Search tips
Search criteria 


Logo of brjsmedBritish Journal of Sports MedicineVisit this articleSubmit a manuscriptReceive email alertsContact usBMJ
Br J Sports Med. 2007 November; 41(11): 739–744.
PMCID: PMC2465259

Shoulder joint kinetics of the elite wheelchair tennis serve



The shoulder is a key joint in wheelchair locomotion and commonly implicated in injury among virtually all wheelchair populations. In tennis, quantification of the shoulder joint kinetics that characterise the wheelchair serve could enhance injury prevention and rehabilitation practices as well as assist coaches evaluate the efficacy of their current technical instruction.


A 12‐camera, 250 Hz Vicon motion analysis system (Oxford Metrics Inc., UK) recorded the 3D flat (WFS) and kick serve (WKS) motions of two male top 30‐ranked international wheelchair players. Mechanical comparisons between wheelchair players, as well as to the previously captured data of 12 high‐performance able‐bodied players executing the same types of serves, were undertaken.


Without the benefit of a propulsive leg action, wheelchair players developed lower peak absolute (~32 m/s) and horizontal (~28 m/s) pre‐impact racquet velocities than able‐bodied players (~42 m/s, ~38 m/s). Wheelchair serve tactics nevertheless necessitated that higher pre‐impact horizontal and right lateral racquet velocities characterised the WFS (~29 m/s, WKS: ~26 m/s) and WKS (~4 m/s, WFS: ~11 m/s) respectively. The shoulder joint kinetics that contributed to the differential racquet velocity profiles were mostly developed independent of wheelchair serve type, but varied with and were likely related to the level and severity of spinal cord injury of the individual players.


Compared with able‐bodied players, wheelchair players experienced matching pre‐ and post‐impact shoulder joint loads, such that wheelchair and able‐bodied playing populations appear subject to similar shoulder joint injury risk.

Current technical instruction of the wheelchair tennis serve is largely intuitive, guided to some extent by the substantiated biomechanical information describing the able‐bodied serve. The link between shoulder pain and wheelchair serve performance has similar origins, especially as wheelchair players, like their able‐bodied counterparts, commonly suffer from shoulder overuse injuries.1,2 Delineation of the shoulder joint kinetics that contribute to the development of racquet velocity in the flat (WFS) and kick (WKS) wheelchair serves, and that are associated with shoulder joint injury, is thus important and was the focus of the present study. Further consideration of this data alongside the shoulder joint kinetics associated with able‐bodied serve performance provided for comparative comment.


Subject performance

The International Tennis Federation (ITF) consider any player to have a medically diagnosed permanent physical disability resulting in a substantial loss of function in one or both lower extremities as eligible to compete on the wheelchair tennis tour. Maximal effort WFSs and WKSs of two top‐30 professionally ranked, right‐handed wheelchair players were recorded by a 12‐camera, 250 Hz Vicon motion analysis system (Oxford Metrics Inc., Oxford, UK). Subject 1, a quarter finalist at the Australian Wheelchair Open in 2005, suffered from an incomplete injury at the T12 level but a complete break at the level of L1, and thus had no lower limb function. Conversely, Subject 2 suffered from an incomplete T10 spinal cord injury, thus exhibiting more trunk and lower limb function, and was able to walk unassisted for brief periods. Participation in the study was contingent on both subjects reading the subject information sheet and signing a consent form, and all experiments were approved by the Ethics in Human Research Committee of the University of Western Australia.

Data from three successful WFSs and WKSs that landed in a 1×1 metre target area bordering the “T” of the first service box were recorded and averaged.

Data treatment and statistical analysis

Both players used the same 0.690 m, 0.380 kg Wilson 6.0 Pro Staff racquet (x: 172.7 kg/cm2, y: 15.3 kg/cm2, z: 188.0 kg/cm2), whose inertial parameters were calculated using procedures outlined by Brody et al.4 The players' individual wheelchairs were used, and segmental masses and moments of inertia used in the inverse dynamics approach to calculate upper limb kinetics were derived from previously published data.5,6

Players were fitted with customised UWA upper‐body marker sets (31 individual retro‐reflective markers, all 16 mm in diameter;3) (fig 11).). Three‐dimensional coordinates were expressed in a right‐handed inertial reference frame, where the origin was at the centre of the baseline. Positive x was pointing forward, positive y was vertical and pointing upwards, while z was perpendicular to x and y and positive to the right. Preliminary analyses established data treatment techniques to minimise end‐point error, as well as a most appropriate level of smoothing (mean squared error, MSE of 25) that best described the 3D service motion (as detailed in Reid et al).3 Filtered outputs were then modelled in the Euler angle decomposition sequence: flexion(+)/extension(−), adduction(+)/abduction(−) and internal(+)/external(−) rotation of the moving segment coordinate system with respect to the fixed segment coordinate system (Z–X–Y), to describe motion at all joints except the shoulder, where a Y–X–Y decomposition was shown as preferable.7,8,9 Previously, the ISB has recommended the Y‐X‐Y decomposition to define position at the shoulder joint (or more specifically, the “thoraco‐humeral” joint).9

figure sm36145.f1
Figure 1 Customised UWA upper‐body marker set for wheelchair tennis players (anterior view). Informed consent was obtained for publicaiton of this figure.

The following phases, reflecting meaningful temporal or kinematic characteristics of the serve, were normalised through customised Matlab software (Mathworks, Natick, Massachusetts, USA): cocking (from the racquet's highest point or peak vertical displacement in the backswing, RHP, to maximum external rotation of the racquet arm, MER), swing (from RHP to 0.004 s prior to racquet–ball impact, IMP), forwardswing (from MER to IMP) and follow‐through (from 0.004 s post‐racquet–ball impact to 0.124 s post‐impact). With the small sample size, comparative descriptive analyses of the mean WFS and WKS kinematics and kinetics related to or considered to represent shoulder joint loading were undertaken. Of note is that for complete interpretation of net joint force data, and more particularly compression‐distraction forces, the assessment of muscle activity is required. Comparison of these data with those reported in Reid et al3 to characterise high‐performance able‐bodied players was also pursued.


Effect of wheelchair serve type on 3D racquet velocity

Comparable mean (SD) maximum absolute racquet velocities of ~31.95 (1.0) m/s were generated across service types. However, higher pre‐impact horizontal racquet velocities characterised the WFS, while the lateral racquet velocities for the WKS were effectively double those generated in the WFS immediately pre‐impact (table 11).). Maximum pre‐impact vertical racquet velocities trended higher in the WKS. Observable differences mark the racquet velocities produced in the wheelchair serves as compared with the able‐bodied serves.

Table thumbnail
Table 1 Comparison of mean linear racquet kinematics between the WFS (n = 3) and WKS (n = 3) as performed by two subjects (S1 and S2), as well as in contrast to the mean able‐bodied flat (FS) and second ...

Upper‐extremity kinematics that describe WFS and WKS

Wheelchair players rotated their upper arms to similar positions of MER irrespective of wheelchair serve type, but through ~20° less external rotation than in the able‐bodied serve (table 22).). At MER, the two wheelchair players recorded plane of elevation angles from 145° to 165° (where, functionally, 180° would see the upper arm horizontally adducted, parallel to the trunk; with lesser values indicative of upper arm position in front of this straight line; fig 22).

figure sm36145.f2
Figure 2 2D functional representation of (A) shoulder alignment lateral flexion, (B) shoulder alignment forward flexion, (C) upper arm–thorax elevation and (D) plane of elevation.
Table thumbnail
Table 2 Mean shoulder joint kinematics for three trials, related to shoulder joint loading, that characterise the WFS and WKS and able‐bodied FS and KS (n = 12)3

The elevation of the upper arm with respect to the thorax at impact was consistent between wheelchair players and independent of wheelchair serve type. During the forwardswing to impact, higher peak upper arm internal rotation velocities were developed by Subject 2, while a similar trend characterised the performance of the WKS, as compared with the WFS.

Notable differences also punctuated the wheelchair players' pre‐impact shoulder alignment and by extension, trunk motion (table 33).). Compared with able‐bodied players, wheelchair players employed less forward trunk flexion, similar amounts of lateral trunk flexion and more variable trunk rotation during the service forwardswing.

Table thumbnail
Table 3 Mean 3D motion of shoulder alignment during the WFS and WKS (n = 3), as well as in comparison with the able‐bodied FS and KS (n = 12)3

Shoulder joint kinetics that characterise the WFS and WKS

As in able‐bodied serves, homogeneous and negligible peak external rotation moments assisted external rotation of the upper arm during the “cocking” phase of the WFS and WKS. Subsequent peak pre‐impact upper arm internal rotation moments were also similar across serves, albeit somewhat variable between subjects. A comparable pattern epitomised the peak external rotation moments of the follow‐through (table 44).

Table thumbnail
Table 4 Mean shoulder joint kinetics that punctuate WFS and WKS performance (n = 3), and values for able‐bodied flat (FS) and second kicker (KS) serves (n = 12)3

Peak compressive forces and the rates at which they were developed prior to impact were similar for the WFS and WKS, but different between subjects. The respective mean compressive and distractive forces that prevailed during the service follow‐throughs of Subject 1 and Subject 2 provide further evidence of some inter‐player mechanical variation.

In absolute terms, wheelchair players harboured less load at the shoulder joint during the serve than able‐bodied players (table 44).). However, when shoulder joint kinetics are expressed relative to maximum pre‐impact absolute racquet velocities:

equation image

participant 2 was subject to loads more commensurate to those generated during the able‐bodied serve (table 55).

Table thumbnail
Table 5 Mean shoulder joint kinetics of the wheelchair and able‐bodied serves (FS and KS) expressed relative to maximum pre‐impact absolute racquet velocity (n = 12)3


Differential 3D racquet velocity in the WFS and WKS

Consistent with expectations, the WFSs were characterised by higher peak pre‐impact horizontal racquet velocities (~29 m/s) than the WKSs (~26 m/s). Further, higher peak right lateral velocities were developed during the forwardswings of the WKSs (~9 m/s) as compared with the WFSs (~4 m/s). These results agree with the dichotomous pre‐impact racquet velocity profiles of different able‐bodied serves.3,10 Interestingly, wheelchair players appear to generate comparable peak pre‐impact absolute racquet velocities to accommodate the varying tactical objectives of these serves. This finding contrasts with the higher absolute racquet velocities generated by high‐performance able‐bodied players in the flat serve (FS) as compared with the kick serve (KS), but concurs with the results of an earlier investigation by Chow et al.10

With both wheelchair players lacking the dynamic leg actions of able‐bodied players and possessing variable but restricted trunk motion, the comparatively lower peak pre‐impact absolute and horizontal racquet velocities of the WFS and WKS were anticipated. In fact, as inferred by Brody,11 due to wheelchair players' low hitting heights, its largely impractical (ie, in terms of serve percentage), if not impossible, for them to generate the same horizontal racquet velocities as their able‐bodied counterparts (ie, 33 m/s;10) and still have their serves land successfully (and consistently) in the court. Nonetheless, the ~33% reduction in the maximum pre‐impact absolute and horizontal racquet velocities developed in the wheelchair serve compares favourably with the 30–50% decreases observed in throwing velocity when throwers experience constrained hip and trunk motion.12,13

Noteworthy is also that contrasting pre‐impact racquet velocities were developed by the individual wheelchair players. That is, Subject 2 generated higher 3D racquet velocities, especially laterally, in the WFS and WKS. Certainly as both players suffered from injuries of similar vertebral level but different severity, some variation in racquet kinematics could be expected. To elaborate, Subject 2 suggested that his comparatively superior leg use enabled him to gain some “push” against the chair to “drive upward” when serving, and more importantly provided a more stable platform for subsequent high‐speed segment coordination (B Weekes, personal communication, 2006). Further, the incomplete nature of Subject 2's injury would typically imply that this subject possessed superior trunk strength as compared with Subject 1. It is probable that these neurophysiological factors account for some of the variation observed in the racquet velocities generated by these players.

Upper‐extremity joint kinematics of the WFS and WKS

Several authors have previously linked the magnitude of humeral MER to the forcefulness of the able‐bodied player's leg drive.14,15 While this assertion appears overly simplistic,3 the hypothesis that wheelchair players would achieve less externally rotated positions of the upper arm than able‐bodied players was supported. That is, without the utility of a leg drive, the upper arms of the two wheelchair players reached ~20° less MER than in the able‐bodied serve. Significantly, this computed external rotation angle is actually a combination of trunk hyperextension, scapulothoracic motion and true glenohumeral rotation, so the disparity between the upper arm MER achieved in the wheelchair and able‐bodied serves might also reflect some variation in trunk hyperextension and scapulothoracic motion; the extent to which is currently unquantifiable using external markers. Further, with the placement of the humeral external markers (triad) near the soft tissue of the upper arm some caution is likely required in interpreting the reported absolute magnitudes of longitudinal upper arm rotation.3

The absolute peak angular velocities of humeral internal rotation during the forwardswing of the wheelchair serves were also lower (~40–80%) than in the able‐bodied FS and KS. Intuitively, a link can be made between the able‐bodied serves' comparatively larger magnitudes of upper arm MER and their augmented peak internal rotation angular velocities. At MER, marginally higher upper arm plane of elevation angles were observed in the serves of Subject 2 as compared with Subject 1. Importantly, both subjects recorded mean plane of elevation angles <180° so that their upper arms remained in front of their shoulder alignments. Similar humeral displacements characterised the high‐performance able‐bodied FS and KS3 as well as the serves of professional ATP players.16 Consequently, the threat of hyperangulation, which can contribute to secondary impingement problems, appears similarly negligible among wheelchair players.

At impact, wheelchair players assume upper arm–thorax elevation angles that are ~15° higher than those observed during the high‐performance and professional able‐bodied tennis serves.3,16 Based on previous research, these more obtuse upper arm–thorax elevation angles could be indicative of injurious shoulder joint kinetics. However, it's likely that the angle that permits maximum wheelchair serve speed with minimal shoulder joint loading is different to the 100°±10° reported by Matsuo et al17 in baseball pitching.

Clear variation pervaded the 3D rotation of shoulder alignment, and by extension the trunk, in the WFS and WKS of the two wheelchair players. During the forwardswing to impact, Subject 2 forward flexed (~25°) and rotated (~50°) his shoulder alignment more than Subject 1 (~11°, ~8° respectively) irrespective of serve type. Further, where both players left laterally flexed their trunks by ~17° during this same phase, Subject 2 impacted the ball in a more left laterally flexed position (S1: ~25°; S2: ~42°). As aforementioned, the contrasting injury pathologies of the two players obviously influenced the extent to which each player was able to actively engage the musculature of his trunk. Simply put, the two wheelchair players explained Subject 2's service motion to more closely resemble that of an able‐bodied player; albeit with the involvement of less trunk flexion. Additionally, as all players went through similar ranges of pre‐impact left lateral flexion and impacted the ball at comparably oblique alignments, the above‐mentioned more obtuse upper arm–thorax elevation angles of wheelchair players at impact are likely related to more elevated positions of the upper arm rather than the differential alignment of the thorax.

Effect of serve type on shoulder joint loading in wheelchair players: implications for injury and performance

Cocking phase

Wheelchair players, whom possess no leg drive, generated negligible peak external rotation moments during the cocking phase of the WFS (0.5–2.2 Nm) and WKS (0.4–1.2 Nm). Similarly insignificant peak moments were reported to assist external rotation of the upper arm in the able‐bodied FS and KS (mean (SD) ~3 (1) Nm). Collectively, these data offer little support to the proposed negative association between leg drive and the role of a player's external rotator musculature in achieving humeral MER during the serve.18 Indeed, it would appear that even wheelchair players produce MER of the upper arm as a consequence of the inertial lag of the forearm, hand and racquet.

Forwardswing phase

Several researchers have demonstrated the ill‐effects of abbreviated or restricted segment coordination on throwing performance.12,13,19,20 For example, the inability to rotate hips and/or generate ground reaction forces has been shown to negatively affect throwing velocity, while McMaster et al19 associated the lack of a base of support in the water polo throw to heightened shoulder joint forces. Correspondingly, as wheelchair tennis players gain virtually no lower limb drive and possess limited trunk function, a relative increase in shoulder joint kinetics was anticipated.

Surprisingly, however, there was little evidence of any compensatory kinetic response. When expressed relative to maximum pre‐impact absolute racquet velocity, similar peak internal rotation moments (~55%) prevailed regardless of serve type or disability, while mean pre‐impact compressive forces were actually lower in the wheelchair serves (S1: 160–200%, S2: 390–400%; able‐bodied FS and KS: ~525%). As compared to the able‐bodied serves (~780%), the relative average rates of pre‐impact maximum compressive force loading were similar for the WFS and WKS of Subject 2 (~820%) but lower for the WFS and WKS of Subject 1 (~265%). These observed differences in the pre‐impact compressive force profile of the two wheelchair players could be related to slight variation in their respective impact heights. That is, where Subject 2 impacted all serves at 1.8× his sitting height, Subject 1 made racquet–ball contact in a less extended position (1.71× his sitting height). In application, these contrastive sitting‐impact heights suggest that Subject 2 possessed a more “up and out” service action,15 which might have necessitated the production of higher compressive forces to maintain the humeral head centred in the gleniod.

Follow‐through phase

No distinctive loading profile characterised the follow‐through of either wheelchair serve. Deceleration of the internally rotating upper arm was facilitated by similar peak WFS and WKS post‐impact external rotation moments. Comparable mean compressive forces were also generated following ball impact in both subjects' WFSs and WKSs. Of note however, was that some variation marked the mean forces applied along the humerus throughout the follow‐throughs of both players' serves. That is, where Subject 2 generated mean (SD) compressive forces (WFS: 50 (9) N; WKS: 39 (7) N) to help resist post‐impact humeral distraction, Subject 1 applied mean distraction forces (WFS: −21 (11) N; WKS: −36 (1) N) to maintain the upper arm in equilibrium. This dichotomous kinetic patterning relates to stylistic differences in the players' follow‐throughs. As is familiar to the able‐bodied serve, Subject 2 followed through “across his body”. Subject 1, by contrast, finished “away from the chair”. Both types of follow‐throughs are commonly used by elite wheelchair players, and ITF Wheelchair Development Officer, Mark Bullock, suggests that while players with more complete spinal cord injuries might find it difficult to follow‐through “across their bodies”, certain coaches prefer players to finish “away” irrespective of disability (M. Bullock, personal communication, 2006). The fact that Subject 1 suffered from a more complete injury than Subject 2, and indicated that he had difficulty in maintaining the necessary balance to comfortably follow‐through “across his body” supports Bullock's affirmation.

What is already known on this topic

  • Wheelchair players, like their able‐bodied counterparts, commonly suffer from shoulder overuse injuries.
  • A strong empirical link between shoulder pain and wheelchair serve performance exists, and some evidence points to the development of racquet velocity increasing along with shoulder joint loading.

What this study adds

  • World‐ranked wheelchair players can generate similar pre‐impact absolute racquet velocities in the wheelchair flat (WFS) and kick (WKS) serves.
  • Higher pre‐impact horizontal and lateral racquet velocities punctuate the WFS and WKS, respectively.
  • Associated shoulder joint kinetics vary between individual players, but appear to remain consistent across wheelchair serve type.

By following through “across his body”, Subject 2 produced mean compressive forces like those previously reported to assist upper arm deceleration in the serve and baseball pitch. In contrast, the “away” finish saw Subject 1 undergo limited 3D trunk rotation and maintain more obtuse shoulder joint elevation and plane of elevation angles such that mean distraction forces were needed to support the mass of the humerus. As injury to the rotator cuff is often associated with its role in generating the compressive forces needed to resist humeral distraction during the deceleration phase of overhand sports skills,21,22 these data could be interpreted to suggest that Subject 1's post‐impact kinetics are less likely to elicit cuff injury. However, upon retrospective inspection, Subject 1 was revealed to apply higher absolute shoulder joint posterior forces (WFS: ~60 N; WKS: ~70 N) to help resist superior translation of the humeral head than Subject 2 (WFS: ~35 N; WKS: ~55 N). Collectively, these results point to both follow‐throughs subjecting wheelchair players' shoulders to high but different loads that appear equally injurious.

Finally, relative post‐impact peak external rotation moments (190–200%) and mean compressive forces (~40%) in the able‐bodied serves remained appreciably higher than the corresponding measures of load in the wheelchair serves (70–120%; ~30%). So, despite lacking the same lower body and trunk rotation, wheelchair players seem no more likely to develop high shoulder joint loading conditions to help decelerate the upper arm and racquet, than able‐bodied players.


Wheelchair players generate similar pre‐impact absolute racquet velocities in the WFS and WKS. Higher pre‐impact horizontal and lateral racquet velocities, however, are produced in the WFS and WKS respectively. The shoulder joint kinetics that contribute to these differential 3D velocity profiles vary between individual players (and level and severity of spinal cord injury) but appear to remain consistent across wheelchair serve type. In comparison to the high‐performance able‐bodied serve, elite wheelchair players are subject to similar or reduced relative pre‐ and post‐impact shoulder joint loading conditions when hitting both types of wheelchair serve. This defies any suggestion that greater shoulder joint loads characterise wheelchair stroke production such that this playing populace is predisposed to elevated shoulder joint injury risk, while also indicating that joint loading is likely to be positively related to the development of racquet velocity.


The authors would like to thank the International Tennis Federation for grant received to complete this project.


WFS - wheelchair flat serve

WKS - wheelchair kick serve


Competing interests: None.

Informed consent was obtained for publication of figure 11.


1. Ferrara M S, Davis R W. Injuries to elite wheelchair athletes. Paraplegia 1990. 28335–341.341 [PubMed]
2. Burnham R S, May L, Nelson E. et al Shoulder pain in wheelchair athletes. The role of muscle imbalance. Am J Sports Med 1993. 21238–242.242 [PubMed]
3. Reid M, Elliott B, Alderson J. Shoulder joint loading in the high‐performance flat and kick serves. Br J Sp Med. Published Online First: 18 May 2007. doi:10.1136/bjsm.2007.036657
4. Brody H, Cross R, Lindsey C. The physics and technology of tennis. Solana Beach, California, USA: Racquet Tech Publishing, 2002
5. Clauser C E, McConville J T, Young J W. Weight, volume and centre of mass segments of the human body. In: AMRL Technical Report 69–70. Dayton, Ohio: Wright‐Patterson Air Force Base, 1969
6. de Leva P. Adjustments to Zatsiorsky‐Seluyanov's segment inertia parameters. J Biomech 1996. 291223–1230.1230 [PubMed]
7. Cheng P L. A spherical rotation coordinate system for the description of three‐dimensional joint rotations. Ann Biomed Eng 2000. 281381–1392.1392 [PubMed]
8. Doorenbosch C A, Harlaar J, Veeger D H. The globe system: an unambiguous description of shoulder positions in daily life movements. J Rehabil Res Dev 2003. 40147–155.155 [PubMed]
9. Wu G, van der Helm F C, Veeger H E. et al ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion‐Part II: shoulder, elbow, wrist and hand. J Biomech 2005. 38981–992.992 [PubMed]
10. Chow J W, Carlton L G, Lim Y T. et al Comparing the pre‐ and post‐impact ball and racquet kinematics of elite tennis players' first and second serves: A preliminary study. J Sports Sci 2003. 21529–537.537 [PubMed]
11. Brody H. Tennis science for tennis players. Philadelphia, Pennsylvania, USA: University of Pennsylvania Press, 1987
12. Alexander R M. Optimum timing of muscle activation for simple models of throwing. J Theor Biol 1991. 150349–372.372 [PubMed]
13. Toyoshima S, Hoshikawa T, Miyashita M. et alContribution of body parts to throwing performance. In: Biomechanics V. Baltimore, Maryland, USA: University Park Press, 1974. 169–174.174
14. Bartlett R M, Piller J, Miller S. A three dimensional analysis of the tennis serves of National (British) and County standard male players. In: Reilly T, Hughes M, Lees A. eds. Science and racket sports. London, UK: E & FN Spon, 1994. 98–102.102
15. Elliott B, Marsh T, Blanksby B. A three‐dimensional cinematographic analysis of the tennis serve. Int J Sport Biomech 1986. 2260–271.271
16. Fleisig G, Nicholls R, Elliott B. et al Kinematics used by world class tennis players to produce high‐velocity serves. Sports Biomech 2003. 251–64.64 [PubMed]
17. Matsuo T, Matsumoto T, Mochizuki Y. et al Optimal shoulder abduction angles during baseball pitching from maximal wrist velocity and minimal kinetics viewpoints. J Appl Biomech 2002. 18306–320.320
18. Elliott B, Fleisig G, Nicholls R. et al Technique effects on upper limb loading in the tennis serve. J Sci Med Sport 2003. 676–87.87 [PubMed]
19. McMaster W C, Long S C, Caiozzo V. Isokinetic torque imbalances in the rotator cuff of the elite water polo player. Am J Sports Med 1991. 1972–75.75 [PubMed]
20. Whiting W C, Puffer J C, Finerman G A. et al Three‐dimensional cinematographic analysis of water polo throwing in elite performers. Am J Sports Med 1985. 1395–98.98 [PubMed]
21. Fleisig G S, Barrentine S W, Escamilla R F. et al Biomechanics of overhand throwing with implications for injuries. Sports Med 1996. 21421–437.437 [PubMed]
22. Jobe F W, Moynes D R, Tibone J W. et al An EMG analysis of the shoulder in pitching: a secondary report. Am J Sports Med 1984. 12218–220.220 [PubMed]

Articles from British Journal of Sports Medicine are provided here courtesy of BMJ Group