Lough b orough Univ ersit y [631700]

Lough b orough Univ ersit y
Institutional Rep ository
The mar gin for err or when
r ele asing the asymmetric
b ars for dismounts
This item w as submitted to Lough b orough Univ ersit y's Institutional Rep ository
b y the/an author.
Citation: HILEY, M.J. and YEADON, M.R., 2005. The margin for error when
releasing the asymmetric bars for dismoun ts. Journal of Applied Biomec hanics,
21 (3), pp.223-235.
A dditional Information:
• This article w as accepted for publication in the Journal of Applied Biome-
c hanics [ c/circlecopyrt Human Kinetics, Inc.] and the denitiv e v ersion is at:
h ttp://hk.h umankinetics.com/JAB/to c.cfm?iss=643
Metadata Record: h ttps://dspace.lb oro.ac.uk/2134/6052
V ersion: A ccepted for publication
Publisher: c/circlecopyrt Human Kinetics, Inc.
Please cite the published v ersion.

This item was submitted to Loughborough’s Institutional Repository
(https://dspace.lboro.ac.uk/ ) by the author and is made available under the
following Creative Commons Licence conditions.

For the full text of this licence, please go to:
http://creativecommons.org/licenses/by-nc-nd/2.5/

Journal Applied Biomechanics 21, 223-235, 2005

The margin for error when releasing the asymmetric bars for dismounts

Michael J. Hiley and Maurice R. Yeadon

School of Sport and Exercise Sciences, Lo ughborough University, Loughborough, UK

Abstract
It has previously been shown that male gymnasts using the “scooped” giant circling
technique were able to flatten the path followed by their mass centre resulting in a larger
margin for error when releasing the high ba r (Hiley and Yeadon, 2003a). The circling
technique prior to performing double layout somersault dismounts from the asymmetric
bars in Women’s Artistic Gymnastics appears to be similar to the “traditional” technique
used by some male gymnasts on the high bar. It was speculated that as a result the female
gymnasts would have margins for error similar to those of male gymnasts who use the
traditional technique. However, it is unclear how the technique of the female gymnasts is
affected by the need to avoid the lower bar. A four segment planar simulation model of the
gymnast and upper bar was used to determine the margins for error when releasing the bar
for nine double layout somersault dismounts at the Sydney 2000 Olympic Games. The elastic properties of the gymnast and bar were modelled using damped linear springs.
Model parameters, primarily the inertia and spring parameters, were optimised to obtain a
close match between simulated and actual performances in terms of rotation angle (1.2°), bar displacement (0.011 m) and release veloc ities (< 1%). Each matching simulation was
used to determine the time window around the actual point of release for which the model
had appropriate release parameters to complete the dismount successfully. The margins for
error of the nine female gymnasts (release window 43 – 102 ms) were comparable with
those of the three male gymnasts using the traditional technique (release window 79 – 84
ms).

Keywords : gymnastics, simulation, release window, angular momentum

Introduction
The double layout somersault dismount is performed in competition by male
gymnasts from the high bar and female gymnasts from the asymmetric bars (Figure 1). In both cases the gymnast uses the preced ing backward giant circle to generate
sufficient angular momentum and flight to co mplete the required number of somersaults
and to travel safely away from the bar. However, during giant circles leading up to a
dismount from the asymmetric bars (a-bar s) a female gymnast’s technique must
incorporate a strategy for avoiding the lower bar (Figure 1b). When performing
backward rotating dismounts, such as the double layout somersault dismount, the gymnast can either straddle th e legs to avoid the lower bar or increase the angle of hip
flexion (Figure 1b).

Figure 1. Double layout somersault dismounts (a) with twist from the high bar and (b) with a hyper-
extended body from th e asymmetric bars.

2
Performing a double layout somersault dismou nt in the straight position requires
enough angular momentum to produce 1.75 some rsaults during flight (or slightly less
since the gymnast is short of the vertical on landing). From the graphics sequences in
Figure 1 it might be expected that the female gymnast requires less normalised angular
momentum than her male equivalent since a hyper-extended configuration is usually
adopted in flight. Although Arampatzis a nd Brüggemann (1999) reported lower mean
normalised (for mass) angular momentum values for dismounts from the a-bars
compared to the high bar, the type of dism ounts performed is unclear. The graphic of
the final giant circle presented suggests that these dismounts were performed with a
tucked body shape rather than a layout.
Hiley and Yeadon (2003a) calculated the ma rgin for error when timing the release
for dismounts from the high bar. The margin for error was quantified in terms of the
release window during which time the gym nast has suitable linear and angular
momentum for performing the double layout di smount. If the gymnast releases at any
point during this window he will have suffici ent angular momentum and flight time to
complete the dismount. The release window was expressed in ms and the margin for
error was calculated as ± half the release window time. For consistency of performance
it is necessary that the margin of error in release timing is large enough to encompass
the timing precision of a gymnast. By definition there are consequences of failing to release the bar within the release window. An early rel ease, before the release window
has started, is likely to be characterised by insufficient time of flight leading to insufficient somersault rotation and excessive tr avel away from the bar. Releasing after
the release window has ended is characterised by insuffici ent angular momentum and
insufficient travel leading to a risk of striki ng the bar during flight. It might be expected
that gymnasts with larger ma rgins for error will be able to land their dismounts more
consistently. Determining the margins for error of performances may give some insight into the required timing precision.
Hiley and Yeadon (2003a) showed that the technique adopted in the backward giant
circle leading up to release influenced the si ze of the release window. It was shown that
the “scooped” backward giant circle techniqu e (Figure 1a) flattened the path of the mass
centre leading up to release, resulting in a larger release window than for a more
“traditional” technique (where the gymnast adopts a more extended configuration when
passing through the highest point of the giant circle). In particular, the size of the
release window was strongly correlated w ith the amount of hype r-extension of the
gymnast when passing through th e lowest point of the final giant circle. Since female
gymnasts must adjust their technique to avoi d the lower bar, it is not clear how the size
of the release windows will be affected.
The aim of this study was to determine the release windows for female gymnasts
performing double layout somersault dismount s from the a-bars at the Sydney 2000
Olympic Games. An additional aim was to determine how the technique of the female gymnasts differ from those of their male counterparts performing the same dismount
and the likely effect on the size of release window.

Methods
Subsections in Methods follow the protocol used to determine the release windows
for female gymnasts. Initially data collection was carried out in which Olympic
performances were recorded and the data pr ocessed for subsequent use with computer
simulation models . The models were then used to obtain matching simulations of the
actual performances and were in turn used to determine the release windows .

3
Data collection
All asymmetric bar performances from the Sydney 2000 Olympic games were
recorded using two digital video came ras (Sony Digital Handycam DCR-VX1000E),
operating at 50 Hz with shutter speeds of 1/ 600 s. The two cameras were located 8 m
above the landing surface and 30 m and 37 m from the a-bars with a camera axis intersection angle of 66° . Prior to the start of each competition a calibration structure
comprising 30 spheres of diameter 0.10 m spanning a volume measuring 3m × 4.5m ×
4m was positioned so as to include a giant circle and dismount from the upper bar.
The centres of the calibration spheres were di gitised in five vide o fields from both
camera views. The performances of the ni ne highest scoring gymnasts who passed the
low bar with legs together in the giant circle before releasing for a double layout
somersault dismount were selected for analysis . The last ¾ backward giant circle and
the dismount were digitised for each subject. In each of the movement fields the centre
of the hand, elbow, shoulder, hip, knee and an kle joint centres and toes on each side of
the body were digitised along with the centre of the gymnast's head and the centre of the
high bar midway between the gymnast's hands. The data obtained from digitising the images of the calibration spheres and their known locations were us ed to calculate the
11 Direct Linear Transformation parameters for each of the cameras (Abdel-Aziz and
Karara, 1971). The two sets of digitised movement data were synchronised using the method of Yeadon and King (1999). Synchronis ed digitised coordinate data from each
camera view along with the camera parameters were used to reconstruct the three-
dimensional locations of the body landmarks using the Direct Linear Transformation.
Joint angles for the left and right sides were averaged to produce input for a planar simulation model. Quintic splines (Wood and Jennings, 1979) were used to fit the
orientation and joint angle time histories so that derivatives could be obtained (Yeadon,
1990a).
A set of anthropometric measurements of a "mean" elite female gymnast was
obtained as the mean measurements taken from eight Romanian international gymnasts.
These mean values were then scaled for each of the nine competitors using segment
lengths and widths obtained from the video digitisation and inertia parameters were calculated using the model of Yeadon ( 1990b). The normalised angular momentum
about the mass centre during the dismount was calculated for each competitor (Yeadon, 1990c). The time of flight was determined fr om the field before there was clear space
between the gymnast’s hands and the bar to the field before contact between the
gymnast’s feet and the landing mat. The hor izontal and vertical displacements of the
mass centre during flight were used to calculate the horizontal and ver tical velocities at
release using a least squares fit and assuming constant acceleration.

Simulation model
A four segment planar model of a gymnast comprising arm, torso, thigh and lower
leg segments was used to simulate the movement around the ba r (Hiley and Yeadon,
2003b). The high bar and the gymnast's shoul der structure were modelled as damped
linear springs (Figure 2). The spring at the shoulder represented the increase in length of the gymnast between the wrist and the hip (i.e. not just the stretch at the shoulder
joint). In addition to the s houlder spring, there was a parameter that governed the extent
to which the torso segment lengthened as th e shoulder elevation angle was increased.

4

Figure 2. The four segment gymnast / high bar simulation model with damped springs representing bar
and shoulder elasticity.

Input to the simulation model comprised the segmental inertia parameters, the
stiffness and damping coefficients of the bar and shoulder springs, the initial displacement and velocity of th e bar, the initial angular velo city of the arm, the initial
orientation of the arm and the joint angle tim e histories in the form of quintic splines
obtained from the video analys is. Output from the model comprised the time histories
of the horizontal and vertical bar displacements, the linea r and angular momentum of
the model and the rotation angle φ (the angle from the vertical of the line joining the
neutral bar position to the mass centre).
The equations of motion were derived using Newton's Second Law and by taking
moments about the neutral bar position a nd the segment mass centres. The angular
momentum of the body about its mass centre was calculated as:
∑
=+φ=4
1 iii ii i ii ))ZX – XZ(m (I h &&&
(1)
where X i = (x i – x cm), Z i = (z i – z cm), (x cm, zcm) = whole body mass centre location,
mi = segmental mass, I i = segmental moment of inertia, iφ&= segmental angular velocity.
The angular momentum at releas e was normalised by dividing by 2 π times the
moment of inertia of the body about its ma ss centre when straight and multiplying by
the flight time to give the equivalent number of straight somersaults in the subsequent
flight phase. The time of flight of a si mulation was calculated from the release and
landing heights of the mass centre and the vertical velocity at release using the equation
for constant acceleration under gravity. The height of the mass centre on landing was taken from the video analysis of each gymnast.

Matching Simulations
In order to determine the release window using the simu lation model a close match
between the simulated and recorded perfor mances was required. The simulation model
was implemented with the Simulated Annealing optimisation algorithm (Goffe et al., 1994). A cost function F was established to measure the difference between the
recorded performance and a simulation of this pe rformance as defined in equation (2):
F = φ + 80(x
b + z b) + 20(h + cmx& + cmz&) + 5φo (2)
where φ = root mean squared (rms) differen ce in degrees between recorded and
simulated rotation angle, x b, zb = the rms differences between recorded and simulated
bar displacements, h = absolute difference in normalised angular momentum at release
between simulation and actual performance, cmx&,cmz& = absolute differences in linear
velocity at release between simu lation and actual performance, φo = absolute difference
in initial rotation angle between simulation a nd actual performance. The weightings of
the cost function F shown in equation (2) we re chosen so that each of the seven

5components of the cost function made appr oximately equal contri butions since they
were considered to be of equal importance.
Since the aim of the matching process was to provide close agreement between the
simulation and the actual performance leadi ng up to release only the last 135° of the
final giant circle was simulated. The subject -specific inertia parameters calculated for
each of the nine gymnasts were used in th e simulation model. The initial conditions,
including the initial angle, angular velocity and bar displacements, for each simulation
were taken from the corresponding video analysis. During the optimisation the
following parameters were allowed to vary in order to improve the match between the
recorded and simulated performance. The ve rtical bar stiffness was allowed to vary
between 13500 N.m-1 and 19286 N.m-1 to conform with the specifications of the
International Gymnastics Federation (FIG , 2000). The horizontal bar stiffness was
allowed to vary between 10800 N.m-1 and 19286 N.m-1 since it has been shown that the
bar can be less stiff in this direction (Ker win and Hiley, 2003). The damping coefficient
of the bar was allowed to vary between 0 N.s.m-1 and 1000 N.s.m-1. The stiffness and
damping coefficients of the shoulder spri ng were allowed to vary over wider ranges
than those of the bar springs, between 0 N.m-1 and 40000 N.m-1 and 0 N.s.m-1 and
10000 N.s.m-1, respectively, since there was less in formation availabl e regarding these
parameters. The masses of the arms and le gs were allowed to vary independently,
since they were based on scaling from segment lengths obtained from the video analysis, and the torso mass was adjusted to maintain whole body mass. The torso
length parameter was allowed to vary by up to 0. 1 m. In addition small variations in the
initial conditions, rotation angle and angular velocity, were permitted to compensate for
any digitisation errors propa gated in their calculation.

Release Windows
Once the optimisation procedure had provi ded a simulation to match the video
performance of the final 135° of rotation leading up to the release, the matching
simulation for each gymnast was continued bey ond the point of release so that a release
window could be determined. It was assumed that the gymnast maintained contact with the high bar and continued with the same joint angle changes that occurred after release.
The release window was defined as the pe riod of time for whic h the model possessed
normalised angular momentum within ± 10% of the range of actual release values,
landed with the mass centre between 1.0 m and 3.0 m from the bar and had a time of flight of at least 0.9 s. The angular momentum limits of ± 10% were chosen so that the
gymnast would be able to make compensato ry configurational ch anges in flight and
successfully land the doubl e layout somersault dismount (Hiley and Yeadon, 2003a)
The range of landing distances was based on th e mean range of recorded performances
± two standard deviations. Similarly th e minimum time of flight was based on the
average flight time minus 10%. One outlier (not included in the average) had a time of flight of 0.9s. It was speculated that this gymnast was close to the limit of performance.
For this individual the lower limit of the time of flight was reduced to 0.89s to allow a
window to be obtained. The release window was allowed to start before and end after
the actual release of the gy mnast so long as the above c onstraints were satisfied.
The relationship between gymnast configur ation and magnitude of release window
was determined by regressing the sum of the fl exion and extension angles at the hip and
shoulder joints at the highest point of the fina l giant circle against the size of the release
window. This was repeated for the lowest poi nt of the giant circle. The highest and
lowest points of the giant circle occurred when the gymnast’s mass centre passed
directly above or below the ne utral bar position, respectively.

6The release windows for the a-bar dismount s were compared with the windows of
male gymnasts in the 2000 Olympic high bar competition.

Results
The information from the video analysis wa s used to give the following results. The
reconstruction error for the video analysis of the 3m × 4.5m × 4m calibration volume
was calculated to be 0.013 m in each coordina te, with the field of view spanning more
than 7 m. The values for the normalised angul ar momentum and the ve rtical velocity at
release for the female gymnasts were generally smaller than those of the male gymnasts
using the traditional and the scooped backward giant circles (Table 1). However, in
some individual cases the normalised a ngular momentum achieved by the female
gymnasts was comparable with that of the ma le gymnasts. The horizontal velocity at
release was smaller than that of the male gymnasts who performed traditional backward
giant circles.

Table 1. Velocity of the mass centre and normalised angular momentum about the mass centre at release

competitor
(no.) horizontal
velocity
(m.s-1) vertical
velocity
(m.s-1) angular momentum
(straight
somersaults)
300 0.95 3.56 1.49
314 0.69 2.99 1.53
331 1.27 3.97 1.58
353 0.96 3.85 1.44
357 0.95 3.59 1.49
364 0.97 3.47 1.49
367 1.74 3.69 1.50
386 1.01 3.80 1.57
390 1.07 4.27 1.59
mean 1.09 3.68 1.52
traditional 2.11 4.89 1.58
scooped 1.27 4.38 1.65

Note: mean values for traditional and scooped circles are
taken from the study on high bar dismounts by Hiley and
Yeadon (2003a)

Over the nine performances studied the simulation model was able to match the
recorded rotation angle during the final 135° leading up to release to within 1.2° rms
difference, and the horizontal and vertical displacements of the bar to within 0.011 m
rms difference (Figure 3). The simulati on model matched the normalised angular
momentum and the linear velocities at releas e to within 1%. For the nine performances
the mean stiffness coefficient (vertical and horizontal combined) of the bar obtained in
the matching procedure was 14669 N.m-1, which lay within the limits as set out by the
FIG (2000). It was found that on average the bar was 19% less stiff in the horizontal
direction. The average damping coefficient for the bar was 55 N.m.s-1. The average
torso length parameter and the average stiffn ess and damping coefficients of the spring

7at the shoulder were less than 0.01 m and 25261 N.m-1 and 1003 N.s.m-1, respectively.
The corresponding values for the male gymnasts were 26129 N.m-1 and 174 N.s.m-1 for
the bar spring, 0.05 m fo r the torso length parameter and 56780 N.m-1 and 12904 N.s.m-
1 for the shoulder spring coefficients.

Figure 3. Typical matches between simulation (solid line) and actual performance (circles) for (a)
whole body rotation angle a nd (b) net bar displacement.

The release windows determined by simulation for the nine female gymnasts along
with the average values calculated for the ma le gymnasts are presented in Table 2. The
mean release window for the female gymnasts (n = 9) was 69 milliseconds (range 43 –
102 ms) whereas the mean windows for the male gymnasts were 81 ms (range 79 – 84
ms) and 127 ms (range 95 – 157 ms) for the traditional (n = 3) and scooped (n = 8)
techniques respectively.

Table 2. Sum of the flexion and extension angles at the hip and
shoulder at the highest and lowest points of the giant circle

competitor
(no.) sum of
angles at
highest
point
(°) sum of
angles at
lowest
point
(°) release
window
(ms) release
window
(°)
300 -16 -26 62 15
314 -8 -15 43 9
331 2 -23 77 21
353 15 -33 98 27
357 0 -29 67 16
364 6 -7 102 25
367 -31 -16 68 15
386 26 -38 61 14
390 3 -19 48 12
mean 3 -23 69 17
traditional 9 -12 81 26
scooped 125 -40 127 46

Note: (a) angles are defined away from a handstand position with
negative angles corresponding to an arched configuration
(b) last column expresses release window in terms of the change
in rotation angle

8There appeared to be no relationship between the sum of the flexion and extension
angles at the hip and shoulder at the highest and lowest points and th e size of the release
window (R2 = 0.07, p = 0.48 and R2 = 0.00, p = 0.87, respectively).

Discussion
Computer simulation is a power ful tool for investigating elite technique in sports
movements. Before simulation can be used fo r this purpose, however, it is essential that
the ability of the model to closely match an actual performance is investigated. Often
such an investigation leads to modifications of the model: for example the way in which
elastic elements are represented. Without a quantitative evaluation of a model, the
confidence that should be placed on the results of simulation is uncertain. In this study the simulation model was able to match the linear and angular momentum of nine giant
circles to within 1%. These simulations gave release windows ranging from 43 ms to
102 ms corresponding to a margin for error in th e range 22 ms to 51 ms. At this point it
is appropriate to note that “margin for er ror” is a misnomer in the sense that all
performances within the release window are acceptable and are theref ore not in error.
Perhaps a better term would be “margin of vari ation” since this would fit in better with
the idea that variation is a necessary pa rt of performance (Newell and Corcos, 1993).
It was speculated that since female gymnast s used backward giant circling techniques
similar to the traditional technique used by a minority of male gymnasts, they would
have comparable release windows. To de termine the release window for a gymnast
requires knowledge of what would have happe ned had the bar been released later than
in the actual performance. Using a com puter simulation model provided a means for
investigating this hypothetical scenario. This approach is limited by the assumption of
configuration changes when re leasing later than in the ac tual performance and also by
the somewhat arbitrary criteria for a succe ssful dismount. Although altering the criteria
may lead to changes in the size of the rele ase windows, it is likely that similar changes
would occur across all gymnasts and so the findings would not change.
The spring parameters obtained for the bar from the matching simulations lay within
the bounds set out for the appa ratus by the FIG norms (2000). The parameter values
obtained for the spring at the shoulder for the female gymnasts were considerably
smaller than those obtained for the male gymnasts. The spring is used to represent the stretch in the gymnast occurring between the wrist and the hip, not just the stretch in the shoulder region. It can also be seen that the amount the torso lengthened during
shoulder elevation was different between the two sets of gymnasts. The optimisation
procedure was used to obtain a close match of the parameters in the cost function
(equation 2) but there appears to have been a degree of interplay between the shoulder
stiffness and damping and the torso lengthening parameters. If individual figures for the
amount the torso lengthened as the shoulder el evation angle increased were available it
is likely a more meaningful comparison c ould be made between the female and male
shoulder spring parameters.
The mean release windows for the female gymnasts and the male gymnasts using the
traditional technique were comparable (Table 2). All of the male traditional release
windows fell within the range of release windo ws obtained for the female gymnasts. It
may be expected that if more than thr ee male gymnasts had used the traditional
technique a wider range of release window s would have been found. The reason why
the a-bar performances had release windows si milar to the traditional male gymnasts
may be understood by looking at the path of the mass centre dur ing the final giant circle
for typical male and female dismounts (Figure 4). The dashed circle in each case was
drawn with its centre at the neutral bar position and passing through the mass centre of

9the gymnast at the lowest point. In the trad itional technique and th e a-bars dismount the
path of the mass centre is almost circular and the velocity direction changes rapidly as the gymnast approaches release. The paths of the mass centre are not perfectly circular
due to the displacement of the bar and the fi nal actions at the hip and shoulder joints.
The mean change in the direction of mass centre velocity during the release window
was 22.2° ± 5.7° for the a-bars and 22.8° ± 1.4° for the traditional technique dismounts.
In contrast in the scooped technique there is a flattening of the path of the mass centre and the velocity direction cha nges less rapidly as the gymnast approaches release with a
mean change over the releas e window of 10.3° ± 6.2°. Th e corresponding changes in
rotation angle over the release window were also similar for the a-bars dismounts and
the traditional technique with mean values of 17.1° ± 5.9° and 26.3° ± 1.5° respectively.
Again the mean value for the scooped techni que (46.1° ± 9.7°) was quite different from
the other two groups and reflects the larger release windows of these gymnasts. The
roughly circular mass centre path of the trad itional circles led to the release window
being limited primarily by the direction of the mass centre velocity (Table 3). That is, at
the upper limit of the window there was insu fficient horizontal velocity to carry the
gymnast safely away from the bar. In cont rast the flattened path of the mass centre for
the scooped circles resulted in less constraints with regard to velocity direction, with the
upper limit of the release window being dete rmined primarily by insufficient angular
momentum. As expected the lower limits to the release windows for the majority of
dismounts corresponded to excessive travel or insufficient time of flight. The limiting
variables for four of the a-bars release windows were the same as those for the
traditional dismounts while the remainder were not characteristic of the majority of
scooped dismounts (Table 3). All the release windows had the upper limit of the
window determined by either insufficient tr avel or insufficien t angular momentum.

Figure 4. Typical paths of the mass centre during the last half of the backward giant circle leading up
to release for one female gymnast, one male gymnast using a traditional technique and one
male gymnast using a scooped technique. The fla tter final part of the scooped path results in
a more consistent flight trajectory and a larger release window.

10
Table 3. The parameters which limited the release window at the start and end of the release window

trial
(no.) release
window
(ms) release
angle
(°) window
start
(°) window
end
(°) release window limited by
start end
300 62 254 243 258 excessive AM insufficient AM
314 43 258 251 261 excessive AM insufficient AM
331 77 250 238 258 excessive travel insufficient AM
353 98 248 230 257 insufficient flight time insufficient travel
357 67 247 237 253 insufficient flight time insufficient travel
364 102 253 235 259 insufficient flight time insufficient travel
367 68 247 242 258 excessive travel insufficient travel
386 61 241 233 247 insufficient flight time insufficient travel
390 48 250 243 255 excessive AM insufficient AM

traditional x 3 insufficient flight time insufficient travel
scooped x 4 excessive travel insufficient AM
scooped x 2 insufficient flight time insufficient AM
scooped x 1 excessive AM insufficient AM
scooped x 1 excessive travel insufficient travel

Note: AM = angular momentum

The angular momentum for the dismount is influenced by the gymnast’s actions at
the hip and shoulder joints during the giant ci rcle leading up to rel ease. It was found
that the female gymnasts’ technique was affect ed by the necessity to clear the lower bar.
For the female gymnasts maximum hip hyper-ex tension occurred after the lowest point
of the circle (Figure 5a), whilst for the male gymnasts it occurred before the lowest point. A similar relationship was found for the shoulder angle (Figure 5b). Flexing the
hip over a larger range of motion may be ex pected to increase the gymnast’s angular
momentum. In the present study none of the female gymnasts used a scooped technique, or a technique that produced a similar effect on the path of the mass centre leading up to release. This may account for the lack of correlation between joint angles
and release window for the female gymnasts.
The scooped technique is characterised by a large hyper-extension at the hip as the
gymnast passes through the lowest point of the circle. The subsequent rapid hip flexion
(Figure 5) helps produce the loading on the bar associated with the flattened path of the
mass centre. It is speculated that achieving a la rger hyper-extension at the hip earlier in
the giant circle may help increase the size of a female gymnast’s window and lead to greater angular momentum. Future inves tigations could be carried out using the
simulation model to determine whether this ca n be achieved given th e constraint of the
lower bar.

11

Figure 5. Typical hip (a) and shoulder (b) joint angle time histories for female (solid line) and male
gymnasts (circles) over the last 135° of the giant circle leading up to release.

The majority of female gymnasts perf orming the double layout dismount adopted a
hyper-extended “arched” position dur ing flight (Figure 1b). As is the trend with male
dismounts the female gymnasts are introduc ing a full twist into the double layout
dismount. As the number of twis ts increases so must the time the gymnast spends in the
straight position which is more efficient fo r twisting than a hyper-e xtended position. In
order to achieve this the gymnast will require either a longer time of flight or more
angular momentum. Increasing the time of flight or the angular momentum of the
gymnast at release is likely to reduce the size of the release window, as shown by Hiley
and Yeadon (in press). Therefore the technique in the final giant circle prior to release
may need to be developed in order to main tain a sufficient release window despite the
increase in angular velocity.
In general the techniques used by the fema le gymnasts produced similar results to
those of the traditional technique used by male gymnasts. However, the limiting factor to the size of the release window was not c onsistent between gymnasts and may have
been the result of variations in the timing of the final shoul der extension and hip flexion.
This raises questions regarding the gymnast’s ability to reproduce the same technique
and release window on repeated trials. We re the gymnasts with the larger release
windows more accomplished and better able to ti me their final actions correctly or was
43 ms a sufficiently large release window fo r consistent timing of the release? This
could not be determined from the present study and would require the calculation of a
gymnast’s release window from repeated trials.

Acknowledgements
The authors wish to acknowledge the suppor t of the British Gymnastics World Class
Programme, Sport England, UK Sport, Pfitzer and the International Olympic Committee
Medical Commission.

References
Abdel-Aziz, Y.I., and Karara, H.M. ( 1971). Direct linear transformation from
comparator coordinates into object space coordinates in close-range
photogrammetry. ASP Symposium on Close-Range Photogrammetry (pp. 1-18). Falls
Church, VA: American Soci ety of Photogrammetry.
Arampatzis, D. and Brüggemann, G-P. (1999). Mechanical and energetic processes
during the giant swing exercise before di smount and flight elements on the high bar
and uneven parallel bars. Journal of Biomechanics , 32, 811-820.
Fédération Internationale de Gymnastique 2000. Manual apparatus norms. Part IV
testing procedures. F.I.G. Moutier. Switzerland.

12Goffe, W.L., Ferrier, G.D. and Rogers, J. (1994). Global optimiza tion of statistical
functions with simulated annealing. Journal of Econometrics , 60, 65-99.
Hiley, M.J. and Yeadon, M.R. (2003a). The ma rgin for error when releasing the high
bar for dismounts . Journal of Biomechanics , 36, 313-319.
Hiley, M.J. and Yeadon, M.R. (2003b). Opt imum technique for generating angular
momentum in accelerated backward giant circles prior to a dismount . Journal of
Applied Biomechanics , 19, 119-130.
Hiley, M.J. and Yeadon, M.R. (in press). Maximal dismounts from high bar . Journal of
Biomechanics .
Kerwin, D.G. and Hiley, M.J. (2003). Estimation of reaction forces in high bar
swinging. Sports Engineering , 6, 21-30.
Newell, K.M. and Corcos, D.M. (1993). Variability and moto r control. Human
Kinetics. Champaign, IL.
Wood, G.A. and Jennings, L.S. (1979). On the use of spline functions for data
smoothing. Journal of Biomechanics , 12, 477-479.
Yeadon, M.R. (1990a). The simulation of aer ial movement – I. The determination of
orientation angles from film data. Journal of Biomechanics , 23, 59-66.
Yeadon, M.R. (1990b). The simulation of aeria l movement – II. A mathematical inertia
model of the human body. Journal of Biomechanics , 23, 67-74.
Yeadon, M.R. (1990c). The simulation of aer ial movement- III. The determination of
the angular momentum of the human body. Journal of Biomechanics , 23, 75-83.
Yeadon, M.R. and King, M.A. (1999). A method for synchronisi ng digitised video
data. Journal of Biomechanics , 32, 983-986.