Gaze behavioral patterns during table tennis forehand rallies

Abstract

This study aimed to clarify gaze behavioral patterns of skilled table tennis players during forehand rallies. Collegiate male table tennis players (n = 7) conducted forehand rallies at a constant tempo (100, 120, or 150 beats per minute) using a metronome. For each tempo condition, the participants performed 30 strokes (three conditions). Gazes targeting specific areas (gaze targets) and saccadic eye movements were detected using an eye-tracking device. We found that participants gazed at the ball only 40% of the time when it approached them. The participants tended to gaze at the ball when the opponent hit it and then moved their gaze away from the ball. Furthermore, saccades were directed toward the opposite side of the court, including the opponent, after tracking the ball. These findings suggest that focusing on the opponent's motion is important for successful forehand table tennis rallies. Thus, skilled table tennis players are likely to direct their gazes and saccades toward the opponent's side of the area during table tennis forehand rallies.

Introduction

In racket sports such as cricket (Mann, Spratford, & Abernethy, 2013), squash (Hayhoe et al., 2012), and tennis (Mann, Nakamoto, Logt, Sikkink, & Brenner, 2019), skilled players direct their gaze to the future position of the approaching ball when it bounces. This is because the ball trajectory can change suddenly during bouncing. They attempt to anticipate the position at which they could better judge the trajectory of the ball as it bounces. The time from bounce to hit has been reported to be 160, 240, and 320 ms in cricket (Mann et al., 2013); 400 to 800 ms in squash (Hayhoe et al., 2012); 300, 550, and 800 ms in tennis (Mann et al., 2019); and 150 to 200 ms in high-pitch rallies in table tennis (Shinkai et al., 2023). Thus, players may not be able to look at the ball from bouncing to hitting, especially in table tennis. Therefore, gaze patterns during table tennis rallies would differ from those during other racket sports. 

In table tennis, skilled players often do not look at the ball immediately after it bounces because they can predict the ball trajectory earlier based on information about the hitting motion of the opposite player. Previous studies have suggested the importance of looking at the racket and swinging arm areas of an opposing player when predicting ball trajectories (Piras, Lanzoni, Raffi, Persiani, & Squatrito, 2016; Piras, Raffi, Lanzoni, Persiani, & Squatrito, 2015; Piras, Raffi, Perazzolo, Malagoli Lanzoni, & Squatrito, 2019). Furthermore, Shinkai, Ando, Nonaka, Kizuka, and Ono (2022) suggested that expert table tennis players look away from the ball earlier than semi-expert players during rallies, indicating that skilled players are able to predict the ball trajectory earlier than semi-skilled players. Thus, we attempted to expand the findings of previous studies by using constant forehand rallies under different tempo conditions. 

Examining saccadic eye movements during intercept performance could be beneficial for the assessment of gaze, which cannot be assessed by fixation alone (Hoffman & Subramaniam, 1995). Previous studies on interceptive sports have demonstrated that participants track a ball approaching them for as long as possible by predictive saccades, indicating that they attempt to predict the future location of the ball and acquire spatiotemporal information about the ball to hit it (Diaz, Cooper, Rothkopf, & Hayhoe, 2013; Mann et al., 2013; Mann et al., 2019). Aoyama et al., (2022) demonstrated that predictive saccades toward a moving target improve visuomotor performance, according to the results of their original psychophysical experiment. Such studies have suggested that predictive saccades to the ball are effective in the interception of a moving target. In contrast, skilled table tennis players would make saccades toward their opponents during forehand rallies, rather than toward the ball approaching players. Thus, saccade analysis can explain where table tennis players direct their gaze during the rallies. 

This study aimed to clarify the gaze behavioral patterns of skilled table tennis players during forehand rallies. We attempted to quantify detailed gaze targeting of a specific area (gaze target), and saccades during rallies. We raised the following three research questions: (a) do skilled table tennis players look at the ball during rallies and, if so, when; (b) is the gaze target only the racket, the swinging arm of the opposing player, or the location on the court where they aim to hit the ball; and (c) do they make saccades toward the opposing player rather than the ball after bouncing? 

Materials and methods

Participants

The participants were seven male college students who had participated in the All-Japan tournament (mean age, 19.7 ± 0.9 years; mean height, 169.9 ± 5.3 cm; mean body mass, 61.1 ± 4.2 kg; mean table tennis experience, 12.1 ± 2.4 years), and they reported having normal or corrected to normal vision and no known motor deficits. They were not diagnosed with stereoscopic problems or strabismus. All of the participants provided informed consent to participate in this study. This study was conducted in accordance with the tenets of the Declaration of Helsinki, and all experimental protocols were approved by the Research Ethics Committee of the Faculty of Health and Sports Sciences, at the University of Tsukuba. Written informed consent was obtained from all participants before participation. 

Experimental procedure

The participants wore an eye-tracking device (Pupil Invisible glasses; Pupil Labs, Berlin, Germany), and the calibration validation was performed before the experiment. During the validation process, the participants fixated the corners of the table as a calibration grid to verify accurate tracking. We confirmed that the error in gaze relative to each corner of the table was less than 1° of the visual angle. The experimenter (Figure 1①) conducted the forehand rallies as experimental tasks with the participants (Figure 1②). The experimenter delivered a ball to one target (diameter, 24 cm, Figure 1③) drawn on the table court of the participant's side, and the participant aimed to hit the ball to the circular target (diameter, 24 cm) Figure 1④) on the experimenter's side. Participants conducted two trials to familiarize themselves with the task. A metronome speaker (Creative MUVO 2c; Creative Technology, Kawasaki, Japan) (Figure 1⑤) was set on the table near the net to accurately control the timing of each stroke. The three tempo conditions were conducted in the order of 100, 120, and 150 beats per minute (bpm) to compare gaze behavioral patterns under different temporal constraints. The higher the tempo, the higher was the speed of the rallies. In practical situations, table tennis players are required to adapt to rallies at different speeds. In each tempo condition, the participants performed 30 strokes (hitting the ball) with the experimenter (three conditions). In addition, four of the seven participants started with the slowest tempo condition and moved onto the faster conditions, whereas the remaining three participants started with the fastest condition and moved onto the slower conditions. 

Figure 1.

 

Overview of the experimental task: ① skilled experimenter, ② participant, ③ participant side of circular target, ④ experimenter side of circular target, ⑤ speaker, ⑥, ⑦ high-speed cameras, ⑧–⑩ LED lights, ⑪ control box for gyro sensor, and ⑫ waveform generator. The red circular point at the corner of the participant side of the court shows the origin of the coordinate system to analyze ball trajectories during rallies.

Apparatus

An eye-tracking device was used to record participant's gaze targets during the experimental tasks. The images from the scene camera of the eye-tracking device were recorded at a sampling frequency of 30 Hz, and the eye movements of the eye cameras were recorded at a sampling frequency of 200 Hz. Two high-speed cameras (frame rate, 240 fps; EX-ZR200, CASIO, Tokyo, Japan) (Figure 1⑥, ⑦) were set at the sides of the court for capturing ball trajectories during experimental tasks. 

An acceleration sensor was attached to the rear of the experimenter's racket to synchronize the hit time of the scene camera of the eye-tracking device and the high-speed cameras. The light-emitting diode (LED) lights (Figure 1⑧, ⑨, ⑩) were set to provide the signal output from the acceleration sensor (MP-A0-01A; MicroStone, Shichika, Japan) (Figure 1⑪) through the waveform generator (SG-4211; IWATSU, Tokyo, Japan) (Figure 1⑫). Therefore, the LED lights flashed with vibrations when the experimenter hit the ball, and the moment of the LED flash reflected the timing of the experimenter's strike. This timing was confirmed by videos from both cameras, indicating data synchronization between them. We manually confirmed the timing using these videos. The delay between the hitting time and the LED flash was <5 ms. Prior to the experiment, this delay was measured using the CASIO high-speed camera (frame rate, 1000 Hz). The hitting movement and LED flash were recorded using the same camera, and the delay was measured. The time at which the LED lights flashed was captured from each image using the CASIO high-speed camera (frame rate, 240 fps) (Figure 1⑥, ⑦). 

Data analysis

Video footage from the scene camera of the eye-tracking device was digitized using motion analysis software (Frame-DIAS Ⅳ; DKH, Tokyo, Japan) to determine the coordinates (pixels) of the ball position relative to the head in each video footage frame. The ball coordinates were resampled from 30 to 200 Hz by using second- or third-order spline interpolation to match them with the eye direction data. There were no errors between the eye and ball direction data for all tempo conditions. The pixel values of the coordinates were converted to visual angles based on the specifications of the eye-tracking device (horizontal angle, 82°/1088 pixels; vertical angle, 82°/1080 pixels). The coordinate origin of the scene camera was defined as the center screen of the video footage. Gazes targeting a specific area (gaze target) during rallies were analyzed for individual participants as described below. 

Duration of gaze target on the ball

The relative time (%) of the gaze target on the ball was calculated based on the above definitions. A gaze target was defined as a gaze being maintained at a single area of interest, either stationary or moving, for a minimum of 120 ms (van Biemen, van Zanten, Savelsbergh, & Mann, 2022) with a visual angle of less than 3° from the center coordinate of the ball (Piras et al., 2016; Rodrigues, Vickers, & Williams, 2002). Although these previous studies did not provide a speed criterion for the detection of gaze targets on the ball, gazing at a single area of interest within the same 3° or for at least 120 ms clearly reflects that the gaze stays on the target. 

Gaze targets on other areas of interest

In addition to the gaze target on the ball, seven other areas of interest where the gaze could be directed were defined: the experimenter's racket, opposite court near the experimenter (near), opposite court on the middle line (middle), opposite court far from the experimenter (far), circular target on the opposite court, space between the circular target and racket, and trunk of the experimenter (Figure 2). We divided the table into three parts to clarify whether the gaze directions were close to the ball or the opponent. All gaze targets were examined to determine whether the difference between the coordinates of the gaze target and the boundary of the area of interest was within 3° of the visual angle. If the ball overlapped the position of the racket when the experimenter hit it, we assigned the gaze to the position of the ball. Also, if the ball overlapped the position of the target, the gaze was assigned to the position of the ball. Furthermore, if the racket overlapped with the position of the experimenter, especially the trunk, we assigned the gaze to the position of the racket. These choices were conservative, because they contradicted our preferred interpretation that the gaze pattern does not look at the ball. Contrary to the definition of how to detect gaze targets on the ball, the gaze was considered to have left one area as soon as it entered the next area, even if it remained within 3° of the original area. Data analysis of the gaze targets was performed by manually encoding frame-by-frame video images using the eye-tracking device. After collecting all data, each gaze target in relative time (0%–100%) was determined using the statistical mode. In other words, we determined the most common region gazed at each percentage of time for each participant, and then we determined the percentage of gaze targets for each region across participants. This process allowed us to identify detailed gaze targets during rallies. It also allowed the comparison of gaze targets among all participants. 

Figure 2.

 

Eight defined areas of interest in this study. Eight areas of interest were defined: the racket, the ball, the opposite court near the experimenter (near), the opposite court on the middle line (middle), the opposite court far from the experimenter (far), the circular target on the opposite court, the space between circular target and racket, and the trunk of the experimenter.

To evaluate the saccades in each shot, the distance between the eye direction in the next and previous samples was divided by the time between those samples to determine the eye velocity at each moment. Based on a previous study of racket sports (Mann et al., 2019), we determined the median absolute acceleration during the trial and defined saccades as portions of the gaze trace between which the acceleration was more than five times this median absolute acceleration or less than negative five times this median absolute acceleration. Moreover, the gaze velocity had to exceed 40°/s for five consecutive frames and had to be at least 20% greater than that of the ball to avoid erroneous detection (Arthur et al., 2021; Arthur, Vine, Wilson, & Harris, 2024). However, we did not apply the latter method when the gaze direction was opposite to that of the ball. Immediately after the saccades, all gaze targets showed one of the eight defined areas of interest. Therefore, the gaze targets immediately after saccades were categorized into eight areas of interest. 

The gaze targets and saccades were defined using the head coordinate system. Although we captured head movements of participants with the built-in gyroscope of the eye-tracking device (sampling rate, 200 Hz), only a small amount of head rotation was detected (average total horizontal range, –4.7° ± 2.7°; average total vertical range, –1.6° ± 2.2°). In this case, negative values indicated a rightward deflection in the horizontal direction and a downward deflection in the vertical direction. 

To evaluate whether the participant's hits were successful in terms of the racket hitting the ball and the ball landing at the target area, the images of the two high-speed CASIO cameras (frame rate, 240 fps) (Figure 1⑥, ⑦) were used for constructing three-dimensional coordinates of the ball trajectories. The Frame-DIAS IV motion analysis software that we used incorporated the direct linear transformation method. The images of the moment the ball landed on the experimenter's side of the court were analyzed to calculate the distance relative to the center of the circular target, reflecting the hitting accuracy of the participants. 

In this study, the data on the gaze targets and ball trajectories are presented relative to the normalized time. The normalized time began at the moment when the experimenter hit the ball toward the participant (time = 0%), and ended one frame before the moment the experimenter hit the ball back again (time = 100%), after the participant has returned it to him (time = 50%). The normalized time was measured using the high-speed camera (Figure 1⑥). The onset of the experimenter's hit was inferred from the LEDs on his racket, whereas the onset of the participant's hit was inferred by manual coding from the images of the high-speed camera, which captured not only the ball trajectories but also the racket movements of the experimenter and the participants. 

Statistical analysis

Statistical analysis was not applied to the collected data because all of the results in this study could be demonstrated without such statistical analysis. The obtained valuables were calculated separately for each trial and then averaged across trials for each condition. 

Properties of ball trajectories during rallies

To confirm the reliability of the test, two components were analyzed: (a) time from one player's hitting to another player's hitting during rallies, and (b) ball positions relative to the circular target. For the ball trajectory, the position at which the ball bounced in all trials was calculated using the Frame-DIAS IV motion analysis software. The two high-speed CASIO cameras (frame rate, 240 fps) (Figures 1⑥, ⑦) were used for capturing ball trajectories. The origin of the coordinate system was the corner of the participant's side of the court (the red point in Figure 1). The times from the experimenter's hitting to the participant's hitting were 576 ± 15.6 ms for 100 bpm, 486 ± 13.8 ms for 120 bpm, and 391 ± 10.7 ms for 150 bpm. The times from the participant's hitting to the experimenter's hitting were 617 ± 14.0 ms for 100 bpm, 507 ± 12.3 ms for 120 bpm, and 401 ± 11.0 ms for 150 bpm. Thus, the forehand rally tasks were successfully conducted in each condition. 

All trials across the participants were successful in terms of the racket hitting the ball; however, not all hits were successfully landed on a circular target. The average numbers of successful balls landing on the circular target in all tempo conditions were 16.4 ± 4.6 hits out of 30 hits in the 100-bpm condition, 17 ± 3.7 hits out of 30 hits in the 120-bpm condition, and 12 ± 4.6 hits out of 30 hits in the 150-bpm condition. Individual hitting accuracy is shown in Table 1. Figure 3 indicates that the plots of ball positions on the court on the experimenter's side were scattered very close to the circular target or within a small area within the circular target. Furthermore, gaze deployment in hit trials did not differ from that in missed trials (Figure 4). 

Table 1.

 

The hitting accuracy of each participant in all tempo conditions. The total potential number of hits at each tempo condition was 30 hits.

 

Figure 3.

 

Ball positions relative to the center coordinate of the circular target at each participant when the ball bounced on the experimentere experimenter (f blue, orange, and dark cyan dots indicate single trial data in the 100-, 120-, and 150-bpm conditions, respectively. The horizontal and vertical dotted lines indicate the mean values of the ball position for each participant. Figure part labels (A–G) indicate indexed name of participants.

Figure 4.

 

Gaze positions relative to the ball position in each normalized time point. The dark blue, orange, and dark cyan dots indicate single trial data for ihit,”iand the salmon red, lime green, and magenta dots indicate single trial data for amiss”iss magenta-, 120-, and 150-bpm conditions, respectively. The horizontal and vertical dotted lines indicate the mean values of the gaze position for each participant.

Results

Duration of gaze targets on the ball

The mean relative durations of gaze targets on the ball approaching participants in each tempo condition were 19.6% ± 2.4% for 100 bpm, 19.5% ± 1.5% for 120 bpm, and 19.4% ± 1.1% for 150 bpm (Figures 5A to 5C). The mean relative durations of gaze targets on the ball approaching the experimenter in each tempo condition were 14.7% ± 4.6% for 100 bpm, 17.8% ± 5.5% for 120 bpm, and 20.5% ± 4.9% for 150 bpm (Figures 5A to 5C). 

Figure 5.

 

Average percentage of gaze targets at each normalized time in all tempo conditions. The x-axis shows normalized time from 0% to 100%. The normalized time began when the experimenter hit the ball toward the participant (time = 0%) and ended when the experimenter hit the ball back again (time = 100%) after the participant has returned it to him (time = 50%). The black solid vertical line along 35% of normalized time indicates the moment the ball bounced on the participantalues of the gaze positiony-axis shows the percentage of gaze targets at each normalized time. The blue shaded areas show averaged durations of gaze targets on the ball. Figure part labels (A–C) correspond to each tempo condition.

Gaze targets during rallies

In this study, the gaze targets during forehand rallies demonstrated eight defined areas of interest. After gazing at the ball approaching the participants, most of the gaze targets stayed on the opposite court in each tempo condition (100 bpm, 99% of total trials; 120 bpm, 100% of total trials; 150 bpm, 97.7% of total trials). Figure 6 shows an example of gaze behavior during rallies. The point of gaze remained at the participant's side of the court in all video frames, indicating that the participant looked away from the ball and approached him soon after the experimenter hit the ball. 

Figure 6.

 

Example of a gaze behavior during one round of rallies. The cross points between two red lines and white lines at each video image indicate the gaze and ball positions in the head coordinate system, respectively.

Figures 5A, 5B, and 5C indicate that the participants directed their gaze at the ball nearly around the same time that the experimenter hit the ball and gradually shifted their gaze to other defined areas of interest. However, after moving away from the ball, the gaze targets varied among participants, even though the gaze was directed to the opposite side of the court during rallies. Figures 7, 8, and 9 show the gaze targets at each normalized time per trial for all tempo conditions. These figures indicate that the gaze targets after moving away from the ball approaching participants varied among individual participants. 

Figure 7.

 

Individual gaze directions at each normalized time in the 100-bpm condition. Figure part labels (A–G) indicate indexed names of participants. The definition of the x-axis is the same as in Figure 5. The y-axis shows the stroke number from 1 to 30 strokes.

Figure 8.

 

Individual gaze directions at each normalized time in the 120-bpm condition. The definitions of the x-axis and y-axis are the same as in Figure 7. Figure part labels (A–G) indicate indexed names of participants.

Figure 9.

 

Individual gaze directions at each normalized time in the 150-bpm condition. The definitions of the x-axis and y-axis are the same as in Figure 7. Figure part labels (A–G) indicate indexed names of participants.

Distribution of saccades direction during rallies

Figures 10, 11, and 12 show polar histograms of the gaze and ball directions in each time bin. The histograms at 0% to 10% and 10% to 20% of the normalized time indicate no saccades because the participants in this study fixated on the ball in these time bins. In contrast, saccades frequently occurred at 30% to 40%, 40% to 50%, 80% to 90%, and 90% to 100% of the normalized time in all tempo conditions. Saccadic directions tended to be opposite to ball directions at 30% to 40% and 40% to 50% of the time because participants in this study made saccades toward the court near the experimenter, not the ball approaching them (Figure 5). 

Figure 10.

 

Saccade direction in the 100-bpm condition. Each graph represents the number of directional distributions of the saccade (cyan) and ball (orange) direction at a normalized time. The ball direction was detected at the same time the saccade was detected.

Figure 11.

 

Saccade direction in the 120-bpm condition (see Figure 10).

Figure 12.

 

Saccade direction in the 150-bpm condition (see Figure 10).

Figure 13 shows the occurrence of saccades for each defined direction at each normalized time, indicating the distribution of when saccades occurred during rallies. The gaze targets after saccades in the ball-approach phase (in the time of 0%–50%) tended to be other defined areas of interest away from the ball, suggesting that not only the gaze target but also the saccadic direction were on the opposite side of the court. Furthermore, the gaze targets after saccades around 90% to 100% (just before the experimenter hit the ball) of normalized time tended to be the target and space. 

Figure 13.

 

Occurrence of saccades for each defined gaze target in each normalized time in all tempo conditions. Figure part labels (A–H) indicate defined gaze targets.

Discussion

This study examined the gaze behavioral patterns of skilled table tennis players during forehand rallies. We found that participants gazed at the ball only 40% of the time when the ball approached. Moreover, the time taken to gaze at the ball during rallies was when the opponent hit the ball. This result reflects the fact that they did not track the ball or approach it with their gaze. However, the gaze target after gazing at the ball was not consistent among individual participants, although their gaze was directed to the opposite side of the court during the rallies. Furthermore, saccades to the ball were infrequent, suggesting that most saccades during rallies were focused in the opposite directions. Thus, skilled table tennis players are likely to direct their gaze and saccades toward the opponent side of the area during table tennis forehand rallies. 

Gaze targets on the ball during rallies

The gaze targets on the ball were observed immediately before and after the experimenter hit the ball under all tempo conditions. Furthermore, the relative duration of the gaze targets on the ball was only approximately 40% of the time the ball approached under all tempo conditions. These results support the importance of table tennis players gazing at the ball during the initial ball-approach phase (Ripoll & Fleurance, 1988, Shinkai et al., 2023; Shinkai, Ando, Nonaka, Kizuka, & Ono, 2024). Rodrigues et al. (2002) also suggested that participants move their gaze away from the ball in the earlier phase of the ball approach (first portion of the ball trajectory), supporting the results of this study. They conducted an experiment in which participants returned the ball to the right or left target cue under three different timing cue conditions (pre-, early, and late cue conditions). In their study, the cue light was illuminated before the serve (pre-cue), during the initial phase of ball flight (early cue), or the last phase of the ball flight (late cue). In particular, they showed that the gaze-ball angles for the early and late cues were greater than 3° after 40% of the trial time. In other words, the participants in their study had to judge which cue was lit by gazing at the opposite court until it was lit. Therefore, the gaze target would remain on the opposite court, even though the ball approached the participants. To our knowledge, this is the first study to demonstrate similar results using a constant forehand rally task. 

Gaze targets during rallies

Most of the gaze targets during rallies pointed to the eight defined areas of interest, indicating that the gaze targets of the participants dwelled on the opposite side of the court, even in the ball-approach phase. The participants directed their gaze to the ball immediately before and after the experimenter hit it and gradually shifted their gaze to other defined areas of interest. This supports the results for the gaze targets on the ball mentioned above. Furthermore, these results suggest that skilled table tennis players direct their gaze after gazing at the ball during the ball-tracing phase. Shinkai et al. (2022) indicated that the gaze-ball angle of expert players was significantly larger than that of semi-expert players, suggesting that the expert players looked away from the ball approaching them. However, it was uncertain in detail where expert players gazed at the ball. Therefore, the present study provides valuable insights for further understanding of gaze behavioral patterns. Moreover, the participants did not move their gaze to where the ball bounced. Previous studies discussed where athletes make predictive saccades after ball bouncing (Diaz et al., 2013; Mann et al., 2013; Mann et al., 2019). These studies have indicated that predictive saccades are made at the point that the ball reaches after bouncing. However, the results of the present study differ from those of previous studies. This discrepancy could be due to the small size of the table tennis court, which creates severe time constraints for players to hit the ball approaching them. Furthermore, the participants in this study had expert skills in rally tasks, because they were able to execute successful rallies even while gazing at the opposite side of the court. 

The importance of looking at an opponent has been highlighted in racket and bat sports. Previous studies on baseball (Nakamoto, Fukuhara, Torii, Takamido, & Mann, 2022) and softball (Takamido, Yokoyama, & Yamamoto, 2022) have indicated that skilled players estimate the ball speed based on the kinematic information of pitching motion. In table tennis, it is likely important to acquire kinematic information about the opponent to estimate not only the ball speed but also the ball direction (right or left). In particular, Piras et al. (2016) suggested that acquiring kinematic information in the hand–racket area improves reaction time and accuracy in the direction of the ball hit by the opponent. Thus, it is suggested that acquiring the kinematic information about an opponent is related to gaze behavioral patterns during forehand table tennis rallies. 

The gaze behavior observed in this study could show one of the general patterns because such rallies are always used in practical situations, regardless of skill level. However, the experimental conditions in this study would be an easily predictable rally compared to a regular match for skilled participants. Our previous study (Shinkai et al., 2022) showed that expert table tennis players of comparable competition level as in this study spent less time looking at the ball approaching the participant than semi-expert table tennis players, even in a set-up similar to that of this study. Thus, such a gaze pattern may not simply be an inherent behavior but may also be a unique gaze pattern by skilled players. 

As shown in Figures 7, 8, and 9, all of the participants tended to direct their gaze on the ball at 0%, 10%, and 100% of the normalized time. Although there were large individual differences in where participants looked, the finding was that participants looked at the ball approaching them for shorter periods because of the prediction of where the ball would come. In particular, visual information on a circular target is utilized for accurate hitting performance. Looking at where players aim is beneficial for improving hitting accuracy during rallies. Large individual differences in gaze targets were associated with differences in the prediction of the ball or types of peripheral vision use during rallies. In the next section, we discuss each participant's gaze patterns. 

Individual gaze patterns

In this section, we discuss the individual gaze patterns based on Figures 7, 8, and 9. 

Participant A. This participant gazed at the experimenter's side of the court immediately after fixating on the ball under all tempo conditions. In the 100-bpm condition, the participant gazed at the circular target when he hit the ball and then at the experimenter's racket at 80% of the normalized time. In the120-bpm and 150-bpm conditions, the percentage of gaze targets in the space (Figure 2) was increased rather than the participant gazing at the circular target or the experimenter's racket. This space is the area of interest between the experimenter's racket and the experimenter's side of the court, indicating that there is no specific target. However, gazing at the space plays a functional role in simultaneously acquiring two visual cues: the ball and the experimenter's movements. This type of gaze pattern is called a gaze anchor and allows a player to monitor the movement of visual targets with peripheral vision (Klostermann et al., 2020). Participant A gazed at the position where the ball was hit by the experimenter. 

Participant B. Gaze targets for Participant B indicated primarily the far side from the experimenter after gazing at the approaching ball, reflecting that this participant could use a visual-pivot strategy to watch the entire visual scene, especially for the opponent area during rallies. This strategy is one of the best strategies for using peripheral vision during performance (Kato, 2020; Williams & Elliott, 1999) and is defined as a gaze position optimally located between the relevant information sources to allow for frequent fixation transitions with minimal saccadic costs (Klostermann, Vater, Kredel, & Hossner, 2020). His gaze strategy was consistent across all of the tempo conditions. Therefore, the visual-pivot strategy does not depend on the tempo of the rally within this range of tempo conditions. 

Participant C. This participant gazed at the experimenter's side of the court and at the experimenter's trunk when the ball approached him; therefore, this participant hit the ball with peripheral vision. Although this participant did not always maintain his gaze position on the circular target during rallies, it is most likely that he tried to aim at the circular target while keeping his gaze targets close to it. 

Participant D. This participant tried to aim for the circular target during rallies because the percentages of gaze targets on the circular target and on the near were clearly higher than other gaze targets when he hit the ball. Although this participant gazed at the near rather than the circular target in the 150-bpm condition, both gaze targets between the circular target and the near are very close to each other. Therefore, the participant aimed for a circular target under all tempo conditions. 

Participant E. This participant maintained his gaze at the experimenter's racket during the rallies; therefore, observing the experimenter's racket can be effective. However, his hitting accuracy was intermediate among all of the participants (Table 1). 

Participant F. This participant clearly gazed at the experimenter's trunk as the ball approached him. In the 100-bpm condition, it is possible that this participant checked whether his hit was accurate within the circular target, because the percentage of the gaze target on the circular target was high after he hit the ball. However, the higher the tempo, the less frequently the participant looked at the circular target in the same period. Instead of gazing at the circular target, the participant gazed at the ball, reflecting that he had lost the opportunity to aim at the circular target during the rallies. In fact, his hitting accuracy decreased gradually from 100 bpm to 150 bpm (Table 1). 

Participant G. This participant had different gaze patterns for each tempo condition. In the 100-bpm and 120-bpm conditions, he attempted to use peripheral vision to monitor the ball, experimenter and circular target simultaneously. In the 150-bpm condition, his gaze was frequently located on the circular target and its surrounding areas, such as the near, and middle areas, reflecting that he aimed at the circular target, even in the highest tempo condition. His hitting accuracy was the highest among all of the participants for the 150-bpm condition. 

Occurrence of saccades during rallies

Participants in this study often made saccades toward the other defined areas of interest (the experimenter's direction) at 30% to 40%, 40% to 50%, 80% to 90%, and 90% to 100% of the normalized time. Hoffman and Subramaniam (1995) suggested that visuospatial attention is an important mechanism for generating voluntary saccadic eye movements. Rolfs, Jonikaitis, and Deubel (2011) suggested that attentional shifts to the future positions are correlated with upcoming saccades to these positions, reflecting a functional correlation between visual attention and saccadic eye movements. Our results indicate that both gaze- and saccade-related visual attention could be directed toward the experimenter's side of the court, even in the ball-approach phase. 

However, our results were not consistent with previous studies in which saccades were directed ahead of the ball to predict the ball trajectories for intercept performance in baseball (Kishita, Ueda, & Kashino, 2020), cricket (Land & McLeod, 2000; Mann et al., 2013), and squash (Hayhoe et al., 2012). Rather, our results show a novel pattern of gaze behavioral strategies in which the skilled table tennis players successfully executed constant forehand rallies without gazing at or paying attention to the entire trajectory of the ball. 

Conclusions

This study examined the gaze behavioral patterns of skilled table tennis players during constant forehand rallies. The results indicate that, although participants tended to gaze at the ball when the experimenter hit it, their gaze remained stationary on the experimenter's side of the court during the ball-approach phase. Furthermore, saccades were made toward the experimenter's side of the court after gazing at the ball. These findings suggest that kinematic information about an opponent is important for successful rallies. Thus, skilled table tennis players are most likely to use visual patterns specific to interceptive sports players to estimate spatiotemporal information about the ball. 

Acknowledgments

The authors thank the members of the table tennis team at the University of Tsukuba for their participation. 

Supported by Japan Society for the Promotion of Science KAKENHI Grant Numbers 23K24749 and 24KJ0480. 

Commercial relationships: none. 

Corresponding author: Seiji Ono. 

Email: [email protected] 

Address: Institute of Health and Sport Sciences, University of Tsukuba, Tennodai, Tsukuba, Ibaraki 305-8577, Japan.