J Neurophysiol 95: 3426 –3437, 2006;
doi:10.1152/jn.00081.2006.
Motor Patterns in Human Walking and Running
G. Cappellini,1 Y. P. Ivanenko,1 R. E. Poppele,2 and F. Lacquaniti1,3,4
1
Department of Neuromotor Physiology, Scientific Institute Foundation Santa Lucia, Rome, Italy; 2Department of Neuroscience,
University of Minnesota, Minneapolis, Minnesota; and 3Department of Neuroscience and 4Centre of Space Bio-medicine,
University of Rome Tor Vergata, Rome, Italy
Submitted 24 January 2006; accepted in final form 1 March 2006
Walking and running are generally considered as distinct
gait modes with strikingly different mechanics and energetics.
Humans change gait to increase locomotion speed while saving
energy. Thus oxygen consumption is lower for walking than
for running below the transition speed, whereas it is higher
above this speed (Margaria 1976). In walking, the body vaults
up and over each stiff leg in an arc, analogous to an inverted
pendulum (Cavagna et al. 1976) (Fig. 1A). Kinetic energy in
the first half of the stance phase is transformed into gravitational potential energy, which is partially recovered as the body
falls forward and downward in the second half of the stance
phase. Running, instead, is analogous to bouncing on a pogo
stick (Full and Koditschek 1999; Raibert 1986) (Fig. 1A). As a
leg strikes the ground, kinetic and gravitational potential energy is temporarily stored as elastic strain energy in muscles,
tendons, and ligaments and then is nearly all recovered during
the propulsive second half of the stance phase. The walking
gait may also be defined by the existence of a double support
phase during stance, whereas running has a “flight” phase
during which neither limb is in ground contact.
Walking and running are the two most common forms of
human gait. Although they share some basic kinetics and
kinematics, the two gaits are also distinctly different so the
transition from walking to running is obvious. In fact, both
kinematics and kinetics change abruptly in going from a
walking gait to a running gait (Hreljac 1993; Minetti et al.
1994; Nilsson et al. 1985). For example, the gait transition is
accompanied by an abrupt decrease in ground contact time by
⬃35% and an ⬃50% increase in peak ground reaction force.
There are also a number of gait parameters that change monotonically with increasing speed during both walking and running, including increased step length and cycle duration and
decreased stance duration (Nilsson et al. 1985). Many of these
changes are associated with increasing intensity of muscle
activation (Ivanenko et al. 2006; Prilutsky and Gregor 2001;
Winter and Yack 1987).
The mechanisms that underlie the changes associated with
speed and gait changes are still not well understood. In general,
the mechanics of locomotion and the associated muscle activity
have been more thoroughly studied in the human than has the
neural control. For the latter, we must rely largely on extrapolations from animal models. A number of experimental findings from animals and humans suggest that different locomotion patterns may result primarily from peripheral factors, like
muscle performance or sensory feedback (Pearson 2004; Smith
et al. 1993) and/or by a reconfiguration of the neural network
(Gillis and Biewener 2001; Grillner 1981) or by modulating the
parameters of a basic network (Collins 2003; de Leon et al.
1994; Golubitsky et al. 1999; Grillner and Zangger 1979;
Orlovsky et al. 1999; Pribe et al. 1997).
An important result from the animal studies was the demonstration that the entire range of speeds and locomotion gaits
can be generated in the cat by varying only the intensity of
stimulation of the mesencephalic locomotor region (MLR)
(Mori et al. 1989; Shik 1983; Shik et al. 1966). The gait
determined by this “central command” was automatically
adapted to the external conditions (such as slope of the road or
body loading) by “peripheral” mechanisms. Increasing only the
strength of the MLR stimulation could increase the speed of
forward progression, and at stronger stimulation levels, the
animal changed the locomotion gait from out-of-phase coordination (walk or trot) to in-phase coordination (run or gallop).
Areas anatomically similar to the MLR in the cat have also
Address for reprint requests and other correspondence: Y. P. Ivanenko,
Dept. of Neuromotor Physiology, Scientific Institute Foundation Santa Lucia,
306 via Ardeatina, 00179 Rome, Italy (E-mail: [email protected]).
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.
INTRODUCTION
3426
0022-3077/06 $8.00 Copyright © 2006 The American Physiological Society
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
Cappellini, G., Y. P. Ivanenko, R. E. Poppele, and F. Lacquaniti.
Motor patterns in human walking and running. J Neurophysiol 95:
3426 –3437, 2006; doi:10.1152/jn.00081.2006. Despite distinct differences between walking and running, the two types of human locomotion are likely to be controlled by shared pattern-generating networks.
However, the differences between their kinematics and kinetics imply
that corresponding muscle activations may also be quite different. We
examined the differences between walking and running by recording
kinematics and electromyographic (EMG) activity in 32 ipsilateral
limb and trunk muscles during human locomotion, and compared the
effects of speed (3–12 km/h) and gait. We found that the timing of
muscle activation was accounted for by five basic temporal activation
components during running as we previously found for walking. Each
component was loaded on similar sets of leg muscles in both gaits but
generally on different sets of upper trunk and shoulder muscles. The
major difference between walking and running was that one temporal
component, occurring during stance, was shifted to an earlier phase in
the step cycle during running. These muscle activation differences
between gaits did not simply depend on locomotion speed as shown
by recordings during each gait over the same range of speeds (5–9
km/h). The results are consistent with an organization of locomotion
motor programs having two parts, one that organizes muscle activation during swing and another during stance and the transition to
swing. The timing shift between walking and running reflects therefore the difference in the relative duration of the stance phase in the
two gaits.
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
3427
METHODS
Subjects
FIG. 1. General characteristics of walking and running. A: schematic representation of walking by vaulting (inverted pendulum) and running by a
“bouncing” gait (leg spring-loaded behavior during stance). B: ankle, knee, and
hip moments of force and vertical ground reaction force (normalized to the
subject’s weight) of the right leg during overground walking (5.4 km/h) and
running (9.4 km/h) in 1 representative subject. C: relative stance duration and
cycle duration (⫾SD) in walking and running. D: foot (fifth metatarsophalangeal joint, VM) trajectory characteristics (mean ⫾ SD, normalized to the
limb length L): horizontal excursion of the VM marker [relative to greater
trochanter (GT)] and vertical VM displacements (in the laboratory reference
frame).
been found to serve a similar function in fish, reptiles, birds,
and primates (Jordan 1991; Mori et al. 1996).
The question remains, however, as to whether this descending control modulates a basic “motor program” for walking and
running (for instance, a set of nonlinear oscillators with bifurcations at critical transitional points) or a separate set of
oscillators for each distinct gait. In either case, a gait transition
might be subject to feedback concerning critical peripheral
factors. A recent line of investigation relevant to the neural
generation of locomotor patterns is based on the identification
of elementary units of muscle activation (Bizzi et al. 1991,
2002; Davis and Vaughan 1993; d’Avella and Bizzi 2005;
Giszter et al. 2001; Hart and Giszter 2004; Hultborn 2001;
Ivanenko et al. 2003; Kargo and Giszter 2000; Patla et al. 1985;
Ting and Macpherson 2005; Tresch and Bizzi 1999). According to this approach, locomotor programs may be considered as
a set of characteristic timings of muscle activation (Ivanenko et
al. 2005). In fact, the same five basic activation components
can account for the electromyographic (EMG) waveforms of
32 ipsilateral leg, trunk, and shoulder muscles during walking
at speeds between 1 and 5 km/h (Ivanenko et el., 2004b). The
same five components were also present when walking was
J Neurophysiol • VOL
Eight healthy subjects [6 males and 2 females, between 26 and 44
yr of age, 69 ⫾ 10 kg (mean ⫾ SD), 1.75 ⫾ 0.07 m] volunteered for
the experiments. All subjects were right-leg dominant. The studies
conformed to the Declaration of Helsinki, and informed consent was
obtained from all participants according to the procedures of the
Ethics Committee of the Santa Lucia Institute.
Experimental setup
Subjects walked or ran on a treadmill (EN-MILL 3446.527, Bonte
Zwolle BV, Netherlands) at different controlled speeds (3–12 km/h).
They were asked to swing their arms normally and to look straight
ahead. Subjects walked with their shoes on. Before the recording
session, subjects practiced for a few minutes in walking and running
on the treadmill at different speeds. In a standard protocol, subjects
were asked to walk at 3, 5, 7, and 9 km/h and to run at 5, 7, 9, and 12
km/h so that we could compare walking and running at the same
speeds (5, 7, and 9 km/h).
We also used a computer-controlled speed program that linearly
increased and then decreased treadmill speed between 1 and 12 km/h
to observe the recorded parameters during continuous speed changes
(ramp speed condition, acceleration, and deceleration was set to 0.4
km 䡠 h⫺1 䡠 s⫺1). Three subjects participated in this experiment. They
were instructed to follow the changes in speed by remaining in place
with respect to the treadmill when the belt velocity was changing. We
also recorded during overground walking [at 5.6 ⫾ 1.1 (SD) km/h]
and running (at 9.6 ⫾ 0.9 km/h). This allowed us to calculate the
moments of forces (see following text) by asking subjects to step on
a force plate located in the middle of a 7-m walkway.
Data recording
We recorded kinematic data bilaterally at 100 Hz by means of the
Vicon-612 system (Oxford, UK) with nine TV cameras spaced around
the walkway. Infrared reflective markers (diameter: 1.4 cm) were
attached on each side of the subject to the skin overlying the following
landmarks: gleno-humeral joint (GH), the midpoint between the
anterior and the posterior superior iliac spine (ilium, IL), greater
trochanter (GT), lateral femur epicondyle (LE), lateral malleolus
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
combined with various voluntary motor tasks that required
significantly different muscle usage (Ivanenko et al., 2005). If
there is a common set of oscillators for walking and running,
we might anticipate that muscle activation during running
might also be accounted for, at least in part, by the same basic
muscle activation components. If, however, separate sets of
oscillators underlie the different gaits, we might expect that
this would be associated with distinctly different timings of
muscle activation. Although the analysis of individual EMG
patterns of human running has been performed in numerous
studies (see Prilutsky and Gregor 2001), the underlying common structure of the motor output has not been previously
investigated.
Therefore the aim of this study was to examine how muscle
activation depends on locomotion speed and on locomotion
gait. The experimental design was to record kinematics and
EMG activity from human subjects as they either walked or ran
on a treadmill at each of several different speeds. We used
statistical methods incorporating a linear decomposition of the
EMG data to determine the general design of the motor output
in walking and running.
3428
G. CAPPELLINI, Y. P. IVANENKO, R. E. POPPELE, AND F. LACQUANITI
Data analysis
BIOMECHANICAL ANALYSIS. The gait cycle was defined with respect to the right leg movement, beginning with right foot contact with
the surface (touch-down). The body was modeled as an interconnected
chain of rigid segments: IL-GT for the pelvis, GT-LE for the thigh,
LE-LM for the shank, and LM-VM for the foot. For walking and
running, the gait cycle was defined as the time between two successive
foot contacts of the right leg corresponding to the local minima of the
HE marker. The timing of the lift-off was determined analogously
(when the VM marker elevated by 3 cm). The touch-down and lift-off
events were also verified from the force plate recordings (when the
vertical ground reaction force exceeded 7% of the body weight), and
we found that the kinematic criteria we used predicted the onset and
end of stance phase with an error ⬍2% of the gait cycle duration (see
also Borghese et al. 1996). The data were time-interpolated over
individual gait cycles on a time base with 200 points.
Raw data were numerically rectified, low-pass filtered with a zero-lag Butterworth filter with cutoff at 10 Hz, timeinterpolated over a time base with 200 points for individual gait cycles
and averaged. Each trial for a given speed included ⱖ10 consecutive
gait cycles (typically 15).
EMG-ANALYSIS.
Joint moments of force
For the overground locomotion records, the moments of forces
(with extensor moments being positive) at the ankle, knee, and hip
J Neurophysiol • VOL
joints of the right leg were calculated using measured kinematics,
force plate data, anthropometric data taken on each subject, and the
traditional Newton-Euler inverse dynamics model (Bresler and
Frankel 1950). The moments were normalized to the body mass.
Factor analysis
We applied a principal component analysis (PCA) to each of
several data sets consisting of normalized EMG patterns over a step
cycle. Factor analysis (FA) used here has been thoroughly described
in our previous papers (Ivanenko et al. 2003, 2004b, 2005). Briefly,
the steps involve calculation of the correlation matrix, extraction of
the initial principal components (PCs), application of the varimax
rotation, calculation of factors scores (referred to as temporal activation components in the text), factor loadings (weighting coefficients
across muscles), and percent of variance accounted for (PV) by each
temporal component in the total data set. The aim of FA is to represent
the original EMG data set E (m ⫻ t matrix, where m: the number of
muscles and t ⫽ 200 for all conditions because EMG data were
time-interpolated over individual gait cycles to fit a normalized
200-point time base) as a linear combination of n basic temporal
components (n ⬍ m), E ⫽ W C ⫹ residual, where W are weighting
coefficients or loadings (m ⫻ n matrix) and C are basic temporal
components (n ⫻ t matrix).
Some of the deeper muscles around the hip have a rather complex
architecture depending on the hip joint angle (Delp et al. 1999), and
indeed we cannot rule out a possibility that there might be some
cross-talk in our recordings of muscles like ILIO. Nevertheless, our
averaged records (the major peak of activity around lift-off) are
consistent with those reported in the literature (Andersson et al. 1997;
Rab 1994) and obtained using a simulated biomechanical model of
bipedal stepping (Zajac et al. 2003). A FA could theoretically be
compromised by electrical cross-talk among adjacent muscle recordings (De Luca and Merletti 1988; Nene et al. 2004). However, if
cross-talk did exist, it would most likely have affected only the
weighting coefficients assigned to each component in accounting for
the activity of a muscle that was contaminated by cross-talk (Ivanenko
et al. 2004b).
In the present study, unless a component explained at least as much
as 3% of the total variance, we dropped it. The higher-order components were generally variable and not significant (see RESULTS). To
assess similarities of activation components across subjects and
speeds, we used the following parameters: timing of the main peak
and the width of the main peak (Ivanenko et al. 2005). The five basic
components were ordered according to the timing of the main peak.
The width of the main peak was estimated by measuring the full-width
at half-maximum (FWHM). The half of maximum was calculated as
a point corresponding to the mid-height between the peak value and
the mean value between 2 minima (1 to the left, another 1 to the right
with respect to the main peak). In the case of “boundary” peaks (close
to the foot contact), we used only descending or ascending part of the
component profile to calculate the half-width at half-maximum and
then multiplied it by 2, under the assumption of the symmetrical
profile.
Independent component analysis and nonnegative
matrix factorization
Different statistical methods have been developed to assess a linear
decomposition of the original set of data based on different assumptions (Tresch et al. 2006). Thus factor analysis with varimax rotation
constrains the analysis to orthogonal (uncorrelated) factors. Although
factor analysis of EMG activity was thoroughly documented for
human locomotion and we could thus compare our results with those
reported in the literature (Davis and Vaughan 1993; Ivanenko et al.
2003, 2004b, 2005; Merkle et al. 1998; Olree and Vaughan 1995), we
also performed two other decomposition-related statistical analyses:
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
(LM), heel (HE), and fifth metatarso-phalangeal joint (VM). The
spatial accuracy of the system is better than 1 mm (root mean square).
We recorded electromyographic activity by means of surface electrodes from 32 muscles simultaneously on the right side of the body.
We used active Delsys electrodes (model DE2.1, Delsys, Boston,
MA) applied to lightly abraded skin over the respective muscle belly.
The signals were amplified (⫻10,000), filtered (20 – 450 Hz; Bagnoli
16, Delsys), and sampled at 1,000 Hz. Sampling of kinematic, force
platform, and EMG data were synchronized.
The following 32 muscles were recorded from all eight subjects:
tibialis anterior (TA), flexor digitorum brevis (FDB), gastrocnemius
lateralis (LG), gastrocnemius medialis (MG), soleus (SOL), peroneus
longus (PERL), vastus lateralis (Vlat), vastus medialis (Vmed), rectus
femoris (RF), sartorius (SART), biceps femoris (long head, BF),
semitendinosus (ST), adductor longus (ADD), tensor fascia latae
(TFL), gluteus maximus (GM), gluteus medius (Gmed), external
oblique (OE), internal oblique (OI), latissimus dorsi (LD), iliopsoas
(ILIO), rectus abdominis, superior portion (RAS), erector spinae
recorded at T1, T9, and L2 (EST1, EST9, ESL2, respectively), biceps
brachii (BIC), triceps brachii (TRIC), deltoideus, anterior and posterior portions (DELTA and DELTP, respectively), trapezius, inferior
and superior portions (TRAPS and TRAPI, respectively), sternocleido
mastoideus (STER), and splenius (SPLE). Electrode placement for the
erector spinae muscle was 2 cm lateral to the spinous process, for
SOL—about 2 cm distal to the medial head of the gastrocnemius, for
RAS—about 3 cm lateral of the umbilicus (also see Winter 1991).
Before the electrodes were placed, the subject was instructed about
how to selectively activate each muscle (Kendall et al. 1993), while
EMG signals were monitored, so as to optimize the EMG signal and
minimize cross-talk from adjacent muscles during isometric contractions.
During overground locomotion, we also recorded the ground reaction forces (Fx, Fy, and Fz) under the right foot at 1,000 Hz by a force
platform (0.9 ⫻ 0.6 m, Kistler 9287B, Zurich, Switzerland). At the
end of the recording session, we made anthropometric measurements
on each subject. These included the mass and stature of the subject,
the length and circumference of the main segments of the body
(Zatsiorsky et al. 1990).
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
Statistics
A factor analysis was performed using Statistica v6.0 (StatSoft).
Statistical analyses (Student’s t-tests) were used to compare the
half-widths of the component peaks (FWHM) in walking and running.
Statistics on correlation coefficients was performed on the normally
distributed, Z-transformed values.
RESULTS
General characteristics of walking and running
Running and walking gaits are usually adopted for different
speeds of locomotion, with a preferred transition occurring at
⬃7 km/h for most human subjects (Nilsson et al. 1985). Some
major differences between the two gaits are illustrated in Fig.
1A. During walking, the leg tends to behave like a rigid strut,
and the joints remain relatively extended throughout the stance
phase (Lee and Farley 1998). In contrast, during running, the
major leg joints undergo substantial flexion and extension
during stance as the leg behaves in a more spring-like manner.
Both running and walking can occur, however, over a wide
range of speeds (Minetti et al. 1994). We examined both gaits
at 5, 7, and 9 km/h and walking alone at 3 km/h and running
alone at 12 km/h. Although increases in locomotion speed in
both gaits were accomplished by increasing both the cadence
and stride length, the duration of the stance phase was always
⬎50% of the gait cycle for walking and ⬍50% for running
(Fig. 1C). This difference corresponds to a change from a
double support phase during walking to a single support during
running, which is then accompanied by a greater ground
reaction force occurring shortly after foot contact at the beginning of the cycle (Fig. 1B).
J Neurophysiol • VOL
Figure 1D shows the normalized values (over all trials and
subjects) of horizontal (VMx) and vertical (VMy) excursion of
the foot plotted as function of speed. These global gait parameters exhibited the well-known monotonic relationship with
increasing speed and different behaviors between walking and
running (Nilsson et al. 1985).
Muscle activity patterns
The overall effect of speed on the average EMG activity was
consistent with that reported in the literature (den Otter et al.
2004; Hogue 1969; Ivanenko et al. 2002; Murray et al. 1984;
Nilsson et al. 1985), and it may be seen in Fig. 2. The intensity
of distal leg activity is more similar across speeds, whereas the
intensity of proximal leg and trunk activity increases much
more with speed. Except for a few muscles like Bic, there is no
obvious distinction in the intensity of muscle activation at the
transition from walking to running (dotted vertical lines in Fig.
2A). However, there are differences in the patterns of activation
at the transition, particularly in the distal leg muscles. The peak
of LG activation shifts to an earlier phase in the cycle. Instead
the changes in Vlat and ST activation are less obvious (Fig.
2B). The single cycle records at the gait transition points show
that the difference in the activation timing of the calf muscles
between walking and running occurs abruptly within a single
step cycle (Fig. 2B). Some high-intensity muscle activation
prior to the walk-to-run transition (see LG activity) is consistent with the idea that the activation of major leg muscles might
act as a trigger for gait transition and explain why locomotion
at nonpreferred speeds requires excessive muscle activation
(Prilutsky and Gregor 2001).
The average EMGs for each muscle we recorded are illustrated in Fig. 3. They show the same basic result illustrated by
Fig. 2 but with more detail. During walking, distal leg activity
had about the same modulation intensity and overall pattern at
all speeds although mean activity increased with speed. Proximal leg activity became more robust at the higher speeds in
walking, and individual patterns of activity were often different
at different speeds (see also Ivanenko et al. 2002; Winter and
Yack 1987). The patterns of activity in trunk and abdominal
muscles also tended to change dramatically as a function of
speed. Cervical muscle activity, which was quite low at the
lowest speeds, became significant at the higher speeds. There is
an increase in the EMG activity of all muscles during running.
In general muscles are most active around the time of foot
contact.
The posterior calf muscles tend to function as one unit.
During walking, the anterior muscles (TA) become active prior
to lift-off and remain active throughout the swing phase and
into the first 10% of the next cycle. The posterior calf muscles
(MG, LG, SOL, PERL) are all active at ⬃40% of the cycle
when the anterior muscles are relatively silent. With running,
this pattern reverses so that the anterior muscles increase their
activity during the mid part of the cycle, and the posterior
muscles are active instead around the time of foot contact.
This, in fact, is the most marked change in muscle coordination
we observed in the transition between walking and running.
These changes are associated with changes in leg and foot
movements. In walking, the foot strikes the ground with the
rear part of the heel and with a marked plantarflexion in the
ankle. In running, on the other hand, initial contact is generally
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
independent component analysis (ICA) and nonnegative matrix factorization (NMF).
We performed an informax ICA (Bell and Sejnowski 1995) using
the function “runica” in the EEGLAB package (v4.5; www.sccn.ucsd.
edu/eeglab/) (Delorme and Makeig 2004) running on Matlab v7. The
rationale for this analysis is that unlike PCA (FA), ICA aims at
extracting unknown hidden components from multivariate data using
only the assumption that the unknown components are mutually
independent (Hart and Giszter 2004).
We also applied a NMF using the matrix multiplication algorithm
described by Lee and Seung (1999) that constrains the temporal
components and weighting coefficients to be nonnegative. The rationale for application of the NMF was that our data consisted of
nonnegative values (rectified EMG activity), and the NMF constraint
allows only additive not subtractive combinations (d’Avella and Bizzi
2005).
All methods were applied to the same data set in each task and the
results across methods and tasks (walking vs. running) were compared. We reduced the dimensionality of the data to the same final
number of temporal components as in the FA for both the ICA and
NMF analyses. Using the correlation matrix for factor analysis (see
preceding text) automatically implies normalization of each EMG
waveform to its SD value. To compare the results of FA with those of
ICA and NMF, the amplitude of EMG waveforms was normalized by
their mean SD values prior to the application of ICA and NMF. This
is reasonable because we were mainly interested in rhythmic patterning elements in the control of locomotion (temporal components)
rather than in muscle synergies (weighting coefficients) and because
the amplitude of EMG activity was roughly similar when walking and
running at the same speed (see RESULTS).
3429
3430
G. CAPPELLINI, Y. P. IVANENKO, R. E. POPPELE, AND F. LACQUANITI
made with a more anterior part of the foot. Individual differences exist in the way the foot is placed on the ground (Chan
and Rudins 1994; Rodgers 1988), which must also lead to
differences in the activity pattern of muscles controlling
the ankle.
EMG components: comparison of analytical methods
In our previous study (Ivanenko et al. 2005), we compared
three different forms of factor analysis to find common components in the EMG patterns across muscles. Because each of
these statistical approaches (FA, ICA, and NMF, see METHODS)
places different restrictions on the outcomes, they might also
extract different sets of activation components. We found,
however, that the different algorithms converged to a similar
solution about the temporal structure of the EMG activation
patterns, both during normal walking and during walking
combined with a voluntary movement.
We repeated this comparison here to determine whether the
same components were also identified for EMG activity recorded during running. As before, we ordered the ICA and
J Neurophysiol • VOL
NMF components so that their waveforms matched the corresponding FA components (Fig. 4). The results were similar to
those we reported previously (Ivanenko et al. 2005; see also
Tresch et al. 2006) in that the three methods gave essentially
the same result for both walking at 5 km/h (average waveform
correlation with FA component r ⫽ 0.92 ⫾ 0.05 for ICA and
0.89 ⫾ 0.09 for NMF) and for running at 12 km/h (r ⫽ 0.94 ⫾
0.03 for ICA and 0.87 ⫾ 0.11 for NMF). In addition, the
loadings or weighting coefficients on particular muscles were
also similar in the three methods. Finally, the total variance
explained by basic temporal components was also similar.
For instance, in Fig. 4, the percent of variance explained by
all components was 87, 90, and 86% during walking and
90, 91, and 89% during running using FA, ICA and NMF,
respectively.
Factor analysis
Given the nearly equivalent results of the previous section,
we again used the same FA approach as we did previously
(Ivanenko et al. 2003, 2004b, 2005) to describe common
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
FIG. 2. Slow changes in treadmill belt speed (ramp speed condition) in 1 representative subject. A: electromyographic (EMG) activity, instantaneous treadmill
belt speed, limb angle, and gait cycle duration. Limb angle was defined as GT-LM elevation angle [the angle between the greater trochanter (GT)-lateral
malleoulus (LM) segment projected on the sagittal plane and the vertical]. B: EMG activity of selected muscles [gastrocnemius lateralis (LG), tibialis anterior
(TA), semitendinosus (ST), and vastus lateralis (Vlat)] computed for single cycles before and after the transition cycle. Vertical dashed lines denote the transition
from walking and running (defined by the presence/absence of the double support phase). The transition typically occurred at the higher speed during gait
acceleration as opposed to gait deceleration, consistent with previous observations (Thorstensson and Roberthson 1987). Note that except for a few muscles like
Bic (associated with an abrupt change in the forearm elevation during running), there is no obvious distinction in the intensity of muscle activation at the transition
from walking to running. However, there are differences at the transition in the pattern of muscle activation.
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
3431
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
FIG. 3. Ensemble averaged activity patterns recorded from 8 subjects for 32 muscles. EMG records were filtered with a low-pass cut-off of 10 Hz. Top: stick
diagrams. Data were plotted vs. a normalized cycle. As the relative duration of stance varied across subjects, a hatched region indicates an amount of variability
in the stance phase duration across subjects.
components in the EMG patterns across muscles. We analyzed
EMG patterns separately for each subject and for each speed
(Fig. 5). The basic result was that five temporal components
accounted for between 83 and 99% of the total EMG waveform
variance for all subjects across all conditions. The most significant component of order higher than five explained only
3.1 ⫾ 1.1% of the total variance and was highly variable across
subjects and conditions (component 6, Fig. 5).
The five components determined from the data recorded
during walking were essentially the same as those we described
J Neurophysiol • VOL
previously (Ivanenko et al. 2004b), but they extend the range of
those results to now include walking at 7 and 9 km/h. Five
temporal components were also consistently found for all
speeds during running (Fig. 5B). They each corresponded to a
component derived from the walking data with the single clear
exception of component 2. This component was substantially
phase-shifted from a peak at ⬃45% of the cycle during walking
to between 20 and 30% of the cycle during running.
As before, we found small but systematic shifts in the timing
of the main peaks of the components to an earlier phase with
95 • JUNE 2006 •
www.jn.org
3432
G. CAPPELLINI, Y. P. IVANENKO, R. E. POPPELE, AND F. LACQUANITI
increasing speed (Fig. 6A). We showed earlier that these shifts
tended to scale with the duration of the stance phase as that
changed with speed (Ivanenko et al. 2004b). At the higher
speeds used here (⬎3 km/h), the duration of stance did not
decrease as much with speed as it did at lower walking speeds
(1–3 km/h), so the component phase shifts were correspondingly less. There were also comparable phase shifts with speed
during running. In fact, the phases of component 1, 3, 4, and 5
were essentially the same as we found at comparable speeds for
walking (Fig. 6A). This might be slightly less clear in the case
of component 5 because there may be a difference in the
running and walking phases at 9 km/h.
The phase of component 2 seems to behave differently
during walking and running. During walking the shift with
speed is comparable to what we observed with the other four
components. During running, however, not only is there a
discrete phase shift with respect to the corresponding walking
component, but the shift with speed seems to be greater.
However, when we consider the phases relative to lift-off (end
of stance) instead of relative to heel strike or touch-down, the
phase of component 2 is consistently ⬃80% of the cycle (20%
prior to lift-off) during both walking and running (Fig. 6B). In
contrast, with lift-off as the reference, the phases of the other
four components are no longer comparable during walking and
running. In fact the phase of component 4 during walking
aligns with component 3 during running and component 1
during walking aligns with the phase of component 5 during
running. Moreover components 5 and 3 during walking and
components 1 and 4 during running were not aligned with
comparable components. Thus all but component 2 may be
better represented by the reference to touch-down as we have
also used in our previous studies.
The half-widths of the component peaks (FWHM) were
essentially the same (⬃13% of the cycle duration) for each
component across speeds (Fig. 6C). There was some tendency
though for the running components 1 and 2 to be slightly wider
(15–17%) at the lower speeds relative to the walking components (P ⫽ 0.06 for 5 km/h and P ⫽ 0.05 for 7 km/h, paired
t-tests).
J Neurophysiol • VOL
The total amount of waveform variance accounted for (PV)
by each component was quite similar for walking and running,
with components 4 and 5 typically accounting for less variance
than the other three during both walking and running (Fig. 6D).
Moreover, some differences that depended on walking speed,
such as an increase in PV by component 3 and decrease by
component 4 with increasing speed (Ivanenko et al. 2004b)
seemed to hold also during running.
The relative strength of the effect of each activation component on a given EMG pattern is given by the component
“loading” or “weighting coefficient” (Fig. 7B). We estimated
any changes in weighting coefficients of the five temporal
components between walking and running at the same speed (7
km/h) by correlating the set of weighting coefficients for
various groups of muscles. We made separate comparisons for
the leg and lower trunk muscles and the upper trunk and
shoulder muscles for each corresponding component in running and walking (Fig. 7B). There was a fairly high consistency
of relatively high correlation coefficients (rLEG ⫽ 0.68 ⫾ 0.17)
indicating that similar muscle groups in the limb and lower
trunk were activated with common timing during walking and
running. Interestingly the best correspondence was for component 2, indicating that the loading of component 2 on leg
muscles was basically the same in walking and running in spite
of the shift in the timing of component 2.
The consistency was much less, however, for the upper trunk
and shoulder muscles (rTRUNK ⫽ 0.28 ⫾ 0.39), where the
biggest differences were seen with components 1 and 4. These
components therefore participate in the activation of different
groups of muscles in walking and running. It is obvious in fact
the upper body and arm motion in particular tend to be much
different in walking and running.
DISCUSSION
We examined muscle activation in human subjects over a
range of speeds during both walking and running. Consistent
with our previous result on the effects of walking speed
(Ivanenko et al. 2004b), we found that speed changes during
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
FIG. 4. Comparison of activation components during walking and running at the same speed (7 km/h), derived from averaged muscle activity patterns (see
Fig. 3) using different statistical methods [factor analysis (FA), independent component analysis (ICA) and nonnegative matrix factorization (NMF)]. The
amplitude of EMG waveforms was normalized by the mean SD before application of ICA and NMF. Correlations between components obtained from ICA and
NMF and corresponding varimax factors (FA) are indicated.
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
3433
both walking and running are associated with increases in the
intensity of muscle activation and only minor changes in their
relative timings. The same five activation components accounted for muscle activity during either walking or running
over the entire speed range tested. Moreover, in spite of
significant differences in the overall characteristics of the two
gaits, we found that major differences in muscle activation
timing occurred only during stance, while walking and running
timing was basically the same during the swing phase.
Speed versus gait effects
Changes in locomotion speed are accompanied by changes
in cycle length and cycle duration as well as in the fraction of
the cycle devoted to stance or swing. Although such changes
J Neurophysiol • VOL
are also accompanied by changes in muscle activation patterns,
most of these scale with the cycle duration so that changes in
activation intensity seem to dominate (Andersson et al. 1997;
den Otter et al. 2004Ivanenko et al. 2004b; Murray et al. 1984;
Prilutsky and Gregor 2001; Winter and Jack 1987). Gait
changes on the other hand, such as the transition from walking
to running, are often considered to involve a different muscle
usage and not just more intense muscle activation. However,
this is not what we found. The major change in muscle
activation during running compared with walking was a phase
shift attributable to a single activation component during
stance, whereas the muscles activated by the shifted component remained the same. Increasing running speeds, from 7 to
12 km/h, was mostly accompanied by a similar activation of
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
FIG. 5. Temporal activation components by subject. A: varimax factors derived from muscle activity patterns by factor analysis for each of 8 subjects (thin
lines). Bold traces show the average of the components across 8 subjects. Five common components across conditions are designated in a “chronological” order
(with respect to the timing of the main peak). The next higher-order component (that explained on average only ⬃3% of variance) was also plotted. It displayed
very high variability across subjects and conditions and, therefore, was dropped from the analysis. Stance phase was somewhat different across subjects as
indicated by hatched area of the stance-swing bar.
3434
G. CAPPELLINI, Y. P. IVANENKO, R. E. POPPELE, AND F. LACQUANITI
One hypothesis proposed to account for the preferred transition
to running with increasing speed predicts this difference in
activation timing. It notes that at higher walking speeds, the
ankle extensors are loaded toward the end of stance in an
unfavorable portion of their force-velocity curve, so their force
production is limited even as levels of activation increase.
Shifting the activation to an earlier phase of stance shifts the
activation to a more favorable force production range (Neptune
and Sasaki 2005). This can also be seen as a shift in the ground
reaction force (Fz) to near the beginning of the cycle (Fig. 1B).
This shift of muscle activation timing from walking to
running and the greater peak activation intensity during walking are both clearly evident in the EMG recordings from the
posterior calf muscles (Fig. 3, MG, LG, PERL, and SOL). The
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
FIG. 6. Characteristics of activation temporal components. A: the timing
(⫾SD) of the peaks of 5 common FA components across speeds and tasks. The
5 basic components are each characterized by a relatively narrow peak of
activation at a particular phase of the cycle. Zero timing refers to the heel
strike. Note the abrupt change in the timing of component 2 between walking
and running. B: the timing of the same peaks referenced to the lift-off event.
Gait cycle and 0 timing refers to the lift-off in this plot. Note the invariance of
the timing of component 2 in walking and running when referenced to the
lift-off event. C: the width (⫹SD) of the main peaks of components estimated
as the full-width at half-maximum (FWHM). D: percent of variance explained
by each component in the data sets obtained from all individual muscles.
leg muscles and somewhat more intense activation of trunk and
shoulder muscles.
Muscle activation patterns recorded at 3 km/h were basically
the same as those recorded at higher walking speeds with only
few exceptions (e.g., TFL, ADD, Fig. 3). Instead, most of the
32 muscles we recorded exhibited more intense activation at
higher speeds, and the activation timing was fully accounted
for by five activation components (Ivanenko et al. 2004b). We
expected though that running, which places different biomechanical demands on the neuromuscular system, would exhibit
different muscle activation patterns. However, many muscle
activation patterns were quite similar to those observed during
walking. In fact, the five components that represent those
patterns were each associated with the activation timing of
nearly the same set of leg and lower trunk muscles during
running and walking. The average correlation across weighting
coefficients was 0.68 (range: 0.38 – 0.82). Thus these components seem to play essentially the same role in both conditions
independently of speed [however, see Ivanenko et al. (2004b)
and Fig. 6B, which shows a decrease in PV for component 4
and an increase for component 3 with increases in speed].
Gait changes
The activation of ankle extensors at an earlier phase of the
locomotion cycle was the major EMG change we found to be
associated with running compared with walking, and it reflects
a difference in the support function of the leg in the two gaits.
J Neurophysiol • VOL
FIG. 7. Activation temporal components derived from averaged muscle
activity patterns (see Fig. 3). A: superimposed activation components in
walking and running. Note a prominent phase shift of the timing of the main
peak of the second component in running. B: weighting coefficients (across 32
muscles) of the 5 basic temporal components during walking and running at
the same speed (7 km/h). The correlation of weighting patterns between
walking and running is shown at the right of each record both for the leg/lower
trunk [tibialis anterior (TA) – rectus abdominis, superior portion (RAS)] and
upper trunk/shoulder [latissimus dorsi (LD) – sternocleido mastoideus
(STER)] muscles separately.
95 • JUNE 2006 •
www.jn.org
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
J Neurophysiol • VOL
This might be expected if component timing (or CPG phase)
can be directly controlled by sensory input or descending
commands (Pearson 2004).
This may also relate to experimental work in frogs suggesting that a small collection of unit burst generators of synergistic muscle activity can be triggered by sensory input to produce
corrections and other adjustments of spinally organized trajectories (Hart and Giszter 2004; Kargo and Giszter 2000). In
humans, several researchers have provided support for a segmental movement generation and superposition of short bursts
or fixed timing elements to control upper limb point to point
and cyclic motions (e.g., Krebs et al. 1999). This is not unlike
our previous observation about the coordination of voluntary
motor acts with walking (Ivanenko et al. 2005). In that case,
the five activation components associated with walking were
also present in the combined motor behavior, and a new
component with task-related timing was superimposed. Thus to
coordinate the stance and swing phases in both walking
and running, four of the activation components remain invariantly timed with respect to the beginning of the stance phase.
The other component (2) is inserted to correspond with the
phase of the lift-off and therefore shifts its phase during
running (Fig. 8).
This interpretation implies that the same basic motor program consisting of five distinct muscle activation phases characterizes both walking and running at various speeds. It also
FIG. 8. Hypothetical motor programs for walking (A) and running (B) in
terms of the characteristic timing of muscle activations. In both programs, 4
activation components (1, 3, 4, and 5) occur relative to the touch-down (TD)
phase of the cycle and component 2 occurs relative to the lift-off (LO) phase,
at about 20% of the cycle preceding LO. As 1 gait mode is switched to the
other, the timing of LO relative to TD changes along with the relative timing
of component 2 with respect to the other 4 components.
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
timing shift is reflected in the activation components as a shift
in the timing of component 2, which is weighted primarily on
those same muscles. That is, component 2 accounts for the
activation timing of basically the same muscles in both gaits.
The preceding explanation about an earlier phase of ankle
extensor activation implied a control based on optimal muscle
performance using feedback information about muscle force
production, length, and velocity. However, muscle performance may not contribute directly to the control of timing. The
subjects in our study were able to use either gait voluntary at
5, 7, and 9 km/h, and their muscle activation exhibited a similar
discrete shift in timing between walking and running at each
speed. It is also evident however, that component 2 timing
shifted to progressively earlier phases in the cycle as running
speed increased, indicating a possible role for muscle performance, as suggested by this hypothesis.
Other proposals to explain the transition to running also
involve parameters that are expected to change continuously
with speed such as the pendulum-related limitations for the
frequency of leg oscillations, metabolic energy expenditure
(Margaria 1976), or an increased “sense of effort” due to the
exaggerated swing-related activation of ankle knee and hip
flexors (Prilutsky and Gregor 2001) or have some critical value
like a critical velocity of ankle flexion (Hreljiac 1995), critical
forces (Farley and Taylor 1991; Raynor et al. 2002), or a
critical angle between the thighs (Minetti et al. 1994; see also
Biewener et al. 2004; Brisswalter and Mottet 1996; Cavagna et
al. 1976; Kram et al. 1997; Neptune and Sasaki 2005; Nilsson
et al. 1985; Thorstensson and Roberthson 1987). Although
such speed-related parameters may account for a bifurcation in
behavior, they do not easily account for a discrete change,
which can occur over a large range of speeds.
Our analysis suggests another possibility, namely that the
motor program activates ankle extensors at a fixed interval
preceding lift-off. We found that this timing, corresponding to
component 2 activation at 20% of the cycle prior to lift-off,
was basically the same over the entire speed range for both
walking and running (Fig. 6B). Thus the propulsive force
generated as a result of this activation timing occurred at the
end of stance for both gaits. Although this timing may produce
a more efficient force production during running due to the
biomechanical factors discussed in the preceding text, it is also
a simple strategy requiring only that the timing of one component be coordinated with lift-off, which may in turn simply
depend on sensory information about the transition from stance
to swing. The effect, however, is to coordinate the swing and
stance phases in the two gaits.
According to this hypothesis, the timing of components 1, 3,
4, and 5 are referenced to the beginning of the cycle at heel
strike or touch-down, whereas component 2 occurs at a phase
that is appropriate for launching the swing phase. This implies
that the invariant relative timing of the five muscle activation
components we have noted for walking does not represent a
basic unit for all locomotion. As we found for running, the
relative timing of component 2 timing can be disassociated
from the timing of the four components because their phases
are referenced to different events in the locomotion cycle. The
relative timing of other components might also be gait dependent in some way so that other forms of locomotion gait (e.g.,
hopping, skipping) could also have different relative component timings that are invariant only for a particular gait pattern.
3435
3436
G. CAPPELLINI, Y. P. IVANENKO, R. E. POPPELE, AND F. LACQUANITI
implies that peripheral factors like the transition to the swing
phase may have essential modulatory roles affecting the relative timing of activation. It is not clear, however, to what extent
these influences might contribute to the basic program to
provide an appropriate and invariant kinematic pattern (Lacquaniti et al. 1999).
Conclusions
GRANTS
This work was supported by the Italian Ministry of Health, Italian Ministry
of University and Research, and Italian Space Agency.
REFERENCES
Andersson EA, Nilsson J, and Thorstensson A. Intramuscular EMG from the
hip flexor muscles during human locomotion. Acta Physiol Scand 161:
361–370, 1997.
d’Avella A and Bizzi E. Shared and specific muscle synergies in natural motor
behaviors. Proc Natl Acad Sci USA 102: 3076 –3081, 2005.
Bell AJ and Sejnowski TJ. An information-maximization approach to blind
separation and blind deconvolution. Neural Comput 7: 1129 –1159, 1995.
Biewener AA, Farley CT, Roberts TJ, and Temaner M. Muscle mechanical
advantage of human walking and running: implications for energy cost.
J Appl Physiol 97: 2266 –2274, 2004.
Bizzi E, D’Avella A, Saltiel P, and Tresch M. Modular organization of spinal
motor systems. Neuroscientist 8: 437– 442, 2002.
Bizzi E, Mussa-Ivaldi FA, and Giszter S. Computations underlying the
execution of movement: a biological perspective. Science 19; 253: 287–291,
1991.
Borghese NA, Bianchi L, and Lacquaniti F. Kinematic determinants of
human locomotion. J Physiol 494: 863– 879, 1996.
Bresler E and Frankel JB. The forces and moments in the leg during level
walking. J Appl Mech 72: 27–36, 1950.
Brisswalter J and Mottet D. Energy cost and stride duration variability at
preferred transition gait speed between walking and running. Can J Appl
Physiol 21: 471– 480, 1996.
Cavagna GA, Thys H, and Zamboni A. The sources of external work in level
walking and running. J Physiol 262: 639 – 657, 1976.
Chan CW and Rudins A. Foot biomechanics during walking and running.
Mayo Clin Proc 69: 448 – 461, 1994.
Collins JJ. Gait transitions. In: The Handbook of Brain Theory and Neural
Networks (2nd ed.), edited by Arbib MA. Cambridge: The MIT Press, 2003,
p. 463– 466.
Davis BL and Vaughan CL. Phasic behavior of EMG signals during gait: use
of multivariate statistics. J EMG Kinesiol 3: 51– 60, 1993.
de Leon R, Hodgson JA, Roy RR, and Edgerton VR. Extensor- and
flexor-like modulation within motor pools of the rat hindlimb during
treadmill locomotion and swimming. Brain Res 654: 241–250, 1994.
Delorme A and Makeig S. EEGLAB: an open source toolbox for analysis of
single-trial EEG dynamics including independent component analysis.
J Neurosci Methods 134: 9 –21, 2004.
Delp SL, Hess WE, Hungerford DS, and Jones LC. Variation of rotation
moment arms with hip flexion. J Biomech 32: 493–501, 1999.
De Luca CJ and Merletti R. Surface myoelectric signal cross-talk among
muscles of the leg. Electroencephalogr Clin Neurophysiol 69: 568 –575,
1988.
J Neurophysiol • VOL
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
In summary, gait changes and speed changes each have
specific effects on muscle activation during locomotion. Nevertheless, the diverse biomechanical demands of running and
walking and of locomoting at different speeds are met by
surprisingly similar general template of muscle activation. The
walking and running motor program in humans may be considered as a sequence of five temporal activation components
(Fig. 8). One gait mode is switched to a distinctly different gait
mode by an interplay between the timings of activation components relative to the stance and swing phase of the locomotion.
den Otter AR, Geurts AC, Mulder T, and Duysens J. Speed related changes
in muscle activity from normal to very slow walking speeds. Gait Posture
19: 270 –278, 2004.
Farley CT and Taylor CR. A mechanical trigger for the trot-gallop transition
in horses. Science 253: 306 –308, 1991.
Full RJ and Koditschek DE. Templates and anchors: neuromechanical
hypotheses of legged locomotion on land. J Exp Biol 202: 3325–32, 1999.
Gillis GB and Biewener AA. Hindlimb muscle function in relation to speed
and gait: in vivo patterns of strain and activation in a hip and knee extensor
of the rat (Rattus norvegicus). J Exp Biol 204: 2717–2731, 2001.
Giszter SF, Moxon KA, Rybak IA, and Chapin JK. Neurobiological and
neurorobotic approaches to control architectures for a humanoid motor
system. Robot Auton Syst 37: 219 –235, 2001.
Golubitsky M, Stewart I, Buono PL, and Collins JJ. Symmetry in locomotor central pattern generators and animal gaits. Nature 401: 693– 695, 1999.
Grillner S. Control of locomotion in bipeds, tetrapods, and fish. In: Handbook
of Physiology. The Nervous System. Motor Control. Bethesda, MD: Am.
Physiol. Soc., 1981, sect. 1, vol. 2, part 1, p. 1179 –1236.
Grillner S and Zangger P. On the central generation of locomotion in the low
spinal cat. Exp Brain Res 34: 241–261, 1979.
Hart CB and Giszter SF. Modular premotor drives and unit bursts as
primitives for frog motor behaviors. J Neurosci 24: 5269 – 82, 2004.
Heglund NC, Cavagna GA, and Taylor CR. Energetics and mechanics of
terrestrial locomotion. III. Energy changes of the centre of mass as a
function of speed and body size in birds and mammals. J Exp Biol 97:
41–56, 1982.
Hogue RE. Upper-extremity muscular activity at different cadences and
inclines during normal gait. Phys Ther 49: 963–972, 1969.
Hreljac A. Preferred and energetically optimal gait transition speeds in human
locomotion. Med Sci Sports Exerc 25: 1158 –1162, 1993.
Hreljac A. Determinants of the gait transition speed during human locomotion: kinematic factors. J Biomech 28: 669 – 677, 1995.
Hultborn H. State-dependent modulation of sensory feedback. J Physiol 533:
5–13, 2001.
Ivanenko YP, Cappellini G, Dominici N, Poppele RE, and Lacquaniti F.
Coordination of locomotion with voluntary movements in humans. J Neurosci 25: 7238 –7253, 2005.
Ivanenko YP, Dominici N, Cappellini G, Dan B, Cheron G, and Lacquaniti F. Development of pendulum mechanism and kinematic coordination
from the first unsupported steps in toddlers. J Exp Biol 207: 3797–3810,
2004a.
Ivanenko YP, Grasso R, Macellari V, and Lacquaniti F. Control of foot
trajectory in human locomotion: role of ground contact forces in simulated
reduced gravity. J Neurophysiol 87: 3070 –3089, 2002.
Ivanenko YP, Grasso R, Zago M, Molinari M, Scivoletto G, Castellano V,
Macellari V, and Lacquaniti F. Temporal components of the motor
patterns expressed by the human spinal cord reflect foot kinematics. J Neurophysiol 90: 3555–3565, 2003.
Ivanenko YP, Poppele RE, and Lacquaniti F. Five basic muscle activation
patterns account for muscle activity during human locomotion. J Physiol
556: 267–282, 2004b.
Ivanenko YP, Poppele RE, and Lacquaniti F. Spinal cord maps of spatiotemporal alpha-motoneuron activation in humans walking at different
speeds. J Neurophysiol 95: 602– 618, 2006.
Jordan IM. Brain stem and spinal cord mechanisms for the initiation of
locomotion. In: Neurobiological Basis of Human Locomotion, edited by
Shimamura M, Grillner S, and Edgerton VR. Tokyo: Japan Scientific
Societies Press, 1991, p. 3–20.
Kargo WJ and Giszter SF. Rapid correction of aimed movements by
summation of force-field primitives. J Neurosci 20: 409 – 426, 2000.
Kendall FP, McCreary EK, and Provance PG. Muscles. Testing and
Function. Baltimore, MD: Williams and Wilkins, 1993.
Kram R, Domingo A, and Ferris DP. Effect of reduced gravity on the
preferred walk-run transition speed. J Exp Biol 200: 821– 826, 1997.
Krebs HI, Aisen ML, Volpe BT, and Hogan N. Quantization of continuous
arm movements in humans with brain injury. Proc Natl Acad Sci USA 96:
4645– 4649, 1999.
Lacquaniti F, Grasso R, and Zago M. Motor patterns in walking. News
Physiol Sci 14: 168 –174, 1999.
Lee CR and Farley CT. Determinants of the center of mass trajectory in
human walking and running. J Exp Biol 201: 2935–2944, 1998.
Lee DD and Seung HS. Learning the parts of objects by non-negative matrix
factorization. Nature 401: 788 –791, 1999.
MOTOR PATTERNS IN HUMAN WALKING AND RUNNING
J Neurophysiol • VOL
Raibert MH. Legged Robots That Balance. Cambridge, MA: MIT Press,
1986.
Raynor AJ, Yi CJ, Abernethy B, and Jong QJ. Are transitions in human gait
determined by mechanical, kinetic or energetic factors? Hum Mov Sci 21:
785– 805, 2002.
Rodgers MM. Dynamic biomechanics of the normal foot and ankle during
walking and running. Phys Ther 68: 1822–1830, 1988.
Shik ML. Action of the brainstem locomotor region on spinal stepping
generators via propriospinal pathways. In: Spinal Cord Reconstruction,
edited by Kao CC, Bunge RP and Reier PJ. New York: Raven, 1983, p.
421– 434.
Shik ML, Severin FV, and Orlovsky GN. Control of walking and running by
means of electrical stimulation of the midbrain. Biophysics 11: 756 –765,
1966.
Smith JL, Chung SH, and Zernicke RF. Gait-related motor patterns and
hindlimb kinetics for the cat trot and gallop. Exp Brain Res 94: 308 –322,
1993.
Thorstensson A and Roberthson H. Adaptations to changing speed in human
locomotion: speed of transition between walking and running. Acta Physiol
Scand 131: 211–214, 1987.
Ting LH and Macpherson JM. A limited set of muscle synergies for force
control during a postural task. J Neurophysiol 93: 609 – 613, 2005.
Tresch MC and Bizzi E. Responses to spinal microstimulation in the chronically spinalized rat and their relationship to spinal systems activated by low
threshold cutaneous stimulation. Exp Brain Res 129: 401– 416, 1999.
Tresch MC, Cheung VCK, and d’Avella A. Matrix factorization algorithms
for the identification of muscle synergies: evaluation on simulated and
experimental data sets. J Neurophysiol 95: 2199 –2212, 2006.
Winter DA. The Biomechanics and Motor Control of Human Gait: Normal,
Elderly and Pathological. Waterloo, Ontario: Waterloo Biomechanics,
1991.
Winter DA and Yack HJ. EMG profiles during normal human walking:
stride-to-stride and inter-subject variability. Electroencephalogr Clin Neurophysiol 67: 402– 411, 1987.
Zajac FE, Neptune RR, and Kautz SA. Biomechanics and muscle coordination of human walking. II. Lessons from dynamical simulations and
clinical implications. Gait Posture 17: 1–17, 2003.
Zatsiorsky V, Seluyanov V, and Chugunova L. In vivo body segment
inertial parameters determination using a gamma-scanner method. In: Biomechanics of Human Movement: Applications in Rehabilitation, Sports and
Ergonomics, edited by Berme N and Cappozzo A. Worthington, OH: Bertec,
1990, p. 186 –202.
95 • JUNE 2006 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 16, 2017
Makeig S, Debener S, Onton J, and Delorme A. Mining event-related brain
dynamics. Trends Cogn Sci 8: 204 –210, 2004.
Margaria R. Biomechanics and Energetics of Muscular Exercise. Oxford:
Clarendon, 1976.
Merkle LA, Layne CS, Bloomberg JJ, and Zhang JJ. Using factor analysis
to identify neuromuscular synergies during treadmill walking. J Neurosci
Methods 82: 207–214, 1998.
Minetti AE, Ardigo LP, and Saibene F. The transition between walking and
running in humans: metabolic and mechanical aspects at different gradients.
Acta Physiol Scand 150: 315–323, 1994.
Mori S, Matsuyama K, Miyashita E, Nakajima K, and Asanome M. Basic
neurophysiology of primate locomotion. Folia Primatol 66: 192–203, 1996.
Mori S, Sakamoto T, Ohta Y, Takakusaki K, and Matsuyama K. Sitespecific postural and locomotor changes evoked in awake, freely moving
intact cats by stimulating the brain stem. Brain Res 505: 66 –74, 1989.
Murray MP, Mollinger LA, Gardner GM, and Sepic SB. Kinematic and
EMG patterns during slow, free, and fast walking. J Orthop Res 2: 272–280,
1984.
Nene A, Byrne C, and Hermens H. Is rectus femoris really a part of
quadriceps? Assessment of rectus femoris function during gait in ablebodied adults. Gait Posture 20: 1–13, 2004.
Neptune RR and Sasaki K. Ankle plantar flexor force production is an
important determinant of the preferred walk-to-run transition speed. J Exp
Biol 208: 799 – 808, 2005.
Nilsson J, Thorstensson A, and Halbertsma J. Changes in leg movements
and muscle activity with speed of locomotion and mode of progression in
humans. Acta Physiol Scand 123: 457– 475, 1985.
Olree KS and Vaughan CL. Fundamental patterns of bilateral muscle activity
in human locomotion. Biol Cybern 73: 409 – 414, 1995.
Orlovsky GN, Deliagina TG, and Grillner S. Neural Control of Locomotion.
From mollusc to man. Oxford, UK: Oxford Univ. Press, 1999.
Patla AE, Calvert TW, and Stein RB. Model of a pattern generator for
locomotion in mammals. Am J Physiol 248: R484 – 494, 1985.
Pearson KG. Generating the walking gait: role of sensory feedback. Prog
Brain Res 143: 123–129, 2004.
Pribe C, Grossberg S, and Cohen MA. Neural control of interlimb oscillations. II. Biped and quadruped gaits and bifurcations. Biol Cybern 77:
141–152, 1997.
Prilutsky BI and Gregor RJ. Swing- and support-related muscle actions
differentially trigger human walk-run and run-walk transitions. J Exp Biol
204: 2277–2287, 2001.
Rab GT. Muscle. In: Human Walking (2nd ed.), edited by Rose J and Gamble
JG. Baltimore, MD: Willaims and Wilkins, 1994, p. 101–102.
3437