See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259588291

The mechanics of jumping over an obstacle during running: A comparison

between athletes trained to hurdling and recreational runners

Article  in  European Journal of Applied Physiology · January 2014

DOI: 10.1007/s00421-013-2805-6 · Source: PubMed

CITATIONS READS

4 1,308

3 authors:

Geoffrey Mauroy Benedicte Schepens

Université Catholique de Louvain - UCLouvain Université Catholique de Louvain - UCLouvain
8 PUBLICATIONS   40 CITATIONS    49 PUBLICATIONS   1,547 CITATIONS   

SEE PROFILE SEE PROFILE

Patrick A Willems
Université Catholique de Louvain - UCLouvain
87 PUBLICATIONS   3,065 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Control of Posture and Movement During Reaching in the Cat View project

Hurdles View project

All content following this page was uploaded by Patrick A Willems on 02 October 2014.

The user has requested enhancement of the downloaded file.

The mechanics of jumping over an
obstacle during running: a comparison
between athletes trained to hurdling and
recreational runners

G. Mauroy, B. Schepens & P. A. Willems

European Journal of Applied

Physiology

ISSN 1439-6319
Volume 114
Number 4

Eur J Appl Physiol (2014) 114:773-784

DOI 10.1007/s00421-013-2805-6

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag Berlin Heidelberg. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784
DOI 10.1007/s00421-013-2805-6
Original Article
The mechanics of jumping over an obstacle during running:
a comparison between athletes trained to hurdling
and recreational runners
G. Mauroy · B. Schepens · P. A. Willems 
Received: 14 December 2012 / Accepted: 16 December 2013 / Published online: 5 January 2014
© Springer-Verlag Berlin Heidelberg 2014
Abstract  subsequently, Wext during contact before the obstacle is 10–
Purpose This study compares the mechanism of running
in trained athletes (TA) experienced in hurdling and in rec-
reational runners (RR), as they approach and jump over an
obstacle.
Methods The movements of the centre of mass of the
body (COM), the external muscular work (Wext) and
the leg-spring stiffness (kleg) were evaluated in athletes
approaching an obstacle at 18 km h−1, from the ground
reaction forces (measured by force-platforms) and the ori-
entation of the lower-limb segments (measured by camera).
These results were compared to those obtained in RR.
Results Two steps before the obstacle, kleg is reduced by
10–20 %; so, the COM is lowered and accelerated forward.
During the step preceding the obstacle, kleg is increased by
40–60 %; so the COM is raised and accelerated upwards,
whereas its forward velocity is reduced. This change in the
running pattern is similar to the one observed in RR while
leaping an obstacle. However, in TA, the change in stiff-
ness is less pronounced. As a result, the orientation of the
velocity vector at the beginning of the aerial phase over the
obstacle is more horizontal than in RR, which involves a
10–20 % greater horizontal velocity and a 40–60 % smaller
vertical excursion of the COM when crossing the obstacle;
Communicated by Jean-René Lacour.
Electronic supplementary material The online version of this
article (doi:10.1007/s00421-013-2805-6) contains supplementary
material, which is available to authorized users.
G. Mauroy · B. Schepens · P. A. Willems (*)  front leg (i.e. the leg that goes first over the hurdle) remains
Laboratoire de physiologie et biomécanique de la locomotion,
Institute of NeuroScience, Université catholique de Louvain,
Place Pierre de Coubertin, 1, 1348 Louvain‑la‑Neuve, Belgium
e-mail: patrick.willems@uclouvain.be
20 % less.
Conclusion  Athletes use the same mechanisms as non-
specialists to cross an obstacle. However, athletes adapt the
mechanism of jumping to reduce the loss in the velocity of
progression when crossing an obstacle.
Keywords Running · Jumping · Hurdle ·
Mechanical work · Leg stiffness
Introduction
In a recent study, Mauroy et al. (2012) have analysed a
situation that often occurs during cross-country running or
jogging: a person moving at a slow or intermediate speed
comes across an obstacle, jumps over it and continues to
run at the same pace. To leap across the obstacle, the run-
ner increases the vertical velocity (Vv) of the centre of mass
of the body (COM) at takeoff to lengthen the duration and
amplitude of the aerial phase. This increase in Vv is done
at the expense of the horizontal velocity (Vf) of the COM.
When a recreational runner crosses a <1-m-high obstacle,
Vv at takeoff is up to 800 % greater than during steady-state
running and Vf is up to 21 % lower (Mauroy et al. 2012).
The higher the obstacle and the faster the approaching
speed, the greater are these differences.
Hurdlers have developed a particular movement pattern
to minimize the vertical excursion of the body over the hur-
dle and so to limit the loss in its horizontal velocity (Mann
and Herman 1985). When athletes cross the hurdle, the
fairly straight and the hip of the back leg is positioned in
abduction with the knee flexed, so that the thigh is paral-
lel to the hurdle, as close as possible to its top. McDonald
13
 Author's personal copy
774
and Dapena (1991b) have evaluated the average velocity
of the COM of athletes during a hurdle sprint, by means of
cameras. At takeoff of the hurdle step, Vv is ~250 % greater
and Vf is ~5 % smaller than during normal running steps. As
compared to recreational runners (RR), trained athletes (TA)
are thus able to lessen the increase in Vv and the loss in Vf.
To jump over an obstacle, RR modify the bouncing mech-
anism of running during the two steps preceding the obstacle
(Mauroy et al. 2012). At the second of the last contact before
the obstacle, the overall stiffness of the leg-spring system
is decreased, as compared to steady-state running. Conse-
quently, the COM is lowered and accelerated forward. Since
Vv at takeoff is close to zero, the following aerial phase is
short and Vv at touchdown of the next step is small. During
the contact before the obstacle, the stiffness is increased and
the velocity vector shifts to a more vertical direction. Con-
sequently, during this last contact, the COM is raised and
accelerated upwards, whereas its forward velocity is reduced.
In TA as in RR, Vv is reduced at the end of penultimate
contact before the hurdle and the following airborne phase is
shorter (McDonald and Dapena 1991a). Consequently, the
magnitude of Vv at the start of the support phase before the
hurdle is small. At the end of this last contact, Vv is increased
and Vf is reduced. However, the increase in Vv and the loss in
Vf are smaller than in RR. To our knowledge, no study has yet
analysed the bouncing mechanism of TA on approaching and
jumping over an obstacle, especially the change in the overall
stiffness of the leg-spring system and the external work done
during the steps preceding the crossing over the obstacle.
The main question raised in this study is: How do trained
athletes modify the mechanism used by untrained runners to
reduce the vertical excursion of the COM and the loss in the
horizontal velocity? Here, we have analysed the bouncing
mechanism of running of athletes trained in hurdling while
approaching and jumping over an obstacle at sub-maximal
speed. We have measured the forces exerted by the feet on
the ground and the orientation of the lower-limb segments
during the steps preceding the leap over an obstacle. From
these data, we have evaluated the movements of the COM, by
time integration of the forces, the external work done to move
the COM relative to the surroundings and the stiffness of the
bouncing system, by computer simulation. The bouncing
mechanism of athletes using the hurdle technique was com-
pared to recreational runners jumping over an obstacle of the
same height while running at the same approaching speed.
Methods
Subjects and experimental procedure
Experiments were carried out on six male athletes,
specialists in hurdle sprint (age 21.6 ± 2.8 years,
13
Eur J Appl Physiol (2014) 114:773–784
mass 73.8 ± 3.4 kg, height 1.85 ± 0.03 m, leg length
0.95  ± 0.03 m). These hurdlers have a Belgian national
level (110 m hurdle race in <15 s) and practise athletics
for at least 10 years. Written informed consent of the sub-
jects was obtained. Experiments involved no discomfort,
were performed according the Declaration of Helsinki and
approved by the local ethics committee.
All athletes wore running shoes. The length of the track
did not allow running at very high speeds and certainly not
at the maximal speed that these athletes could reach. There-
fore, they were asked to run at 18 km h−1 across a force
platform and to jump over a hurdle placed at three different
heights: 0.85 m (for comparison with recreational runners),
0.91 m (height of the hurdles in the men’s 400 m race) and
1.07 m (height of the hurdles in the men’s 110 m sprint).
They were also instructed to continue to run at the same
pace after the obstacle (note that there was no control on
the speed after the obstacle). Four to five steps before the
obstacle were recorded. Traces were analysed if the aver-
age velocity of the first step(s) ranged between 17.5 and
18.5 km h−1. Control experiments, i.e. runs without any
obstacle, were also recorded on the same subjects at the
same speed. For each hurdle height, two to six trials per
subject were recorded in each condition.
The results of the TA were compared with those of seven
RR, not familiar with the hurdle technique who participated in
a previous study (Mauroy et al. 2012). In this set of “control
data”, RR were running at the same speed as TA (18 km h−1)
and leaped 0.45-, 0.65- and 0.85-m-high obstacles.
Experimental setup
The experimental setup and the data analysis are explained
in detail in Mauroy et al. (2012) and only described
briefly here. The three components of the ground reaction
force under the feet were measured each 1 ms by means
of a 13-m-long force platform placed in the middle of a
33-m-long running track. The platform was composed of
13 plates (1 m× 1 m) described in detail by Genin et al.
(2010). A hurdle was mounted 3 m before the end of the
force platform. The distance between the start of the track
and the hurdle was 23 m. Two pairs of photocells were
placed at each end of the plates at the level of the neck of
the subject. A vertical mattress helped the subject to decel-
erate at the end of the running track.
The orientation of the lower limbs in the sagittal plane
at touchdown and at takeoff was measured by means of a
high-speed video camera (BASLER A501k, resolution
1,280  × 1,024 pixels, aperture time 3 ms). Images were
sampled at a rate of 100 frames s−1. Camera images were
synchronized with the force traces by means of a lead that
switched on, when the subject passed in front of the first
pair of photocells. Reflective markers were attached with
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784 775
Fig. 1  Typical traces from a runner during the five steps preceding a
jump over a 0.85-m-high obstacle at 18 km h−1. The right traces are
experimental data from one trained athlete (71 kg, 1.87 m, 25-year-
old) and the left traces from one recreational runner (64 kg, 1.83 m,
27-year-old). The Ekf-curve represents the kinetic energy due to the
forward velocity of the COM; the Ep  +  Ekv-curve (continuous line)
is the energy due to the vertical movement of the COM: it is the sum
of its gravitational potential energy (Ep, dashed line) plus its kinetic
self-adhesive tape on each limb, as described in Mauroy
et al. (2012). The angle made by the lower limbs with the
vertical axis was measured on each frame from the coordi-
nates of the reflectors, using the Lynxzone software (Arsa-
lis, Belgium).
Data analysis
Steps were numbered as follows: step 0 corresponds to
the last contact before the obstacle followed by the aerial
phase over the obstacle, step −1 is the step before step 0,
etc. (Fig. 1). Step −4 includes the average values of all the
steps prior to step −3. The control step includes the aver-
age value of all the steps obtained during running without
obstacle.
Calculations and data processing were performed using
custom software (LABVIEW 8.2, National Instruments,
Austin, TX, USA). The acceleration, velocity and displace-
ment of the COM were measured from the vertical and
fore–aft components of the ground reaction force using the
method of Cavagna (1975) and Schepens et al. (1998). The
lateral component of this force was neglected. From these
variables, the kinetic energy due to the horizontal velocity
of the COM (Ekf), to its vertical velocity (Ekv) and to its
gravitational potential energy (Ep) was then calculated. The
energy due to its vertical velocity (Ekv). The Ecom-curve is the exter-
nal mechanical energy of the COM, i.e. the sum Ekf + Ep + Ekv. Each
step is delimited by heel strike (vertical dotted line) and begins with a
contact phase followed by an aerial phase. Step 0 contains the contact
phase of the impulse before the obstacle and the aerial phase over the
obstacle, step −1 is the step before step 0, step−2 is the step before
step −1, etc
total energy of COM is given by Ecom = Ekf + Ep + Ekv.
The increments in the Ekf-curve, the Ep + Ekv-curve and the
Ecom-curve represent, respectively, the muscular positive
work done to accelerate the COM forward (W + f ), to elevate
and accelerate the COM vertically (W + v ) and to sustain the
movements relative to the surroundings (external work,
+
Wext ). Similarly, the decrements of the Ekf-curve, Ep + Ekv-
curve and Ecom-curve represent the negative work done to
decelerate the COM forward (W − f ), to lower and decelerate
the COM vertically (W − v ) and the negative external work
(W ext). The net muscular work (Wnet) done during the step
−
was calculated as the difference between the positive and
the negative external work done to move the COM during
the contact phase, i.e. as the difference between Ecom at the
end and at the beginning of the contact period.
To estimate the overall stiffness (kleg) generated by the
muscle–tendon units, the lower limb was modelled as a
simple linear spring that swept on an arc during the con-
tact (Mauroy et al. 2012). In this model, initial and final
conditions were determined from the experimental traces:
orientation (θtd) and length (ltd) of the leg spring at touch-
down, the instantaneous forward and vertical veloci-
ties of the COM at touchdown and the orientation (θto) of
the leg spring at takeoff. The trajectory of the COM dur-
ing the contact time was then calculated using computer
13
 Author's personal copy
776
simulation: during the first half of the contact, the leg
spring shortened (loading phase), and during the second
half of the contact, the leg spring lengthened (unloading
phase). The simulation was stopped when the leg reached
its initial length, ltd. The stiffness kleg was estimated by
successive approximations. At the end of the simulation,
the angle between the leg spring and the vertical should
reach a value of θto (± 0.002°) as measured experimen-
tally. If the value of θto was not reached, the simulation was
redone with a new value of kleg, and so on until a proper
value of θto was reached. This model was validated in Mau-
roy et al. (2012): we showed that the values of leg stiffness
kleg obtained during the control steps of RR were in good
agreement with those of Farley and Gonzalez (1996), He
et al. (1991), McMahon and Cheng (1990) and Morin et al.
(2005). Furthermore, the numerical analysis reproduced
accurately the trajectory of the COM during contact when
running in a steady state and when approaching the obsta-
cle (see upper panels of Fig. 2 in Mauroy et al. 2012).
To validate the model during control steps, step −1 and
step 0, the experimental values were compared with those
obtained by simulation. For most of the variables tested, there
was a good agreement between simulation and experiment
(see electronic supplementary material). For control steps, the
maximal ground reaction force obtained in the model differed
from the experimental value by −4.1 ± 12.8 % (mean ± SD,
n = 62) and the time of contact (tc) by −7.5 ± 4.4 %. The
horizontal and vertical velocities of the COM at takeoff were,
respectively, 1.4 ± 2.0 % greater and 12.7 ± 5.18 % smaller
than the experimental value. When approaching an obstacle,
the maximal ground reaction force obtained by the simula-
tion differed from the experimental value by −4.6 ± 18.4 %
(n = 82) for step −1 and by 0.5 ± 17.6 % (n = 82) for step 0;
tc presented a difference of −6.9 ± 3.7 % for step −1 and of
15.3 ± 4.1 % for step 0. The horizontal velocity of the COM
at takeoff presented a difference of −0.2 ± 5.9 % for step −1
and −1.6 ± 4.9 % for step 0. During step −1, the difference
in the vertical velocity of the COM at takeoff was less than
−0.17 ± 0.42 m s−1 (since the velocity at takeoff is close to
zero, the difference cannot be expressed in %). During step 0,
this difference was −3.7 ± 25.2 %. The discrepancy observed
in tc, especially during step 0, arises mainly from the absence
of an ankle joint in the model, which implies that the leg-
spring length at the end of the contact phase cannot exceed
its initial length at touchdown. In reality, our experimental
results show that the leg extends beyond the initial length;
for all the steps analysed, the distance measured at takeoff
between the great trochanter and the head of the fifth meta-
tarsal joint was 2.7 ± 4.6 % (n = 226) greater than the initial
length. As a consequence, the simulation underestimated the
duration of the push and in turn the time of contact. Another
reason for the inaccuracies of the model is that it does not
take into account the movement of the centre of pressure
13
Eur J Appl Physiol (2014) 114:773–784
under the foot during the contact phase. Furthermore, the
model supposes a totally passive spring–mass system, which
stores and releases mechanical energy during contact. In real-
ity, as shown in “Results”, additional muscular work is done
during the last contact to increase the energy at takeoff before
the obstacle.
A sensitivity analysis was realized to test the robust-
ness of the results of the simulation. This analysis was
performed on both groups on the control steps, on step −1
and step 0 while jumping over a 0.85-m-high barrier. We
used the OFAT (one-factor-at-a-time) method: assuming a
typical error of ±10 % in the measurements. The following
factors were changed one at a time by ±10 % of their aver-
age value: orientation (θtd) and length (ltd) of the leg spring
at touchdown, the instantaneous forward (Vf,td) and vertical
(Vv,td) velocities of the COM at touchdown and the orienta-
tion (θto) of the leg spring at takeoff. When changing one
variable at a time, the average change in kleg is ~7 % of the
experimental value and the maximal change is in the order
of ±1 SD in each group and for each type of step.
Statistics
The analysed steps were classified according to the step
number (control step, step 0 and step −1). For each subject,
the trials obtained on the same step number in each height
class were averaged. The results of all subjects obtained
in the same class were then averaged. The results are pre-
sented in classes according to the step number and the
obstacle height. Statistics were realized with the program
IBM SPSS Statistics 21. All variables tested were distrib-
uted normally (Kolmogorov–Smirnov test). Depending on
the case, a Student’s t test, a one-way ANOVA (Bonferroni
post hoc) or a two-way repeated measures ANOVA was
performed to evaluate the effect of the step number or of
the obstacle height on the variables studied.
Results
Figure  1 presents a typical trace of the energy–time
curves of the COM when approaching a 0.85-m obstacle
at 18 km h−1 and leaping over it: the experimental traces
obtained on one TA familiar with the hurdling technique
are presented in the right panel and on one RR who never
learned hurdling in the left panel.
In TA, before step −1, the Ekf, Ep + Ekv and Ecom-curves
are in phase and are comparable to the energy-time curves
of a spring–mass system bouncing on the ground (Cavagna
et al. 1976). The phases during which the Ekf, Ep + Ekv and
Ecom-curves are constant represent the aerial phase of the
step. During the contact period of step −1, the muscle–
tendon units perform negative work when the Ecom-curve
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784 777
Fig. 2  Changes in the approaching speed during the steps preced-
ing the obstacle. On the abscissa, step 0 corresponds to the forward
velocity of the COM during the aerial phase over the obstacle, step
−1 to the velocity during the last aerial phase before the obstacle, etc.
The filled circles, diamonds and squares represent the mean value
of the data obtained on trained athletes approaching, respectively, a
0.85-, 0.91- and 1.07-m-high obstacle at 18 km h−1 (~5 m s−1). The
‘×’ symbols, ‘+’ symbols and circles represent the mean value of
the data obtained by Mauroy et al. (2012), on recreational runners
approaching a 0.45-, 0.65- and 0.85-m-high obstacle at 18 km h−1
(~5 m s−1) and 9 km h−1 (~2.6 m s−1). The points corresponding to
step−4 are the average of all the steps before step −3. Vertical bars
indicating SD are drawn when they exceed the size of the symbol.
The horizontal interrupted lines indicate the set-point speed
decreases. During this phase, the COM barely decelerates
(i.e. the Ekf-curve decreases little) while it is lowered (i.e.
the Ep  +  Ekv-curve decreases). Then the muscle–tendon
units perform positive work to accelerate the COM in the
forward direction (i.e. the Ekf-curve increases), without
significant change in its height (i.e. limited change in the
Ep + Ekv-curve). During the main part of the contact period
of step 0 the COM gains height and vertical velocity (i.e.
the Ep + Ekv-curve increases), whereas its forward velocity
is reduced (i.e. the Ekf-curve decreases). The energy-time
curves of TA are similar to those of RR (Fig. 1, left), as
observed by Mauroy et al. (2012). However, some differ-
ences exist between the two groups.
First, when approaching the obstacle, TA increase their
speed of progression from step −3 to step −1 (see Ekf in
Fig.  1), then lose speed during the contact of step 0 and
leap the obstacle at a speed close to the set-point speed
(one-way ANOVA P > 0.952) (Fig. 2). On the contrary,
the RR group accelerates barely from step −3 to step −1
and lose speed during the contact of step 0. Consequently,
the horizontal velocity over the obstacle is less than the
set-point speed (one-way ANOVA P < 0.001). Note that in
both groups, the higher the obstacle, the greater is the loss
in speed during the contact of step 0; however, this loss is
less important in the TA than in the RR group.
A second major difference is the external work done dur-
ing the negative and positive work phases of the contact of
step −1 and step 0. During the negative work phase of step
−1 (Fig. 3), TA do less work W − ext than RR because they
lose less speed (i.e. W −
f is smaller). During the positive work
phase of step −1, TA accelerate more than RR (Fig. 3), so
W+ f and consequently Wext are greater. When the obstacle
+
height increases, W f and Wext
+ +
increase in TA, but decrease in
RR. During the negative work phase of step 0, W − f and W ext
−
increase with the obstacle height both in TA and in RR, but
TA do less negative work than RR. During the positive work
phase of step 0, W +v increases with the obstacle height in both
groups. However for a 0.85-m-high obstacle, W + v and there-
fore Wext
+
are ~30 % smaller in TA than in RR. During most
part of the contact phase of step 0, the COM is decelerated
forward, while it is elevated and accelerated upwards (Fig. 1).
There is thus an exchange between the Ekf- and the Ep + Ekv-
curves, i.e. the energy lost during the horizontal movement
of the COM is transferred into energy to sustain its vertical
movement. Consequently, the positive work actually done
by the muscle–tendon units (Wext +
) is smaller than the work
done to elevate the COM (W v ). The net work done during
+
the two last steps is also different in TA than in RR (Fig. 4).
During the contact step −1, the net positive work done by TA
to accelerate the COM (Wf,net) is greater than the net negative
work done to lower it (Wv,net); consequently, Wext,net is posi-
tive. On the contrary, in RR |Wf,net| < |Wv,net| and Wext,net are
negative. During the contact of step 0, whatever the obstacle
height, the positive external work (Wnet,ext) done is greater in
TA than in RR (two-way ANOVA P < 0.012).
The third major difference concerns the vertical velocity
of the COM at takeoff (Vv,to) of step 0. During the contact
of step 0, W +v is smaller in TA than in RR (Fig. 3); for a
0.85-m-high obstacle, Vv,to (Fig. 5) is ~30 % smaller in TA
than in RR. Consequently, the aerial phase over the hurdle
(Fig. 6) is shorter in athletes and the vertical excursion of
the COM upwards (Sv,up = Vv,to 2 /2g) is ~50 % smaller. In
both groups, Vv,to and ta increase with the obstacle height
(two-way ANOVA P < 0.001). However, the increase is
smaller in TA than in RR.
The fourth difference concerns the angle swept by the
lower limb during contact (i.e. θto -θtd in Fig. 5). What-
ever the obstacle height, the angle swept during step −1
is smaller in TA than in RR (two-way ANOVA P < 0.001).
This is because the orientation of lower limb of the ath-
letes at touchdown (θtd) and takeoff (θto) is more vertical.
At touchdown of step 0, θtd is also more vertical in TA,
whereas at takeoff, θto is more horizontal. In front of a
0.85-m-high obstacle, the angle swept during the contact of
step 0 is not statistically different in TA and in RR (Stu-
dent’s t test P > 0.562); however on average, the leg spring
is oriented more horizontally in athletes. In both groups
when the obstacle height increases, θtd increases (i.e. the
lower limb is more horizontal) whereas θto decreases (i.e.
the lower limb is more vertical).
13
 Author's personal copy
778
Fig. 3  Mass-specific mechanical work done to move the COM dur-
ing step −1 and step 0 as a function of the obstacle height. From
top to bottom W − f and W f are, respectively, the negative and posi-
+
tive work done to decelerate and accelerate the COM forward; W − v data obtained by Mauroy et al. (2012) on recreational runners running
and W +v are, respectively, the negative and positive work done to
lower and lift the COM; W ext and Wext are, respectively, the negative
− +
and positive external work done by the muscle–tendon units to sus-
tain the movements of the COM relative to the surroundings. On the
abscissa, “C” represents the control steps when running on the level
without obstacles. Filled circles represent the mean value of the data
Finally, the last major difference between the two
groups concerns the overall stiffness of the leg-spring
system (kleg) generated by the lower-limb muscle–tendon
units. The stiffness kleg is greater in athletes than in RR
during steady-state running (Fig. 7), during step −1 (two-
way ANOVA P < 0.007) and step 0 (two-way ANOVA
P < 0.029). In both groups, during step −1, kleg decreases
as compared to steady-state running. For the same obsta-
cle height (0.85 m), the decrease is smaller in TA (−12 %)
than in RR (−25 %). During step 0, kleg increases as com-
pared to steady-state running; for the same obstacle height
(0.85 m) the increase is smaller in TA (+43 %) than in
RR (+55 %).
13
Eur J Appl Physiol (2014) 114:773–784
obtained on trained athletes during steady-state running at 18 km h−1
and while approaching a 0.85-, 0.91- and 1.07-m-high obstacle at the
same speed. Similarly, empty circles and dotted lines are from the
at the same speed and approaching a 0.45-, 0.65-, and 0.85-m-high
obstacle. Vertical bars indicating SD are drawn when they exceed the
size of the symbol. Stars are placed above the symbols if the results
obtained in TA are significantly different from the ones obtained in
RR (Student’s t test, *P < 0.05 and **P < 0.001)
Discussion
Several studies have been performed on the mechanism of
hurdling during hurdle sprint, mainly using a cinemato-
graphic method. Mann and Herman (1985) have shown
that a high Vf during the clearance and a low Vv at take-
off are indicators of a good hurdle performance. In men’s
hurdle sprint, Vv must be >1.7 m s−1 to clear the hurdle
successfully (Salo et al. 1997). McDonald and Dapena
(1991a) have evaluated the contribution of each body seg-
ment on the total angular momentum of the body during
hurdling. Coh and Iskra (2012) have measured the ground
reaction forces during the contact before and after the
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784 779

Fig. 4  Net muscular work

(Wnet) and recovery during step
−1 and step 0 as a function
of the height of the obstacle.
Wnet represents the differ-
ence between the positive and
negative work done during the
contact. Wf,net and Wv,net are,
respectively, the net work to
accelerate or decelerate the
COM forward and to raise or
lower it. Wext,net represents the
net work done by the muscle–
tendon units to move the COM
relative to the surroundings. The
recovery evaluates the energy
exchange between the Ekf-curve
and the Ep + Ekv-curve and vice
versa. Other indications are as
in Fig. 3

hurdle. They found that an effective hurdle technique is
characterized by an optimal ratio between brake and pro-
pulsion at takeoff, a short aerial period and a short brak-
ing phase at landing.
In the present study, we compare the bouncing mecha-
nism of running while approaching and leaping over an
obstacle in TA familiar with the hurdling technique and in
RR unfamiliar with this technique.The aim is to analyse
how athletes trained to hurdling limit the loss in the for-
ward velocity while crossing over the obstacle. Therefore,
we analysed the movements of the COM of TA during the
four running steps preceding the jump over an obstacle of
different heights, estimated the overall stiffness kleg of the
leg-spring system and compared these results with those
obtained on RR by Mauroy et al. (2012).
The overall stiffness kleg generated by the muscle–ten-
don units was estimated using the model of a mass mounted
on a linear spring that swept on an arc during the contact
(McMahon and Cheng 1990). This model was adapted to
take into account the difference in orientation of the leg at
touchdown and takeoff (Mauroy et al. 2012). Even if inac-
curacies were found between the parameters evaluated by
the simulation and the experimental values (see the end of
“Methods”), we think that the model gives a general idea of
how the central nervous system (CNS) controls the move-
ment of the COM during the two steps before the obstacle.

13
 Author's personal copy
780
Fig. 5  Orientation of the leg spring and vertical velocity of the COM
at the beginning and end of the contact phase during step −1 and step
0 as a function of the obstacle height. Top panels Vv is the instantane-
ous vertical velocity of the COM at touchdown (subscript td) and at
As pointed out by Lacquaniti et al. (2012), if the CNS
controlled independently each individual muscle, as well
as the motion and the stiffness of each articular degree of
freedom, motor control would be extremely fractionated
and difficult to implement. Even if the spring–mass model
is not able to predict accurately all the mechanical variables
during running (Bullimore and Burn 2007), this model sug-
gests that the CNS controls the different lower limb mus-
cles in a global way to generate an overall leg-spring stiff-
ness that imitates a linear spring–mass system bouncing
and sweeping forward during ground contact.
In 1990, McMahon and Cheng showed that during
steady-state running, the stiffness is modulated so that the
orientation of the leg spring at takeoff and at touchdown
is symmetric relative to the vertical, and that the maximal
compression of the system is reached when the leg spring is
vertical. These authors predicted that if the stiffness is too
low, the leg is oriented more horizontally at takeoff than
at touchdown. On the contrary, if the stiffness is too high,
the orientation of the leg is more vertical at takeoff than at
touchdown.
13
Eur J Appl Physiol (2014) 114:773–784
takeoff (subscript to). Bottom panels θ is the angle made by the leg
spring and the vertical at touchdown (subscript td) and takeoff (sub-
script to). Other indications are as in Fig. 3
In recreational runners, the trajectory of the COM is
modified during the two steps preceding the obstacle by
modulating the overall stiffness of the leg-spring system
(Mauroy et al. 2012): the lowering and forward accelera-
tion of the COM during step −1 is obtained by reducing
the overall stiffness of the lower limb muscles, whereas
during step 0 the COM is lifted and decelerated forward
due to a stiffening of the lower limb muscles.
The results of the present paper show that, even if
the kinematics of the lower limb movement is different
between TA and RR, the basic mechanisms used by hur-
dlers and non-specialists crossing an obstacle are compa-
rable. The effect of the obstacle height on the variables
studied is also similar in the two groups, although less pro-
nounced in TA than in RR.
Note that our experiments do not reproduce the hur-
dle race situation: here, TA approach the hurdle at a sub-
maximal speed (i.e. 18 km h−1) and cross only one hurdle.
This experimental protocol was chosen: (1) for a practical
reason—our track is too short to run at speeds higher than
~20 km h−1 and to cross more than one hurdle—and (2) to
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784 781

Fig. 6  Step durations of step

−1 and step 0 as a function of
the obstacle height. From top to
bottom T is the step period, ta
the aerial time and tc the contact
time. Other indications are as
in Fig. 3

Fig. 7  Leg-spring stiffness
(kleg) during step −1 and step 0
as a function of the height of the
obstacle. The schema illus-
trates the spring–mass model at
touchdown, maximal compres-
sion and takeoff. At maximal
compression, the leg-spring
system is vertical during steady-
state running, oriented forward
during step −1 and backward
during step 0. Other indications
are as in Fig. 3

13
 Author's personal copy
782
allow comparison with the results obtained on RR. Thus,
since TA do not run at their maximal speed, our results
do not give any information about the maximal muscular
force, work and power developed by athletes during a hur-
dle race. However, if the athletes ran at maximal speed,
they would probably modify the same parameters used at
the submaximal speeds examined here. In other words, the
observation made at 18 km h−1 could be transposed, muta-
tis mutandis, to the maximal speed of the hurdlers.
The increase in stiffness observed during the contact
of step 0 is consistent with the results of Grimmer et al.
(2008). These authors analysed the bouncing mechanism
of running on uneven ground; they showed that during the
contact prior to a 0.05–0.15-m-high step, the stiffness gen-
erated by the muscle–tendon unit is increased. These results
are also consistent with those of Mauroy et al. (2012).
Even if TA used the same mechanism as RR, athletes
adjusted their running pattern to maintain the pace over
the obstacle as high as possible. To leap the obstacle, the
runner shifts the velocity vector (V) at takeoff in a more
vertical orientation, to increase the duration of the aerial
phase. This change in orientation leads to a smaller hori-
zontal component of V. As a consequence, the forward
velocity during the aerial phase is lower than during the
contact phase (Figs. 1, 2). To race as fast as possible, hur-
dlers limit the increase in the aerial time over the obstacle
and the decrease in the forward velocity over the obstacle.
Mann and Herman (1985) show that the high performance
in hurdlers is correlated, among others, to a low Vv at take-
off of step 0 and a high Vf over the hurdle. Our results con-
firm these observations: in TA the speed of progression is
reduced less and the COM is elevated less than in RR. Dur-
ing step 0 the positive work Wv,net and the negative work
Wf,net are smaller in TA than in RR (Fig. 4).
To reduce the vertical excursion of the COM over the
obstacle, trainers advise athletes to “run the hurdle” and not
to jump over it. This is done by limiting the change in the
leg-spring stiffness during step −1 and step 0; our results
show that for a same obstacle height, TA change their stiff-
ness less than RR (Fig. 7). Consequently, the orientation of
the leg and the beginning and the end of contact of step −1
and step 0 are closer to the values observed during steady-
state running (Fig. 5). At the end of the contact phase of
step 0, the leg spring is thus oriented more horizontally in
TA than in RR and the orientation of the velocity vector (V)
at takeoff is more horizontal in TA than in RR (Fig. 8). As
pointed out by Seyfarth et al. (2002), this strategy is prob-
ably adopted to keep the bouncing mechanism of running
as stable as possible.
The “run the hurdle” recommendation is also evidenced
by the recovery (lower panel of Fig. 4). This variable repre-
sents the relative amount of energy transferred between the
Ekf- and the Ep + Ekv-curves. During normal running, the
13
Eur J Appl Physiol (2014) 114:773–784
Fig. 8  Orientation and magnitude of the velocity vector of the COM
at takeoff of step 0. On each graph, the three vectors are the average
of the data obtained on all subjects, for a given obstacle height. On
the right side, filled circles, diamonds and squares represent the mean
value of the data obtained on trained athletes hurdling, respectively,
a 0.85-, 0.91- and 1.07-m-high obstacle at 18 km h−1. On the left
side, ‘×’ symbols, ‘+’ symbols and circles represent the mean value
of the data obtained by recreational runners leaping, respectively, a
0.45-, 0.65- and 0.85-m-high obstacle at the same speed. The veloc-
ity vector at takeoff of steady-state running step is illustrated by the
interrupted line in bold. The interrupted curve in bold corresponds to
the norm of the velocity vector during steady-state running. Since the
vertical scale is twice as big as the horizontal scale, this curve does
not follow a circle trajectory. Note that the orientation of the vector is
more horizontal in TA than in RR (compare, for example, the vectors
for an obstacle of 0.85 m)
Ekf- and the Ep + Ekv-curves are more or less in phase, and
the recovery is typically <5 % (Cavagna et al. 1976). When
approaching the obstacle at 18 km h−1, the recovery during
step 0 is ~10 % lower in TA than in RR, suggesting that the
mechanism used by TA to leap the hurdle is closer to the
bouncing mechanism of running in steady state.
One of the goals of training is to increase the capacity of
the lower-limb muscle–tendon units to develop high forces
at high speeds. Here, we observe that the leg-spring stiff-
ness of athletes running at steady state is increased, as com-
pared to RR (Fig. 7). These results corroborate the observa-
tions of Morin et al. (2005) who found higher leg stiffness
during running in highly trained middle-distance runners,
as compared to physical education students. Moreover
during hopping in place, endurance-trained athletes dem-
onstrated significantly higher leg stiffness than untrained
subjects (Hobara et al. 2010). According to Mero et al.
(1992) and Butler et al. (2003), a stiffer muscle–tendon
unit would augment the rate of force development, which
in turn would increase the sprint performance (Brughelli
and Cronin 2008). Furthermore, Komi (1986) suggests
that the purpose of strength training is to increase the stiff-
ness of the muscle–tendon unit. As said by Hobara et al.
(2007), the co-contraction between agonist and antagonist
muscles plays a minor role in the regulation of kleg. These
 Author's personal copy
Eur J Appl Physiol (2014) 114:773–784 783
authors have observed that in hopping, kleg is regulated by a
change in centrally programmed muscle pre-activation and
by a concomitant change in the short-latency stretch reflex
response of the triceps surae muscles.
In a recent study, Morin et al. (2012) have shown that
the 100-m sprint running performances present a higher
correlation with the variables associated with the capacities
of the muscles to develop velocity than to develop force.
When comparing the force–velocity curves of world-class
sprinters with those of non-specialists, these authors show
that the maximal forces during contact are ~40 % greater
in athletes and that the maximal velocities at which this
force is developed are more than 100 % higher; so, the
maximal power developed by sprinters during the contact
phase is ~220 % greater and occurs at a velocity ~60 %
higher than non-specialists. Our results suggest that hur-
dlers also shifted their force–velocity curves to higher
velocities. At 18 km h−1, TA are still able to accelerate sig-
nificantly during the steps preceding the jump, whereas RR
barely increase their speed (Fig. 2). Note that RR running
at 9 km h−1 have the same behaviour as TA at 18 km h−1.
This observation suggests that, due to the force–velocity
relation of the muscles, RR may not be able to accelerate
when approaching the obstacle at high speeds, whereas,
due to training, TA are able to develop greater forces and
accelerations.
The effect of training on the muscle power is also evi-
denced in Fig. 4: at 18 km h−1, TA are still able to develop
a net positive muscular work (Wext,net) both during the con-
tact of step −1 and step 0. Due to training, hurdlers are
thus able to develop higher muscular power at high veloci-
ties than untrained subjects, both during step −1 and step
0. Consequently in TA, the magnitude of velocity vector
at takeoff of step 0 is increased (Fig. 8), as compared to
steady-state running, whereas in RR, the magnitude of this
vector is decreased.
Conclusion
The results of the present study evidence what trainers
intuitively know: the hurdler technique adjusts the trajec-
tory of the COM to keep the velocity of progression as high
as possible, while leaping the obstacle. To increase the ver-
tical velocity of the COM at takeoff, runners change the
orientation of the velocity vector of the COM during the
contact preceding the jump. As a consequence, the forward
velocity over the obstacle is lower than during the preced-
ing contact. To maintain the speed of progression as high
as possible, hurdlers minimize the duration and excursion
of the COM over the obstacle by maintaining the bounc-
ing mechanism as close as possible to the normal running
pattern. Therefore, they limit the change in orientation of
the velocity vector of the COM at takeoff by minimizing
the modifications of the leg-spring stiffness during the
two last steps preceding the hurdle. In this way, athletes
barely brush the hurdle so that their front leg comes rap-
idly in contact with the ground to reaccelerate the COM.
Our results also suggest that, most probably due to training,
athletes modify their force–velocity relationship to develop
higher forces at high velocity of contraction. Further stud-
ies during a 110-m hurdle sprint at maximal speed should
confirm these observations.
Acknowledgments The authors are grateful to the athletes who
accepted to spend some of their precious time to participate in our
experiments. This study was funded by the Université catholique de
Louvain, Belgium.
References
Brughelli M, Cronin J (2008) Influence of running velocity on verti-
cal, leg and joint stiffness: modelling and recommendations for
future research. Sports Med 38(8):647–657
Bullimore SR, Burn JF (2007) Ability of the planar spring–mass
model to predict mechanical parameters in running humans. J
Theor Biol 248(4):686–695. doi:10.1016/j.jtbi.2007.06.004
Butler RJ, Crowell Iii HP, Davis IM (2003) Lower extremity stiff-
ness: implications for performance and injury. Clin Biomech
18(6):511–517
Cavagna GA (1975) Force platforms as ergometers. J Appl Physiol
39(1):174–179
Cavagna GA, Thys H, Zamboni A (1976) The sources of external
work in level walking and running. J Physiol 262(3):639–657
Coh M, Iskra J (2012) Biomechanical studies of 110 m hurdle clear-
ance technique. Sport Sci 5(1):10–14
Farley CT, Gonzalez O (1996) Leg stiffness and stride frequency in
human running. J Biomech 29(2):181–186
Genin JJ, Willems PA, Cavagna GA, Lair R, Heglund NC (2010)
Biomechanics of locomotion in Asian elephants. J Exp Biol
213(5):694–706. doi:10.1242/jeb.035436
Grimmer S, Ernst M, Gunther M, Blickhan R (2008) Running on une-
ven ground: leg adjustment to vertical steps and self-stability. J
Exp Biol 211(Pt 18):2989–3000. doi:10.1242/jeb.014357
He JP, Kram R, McMahon TA (1991) Mechanics of running under
simulated low gravity. J Appl Physiol 71(3):863–870
Hobara H, Kanosue K, Suzuki S (2007) Changes in muscle activ-
ity with increase in leg stiffness during hopping. Neurosci Lett
418(1):55–59
Hobara H, Kimura K, Omuro K, Gomi K, Muraoka T, Sakamoto
M, Kanosue K (2010) Differences in lower extremity stiffness
between endurance-trained athletes and untrained subjects. J Sci
Med Sport 13(1):106–111
Komi PV (1986) Training of muscle strength and power: interaction
of neuromotoric, hypertrophic, and mechanical factors. Int J
Sports Med 7(SUPPL. 1):10–15
Lacquaniti F, Ivanenko YP, Zago M (2012) Patterned control of
human locomotion. J Physiol 590(Pt 10):2189–2199. doi:10.111
3/jphysiol.2011.215137
Mann R, Herman J (1985) Kinematic analysis of Olympic hurdle per-
formance: women’s 100 meters. Int J Sport Biomech 1:163–173
Mauroy G, Schepens B, Willems PA (2012) The mechanics of run-
ning while approaching and jumping over an obstacle. Eur J Appl
Physiol 113(4):1043–1057. doi:10.1007/s00421-012-2519-1
13
 Author's personal copy
784
McDonald C, Dapena J (1991a) Angular momentum in the men’s
110-m and women’s 100-m hurdles races. Med Sci Sports Exerc
23(12):1392–1402
McDonald C, Dapena J (1991b) Linear kinematics of the men’s
110-m and women’s 100-m hurdles races. Med Sci Sports Exerc
23(12):1382–1391
McMahon TA, Cheng GC (1990) The mechanics of running: how
does stiffness couple with speed? J Biomech 23(Suppl 1):65–78
Mero A, Komi PV, Gregor RJ (1992) Biomechanics of sprint running.
A review. Sports Med 13(6):376–392
Morin JB, Dalleau G, Kyrolainen H, Jeannin T, Belli A (2005) A sim-
ple method for measuring stiffness during running. J Appl Bio-
mech 21(2):167–180
13
View publication stats
Eur J Appl Physiol (2014) 114:773–784
Morin JB, Bourdin M, Edouard P, Peyrot N, Samozino P, Lacour JR
(2012) Mechanical determinants of 100-m sprint running perfor-
mance. Eur J Appl Physiol. doi:10.1007/s00421-012-2379-8
Salo A, Grimshaw PN, Marar L (1997) 3-D biomechanical analysis
of sprint hurdles at different competitive levels. Med Sci Sports
Exerc 29(2):231–237
Schepens B, Willems PA, Cavagna GA (1998) The mechanics of run-
ning in children. J Physiol 509(Pt 3):927–940
Seyfarth A, Geyer H, Gunther M, Blickhan R (2002) A movement cri-
terion for running. J Biomech 35(5):649–655