Nicolas Vignais* Virtual Thrower Versus Real M2S Laboratory, UFR APS, Rennes 2-ENS Cachan University, Goalkeeper: The Influence 35044 Rennes, France of Different Visual Conditions

Richard Kulpa on Performance M2S Laboratory, UFR APS, Rennes 2-ENS Cachan University, 35044 Rennes, France and IRISA BUNRAKU Project, Abstract Campus de Beaulieu, 35042 Rennes, France In order to use virtual reality as a sport analysis tool, we need to be sure that an immersed athlete reacts realistically in a virtual environment. This has been validated Cathy Craig for a real handball goalkeeper facing a virtual thrower. However, we currently ignore School of Psychology, Queen’s which visual variables induce a realistic motor behavior of the immersed handball goal- University Belfast, Northern Ireland keeper. In this study, we used virtual reality to dissociate the visual information related to the movements of the player from the visual information related to the trajectory of Benoit Bideau the ball. Thus, the aim is to evaluate the relative influence of these different visual infor- M2S Laboratory, UFR APS, mation sources on the goalkeeper’s motor behavior. We tested 10 handball goalkeep- Rennes 2-ENS Cachan University, ers who had to predict the final position of the virtual ball in the goal when facing the 35044 Rennes, France following: only the throwing action of the attacking player (TA condition), only the resulting ball trajectory (BA condition), and both the throwing action of the attacking player and the resulting ball trajectory (TB condition). Here we show that performance was better in the BA and TB conditions, but contrary to expectations, performance was substantially worse in the TA condition. A significant effect of ball landing zone does, however, suggest that the relative importance between visual information from the player and the ball depends on the targeted zone in the goal. In some cases, body- based cues embedded in the throwing actions may have a minor influence on the ball trajectory and vice versa. Kinematics analysis was then combined with these results to determine why such differences occur depending on the ball landing zone and conse- quently how it can clarify the role of different sources of visual information on the motor behavior of an athlete immersed in a virtual environment.

1 Introduction

Virtual reality is now being used as a tool to analyze and understand motor behavior in sport (Katz et al., 2006; Bideau et al., 2010). This promising technology has a number of advantages over video presentation or real-game situations. Firstly, all factors can be controlled and manipulated in a systematic manner, ensuring reproducibility between trials (Tarr & Warren, 2002). Sec- ondly, the effects of these modifications on the resulting behavior can be moni- tored in real time. Thirdly, the immersion of the subject in the virtual scene in an egocentric position allows the optical information gleaned from the virtual world to correspond directly to what the participant would see in a real sporting

Presence, Vol. 19, No. 4, August 2010, 281–290 ª 2010 by the Massachusetts Institute of Technology *Correspondence to [email protected].

Vignais et al. 281

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 282 PRESENCE: VOLUME 19, NUMBER 4

situation (Cutting, 1997). Finally, the participant’s view- nent’s next move (Abernethy, Wood, & Parks, 1999). point is stereoscopic which has previously been high- While some studies have suggested that successful antici- lighted as an important factor when performing inter- pation is determined by the perception of spatial-tempo- ceptive tasks (Mazyn, Lenoir, Montagne, & Savelsbergh, ral information in the trajectory of the ball (Bastin, Craig, 2004). Given these advantages, this technology has been & Montagne, 2006; Craig et al., 2006; Lenoir, Musch, exploited to analyze and understand players’ perform- Thiery, & Savelsbergh, 2002; McLeod, Reed, & Dienes, ance in different sports such as soccer (Craig, Berton, 2006), others have shown that successful anticipatory Rao, Fernandez, & Bootsma, 2006), tennis (Molet behavior depends on the accurate perception of body- et al., 1999), handball (Bideau et al., 2004; Vignais based cues (Jackson, Warren, & Abernethy, 2006; Pan- et al., 2009), and rugby (Brault, Bideau, Kulpa, & Craig, chuk & Vickers, 2006; Savelsbergh, Williams, Van der 2009). Moreover this technology can be employed to Kamp, & Ward, 2002; Williams, Davids, Burwitz, & Wil- test team sport strategies (Metoyer & Hodgins, 2000) liams, 1994). The general consensus suggests that when or to train athletes in a simulator (Kelly & Hubbard, the temporal constraints are too tight (e.g., a penalty kick 2000). With respect to this latter option, Kelly and Hub- in soccer), players do not have enough time to act on ball bard have successfully designed a complete bobsled sim- trajectory information and consequently use body-based ulator which displays a graphical representation from the cues (Savelsbergh, Van der Kamp, Williams, & Ward, driver’s viewpoint on a cockpit-mounted monitor. 2005). In spite of this conclusion, the relative contribu- In spite of these advantages, using virtual reality as a tion of body-based versus ball flight information in per- sports analysis tool does raise new questions. One of the formance has not been properly tested. most important is the motor response of the athlete Here, we present body-based cues and ball trajectory immersed in a virtual environment. In fact, the motor information in isolation in order to better assess the rela- behavior of a subject can be considered as one of the tive importance of both types of visual information for physical measures of presence (Slater, 2009; Slater, performance. Using immersive interactive virtual reality, Khanna, Mortensen, & Yu, 2009). In sport, this concept this study allows dissociation between body-based cues has been used to validate a virtual sport situation. and ball flight information. The relative importance of Indeed, Bideau et al. (2003) compared the movement of these two visual elements in performance will be assessed handball goalkeepers during the interception of the same in order to understand why an athlete tends to respond throws made by real and virtual opponents. The results realistically to a virtual sport situation. To this end, we showed that the goalkeepers’ gestures in both environ- investigate the effects of different visual conditions of a ments were similar. Although we know that the virtual virtual throwing action on the performance of immersed sport situation used in this case study induced handball handball goalkeepers. goalkeepers to respond realistically, we do not yet know why they did. Are some visual elements displayed in the 2 Overview virtual scene more important than others to enable handball goalkeepers to react in a realistic way? In order to evaluate the influence of the different In most ball sports, successful anticipation of what an visual conditions of a throwing action on the handball opponent is going to do next can give a substantial com- goalkeepers’ performance, we developed a three-step petitive edge. The temporal constraints within which a process (see Figure 1). player must act are often very tight, and so early recogni- Step 1. First of all, to realistically animate the tion of an opponent’s actions or anticipation of where synthetic thrower and the ball in the virtual the ball is going can increase the player’s chances of environment, we captured real throwing making a successful counteracting move. Much of the actions. This led to the creation of a large research to date has focused on identifying the sources of motion database with throws to various visual information that players use to anticipate an oppo- places in the goal.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 Vignais et al. 283

Figure 1. The three-step process includes data acquisition, real time experiment, and performance comparison.

Step 2. Secondly, these captured throws were used segment, with six markers being reserved for the ball. by the MKM animation engine (Kulpa, Each player was asked to throw the ball 12 m from the Multon, & Arnaldi, 2005; Multon, Kulpa, goal, aiming for different prespecified target zones & Bideau, 2008) in order to realistically within the goal (no goalkeeper was present). For each animate the virtual character. The rendering of throw, the participant and ball movement were captured these animations was done with three different at 200 Hz using 12 infrared cameras. Nine trajectories visual conditions. that arrived close to the center of each zone were Step 3. Finally, the movements of the goalkeepers selected and used in the subsequent virtual animation were captured in real time. These data were part of the study (see Figure 2). then analyzed to determine which responses Kinematic throwing data, captured using the Vicon were successful. system, were then incorporated into the MKM (Manage- able Kinematic Motion) animation engine (Kulpa et al., 3 Motion Capture and Animation Process 2005; Multon et al., 2008) to animate the virtual hand- ball player in real time. Based on these captured motions, In order to animate the virtual character and the MKM can automatically synchronize, blend, and adapt ball, real handball throwing actions were recorded. actions to different morphologies and to external con- A Vicon motion capture system (Oxford Metrics, straints such as foot contacts. Virtual ball trajectories for Oxford, UK) was used to record movement kinematics the nine zones were calculated using another software from three top-level handball players. All the throwers module that incorporates data obtained from the motion were right-handed and had at least 10 years experience capture part of the study. Ball velocities were similar for playing handball in the top French league at the time of all nine trajectories (20 6 0.2 ms1), making ball flight the study. Each player was equipped with 36 reflective time very nearly the same for all nine zones. markers placed on anatomical landmarks to precisely A realistic handball stadium was created using 3dsmax reconstruct the 3D position and orientation of each limb (Autodesk Inc., San Rafael, CA). The size of the virtual

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 284 PRESENCE: VOLUME 19, NUMBER 4

Figure 2. A schematic view of the handball goal. It has been divided into nine zones but only zones 1, 3, 4, and 6 were used in the current analysis. Figure 3. A goalkeeper immersed in the virtual environment.

stadium was scaled so that one unit in the virtual envi- ronment corresponded to 1 m in reality. The virtual onto a large cylindrical screen (3.80 m radius, 2.38 m court was 20 m wide and 40 m long with a clearly height, and 1358 field of vision). A set of glasses marked semicircular area (6 m radius) corresponding to synchronized with the system enabled stereovision (60 the goal zone. Hz). The Vicon motion capture system (12 cameras) was used to record the goalkeeper’s movements and was 4 Evaluation of Goalkeepers’ coupled to the virtual reality display. As the two systems Performance were linked and the goalkeeper’s head was tracked, it was possible to change the goalkeeper’s perspective in After animating the handball throwing actions, the virtual world in real time (delay < 20 ms). 10 elite male handball goalkeepers (playing in the top Three experimental conditions representing the differ- league and at the national level) were placed in the vir- ent types of visual information were randomly presented tual environment (see Figure 3). to the participants. The ball only condition (BA) The mean age of the participants was 22.6 years involved the goalkeeper judging the arrival position of (SD ¼ 4.9 years), mean height was 1.81 m (SD ¼ the ball while seeing only the trajectory of the ball from 0.09 m) and mean weight was 78.2 kg (SD ¼ 13.1 kg). the point of release (12 m) to the cut-off point (6 m; see All participants had normal or corrected-to-normal Video1.avi); the thrower only (TA) condition involved vision. All participants gave their informed consent the goalkeeper making an action when presented with before the study began. only the complete throwing action of the thrower (pre- In order to enhance the behavior of subjects, a real release to release point; see Video2.avi); and the thrower goal (3 m 2 m) was placed where it was virtually repre- and ball condition (TB) involved the goalkeeper making sented in the computer generated environment. Goal- a judgment when presented with both the throwing keepers were encouraged to touch the real goalposts to action and the resulting ball trajectory (see Video3.avi). enhance the correspondence between virtual and real These three experimental conditions allowed us to sys- distances. Three synchronized video projectors, Barco tematically isolate the two visual information sources a 1208S (Barco, Courtrai, Belgium) driven by a SGI 83 goalkeeper might use to guide his actions. Onyx2 Infinite Reality (Silicon Graphics, Sunnyvale, Two response modes were then performed: a motor CA) were used to project the 3D sports hall environment task condition and a judgment task condition.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 Vignais et al. 285

Figure 4. A schematic representation of the location of the optoelectronic markers (a) in the judgment task condition and (b) in the motor task condition. These different positions permitted us to reconstruct the segments and to compute the radial error.

Judgment Task Condition. In this experiment, the three visual information conditions (three BA the goalkeepers were equipped with only 11 reflec- throws, three TA throws, and three TB throws). Trials tive markers: four on their head for head tracking, from the training period were not included in the subse- three on their right hand, and four on their left hand quent analysis. to record final hand positions. The markers were A total of four different trajectories were randomly placed on the hands in such a way that the central presented for the different visual conditions. They corre- position of the hand could be easily calculated, with sponded to zones 1, 3, 4, and 6 (see Figure 2). The four the additional marker allowing the left hand and different ball trajectories were randomly repeated five right hand to be distinguished during analysis (see times for each condition giving a total of 20 throws per Figure 4[a]). After the end of the throwing action, condition. Ten other throws to different target locations goalkeepers were asked to then move their hand as in the goal were randomly included so as to create more quickly as possible to the position in the goal where variability in end arrival position for the goalkeeper. In they thought the ball would have ended up. fact, a total of 140 trials ([(5 repetitions per zone 4 Motor Task Condition. This experiment zones aimed 3 visual conditions) þ 10 other throws] required the goalkeepers to get their limb to the right 2 response modes) were presented to each goalkeeper. place at the right time to stop the virtual ball. The A short break was given after each block of 30 trials. goalkeepers reacted to the same randomized throws In order to estimate and compare the goalkeeper’s displayed in the judgment condition. The goalkeep- perceptual and motor responses, we performed two ers were asked to react as if they were in a real match kinds of data analysis. situation. Each goalkeeper was equipped with 36 optoelectronic markers placed on the same anatomi- Judgment Task Condition. In this condition, the cal landmarks as in the motion capture part of the hand and the ball were both represented by spheres real handball throwing actions (see Figure 4[b]). (see Figure 4[a]). This permitted us to calculate the Before the experiment began, each participant was radial error which corresponds to the difference given a training period to become familiar with the envi- between the final position of the hand and the ronment and the task. During this time, the participants virtual ball’s arrival position in the goal opening. randomly viewed a sample of three throws from each of The percentage of successful judgments was

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 286 PRESENCE: VOLUME 19, NUMBER 4

Figure 5. Mean percentage of correct responses for all goalkeepers (a) during the judgment task, (b) for each zone aimed, and (c) during the motor task (d) for each zone aimed (***p < .001).

computed on the basis of the zero radial error which centage of successful responses. During the judgment equated to a correct response. task, a prediction was considered successful if the final Motor Task Condition. In this condition, we position of the hand, represented by a sphere, was in represented the goalkeeper’s limbs as cylinders contact with the final position of the ball sphere (see Fig- (trunk, arms, forearms, thighs, calves, and feet) and ure 4[a]). Concerning the motor task, an interception spheres (head and hands) from joint center positions was considered successful if the representation of one of (see Figure 4[b]). This full body representation the goalkeeper’s limbs was in contact with the ball sphere enabled us to determine whether there was a colli- (see Figure 4[b]). sion between the virtual ball and the goalkeeper in As expected, the participants were most successful real time (a visual feedback was displayed after each when they were presented with TB information during throw). This provided us the percentage of success- the judgment task, 21.3 6 9.7%, and during the motor ful interceptions. task, 31.2 6 5.9% (see Figures 5[a] and 5[c]). The par- ticipants were only marginally less successful for the BA 5 Results condition, judgment: 21.1 6 7.3%, motor: 29.1 6 6.7%, but performance was statistically worse with the TA In the first part of the analysis, we focused on the information, judgment: 8.6 6 7.9%, motor: 14.5 6 influence of the different visual conditions on the partici- 8.6%, after a two-way repeated measures analysis of pants’ effectiveness. To this aim, we analyzed the per- variance, judgment: F(2, 54) ¼ 46.082, p < .001,

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 Vignais et al. 287

Figure 6. Mean radial error and its lateral and vertical components (a) for all goalkeepers during the judgment task (b) for each zone aimed, and (c) during the motor task (d) for each zone aimed (*p < .05; **p < .01; ***p < .001). Lateral and vertical components are not represented in the goal (b and d).

motor: F(2, 54) ¼ 100.226, p < .001. This marked 6 Discussion reduction in performance for the TA condition was seen in all zones except zone 1 (see Figures 5[b] and 5[d]). The aim of this study was to see how different The second part of the analysis focuses on the influ- sources of visual information influence the goalkeepers’ ence of the different visual conditions on the goalkeep- motor behavior in a virtual environment. By separating ers’ accuracy. With this aim, the radial error was calcu- the presentation of body-based movement cues and lated and examined (see Figure 6). ball trajectory information, we were able to look at the A repeated measures analysis of variance revealed a relative contribution of each of them individually and significant difference for the radial error between the together. The results show that body-based cues alone different visual conditions during the judgment task, do not provide sufficient information for expert goal- F(2, 54) ¼ 803.501, p < .001, and during the motor keepers to accurately predict where the ball is going in task, F(2, 54) ¼ 257.762, p < .001. A Tukey HSD post the goal. Indeed, regarding the final limb position hoc analysis shows that the radial error obtained for the results, we obtained far greater errors for the TA TA condition, judgment: 61.3 6 5.3 cm, motor: 47.8 6 condition than for the BA and TB conditions. Although 7.3 cm, was significantly greater than the BA condition, the BA condition was significantly more successful than judgment: 31.8 6 1.4 cm, motor: 18.2 6 2.9 cm, and the TA condition, it does not seem that seeing the the TB condition, judgment: 26.2 6 1.8 cm, motor: throwing action with the ball trajectory (TB condition) 13.6 6 4.4 cm. improved performance compared to the BA condition,

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 288 PRESENCE: VOLUME 19, NUMBER 4

Figure 7. (a) Segment configuration of the throwing action at ball release in frontal plane for each zone aimed and (b) histogram of related joint angles. The different joints represented are: left shoulder (LS), right shoulder (RS), right elbow (RE), and right wrist (RW).

except for the goalkeepers’ accuracy during the judg- shoulder. Although biomechanically these differences in ment task. limb segment orientation partly determine where the However, we observed a decrease in successful ball will land in the goal, they are probably too subtle for responses in all zones, except zone 1, for the TA condi- the goalkeeper to use them as reliable visual information, tion. This finding could be explained by the biomechani- except for throws in zone 1. cal analysis of the throwing actions (see Figure 7). These findings are very interesting when considered in As it can be seen from Figure 7, the patterns of final light of other research in this area. Previous studies have limb positions are very similar for throwing actions in attempted to show that expertise in goalkeeping resides zones 1 and 3. Segmental positions and joint angle val- in the ability to pick up pertinent information from the ues of both throwing actions seem very close. It is possi- movement kinematics of the attacking player (Savels- ble that participants have confused the throwing actions bergh, Williams, Van der Kamp, & Ward, 2002; Savels- for zone 3 with zone 1 (see Figure 7[a]). Moreover, it is bergh et al., 2005; Franks & Hanvey, 1997; Williams possible that goalkeepers were able to identify kinematic et al., 1994). In a virtual environment, it seems that cues of the throwing action for zone 1. Given these simi- handball goalkeepers are able to use body-based cues larities in the final limb position, the last part of the anal- alone for specific throwing actions. Overall, it would ysis aims to see how the throwing actions differed. In appear that body-based information on its own is not other words, what biomechanical parameters changed sufficient to make accurate judgments about (and conse- most significantly from one throw to another? The analy- quently movements toward) where the ball will end up sis of body segment angles showed greatest differences in the handball goal. Information picked up from the in the elbow angle (see Figure 7[b]). Differences, ball trajectory is necessary to refine judgment and though small, were also found for dipping of the motor tasks. This kind of visual information helps the shoulder and elevation of the elbow respective to the handball goalkeeper react in a realistic way in the virtual

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 Vignais et al. 289

environment. A recent study by Jackson and Mogan Bastin, J., Craig, C., & Montagne, G. (2006). Prospective (2007) also revealed the importance of ball flight infor- strategies underlie the control of interceptive actions. mation in making anticipatory judgments in tennis. Human Movement Science, 25, 718–732. Using a spatial occlusion technique to selectively omit Bideau, B., Kulpa, R., Menardais, S., Fradet, L., Multon, F., key components of the tennis serve, they found the most Delamarche, P., et al. (2003). Real handball goalkeeper vs. virtual handball thrower. Presence: Teleoperators and Virtual detrimental effects in performance occurred when no Environments, 12(4), 411–421. ball information was present. Expert tennis players’ per- Bideau, B., Kulpa, R., Vignais, N., Brault, S., Multon, F., & formance in judging where the ball would land on the Craig, C. M. (2010). Virtual reality, a serious game for court was severely compromised, with the percentage of understanding performance and training players in sport. correct judgments falling to chance level. Interestingly, IEEE Computer Graphic Applications, in press. when these same players were asked to judge the utility Bideau, B., Multon, F., Kulpa, R., Fradet, L., Arnaldi, B., & of different sources of information (body-based vs. ball), Delamarche, P. (2004). Using virtual reality to analyse links they claimed positional information pertaining to the between handball thrower kinematics and goalkeeper’s reac- arm and racquet was most informative. tions. Neuroscience Letters, 372(1–2), 119–122. Brault, S., Bideau, B., Kulpa, R., & Craig, C. (2009). Detect- 7 Conclusion ing deceptive movements in 1 vs. 1 based on global body dis- placement of a rugby player. International Journal of Virtual The results presented in this study help to clarify Reality, 8(4), 31–36. Craig, C. M., Berton, E., Rao, G., Fernandez, L., & Bootsma, the role that different sources of visual information play R. J. (2006). Judging where a ball will go: The case of curved on the motor behavior of an athlete immersed in a vir- free kicks in football. Naturwissenschaften, 93, 97–101. tual environment. They reinforce the importance of Cutting, J. E. (1997). How the eye measures reality and virtual being able to track at least part of the ball trajectory to reality. Behavior Research Methods, Instruments, & correctly predict where the ball is going to end up. How- Computers, 29, 27–36. ever, body-based cues can be significant depending on Franks, I. M., & Hanvey, T. (1997). Cues for goalkeepers; the throwing action. Future studies will attempt to see High-tech methods used to measure penalty shot response. which kind of visual cues of the throwing movement can Soccer Journal, 42, 30–38. be detected. Moreover, these findings should be com- Jackson, R. C., & Mogan, P. (2007). Advance visual informa- pared with other virtual sport situations. The influence tion, awareness and anticipation skill. Journal of Motor of the handedness of the goalkeeper could also be inves- Behavior, 39, 341–351. tigated. Jackson, R. C., Warren, S., & Abernethy, B. (2006). Anticipa- tion skill and susceptibility to deceptive movement. Acta Psy- chologica, 123, 355–371. Acknowledgments Katz, L., Parker, J., Tyreman, H., Kopp, G., Levy, R., & Chang, E. (2006). Virtual reality in sport and wellness: Prom- The authors want to thank the Bunraku team for all the sup- ise and reality. International Journal of Computer Science in port and resources made available during this study and in par- Sport, 4(1), 4–16. ticular Julien Bilavarn for the work done. The authors received Kelly, A., & Hubbard, M. (2000). Design and construction of written permission from JVRC to reuse two sentences verbatim a bobsled driver training simulator. Sports Engineering, 3(1), from Vignais et al. (2009). 13–24. Kulpa, R., Multon, F., & Arnaldi, B. (2005). Morphology- References independent representation of motions for interactive human-like animation. Computer Graphics Forum, 24(3), Abernethy, B., Wood, J. M., & Parks, S. (1999). Can the antic- 343–352. ipatory skills of experts be learned by novices? Research Lenoir, M., Musch, E., Thiery, E., & Savelsbergh, G. J. P. Quarterly for Exercise and Sport, 70, 313–318. (2002). Rate of change of angular bearing as the relevant

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021 290 PRESENCE: VOLUME 19, NUMBER 4

property in a horizontal interception task during locomo- behaviour in expert soccer goalkeepers. Ergonomics, 48, tion. Journal of Motor Behavior, 34, 385–401. 1686–1697. Mazyn, L. I. N., Lenoir, M., Montagne, G., & Savelsbergh, Savelsbergh, G. J. P., Williams, A.M., Van der Kamp, J., & G. J. P. (2004). The contribution of stereo vision to one- Ward, P. (2002). Visual search, anticipation and expertise handed catching. Experimental Brain Research, 157(3), in soccer goalkeepers. Journal of Sports Sciences, 20, 279– 383–390. 287. McLeod, P., Reed, N., & Dienes, Z. (2006). The generalized Slater, M. (2009). Place illusion and plausibility can lead to optic acceleration cancellation theory of catching. Journal of realistic behaviour in immersive virtual environments. Philo- Experimental Psychology: Human Perception and Perform- sophical Transactions of the Royal Society, B 12, 364(1535), ance, 32, 139–148. 3549–3557. Metoyer, R., & Hodgins, J. (2000). Animating athletic motion Slater, M., Khanna, P., Mortensen, J., & Yu, I. (2009). Visual by example. Proceedings of Graphics Interface 2000. realism enhances realistic response in an immersive virtual Molet, T., Aubel, A., Capin, T., Carion, S., Lee, E., environment. IEEE Computer Graphic Applications, 29(3), Magnenat-Thalmann, N., et al. (1999). Anyone for tennis? 76–84. Presence: Teleoperators and Virtual Environments, 8(2), Tarr, M. J., & Warren, W. H. (2002). Virtual reality in behav- 140–156. ioral neuroscience and beyond. Nature Neuroscience, 5, Multon, F., Kulpa, R., & Bideau, B. (2008). MKM: A global 1089–1092. framework for animating humans in virtual reality applica- Vignais, N., Bideau, B., Kulpa, R., Craig, C., Brault, S., & tions. Presence: Teleoperators and Virtual Environments, Multon, F. (2009). Virtual environments for sport analysis: 17(1), 17–28. Perception-action coupling in handball goalkeeping. Inter- Panchuk, D., & Vickers, J. N. (2006). Gaze behaviors of goal- national Journal of Virtual Reality, 8(4). tenders under spatial-temporal constraints. Human Move- Williams, A. M., Davids, K., Burwitz, L., & Williams, J. G. ment Science, 25, 733–752. (1994). Visual-search strategies in experienced and inexper- Savelsbergh, G. J. P., Van der Kamp, J., Williams, A. M., ienced soccer players. Research Quarterly for Exercise and & Ward, P. (2005). Anticipation and visual search Sport, 65, 127–135.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/PRES_a_00003 by guest on 27 September 2021