Microsaccades and Prediction of a Motor Act Outcome in a Dynamic Sport Situation

Abstract

Purpose: Microsaccades could indicate the place where our mind is unconsciously focusing, although our gaze is directed elsewhere. Many studies report the importance of microsaccades in visual scene perception, but none of them has addressed their relationship with the perception of a dynamic action and the prediction of its outcome.

Methods: Expert and novice table tennis players were asked to fixate their gaze on a precise spot while viewing the launch of a ball whose final landing had to be predicted. Four separate epochs of the action were considered for their information content. The correctness of the prediction and microsaccade statistics were measured in order to estimate the relationship between covert attention and predictions.

Results: Microsaccades rate showed a time course modulated by the different epochs, with a significant enhancement during the post-bounce. In this epoch, novices showed a significantly higher rate than experts when the responses were correct. Duration and amplitude were highest in the pre- and post-bounce periods and lowest in the other two. Mean microsaccades direction was toward the stimuli that most probably attracted the visual attention (ball or racket), whereas there was no relationship with the predicted side of the final bounce.

Conclusions: Distribution of microsaccades can be influenced by attentional cues in a task-specific situation, revealing links between visuomotor performance and covert attention shifts in fast visuomotor perception. Microsaccade orientation is conditioned by objects that attract visual attention and not by the direction in which action is expected to be performed.

Attention is the ability to select salient stimuli. Such stimulus selection involves the orientation of attention, which can be overt, when the gaze is directed toward an object of interest, or covert, when attention is shifted in the absence of gaze shifts. Studies of attention and eye movements have helped to clarify the functional and neural interactions between these two systems.¹ The relationship between covert attention and oculomotor system is demonstrable through microsaccades, which are small and rapid eye movements that could indicate the place where our mind is unconsciously focusing, although our gaze is directed elsewhere.² 

Previous studies have shown a relationship between microsaccades and visual perception in natural scenes and visual search tasks.^3–6 It has been hypothesized that the spatial location of attention strongly influences the rate and/or direction of microsaccades during visual fixation. Free-viewing tasks that do not require the subject's attentive fixation may lead to reduced microsaccade production. Microsaccade rate is correlated with the perceptual state of target visibility, suggesting that measured microsaccade rate and direction are reliable indicators of the perception of a visible target.⁷ 

Visual attention plays an important role in sports, in which players must look at the activities and positions of multiple opponents and playmates simultaneously.⁸ Experts require knowledge of the positions of other players as well as the prediction of how such positions will change over time. A greater ability to attend to elements of the game or to encode the scene into more meaningful patterns might explain how basketball or soccer players make ‘‘no look'' passes to an empty location knowing that the receiver will be there to catch the ball. To date, no research has investigated the role of microsaccades in a task in which a subject must predict the outcome of an action. In racket ball, the opponent's overt attention is directed mostly to hand, racket, and ball areas, especially when the ball is hit by the racket.^9,10 Thus, we can hypothesize from these elements that the player can predict the outcome of the throw, and covert attention would be directed toward these areas when one fixates on the opponent's body center. 

Given the relationships among microsaccades, visual perception, and direction of covert attention, our main questions were: (1) which relationship exists between microsaccades and prediction of outcome; (2) which are the most important action phases for a correct prediction? and (3) how does expertise modify the strategy of information acquisition? 

To address these questions we recorded eye movements from naive and expert tennis table players, who had to predict the outcome of an opponent's ball throw. Then, we related the microsaccade number, time distribution, and direction to the time interval preceding the throw, as well as, to the success of the prediction. 

Methods

Participants

Eye movements and behavioral responses were recorded in 22 healthy volunteers. The participants were divided in two groups: a novice group of 15 students (4 female and 11 male; mean 26.8 ± 3.7 years of age) and an “expert players” group of seven elite athletes (all male, mean 23.1 ± 7.6 years of age). The experts had played table tennis at a professional level for an average of 10 (±2.6) years. All participants received a verbal explanation of experimental procedures, and informed consent was obtained before the beginning of the experiment. All subjects had normal or corrected-to-normal vision, and none reported any uncompensated visual deficit. The experimental protocol was approved by the Institutional Ethics Committee of the University of Bologna, Italy. 

Stimuli and Procedure

A professional table tennis coach (whom we call “coach”) was filmed from participants' perspective with a digital video camera (300 frames/s; maximum resolution of 1280 × 960 pixels; Casio, Montrose, CA, USA) while responding to a ball emerging from a throwing machine (speed ∼120 km/h). The coach used the backhand drive (BD) technique, one of the four basic table tennis strokes, which is performed with the paddle held at a slightly closed angle, so that, using the elbow, the player can move the paddle forward and in an upward direction, hitting the ball at the peak of the bounce. We filmed 10 BD strokes, subdivided into five strokes directed to the right and five to the left side of the participant's game field. These videos were used as stimuli and were shown to the participants on a video screen. 

The experiments were performed in a darkened room. Stimuli were presented by a retro-video projector (720 × 486 resolution; frame rate, 60 Hz; model EB-W12; Epson, Long Beach, CA, USA) positioned 300 cm from a translucent screen. The screen covered 38° × 29° of visual field and was placed 180 cm from the subjects' eyes. 

Subjects sat in a chair, with the head immobilized by a chin rest. They were instructed to fixate on a red dot (0.6° in diameter), which was placed in the middle of the coach's chest (Fig. 1). Subjects were enforced to fixate on the target for the duration of the video. Video sequences in which participants' eyes were out of the fixation dot were discarded from subsequent analysis. The video started with a still image of the coach, lasting 2000 ms (Fig. 1A). Then the ball appeared in the lower portion of the screen (subject's point of view) (Fig. 1B), bouncing on a portion of the coach's table (Fig. 1C). The video ended when the ball made contact with the coach's racket, and the subjects had to predict the bouncing side of the ball (i.e., in the left or in the right portion of the game field on the participant's side) by pressing one of two buttons on a gamepad (Fig. 1D). The whole action was divided in four time epochs, as follows. 

Figure 1

 

Backhand drive (BD) technique. The four video epochs show the “coach” attending the ball emerging from a projection machine and responding to it with BD technique. (A) No ball phase with duration of 2000 ms; (B) pre-bounce phase with a duration of 334 ms shows the appearance of the ball; (C) post-bounce phase with a duration of 134 ms shows the ball-table contact, and (D) the ball-racket contact, that marks the beginning of the response phase, lasting between 150 and 1000 ms.

No Ball.

This epoch occurred from the first frame showing the participants' eyes on the red dot to the last frame before the appearance of the ball. The distance between the center of the red dot and the center of the racket was approximately 1° (Fig. 1A). 

Pre-Bounce.

The pre-bounce epoch lasted from the first frame showing the ball to the last frame before the ball bounced in the portion of the coach's table. The distance between the center of the red dot and the center of the ball, when it appeared on the screen, was approximately 14° (Fig. 1B). 

Post-Bounce.

Post-bounce epoch was from the first frame showing the ball–table contact to the last frame before the coach's racket hit the ball. The distance between the center of the red dot and the center of the ball, when it bounced on the table, was approximately 6.4° (Fig. 1C). 

Response.

The final epoch lasted from the first frame showing the racket–ball contact to the participants' key-press response. The distance between the center of the red dot and the center of the racket–ball contact was approximately 1° (Fig. 1D). 

Each video lasted 2468 ms. The 10 videos were randomly presented and were shown 10 times to each participant. 

Eye Movement Recording

Horizontal and vertical eye movements were recorded binocularly by a video-based eye tracking system (EyeLink II; SR Research Ltd., Mississauga, Ontario, Canada). The system consisted of two miniature cameras mounted on a leather-padded headband. Pupil tracking was performed at 500 samples/s, with high spatial resolution (<0.005°) and low noise (<0.01°). 

Before each participant was tested, eye tracking calibration was carried out by letting the participant fixate on a target presented in random order on each of a nine-point square grid. Then, data validation was performed, and drift correction was executed by applying a corrective offset to the raw eye position data after every movie. Calibration and validation of the system was repeated for each trial to avoid measurement errors due to participants' repositioning movements. 

Data Analysis

Anticipation Test.

This test was made in order to find out the ability to make accurate predictions and the time needed to get it. It was assessed as follows. 

1.  
   Response accuracy (RA). The percentage of trials in which the subject's response was correct. A 2 × 2 repeated measures ANOVA was used to analyze the percentage of correct trials in which response accuracy (correct, incorrect) was the within-subjects factor and expertise (experts, novices) was the between-subjects factor.
2.  
   Key-press response time (RT). The time (in milliseconds) from the coach's racket–ball contact to the button pressed by the participant. A 2 × 2 repeated measures ANOVA was performed to analyze the key-press response time in which response accuracy (correct, incorrect) was the within-subjects factor and expertise (experts, novices) was the between-subjects factor.

Microsaccade Dynamics.

Microsaccades were defined as movements less than 1° and with the same peak velocity versus amplitude curve as large saccades.¹¹ To identify microsaccades, we applied the Engbert-Kliegl algorithm.¹² To reduce the amount of potential noise, we considered only binocular microsaccades lasting at least three data samples (6 ms), with velocity threshold detection set at 6. Trials with incorrect fixations, eye blinks, or behavioral errors were discarded. We also removed portions of data where very fast decreases and increases in pupil area occurred (>50 units/sample; such periods are probably semi-blinks, where the pupil is never fully occluded), and disregarded the 200 ms before and after each blink or semi-blink to eliminate the initial and final parts, where the pupil was still partially occluded.¹³ When subjects first performed the tasks, they tended to move their eyes, and this tendency was dramatically reduced after practice. Such trials were discarded from analysis. 

Microsaccade amplitude, duration, and peak velocity were first calculated for each subject, under each condition, and in each epoch separately. Then, values for all subjects under each condition and group were averaged. Microsaccade rates were calculated taking into account only the time spent in fixation periods: the total number of microsaccades in each subject under each condition and epoch was divided by the total time spent in fixation during that condition in each epoch. 

A 4 × 2 × 2 repeated measures ANOVA was performed separately to analyze microsaccade rates, amplitudes, durations, and peak velocities. Epochs (no ball, pre-bounce, post-bounce, response) and response accuracy (correct, incorrect) were the within-subject factors and expertise (experts, novices) the between-subjects factor. 

We computed the two-dimensional distribution of all microsaccade directions in the different epochs described above. We performed the Watson-Williams test for homogeneity of means in which the null hypothesis was that the orientations of microsaccades between epochs (no ball versus pre-bounce; pre-bounce versus post-bounce; and post-bounce versus response) of each group (experts, novices) have similar continuous distribution at the 5% level of significance. Furthermore, analysis took into account response accuracy (correct, incorrect) as a dependent variable to reveal any relationships between correct responses and microsaccade orientation. ANOVA (SPSS version 13.0 software, Chicago, IL, USA) and circular statistics (Oriana version 4.0 software [Kovach Computing Services, Anglesey, Wales, UK] for Windows [Microsoft, Redmond, WA, USA]) were used to determine possible shifts of visual attention in this type of task. Results were considered significant at a P value of <0.05. Post hoc tests were corrected with a Bonferroni procedure (P < 0.006). 

Results

Data Preprocessing

The total number of trials analyzed is shown in Figure 2. Responses with RT shorter than 150 ms and longer than 1000 ms (early or delayed responses) were discarded.¹⁴ After pre-processing (see Methods for description), we used 624 videos for athletes (of a total of 700; 76 videos were excluded) and 1331 for novices (of a total of 1500; 169 videos were excluded). Response time was also related to response accuracy (correct, incorrect). 

Figure 2

 

Pre-processing data. Flow chart shows the number of trials counted (sum and percentage) from each group and subdivided into correct and incorrect responses.

Anticipation Test

For response accuracy, ANOVA showed an interaction effect for expertise × response accuracy (F_1,21 = 5.77, P = 0.026, partial η² [ηp²] = 0.22). Experts performed 67% of trials correctly, failing in 33% of them, whereas novices gave correct responses in 56% of trials, failing in 44% of them (Fig. 3A). 

Figure 3

 

Response accuracy and key-press response time. (A) Pie charts show percentages of correct (black) and incorrect (gray) trials for experts and novices. (B) Values are means ± SE of key-press reaction times (milliseconds) of experts (black bars) and novices (gray bars). *Significant differences between groups.

For key-press response time analysis, ANOVA showed a significant main effect for expertise (F_1,21 = 4.50, P = 0.045, ηp² = 0.16), in which experts showed shorter RT than novices did (mean = 434.25 ± 35.38 vs. 516.58 ± 24.17). Although no interaction effect was observed, experts responded faster in both correct (mean = 455.03 ± 25.06 ms) and incorrect (mean = 413.49 ± 24.88 ms) trials than novices, who gave correct responses in 519.02 ms (±25.83) and incorrect responses in 514.14 ms (±26.46). 

Microsaccade Rate

A comparison of cumulative microsaccade rates in each epoch, for expertise and response accuracy was done by ANOVA, which showed a significant main effect for epoch (F_3,60 = 74.69, P = 0.000, ηp² = 0.79). An interaction effect was observed for expertise × response accuracy (F_1,21 = 4.43, P = 0.048, ηp² = 0.18) and epoch × response accuracy × expertise (F_3,60 = 4.01, P = 0.037, ηp² = 0.17). 

Epoch main effect analysis revealed that participants made the greatest number of microsaccades during post-bounce epoch (mean = 2.48 ± 0.16), that the lowest occurred during response period (mean = 0.61 ± 0.04), and that this was observed under all conditions (Fig. 4). We found a significantly higher mean rate in post-bounce with respect to other epochs, in both correct and incorrect trials. On the other hand, mean rate was significantly lower in response epoch than in all the others. 

Figure 4

 

Microsaccade rate. Plots represent microsaccade rates (N/s; computed within a moving window of 100 ms) subdivided into correct (A) and incorrect (B) responses during the four epochs (no ball; pre-bounce; post-bounce; response) across groups (black dashed lines = experts; gray dotted lines = novices). Timeline starts with the last 300 ms of no-ball epoch, zero represents the appearance of the ball and ends after the first 300 ms of response epoch. (Right panel) Mean (N/s ± SE) microsaccade rate subdivided into correct (C) and incorrect (D) responses during the four epochs across groups (black bars = experts; gray bars = novices). *Significant differences between groups; †significant differences between epochs (P < 0.006).

In order to carry out a more detailed analysis of microsaccade time courses across the epochs, we drew rate/time curves, averaged over all trials of 22 participants, subdivided by expertise and response accuracy, and computed within a moving window of 100 ms. The no-ball epoch was taken as baseline, including in the graph the last 300 ms of it. In both the correct and incorrect trials, there was a reduction in rate, with a minimum of approximately 30 ms after ball appearance (start of pre-bounce epoch) and a second inhibition trough at the onset of post-bounce epoch. After that, a clear increase in rate followed for the entire post-bounce epoch, ending at the end of it (at the beginning of the response epoch), when it dropped to its minimum value (Figs. 4A, 4B). Post hoc test for interaction effects demonstrated that experts, during post-bounce epoch, made fewer microsaccades than novices when the responses were correct (Fig. 4). 

Microsaccade Duration

ANOVA of microsaccade duration showed a significant main effect for expertise (F_1,21 = 5.90, P = 0.025, ηp² = 0.23) and epoch (F_3,60 = 125.43, P = 0.000, ηp² = 0.86). Interaction effect was observed for expertise × epoch × response accuracy (F_3,60 = 3.05, P = 0.035, ηp² = 0.13). 

Analysis of expertise main effect showed that experts made microsaccades with longer duration than those of novices, and all subjects, taking correct and incorrect responses together, had shorter microsaccade duration during no-ball and response than during the other epochs. Post hoc test for interaction effects demonstrated that, in incorrect responses, experts had longer microsaccade durations during the response epoch [t (21) = 2.92; P = 0.005] than novices (Fig. 5B). 

Figure 5

 

Microsaccade duration. Histograms represent mean (± SE) microsaccade duration (milliseconds) subdivided into correct (A) and incorrect (B) responses during the four epochs (no ball; pre-bounce; post-bounce; response) across groups. Black bars = experts; gray bars = novices. *Significant differences between groups; †significant differences between epochs (P < 0.006).

Microsaccade Amplitude

ANOVA of microsaccade amplitudes showed a significant main effect for expertise (F_1,21 = 6.61, P = 0.018, ηp² = 0.25) and epoch (F_3,60 = 6.30, P = 0.001, ηp² = 0.24). Interaction effect was observed for expertise × epoch (F_3,60 = 3.67, P = 0.022, ηp² = 0.16). No significant differences were observed between correct and incorrect responses. 

Analysis of expertise main effect showed that experts had higher microsaccade amplitude with respect to novices and that all subjects, taking correct and incorrect responses together, revealed higher microsaccade amplitude during post-bounce than during the other epochs (no ball, pre-bounce, and response). Post hoc test for interaction effects demonstrated that experts made microsaccades of greater amplitude during pre-bounce epoch [Student's t-test (21) = 3.21; P = 0.004] with respect to novices (Fig. 6). 

Figure 6

 

Microsaccade amplitude. Histogram represents mean (± SE) microsaccade amplitude (degrees) during the four epochs (no ball; pre-bounce; post-bounce; response) across groups. Black bars = experts; gray bars = novices. *Significant differences between groups; †significant differences between epochs (P < 0.006).

Microsaccade Peak Velocity

ANOVA showed a significant main effect for epoch (F_3,60 = 195.90, P = 0.000, ηp² = 0.91). Interaction effect was observed for expertise × epoch (F_3,60 = 11.78, P = 0.000, ηp² = 0.37). No significant differences were observed between correct and incorrect responses. 

Analysis of epoch main effect showed that, taking correct and incorrect trials together, all subjects exhibited the highest microsaccade peak velocity during the no-ball epoch (Fig. 7B). Post hoc test for interaction effects demonstrated that experts performed microsaccades with higher peak velocity during no-ball [Student's t-test (21) = 3.06; P = 0.005] than novices, who showed higher microsaccade peak velocity during the response epoch [t(21) = 3.41; P = 0.003] (Fig. 7). Therefore, analysis of main effect showed that no-ball epoch was significantly different from the others, whereas interaction effect showed highest peak velocity during no-ball in experts and during response epoch in novices (P < 0.006). 

Figure 7

 

Microsaccade peak velocity. Histogram represents mean (±SE) microsaccade peak velocity (degrees/second) during the four epochs (no ball; pre-bounce; post-bounce; response) across groups. Black bars = experts; gray bars = novices. *Significant differences between groups; †significant differences between epochs (P < 0.006).

Orientation of Microsaccades

The orientation of microsaccades is known to be influenced by shifts of covert visual attention. To study possible changes in the orientation of attention during the different epochs of our task, we analyzed the angular distribution of microsaccade directions during each epoch in each group (see Methods). Watson-Williams test in the experts group showed significant differences between no-ball and pre-bounce (F_1,18 = 18.58, P = 0.000) and between pre-bounce and post-bounce (F_1,18 = 27.28, P = 0.000) epochs (Fig. 8A). 

Figure 8

 

Microsaccade orientation. Distributions of preferred microsaccade directions (degrees/second) in a bidimensional space centered on the fixation point. Rose diagrams show the frequency distribution of the mean vectors subdivided into groups, experts (A) and novices (B), and epochs (no ball; pre-bounce; post-bounce; response). Each bar is 22.50° in width.

During the no-ball period, microsaccades were distributed over a wide angle in the upper and lower left quadrants of the visual field, whereas in the pre-bounce epoch, the directions were restricted to a narrower angle in the lower left quadrant (where the ball appeared). During post-bounce epoch, directions were oriented to the left, likely toward the ball–racket target, and, finally, in the response period, microsaccades were mainly directed toward the lower left quadrant. 

Also, novices reported significant differences between the no-ball and pre-bounce (F_1,28 = 73.90, P = 0.000) epochs and between the pre-bounce and post-bounce (F_1,28 = 35.09, P = 0.000) epochs (Fig. 8B). The pattern of orientations seemed the same as that for the experts. No differences were observed when post-bounce and response epochs, in both groups, were compared (P > 0.05) (Figs. 8A, 8B). 

Furthermore, Watson-Williams test was used in both groups to analyze the relationship between microsaccade orientation and prediction of outcome. All participants showed microsaccades directed to the left of the fixation point (Figs. 8A, 8B). Analyses revealed no relationship between microsaccade orientation and predicted stroke direction, either correct or incorrect. 

Discussion

The importance of microsaccades in visual perception has received much attention in recent years.² Several functions have been proposed for microsaccades: counteracting perceptual fading,^15–18 improving performance in acuity visual tasks,¹⁹ and varying along a continuum as a function of size of the scene to be scanned.²⁰ Microsaccades recorded during visual fixation before a motor response to a peripheral stimulus provide an important tool for the analysis of vision, attention, and eye movements.²¹ Recently, McCamy et al.¹⁸ found that task-relevant information is associated with more microsaccades and longer fixations, so that the visual system actively uses microsaccades to acquire information. 

Few research reports have linked microsaccade production to information acquisition during perceptual tasks in which a dynamic situation requiring a fast and precise response was presented.²² Our experiment, with the intention of defining the functional significance of microsaccades under a demanding condition like table tennis indicated that the distribution of microsaccade directions can be influenced by attentional cues in a sport-specific situation. 

By forcing the participants to constantly fixate on a red dot in the middle of the opponent's chest, we found that the brain produces more microsaccades to acquire information, from video frames in which task-relevant information is richest. Indeed, the highest microsaccade values of duration, rate, and amplitude were reached during the post-bounce epoch. It has been demonstrated that larger and multiple microsaccades were more effective than smaller or single ones to restore vision, due to their ability to bring the neuronal receptive fields to regions not correlated with the target stimulus.¹⁶ The post-bounce period is probably the time in which the subject tries to figure out the coach's motor action. The highest value of microsaccade amplitude during this epoch with respect to the others suggests that microsaccade amplitude is related to the amplitude of attention shifts, as proposed by Hafed and Clark.²³ This was also found during the pre-bounce epoch, probably due to the appearance of the ball in the lower portion of the subjects' visual field. Moreover, during pre-bounce epoch in which the ball appeared, the microsaccade rate was lower than that in the post-bounce epoch (Fig. 4). After this “low microsaccadic rate,” an epoch of enhancement started approximately 334 ms after the display change (ball appearance) and extended to 468 ms (end of post-bounce epoch). This pattern of rate modulation is qualitatively similar to saccadic inhibition (i.e., the decrease in saccade rate following display changes).^12,24 The distribution of microsaccade rate during the perception of a dynamic action may be related to the “double-phase effect” in the absolute frequency of microsaccades. In our task, a microsaccade inhibition occurs at the two salient visual stimuli: the appearance of the ball in the visual field (pre-bounce epoch) and when the ball bounces on the table. This second inhibition is followed by a robust rate enhancement during the post-bounce epoch, ending in a return to the initial value (response epoch). The enhancement can be related to the attentional load. Actually, it has been described that task difficulty can influence microsaccade production. Pastukhov and Braun²⁵ found that, during visual search task, microsaccade rate depends on both the nature of the visual stimulation and the condition under which it occurs. Thus, increased microsaccade production may be due to increased attentional load (post-bounce epoch, in which prediction of the coach's motor action could be built up) and decreased microsaccade production means decreased attentional load (response epoch).⁵ 

Studies of perception and action in sports have shown that players fixate longer on task-relevant areas as they choose to “anchor” the fovea close to these key locations, so that they use the parafovea and the retinal periphery to pick up relevant information.^26,27 The effective use of such “visual pivots,” in which the gaze is centrally located between different interest areas (i.e., hands, racket, ball), enables optimal use of both the foveal and the parafoveal vision.²⁸ For this reason, when our experimental subjects maintained their fixation on the middle of the coach's chest, their “gaze pivot” fostered the shift of covert attention in order to predict the development of the action. 

Prediction is a fundamental aspect of visual perception. Land and McLeod²⁹ found that experienced cricket batsmen made a saccade to the anticipated bounce point of the ball, arriving 100 to 200 ms before the ball bounced. Given that saccades and microsaccades share not only dynamic properties but also a common oculomotor origin,^2,30–32 it could be possible that players do the same microsaccade movements toward the future bounce point of the ball or toward the moving stimulus (ball, hand–racket) around the fixation point. Indeed, microsaccade orientation analysis showed a shift of visual attention from the no-ball to pre-bounce epoch (Fig. 8A), with microsaccades directed toward the appearance of the ball and/or toward the anticipated bounce point of it. We can consider “the ball” as a stimulus that appears abruptly, being processed by an automatic and rapid shift of ‘‘exogenous'' attention.³³ If an abrupt onset activates saccade cells in the superior colliculus, their activation might influence the activity of fixation cells through their mutual inhibition. Thus, even if a saccade is not generated, evidence of the abrupt onset may be visible in the pattern of eye micromovements made during fixation.³⁴ 

The present results also show a clear relationship between some microsaccade parameters and the success of prediction. This is documented by the fact that microsaccade rate and duration change from correct to incorrect responses (Figs. 4–6). Correct responses were characterized by lower microsaccade rate in the post-bounce epoch and shorter duration in the response period. Hence, the present findings indicate that the success of the prediction, especially in expert players, is associated with diminished values of microsaccade parameters. Some degree of interference between oculomotor activity and execution of a manual response has already been described. For instance, Pashler et al.³⁵ showed that saccadic latency is affected by a concomitant manual response choice. Moreover, Betta and Turatto³⁶ showed that microsaccadic activity depends on whether or not participants are preparing to manually respond to an upcoming visual stimulus. They suggest the existence of a link between the unconscious oculomotor activity, which is evident during fixation, and the cognitive processes involved in preparing to respond to an upcoming visual target. Our results demonstrate that the manual response interferes with the oculomotor fixational activity, and the success of the prediction enhances such interaction. 

Although the main sequence for the whole data set of our experiment is in line with the relationship between peak velocity and amplitude, the apparent discrepancy appears between amplitude and peak velocity in the pre-bounce and post-bounce epochs, where amplitudes and durations were highest, whereas peak velocities were lowest. A possible explanation for this result may be the high attentional load during these epochs, according to Di Stasi et al.,³⁷ who reported a decreased saccade peak velocity with the increased mental workload. Actually, in our task the pre-bounce and, even more, the post-bounce are the most attention demanding periods, and this could explain the amplitude/velocity dissociation. Durations that are longer than usual might be the effect of increased amplitudes. 

One final note concerns the comparison between microsaccade patterns of expert players and those of novices. Taking into account both correct and incorrect responses, experts made longer and wider microsaccades than those of novices. Given that microsaccade amplitude is related to the amplitude of attention shifts and that athletes are better able to use peripheral vision than novices,^27,38 the greater amplitude value found in the expert group could be explained by the fact that they are talented in making longer fixations and “anchor” the fovea close to relevant interest areas in order to use the parafovea and the retinal periphery to pick up relevant information. Microsaccades are relevant to human perception across the entire retina and can restore both foveal and peripheral vision in an analogous fashion.^16,17 

Adaptive behavior in many situations, such as driving a car or playing sports, requires the ability to continuously monitor multiple independently moving objects at different locations in the visual field. In this study, we tried to assess if and how microsaccades could be related to action perception in expert players. Experts, compared to novices, show better action anticipation in a variety of sports, such as soccer³⁹ and volleyball, and present results show the clear effect of expertise in table tennis athletes as well. It is well known that visual perception improves as a function of experience. Perceptual learning takes place during motion perception, when participants are exposed to motion stimuli without eye movements.⁴⁰ Athletes who play ball games are repeatedly exposed to motion stimuli during their training. Such exposure likely improves their perception of moving objects. Our experiment indicated that the distribution of microsaccade direction can be influenced by attentional cues in a task-specific situation. Microsaccade studies reveal links between visuomotor performance and covert attention shifts. The potential impact of these findings upon specific situations like sports could be that detailed assessment of visual performance may help to acknowledge potential elite skills and that vision training offers a means to further improve performance. 

Acknowledgments

The authors thank Andrea Giovanardi for technical assistance in the experiments. 

Supported by University of Bologna and Italian Ministry for University and Scientific Research. 

Disclosure: A. Piras, None; M. Raffi, None; I. Malagoli Lanzoni, None; M. Persiani, None; S. Squatrito, None