Scand J Med Sci Sports 2007: 17: 649–655
Printed in Singapore . All rights reserved
DOI: 10.1111/j.1600-0838.2006.00625.x
A comparison of sprinting kinematics on two types of treadmill and
over-ground
M. McKenna, P. E. Riches
Department of Applied Physiology, University of Strathclyde, Glasgow, UK
Corresponding author: Dr. Philip Edward Riches, Department of Applied Physiology, University of Strathclyde, 199 Cathedral
Street, Glasgow, G4 0QU, UK. Tel: 144 141 548 5703, E-mail: philip.riches@strath.ac.uk
Accepted for publication 9 November 2006
Conventionally motorized treadmills elicit different sprint-
ing kinematics to the over-ground condition. Treadmills
powered by a torque motor have been used to assess
sprinting power; yet, the kinematics of sprinting on the
torque treadmill are unknown. This study compares the
sprinting kinematics, during the constant velocity phase,
between a conventional treadmill, a torque treadmill and the
over-ground condition to assess the suitability of each
treadmill for sprinting studies and training. After familiar-
ization, 13 recreationally active males performed multiple
sprints at various experimental settings on each surface.
Ninety sprints, which attained mean velocities over 7.0 m/s,
had their lower-body sagittal plane joint angles during
Conventionally motor-driven treadmills can ensure
that an athlete attains the same velocity as over-
ground sprinting during the constant velocity phase,
but are known to result in different sprinting kine-
matics (Frishberg, 1983). Alternatively, a non-
motorized treadmill enables athletes to control the
velocity of the belt; however, athletes typically only
attain 80% of their maximum over-ground velocity
(Lakomy, 1987). Treadmills powered by a torque
motor (torque treadmill) may have the advantages of
both systems: they allow athletes to have some
control over the velocity of the belt and, as being
powered, aid the athlete in reaching their maximum
velocity. However, the kinematics of sprinting on the
torque treadmill at constant velocity is unknown.
On the torque treadmill, the belt is powered at a
given applied motor torque. If the applied motor
torque is greater than the resistive moments applied
to the belt, the belt will accelerate. Conversely, if the
resistive moments are greater than the applied tor-
que, then the belt will decelerate or remain station-
ary. The resistive moments arise from two sources:
the inherent system friction (e.g. friction associated
with the unloaded belt rubbing on the base plate),
and moments from the forces applied to the belt from
the subject. The vertical component of force aug-
Copyright & 2007 The Authors
Journal compilation & 2007 Blackwell Munksgaard
ground contact captured at 250 Hz. These data were low-
pass filtered at 10 Hz, and compared with respect to surface,
subject and velocity using an ANCOVA statistical model.
Sprinting on the conventional treadmill elicited a longer
ground contact time, a longer braking phase, a more
extended knee at foot strike and a faster extending hip
than the torque treadmill and over-ground (all Po0.05).
The torque treadmill obtained an equivalent sprinting tech-
nique to the over-ground condition, with the exception of a
less extended hip at toe-off, suggesting that it is more
appropriate for laboratory sprinting analyses and training
than the conventional treadmill.
ments the belt’s friction via Coulomb’s law of fric-
tion, and will thus always act to retard belt motion.
The horizontal component of force can either con-
tribute positively or negatively to the belt’s accelera-
tion depending on its direction. By changing the
vertical and horizontal components of the force
being applied to the belt, an athlete has some control
over the belt’s acceleration and thus velocity.
The kinematic studies that have compared con-
stant velocity phase running on conventional tread-
mills with over-ground running have demonstrated a
lack of consistency. At a given velocity, an athlete
may have a longer stride length, and shorter stride
rate while running on a treadmill (Nelson et al., 1972)
or perhaps the opposite (Elliot & Blanksby, 1976;
Wank et al., 1998), or there may be no difference
(Frishberg, 1983). Furthermore, athletes may have a
longer (Nelson et al., 1972) or shorter (Wank et al.,
1998) ground contact phase on the treadmill com-
pared with over-ground. However, these disparities
may reflect the measurement technique and indivi-
duality of running gait (Saito et al., 1974; Frishberg,
1983; Moravec et al., 1988; Ae et al., 1992), or the
fact that running kinematics is highly specific with
regard to velocity (Luhtanen & Komi, 1978; Mero
et al., 1992). Similar to over-ground running (Mero
649
McKenna & Riches
et al., 1992), stride rate increased and ground contact
duration decreased with increasing velocity on the
conventional treadmill (Kivi et al., 2001). However,
it was stated that ‘‘mechanical form deteriorated’’ at
95% of maximal speed, and it was subsequently
suggested that training above 90% be seldom used
(Kivi et al., 2001). Nonetheless, while running at
‘‘constant velocity,’’ the ground contact phase can be
subdivided into braking and propulsion phases,
relating to periods of horizontal deceleration and
acceleration, respectively (Novacheck, 1998).
Frishberg (1983) is the only study that has com-
pared the kinematics of maximum velocity sprinting
(9.21  0.11 m/s) on the conventional treadmill and
over-ground. It was found that, compared with over-
ground, on the conventional treadmill the foot was
further forward at foot strike (eliciting a greater
braking effect); and throughout the ground contact
phase, the knee joint moved through a greater range
of motion with a faster angular velocity, and the hip
joint was displaced at a smaller angle with a slower
angular velocity. These differences were attributed to
the fact that the belt forces the supporting leg back
beneath the body.
The torque treadmill and non-motorized tread-
mills have been used more commonly in the estima-
tion of over-ground sprint running power output
(Belli & Lacour, 1989; Paavolainen et al., 1994;
Falk et al., 1996; Jaskolski et al., 1996; Sutton
et al., 2000; Chelly & Denis, 2001; Funato et al.,
2001; Tong et al., 2001). However, a kinematic
comparison of torque treadmill sprinting with over-
ground sprinting must be made if power output
measurements taken on the torque treadmill are to
be considered equivalent to over-ground power
output, or if the torque treadmill is to be used as
a sprint analysis tool or training aid. As stride
length and frequency measurements may be highly
individual, and that interesting differences may be
occurring during the ground contact phase with
respect to contact time and joint angular displace-
ments and velocities, this paper compares these latter
variables between over-ground, conventional tread-
mill and torque treadmill sprinting, to identify the
kinematic appropriateness of each treadmill for
sprinting studies.
Methods
Thirteen recreationally active males (age: 23  2.1 years;
height 1.79  0.03 m; mass 72.5  4.8 kg), all regularly parti-
cipating in sports involving sprinting gait, e.g. football, rugby,
athletics, hockey, volunteered for the study. The University’s
Ethics Committee approved all the procedures undertaken,
and informed consent was also obtained before testing com-
mencement.
Each subject completed four visits to the laboratory.
Initially, each subject was familiarized with the sensation of
650
sprinting on the torque and conventional treadmills. This
familiarization was required due to the unusual sensation of
treadmill sprinting, especially on the torque treadmill (Jas-
kolski et al., 1996). Treadmill velocities were incrementally
increased as the subject self-reported competence and con-
fidence, up to and including the experimental settings. On the
next visit, the subjects were assessed on the torque treadmill.
This was to enable the range of velocities attained to be
matched both over-ground and on the conventional treadmill,
whose sessions were applied randomly. Typically, a week
expired between sessions.
The torque setting on the treadmill (Sprint Club, Médical
Developpement, France) was defined in terms of the max-
imum torque of the treadmill motor, Cmax. With each subject
standing still on the treadmill, the torque was increased in
increments of 0.01Cmax, until the belt just started to move.
This setting was denoted the slide limit (SL) and it depended
linearly on the mass of the subject. It was unknown as to what
fraction of Cmax aided the fastest sprinting velocity: too little
torque and it became too difficult to run, too much torque and
the belt would over-accelerate the legs in a manner that
deteriorated form. Therefore, five settings were investigated
from SL 0.05Cmax to SL10.15Cmax in increments of
0.05Cmax. The horizontal velocity of the subject on the torque
treadmill was taken to be the velocity of the belt during the
constant velocity phase of the trial, while over-ground, the
horizontal velocity was taken as the mean horizontal velocity
of the hip marker throughout the ground contact phase. To
facilitate a variation of velocity in the over-ground condition,
subjects were asked to run at speeds defined verbally as ‘‘400 m
pace,’’ ‘‘200 m pace’’ and ‘‘100 m pace,’’ achieved along a long
indoor corridor with adequate acceleration and deceleration
space to avoid end effects. The conventional treadmill (Pulsar
4.0, h/p/cosmos, Nussdorf-Traunstein, Germany) was set at
7.5, 8.0 and 8.5 m/s, and any variation in the mean horizontal
velocity of the hip marker was subtracted from this value.
To determine the kinematics of motion, 14 mm reflective
markers were placed on the left side of the body at the fifth
metatarsal head (toe), the lateral malleolus (ankle), the lateral
femoral epicondyle (knee) and the tip of the greater trochanter
(hip). The same experimenter positioned all the markers on all
the testing days. The positions of these markers were deter-
mined in 3D using a six-camera motion analysis system (Vicon
612 with M2 cameras, Oxford, UK) operating at 250 Hz,
calibrated statically and dynamically so that the standard
deviation of the residual was less than 1 mm. The speed of
the torque treadmill belt was determined by using four
individually shaped clusters of markers attached to the belt.
The clusters varied in the number of markers between four and
seven, and were positioned so that at least one cluster was in
view on the top surface of the belt at all times. Body and belt
marker coordinate data were reconstructed using the direct
linear transform method, reduced to the sagittal plane and
subsequently low-pass filtered using a zero-lag fourth-order
Butterworth filter (10 Hz cut-off) and post-processed using a
custom-written routine in Matlab (Mathworks, Massachu-
setts, USA).
Sprinting was defined in this study as running at a mean
constant velocity equal to or over 7.0 m/s and trials were only
included if this criterion was satisfied. Constant velocity
sprinting was maintained for 4 s on each of the treadmills,
and for as long as possible in the over-ground condition,
typically about 3 s before the subject had to decelerate. As the
capture volume of the over-ground condition was 5 m in
length, only one ground contact phase per trial was available
for analysis. Therefore, as stride-to-stride variance may be
assumed to be small (Morgan et al., 1991), and for a consistent
methodology between surfaces, only one ground contact

phase was randomly selected, from the constant velocity phase
of each trial on the torque and conventional treadmills, for
detailed subsequent analysis.
The following variables were used to identify differences
between conditions in the sprinting gait: ground contact time
(TGC), which was further subdivided into braking time (Tb)
and propulsion time (Tp), hip, knee and ankle angles at foot
strike (a0, b0, g0, respectively), hip, knee and ankle angles at
toe-off (a1, b1, g1, respectively),
. maximum knee flexion velo-
city during
. ground contact (bflx ) and knee and hip extension
.
velocity (bext and aext ) during ground contact also. The marker
locations and angle definitions are shown in Fig. 1.
We wished to analyze the multiple kinematic variables with
respect to the type of running surface (condition) and velocity.
As subjects attained different velocities over-ground and on
the torque treadmill, the between-subject covariate (velocity)
was non-homogenous for the different levels of the within-
subject variable (condition). Thus, in order to retain both
variables in the analysis, we assumed each sprint to be an
independent event and we did not perform a repeated mea-
sures analysis, but rather an independent analysis with an
acknowledged loss of statistical power. However, as multiple
sprints per subject per condition were measured, we investi-
gated the inter-individual variability in the analysis by includ-
ing the subject label in the analyses. Consequently, a
MANCOVA analysis was performed, with condition and
subject as factors and velocity as a covariate, to identify the
factors and covariates and interactions that affected the
kinematic variables. The identified variables were subse-
quently used in univariate ANCOVA models, to see the effects
of the independent variables on each individual dependent
kinematic variable. Any significant effects of condition were
further explored using Bonferroni-adjusted pairwise compar-
Fig. 1. Marker locations and joint angle definitions.
Sprinting kinematics
isons. The means for these conditions were estimated at a
specific velocity so that differences could be attributed to the
condition independent of velocity. Angles were measured as
shown in Fig. 1, with presented means accompanied by the
appropriate 95% confidence interval. Throughout the ana-
lyses, any differences were deemed significant if the P-value
was less than 0.05, and actual P values (unless they are less
than 0.001), effect sizes (ES) and power are reported.
Results
The mean maximum over-ground velocities (100 m
pace) of our subjects were 7.77  0.29 m/s (range 7.0–
8.64 m/s). As the applied torque on the treadmill
increased, the mean velocity of the subjects also
increased (Fig. 2). Figure 2 suggests that the mean
maximum over-ground velocity best corresponds to
the setting that applied 10% of the Cmax over and
above the SL. As the ability of the subject to
influence the speed of the belt reduces with applied
torque, so too does the 95% confidence interval.
Ninety sprints satisfied the criteria of being at or
over 7.0 m/s. These corresponded to 20 over-ground
sprints, 38 conventional treadmill sprints (one sub-
ject was not deemed competent to attempt the 8.5 m/s
conventional treadmill sprint) and 32 torque tread-
mill sprints. On the torque treadmill, all subjects
achieved an average speed of  7.0 m/s at SL110%
and SL115%, while four subjects attained an aver-
age velocity  7.0 m/s at SL15%, and one at SL
too. The MANCOVA indicated that condition, sub-
ject, velocity and all their two-way interactions had a
significant effect on a combination of the kinematic
variables (Wilks’ l, all Po0.001 and power 5 1).
Thus, the same statistical model was used to inves-
tigate each kinematic variable independently using
univariate ANCOVA. A summary table showing
mean values for each kinematic variable in each
condition, evaluated at the pooled mean velocity
over all conditions (7.88 m/s), is given as Table 1.
Fig. 2. Variation in mean velocity for sprinting on the
torque treadmill at the different applied motor torques
(SL, slide limit, the torque setting that just initiates belt
movement with the subject stationary; percentages indicate
percentage of the maximum motor torque). One sprint was
randomly selected from each subject (N 5 13) for analysis.
651
McKenna & Riches
Table 1. Estimates of mean kinematic values for each condition

Variable Condition Mean 95% lower 95% upper

bound bound

TGC (s) Over-ground 0.140w 0.133 0.147

Conventional 0.162* 0.159 0.165
Torque 0.134w 0.131 0.138
Tb (s) Over-ground 0.074w 0.067 0.080
Conventional 0.092* 0.089 0.095
Torque 0.071w 0.068 0.074
Tp (s) Over-ground 0.066 0.060 0.072
Conventional 0.070 0.068 0.073
Torque 0.063 0.061 0.066
a0 (1) Over-ground 40.4 38.7 42.0
Conventional 39.5 38.8 40.3 Fig. 3. 95% confidence interval for the knee flexion angle
Torque 38.4 37.5 39.2 during the ground contact phase for the three conditions.
b0 (1) Over-ground 35.4w 31.6 39.1 One sprint was randomly selected from each subject (N 5 13)
Conventional 15.0* 13.4 16.6 for analysis. The confidence intervals were determined from
Torque 31.2w 29.4 33.1 when (i) the conventional treadmill was set at 8.0 m/s, (ii) the
g0 (1) Over-ground 120.9 117.1 124.7 torque treadmill at SL10.1Cmax and (iii) each individual’s
Conventional 120.3 118.7 122.0 fastest over-ground sprint.
Torque 123.3 121.5 125.3
a1 (1) Over-ground 24.0 26.3 21.8
Conventional 21.6 22.6 20.6 surfaces. The extended knee at foot strike can clearly
Torque 19.9* 21.1 18.9 be seen on the knee angle traces throughout the
b1 (1) Over-ground 36.3 33.2 39.4
Conventional 31.9 30.6 33.3 ground contact phase (Fig. 3). The hip angle at toe-
Torque 32.5 31.0 34.1 off, a1, on the torque treadmill was significantly less.
g1 (1) Over-ground 146.4 142.5 150.2 extended than while on the other surfaces, and bflx
Conventional 141.9 140.2 143.6 was greater while on the conventional treadmill than
. Torque 139.5* 137.5 141.4
bflx (rad/s) Over-ground 9.08 6.60 11.6 while on the torque treadmill. All other variables
Conventional 11.2 10.1 12.3 showed no difference with condition.
. Torque 6.56w 5.33 7.80 Velocity was a significant covariate for TGC
bext (rad/s) Over-ground 7.52 6.45 8.59 (Po0.001, ES 5 0.69, power 5 0.98), Tp (P 5 0.003,
Conventional 7.06 6.59 7.52
Torque 6.88 6.35 7.41 ES 5 0.53, power 5 0.87), a0 (P 5 0.03, ES 5 0.38,
.
aext (rad/s) Over-ground 16.8w 11.6 22.1 power 5 0.59),
. g0 (P 5 0.029, ES 5 0.38, power 5
Conventional 27.9* 25.7 30.3 0.60) and bflx (Po0.047, ES 5 0.34, power 5 0.52).
Torque 19.5w 16.8 22.1
As horizontal velocity increases from 7.0 m/s, the
All the mean values were evaluated at V 5 7.88 m/s (the pooled mean analysis suggests that, on average, TGC and Tp
velocity of all sprints under all conditions), and presented to three decrease (both at approximately 0.01 s per m/s), while
significant figures. the hip is more flexed at foot strike, and the ankle
*Significantly different from the over-ground condition. angle is more plantar-flexed at toe-off. .
w
Significantly different from the conventional treadmill condition Subject variation was a main effect for bflx
(Bonferroni-adjusted). .
(P 5 0.022, ES 5 0.89, power 5 0.90) and aext (P 5
0.01, ES 5 0.95, power 5 0.94); however, instead of
appearing as a main effect, the inherent inter-subject
The type of running surface (condition) had a variance possibly located itself within the interaction
significant effect on TGC (P 5 0.001, ES 5 0.69, between subject and condition, which exhibited sig-
power 5 0.95), Tb (Po0.001, ES 5 0.92, power 5 nificant results for all .kinematic variables with the
0.99), b0 (Po0.001, ES 5 1.08, power . 5 1), a1 (P 5 exception of Tp and bext . This implies that, while
0.038, ES 5 0.45, power 5 0.63), bflx (Po0.001, there were clear main effects of condition and velocity,
.
ES 5 0.75, power 5 0.98) and aext (P 5 0.043, subjects reacted idiosyncratically to the various run-
ES 5 0.44, power 5 0.61). Subsequent Bonferroni- ning surfaces.
adjusted pairwise comparisons showed that for Finally, the relationship between velocity and some
.
TGC, Tb, b0 and aext , the conventional treadmill kinematic variables varied with condition for TGC
elicited significantly different effects compared with (P 5 0.002, ES 5 0.63, power 5 0.91), Tb (Po0.001,
the torque treadmill and over-ground conditions, but ES 5 0.87, power 5 1), b0 (P 5 0.001, ES 5 1.02,
the torque treadmill and over-ground conditions power. 5 1), a1 (P 5 0.028, ES 5 0.47, power 5 0.67)
were kinematically equivalent for these variables. and bflx (P 5 0.001, ES 5 0.72, power 5 0.97). For
On the conventional treadmill, subjects had longer example, the knee angle at foot strike becomes
TGC, longer Tb, a more extended knee at foot strike increasingly flexed with increasing velocity for the
and a faster extending hip than when on the other conventional and over-ground conditions, but on the

652

Fig. 4. Interaction between the three surface conditions and
velocity complete with 95% confidence intervals for the
prediction of the population mean knee angle at ground
contact (GC). Also indicated is V 5 7.88 m/s, the mean
velocity of all sprints under all conditions, and the velocity
at which condition comparisons are made.
torque treadmill, the knee becomes more extended at
foot strike as the velocity increases (Fig. 4). The
combined effect of these variations is a null result for
the main effect of b0 with velocity.
Discussion
Sprinting on the torque treadmill is equivalent to
over-ground sprinting for the majority of kinematic
variables measured in this study. Therefore, allowing
the subject to have a degree of control over the
acceleration of the torque treadmill belt facilitates
similar kinematics compared with over-ground
sprinting. Moreover, as the maximum speeds at-
tained on the torque treadmill can actually be greater
than those over-ground, the torque treadmill has the
potential not only to be the most suitable treadmill
for sprint analyses but also for over-speed training,
an ability that Frishberg (1983) ascribed to the
conventional treadmill too. One difference between
the torque treadmill and the over-ground condition
was that of the hip angle at toe-off, which was
significantly less extended on the torque treadmill,
albeit by 41. It may be possible that by fine tuning the
applied torque for each individual so that their
maximum treadmill velocity equals that over-
ground, a sprinting environment synonymous to
the ‘‘natural’’ may be able to be reproduced under
laboratory conditions.
Table 2 facilitates a comparison between the re-
sults of different studies on two key variables: the
knee angle at touchdown and the hip extension at
toe-off. The differences that exist may be attributable
to the different experimental equipment and proce-
dures and/or differences in velocity and subjects: the
recreational males in the present study were unable
Sprinting kinematics
Table 2. Comparison across studies of the knee angle and hip extension
at touchdown and toe-off
Variable Condition Parameter value
Knee flexion at touchdown (1) Conventional 16.8  0.8*
15.0  0.8w
Over-ground 27.2  1.1*
35.4  0.9w
Hip extension at toe-off (1) Conventional 25.8  1.0*
21.6  0.5w
Over-ground 29.3  1.1*
24.0  1.1w
*Refers to the study of Frishberg (1983), and
w
To this study.
to attain the velocities of Frishberg’s subjects, who
collectively achieved an average over-ground velocity
of 9.21  0.11 m/s (Frishberg, 1983). The fact that
Frishberg’s subjects demonstrated greater hip exten-
sion at toe-off is indicative of the difference in
sprinting ability between the two sets of subjects.
Thus, our results may only be applicable to the sub-
elite population Nonetheless, both studies show an
approximate 2.51 difference in hip extension at toe-
off between the two conditions, irrespective of abil-
ity, and such agreement encourages confidence in our
experimentation, and thus the kinematic data for the
torque treadmill.
The current results not only highlight the equiva-
lence of torque treadmill and over-ground sprinting
but also demonstrate the inappropriateness of the
conventional treadmill for sprint running. Moreover,
numerous studies in the whole range of velocity of
human locomotion from walking (Alton et al., 1998)
through jogging (Nelson et al., 1972; Elliot &
Blanksby, 1976) to sprinting (Frishberg, 1983) have
found differences between conventional treadmill
and over-ground kinematics. The majority of differ-
ences between sprinting on the conventional tread-
mill and the other two surfaces have been found, in
this study, to be around foot strike. At foot strike,
the knee is more extended on the conventional tread-
mill, which means that the foot strike locus is
probably more anteriorly placed compared with the
torque and over-ground conditions, resulting in the
observed longer braking phase. However, only using
kinematics to describe the braking and propulsion
phases is difficult and kinetic analysis of the ground
contact phase is essential to detail fully the forces
acting on the body during ground contact. None-
theless, the relationship between kinematic variables
and sprinting performance has been well-documen-
ted (Frishberg, 1983; Mann & Herman, 1985; Mero
et al., 1992; Kivi et al., 2001). It suffices to say that
the kinematic pattern observed on the conventional
treadmill, if replicated over-ground, would result in a
reduced sprint performance.
653
McKenna & Riches
It has been argued that the mechanics of locomo-
tion on a treadmill and over-ground are similar as
long as the velocities are constant and the same (van
Ingen Schenau, 1980). With regard to either sprinting
over-ground or on a treadmill, neither requirement is
satisfied. It is known that a subject running at 2.3 m/s
has an effect on the intra-stride speed variation of the
conventional treadmill’s belt (Savelberg et al., 1998),
with each foot strike acting to retard the belt (Frish-
berg, 1983). Such speed variations may be as high as
10% with the kinematic differences becoming more
pronounced the faster the treadmill velocity (Savel-
berg et al., 1998). Cavanagh and LaFortune (1980)
have shown a decrease of 0.18 m/s during the braking
phase while running at 4.47 m/s over-ground, fol-
lowed by a subsequent increase of 0.27 m/s. The
reasons for this variation have been associated with
the braking force at foot strike, air resistance and
propulsion (Mero et al., 1992). Furthermore, the
torque treadmill, by design, has an intentional oscil-
latory velocity during sprinting (Jaskolski et al.,
1996; Chelly & Denis, 2001).
The expected subject variation (Nigg et al., 1995)
did not appear for the majority of kinematic variables
as a main effect, but the variation resided in the
interaction between condition and subject. While con-
siderable variation occurred associated with indivi-
duality of gait, highly significant main effects
associated with both condition and velocity exist,
which indicates the strength of the effect of these
variables. Large subject variability is known to occur
at velocities between 3.0 and 6.0 m/s (Nigg et al., 1995)
and we have demonstrated a large subject variance
(albeit in an interaction form) for velocities between
7 and 8.6 m/s. Therefore, it may be suggested that
individuality of gait does not reduce as velocity
increases: as one approaches the subject’s maximum
velocity, the kinematic strategy for attaining max-
imum speed is still idiosyncratic to the individual.
In conclusion, running on the torque treadmill can
mimic over-ground sprinting performance with re-
spect to velocity and kinematic variables. An approx-
imate 201 difference in knee flexion occurs at foot
strike between the conventional treadmill and over-
ground, while the torque treadmill does not elicit any
such difference. Thus, the torque treadmill is more
appropriate for laboratory constant velocity phase
sprinting analyses compared with the conventional
treadmill; however, the applied torque setting should
be adjusted to ensure equivalence with the over-
ground performance.
Perspectives
The torque treadmill has been used predominantly in
athlete power analysis (Belli & Lacour, 1989; Jas-
kolski et al., 1996; Chelly & Denis, 2001). In these
studies, the applied torque is generally set at or below
that of the SL, to increase the resistance of running.
In this respect, while they may facilitate a maximum
peak power to be reached, the kinematics may not be
equivalent to the over-ground condition, and there-
fore may not be directly related to the track. This
study has shown that the torque treadmill can mimic
the kinematics of over-ground constant velocity
sprinting; however, further work must be carried
out to identify whether the acceleration phase (the
phase in which peak power is produced) is also
kinematically equivalent. Thus, we suggest that a
prior process of determining the most appropriate
torque setting to best match over-ground and torque
treadmill maximum velocities will result in a more
representative value for athlete power.
Key words: sprinting, biomechanics, motion analysis,
sagittal plane.

References
Ae M, Ito A, Suzuki M. The men’s 100
metres. New Stud Athletics 1992: 7: 47–
52.
Alton F, Baldey L, Caplan S, Morrissey
MC. A kinematic comparison of
overground and treadmill walking.
Clin Biomech 1998: 13: 434–440.
Belli A, Lacour R. Treadmill ergometer
for power measurement during
sprint running. J Biomech 1989:
22: 985.
Cavanagh CR, LaFortune MA. Ground
reaction forces in distance running. J
Biomech 1980: 13: 397–406.
Chelly SM, Denis C. Leg power and
hopping stiffness: relationship with
running performance. Med Sci Sports
Exerc 2001: 33: 326–333.
Elliot BC, Blanksby BA. A
cinematographical analysis of over-
ground and treadmill running by males
and females. Med Sci Sports 1976: 8:
84–87.
Falk B, Weinstein Y, Dotan R,
Abramson DA, Mann-Segal D,
Hoffman JR. A treadmill test of sprint
running. Scand J Med Sci Sports 1996:
6: 259–264.
Frishberg BA. An analysis of over-
ground and treadmill sprinting.
Med Sci Sports Exerc 1983: 15:
478–485.
Funato K, Yanagiya T, Fukunaga T.
Ergometry for estimation of
mechanical power output in sprinting
in humans using a newly developed
self-driven treadmill. Eur J Appl
Physiol 2001: 84: 169–173.
Jaskolski A, Veenstra P, Goossens P,
Jaskolska A, Skinner JS. Optimal
resistance for maximal power during
treadmill running. Sports Med
Training Rehab 1996: 7: 17–30.
Kivi DMR, Maraj BKV, Gervais P. A
kinematic analysis of high-speed
treadmill sprinting over a range of
velocities. Med Sci Sports Exerc 2001:
34: 662–666.

654

Lakomy HKA. The use of a non-
motorised treadmill for analysing
sprint performance. Ergonomics 1987:
30: 627–637.
Luhtanen P, Komi PV. Mechanical
factors influencing running speed. In:
Asmussen E, Jorgensen K, eds.
Biomechanics VI-B. Baltimore:
University Park Press, 1978: 23–29.
Mann RV, Herman J. Kinematic analysis
of Olympic sprint performance: men’s
200 meters. Int J Sports Biomech 1985:
1: 151–162.
Mero A, Komi PV, Gregor RJ.
Biomechanics of sprint running: a
review. Sports Med 1992: 13: 376–392.
Moravec P, Ruzicka J, Susanka P, Dostal
E, Kodejs M. The 1987 International
Athletic Foundation/IAAF Scientific
Project Report: time analysis of the 100
metres events at the II World
Championships in Athletics. New Stud
Athletics 1988: 3: 61–96.
Morgan DW, Martin PE, Krahenbuhl
GS, Baldini FD. Variability in running
economy and mechanics among trained
male runners. Med Sci Sports Exerc
1991: 23: 378–383.
Nelson RC, Dillman CJ, Lagasse P,
Bickett P. Biomechanics of over-
ground versus treadmill running. Med
Sci Sports 1972: 4: 233–240.
Nigg BM, De Boer RW, Fisher V. A
kinematic comparison of overground
and treadmill running. Med Sci Sports
Exerc 1995: 27: 98–105.
Novacheck TF. The biomechanics of
running. Gait Posture 1998: 7: 77–95.
Paavolainen L, Hakkinen K, Nummela
A, Rusko H. Neuromuscular
characteristics and fatigue in endurance
and sprint athletes during a new
anaerobic power test. Eur J Appl
Physiol 1994: 69: 119–126.
Saito M, Kobayahi K, Miyashita M,
Hoshikawa T. Temporal patterns
in running. In: Nelson RC, Morehouse
CA, eds. Biomechanics IV. Baltimore:
University Park Press, 1974:
106–111.
Sprinting kinematics
Savelberg HHCM, Vorstenbosch MATM,
Kamman EH, van de Weijer JGW,
Schambardt HC. Intra-stride belt-speed
variation affects treadmill locomotion.
Gait Posture 1998: 7: 26–34.
Sutton NC, Childs DJ, Bar-Or O,
Armstrong N. A non-motorised
treadmill test to assess children’s short-
term power output. Pediatr Exerc Sci
2000: 12: 91–100.
Tong RJ, Bell W, Ball G, Winter EM.
Reliability of power output
measurements during repeated
treadmill sprinting in rugby players. J
Sports Sci 2001: 19: 289–297.
van Ingen Schenau GJ. Some
fundamental aspects of the
biomechanics of overground versus
treadmill locomotion. Med Sci Sports
Exerc 1980: 12: 257–261.
Wank V, Frick U, Schmidtbleicher D.
Kinematics and electromyography of
lower limb muscles in over-ground and
treadmill running. Int J Sports Med
1998: 19: 455–461.
655