Gait & Posture 38 (2013) 674–678

Contents lists available at SciVerse ScienceDirect

Gait & Posture

journal homepage: www.elsevier.com/locate/gaitpost

Kinematic and kinetic analysis during forward and backward walking

Minhyeon Lee, Jungyoon Kim, Jongsang Son, Youngho Kim *
Department of Biomedical Engineering, Yonsei University, Room 204, Medical Industry Techno Tower, Heungup, Wonju, Gangwon 220-710, South Korea

A R T I C L E I N F O A B S T R A C T

Article history: Backward walking (BW) is a recently emerging exercise. However, limited studies exist regarding the
Received 13 June 2011 motion analysis of BW compared with that of forward walking (FW). The present study identified the
Received in revised form 31 January 2013 mechanisms of BW through kinetic analysis and focused on BW time-reversed data. A three-dimensional
Accepted 24 February 2013
motion capture system was used to acquire the joint movements and to calculate the joint moments and
the powers during walking. Ground reaction force curves were acquired from force plates. Each
Keywords: participant performed 10 FW trials and 40 BW trials with bare feet. All data were analyzed using paired t-
Backward walking
tests (p < 0.05) to verify the significant differences between FW and BW. In BW, since the progress is in
Normal gait
Human locomotion
the direction in which the person cannot see, the walker’s speed is generally decreased compared to FW.
Gait analysis As a result, the stride characteristics for each respective activity showed significant differences. The
Motion capture characteristics of angular displacement in all joints were almost identical in FW and time-reversed BW.
However, selected crucial points of joint angles were significantly different. The moment pattern of the
ankle joint was very similar in FW and time-reversed BW. In the knee and the hip joint, the joint moment
pattern of time-reversed BW was simpler than FW. The joint power patterns of the ankle, the knee and
the hip were different in FW and BW. An original finding of this study was that the main propulsion and
shock absorption joint during BW is the ankle joint. The knee and hip joint did not generate propulsion
power.
ß 2013 Elsevier B.V. All rights reserved.

1. Introduction Adopting the motor/system control perspective, research in

human walking has classified the aforementioned retro-locomo-
Backward walking (BW) is a recently emerging exercise that tion as a type of ‘‘reversible movement’’. Researchers have asserted
helps enhance the balance and muscle strength of the lower that the joint motions of FW, especially at the hip and ankle, are
limbs [1–6]. BW provides a stronger musculoskeletal system and similar to the time-reversed counterpart of BW [6–9]. However,
cardiovascular function [1]. BW is also beneficial to patients researchers differ in their statements on muscle activation during
with overuse injuries in the lower extremities since lower BW. Thorstensson [7] reported that electromyelography (EMG)
patellofemoral joint reaction forces were observed in BW patterns of muscle activity in BW show a poor relation to those in
compared to forward walking (FW) [2]. This activity is FW. Grasso et al. [8] found that the difference was the origin of
recommended for hemiplegic patients to realize or improve propulsion by EMG analysis; while the main FW propulsion is
proper motor control and gait performance [3]. Yang et al. [4] generated by the ankle plantarflexors, the principal BW propulsion
demonstrated that asymmetric gait patterns of post-stroke is provided by the hip and knee extensors.
patients could be improved after receiving BW therapy. Hackney Limited studies exist regarding the motion analysis of BW
and Earhart [5] reported BW’s efficacy in improving functional compared with that of FW. Though some kinematic analyses of BW
mobility, gait performance and balance of patients with have been made, the lack of research on BW lies prominently in its
Parkinson’s disease. kinetic analysis. BW is an instinct of human locomotion based on
FW, so studies in BW have substantial potential for understanding
the control of human locomotive behavior [10].
The present study identified the mechanisms of BW through
kinetic analysis and concentrated on BW time-reversed data.
It focused on comparing the spatiotemporal parameters and
* Corresponding author. Tel.: +82 33 760 2859; fax: +82 33 760 2806.
time-reversed data of BW kinematics and kinetics to those of
E-mail addresses: [email protected] (M. Lee), [email protected]
(J. Kim), [email protected] (J. Son),
non-reversed FW, and the results should contribute to the
[email protected] (Y. Kim). understanding of the mechanism of gait during BW.

0966-6362/$ – see front matter ß 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.gaitpost.2013.02.014
 M. Lee et al. / Gait & Posture 38 (2013) 674–678 675

2. Methods the gait event itself [11]. The power curves of BW also were not
time-reversed to describe BW’s mechanism. As referenced in
2.1. Participants Benedetti et al. [12], crucial points among the joint angles,
moments, and power curves were chosen as gait parameters to
Thirty-one healthy subjects 22.4  3.2 years of age, compare FW and BW.
171.5  5.5 cm in height, and weighing 70.0  10.4 kg participated Gait cycles were normalized entirely from 0% to 100% of the gait
in this study. These twenty-six male and five female participants had cycle to clearly distinguish both major and minor variations in the
no evidence or history of lower-limb diseases nor any record of patterns of any individual trial [13]. Spatiotemporal parameters,
surgery to the lower limbs. All participants gave informed consent kinematics, kinetics, and GRF data were determined from each
before taking part in the experiments. participant during both forward and backward level walking.
Sagittal plane motions were analyzed as the majority of the forces
2.2. Instrumentation and motions occur in this plane [14].

A three-dimensional motion capture system (VICON 612, 2.5. Statistical analysis

VICON Motion Systems, Oxford, UK) was used to acquire joint
movements and to calculate joint moments and power during The spatiotemporal parameters were also analyzed using
walking using inverse dynamics. Calibration of the system was paired t-tests (p < 0.05) to verify the significant differences
performed before the gait trials. Sixteen retro-reflective markers between forward and backward walking. Paired t-tests
(14 mm in diameter) were attached with double-sided tape to the (p < 0.05) were used to detect significant differences in gait
participants’ lower limbs according to the Plug-In-Gait marker set. parameters. All data were analyzed with the Statistical Package for
Motion data were collected at 120 samples per second. All marker the Social Sciences (SPSS) 19 statistical software.
coordinates were smoothed with a Woltring filter (MSE = 15).
Ground reaction force (GRF) curves were acquired from two 3. Results
Kistler (5233A2, Kistler, Winterthur, Switzerland) and two AMTI
(OR6-6, AMTI, Watertown, US) force plates. The GRF data were 3.1. Stride characteristics
sampled at 1080 Hz. The marker and force data were time-
synchronized. Significant reductions in walking speed (1.3  0.1 m/s vs.
1.1  0.1 m/s, p < 0.05) and cadence (111.4  5.2 steps/min vs.
2.3. Procedures 98.1  8.1 steps/min, p < 0.05) were observed in BW compared with
FW. However, no significant differences were found in the stance
Before the gait analysis, the participants’ age, height, weight, phase percentage (60.1  1.4% gait cycle vs. 60.4  1.6% gait cycle,
and lower-extremity anthropometric data were measured. Each p = 0.321). Stride time (1.1  0.1 s vs. 1.2  0.1 s, p < 0.05) increased
participant performed 10 FW trials and 40 BW trials on a 10-m significantly during BW. Stride length was also significantly different
walkway with bare feet. For the trials, stride characteristics, between BW and FW (1.4  0.1 m vs. 1.3  0.1 m, p < 0.05) (Table 1).
marker coordinates, and GRFs were recorded simultaneously. The
participants practiced BW prior to the actual experiments for 3.2. Kinematics
successful adaptation to the new environment and walking
pattern. To reflect their natural stride length and unique gait The characteristics of angular displacement in the ankle, knee
characteristics, participants were required to walk at a comfortable and hip were almost identical in FW and time-reversed BW (Fig. 1).
pace without an awareness of the position of the force plates. However, selected crucial points of joint angles were significantly
different (Table 2).
2.4. Data analysis There was no significant difference in the maximum plantar-
flexion at the loading response during FW and pre-swing during
Data pairs acquired from the markers and force plates were BW (A1, p = 0.335; Fig. 1a, Table 2). The ankle joint was
used to calculate joint angles, moments, and powers through the significantly dorsiflexed during the swing phase to loading
Plug-In-Gait Biomechanical Modeler Pipeline (Oxford Metrics, response in BW than pre-swing to the swing phase in FW (A2
Oxford, UK). Spatiotemporal parameters were computed from the and A3, p < 0.05; Fig. 1a, Table 2). The total range of motion (A4,
marker coordinate data using a developed code (MATLAB, Math- p < 0.05) was also significantly different.
Works Inc., US). The knee showed significant flexion at the loading response
The gait patterns of BW were divided into the following: (1) toe- during FW compared to pre-swing during BW (K1, p < 0.05; Fig. 1b,
contact to heel-off pattern (heel-off) and (2) toe-contact to toe-off Table 2). The knee was less flexed at the initial swing during FW
pattern (toe-off). The toe-off pattern often was evident in the early (K2, p < 0.05) than it was for the terminal swing during BW. The
trials of the experiments. The moment and power data of the toe- total range of motion of the knee was greater in FW (K3, p < 0.05).
off pattern were smaller than those of the heel-off pattern. The toe-
off pattern resulted from a lack of training, so these data were Table 1
excluded. FW consists of heel contact to toe off, which means that Stride characteristics: forward walking vs. backward walking.
FW and BW (heel off) had opposite contact positions (toe or heel) Stride characteristics Forward Backward p-value
for the same event (contact or push off). To adhere to the purpose walking walking
of this study, which was to compare the kinematic and kinetic Mean  SD Mean  SD
patterns of FW and BW, joint angle curves and joint moments for
Walking speed (m/s) 1.3  0.1 1.1  0.1 <0.001**
BW (heel off) were time-reversed to equalize the contact position
Cadence (steps/min) 111.4  5.2 98.1  8.1 <0.001**
as well as the type of event. First, the whole stance phase was Stance phase percentage 60.1  1.4 60.4  1.6 0.321
reversed. Then, the remaining gait cycle (the swing phase) was in gait cycle (%)
reversed to form an entirely new gait cycle with each phase Stride time (s) 1.1  0.1 1.2  0.1 <0.001**
reversed. The GRF curves of BW were not time-reversed because Stride length (m) 1.4  0.1 1.3  0.1 0.016*

they are affected more by the participant’s body weight than *p < 0.05; **p < 0.01.
676 M. Lee et al. / Gait & Posture 38 (2013) 674–678

Fig. 1. Mean joint angles, moments, and powers of ankle, knee and hip during FW and BW. (a) Ankle joint, (b) knee joint, and (c) hip joint.

The hip showed significantly less flexed at the initial contact phase during FW was significantly larger than that observed during
during FW than the heel off during BW (H1, p < 0.05; Fig. 1c, BW (HM2, p < 0.05).
Table 2) and was less extended at the loading response of BW than
the pre-swing of FW (H2, p < 0.05; Fig. 1c, Table 2). The total range 3.3.2. Joint powers
of motion of the knee was greater in FW (H3, p < 0.05). The characteristics of joint power in the ankle, knee and hip
were different in FW and BW (Fig. 1). The peak power generation,
3.3. Kinetics absorption timing, and amplitude were different in the ankle joint,
while the numbers of peak power generation and absorption were
3.3.1. Joint moments different in knee and hip joints.
The characteristics of joint moment in the ankle were almost In the ankle joint, the maximum power generation during FW
identical in FW and time-reversed BW (Fig. 1a). In the knee and hip was significantly larger than during BW (AP1, p < 0.05; Fig. 1a,
joint, joint moment characteristics of time-reversed BW were Table 2). The maximum power absorption during FW was
simpler than FW (Fig. 1b and c). significantly smaller than during BW (AP2, p < 0.05). There was
There was no significant difference in the maximum dorsi- no significant difference in the maximum power generation for
flexor moment of the ankle joint in the loading response phase the knee joint (KP1, p = 0.12; Fig. 1b and Table 2). The maximum
during FW and in the pre-swing phase during BW (AM1, p = 0.056; power absorption during BW for the knee joint was significantly
Fig. 1a and Table 2). There was no significant difference in the smaller than during FW (KP2, p < 0.05; Fig. 1b and Table 2). In
maximum plantarflexor moments during the stance phase (AM2, the hip joint, the maximum power generation at loading
p = 0.226). response during FW was significantly larger than that observed
No significant difference existed in the maximum extensor during BW (HP1, p < 0.05; Fig. 1c and Table 2). The maximum
moment of the knee joint in the loading response phase during power generation in the swing phase during FW was also
FW and the pre-swing phase during BW (KM1, p = 0.103; Fig. 1b significantly larger than that produced during BW (HP2,
and Table 2). The maximum flexor moment in the stance phase p < 0.05).
during FW was significantly larger than during BW (KM2,
p < 0.05). 3.3.3. Ground reaction forces
The hip showed significantly larger extensor moment in the The patterns of the vertical GRF curves also differed between
stance phase during FW than it did during BW (HM1, p < 0.05; FW and BW (Fig. 2). The GRFs on both FW and BW exhibited two
Fig. 1c and Table 2). The maximum flexor moment in the stance main peaks during the loading response and pre-swing phase.
 M. Lee et al. / Gait & Posture 38 (2013) 674–678 677

Table 2
Comparison of joint angles and moments variables: forward walking vs. backward walking.

Forward walking Backward walking p-value

Mean  SD Mean  SD

Ankle variable
A1 Max. plantarflexion in loading response (pre swing in BW) (8) 4.0  2.4 4.1  3.5 0.335
A2 Max. dorsiflexion in stance (8) 14.8  2.3 17.9  2.0 <0.001**
A3 Max. plantarflexion in swing (8) 17.7  4.8 3.0  3.4 <0.001**
A4 Total range of motion (8) 32.5  4.0 21.4  3.3 <0.001**
AM1 Max. dorsiflexor moment (N m/kg) 0.15  0.07 0.24  0.08 0.056
AM2 Max. plantarflexor moment (N m/kg) 1.67  0.08 1.60  0.11 0.226
AP1 Max. power generation (W/kg) 4.42  0.38 1.25  0.18 <0.001**
AP2 Max. power absorption (W/kg) 0.9  0.2 2.9  0.6 <0.001**

Knee variable
K1 Max. flexion in loading response (pre swing in BW) (8) 14.6  2.2 12.4  4.7 <0.001**
K2 Max. flexion in swing (8) 59.6  3.3 45.7  8.4 <0.001**
K3 Total range of motion (8) 61.4  3.6 47.9  8.6 <0.001**
KM1 Max. extensor moment (N m/kg) 0.54  0.06 0.49  0.10 0.103
KM2 Max. flexor moment in stance (N m/kg) 0.38  0.11 0.01  0.09 <0.001**
KP1 Max. power generation (W/kg) 1.05  0.36 0.85  0.23 0.120
KP2 Max. power absorption in loading response (W/kg) 0.79  0.28 0.17  0.16 <0.001**

Hip variable
H1 Flexion at heel strike (heel off in BW) (8) 31.8  3.7 29.5  3.8 <0.001**
H2 Max. extension in stance (8) 12.3  3.9 6.8  4.8 <0.001**
H3 Total range of motion (8) 45.2  2.8 37.7  4.3 <0.001**
HM1 Max. extensor moment in stance (N m/kg) 0.50  0.19 0.12  0.07 <0.001**
HM2 Max. flexor moment in stance (N m/kg) 0.69  0.14 0.50  0.15 <0.001**
HP1 Max. power generation in loading response (W/kg) 0.69  0.13 0.54  0.21 <0.001**
HP2 Max. power generation in swing (W/kg) 1.33  0.17 0.14  0.07 <0.001**

*p < 0.05; **p < 0.01.

However, the first peak of BW was higher than FW, and the second BW, the ankle dorsiflexed after the toe-strike and then plantar-
BW peak was smaller than that for FW. flexed during most of the support phase, and the knee extended
throughout the stance phase and flexed during most of the swing
phase, while the hip flexed during the stance phase and extended
4. Discussion
during the swing phase. In addition, the moment pattern of the
ankle joint was almost identical in FW and time-reversed BW. In
In BW, since the person cannot see the direction of the progress,
knee and hip, the joint moment patterns of time-reversed BW were
the speed generally decreased compared to FW. As a result, the
simpler than those of FW, because of few peak points disappeared
stride characteristics showed significant differences between FW
in time-reversed BW.
and BW [16]. However, unlike the differences in cadence and stride
In the ankle joint, the maximum plantarflexion moment was
length found in this study, previous researchers have noted only
generated during loading response during BW. In this phase, the
slight differences in average speed between FW and BW [6,8].
ankle decelerated to play an important part in shock absorption;
Stride length differs significantly because the average range of
thus, the maximum plantarflexion moment and the greatest power
motion in the hip joint between flexion and extension varies
absorption occurred. According to several previous studies [6–8],
greatly; the flexion angle was 908, and the extension angle was 208
the plantarflexor muscles, the gastrocnemius and soleus, are
[15].
activated in this phase. These muscles are activated in order to
Our results compare favorably with previous reports [6–8]. The
decelerate the ankle and absorb power. In midstance, the ankle
angular displacement patterns in the ankle, the knee and the hip
joint plantarflexed to move the trunk backward, causing the
were almost identical in FW and time-reversed BW. Therefore, in
greatest power generation during BW. Grasso et al. reported that
the main thrust is provided by knee and hip extensors during BW
[8]. However, during BW, the ankle joint generated much greater
power than the knee and the hip joints. Based on the results of our
power analysis, the ankle might be considered as a significant joint
for propulsion and shock absorption.
In the knee joint, the maximum power absorption at the loading
response was significantly smaller during BW than during FW. This
result indicated that the shock absorption at foot strike in the knee
joint was reduced during BW, because of the greater shock
absorption in the ankle joint. The power amplitude of the knee
joint decreased during BW because of a simpler pattern of knee
joint displacement. During initial contact in BW, the knee flexed
(60% gait cycle of Fig. 1b), and the knee extended during initial
contact to mid stance (60–25% gait cycle of Fig. 1b). The knee
slightly flexed in terminal stance to lower the body for opposite
foot contact and to prepare the backward propulsion of the body.
According to [6–8], the knee extensor muscles, the vastus lateralis,
Fig. 2. Mean GRF curves during FW and BW. vastus medialis, and rectus femoris, are activated from initial
678 M. Lee et al. / Gait & Posture 38 (2013) 674–678

contact to terminal stance. Because of these activations, the knee of time-reversed BW was simpler than that for FW. The joint power
joint is extended during this stance phase in order to prevent patterns of the ankle, knee and hip were different during FW and
descent of the body’s center of mass during BW. The knee joint BW. An original finding of this study was that the main propulsion
extended in pre-swing phase and flexed in the initial swing phase and shock absorption joint during BW was the ankle joint. The knee
to propel the foot backward, clear the foot over the ground, and and hip joints did not generate propulsion power. Further study
generate joint power. using EMG will help to determine the rehabilitation advantage of
Similar to previous reports [6,8], the hip joint was less flexed BW.
and less extended during the entire gait cycle in BW. In loading
response, the hip joint flexed, and the hip joint generated the
Acknowledgements
maximum power during BW. The flexion moment was converted
to an extension moment during the terminal stance to maintain
This research was financially supported by the Technology
the trunk vertically during BW. According to [6–8], the hamstring
Innovation Program (Industrial Strategic Technology Development
is activated during pre-swing to initial contact. This muscle is
Program, 10032055) funded by the Ministry of Knowledge
activated to hip extension during swing phase and braking knee
Economy (MKE, Korea), and was also supported by the Ministry
extension during swing phase.
of Knowledge Economy (MKE) and Korea Institute for Advance-
The pattern of the vertical GRF curves differed between FW and
ment of Technology (KIAT) through the Research and Development
BW. GRFs increased rapidly to support the whole body weight
for Regional Industry (70011192).
during the loading response. The knee flexed during mid-stance;
the force plate briefly unloaded, and the GRFs dropped below body
weight [12]. Similar to the findings of a previous report [8], the two Conflict of interest statement
peaks were roughly symmetrical in FW, whereas in BW, the first
peak caused by the loading of the body weight was always greater The authors have no conflict of interest regarding any of the
than the second peak caused by the heel push off. During BW, the material in the manuscript and meet the criteria for authorship as
second peak of GRF was smaller than the first since the knee and defined in your submission guidelines.
hip joints were merely lifting the limb and moving it backward.
Thus, a plateau shape not typically seen in FW was observed during References
the pre-swing phase in BW.
BW is more difficult and demanding than FW due to postural [1] Flynn TWC, Smutok SM, Michael A, Zeballos RJ, Weisman IM. Comparison of
instability and no visual cue to progress [16]. However, compared cardiopulmonary responses to forward and backward walking and running.
Medicine and Science in Sports and Exercise 1994;26:89–94.
to FW, BW is potentially useful in rehabilitation in terms of greater [2] Flynn TWC, Soutas-Little RW. Patellofemoral joint compressive forces in
muscle and neural activities [6,8,17], higher heart rate and oxygen forward and backward running. Journal of Orthopaedic and Sports Physical
consumption [1], and lower patellofemoral joint reaction forces Therapy 1995;21:277–82.
[3] Bobath B. Adult hemiplegia: evaluation and treatment. London: Butterworth-
[2]. Motor learning mechanism is important to re-educate Heinemann; 1990.
impaired musculoskeletal function [18]. Hackney and Earhart [4] Yang YR, Wang RY, Yen LL, Lieu FK. Gait outcomes after additional backward
[17] demonstrated that backward direction generally more walking training in patients with stroke: a randomized controlled trial. Clinical
Rehabilitation 2005;19:264–73.
impacted to people with Parkinson disease (PD) than a secondary [5] Hackney ME, Earhart GM. Tai Chi improves balance and mobility in people
cognitive task, i.e. mental arithmetic task, during walking. with Parkinson disease. Gait and Posture 2008;28:456–60.
Especially, they also found that the gait asymmetry linked to [6] Van Deursen RWM, Flynn TW, McCrory JL, Morag E. Does a single control
mechanism exist for both forward and backward walking? Gait and Posture
deprived attentional resources allocated to gait [19] was notable in 1998;7:214–24.
those with PD in the simple backward direction. In addition, many [7] Thorstensson A. How is the normal locomotor program modified to produce
studies reported that joint angle and moment patterns were backward walking? Experimental Brain Research 1986;61:664–8.
[8] Grasso R, Bianchi L, Lacquaniti F. Motor patterns for human gait: backward
similar between FW and time-reversed BW [6–8]. EMG pattern
versus forward locomotion. Journal of Neurophysiology 1998;80:1868–85.
shift [7,8] and simpler shape of knee and hip joint moments could [9] Muddasir A, Nanda KTR. The effect of backward walking treadmill training on
be explained as the modification of neural mechanism [6]. Thus, kinematics of the trunk and lower limbs. Serbian Journal of Sports Sciences
BW could be helpful to the motor learning caused by the 2009;3:121–7.
[10] Vilensky JA. A kinematic comparison of backward and forward walking in
modification of the neural mechanism. humans. Journal of Human Movement Studies 1987;13:29–50.
A limitation of this study is that electromyography (EMG) of the [11] Winter DA. Biomechanics and motor control of human movement. Hoboken:
lower limb muscles was not included to analyze the BW Wiley; 2009.
[12] Benedetti MG, Catani F, Leardini A, Pignotti E, Giannini S. Data management in
mechanism. Further studies should be conducted to determine gait analysis for clinical applications. Clinical Biomechanics 1998;13:204–15.
the BW mechanism while considering EMG data of the lower limbs. [13] Winter DA. Kinematic and kinetic patterns in human gait: variability and
compensating effects. Human Movement Science 1984;3:51–76.
[14] Scott SH, Winter DA. Talocrural and talocalcaneal joint kinematics and kinetics
5. Conclusions during the stance phase of walking. Journal of Biomechanics 1991;24:743–52.
[15] Perry J, Gait Analysis:. Normal and pathological function. Thorofare: Slack
The aim of this study was to analyze gait mechanisms through Incorporated; 1992.
[16] Katsavelis D, Mukherjee M, Decker L, Stergiou N. Variability of lower extremity
acquisition of data on the kinematic and kinetic patterns of BW and joint kinematics during backward walking in a virtual environment. Nonlinear
FW. The characteristics of angular displacement in all joints were Dynamics Psychology and Life Sciences 2010;14:165–78.
almost identical in FW and time-reversed BW. However, selected [17] Hackney ME, Earhart GM. The effects of a secondary task on forward and
backward walking in Parkinson disease. Neurorehabil Neural Repair
crucial points of joint angles were significantly different. In BW,
2010;24:97–106.
range of motion of all joints was significantly smaller than in FW. [18] Purves D. Principles of neural development. Sunderland: Sinauer Associates;
The kinetic analysis of BW, which has rarely been studied 1989.
previously, was implemented in the current study. The moment [19] Yogev G, Plotnik M, Peretz C, Giladi N, Hausdorff JM. Gait asymmetry in
patients with Parkinson’s disease and elderly fallers: when does the bilateral
pattern of the ankle joint was almost identical in FW and time- coordination of gait require attention? Experimental Brain Research
reversed BW. In the knee and hip joints, the joint moment pattern 2007;177:336–46.