Thehttp://nro.sagepub.

com/
Neuroscientist

How we Walk: Central Control of Muscle Activity during Human Walking

Jens Bo Nielsen
Neuroscientist 2003 9: 195
DOI: 10.1177/1073858403009003012
The online version of this article can be found at:
http://nro.sagepub.com/content/9/3/195

Published by:
http://www.sagepublications.com

Additional services and information for The Neuroscientist can be found at:
Email Alerts: http://nro.sagepub.com/cgi/alerts
Subscriptions: http://nro.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://nro.sagepub.com/content/9/3/195.refs.html

>> Version of Record - Jun 1, 2003

What is This?

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

REVIEW

How We Walk: Central Control of

Muscle Activity during Human Walking
JENS BO NIELSEN

Recent advances in noninvasive electrophysiological and brain imaging techniques have made investigation of the central control of human walking possible. We are thus now able to ask in what way the motor
control circuitries in the human brain and spinal cord have been modified in order to control bipedal walking. This information is of importance not only for our understanding of basic control strategies and paradigms but also for future attempts at rehabilitating the gait ability of patients after lesions of the brain and
spinal cord. NEUROSCIENTIST 9(3):195204, 2003. DOI: 10.1177/1073858403251978
KEY WORDS Human, Walking

Human walking is a very complex task that requires the

coordination of a number of different muscles acting on
different joints. The muscle activities tend to be only a
little coupled with each other leaving significant room
for flexibility and adaptability in the gait pattern. This
flexibility is the result of the integrated activity of spinal
neuronal circuitries, sensory feedback signals, and
descending supraspinal motor commands.
The available evidence suggests that humanslike
other animalspossess a network in the spinal cord,
which is capable of generating the basic rhythmic walking activity. However, the activity of this network
depends much more on supraspinal influences than in
other animals, and it has been importantly modified to
meet the functional requirements of bipedal walking.
Some corrective reflexes, which in the cat are integrated
at a spinal level, tend to be mainly integrated at a
supraspinal level in man and to depend crucially on
intact corticospinal transmission. In primates and man, a
direct monosynaptic corticospinal pathway from the
motor cortex to spinal motoneurons has developed. This
pathway contributes significantly to the muscle activity
during walking.
Although there are many similarities in the central
control of walking in man and other animals, there are
thus also important differences. It is these differences
that make human bipedal walking unique and that we
need to understand better to provide a sound basis for
future attempts at rehabilitating gait in patients with
lesions of the central nervous system.
We are separated from our closest primate relatives,
the chimpanzees, by around 5 to 8 million years according to current estimates (Gibbons 2002). The first memDivision of Neurophysiology, Department of Medical Physiology, The
Panum Institute, University of Copenhagen, Denmark.
Address correspondence to: Professor Jens Bo Nielsen, Division of
Neurophysiology, Department of Medical Physiology, The Panum
Institute, Copenhagen University, Blegdamsvej 3, 2200 Copenhagen N
(e-mail: j.b.nielsen@mfi.ku.dk).

bers of the hominid family did not have much larger

brains than the chimpanzees, their use of the hand was
not significantly more sophisticated, and they did not yet
possess the skill of language. What initially separated
themand usfrom other primates was the development of a habitual bipedal walking behavior. Bipedalism
thus stands out as the initiating factor in the evolution of
the hominid lineage and eventually the evolution of
modern man. We have little way of knowing which
advantages habitual bipedalism provided for the early
members of the human family, but the most widely
accepted theory is still that freeing of the hands from
locomotion tasks enabled the hands to be used to carry
food and children, to develop tools, and in other ways
manipulate the environment. This appears to have been a
continually important factor in the success of the
hominid lineage.
So humans walk on two legs, but even the species
with which we are most closely related prefer to walk on
all four. This creates some obvious problems when looking for an animal model of human walking. Most of our
understanding of how walking and other types of locomotion are controlled is based on experiments in the cat
and other vertebrates with which we are even more distantly related; evolution separates us from the cat by
probably more than 60 million years. This long span of
time, during which our ancestors first adapted to the
demands of an arboreal life (at the point of separation
between the lines leading to modern primates and modern cats) and then later to a terrestrial life and bipedal
gait (at the point of separation of hominids from other
primates), must be sufficient for quite pronounced neural adaptations to have taken place. The brain possesses
a remarkable plasticity and flexibility and has evolved in
a very remarkable way in human subjects as compared
with other species, including the cat. Should we assume
that it is only the higher cognitive capabilities that have
diverged so remarkably and that 60 million years of separate evolution have not resulted in some evolutionary
changes in the motor centers of the brain that control

Volume 9, Number 3, 2003

Copyright 2003 Sage Publications
ISSN 1073-8584

THE NEUROSCIENTIST

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

195

walking? This seems unlikely and underscores more

than anything the necessity of studying human bipedalism in the only species in which this is possible: Man.
Within recent years, the development of noninvasive
electrophysiological and brain imaging techniques have
allowed us to approach the neural control of walking in
man. This makes it possible for the first time to gain
direct information about the way we all walkin what
ways we are similar toand in what ways we are different from other animals. In this review, I hope to demonstrate that many basic principles of neural control of
locomotion are remarkably preserved during evolution
and essentially comparable across species. However, I
will also stress that man is not simply a cat or monkey
walking on two legs but that a very considerable adaptation and refinement of the motor control centers
involved in human bipedal walking has taken place.
Kinematics of Human Bipedal Walking
Bipedal walking is in itself not unique in the animal
kingdom. Birds walk on two legs with great ease and so
do many lizards, to name just two examples. Many monkeys intermittently walk on their hind limbs, and dogs
and elephants can be trained to do so as well. However,
what is unique is the very special way of bipedal walking that has evolved in modern humans. The characteristic features of this walking pattern is the lean body
placed above the unstable support of two long legs, the
heel strike in the early part of the stance phase, the
lengthening contraction of the ankle dorsiflexors in the
early stance phase, the lengthening contraction of the
ankle plantarflexors throughout most of the stance
phase, the controlled forward shift of the center of body
mass by the propulsive power mainly of the ankle plantar flexors, and the subsequent controlled fall of the
body, which is only stopped by the initiation of the next
stance phase (Fig. 1). These features as well as many
others make human gait distinct and unique in the animal kingdom.
Musculoskeletal Adaptations
to Bipedal Walking
This unique pattern of walking is surprisingly effectivein terms of energy consumption easily comparable
to quadrupedal animalsand in any case far more effective than bipedal walking in any of our primate relatives
(Rodman and McHenry 1980; Abitbol 1988). To obtain
this efficiency, a number of musculoskeletal adaptations
have taken place during evolution. The skeletal changes
are evident from comparison of monkey and human
skeletons. The most apparent of these changes are 1)
repositioning of the foramen magnum, so that the head
is balanced on the spine; 2) curving of the spine to keep
the trunk and weight centered above the pelvis and to
absorb force when the feet strike the ground; 3) change
of the angle between the pelvis and the spine so that the
legs and spine are in line with each other; 4) changes in
the pelvis to support the internal organs and to make

196

efficient attachments for the muscles that ensure the

upright position; 5) elongation of the lower limb; 6)
inward angling of the femur so that the legs are positioned more directly under the body; 7) modification of
the knee anatomy to allow full extension; and finally 8)
modifications of the feet (big toe is enlarged and brought
in line with the other toes, development of the longitudinal arch) to support the weight of the body and to ensure
an effective ground reaction force in the stance phase of
walking (Jurmain and others 1997).
The muscular changes are less apparent and are mainly revealed when analyzing the muscle activities during
walking. To mention a few major changes: The largest
muscle in the human body is the gluteus maximus. In
monkeys this muscle is quite small and is involved in hip
extension, but in man it has gained its large size to prevent the body from moving forward when the foot strikes
the ground. Similarly, the abductor muscles (m. gluteus
medius and minimus) have been modified to prevent tilt
of the body when the opposite leg is lifted from the
ground. Characteristically, many extensors and flexors
are activated at the time of heel strike to help ensure the
upright position and that the leg is fully extended at the
critical moment when all of the body weight is shifted
onto that leg. During human walking, the ankle extensors play a pivotal role for the forward propulsion in the
late stance phase, whereas hip, knee, and ankle extensors
contribute importantly during quadrupedal animal walking.
Human Walking Is a Complex Task Requiring
Coordination of Many Muscles
The task of human walking is an immensely complicated one. To take just one example: It is essential that the
distal part of the foot in the swing phase is lifted sufficiently above the ground to prevent stumbling, but at the
same time not more than necessary. In fact, during normal human walking, the foot is elevated by only 1 to 2
cm above the ground and the position of the foot varies
by less than 4 mm from step to step (Winter 1992). This
is a quite remarkable precision when considering that the
position of the foot in the swing phase is determined by
the position of five individual joints and that 15 muscles
act on the knee joint alone (Winter and Eng 1995). The
activity of all these muscles has to be scaled exactly
according to each other in order to obtain a given position of the foot, and as can be realized, the number of
different combinations of muscle activities, which could
lead to the same position of the foot, is almost infinite.
How does the brain solve this phenomenal computing
task? The answer to this was provided already by the
Russian scientist Bernstein (1967): He suggested that it
may be a fundamental strategy of the brain to reduce the
number of degrees of freedom and only control the position of a joint, leaving a considerable flexibility for specific muscle activities. The individual muscle activities
are inessential as long as they are all scaled to each other
so that the end-point, the position of the foot, is within
the desired range. This flexibility in muscle activation is

THE NEUROSCIENTIST

How We Walk

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

Fig. 1. Muscle coordination during human walking. A, the position of the ankle, knee, and hip joints as well as the rectified EMG activity from the ankle dorsi- and plantarflexors during the human gait cycle. The times of the stance and swing phases are indicated above
the recordings. A stick diagram at the top indicates the various joint and limb positions during the gait cycle. The recorded leg is indicated in green. B, the principles of the cross-correlation for the study of motor unit coupling during human walking. If two motoneurons receive a common drive from branches of last-order neurons, they will tend to discharge synchronously; that is, when an excitatory postsynaptic potential (EPSP) is elicited from the common last-order neurons, it will simultaneously bring the two motoneurons
closer to their firing thresholdthey will therefore tend to discharge simultaneously when this common EPSP arrives. This may be
visualized by measuring how often one of the two motor units discharges in relation to the discharges of the other motor unit. In this
way, a cross-correlogram showing the proportion of discharges of one motor unit at different times in relation to the discharges of
another motor unit may be constructed. A short-lasting central peak indicates that the two peaks have a tendency to discharge synchronously. This short-term synchrony has a duration of 5 to 15 ms. C, cross-correlation has been calculated for two wire recordings
of motor unit activity in the tibialis anterior muscle (an ankle dorsiflexor muscle) in the swing phase during walking. The short-lasting
peak of synchrony indicates that the motor units receive a common synaptic drive from central last-order neurones.

in fact also what is seen when comparing gait in different subjects and individual steps on different occasions
in the same subject (Winter 1992; Winter and Eng 1995).
This indicates that the neuronal networks in the central nervous system, which are responsible for the generation and control of the muscle activity during walking,
must be organized to ensure that the overall activity of
the muscles is scaled correctly, yet at the same time that
considerable flexibility of the individual muscle activities is allowed. This is also what was found in a recent
study by Hansen and others (2001). They analyzed the
coupling between paired recordings of motor unit activity from the same or different leg muscles when subjects
walked on a treadmill. Motor units recorded within a
muscle or from close synergists were found to have a
significant tendency to discharge within a few milliseconds of each other (Fig. 1). This short-term synchrony is
thought to be caused by a common synaptic drive to the
motoneurons from branches of last-order neurons (Sears

and Stagg 1976; Kirkwood and Sears 1978; Datta and

Stephens 1990). The presence and size of this short-term
synchrony thus reflects the extent to which the synaptic
drive to the motor units originates from a common
source in the central nervous system. Importantly,
Hansen and others (2001) never found this type of coupling between muscles acting on different joints, suggesting that the common last-order drive of muscles during walking is restricted to motor units within the same
muscle and close synergists. This indicates that the scaling of muscle activities in order to ensure the correct
movement trajectory is not performed at the last-order
stage, thereby allowing a considerable flexibility in the
activity pattern of individual muscles.
From a functional point of view, this flexibility is also
desirable when considering that human subjects have to
walk on surfaces that may be more or less even (stepping
on stones and tree branches), have different compliances
(from sand to rock), and have different slopes (uphill-

Volume 9, Number 3, 2003

THE NEUROSCIENTIST

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

197

downhill). A too rigid coupling of motor unit activities

would make the necessary adjustments of the walking
pattern and muscle activities impossible.
It is the task of the whole central nervous system
(CNS) to generate this muscle activity, to ensure that it
is optimally coordinated, to ensure that it is adjusted to
the immediate environment, and to modify it when
required. Several different control systems at different
levels of the CNS contribute to this control. In the following sections, I will focus on three control systems as
outlined in Figure 2: The possible existence of a rhythmgenerating network in the spinal cord (a spinal stepping
generator), the role of sensory feedback in the control of
walking, and finally the role of descending control from
the motor cortex.
Do We Have a Spinal Rhythm Generator?
It was suggested already by Graham Brown (1911) that
the cat spinal cord possesses a network that is capable of
eliciting stepping movements. He demonstrated that
alternating activity in antagonistic muscles could be
evoked in the chronic spinalized animal even after
deafferentation, which demonstrates that the rhythmicity
must have originated in the spinal cord. However, the
main body of evidence on the spinal control of animal
walking has come from work within the past 30 years
from several different research groups (Shik and
Orlovsky 1976; Lundberg 1979; Grillner 1985; Jordan
1998; Rossignol 2000). This work has not only confirmed that the spinal network has the capacity of generating the basic locomotor rhythmicity in the absence of
any supraspinal or sensory input to the spinal cord in the
chronic spinalized animal, but also that the network may
be activated in the decerebrated animal by tonic
descending activation from centers in the brain stem
(mesencephalic and subthalamic locomotor regions;
MLR and SLR) or by application (intravenously or
directly on the cord) of a number of different monoaminergic neurotransmitters. This suggests that the spinal
rhythm generating network is in all likelihood activated
in the intact animal by a nonpatterned and nonspecific
descending drive from the brain stem and that the network is in itself capable of recruiting muscles in the correct sequence to obtain a coordinated walking pattern.
The main question of interest in relation to this review
is whether a similar spinal network exists in man and
whether it is used in the same way as in other species.
These questions are pertinent, in that humans are
much more debilitated by lesions of supraspinal motor
pathways than are other animals. Whereas the functional deficit following motor cortical lesion is rather limited in animals such as the cat and rat, complete or nearly
complete paralysis is the rule in human subjects (Porter
and Lemon 1993). Lesions of the pyramidal tract also
have much more devastating effects in human subjects
than in other animals, including other primates (Porter
and Lemon 1993). It is therefore not unreasonable to
believe that the motor cortex and pyramidal tract have
achieved their important role for human motor control at

198

Fig. 2. Overview of central control centers involved in the control of human walking. The central control of human walking is
based on the integrative function of a number of different neuronal circuitries at different levels of the central nervous system.
There is a network in the spinal cord that has the capacity of
generating the basic locomotor rhythm (spinal central pattern
generator; CPG). This network is controlled by descending
motor commands from the brain stem and motor cortex and
receives sensory afferent feedback from the moving joints,
muscles, and skin. The descending motor tracts also contribute
directly to the motoneuronal drive, and the sensory feedback is
integrated at all levels of the central nervous system.

the expense of spinal control mechanisms, such as a

spinal stepping generator. Indeed, the lack of recovery of
functional motor ability following complete spinal cord
lesion in human subjects makes it easy to ignore a role
of the spinal cord motor circuitry in human motor control, as is often done in clinical accounts of the physio-

THE NEUROSCIENTIST

How We Walk

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

logical basis of neurological symptoms following brain

lesions (cf. the popular use of the terms upper and lower
motor neuron diseases in neurological literature and
clinical practice, which ignores a role of nonpyramidal
pathways and spinal cord circuitries). However, this
point of view is unnecessarily simplisticif not plain
wrongas demonstrated by recent studies in both monkey and man.
The view that humans may not possess a functioning
spinal rhythmgenerating network found support in
some initial failures to demonstrate locomotion in
macaque monkeys following complete spinal cord lesion
(Fulton and Sherrington 1932; Eidelberg and others
1981). However, this may have been due to inadequate
support of the animal, because some stepping movements have been observed in a single macaque monkey
in which care was taken that the hindlimbs were sufficiently loaded (Vilensky and others 1992; Vilensky and
OConnor 1998). Furthermore, Fedirchuk and others
(1998) demonstrated rhythmic alternating activity in
antagonistic muscles in acute spinalized and decerebrated marmoset monkeys following application of the
monoaminergic drug clonidine or the excitatory amino
acid NMDA. Although this rhythmic activity was not as
persistent and robust as that seen in the cat, it nevertheless provided conclusive evidence that a spinal
rhythmgenerating network does exist in the monkey
spinal cord. Fedirchuk and others (1998) observed a
much more robust alternating activity when the spinal
cord was left intact and stimulation was applied in the
brain stem. In line with Eidelberg and others (1981),
they therefore suggested that the spinal network in the
monkey spinal cord depends more on supraspinal control to work properly than what is the case in other animals.
The available data suggest that this also applies to
humans. There is now accumulating evidence that
demonstrates that rhythmic activity may be generated
from the human spinal cord under some circumstances.
There are a few reports in the older literature of observations of rhythmic activity in patients with complete
spinal cord injury (Holmes 1915; Kuhn 1950). However,
the patients studied by Holmes (1915) turned out to have
incomplete lesions when autopsy was performed some
time later, and the patient studied by Kuhn (1950) only
showed short-lasting activity following elicitation of
flexor reflexes. More recently, Bussel and others (1988,
1996) reported of a patient in whom bursts of extensor
activity were observed spontaneously and which turned
into an alternating flexion-extension pattern following
elicitation of flexor reflexes. Calancie and others (1994)
also observed involuntary stepping in a patient who
was unable to produce voluntary movements following
severe incomplete spinal cord injury. Finally,
Dimitrijevic and others (1998) reported that electrical
stimulation of the spinal cord could elicit alternating
rhythmic activity in leg muscles in patients with complete spinal cord injury.
The combined evidence from these studies suggests that
the human spinal cord does possess a rhythm-generating

network, but at the same time it also has to be concluded that spontaneous rhythmic activity following complete spinal cord injury is rare and that considerable
effort is generally necessary to activate the network in
the absence of supraspinal input. This difference
between patients with complete and incomplete spinal
cord lesion is also evident from recent attempts at rehabilitating their gait ability by treadmill training. Whereas
rehabilitation of gait in patients with incomplete lesions
with some survival of descending motor pathways shows
great promise (Wernig and others 1995; Barbeau and
Fung 2001; Wirz and others 2001), functional recovery
of gait is not seen in patients with complete lesions,
although some EMG activity may be seen when the
patients are trained on a treadmill with body-weight support (Dietz and others 1995; Wirz and others 2001).
Thus, we do have a spinal rhythmgenerating network, which can be activated under some circumstances,
but do we also use it during normal walking? There is
actually not much evidence of this, and what there is is
of an indirect nature.
It has been known for some time that the swing of the
arms during walking is not simply a passive movement
induced by the mechanical displacement of the body, but
that the muscles are activated cyclically by the CNS
(Fernandez and others 1965). Although it is tempting to
interpret this cyclic activity as reminiscent of the distant
past when our ancestors walked on four legs, this observation does not in itself prove that a spinal rhythm generator is involved. Recently, Dietz and others (2001),
however, demonstrated that perturbation of the legs
induced responses in the active arm muscles. These
responses were only seen during walking and had a sufficiently short latency to be mediated by a spinal pathway. This observation provides support to the idea that
the cyclic activity in the arms (and the legs) is generated
by a spinal rhythm generator.
Role of Sensory Feedback
in the Control of Human Walking
Sensory information plays a crucial role in all forms of
motor control, including walking. It is useful to consider this sensory information as having at least three different roles in motor control: 1) The sensory feedback
may help to drive the active motoneurons. 2) The sensory feedback may contribute to corrective reflexes following sudden perturbations. 3) The sensory feedback
may provide essential error signals that inform the brain
of differences between the intended movement and the
actually executed movement, which may be used for
updating of the future movements (i.e., motor learning).
Sensory feedback also plays all of these three roles in
relation to the control of walking. In the following, the
first two of these roles will be described in more detail,
whereas it will only be mentioned here that Dietz and
colleagues have elegantly demonstrated how sensory
afferent information may be used to adapt and update the
gait pattern (Prokop and others 1995; Erni and Dietz
2001).

Volume 9, Number 3, 2003

THE NEUROSCIENTIST

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

199

Sensory Drive to Motoneurons

during Walking
Sensory feedback has been shown to be crucial for
entraining and regulating the activity of the spinal
rhythmgenerating network in the cat. Positive feedback
signals from the active ankle extensors thus determine
the duration of the stance phase (Pearson and others 1998).
When the ankle joint is unloaded in late stance, this positive feedback decreases and the swing phase is initiated.
Similarly, feedback from hip joint afferents contributes
to the timing of the different phases of the gait cycle
(Grillner and Rossignol 1978). Does sensory feedback
play similar roles during human walking? The answer to
this is Yes, sensory feedback does play a role, but there
are important differences as compared to the cat.
A striking demonstration of the important contribution of sensory feedback to the motoneuronal drive during human walking was presented in a recent study in
which the ankle extensors were suddenly unloaded at
different times during the stance phase of walking
(Sinkjaer and others 2000). This was done by a portable
mechanical device, which could suddenly impose a
dorsi- or plantarflexion movement of the ankle joint
(Fig. 3). When the device imposed a plantarflexion
movement in the stance phase of walking, the ankle
plantarflexors were shortened and the muscles unloaded.
Following this unloading, a very significant drop in the
EMG activity recorded from the active ankle plantarflexors was observed at a latency of around 60 ms (Fig. 3).
This drop in EMG activity was still observed when the
common peroneal nerve, which innervates the ankle dorsiflexors, was blocked by injection of a local anesthetic
(Fig. 3) as well as when the skin of the foot was anesthetized. This demonstrates that the drop in the EMG
activity must be caused by an unloading-induced
decrease of activity in tibial nerve afferents from the
active ankle plantarflexors. The finding thus demonstrates that positive feedback from the active muscles
contributes to the motoneuronal drive during normal
walking much the same way as has been shown in the cat
(Pearson and others 1998). However, in the cat this positive feedback is mainly carried by force-sensitive group
Ib afferents from golgi tendon organs. Although positive
force feedback probably also makes some contribution
during human walking, it seems likely that lengthsensitive group II afferents, provide a very substantial
part of the positive feedback. There are two findings in
the study by Sinkjaer and others (2000) that support this
idea: First, the size of the EMG depression did not parallel changes in the load on the muscles as would have
been expected if force-sensitive afferents were responsible for the observed drop in the EMG activity. Rather, it
paralleled changes in muscle length in the early stance
phase as would have been expected if length-sensitive
afferents were responsible. Second, when ischemia was
induced in the leg, the drop in EMG activity could still
be observed. Ischemia has been shown to selectively
block transmission in large-diameter group I afferents in
a time window of 25 to 30 min following the onset of

200

ischemia, leaving transmission in smaller diameter afferents such as group II afferents intact. As already mentioned, one of the characteristics of human walking is the
lengthening contraction of the ankle plantarflexors
throughout a large part of the stance phase. Such a
lengthening contraction is not seen in the cat (Engberg
and Lundberg 1969). It seems likely that this change in
movement kinematics is related to the important role of
group II afferent feedback in human walking.
Sensory Feedback Participates
in Corrective Reactions
As already mentioned, the completely spinalized cat has
the ability of making corrective movements when the
paw hits an obstacle during treadmill walking. Evoked
reflex activity thus tends to be mainly integrated at a
spinal level in the cat. Activation of cutaneous afferents
may also evoke reflexes of a definite spinal origin in
human subjects while sitting down and contracting their
muscles voluntarily (Aniss and others 1992). Thus,
short-latency spinal pathways that convey reflex activation of lower limb muscles do exist in man, and these
seem to be organized in much the same way as in the cat.
Surprisingly, these short-latency reflexes are effectively
depressed during human walking (Christensen and others 1999). When stimulating afferents from the skin of
the foot during human walking, only responses at relatively long latency are observed (Duysens and others
1990; Yang and Stein 1990). The latency of these
responses is sufficiently long for them to be caused by a
long-loop pathway through the brain stem or even the
motor cortex, although a spinal origin certainly cannot
be disregarded based on the latency alone. Recent data
suggest that these reflexes may in fact be at least partly
transcortical in origin. First, the responses are not present on the lesioned side in patients with hemiplegia
(Zehr and others 1998). Second, when the motor cortex
is activated by transcranial magnetic stimulation (TMS)
during walking, a convergence between the input from
the sensory afferents and the activation of the corticospinal cells by TMS was demonstrated (Christensen
and others 1999). This provides rather strong evidence in
support of the idea that at least part of the responses to
cutaneous stimulation during walking is caused by a
transcortical reflex pathway in man. This suggests that at
least some of the reactions to external perturbations during walking are integrated to a larger extent at a
supraspinal level in man than in the cat. This allows that
the correction of the movement becomes more flexible
and that visual input and other information may be taken
into account before the muscles are activated. This may
be more crucial when standing on only one leg as is the
case throughout most of the human gait cycle than during quadrupedal walking.
Role of the Motor Cortex in Human Walking
Cats are quite capable of walking over flat ground following a complete lesion of the motor cortex

THE NEUROSCIENTIST

How We Walk

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

Fig. 3. Muscle afferents contribute to the muscle activity

during human walking. When a sudden perturbation of the
ankle joint in plantarflexion direction is suddenly made during the stance phase of human walking, a drop in the EMG
activity recorded from the soleus muscle is observed. The
subjects were walking on a treadmill with a portable stretch
device that has the ability of imposing sudden stretches in
either ankle plantarflexion or dorsiflexion direction (A). In a
large part of the stance phase, the ankle joint is moving in
dorsiflexion direction (downward in B) while the soleus muscle is active (C). There is thus an active lengthening of the
muscle in this phase of walking. The heavy lines in B and C
show the movement of the ankle joint (A) and the soleus
EMG activity (C) in control steps (heavy lines; average of 10
steps) and steps in which a stretch in platarflexion direction
was induced (thin lines). In this way, an effective unloading of
the ankle plantarflexors was induced, which resulted in a drop
in the soleus EMG activity around 60 ms after the perturbation. Positive degrees are plantarflexion direction. Ten trials
per average. (Modified from Sinkjaer and others 2000; the
drawing on top is modified from Andersen and Sinkjaer 1995).

(Armstrong 1988; Drew and others 1996). Only when

the cats are required to avoid obstacles or to walk in a
more complicated environment such as a horizontal ladder do they encounter difficulties. Furthermore,
although corticospinal cells are modulated during the
gait cycle when cats are walking on flat ground or on a
treadmill, the main activity is seen when the cats walk on
a horizontal ladder or when they have to avoid obstacles.
This has led to the conclusion that the motor cortex and
the pyramidal tract play only a facultative role during
normal walking on a flat surface, whereas they play an
obligatory role in modifications of gait in response to
environmental or motivational influences.
What is the situation in man? Our knowledge about
this has mainly been gained within the last 4 to 5 years,
and it is still greatly lacking. The evidence comes from
three different sources: lesion studies, brain imaging
studies, and electrophysiological studies.
There is only very limited data regarding the effect of
selective pyramidal tract lesion in man. Nathan (1994)
published a study in which he reported that a patient who
had a selective lesion of the pyramidal tracts bilaterally
was incapable of walking for up to 5 months after the
lesion but then regained the capacity of walking. This

illustrates that the pyramidal tract does play a role in the

control of walking, but also that, as in the cat, there are
other ways of ensuring the same task in the absence of
an intact pyramidal tractother descending pathways
may take over. In contrast to the cat, it is, however, clear
that an intact motor cortex is a prerequisite for walking
to occur, although the transmission of the command signals from the motor cortex to the spinal cord obviously
does not need to be transmitted directly through the corticospinal tract but may also be channeled through other
pathways. Another question is whether these other pathways are necessary for human walking to occur. We do
not know much about this, but in his seminal review,
Wiesendanger (1969) mentioned one subject who had a
tumor that had destroyed most of the brain stem except
the pyramidal tract. Wiesendanger mentioned that this
subject was capable of performing normal voluntary
movements with normal force up to the day before his
death. Unfortunately, it is unclear whether voluntary
movements also included walking.
Recently, it has also become possible to obtain evidence of the activity of the pyramidal tract during walking by either functional imaging techniques or transcranial magnetic stimulation (TMS). Fukuyama and others

Volume 9, Number 3, 2003

THE NEUROSCIENTIST

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

201

(1997) used single photon emission tomography

(SPECT) to demonstrate that there is a very significant
increased blood flow in the leg area of the sensorimotor
cortex, the cerebellum, and the frontal cortex during
treadmill walking in healthy human subjects. However, it
is unclear how much of this increased blood flow was
caused by neuronal activity involved in the generation of
muscle activity and how much might be explained by
afferent sensory feedback to the various brain areas.
Indeed, Christensen and others (2000) used positron
emission tomography (PET) to show that passive bicycle
movements (induced by an experimenter) may be as
effective as active movements (generated by the subjects) in producing changes in the cerebral blood flow
(Fig. 4). In both tasks, there was a very significant
increased cerebral blood flow in the same areas as
described by Fukuyama and others (1997). Larger cerebral blood flow during the active cycling task as compared to the passive task was only found in a restricted
area corresponding to the leg representation in the primary motor cortex. Most of the cerebral activity during
walking and bicycling thus appears to be generated by
sensory afferent feedback rather than the active internal
generation of muscle activity.
However, the increased blood flow in the primary
motor cortex in the active task must reflect the involvement of this area in controlling the rhythmic movements.
Using TMS, several groups have demonstrated that the
transmission in the corticospinal tract is greatly modulated during the walking cycle (Schubert and others
1997; Petersen and others 1998; Capaday and others
1999); for instance, motor evoked potentials (MEPs) in
the soleus muscle evoked by TMS are large in the stance
phase and absent in the swing phase. Importantly,
Petersen and others (1998) demonstrated that at least
part of this modulation of the MEPs is caused by
changes in excitability of corticospinal cells with direct
monosynaptic projections to the spinal motoneurons.
Such direct monosynaptic corticospinal projections or
corticomotoneuronal projections are only found in primates including man and are generally thought to be
responsible for the great manual dexterity in these
species, and especially in man (Porter and Lemon 1993).
However, in both monkeys and man, corticomotoneuronal projections are observed not only for hand muscles
but also for the large muscles in the leg. This newly
developed descending motor pathway, which is really
one of the most remarkable differences in the organization of the central motor system in primates as compared
to, for instance, the cat, must therefore play other roles
in the control of movement also. The evidence from
Petersen and others (1998) suggests that one of these
roles may be in the control of walking. In a more recent
study, this has been confirmed (Petersen and others
2001). These authors used weak TMS to knock out the
corticospinal contribution to the EMG activity during
human walking. Such weak stimuli have been shown to
activate only inhibitory cortical interneurons, which
inhibit the activity of the corticomotoneuronal cells.
Petersen and others (2001) observed that such weak

202

stimuli depressed the ongoing EMG activity in leg muscles during walking as a sign that the corticospinal contribution to the EMG activity had been knocked out.
In contrast to the cat, the human motor cortex thus
makes a significant contribution to the activation of the
muscles through the direct monosynaptic projections to
the spinal motoneurons even during locomotion on a
treadmill or on a flat surface. It seems likely that an even
more significant contribution is made in relation to visually guided modification of the gait as seen in the cat,
but little is known about this. Schubert and others (1999)
demonstrated that MEPs are larger when subjects have
to use visual cues to make precise steps as compared to
ordinary treadmill walking. It seems likely that this
reflects a greater involvement of the motor cortex in this
visually guided walking task, as has been found in the
cat. However, the exact mechanism for the increase is
unclear, and this and many other unresolved issues
regarding the exact role of the motor cortex in the control of walking remain a challenge for future research.
Conclusions and Prospects
I hope in this review to have shown that the control of
human walking rests on many of the same principles that
also govern the control of locomotion in other animals,
but at the same time that there have been some very significant modifications and adaptations in the central
neuronal circuitries involved.
There is no reason to suggest that human walking is
generated only by the spinal cord and that the corticospinal tract only makes a significant contribution in
relation to voluntary modifications of the gait pattern.
Nor is there any reason to suggest that the motor cortex
alone is responsible for the activation of the muscles during walking without any contribution from the spinal
cord neuronal circuitries. Human bipedal walking, as
animal quadrupedal walking, is based on an integration
of the activity of spinal neuronal circuitries with sensory feedback signals and descending motor commands
(cf. Fig. 2). However, to meet the functional requirements of bipedal walking, the activity and interaction of
these different central circuitries have been modified in
various ways. Probably the most important of these modifications is the greater dependency on supraspinal control and especially the significant contribution to the
control of the muscle activity by monosynaptic projections directly from the motor cortex to the spinal
motoneurons. In addition, the greater involvement of
transcortical reflex pathways in the reactions to perturbations of the walking pattern also seems significant.
However, modifications have also taken place at a spinal
level as evidenced from the significant motoneuronal
drive from length-sensitive muscle afferents when the
ankle plantarflexors are lengthened in the stance phase
of walking. Such modifications are the basis for the
unique walking pattern in human subjects. This has to be
taken into account in future attempts at rehabilitating
gait in patients with brain lesions. As a result of these
modifications, it has to be realized that although animal

THE NEUROSCIENTIST

How We Walk

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

Fig. 4. The motor cortex is

involved in the control of cyclic
leg movements. Subjects were
lying in a positron emission
tomography (PET) scanner
while performing bicycle movements using pedals attached to
the scanner bed. Oxygen
labeled H2O was used to
demonstrate changes in local
cerebral blood flow in relation
to the bicycle movements. The
rate of bicycling was 60 revolutions per minute. The PET data
were superimposed on magnetic resonance (MR) scans of
the subjects brains. A and B
show data recorded during
active and passive bicycling,
respectively. In the latter case,
the subjects were instructed to
relax and one of the experimenters moved the pedals at
the same rate as in A. C shows
the image obtained when the
images in A and B were subtracted. It is seen that active
and passive bicycling resulted
in a similar pattern of activity in
the sensory-motor cortex and
cerebellum. However, in the
subtracted image (C), it is seen
that active bicycling was associated with more activity bilaterally in the primary motor cortex. D shows the cerebral activity recorded while the subjects
were imagining to be bicycling.
This was associated with
increased activity in the supplementary motor cortex. (Modified
from Christensen and others
2000.)

experiments are valuable for an understanding of fundamental principles of neural control, they have to be supplemented by experimental evidence from human subjects. With the new developments in electrophysiological
and imaging techniques, it is now within our reach to
obtain this evidence.
References
Abitbol MM. 1988. Effect of posture and locomotion on energy expenditure. Am J Physiol Anthropol 77(2):1919.

Andersen JB, Sinkjaer T. 1995. An actuator system for investigating

electrophysiological and biomechanical features around the human
ankle joint during gait. Trans Rehabil Eng 3(4):299306.
Aniss AM, Gandevia SC, Burke D. 1992. Reflex responses in active
muscles elicited by stimulation of low-threshold afferents from the
human foot. J Neurophysiol 67(5):137584.
Armstrong DM. 1988. The supraspinal control of mammalian locomotion. J Physiol 405:137.
Barbeau H, Fung J. 2001. The role of rehabilitation in the recovery of
walking in the neurological population. Curr Opin Neurol
14(6):73540.
Bernstein NA. 1967. The coordination and regulation of movements.
Oxford (UK): Pergamon.

Volume 9, Number 3, 2003

THE NEUROSCIENTIST

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013

203

Bussel B, Roby-Brami A, Azouvi P, Biraben A, Yakovleff A, Held JP.

1988. Myoclonus in a patient with spinal cord transection. Possible
involvement of the spinal stepping generator. Brain 111(Pt
5):123545.
Bussel B, Roby-Brami A, Neris OR, Yakovleff A. 1996. Evidence for a
spinal stepping generator in man. Paraplegia 34(2):912.
Calancie B, Needham-Shropshire B, Jacobs P, Willer K, Zych G, Green
BA. 1994. Involuntary stepping after chronic spinal cord injury.
Evidence for a central rhythm generator for locomotion in man.
Brain 117(Pt 5):114359.
Capaday C, Lavoie BA, Barbeau H, Schneider C, Bonnard M. 1999.
Studies on the corticospinal control of human walking: I.
Responses to focal transcranial magnetic stimulation of the motor
cortex. J Neurophysiol 81(1):12939.
Christensen LO, Johannsen P, Sinkjaer T, Petersen N, Pyndt HS,
Nielsen JB. 2000. Cerebral activation during bicycle movements in
man. Exp Brain Res 135(1):6672.
Christensen LO, Morita H, Petersen N, Nielsen J. 1999. Evidence suggesting that a transcortical reflex pathway contributes to cutaneous
reflexes in the tibialis anterior muscle during walking in man. Exp
Brain Res 124(1):5968.
Datta AK, Stephens JA. 1990. Synchronization of motor unit activity
during voluntary contraction in man. J Physiol 422:397419.
Dietz V, Colombo G, Jensen L, Baumgartner L. 1995. Locomotor
capacity of spinal cord in paraplegic patients. Ann Neurol
37(5):57482.
Dietz V, Fouad K, Bastiaanse CM. 2001. Neuronal coordination of arm
and leg movements during human locomotion. Eur J Neurosci
14(11):190614.
Dimitrijevic MR, Gerasimenko Y, Pinter MM. 1998. Evidence for a
spinal central pattern generator in humans. Ann N Y Acad Sci
860:36076.
Drew T, Jiang W, Kably B, Lavoie S. 1996. Role of the motor cortex in
the control of visually triggered gait modifications. Can J Physiol
Pharmacol 74(4):42642.
Duysens J, Trippel M, Horstmann GA, Dietz V. 1990. Gating and reversal of reflexes in ankle muscles during human walking. Exp Brain
Res 82(2):3518.
Eidelberg E, Walden JG, Nguyen LH. 1981. Locomotor control in
macaque monkeys. Brain 104(Pt 4):64763.
Engberg I, Lundberg A. 1969. An electromyographic analysis of muscular activity in the hindlimb of the cat during unrestrained locomotion. Acta Physiol Scand 75(4):61430.
Erni T, Dietz V. 2001. Obstacle avoidance during human walking:
learning rate and cross-modal transfer. J Physiol 534(Pt 1):30312.
Fedirchuk B, Nielsen J, Petersen N, Hultborn H. 1998.
Pharmacologically evoked fictive motor patterns in the acutely
spinalized marmoset monkey (Callithrix jacchus). Exp Brain Res
122(3):35161.
Fernandez JM, Ballestros ML, Buchthal F, Rosenfalck P. 1965.
Electromyographic study of the muscles of the upper arm and
shoulder during walking. Acta Physiol Scand 63:296.
Fukuyama H, Ouchi Y, Matsuzaki S, Nagahama Y, Yamauchi H, Ogawa
M, and others. 1997. Brain functional activity during gait in normal subjects: a SPECT study. Neurosci Lett 228(3):1836.
Fulton JF, Sherrington CS. 1932. State of the flexor reflex in paraplegic
dog and monkey respectively. J Physiol 75:1722.
Gibbons A. 2002. Becoming human. In search of the first hominids.
Science 295(5558):12149.
Graham Brown T. 1911. The intrinsic factors in the act of progrssion in
the mammal. Proc R Soc Lond B 84:30819.
Grillner S. 1985. Neurobiological bases of rhythmic motor acts in vertebrates. Science 228(4696):1439.
Grillner S, Rossignol S. 1978. On the initiation of the swing phase of
locomotion in chronic spinal cats. Brain Res 146(2):26977.
Hansen NL, Hansen S, Christensen LO, Petersen NT, Nielsen JB.
2001. Synchronization of lower limb motor unit activity during
walking in human subjects. J Neurophysiol 86(3):126676.
Holmes G. 1915. Spinal injuries of warfare: II. The clinical symptoms
of gunshot injuries of the spine. Br Med J 2:81521.

204

Jordan LM. 1998. Initiation of locomotion in mammals. Ann N Y Acad

Sci 860:8393.
Jurmain R, Nelson H, Kilgore L, Trevathan W. 1997. Introduction to
physical anthroplogy. 7th ed. New York: Wadsworth.
Kirkwood PA, Sears TA. 1978. The synaptic connexions to intercostal
motoneurones as revealed by the average common excitation
potential. J Physiol 275(February):10334.
Kuhn RA. 1950. Functional capacity of the isolated human spinal cord.
Brain 73:151.
Lundberg A. 1979. Multisensory control of spinal reflex pathways.
Prog Brain Res 50:1128.
Nathan PW. 1994. Effects on movement of surgical incisions into the
human spinal cord. Brain 117(Pt 2):33746.
Pearson KG, Misiaszek JE, Fouad K. 1998. Enhancement and resetting
of locomotor activity by muscle afferents. Ann N Y Acad Sci
860:20315.
Petersen N, Christensen LOD, Nielsen J. 1998. The effect of transcranial magnetic stimulation on the soleus H reflex during human
walking. J Physiol (Lond) 513(Pt 2):599610.
Petersen NT, Butler JE, Marchand-Pauvert V, Fisher R, Pyndt HS,
Hansen NL, and others. 2001. Suppression of EMG activity by
transcranial magnetic stimulation in human subjects during walking. J Physiol 537(Pt 2):6516.
Porter R, Lemon R. 1993. Corticospinal function and voluntary movement. New York: Oxford University Press.
Prokop T, Berger W, Zijlstra W, Dietz V. 1995. Adaptational and learning processes during human split-belt locomotion: interaction
between central mechanisms and afferent input. Exp Brain Res
106(3):44956.
Rodman PS, McHenry HM. 1980. Bioenergetics and the origin of
hominid bipedalism. Am J Physiol Anthropol 52(1):1036.
Rossignol S. 2000. Locomotion and its recovery after spinal injury.
Curr Opin Neurobiol 10(6):70816.
Schubert M, Curt A, Colombo G, Berger W, Dietz V. 1999. Voluntary
control of human gait: conditioning of magnetically evoked motor
responses in a precision stepping task. Exp Brain Res
126(4):5838.
Schubert M, Curt A, Jensen L, Dietz V. 1997. Corticospinal input in
human gait: modulation of magnetically evoked motor responses.
Exp Brain Res 115(2):23446.
Sears TA, Stagg D. 1976. Short-term synchronization of intercostal
motoneurone activity. J Physiol 263(3):35781.
Shik ML, Orlovsky GN. 1976. Neurophysiology of locomotor automatism. Physiol Rev 56(3):465501.
Sinkjaer T, Andersen JB, Ladouceur M, Christensen LO, Nielsen JB.
2000. Major role for sensory feedback in soleus EMG activity in
the stance phase of walking in man. J Physiol 523(Pt 3):81727.
Vilensky JA, Moore AM, Eidelberg E, Walden JG. 1992. Recovery of
locomotion in monkeys with spinal cord lesions. J Motor Behav
24:28896.
Vilensky JA, OConnor BL. 1998. Stepping in nonhuman primates
with a complete spinal cord transection: old and new data, and
implications for humans. Ann N Y Acad Sci 860:52830.
Wernig A, Muller S, Nanassy A, Cagol E. 1995. Laufband therapy
based on rules of spinal locomotion is effective in spinal cord
injured persons. Eur J Neurosci 7(4):8239.
Wiesendanger M. 1969. The pyramidal tract: recent investigations on
its morphology and function. Ergeb Physiol 61:72136.
Winter DA. 1992. Foot trajectory in human gait: a precise and multifactorial motor control task. Phys Ther 72:4553.
Winter DA, Eng P. 1995. Kinetics: our window into the goals and
strategies of the central nervous system. Behav Brain Res
67:11120.
Wirz M, Colombo G, Dietz V. 2001. Long term effects of locomotor
training in spinal humans. J Neurol Neurosurg Psychiatry
71(1):936.
Yang JF, Stein RB. 1990. Phase-dependent reflex reversal in human leg
muscles during walking. J Neurophysiol 63(5):110917.
Zehr EP, Fujita K, Stein RB. 1998. Reflexes from the superficial peroneal nerve during walking in stroke subjects. J Neurophysiol
79(2):84858.

THE NEUROSCIENTIST

How We Walk

Downloaded from nro.sagepub.com at St Petersburg State University on December 16, 2013