SEARCH

SEARCH BY CITATION

Keywords:

  • horse;
  • electromyography;
  • treadmill;
  • speed;
  • gradient

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

Reasons for performing study: Locomotion requires successful negotiation of different terrains, but we currently know little of how the musculoskeletal system adapts to cope with positive and negative slopes.

Objective: To compare the effects of treadmill speed and gradient on equine hindlimb muscle mean electromyographic (EMG) intensity.

Methods: Surface EMG recorded the activity of gluteus medius (GM), biceps femoris (BF), vastus lateralis (VL), gastrocnemius lateralis (GL) and extensor digitorum longus (EDL) in 6 horses at walk (1.4–1.6 m/s) and trot (2.6–3 m/s) on 3 different treadmill gradients (0, 10% and −10%). Significant differences in mean EMG intensity and kinematic data were determined using Friedman and Wilcoxon signed rank tests (P<0.05).

Results: Increasing velocity increased the mean EMG intensity of GM, BF and GL regardless of gradient. Treadmill incline increased the mean EMG intensity for all muscles at the walk and that of GM at trot. Treadmill decline reduced the mean EMG intensity of GM at both the walk and the trot and that of BF at the walk, but not the trot. The mean EMG intensity of EDL, VL and GL remained similar at both gaits when compared to the horizontal.

Conclusions: The hip retractors are the primary muscles responsible for powering equine locomotion in response to increasing workload.

Potential relevance: A better understanding of the effects of speed and gradient on the functional activity of the horses' locomotor muscles will enable the development of more effective training programmes pre- and post injury.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

Steady state locomotion rarely takes place other than in a laboratory setting. Most of the time, limb muscles are required to alter their activity and power output in response to changes in surface gradient, gait and velocity. Previously, studies have tended to focus on how these changes affect the metabolic cost of locomotion (MCL). Typically, horses' MCL increases with velocity and treadmill inclination (Wickler et al. 2000), doubling on a 10% gradient (Eaton et al. 1995) and almost halving on a 10% decline (Hoyt et al. 2006). Such shifts in locomotor energetics relate to changes in muscle recruitment and activation levels and whilst previous research has looked at the effects of velocity and incline on muscle activity, very little is known about exercising on a decline.

A better understanding of the effects of speed and gradient on the functional activity of the horses' locomotor muscles would provide an evidence base to develop effective training programmes pre- and post injury.

The purpose of this study was to measure the linear and temporal kinematics of the equine hindlimb and relate them to hindlimb muscle activity. Our aim was to compare the effect of treadmill speed and gradient (0, 10% and −10%) on the mean electromyographic (EMG) intensity in the equine hindlimb muscle. We made the following hypotheses: firstly, that the mean EMG intensity of the hip retractors (gluteus medius, GM and biceps femoris, BF) and the antigravity muscle (vastus lateralis, VL) would increase with speed and a 10% treadmill incline and would decrease on a 10% declined treadmill; and secondly, that the mean EMG intensity of gastrocnemius lateralis (GL) and extensor digitorum longus (EDL) would remain unaltered by speed, regardless of gradient, but that the relative mean EMG intensity of these muscles would increase on an incline and decrease on a decline.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

This study was approved by the Royal Veterinary College (RVC) Animal Ethics Committee and conducted at the RVC Structure and Motion Laboratory.

Subjects

Data were collected from 6 adult Welsh Mountain mares of varying wither height (1.30 ± 0.10 m) and weight (378 ± 89 kg). All horses were deemed sound by an attending veterinarian and had previously undergone high-speed treadmill training sessions at walk, trot and canter so that they were habituated to the treadmill as recommended by Buchner et al. (1994).

Kinematic data

Hemispherical 25 mm polystyrene markers covered in Scotchlite reflective tape were glued to the skin on the lateral aspect of the limb, over standard palpable bony landmarks (Back et al. 1993). Data were collected (200 Hz) from 8 Opus 300 series cameras1 positioned at a height of 1.9 m, 4 m away from the treadmill2.

EMG data

Prior to the application of self-adhesive electrodes3 (interelectrode distance of 15 mm), which were positioned parallel to the direction of the muscle fibres (Crook et al. 2008), an area of skin over the midpoint of each of the muscle bellies of GM, BF, VL, GL and EDL was clipped, shaved and cleaned with ethyl acetone and isopropyl solution. The ground electrode was positioned over the left tuber coxae. The EMG signal (2000 Hz) was amplified at source close to the recording electrode and filtered using a third-order high-pass Butterworth filter with a 20 Hz cut-off point.

A solid-state capacitive accelerometer (ADXL50)4 (dynamic range of ± 50 g), encased in epoxy impregnated Kevlar fibres, was secured with hot glue to the dorsal midline of the left hind hoof wall and signalled hoof impact acceleration (Parsons and Wilson 2006). A microcomputer (Q1)5 running a custom-made Labview6 script was attached to the horses' surcingle and simultaneously logged accelerometer data, sampled at a bit rate of 128 kbps (44.1 kHz), and EMG data.

Treadmill protocol

Prior to data collection, the horses walked on the horizontal treadmill at their preferred speed for 5 min. Following this, the horses completed a designated exercise protocol consisting of 3 min each of walk (1.4–1.6 m/s) and trot (2.6–3.0 m/s) on each of the 3 gradients (horizontal [0%], incline [10%] and decline [-10%]). The sequence of these conditions was randomised between subjects. Accelerometer and EMG data were collected for 120 s once the horses' gait had stabilised.

Froude number calculations were performed for each horse; this enabled standardisation of walking and trotting speed according to their wither height (Alexander 2005).

Data analysis

Temporal kinematic data: Kinematic data were determined for each animal for 10 consecutive strides for each condition. The vertical position of the distal hoof marker was used to determine toe on and toe off, with initial ground contact being the lowest point reached by the marker and toe off when the position of the marker started to rise. Stride duration was established as the time taken for successive impacts of the hindlimb to occur, with the movement of the distal hoof marker being used to determine the duration of the stance and swing phases (Hodson-Tole 2006). Stride length (the product of stride duration and treadmill speed) and duty factor (the percentage of the stride that the foot is in contact with the treadmill), were determined in Excel 20027.

EMG data: Electromyographical data were determined for each animal for a minimum of 50 strides for each condition; ‘atypical’ strides were removed prior to analysis. Wavelet analysis was used to simultaneously resolve these data into time and frequency components, using a filter bank of nonlinearly scaled wavelets with physiologically relevant time-resolutions. Each of the wavelets acted as a band-pass filter characterised by its centre frequency, bandwidth and time resolution (von Tscharner 2002; Wakeling et al. 2002; Wakeling and Rozitis 2004). The raw EMG signal was convoluted with wavelets 0–9, with central frequencies spanning 6.90–330.63 Hz and the EMG intensity was calculated from the slope and magnitude of the square of the signal.

Statistics

Mean and s.d. were calculated for the kinematic variables and EMG intensities. Normality of data distribution was determined using a one-sample Kolmogorov-Smirnov test. Comparisons between conditions were made using the Friedman test. Where significant differences (P<0.05) were identified, pair-wise Wilcoxon signed rank tests were applied.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

Results of analysis of kinematic data

Data for temporal and linear kinematics were remarkably consistent between horses (Table 1).

Table 1. Hindlimb linear and temporal kinematic values for each gait, showing mean ± s.d. (n = 6)
VariableWalk (1.4–1.6 m/s)Trot (2.6–3.0 m/s)
0%10%-10%0%10%-10%
  • *

    Values significantly (P<0.05) different from corresponding horizontal treadmill values.

  • Values significantly (P<0.05) different from corresponding walk values.

Stride length (m)1.49 ± 0.181.5 ± 0.151.52 ± 0.191.84 ± 0.161.84 ± 0.201.85 ± 0.16
Stride duration (s)0.98 ± 0.080.99 ± 0.061.00 ± 0.080.65 ± 0.020.64 ± 0.040.65 ± 0.02
Stance duration (s)0.64 ± 0.050.66 ± 0.030.64 ± 0.050.32 ± 0.020.33 ± 0.01*0.32 ± 0.01
Swing duration (s)0.34 ± 0.030.33 ± 0.020.36 ± 0.030.33 ± 0.020.31 ± 0.030.33 ± 0.01
Stride frequency (strides/s)1.03 ± 0.081.01 ± 0.051.01 ± 0.081.55 ± 0.051.56 ± 0.091.54 ± 0.05
Duty factor0.66 ± 0.010.67 ± 0.010.64 ± 0.000.50 ± 0.020.51 ± 0.020.49 ± 0.01

Velocity: An increase in treadmill velocity resulted in a change of gait from the walk to the trot for all the horses. This was accompanied by a significant (P<0.05) increase in stride length and stride frequency and a significant decrease in stride duration, stance duration and duty factor. Swing duration remained similar at the walk and trot.

10% incline: Exercising on the 10% inclined gradient had no significant effect on any of the kinematic variables: stance duration (P = 0.102 walk; P = 0.180 trot); stride duration (P = 0.593 walk; P = 1.00 trot); stride frequency (P = 0.465 walk; P = 0.786 trot); stride length (P = 0.465 walk; P = 0.893) and duty factor (P = 0.063 walk; P = 0.063 trot).

10% decline: Exercising on the 10% declined gradient had no significant effect on any of the kinematic variables: stance duration (P = 0.581 walk; P = 0.783 trot); stride duration (P = 0.285 walk; P = 0.083 trot); stride frequency (P = 0.276 walk; P = 0.216 trot); stride length (P = 0.225 walk; P = 0.138) and duty factor (P = 0.063 walk; P = 0.680 trot).

Results of analysis of EMG data (Fig 1)

image

Figure 1. Mean±s.d. intensity of the EMG signal at walk (1.4–1.6 m/s) and trot (2.6–3.0 m/s) on a 0% and 10% incline and 10% decline (mean EMG intensities for each muscle are normalised to their mean EMG intensity at the walk at 0% gradient) (n=6). a) gluteus medius; b) biceps femoris; c) vastus lateralis; d) gastrocnemius lateralis; e) extensor digitorum longus.

Download figure to PowerPoint

Velocity: An increase in treadmill velocity significantly increased the mean EMG intensity of GM (P≤0.046), BF (P≤0.028) and GL (P≤0.028) for all of the 3 gradients. Vastus lateralis mean EMG intensity increased with speed on the inclined (P = 0.028) and declined treadmill (P = 0.043) but not on the horizontal (P = 0. 116). Extensor digitorum longus increased its mean EMG intensity with an increase in velocity on the horizontal (P = 0.028) and inclined gradient (P = 0.046), but not on the decline (P = 0.80).

10% incline: When compared to the horizontal, the mean EMG intensity increased for GM (P = 0.028), BF (P = 0.028) and GL (P = 0.028) at the walk on the inclined treadmill. The mean EMG intensity of GM (P = 0.028) also significantly increased when trotting on the inclined treadmill when compared to the horizontal. However, treadmill incline made no significant difference to the mean EMG intensity of the remaining muscles at either gait.

10% decline: In comparison to the results from the horizontal treadmill, the mean EMG intensity of GM (P = 0.028), BF (P = 0.028) reduced for horses walking on the declined treadmill; the mean EMG intensity of GM also decreased (P = 0.28) when the horses trotted on the decline. There was not a significant difference in the mean EMG intensity for VL, GL or EDL at either the walk (VL: P = 0.917; EDL: P = 0.08, GL: P = 0.249) or trot (VL: P = 0.50; EDL P = 0.753, GL: P = 0.753).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

This study examined the effects of treadmill speed and gradient on the mean EMG intensity of GM, BF, VL, GL and EDL and demonstrated that the horse alters its hindlimb kinematics and muscle recruitment patterns to accommodate for differences in treadmill velocity and gradient.

The effects of velocity

An increase in horizontal treadmill velocity resulted in a change in gait from walk to trot and was also associated with an increase in stride length and stride frequency and a reduction in stride duration, stance duration and duty factor. As hypothesised, the mean EMG intensity of GM and BF increased with velocity on all 3 gradients. Regardless of gradient, GM almost trebled its mean EMG intensity. Biceps femoris mean EMG intensity was almost 15 times greater on the horizontal and 5 times greater on both the inclined and declined treadmill at the faster speed (Fig 1). A velocity-dependent increase in both muscles' EMG amplitude has previously been reported (Wentink 1978; Robert et al. 2000) and probably results from the observed changes in the limb kinematics that occur with speed. The increase in stride length and stride frequency and reduction in duty factor at the faster gait mean that both muscles have to move the hip joint through a greater range of motion at a faster rate, whilst withstanding an increase in peak vertical forces.

Contrary to our hypotheses, VL only increased its mean EMG intensity with an increase in velocity on the inclined and declined treadmill. One possible explanation is that muscles other than VL increase their activity to meet the increased power demands of the faster velocity on the horizontal treadmill. Indeed, studies concerning inverse dynamics indicate that most of the positive work done at the trot is by the proximal muscles such as GM and BF and that the stifle muscles are most likely to be involved in stabilising the stifle joint, allowing efficient energy transfer from the hip musculature to the ground (Dutto et al. 2006). As such, the increase in its EMG activity, with increasing velocity on a gradient, is probably more a reflection of a significant increase in both concentric and eccentric muscle work necessitated by the faster gait on an inclined and declined treadmill, respectively (Wickler et al. 2005).

We had hypothesised that the mean EMG intensity of GL and EDL would remain similar, with an increase in velocity reflecting the roles of the distal limb muscles as elastic spring-like mechanisms. Contrary to our hypothesis, the mean EMG intensity of both GL and EDL increased with velocity on all 3 gradients (although this was not significant for EDL on the decline, where P = 0.80). It appears that increasing velocity increases limb forces independent of the treadmill gradient.

The effects of a 10% incline

The observed (but not significant) increase in stance time (Robert et al. 2002) and duty factor at both gaits allowed the propulsive muscles a greater time for power generation. Correspondingly, we saw an increase in the mean EMG intensity of GM, BF and VL at the walk and at the trot (although this was only significant for GM, P = 0.028 at trot; BF, P = 0.345; VL, P = 0.753), indicating that these muscles are contributing to the increase in hindlimb power output. This increase in muscle workload is because of an increase in gravitational potential energy and a shift in the force distribution between the fore- and hindlimbs associated with locomotion on an inclined gradient (Dutto et al. 2004).

As we had hypothesised, the distal limb muscles LG and EDL muscles more than doubled their activity at walk on the incline when compared to the flat (although this was not significant for EDL P = 0.075) emphasising their important extensor role. However, only LG significantly increased its mean EMG intensity at the trot on the incline when compared to trotting on the horizontal. This was surprising as we expected both muscles to increase their activity to at both gaits. It is possible that an increase in the storage and return of elastic strain energy at the trot on the incline negated the need for greater muscle activity.

The effects of 10% decline

On the declined gradient, the horses did not significantly reduce their stride length, stance time or duty factor. Although it appears, the horses negotiated the 10% decline at the walk by taking a greater length of stride and reducing the amount of time that their foot spent in contact with the treadmill belt when applying a braking force. As the walk is a 4 beat gait and has less kinetic energy associated with it when compared to the trot, it is likely that the reduction in duty factor still allowed sufficient braking forces to be developed by the 3 remaining limbs in contact with the treadmill. Swing duration was similar between walk and trot on all 3 gradients, probably owing to the passive nature of limb protraction (Wilson et al. 2003), which is similar over a range of speeds.

As hypothesised, GM was less active on the decline when compared to the horizontal treadmill at both the walk and at the trot (P = 0.028).

The declined treadmill negates the need for active hip retraction and passive limb retraction occurs owing to hoof contact with moving treadmill surface. Therefore, we suggest that the recorded EMG activity results from the muscle activity that is required to stabilise the hip joint and allow other muscles (such as BF) to act about it.

Biceps femoris mean EMG intensity reduced on the decline at both the walk (P<0.028) and the trot when compared to the horizontal gradient, although this reduction was not significant at the faster gait (P<0.345). The observed differences in the 2 hip retractors' relative EMG intensities probably relates to the biarticular nature of BF. As previously noted, the passive nature of hip retraction on the negative gradient reduces its extensor role at this joint, but its distal attachment means that it probably exerts an extensor moment at the stifle, working eccentrically to slow the animals' rate of decent. Eccentric muscle work enables a muscle to produce greater peak forces with a lesser volume of active muscle fibres and could explain the reduction in EMG intensity on the decline when compared to the horizontal gradient (Gillis and Biewener 2002; Gabaldon et al. 2004). Eccentric muscle work may also explain why there was no significant difference in the mean EMG intensity of VL on the decline when compared to the horizontal gradient at either gait. Studies of other mammals have shown VL is more suited to eccentric muscle work on a decline when compared to the other quadriceps muscles (Gillis and Biewener 2002).

The mean EMG intensity of GL on the declined gradient was similar to that on the horizontal at both the walk and the trot. The mean EMG intensity of EDL was 6 times greater at walk on the decline when compared to the horizontal gradient and slightly greater at the trot but not significantly so. Although we are unable to confirm conclusively (as is the case with other extensor muscles), it is likely that these differences also reflect an increase in eccentric muscle work associated with work on a decline.

The relatively small sample size used in this study, together with some of the reported standard deviations in the data set (particularly with regard to vastus lateralis) possibly explains why statistical differences between conditions were not more marked. It is interesting to note that recruitment patterns appear to vary between individual animals particularly at the trot. This may be a true indication of individual differences in motor recruitment or may reflect some of the limitations associated with collecting surface EMG such as skin displacement and crosstalk between adjacent muscles. Every attempt was made to limit cross talk by careful siting of the electrodes, but the possibility that some interference from adjacent muscles and electrode movement can not be ruled out completely. As such, the intention of this study is to provide a preliminary insight into the effects of speed and gradient on muscle activity and it is hoped that further research will build on our findings.

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

This study confirms the important role of the hip retractors as the primary muscles responsible for powering equine locomotion in response to increasing workload and demonstrates that gradient has minimal effect on the work done by the intrinsic leg muscles of the hindlimb.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

Special thanks to the students and staff of the Structure and Motion Laboratory at The Royal Veterinary College, London.

Manufacturers' addresses

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References

1 Qualysis Medical AB, Gothenburg, Sweden.

2 Sato AB, Knivsta, Sweden.

3 Norotrode Myotronics, Inc. Kent, Washington, USA.

4 Analog Devices Inc., Norwood, Massachusetts, USA.

5 Samsung Electronics, Chertsey, Surrey, UK.

6 Labview National Instruments Inc., Houston, Texas, USA.

7 Microsoft Corporation, Redmond, Washington, USA.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. Conflict of interest
  10. Manufacturers' addresses
  11. References
  • Alexander, R.M. (2005) Models and the scaling of energy costs for locomotion. J. expt. Biol. 208, 1645-1652.
  • Back, W., Van Den Bogert, A.J., Van Weeren, P.R., Bruin, G. and Barneveld, A. (1993) Quantification of the locomotion of Dutch Warmblood foals. Acta Anat. (Basel) 146, 141-147.
  • Buchner, H.H., Savelberg, H.H., Schamhardt, H.C., Merkens, H.W. and Barneveld, A. (1994) Kinematics of treadmill versus overground locomotion in horses. Vet. Quart. 16 ( Suppl. 2), S87-S90.
  • Crook, T.C., Cruickshank, S.E., McGowan, C.M., Stubbs, N., Wakeling, J.M., Wilson, A.M. and Payne, R.C. (2008) Comparative anatomy and muscle architecture of selected hind limb muscles in the Quarter Horse and Arab. J. Anat. 212, 144-152.
  • Dutto, D.J., Hoyt, D.F., Clayton, H.M., Cogger, E.A. and Wickler, S.J. (2006) Joint work and power for both the forelimb and hindlimb during trotting in the horse. J. expt. Biol. 209, 3990-3999.
  • Dutto, D.J., Hoyt, D.F., Cogger, E.A. and Wickler, S.J. (2004) Ground reaction forces in horses trotting up an incline and on the level over a range of speeds. J. expt. Biol. 207, 3507-3514.
  • Eaton, M.D., Evans, D.L., Hodgson, D.R. and Rose, R.J. (1995) Effect of treadmill incline and speed on metabolic rate during exercise in Thoroughbred horses. J. appl. Physiol. 79, 951-957.
  • Gabaldon, A.M., Nelson, F.E. and Roberts, T.J. (2004) Mechanical function of two ankle extensors in wild turkeys: shifts from energy production to energy absorption during incline versus decline running. J. expt. Biol. 207, 2277-2288.
  • Gillis, G.B. and Biewener, A.A. (2002) Effects of surface grade on proximal hindlimb muscle strain and activation during rat locomotion. J. appl. Physiol. 93, 1731-1743.
  • Hodson-Tole, E. (2006) Effects of treadmill inclination and speed on forelimb muscle activity and kinematics in the horse. Cambridge Journals Online .
  • Hoyt, D.F., Wickler, S.J. and Garcia, S.F. (2006) Oxygen consumption (VO2) during trotting on a 10% decline. Equine vet. J., Suppl. 36, 573-576.
  • Parsons, K.J. and Wilson, A.M. (2006) The use of MP3 recorders to log data from equine hoof mounted accelerometers. Equine vet. J. 38, 675-680.
  • Robert, C., Valette, J.P. and Denoix, J.M. (2000) The effects of treadmill inclination and speed on the activity of two hindlimb muscles in the trotting horse. Equine vet. J. 32, 312-317.
  • Robert, C., Valette, J.P., Pourcelot, P., Audigie, F. and Denoix, J.M. (2002) Effects of trotting speed on muscle activity and kinematics in saddlehorses. Equine vet. J., Suppl. 34, 295-301.
  • Von Tscharner, V. (2002) Time-frequency and principal-component methods for the analysis of EMGs recorded during a mildly fatiguing exercise on a cycle ergometer. J. electromyogr. Kinesiol. 12, 479-492.
  • Wakeling, J.M. and Rozitis, A.I. (2004) Spectral properties of myoelectric signals from different motor units in the leg extensor muscles. J. expt. Biol. 207, 2519-2528.
  • Wakeling, J.M., Kaya, M., Temple, G.K., Johnston, I.A. and Herzog, W. (2002) Determining patterns of motor recruitment during locomotion. J. expt. Biol. 205, 359-369.
  • Wentink, G.H. (1978) Biokinetical analysis of the movements of the pelvic limb of the horse and the role of the muscles in the walk and the trot. Anat. Embryol. (Berl) 152, 261-272.
  • Wickler, S.J., Hoyt, D.F., Biewener, A.A., Cogger, E.A. and De La Paz, K.L. (2005) In vivo muscle function vs speed. II. Muscle function trotting up an incline. J. expt. Biol. 208, 1191-1200.
  • Wickler, S.J., Hoyt, D.F., Cogger, E.A. and Hirschbein, M.H. (2000) Preferred speed and cost of transport: the effect of incline. J. expt. Biol. 203, 2195-2200.
  • Wilson, A.M., Watson, J.C. and Lichtwark, G.A. (2003) Biomechanics: A catapult action for rapid limb protraction. Nature 421, 35-36.