevant in stumbling reac-
tions (Van Wezel et al. 1997; Zehr et al. 1997). However, the
EMG and the accompanying kinesiologic responses occurring
after more realistic perturbations (e.g., stumbling over an ob-
ject, such as a doorstep or a paving stone) have not been
studied very extensively. Hence it is of importance to describe
the compensatory reactions during stumbling and to study the
functional significance of the observed responses.
To evoke natural stumbling reactions in an experimental
setting, mechanical perturbations were induced by an obstacle
suddenly rising above the ground and perturbing the forward
swinging foot of subjects walking on a walkway (Eng et al.
1994). When a perturbation occurred in early swing, an “ele-
vating strategy” was performed, during which the flexion an-
gles of the hip, knee, and ankle of the swinging leg increased
after the perturbation. In contrast, during late swing, mostly a
“lowering strategy” was performed, in which the foot of the
swinging leg was rapidly lowered to the ground causing a
shortened step length. The reflex responses in the leg muscles
during these recovery strategies had latencies varying from 60
to 140 ms.
It cannot be determined whether the responses described
above were mainly related to the phase of perturbation in the
step cycle or to the strategy performed. Hence a method was
developed in which the perturbations can be induced in all
parts of the swing phase, including mid swing, in which both
strategies could occur (Schillings et al. 1999a), allowing for a
comparison of the two strategies in the same phase. Perturba-
The costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked “advertisement”
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
20930022-3077/00 $5.00 Copyright © 2000 The American Physiological Society
tions are caused by an obstacle put on a treadmill, which
unexpectedly obstructs the forward swinging foot (Schillings et
al. 1996). Because of its weight (2.2 kg), the obstacle has a
considerable impact on the ongoing movement of the forward
swinging leg. A previous study showed that after perturbations
with this obstacle, short-latency stretch reflexes form a consis-
tent part of the stumbling reactions (Schillings et al. 1999b).
The aim of the present study is to describe the responses with
longer latencies and the coordination of leg muscle activity
compensating for this natural unexpected perturbation. The
questions whether these responses are dependent on the phase
of perturbation in the step cycle and/or whether the responses
are functionally related to the stumble strategy performed
(elevating or lowering) will be discussed.
M E T H O D S
Eight healthy subjects (5 male, 3 female) between 20 and 47 yr of
age (mean age 27) participated in the experiment. They had no known
history of neurological or motor disorder. The experiments were
carried out in conformity with the declaration of Helsinki for exper-
iments on humans. All subjects gave informed consent, and the study
was approved by the local ethical committee.
A detailed account of the experimental setup can be found in
Schillings et al. (1996). While subjects walked on a treadmill (speed,
4 km/h), an obstacle (length, width, and height, 40.0, 30.0, and 4.5 cm,
respectively; weight, 2.2 kg) was held by an electromagnet above the
treadmill in front of the subject (Fig. 1A). To induce perturbations, the
obstacle was dropped on the belt, thereby unexpectedly obstructing
the forward sway of the left (ipsilateral) leg. Release of the obstacle
occurred at a predetermined delay after ipsilateral or contralateral heel
strike. In the thin flexible shoes, the toes were covered with a piece of
cotton to protect them. A pressure-sensitive strip attached to the front
of the obstacle measured the time at which the foot hit the obstacle.
The subjects wore a pair of glasses, which blocked downward sight
(and thus blocked the view of the obstacle). Earplugs eliminated most
of the sound perception of the obstacle landing on the treadmill. In
addition, the sound was masked by music through headphones. Fur-
ther, to avoid that the subjects could feel the vibration of the obstacle
landing on the treadmill, a heavy metal object was put on the treadmill
at irregular intervals (imitating the landing of the obstacle). As a result
of these measures, subjects were not able to perceive the obstacle
before the collision with the foot. Subjects were instructed to keep the
same position on the treadmill before the perturbation, but after the
collision they were free to react without restrictions. The subjects
wore a safety harness, fixed to a safety brake on the ceiling that would
hold the subject and stop the treadmill in case a subject should start to
fall. In practice, this never occurred because none of the subjects
really started to fall. The harness was loosely suspended and did not
provide extra stability during the experiment.
Bipolar surface electromyogram (EMG) activity of the biceps fem-
oris (BF), rectus femoris (RF), tibialis anterior (TA), and soleus (SO)
of both legs was measured. Laterally placed goniometers were used to
measure the joint angles of the knee and ankle of the ipsilateral leg.
Thin insole foot switches measured foot contact with the treadmill.
Data were sampled in a time interval starting 100 ms before triggering
the electromagnet and lasting for 2,100 ms. For the control trials the
same intervals were sampled, but no obstacle was dropped after the
trigger. The EMG was (pre-)amplified, high-pass filtered (?3 Hz),
full-wave rectified, low-pass filtered (?300 Hz), AD-converted (500
Hz), and stored on hard disk along with the signals of the goniometers,
foot switches, and pressure-sensitive strip. In practice, this sampling
rate appeared to be sufficiently high. Increasing the sampling rate to
1,000 Hz did not lead to appreciable improvement of the signals for
the purpose of this study. In addition, the subjects were recorded on
video (25 Hz).
Each experiment consisted of three parts. Part one (5 min) consisted
of the registration of unperturbed walking. This control condition
enabled to check whether the presentation of the obstacles (in follow-
ing parts) affected baseline-walking characteristics (because of pos-
sible effects due to anticipation or fear of stumbling).
In part two (20 min), the effect of the timing of the perturbation on
the behavioral response “strategies” (elevating or lowering) was stud-
ied for a wide variety of delays after onset of swing. For this purpose
the computer triggered the electromagnet to drop the obstacle on the
treadmill after fixed delays (0, 40, 80, . . . , 600 ms) after heel strike.
Each delay condition was randomly applied only once. A perturba-
tion-free period of at least 10 s was taken between two succeeding
trials to be sure that the subject was walking normally again at the
time of the next perturbation. The normal walking pattern was usually
regained within approximately two step cycles. The behavioral re-
sponses were classified in two categories (Fig. 2) on the basis of video
analysis. A response was labeled as “elevating strategy” (Eng et al.
the obstacle above the treadmill in front of the subject’s left (ipsilateral) foot
(reprinted from Schillings et al. 1996, with permission from Elsevier Science).
B: example of an averaged subtracted electromyographic (EMG) response of
the tibialis anterior during stumbling (averaged stumbling response minus
averaged control EMG) showing the 4 response peaks (RP1-RP4). These 4
response peaks were not always clearly observable in every muscle during all
phases. Window settings, which were used to determine the mean response
amplitudes in this example, are indicated with gray. Zero time is the time the
foot collided with the obstacle.
Experimental methods. A: the electromagnet (black object) holds
2094 SCHILLINGS, VAN WEZEL, MULDER, AND DUYSENS
1994) when the ipsilateral foot was lifted over the obstacle during the
perturbed swing. The stumble response was classified as “lowering
strategy” when the foot was first placed on the treadmill and then
lifted over the obstacle.
In part three (30 min) stumbling reactions were repeatedly and
randomly introduced during early swing (5–25%, time of obstacle
contact with respect to control swing duration), mid swing (30–50%),
and late swing (55–75%) to construct averages. On average eight trials
(minimal 5 trials) were obtained for each phase of perturbation. The
responses during these perturbed cycles were compared with unper-
turbed control trials obtained in between the perturbation trials (per-
turbation-free period between trials ?10 s).
The stumble responses of each subject occurring in the same phase
of the step cycle were averaged. In addition, the corresponding control
trials were averaged. Then the averaged control activity was sub-
tracted from the averaged stumbling trials. To quantify the amplitudes
of the responses, the mean EMG activity was calculated in the period
between the beginning and end of the response. For this purpose,
windows were set around the individual response peaks occurring
within the first 200 ms (see Fig. 1B). To enable a proper intersubject
comparison of the response amplitudes, the resulting data of each
muscle were normalized with respect to the maximal EMG activity
during the control step cycles. The normalized responses of all sub-
jects were averaged. This type of analysis was performed on four
response peaks, namely RP1 (latency ?40 ms), RP2 (latency ?75
ms), RP3 (latency ?110 ms), and RP4 (latency ?160 ms, see Fig.
1B). The RP1 responses have already been described in a previous
publication (Schillings et al. 1999b) and are only included in the
present paper as a basis for comparison with the later responses. The
Wilcoxon matched-pairs signed-rank test was used to test whether the
response amplitudes during stumbling were significantly different
from the control EMG activity. The Friedman two-way ANOVA was
used to test whether the subtracted response amplitudes were different
for the three phases of perturbation (8 triples of comparison: 8
subjects, 3 phases; P ? 0.05). The Wilcoxon rank sum test was used
to compare the response amplitudes of the two strategies in mid swing
(P ? 0.05). The choice for nonparametric tests was based on the low
number of averages that were compared (8 subjects).
R E S U L T S
To check whether the normal walking pattern was affected
by the knowledge that a stumble over an obstacle could occur,
the normal walking pattern measured in between the stumble
trials (control trials of part 3) was compared with the normal
walking pattern measured during unperturbed walking (part 1,
see METHODS). None of the subjects clearly changed his/her
normal walking pattern during the stumble experiment.
Strategies in general
The choice for the behavioral strategy depended on the
timing of the perturbation in the step cycle. This is shown for
all subjects in Fig. 2C (data of 2nd part of the experiment).
When perturbations were caused in early swing (5–25%, time
of obstacle contact with respect to control swing duration), all
subjects showed the elevating strategy (see Fig. 2A). After
perturbations in late swing (55–75%), all subjects showed the
lowering strategy (see Fig. 2B). In mid swing (30–50%) both
strategies could occur. It can be seen that some subjects
showed a distinct transition in choice from elevating to low-
ering strategy (subjects 3–5, 7, and 8, Fig. 2C), whereas others
showed a zone of overlap of the two strategies (for example,
49–54% in subject 2; Fig. 2C). For each subject the transition
a function of perturbation onset. A: sche-
matic picture of the characteristics of the foot
trajectory during the elevating strategy. After
the collision with the obstacle, the foot was
directly lifted over the obstacle during the
perturbed swing. B: characteristics of the
foot trajectory during the lowering strategy.
After the perturbation, the ipsilateral foot
was quickly placed on the treadmill without
clearing the obstacle. The foot was lifted
over the obstacle in the swing phase that
succeeded the perturbed swing. C: strategies
performed as a function of perturbation onset
in all subjects (% of swing indicates time of
obstacle contact with respect to control
swing duration). These data are based on the
2nd part of the experiment (see METHODS).
Behavioral response strategies as
2095MUSCULAR RESPONSES DURING STUMBLING OVER OBSTACLES
point from the elevating to the lowering strategy was defined as
the time for one-half the interval between the last elevating and
the first lowering strategy occurrences. This transition point
varied for all subjects from 35% into the swing phase (subject
7, Fig. 2C) to 52% (subject 2, Fig. 2C) and was on average
44 ? 5% (mean ? SD, n ? 8 subjects).
Passive joint movements during both the elevating and the
During both the elevating and the lowering strategy, the
collision of the foot with the obstacle induced small passive
movements in the ipsilateral knee and ankle joint. These move-
ments are considered to be passive because they start before the
occurrence of the first EMG responses in the muscles that
could influence these joints. As seen in both the elevating
strategy of Fig. 3 and the lowering strategy of Fig. 4 the ankle
is first plantar flexed due to the collision (amplitude for all
subjects between 1 and 10°) with a latency of ?15 ms, whereas
the first responses in ipsilateral soleus and tibialis anterior (iSO
and iTA) occurred with a latency of ?40 ms. Second, the knee
was flexed due to the collision with a latency of 46 ms and an
amplitude of 13° during the elevating strategy (see Fig. 3) and
34 ms and 9° during the lowering strategy (see Fig. 4). In the
muscles ipsilateral biceps femoris and rectus femoris (iBF and
iRF), which could influence the knee joint angle, no responses
were observed before this early flexion.
Responses during the early swing elevating strategy
After the collision with the obstacle, active ipsilateral knee
and ankle flexion assisted the elevation of the swing foot to
step over the obstacle. The active knee flexion started ?160 ms
after perturbation onset (see sudden increase of the knee flex-
ion in Fig. 3; mean latency of active knee flexion for all
subjects was 169 ? 47 ms). The maximum knee flexion
reached during the elevating swing was considerably larger
than during normal swing (96 vs. 58° for the subject of Fig. 3).
The upper leg muscles typically showed first a large iBF burst
(latency, 64 ms; Fig. 3) assisting knee flexion, followed by a
large iRF burst (latency, 154 ms; Fig. 3) extending the knee
before touch down.
The ankle dorsiflexion started 90 ms after perturbation, and
the maximum dorsiflexion reached during the movement over
the obstacle was ?17° larger than the maximum dorsiflexion
during the unperturbed swing (Fig. 3). Facilitatory iTA re-
sponses (latency, ?75 ms) assisted this dorsiflexion. In two of
eight subjects, suppressive iTA responses (latency, ?80 ms;
suppressive response: the response EMG activity is lower than
the control EMG activity) preceded the main facilitatory re-
sponses in iTA (latency, 112 and 214 ms). This suppression
possibly had the function to allow for ankle plantar flexion to
avoid that the foot got hooked behind the obstacle. The change
in foot trajectory caused a lengthening of the ipsilateral swing
during the early swing elevating strategy (n ? 8 trials, subject 4). EMG
responses (mV) are shown for the ipsilateral biceps femoris (iBF), rectus
femoris (iRF), tibialis anterior (iTA), and soleus (iSO) as well as for the
contralateral biceps femoris (cBF). Joint angle changes (degrees; not sub-
tracted) are shown for the ipsilateral knee (iKnee) and ankle (iAnkle). Angle
at standing position is zero. Two bottom traces show stance phases of the
ipsilateral (iFoot) and contralateral foot (cFoot). Goniometer and foot signals:
—, stumble responses; – – –, control data (n ? 15). Zero time is the time the
foot collided with the obstacle. Ext., extension; Fl., flexion; Plant. Fl., plantar
flexion; Dorsi Fl., dorsiflexion.
Typical average subtracted EMG responses and joint angle changes
changes (degrees) during the late swing lowering strategy (n ? 10 trials,
subject 4). Two bottom traces: stance phases of the ipsilateral (iFoot) and
contralateral foot (cFoot). The same format is used as in Fig. 3.
Typical average subtracted EMG responses (mV) and joint angle
2096SCHILLINGS, VAN WEZEL, MULDER, AND DUYSENS
and concomitant of the contralateral stance with on average
128 ? 30 ms and 81 ? 41 ms, respectively (see Table 1).
Contralaterally, a large burst of activity in the cBF appeared
with a latency of 66 ms (subject of Fig. 3) after the perturba-
tion. Because, during stance the biarticular BF serves as a hip
extensor (Winter 1987), the cBF response could contribute to
stabilizing the upper body by the standing leg after perturba-
tions of the swinging leg, as was also suggested by other
authors (Dietz et al. 1986b; Eng et al. 1994). In the other
contralateral muscles (cRF, cTA, and cSO), the responses were
too variable or small to give a detailed description.
Responses during the late swing lowering strategy
During the lowering strategy, the foot was quickly placed on
the treadmill by shortening the forward sway and slightly
extending the knee. The knee extension needed to place the
foot was small in comparison with the knee extension before
touch down during normal walking (see Fig. 4). This was first,
because the foot was already close to the treadmill at the time
of the perturbation (?2–4 cm above the treadmill) and second
because the position of the foot during the landing (forefoot or
flat foot landing) was different from normal (heel landing). The
knee extension started 94 ms after perturbation and was pre-
sumably related to the large iRF burst (latency, 62 ms; see Fig.
4). For all subjects, the iRF burst occurred on average 53 ? 36
ms before the average foot placing in late swing. In three of
eight subjects, responses in iBF occurred approximately simul-
taneously with the responses in iRF (see Fig. 4). In the other
five subjects the onset of the main responses in the iBF oc-
curred ?25 ms later than the onset of the main iRF burst. The
iBF activity could slow down the forward swing in preparation
of the early foot placement (hip joint angles were not mea-
In the lower leg muscles, first short-latency responses oc-
curred with a latency of 40 ms in both iTA and iSO, possibly
transiently enhancing ankle-joint stiffness (see Schillings et al.
1999b). Subsequently, a large activity burst was observed in
iTA (latency, 66 ms after perturbation; Fig. 4), which could
participate in the ankle dorsiflexion (latency, ?90 ms; Fig. 4).
A large iSO burst appeared with a latency of 111 ms and was
well timed to take up body support during the preliminary
stance phase. This sequence of iTA and iSO responses was
observed in seven of eight subjects and could support an initial
movement away from the obstacle (iTA activity and ankle
dorsiflexion) followed by foot placement (iSO activity). The
mean latency of the premature placing off all subjects was
125 ? 35 ms after the collision. The ipsilateral swing phase
and the contralateral stance phase were shortened with, on
average, 35 ? 25 ms and 62 ? 29 ms (Table 1), respectively.
Contralaterally, the main consistent responses occurred in the
cBF, which showed a large response with a latency of 62 ms
(subject of Fig. 4).
Responses during mid swing
After mid swing perturbations (30–50% of swing), both
strategies could occur. For the mid swing elevating strategy,
the major characteristics were the same as for the early swing
elevating strategy (lengthening of swing phase duration, in-
creased knee flexion and ankle dorsiflexion, first iBF activation
then iRF activation, large iTA activity). However, some small
differences could be observed. For example, the duration of the
perturbed swing was lengthened in both phases, but the swing
phase duration increased on average 55 ? 42 ms more in mid
swing than in early swing (average of subjects 1–5, see Table
1). Comparing the mid swing lowering with the late swing
lowering strategy, it was found that the foot was placed later
after mid swing (mean latency, 246 ms; subjects 5–7) than after
late swing perturbations (mean latency, 125 ms; subjects 1–7).
The differences between responses in early and late swing
could be related to the strategy performed. However, some of
these differences may be related more to variations in the
timing of the perturbation within the step cycle (“phase depen-
dency”; see INTRODUCTION) than to changes in strategy. This
complication does not occur for some of the data related to mid
swing perturbations. Three subjects performed an elevating
strategy in one trial, whereas a lowering strategy was used in
another trial, despite the same timing of the perturbation.
Hence in these cases it was possible to study which parts of the
responses were strictly coupled to either lowering or elevating
strategy. In Fig. 5 the signals of an elevating (thin lines) and a
lowering strategy (heavy lines) occurring in the same phase are
Changes in ipsilateral swing and contralateral stance phase durations
Early Swing ElevatingMid Swing Elevating Mid Swing Lowering Late Swing Lowering
iSwingcStance iSwingcStanceiSwingcStance iSwingcStance
?81 ? 41
?62 ? 29
?90 ? 11
?128 ? 30
?172 ? 49
?70 ? 20
?1 ? 0
?35 ? 25
Values in Mean are the pooled averages of all subjects ? SD; number in parentheses is number of times the stumble strategy was observed in each subject.
Mean differences (in ms) between stumble and control trials (stumble minus control) of the ipsilateral swing-phase durations and the contralateral stance-phase
durations. The data of early, mid, and late swing perturbations are shown for all subjects separately and averaged. For the stumble trials the ipsilateral swing
(iSwing) is the swing in which the ipsilateral foot hits the obstacle, and the contralateral stance (cStance) is the corresponding stance phase of the contralateral
leg. After mid swing perturbations, subjects 1–5 repeatedly performed an elevating strategy; whereas subjects 6 and 7 performed a lowering strategy (data of
3rd part of the experiment). The number of trials was always between 5 and 10. * Based on 3 trials; † based on 1 trial because of the absence of some foot switch
data, which was also the reason for excluding subject 8 (n ? 9 trials for each phase).
2097 MUSCULAR RESPONSES DURING STUMBLING OVER OBSTACLES
The knee goniometer signals of both the elevating and the
lowering strategy started deviating from the control (– – –) in
the direction of knee flexion with a latency of 60 ms. This knee
flexion increased during both strategies till ?150 ms. After this
common movement, the knee flexed further to lift the foot over
the obstacle during the elevating strategy, whereas the knee
started to extend to place the foot on the ground during the
It was observed that the responses in the first 100 ms after
the perturbation were similar for both strategies (Fig. 5). In this
interval, the most obvious responses were observed in iTA
with a latency of ?70 ms in both the elevating and the
lowering strategy. In the period after 100 ms, the first differ-
ence between the two strategies occurred in iBF, namely after
104 ms (subject of Fig. 5). In this subject, the iBF (knee flexor)
showed a burst, which was followed by a burst in the iRF
(latency, 148 ms; Fig. 5) during the elevating strategy. In
contrast, during the lowering strategy the iRF was activated
(latency, 136 ms; Fig. 5) before the iBF. Similar results were
observed in two other subjects. In each case there was a
common initial movement in the knee (mean onset of differ-
ence in knee trajectory was 136 and 186 ms, respectively).
Correspondingly, these two subjects showed a common pattern
of EMG responses during the first 100 ms for the two strate-
Delayed lowering strategy
In the previous section it was shown that the EMG responses
during the first 100 ms did not determine the ensuing behav-
ioral response (strategy). Later EMG bursts (between 100 and
150 ms) were characteristic for the strategies, but the question
remains how predetermined these later responses were. If these
responses were completely defined from the moment of colli-
sion onward, a fixed response pattern could be expected and no
changes should occur during the course of this reaction. The
example shown in Fig. 6, however, illustrates that this is not
the case. In this exceptional trial, the subject started with an
elevating strategy after the early swing perturbation, but the
obstacle stuck to the toes and the subject was unable to clear
the obstacle. Instead, he extended the knee (latency, 236 ms)
and placed the foot on the treadmill (latency, 416 ms) without
grees) of the elevating (thin lines) vs. the lowering strategy (heavy lines) after
perturbations in mid swing for subject 5 (perturbation onset 41% of swing; data
from 3rd part of the experiment). For each muscle 2 separate traces show the
subtracted EMG responses for the 2 strategies. EMG calibration, 1 mV (the
calibration is the same for the 2 strategies). The joint angle changes (not
subtracted) of the elevating strategy (n ? 1), the lowering strategy (n ? 1), and
the averaged control (dashed line; n ? 15 trials) are superimposed. Angle at
standing position is zero. Calibration goniometer traces, 10°. Two bottom
traces show stance phases of the ipsilateral (iFoot) and contralateral foot
(cFoot). Zero time is the time the foot collided with the obstacle. Ext.,
extension; Fl., flexion; Plant. Fl., plantar flexion; Dorsi Fl., dorsiflexion.
Subtracted EMG responses (mV) and averaged joint angles (de-
grees) of the elevating (thin lines, n ? 1) vs. the delayed lowering strategy
(heavy lines, n ? 1) after perturbations in early swing for subject 6 (pertur-
bation onset 16 and 14% of swing, respectively; data from 3rd part of the
experiment). The same format is used as in Fig. 5.
Subtracted EMG responses (mV) and averaged joint angles (de-
2098 SCHILLINGS, VAN WEZEL, MULDER, AND DUYSENS
clearing the obstacle (“delayed lowering”). In the succeeding
swing phase the foot was lifted over the obstacle. In this
situation, in which the reaction was in fact a mixture of the two
strategies, the EMG responses of the two strategies (normal
elevating and delayed lowering) showed similar responses till
?120 ms after perturbation. From then on the iRF showed a
large response during the delayed lowering strategy that was
absent during the elevating strategy. This iRF burst occurred
116 ms before the onset of the knee extension, which resulted
in foot placement.
Modulation of the response amplitudes
In the EMG traces the main responses of the stumbling
reactions occurred with four peaks (RP1-RP4) within the
first 200 ms. To study the amplitudes of the EMG peaks,
time windows were set in these four periods, and the mean
EMG activity was calculated within these time windows
(see Table 2).
In Fig. 7, the amplitudes of the mean normalized control and
response activity of all subjects are shown during early swing
elevating (n ? 8 subjects), mid swing elevating (n ? 5), mid
swing lowering (n ? 3), and late swing lowering reactions
(n ? 8). Below the bars is indicated whether the pooled
average response amplitudes of all subjects were significantly
different from the average control activity (*) or not (E).
Representing the data of all subjects together (Fig. 7) en-
abled us to study phase and strategy dependency for the data of
all subjects. First, to study whether different response peaks are
phase dependent, the responses occurring in the three phases
[early swing (ES) vs. mid swing (MS) vs. late swing (LS), Fig.
7] were compared, irrespective of the strategy performed.
Almost all responses were phase-dependent as determined with
the Friedman two-way ANOVA (P ? 0.05). The only re-
sponses that did not show a significant phase effect were the
RP2 of iSO, RP3 and RP4 of iBF, and RP1 and RP2 of iRF.
In general for all muscles, the response amplitudes were not
strictly related to the background EMG activity (see Fig. 7).
For a clear example, compare the response amplitudes of the
iTA RP3 that showed largest responses in mid swing, when the
background activity was lowest, and the smallest responses in
late swing, when the background activity was largest.
Second, to study whether the responses are strategy de-
pendent, the responses occurring during the two strategies
[elevating (el) vs. lowering (lo), Fig. 7] in mid swing were
compared. As expected on the basis of the observation of
Fig. 5, no significant differences in the RP1 and RP2 am-
plitudes between the elevating and the lowering strategy in
mid swing were observed (Wilcoxon rank sum test). The
amplitudes of the late responses (RP4) in iSO were signif-
icantly related to the strategy. A general characteristic of the
late iSO responses in mid swing was that they were small
during the elevating strategy and large during the lowering
D I S C U S S I O N
The present study demonstrates that stumbling over a rigid
object on a treadmill induces a sequence of EMG and behav-
ioral responses. The elevating and lowering stumbling strate-
gies presently described show basic similarities with previ-
ously described stumbling reactions, which were induced by
more flexible obstacles (Eng et al. 1994). However, some
differences were observed. First, the short-latency stretch re-
flexes in the ipsilateral leg muscles were not present after
perturbations with the flexible metal strip (Eng et al. 1994;
Rietdyk and Patla 1998). The larger impact of the collision,
influenced by the weight and flexibility of the obstacle, in the
present study might be the cause of this difference. The occur-
rence of short-latency stretch reflexes in both flexors and
extensors can be understood because these responses are
caused by a sudden jar through the ipsilateral leg due to the
collision of the foot with the obstacle (see Schillings et al.
Second, in the late swing lowering strategy, the quick foot
placement after the perturbation was accompanied by iTA and
iRF facilitations (latency, ?75 ms). Eng et al. (1994) described
suppressive responses during the same period in these muscles,
and they speculated that this permitted gravity to assist the
lowering of the foot. Again, the observed difference might be
related to the characteristics of the perturbing obstacle. With
the presently used obstacle, subjects tried to avoid that they
would step on the obstacle. So they tended to move away from
the obstacle (iTA activity) before the quick touchdown assisted
by iRF. In the study of Eng et al. (1994), there was no chance
to step on the obstacle, because the flexible metal strip disap-
peared directly after the collision (the strip turned to a flat
position). Hence it was of less importance to actively control
the quick foot placement.
Third, Eng et al. (1994) described a “reaching strategy” after
late swing perturbations, whereas in the present study this
strategy was not observed. During this strategy, the foot is
directly lifted over the obstacle, primarily due to hip flexion
rather than knee flexion. Apparently the reaching strategy was
avoided in the present study because it requires considerably
more hip flexion to cross a long obstacle than a short one.
Mean onset and end of windows around the four response peaks
RP1 RP2 RP3 RP4
Number in parentheses is ?SD of the window onset. Mean onset and end of windows, which were set around the 4 response peaks (RP1–RP4) in the muscles
iSO, iTA, iBF, iRF, and cBF. These values are means of all subjects (n ? 8) in ms after perturbation. i, ipsilateral; SO, soleus; TA, tibialis; BF, biceps femoris;
RF, rectus femoris; c, contralateral.
2099MUSCULAR RESPONSES DURING STUMBLING OVER OBSTACLES
Phase dependency of the EMG responses
Differences in response amplitudes in the three phases
(early, mid, and late swing) could either be due to changes in
stimulation or to modulation of reflexes by the nervous system.
Changes in the intensity of the collision of the foot with the
obstacle could influence the amplitude of the EMG responses.
The impact of the collision on the foot is mainly dependent on
the horizontal toe velocity, which varies during the swing
phase. The horizontal toe velocity is ?1.5 times lower in early
swing than in either mid or late swing (Winter 1992). If the
amplitudes of the responses were completely determined by the
impact of the collision, one would always expect the smallest
responses in early swing. This was not the case. For example,
the RP2 response of iBF was largest in early swing and
smallest in late swing (see Fig. 7). Furthermore, one would
expect about equal amplitudes in mid and late swing. However,
different response amplitudes were often observed in these two
phases. Another possible cause for the amplitude modulation
observed could be that the responses were related to the back-
ground activity (Matthews 1986). However in general, there
was no strict relation between reflex amplitude and background
activity. This leaves the possibility that premotoneuronal
mechanisms might contribute to the phase-dependent modula-
tion of the response amplitudes.
Initial common reaction of the two strategies in mid swing
The occurrence of the two strategies after perturbations in
mid swing offered the unique possibility to study which parts
of the responses were strictly coupled to either the lowering or
the elevating strategy. Both strategies in mid swing started with
and cBF. Response amplitudes are shown with respect to the phase of perturbation and with respect to the behavioral strategy: early
swing elevating (ES el), mid swing elevating (MS el), mid swing lowering (MS lo), and late swing lowering (LS lo). Dark bars
show the normalized background activity (normalization with respect to the maximum background locomotor activity). Light bars
show the normalized EMG response amplitudes during the stumbling reactions. In this way, the part of the light bars above the dark
bars indicate the amplitude of the subtracted responses. In case the response was suppressive, the dark bar was larger than the light
bar (see, for example, iTA, LS lo, RP4). The SE of the control activity and the subtracted response activity is shown by error bars.
Number of trials per subject: 5–10 (see Table 1 between brackets). Averaged responses were calculated by averaging the mean
responses of all subjects: n ? 8 subjects for ES el and LS lo, n ? 5 for MS el, and n ? 3 for MS lo. *, pooled average response
amplitudes of all subjects were significantly different from the averaged control activity (Wilcoxon matched-pairs signed-rank test,
P ? 0.05). E, nonsignificant responses (P ? 0.05). The responses of the mid swing lowering (MS lo) were not analyzed with the
Wilcoxon matched-pairs signed-rank test (P ? 0.05) because the number of subjects was too small (n ? 3).
Mean normalized EMG amplitudes (average of all subjects) of the 4 response peaks (RP1-RP4) in iSO, iTA, iBF, iRF,
2100 SCHILLINGS, VAN WEZEL, MULDER, AND DUYSENS
the same EMG responses during the first 100 ms after the
perturbation. Neither the RP1 nor the RP2 responses predicted
the choice for one of the two strategies. Only the RP3 and RP4
responses (?100 ms) of some muscles (iBF and iSO) corre-
lated well with the behavioral strategy performed (e.g., see the
large iBF RP3 during the mid swing elevating strategy in Fig.
5, and see the large iSO RP4 during the mid swing lowering
strategy in Fig. 7). The initial common reaction of the two
strategies in mid swing first consists of the short-latency stretch
reflex (RP1), which might contribute to a temporary stiffening
of the joint (Schillings et al. 1999b). Second, it involves the
RP2 response in muscles such as iTA. The iTA RP2 activation
could contribute to the observed ankle dorsiflexion to move the
foot away from the obstacle. The same initial reaction of the
two strategies possibly provides the CNS sufficient time to
integrate information obtained by various sensory receptors
(Brooke et al. 1997; Dietz 1992; Jankowska 1992) and su-
praspinal sources (see review Dietz 1992) to make an appro-
priate decision about the final behavioral strategy.
Behavioral responses are not predetermined at the time
The occurrence of both behavioral strategies in mid swing
indicates that the decision about the final behavioral strategy is
not tightly linked to the time of impact. Further support for this
idea comes from the example of a lowering strategy performed
when a subject’s foot got hooked behind the obstacle during an
early swing perturbation. The foot was first lifted as in the
elevating strategy (initial reaction plus onset of the elevating
strategy), but because the obstacle stuck to the toes (continuing
mechanoreceptor feedback information), the subject decided to
place the foot on the ground again and finally completed a
delayed lowering strategy. Apparently, on-line afferent infor-
mation during the stumble response is integrated in the final
reaction and can be used to adjust the strategy to the demands
of the moment. In the example of the delayed lowering strat-
egy, the earliest adaptive response to extend the leg for the foot
placement on the treadmill was observed after 120 ms in the
iRF. This is too early to be a voluntary reaction, because the
earliest EMG changes during voluntary reactions occur after
?150 ms. For example, in a study of Hase and Stein (1998), in
which subjects were instructed to stop walking as soon as they
got a cue by electrical stimulation of the superficial peroneal
nerve, the earliest voluntary changes in EMG activity of leg
muscles occurred 150–200 ms after stimulation.
The idea that a corrective response can be adjusted en route
has found support in some earlier observations as well. While
subjects were walking on a treadmill, Dietz et al. (1986a)
applied a holding impulse by a cord attached to the swinging
leg, which was followed by a second perturbation, i.e., a
treadmill deceleration. On the basis of their results, they sug-
gested that the first part of the compensatory reaction is re-
leased as an immutable pattern within the spinal cord. In
contrast, the later part of the response (in the order of 120 ms
after the 1st perturbation, see their Fig. 2) can be modified by
external factors and adjusted to the actual nature of the task.
Furthermore, the present data are in line with the finding of
bimodal responses in subjects who were tripped while they
were taking a single step forward (Rietdyk and Patla 1998). In
this situation, the perturbation always induced an initial change
in ankle trajectory (elevation of the ankle), which could or
could not be followed by a second elevation of the ankle. Thus
the initial response could be followed by a later correction,
which resulted in an enhancement of the initial movement. In
the present study it was demonstrated that the later correction
is not always an enhancement of the first movement as ob-
served in the study of Rietdyk and Patla (1998), but can also be
a reversal of the first movement.
Origin of the responses
On the basis of several studies on cats, it has been assumed
that the responses observed during “stumbling corrective reac-
tions” (Forssberg 1979) are mainly cutaneous in origin (Forss-
berg 1979; Forssberg et al. 1975, 1977; Prochazka et al. 1978;
Wand et al. 1980). In addition, for humans it has been sug-
gested that the medium-latency EMG responses and the ac-
companying joint angle changes after electrical cutaneous
stimulation might be functionally relevant in the context of
stumbling reactions (Van Wezel et al. 1997; Zehr et al. 1997;
see, however, Duysens et al. 1992). There are indeed some
similarities between the modulation of medium-latency cuta-
neous responses and the RP2 responses observed during stum-
bling. For example, the iTA facilitation (RP2) with ankle
dorsiflexion observed during the early swing elevating strategy
was also observed after sural nerve stimulation in early swing
(Duysens et al. 1992; Van Wezel et al. 1997). However,
cutaneous stimulation evoked suppression of the iTA activity
in late swing (Duysens et al. 1990; Yang and Stein 1990; Zehr
et al. 1997), whereas the present mechanical perturbation
evoked facilitatory iTA (RP2) responses. Hence these differ-
ences indicate that the RP2 observed during stumbling in
humans cannot be fully attributed to cutaneous responses (al-
though we cannot rule out the possibility that cutaneous re-
sponses after mechanical perturbations are different from cu-
taneous responses after electrical stimulation).
Alternatively, proprioceptive afferents might contribute to
the RP2 responses observed. Medium-latency stretch responses
(M2 or MLR) in leg muscles with latencies similar to the RP2
responses (?75 ms) have been described by many authors after
joint rotations during various conditions (Fellows et al. 1993;
Nielsen et al. 1998; Schieppati and Nardone 1997; Schieppati
et al. 1995; Sinkjaer et al. 1988; Toft et al. 1989). Although
some authors suggested that Ia afferents could mediate me-
dium-latency stretch reflexes in leg muscles (Berardelli et al.
1982; Fellows et al. 1993), most evidence points in the direc-
tion of a contribution of muscle proprioceptive group II affer-
ents in these responses (see Corna et al. 1995; Dietz 1992;
Nardone et al. 1996; Nielsen et al. 1998; Schieppati and Nar-
done 1997; Schieppati et al. 1995). Even activations of Ib
afferents cannot be excluded because of the strong impact of
the obstacle (for review see Duysens et al. 2000). The contri-
bution of vestibular afferents in the RP2 responses during
stumbling is probably small because vestibular responses have
much smaller amplitudes than somatosensory responses
(Horstmann and Dietz 1988).
For the short-latency responses, there is little doubt that
these are spinal stretch reflexes mediated by Ia afferents (see
Schillings et al. 1999b). The responses with longer latencies
(RP2-RP3) during stumbling could follow both spinal or su-
praspinal neural pathways. In principle, these responses might
2101MUSCULAR RESPONSES DURING STUMBLING OVER OBSTACLES
be polysynaptic EMG responses of spinal origin and could be Download full-text
related to activation of slower conducting afferents (see review
Dietz 1992). However, the latency of these responses is also
long enough to be compatible with long-loop reflexes through
the cortex (Christensen et al. 1999; Nielsen et al. 1997; Pe-
tersen et al. 1998; Pijnappels et al. 1998). Although the same
neural pathways could contribute to the RP4 responses, these
responses are likely to be at least partly under voluntary control
because they have latencies above the voluntary reaction time
of ?150 ms in leg muscles.
We thank P.H.J.A. Nieuwenhuijzen and H.W.A.A. Van de Crommert for
help with the experiments and the analysis software. We also acknowledge G.
Windau for developing the software and A. M. Van Dreumel and J.W.C.
Kleijnen for technical assistance.
This study was supported by the Dutch Science Foundation (NWO).
Address for reprint requests: A. M. Schillings, Dept. of Medical Physics and
Biophysics, University of Nijmegen, P.O. Box 9101, NL-6500 HB Nijmegen,
Received 9 August 1999; accepted in final form 13 December 1999.
BERARDELLI, A., HALLETT, M., KAUFMAN, C., FINE, E., BERENBERG, W., AND
SIMON, S. R. Stretch reflexes of triceps surae in normal man. J. Neurol.
Neurosurg. Psychiatry 45: 513–525, 1982.
BROOKE, J. D., CHENG, J., COLLINS, D. F., MCILROY, W. E., MISIASZEK, J. E.,
AND STAINES, W. R. Sensori-sensory afferent conditioning with leg move-
ment: gain control in spinal reflex and ascending paths. Prog. Neurobiol. 51:
CHRISTENSEN, L. O., MORITA, H., PETERSEN, N., AND NIELSEN, J. Evidence
suggesting that a transcortical reflex pathway contributes to cutaneous
reflexes in the tibialis anterior muscle during walking in man. Exp. Brain
Res. 124: 59–68, 1999.
CORNA, S., GRASSO, M., NARDONE, A., AND SCHIEPPATI, M. Selective depres-
sion of medium-latency leg and foot muscle responses to stretch by an
?2-agonist in humans. J. Physiol. (Lond.) 484: 803–809, 1995.
DIETZ, V. Human neuronal control of automatic functional movements: inter-
action between central programs and afferent input. Physiol. Rev. 72: 33–69,
DIETZ, V., QUINTERN, J., AND BERGER, W. Stumbling reactions in man: release
of a ballistic movement pattern. Brain Res. 362: 355–357, 1986a.
DIETZ, V., QUINTERN, J., BOOS, G., AND BERGER, W. Obstruction of the swing
phase during gait: phase-dependent bilateral leg muscle coordination. Brain
Res. 384: 166–169, 1986b.
DUYSENS, J., CLARAC, F., AND CRUSE, H. Load regulating mechanisms in gait
and posture: comparative aspects. Physiol. Rev. 80: 83–133, 2000.
DUYSENS, J., TAX, A.A.M., MURRER, L., AND DIETZ, V. Backward and forward
walking use different patterns of phase-dependent modulation of cutaneous
reflexes in humans. J. Neurophysiol. 76: 301–310, 1996.
DUYSENS, J., TAX, A.A.M., TRIPPEL, M., AND DIETZ, V. Phase-dependent
reversal of reflexly induced movements during human gait. Exp. Brain Res.
90: 404–414, 1992.
DUYSENS, J., TRIPPEL, M., HORSTMANN, G. A., AND DIETZ, V. Gating and
reversal of reflexes in ankle muscles during human walking. Exp. Brain Res.
82: 351–358, 1990.
ENG, J. J., WINTER, D. A., AND PATLA, A. E. Strategies for recovery from a trip
in early and late swing during human walking. Exp. Brain Res. 102:
FELLOWS, S. J., D¨ OMGES, F., T¨ OPPER, R., THILMANN, A. F., AND NOTH, J.
Changes in the short- and long-latency stretch reflex components of the
triceps surae muscle during ischaemia in man. J. Physiol. (Lond.) 472:
FORSSBERG, H. Stumbling corrective reaction: a phase-dependent compensa-
tory reaction during locomotion. J. Neurophysiol. 42: 936–953, 1979.
FORSSBERG, H., GRILLNER, S., AND ROSSIGNOL, S. Phase dependent reflex
reversal during walking in chronic spinal cats. Brain Res. 85: 103–107,
FORSSBERG, H., GRILLNER, S., AND ROSSIGNOL, S. Phasic gain control of
reflexes from the dorsum of the paw during spinal locomotion. Brain Res.
132: 121–139, 1977.
HASE, K. AND STEIN, R. B. Analysis of rapid stopping during human walking.
J. Neurophysiol. 80: 255–261, 1998.
HORSTMANN, G. A. AND DIETZ, V. The contribution of vestibular input to the
stabilization of human posture: a new experimental approach. Neurosci.
Lett. 95: 179–184, 1988.
JANKOWSKA, E. Interneuronal relay in spinal pathways from proprioceptors.
Prog. Neurobiol. 38: 335–378, 1992.
MATTHEWS, P.B.C. Observations on the automatic compensation of reflex gain
on varying the pre-existing level of motor discharge in man. J. Physiol.
(Lond.) 374: 73–90, 1986.
NARDONE, A., GRASSO, M., GIORDANO, A., AND SCHIEPPATI, M. Different effect
of height on latency of leg and foot short- and medium-latency EMG
responses to perturbation of stance in humans. Neurosci. Lett. 206: 89–92,
NIELSEN, J., PETERSEN, N., AND FEDIRCHUK, B. Evidence suggesting a transcor-
tical pathway from cutaneous foot afferents to tibialis anterior motoneurones
in man. J. Physiol. (Lond.) 501: 473–484, 1997.
NIELSEN, J., SINKJAER, T., BAUMGARTEN, J., ANDERSEN, J. B., TOFT, E., CHRIS-
TENSEN, L.O.D., LADOCEUR, M., AND MORITA, H. Modulation of the soleus
stretch reflex and H-reflex during human walking after block of peripheral
feedback. Soc. Neurosci. Abstr. 24: 2103, 1998.
PETERSEN, N., CHRISTENSEN, L.O.D., MORITA, H., SINKJAER, T., AND
NIELSEN, J. Evidence that a transcortical pathway contributes to stretch
reflexes in the tibialis anterior muscle in man. J. Physiol. (Lond.) 512:
PIJNAPPELS, M., VAN WEZEL, B.M.H., COLOMBO, G., DIETZ, V., AND DUYSENS,
J. Cortical facilitation of cutaneous reflexes in leg muscles during human
gait. Brain Res. 787: 149–153, 1998.
PROCHAZKA, A., SONTAG, K. H., AND WAND, P. Motor reactions to perturba-
tions of gait: proprioceptive and somesthetic involvement. Neurosci. Lett. 7:
RIETDYK, S. AND PATLA, A. E. Context-dependent reflex control: some insights
into the role of balance. Exp. Brain Res. 119: 251–259, 1998.
SCHIEPPATI, M. AND NARDONE, A. Medium-latency stretch reflexes of foot and
leg muscles analysed by cooling the lower limb in standing humans.
J. Physiol. (Lond.) 503: 691–698, 1997.
SCHIEPPATI, M., NARDONE, A., SILIOTTO, R., AND GRASSO, M. Early and late
stretch responses of human foot muscles induced by perturbation of stance.
Exp. Brain Res. 105: 411–422, 1995.
SCHILLINGS, A. M., VAN WEZEL, B.M.H., AND DUYSENS, J. Mechanically
induced stumbling during human treadmill walking. J. Neurosci. Methods
67: 11–17, 1996.
SCHILLINGS, A. M., VAN WEZEL, B.M.H., MULDER, TH., AND DUYSENS, J.
Stumbling and stepping over obstacles during human treadmill walking in
young and older adults (Abstract). Gait Pos. 9: S30, 1999a.
SCHILLINGS, A. M., VAN WEZEL, B.M.H., MULDER, TH., AND DUYSENS, J.
Widespread short-latency stretch reflexes and their modulation during stum-
bling over obstacles. Brain Res. 816: 480–486, 1999b.
SINKJAER, T., TOFT, E., ANDREASSEN, S., AND HORNEMANN, B. C. Muscle
stiffness in human ankle dorsiflexors: intrinsic and reflex components.
J. Neurophysiol. 60: 1110–1121, 1988.
TAX, A.A.M., VAN WEZEL, B.M.H., AND DIETZ, V. Bipedal reflex coordination
to tactile stimulation of the sural nerve during human running. J. Neuro-
physiol. 73: 1947–1964, 1995.
TOFT, E., SINKJAER, T., AND ANDREASSEN, S. Mechanical and electromyo-
graphic responses to stretch of the human anterior tibial muscle at different
levels of contraction. Exp. Brain Res. 74: 213–219, 1989.
VAN WEZEL, B.M.H., OTTENHOFF, F.A.M., AND DUYSENS, J. Dynamic control
of location-specific information in tactile cutaneous reflexes from the foot
during human walking. J. Neurosci. 17: 3804–3814, 1997.
WAND, P., PROCHAZKA, A., AND SONTAG, K. H. Neuromuscular responses to
gait perturbations in freely moving cats. Exp. Brain Res. 38: 109–114, 1980.
WINTER, D. A. The Biomechanics and Motor Control of Human Gait. Water-
loo, Ontario: Univ. of Waterloo Press, 1987.
WINTER, D. A. Foot trajectory in human gait: a precise and multifactorial motor
control task. Phys. Ther. 72: 45–56, 1992.
YANG, J. F. AND STEIN, R. B. Phase-dependent reflex reversal in human leg
muscles during walking. J. Neurophysiol. 63: 1109–1117, 1990.
ZEHR, E. P., KOMIYAMA, T., AND STEIN, R. B. Cutaneous reflexes during human
gait: electromyographic and kinematic responses to electrical stimulation.
J. Neurophysiol. 77: 3311–3325, 1997.
2102SCHILLINGS, VAN WEZEL, MULDER, AND DUYSENS