Accepted Manuscript

Title: Quantifying Coordination and Coordination Variability

in Backward versus Forward Running: Implications for
Control of Motion

Author: Sina Mehdizadeh Ahmed Reza Arshi Keith Davids

PII: S0966-6362(15)00466-X
DOI: http://dx.doi.org/doi:10.1016/j.gaitpost.2015.05.006
Reference: GAIPOS 4485

To appear in: Gait & Posture

Received date: 10-3-2015

Revised date: 10-5-2015
Accepted date: 11-5-2015

Please cite this article as: Mehdizadeh S, Arshi AR, Davids K, Quantifying
Coordination and Coordination Variability in Backward versus Forward
Running: Implications for Control of Motion, Gait and Posture (2015),
http://dx.doi.org/10.1016/j.gaitpost.2015.05.006

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Highlights

 This study compared coordination and coordination variability in backward and forward running

 Coordination of lower limb couplings in sagittal plane was quantified using CRP method.

t
 Coordination pattern in a gait cycle of backward running was in reverse to that of forward

ip
running

cr
 A single mechanism might be responsible for generating patterns in forward and backward

us
running

an
M
d
p te
ce
Ac

Page 1 of 22
Quantifying Coordination and Coordination Variability in Backward versus
Forward Running: Implications for Control of Motion

Sina Mehdizadeh (Corresponding Author) a, Ahmed Reza Arshi b, Keith Davids c

t
ip
a
Biomechanics and Sports Engineering Groups, Biomedical Engineering Department, Amirkabir

cr
University of Technology, Tehran, Iran. Address: Hafez Ave., Tehran, Iran. E-mail:

s.mehdizadeh@aut.ac.ir. Tel: +98-21-64542374.

us
an
b
Biomechanics and Sports Engineering Groups, Biomedical Engineering Department, Amirkabir

University of Technology, Tehran, Iran. Address: Hafez Ave., Tehran, Iran. E-mail: arshi@aut.ac.ir. Tel:
M
+98-21-64542377. Fax: +98-21-66468186.
d

c
Centre for Sports Engineering Research, Sheffield Hallam University UK and FiDiPro Programme,
te

University of Jyväskylä, Finland. Address: Centre for Sports Engineering Research, Faculty of Health and
p

Wellbeing, A217, Collegiate Hall, Collegiate Crescent, Sheffield S10 2BP, UK. Email:
ce

k.davids@shu.ac.uk. Tel: +61-7-31388744. Fax: +61-7-31383980.

Ac

Acknowledgment

The tests conducted for this study were supported by the Sports and Health Engineering Center and

performed at the Biomechanics Lab, University of Social Welfare and Rehabilitation Sciences.

Page 2 of 22
Abstract

The aims of this study were to compare coordination and coordination variability in backward and

forward running and to investigate the effects of speed on coordination variability in both backward and

t
forward running. Fifteen healthy male participants took part in this study to run forwards and backwards

ip
on a treadmill at 80%, 100% and 120% of their preferred running speeds. The coordinate data of passive

cr
reflective markers attached to body segments were recorded using motion capture systems. Coordination

of shank-foot and thigh-shank couplings in sagittal plane was quantified using the continuous relative

us
phase method. Coordination variability was calculated as the standard deviation of a coordination pattern

an
over 50 strides. Cross-correlation coefficients and associated phase shifts were determined to quantify

similarity in coordination patterns between forward and backward running. Our results demonstrated that

the coordination pattern in a gait cycle of backward running was in reverse to that of forward running at
M
all speeds implying that the same neural circuitry is responsible for regulating both forward and backward

running gaits. In addition, results demonstrated that there was an average of approximately 11% phase
d

shift between the coordination patterns of backward and forward running which indicates that a single
te

underlying mechanism might be responsible for generating motor patterns in both forward and backward
p

running. Finally, backward running had significantly higher magnitude of coordination variability
ce

compared to forward running, signifying that more degrees of freedom were involved in backward

running. Speed however, did not affect coordination variability in either task.
Ac

Keywords: Backward running; Coordination; Variability; Central pattern generators; Gait.

Page 3 of 22
Introduction

Backward running (BR) has been a popular rehabilitation and training routine or program to improve

t
coordination, endurance [1], muscular balance and strength [1, 2]. BR routines have also been used to

ip
enhance forward running (FR) performance [2]. BR is a complex motor skill with many interacting

cr
degrees of freedom which requires intralimb coordination to regulate performance of the task. Movement

coordination has been defined as the ability to assemble and maintain proper relations between segments

us
during motion [3]. However, while a number of studies have investigated coordination patterns in FR [4-

8], there are no studies, to the best of our knowledge, on movement coordination in BR and its relation to

FR.

an
M
Studying coordination in BR could reveal two important aspects of neuro-motor control of gait. First,

understanding coordination as an intrinsic property of the spinal neural circuitry (i.e. central pattern
d

generators, CPGs), could provide tangible insights on movement control [3, 9, 10]. Studies on human
te

walking have suggested that the same CPG network controlling forward locomotion may be reversed

during backward locomotion [11-14]. This suggestion is based on the fact that kinematic profiles for
p

forward locomotion are time-reversed during backward locomotion [13, 15]. However, most of these
ce

studies have considered the individual limb kinematics in their investigations and the interaction of limbs

and, thus, the intralimb coordination patterns during backward locomotion have rarely been investigated.
Ac

By looking at coordination patterns, one can understand how the movement system adapts to changing

task constraints (e.g. direction of motion and/or speed) [16]. The implication is that, if backward

locomotion is the reversal of forward locomotion, then the coordination of backward locomotion will be a

reverse of forward locomotion. This idea implies that the CPG neural circuitry regulates an appropriate

order parameter (e.g. intralimb coordination) to cause a transition in the direction of motion from forward

to backward.

Page 4 of 22
In a previous study, Grasso, Bianchi, and Lacquaniti [13] used planar covariation and cross-correlation

methods to compare coordination between forward and backward walking. Their results indicated that

spatial coordination is similar in both forward and backward walking. In addition, similar phase shifts

between the kinematics of the two walking patterns signified that a single underlying mechanism might

t
be responsible for generating both forward and backward walking. However, Grasso et al. [13] reported

ip
data on segmental angles and did not consider velocity information in their analysis. Moreover, the

cr
method of cross-correlation only quantifies the temporal relationship between the two segmental angles.

us
In addition to the coordination pattern, quantifying coordination variability in BR, and examining any

possible differences to FR, could provide substantial information on the regulation of gait. Strong

an
evidence suggests that coordination variability has a functional role in harnessing coordinative structures

in a movement system with redundant degrees of freedom providing adaptability to changing task
M
constraints [17, 18]. Because of the lack of visual feedback on heading orientation in BR, during

performance one might adopt different organizations of body segments leading to observations of higher
d

coordination variability in BR. This theoretical notion has some support from previous studies, which
te

have shown that backward walking is more variable than forward walking [19-21]. Furthermore, speed as

a task constraint, has been shown to affect coordination variability during forward locomotion in humans
p

[3, 22-25]. However, it is not currently well understood how speed affects coordination variability during
ce

BR and whether that differs from FR.

Ac

The aim of this study was, therefore, to investigate movement coordination and coordination variability in

BR and whether these processes differ from FR. We also investigated the effect of speed on coordination

variability in both BR and FR. Based on the finding of the previous studies, we proposed three

hypotheses: 1) the intralimb coordination pattern would be reversed in BR, 2) variability of coordination

would be higher in BR compared to FR and 3) speed change could affect coordination and its variability

in both BR and FR.

Page 5 of 22
Methods

Participants

t
ip
15 healthy male participants took part in this study. Their average age was 24.1±1.0 years with average

mass and height of 68.8±3.8 kg and 1.76±0.04 m, respectively. None of the participants suffered from any

cr
musculoskeletal injuries at the time of the experiment. All participants provided written informed consent

us
before participation in the study. The ethics committee of Amirkabir University of Technology approved

the experimental procedure.

an
Marker placement M
Seventeen passive reflective markers (14 mm diameter) were attached to the skin of each participant at

the right and left bony landmark on the second metatarsal head (toe), calcaneus (heel), lateral malleolus
d
(ankle), mid-tibia, lateral epicondyle of knee (knee), midthigh, anterior superior iliac spine and also on the

sacrum, midway between posterior superior iliac spines, 10th thoracic vertebrae (T10) and 7th cervical
te

vertebrae (C7).
p
ce

Task
Ac

Before starting the experiment, participants had enough time to familiarize themselves with running in

forward and backward directions on a treadmill. During the actual tests, all participants ran in forward and

backward directions on a motorized treadmill (Cosmed® T150, Rome, Italy) at their preferred running

speeds (PRS). PRS in each direction was recorded following a top-down and bottom-up approach similar

to protocols described in previous studies [25, 26]. During acquisition of the PRS data, participants were

not allowed to view the speed at which they were running on the treadmill [27]. The average PRS over all

Page 6 of 22
participants was 7.48±1.14 and 3.96±0.02 km/h for forward and backward running tasks, respectively.

Each participant was then asked to run for 2 minutes in each direction. Participants were not allowed to

use handrails in any of the running patterns. Sufficient rest periods were allocated between the tests to

allow participants to recover.

t
ip
Data recording

cr
The three-dimensional coordinate data of the markers were recorded using five Vicon® VCAM motion

capture calibrated cameras (Oxford Metrics, Oxford, UK) at the sampling frequency of 100

us
samples/second. Reconstruction and labeling were performed using Vicon® Workstation software

(Oxford Metrics, Oxford, UK).

Data analysis
an
M
Position data were first filtered using a fourth-order low-pass Butterworth filter with 10 Hz cutoff

frequency. Filtered data were used to calculate foot, shank and thigh angles with respect to the horizontal
d

in the sagittal plane as   arctan[( y d  y p ) / ( x d  x p )] , where subscripts p and d indicate proximal

te

and distal markers of the segment, respectively. Horizontal and vertical coordinates in the sagittal plane
p

are also represented by x and y, respectively. For all participants, segmental angles for individual strides
ce

were time-normalized to 100 data points. Each stride cycle was defined from heel contact to heel contact.

Heel contacts were identified as the minima in the ankle time series [28].
Ac

To quantify coordination, continuous relative phase was calculated using Hilbert transform as introduced

in a study by Lamb and Stöckl [29]. According to this method, segmental angles were first centered on

zero using equation (1):

centered (t i )   (t i )  min( (t ))  (max( (t ))  min( (t ))) / 2; (1)

Page 7 of 22
where centered (t i ) and  (t i ) are the zero-centered and original segment angles at data point

t i , respectively. In addition, min( (t )) and max( (t )) are minimum and maximum values of the

segmental angle, respectively. Second, each zero-centered time-series is transformed to an analytic signal

using Hilbert transform (equation (2)):

t
ip
 (t )  centered (t )  iH (t ); (2)

cr
where the Hilbert transform H (t ) of centered (t ) serves as the imaginary part of the analytic signal. Third,

us
phase angle at each time point t i ,  (t i ), is calculated using equation (3):

an
 (t i )  arctan(H (t i ) / centered (t i )); (3)

Finally, the continuous relative phase, CRP(t ), between two angle time series 1 (t ) and 2 (t ) is calculated
M
using equation (4):
d

CRP(t i )  1 (t i )  2 (t i ); (3)
te

The Hilbert transform produces analytic signals from non-sinusoidal signals, thus removing frequency
p

artifacts. This procedure makes the signal appropriate for studying inter- and intralimb coordination in
ce

human movement [29]. To quantify coordination variability, between-stride standard deviation of CRP

curve data points were calculated and averaged over the stride period (over all data points t i ) . For the
Ac

purpose of this study, CRP was calculated for the shank-foot and thigh-shank couplings by subtracting the

phase angles of distal joint from that of the proximal joint. A total number of 50 strides was considered

for the calculation of CRP and CRP variability.

To quantify similarity of CRP patterns, cross-correlation coefficients were calculated for each participant

[3]. The cross-correlation coefficient quantifies the temporal differences between CRP curves. A higher

cross-correlation coefficient indicates a better similarity between the two CRP curves [3, 30]. To compare

Page 8 of 22
coordination patterns between BR and FR, cross-correlation coefficients between the CRP curves of BR

and FR were determined at each running speed. In addition, the phase-shift (i.e. time lag) at which the

highest cross-correlations were obtained, was also determined [3, 12]. Finally, CRP variability was

calculated at each running speed and compared between BR and FR.

t
ip
Statistical analyses

cr
Separate one-way repeated measure analyses of variance (ANOVA) were performed to determine the

effect of speed on the cross-correlation of CRP curves of BR and FR in both shank-foot and thigh-shank

us
couplings. Here, the cross-correlation coefficient value and the time lag were set as dependent variables.

In addition, a two-way, mixed model (speed×direction) ANOVA, with repeated measures on the factor of

an
speed, was performed to compare CRP variability values between BR and FR at each speed. Here, CRP

variability was set as the dependent variable and direction (forward vs. backward) and speed were set as
M
factors. Statistical significance levels were set at P<0.05.
d
te

Results
p

Comparison of CRP patterns between BR and FR and the effect of speed

ce

According to Table 1, for both shank-foot and thigh-shank couplings, cross-correlation coefficients of

CRP patterns between BR and FR were close to -1 at all speeds, indicating that coordination patterns of
Ac

BR were the reverse of FR. The effect of speed was only significant in analyses of shank-foot coupling

(P=0.004). Post hoc analysis revealed that the difference between cross-correlation coefficients was only

significant between 80% and 120% PRS. For thigh-shank coupling, on the other hand, the influence of

running speed was not statistically significant (P=0.2). Moreover, there was no significant effect of speed

on time lag (P>0.05) in either of the couplings (Table 1). The range of time lags was -9 to -13% with the

Page 9 of 22
average of -10.9%. The CRP curves for shank-foot and thigh-shank couplings for the speed of 100% PRS

are depicted in Figure 1.

Comparison of CRP variability between BR and FR and the effect of speed

t
According to Table 2 and Figure 2, for shank-foot coupling, both speed and direction, and their

ip
interaction, revealed significant effects on CRP variability (P<0.05). Simple main effect analyses revealed

cr
that at all running speeds, BR resulted in significantly higher values of CRP variability compared to FR

(P<0.001; Figure 2). In addition, further simple main effect analyses indicated that there was no effect of

us
speed on CRP variability in FR (P>0.05). For BR, however, pairwise comparisons revealed that CRP

variability decreased from the speed of 100% to 120% PRS (P=0.001).

an
For thigh-shank coupling, the results demonstrated that there were no interaction effects (P=0.4). In
M
addition, while effects of speed were not statistically significant (P=0.05), the main effect of direction

was significant for CRP variability (P<0.001). Post hoc analysis revealed that, at all running speeds, BR
d

had significantly higher levels of CRP variability compared to FR (P<0.05; Figure 2).
te

Discussion
p

This study investigated coordination and coordination variability in BR, and how they differ with FR. We
ce

also investigated the effect of speed on variability of coordination in both BR and FR. Our findings

demonstrated that the pattern of coordination in BR was the reverse of FR at all speeds. In addition, BR
Ac

revealed higher magnitudes of coordination variability compared to FR. Speed, however, did not affect

coordination variability in either of the tasks.

Comparison of coordination patterns during BR and FR

Although, there are differences in opinion on the issue [31], several studies that have compared

kinematics of forward with backward locomotion suggest that the same neural circuit (i.e. CPG) that

generates forward locomotion, is responsible for generating backward locomotion as well [12, 13].

10

Page 10 of 22
However,it has been suggested that coordination is an intrinsic property of spinal neural circuitry for

movement control [9]. Therefore, studying coordination could provide substantial information about the

neural control of locomotion [32]. Our results which indicated that coordination patterns are reversed in

BR compared to FR (Table 1), complement those of previous studies. That is, it has been suggested that

t
the multi-segment motion of mammalian locomotion is controlled by a network of coupled oscillators in

ip
central pattern generators, and that any changes in inter-oscillator phase coupling, would generate

cr
changes in gait such as a reversal in gait direction [13, 33]. Our results suggest that such changes in the

phase coupling could be exemplified by the CRP pattern. This result is also in line with previous

us
arguments which suggested that legs are typically controlled in a coordinative structure [12], the

an
regulation of which could be reversed to generate backward locomotion.

Furthermore, the data reported here demonstrated that there was an average of approximately 11% phase
M
shift between the coordination patterns of BR and FR. A previous study on forward and backward

walking [12], reported that there were phase shifts of 14.3% and 25% for movement and muscle
d

activation patterns of the knee joint, respectively. Our study suggested that small phase shifts emerging
te

between forward and backward locomotion might be indicative of a single underlying mechanism for

generating both forward and backward locomotion, which is supported by the findings of van Deursen,
p

Flynn, McCrory, and Morag [12]. This result was also reported by Grasso et al., [13] and Lacquaniti,
ce

Grasso, and Zago [34] which indicated that the phase coupling between segmental angles remained

unaltered when walking direction was reversed. However, while Grasso et al., [13] and Laquintini et al.
Ac

[34] used cross-correlations of two angle time series in their study, we used cross-correlations of CRP

patterns of FR and BR. Our study thus adds to previous findings by demonstrating that the coordination

patterns are also maintained with the reversal of movement direction.

Comparison of CRP variability between BR and FR and the effect of speed

11

Page 11 of 22
Our results indicated that BR displayed higher levels of CRP variability compared to FR in both shank-

foot and thigh-shank couplings (Table 2 and Figure 2). This observation is in line with data from previous

studies where it has been shown that the behaviour of backward locomotion has more variable than

forward locomotion [19-21]. One possible reason for higher variability in backward locomotion might be

t
due to the lack of availability of visual information on heading orientation in backward locomotion [19].

ip
However, in one study [21], it was shown that availability of visual feedback had no significant effects on

cr
levels of variability in backward walking. Therefore, our data could be interpreted to suggest that

backward locomotion is inherently more variable than forward locomotion due to backward locomotion

us
being a more challenging task for the nervous system to regulate [20]. It could be suggested that the

an
regulation of BR requires more degrees of freedom to re-organise, which results in higher CRP

variability. This finding demonstrates the functional role of movement pattern variability in adapting to
M
changing task constraints, which have some novelty for participants. This interpretation of the data is

supported by results of studies comparing electromyographic (EMG) activities during forward and
d
backward locomotion, which revealed greater EMG activity during backward locomotion [11, 13].
te

One possible reason of this greater muscular activity might thus be due to the greater number of degrees

of freedom, which in turn, was manifested by higher variability. In contrast to data from the study of
p

Katsavelis, et al. [21], which argued that greater muscular activity results in higher magnitude of
ce

variability, we believe that the direction of this relationship might be reversed. That is, the need for

controlling excessive degrees of freedom in backward locomotion, as realized by higher variability,

Ac

requires greater muscular activity. The higher magnitude of CRP variability for BR in our study might

thus indicate greater muscular activity. This finding is important when considering data from studies

which have suggested that BR should be part of an appropriate rehabilitation and training program to

increase muscular balance and strength [1, 2]. To summarize, as mentioned earlier, regulation of

backward running is facilitated by coordination variability which has a functional role in providing

adaptability to changing task constraints [17, 18]. For example, a study on backward walking [35]

12

Page 12 of 22
reported that people with anterior cruciate ligament (ACL) deficiency had less variability in their knee

joints compared to healthy individuals. The authors concluded that less variability of ACL deficient

people could imply decreased functional responsiveness to the environmental changes which could lead

to knees being more prone to injury. Therefore, the higher level of variability of backward running in our

t
study could allow the movement system harness coordination variability in adapting an available

ip
coordination pattern to prevent further injuries. Furthermore, since our results showed that CRP

cr
variability decreased from the speed of 100% to 120% PRS of shank-foot coupling only (Table 2), it

could not be concluded that speed affected coordination variability in BR.

us
Study limitations

an
In this study, coordination, and its variability, was quantified under the task constraints of treadmill

running. In a recent study, it was shown that using a treadmill might affect the pattern and variability of
M
coordination in walking [36]. However, no published results have so far been elicited in studies

comparing coordination and its variability in running under the specific task constraints of both treadmill
d

and over-ground running. In addition, although participants of this study had enough time to become
te

familiar with the BR task, the pattern and variability of coordination of backward running might have
p

been influenced by the novelty of this task.

ce

Conclusion
Ac

The findings of this study suggest that the same neural circuitry might be responsible for generating the

pattern dynamics to regulate forward and backward running gaits. In addition, the small phase shift

between the coordination patterns of backward and forward running indicates that a single underlying

mechanism is likely responsible for generating both forward and backward running. Finally, the finding

of greater levels of variability for backward running potentially demonstrates that excessive degrees of

freedom are involved in the performance of backward running which might require greater muscular

13

Page 13 of 22
activity to be controlled. This observation could explain the underlying reason why backward running has

been established as an appropriate rehabilitation and training program to increase muscular strength.

Conflict of interest statement

t
There is not any type of conflict of interest associated with this manuscript.

ip
cr
References

us
[1] Flynn TW, Soutas-Little RW. Mechanical power and muscle action during forward and backward

running. J Orthop Sports Phys Ther 1993; 17(2): 108-12.

an
[2] Threlkeld AJ, Horn TS, Wojtowicz G, Rooney JG, Shapiro R. Kinematics, ground reaction force,

and muscle balance produced by backward running. J Orthop Sports Phys Ther 1989; 11(2): 56-
M
63.

[3] Chiu SL, Chou LS. Effect of walking speed on inter-joint coordination differs between young and
d

elderly adults. J Biomech 2012; 45(2): 275-80.

te

[4] Hein T, Schmeltzpfenning T, Krauss I, Maiwald C, Horstmann T, Grau S. Using the variability of

continuous relative phase as a measure to discriminate between healthy and injured runners. Hum
p

Mov Sci 2012; 31: 683-94.

ce

[5] Miller RH, Meardon S, Derrick TR, Gillette JC. Continuous relative phase variability during an

exhaustive run in runners with a history of iliotibial band syndrome. J Appl Biomech 2008; 24(3):
Ac

262-70.

[6] Dierks TA, Davis I. Discrete and continuous joint coupling relationships in uninjured recreational

runners. Clin Biomech 2007; 22(5): 581-91.

[7] Cunningham TJ, Mullineaux DR, Noehren B, Shapiro R, Uhl TL. Coupling angle variability in

healthy and patellofemoral pain runners. Clin Biomech (Bristol, Avon) 2014; 29(3): 317-22.

14

Page 14 of 22
[8] Ivanenko YP, Cappellini G, Dominici N, Poppele RE, Lacquaniti F. Modular control of limb

movements during human locomotion. J Neurosci 2007; 27(41): 11149-61.

[9] Eilam D, Shefer G. Reversal of interleg coupling in backward locomotion implies a prime role of

the direction of locomotion. J Exp Biol 1992; 173: 155-63.

t
[10] Barela JA, Whitall J, Black P, Clark JE. An examination of constraints affecting the intralimb

ip
coordination of hemiparetic gait. Hum Mov Sci 2000; 19: 251-73.

cr
[11] Thorstensson A. How is the normal locomotor program modified to produce backward walking?

Exp Brain Res 1986; 61: 664-8.

us
[12] van Deursen RW, Flynn TW, McCrory JL, Morag E. Does a single control mechanism exist for

an
both forward and backward walking? Gait Posture 1998; 7(3): 214-24.

[13] Grasso R, Bianchi L, Lacquaniti F. Motor patterns for human gait: Backward versus forward
M
locomotion. J Neurophysiol 1998; 80: 1868-85.

[14] Earhart GM, Jones GM, Horak FB, Block EW, Weber KD. Forward versus backward walking:
d
Transfer of podokinetic adaptation. J Neurophysiol 2001; 86: 1666-70.
te

[15] Meyns P, Molenaers G, Desloovere K, and Duysens J. Interlimb coordination during forward

walking is largely preserved in backward walking in children with cerebral palsy. Clin
p

Neurophysiol 2013; 125(3): 552-61.

ce

[16] Stergiou N, Jensen JL, Bates BT, Scholten SD, Tzetzis G. A dynamical systems investigation of

lower extremity coordination during running over obstacles. Clin Biomech (Bristol, Avon) 2001;
Ac

16(3): 213-21.

[17] Seifert L, Button C, Davids K. Key properties of expert movement systems in sport: An

ecological dynamics perspective.Sports Med 2013; 43(3): 167-78.

[18] Davids K, Glazier P, Araju D, Bartlett RM. Movement systems as dynamical systems. Sports

Med 2003; 33(4): 245-60.

15

Page 15 of 22
[19] Hoogkamer W, Massaad F, Jansen K, Bruijn SM, Duysens J. Selective bilateral activation of leg

muscles after cutaneous nerve stimulation during backward walking. J Neurophysiol 2012;

108(7): 1933-41.

[20] Kurz MJ, Wilson TW, Arpin DJ. Stride-time variability and sensorimotor cortical activation

t
during walking. Neuroimage 2012; 59(2): 1602-7.

ip
[21] Katsavelis D, Mukherjee M, Decker LM, Stergiou N. Variability of lower extremity joint

cr
kinematics during backward walking in a virtual environment. Nonlinear Dynamics Psychol Life

Sci 2010; 14(2): 165-78.

us
[22] Seay JF, Van Emmerik RE, Hamill J. Low back pain status affects pelvis-trunk coordination and

an
variability during walking and running. Clin Biomech (Bristol, Avon) 2011; 26(6): 572-8.

[23] Lamoth CJC, Daffertshofer A, Meijer OG, Beek PJ. How do persons with chronic low back pain
M
speed up and slow down? Trunk-pelvis coordination and lumbar erector spinae activity during

gait. Gait Posture 2006; 23(2): 230-9.

d
[24] Lamoth CJC, Daffertshofer A, Huys R, Beek PJ. Steady and transient coordination structures of
te

walking and running. Hum Mov Sci 2009; 28(3): 371-86.

[25] Jordan K, Challis JH, Cusumano JP, Newell KM. Stability and the time-dependent structure of
p

gait variability in walking and running. Hum Mov Sci 2009; 28(1): 113-28.
ce

[26] Dingwell JB, Marin LC. Kinematic variability and local dynamic stability of upper body motions

when walking at different speeds. J Biomech 2006; 39(3): 444-52.

Ac

[27] Jordan K, Challis JH, Newell KM. Speed influences on the scaling behavior of gait cycle

fluctuations during treadmill running. Hum Mov Sci 2007; 26(1): 87-102.

[28] Li L, van den Bogert ECH, Caldwell GE, van Emmerik RE, Hamill J. Coordination patterns of

walking and running at similar speed and stride frequency. Hum Mov Sci 1999; 18(1): 67-85.

[29] Lamb PF and Stöckl M. On the use of continuous relative phase: Review of current approaches

and outline for a new standard. Clin Biomech (Bristol, Avon) 2014; 29(5): 484-93.

16

Page 16 of 22
[30] Haddad JM, van Emmerik RE, Whittlesey SN, Hamill J. Adaptations in interlimb and intralimb

coordination to asymmetrical loading in human walking. Gait Posture 2006; 23(4): 429-34.

[31] Meyns P, Desloovere K, Molenaers G, Swinnen SP, Duysens J. Interlimb coordination during

forward and backward walking in primary school-aged children. PloS One 2013; 8(4): e62747-

t
e62747.

ip
[32] Kelso JAS. Dynamic patterns: The self-organization of brain and behavior. 1995, Cambridge,

cr
MA: MIT.

[33] Musienko PE, Zelenin PV, Lyalka VF, Gerasimenko YP, Orlovsky GN, Deliagina TG. Spinal

us
and supraspinal control of the direction of stepping during locomotion. J Neurosci 2012; 32(48):

an
17442-53.

[34] Lacquaniti F, Grasso R, Zago M. Motor patterns in walking. News Physiol Sci 1999; 14: 168-74.
M
[35] Zampeli F, Moraiti CO, Xergia S, Tsiaras VA, Stergiou N, Georgoulis AD. Stride-to-stride

variability is altered during backward walking in anterior cruciate ligament deficient patients.
d

Clin Biomech (Bristol, Avon) 2010; 25(10):1037-41.

te

[36] Chiu SL, Chang CC, Chou LS. Inter-joint coordination of overground versus treadmill walking in
p

young adults. Gait Posture 2015; 41(1): 316-8.

ce
Ac

17

Page 17 of 22
Tables:

Table 1: Mean (standard deviation) and ANOVA results of cross-correlation coefficients and phase shifts

t
ip
for shank-foot and thigh-shank couplings.

cr
speed (%PRS) ANOVA
80 100 120 F P η2

us
Cross-correlation coefficient
shank-foot -0.90(0.04) -0.86(0.08) -0.85(0.06) 6.92 0.004 0.33
thigh-shank -0.87(0.04) -0.88(0.03) -0.87(0.03) 1.35 0.2 0.08
Phase shift (%)

an
shank-foot -12.00(6.58) -10.73(7.76) -12.13(6.25) 0.56 0.5 0.03
thigh-shank -9.53(5.22) -9.00(5.32) -12.33(6.66) 2.52 0.1 0.15
M
d
p te
ce
Ac

18

Page 18 of 22
 Table 2: Mean (standard deviation) and ANOVA results of continuous relative phase (CRP) variability of

shank-foot and thigh-shank couplings for forward (FR) and backward (BR) running.

t
ip
speed (%PRS) ANOVA
speed direction speed×direction

cr
80 100 120
F P η F
2
P η 2
F P η2
FR 3.23(0.79) 3.13(1.69) 2.75(1.45)
shank-foot 11.36 <0.001 0.44 83.44 <0.001 0.85 9.93 0.001 0.41

us
BR 9.84(3.43) 9.18(2.22) 7.09(1.39)
FR 5.21(1.29) 4.56(1.37) 4.84(2.40)
thigh-shank 3.78 0.05 0.21 169.14 <0.001 0.92 0.76 0.4 0.05
BR 12.48(3.12) 11.45(2.15) 10.79(2.18)

an
M
d
p te
ce
Ac

19

Page 19 of 22
Figure captions:

Figure 1: Continuous relative phase (CRP) curves of forward (FR) and backward (BR) running over the

t
gait cycle for shank-foot (left) and thigh-shank (right) couplings for one participant running at preferred

ip
running speed.

cr
us
Figure 2: Continuous relative phase (CRP) variability of forward (FR) and backward (BR) running for

shank-foot (left) and thigh-shank (right) couplings. Error bars are standard deviations. Asterisks show

an
significant difference between FR and BR.
M
d
p te
ce
Ac

20

Page 20 of 22
 350
BR

t
FR 300

ip
100
250

200
50
) )

cr
ees ees 150
rg gr
e e
d( 0
d( 100
P P
R R
C C 50

us
-50 0

-50

-100 -100

an
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Gait cycle (%) Gait cycle (%)

Figure 1
M
d
p te
ce
Ac

21

Page 21 of 22
 t
ip
cr
us
an
M
Figure 2
d
te
p
ce
Ac

22

Page 22 of 22