1. INTRODUCTION

Optokinetic nystagmus (OKN) is induced when a continuously moving stimulus appears in a visual field [1]. OKN consists of a slow phase (pursuit movements in the direction of the stimulus motion) and a fast phase (saccadic return movements in the opposite direction to the stimulus motion), and serves to stabilize the image of a moving stimulus on the retina. When two visual patterns moving either in opposite directions or with different velocities are superimposed, it is perceived as two transparent surfaces sliding over each other [2–5]. This phenomenon is called transparent motion perception. Additionally, the direction of the slow phase of OKN corresponds to the direction of motion that dominates perceptual appearance [6–8]. For example, the slow phase of OKN is directed to the left when observers report a leftward moving percept to be dominant. Some studies have shown that OKN is modulated by visual attention in transparent motion. The motion to which the observer attends elicits OKN corresponding to the attended motion direction when two patterns moving in different directions are superimposed on the same depth plane [7], and when a motion stimulus containing multiple motion areas with different velocities is presented [9]. Hence, OKN transitions corresponding to perceptual and attentional switches will occur.

It has also been shown that pupil size is modulated when an individual’s perception switches between stimuli with different characteristics [10–14]. Pupil size increases just before the reported perceptual switch of bistable illusions such as a Necker cube and a plaid stimulus [10]. Similarly, pupil dilation can be elicited by changes in perceptual content (i.e., the subjective disappearance and reappearance of a physically constant visual target in a motion-induced blindness illusion) and surprise in response to the perceptual switch [11]. Additionally, pupil size is modulated by the luminance of either the spatially attended location [15–19] or the attended object [20]. When observers attend to one of two surfaces consisting of either bright or dark dots moving in different directions, presented at the same or different locations, pupil size will be smaller when the observer’s attention is directed toward the surface with bright dots than when directed toward the surface with dark dots [20]. These findings suggest that pupil size is modulated when observers switch their perception.

Furthermore, it has been shown that spatial attentional shifts are associated with saccadic eye movements [21,22] and modulated pupil size [23–25]. One previous study combined these effects and showed that pupil response corresponded to the brightness of the background of a saccadic target before saccade initiation, regardless of whether the target background’s level of brightness changed at the timing of the saccade initiation [23]. Furthermore, pupil size changes in accordance with the level of brightness in the location to which attention shifts were related to the fast-phase direction of OKN produced by unidirectional field motion [24]. How the switch of object-based attention associated with OKN relates to modulating pupil size in transparent motion reminds unclear, but object-based attention [26,27] is associated with perceptual switching in motion transparency [28]. Previous studies [23–25] have shown that pupil size changes immediately after attentional shifts associated with saccadic eye movements; therefore, the pupil may change immediately after an OKN transition in transparent motion. If such a close relationship between OKN and pupil response can be found, it would suggest a robust connection between the object-based attentional mechanism of OKN and that mechanism of pupil modulation.

 

Fig. 1. Time course of stimulus presentation in a trial. After the stationary stimulus presentation (middle panel), the white and black dots started moving in opposite directions (right panel). During the motion stimulus presentation, the observer was instructed to report the dominant percept by pressing one of two buttons (left or right) immediately after a perceptual transition occurred. The size of the dots in these images is different from those in the actual stimuli to make this stimulus easier to see.

Download Full Size | PDF

This study aimed to investigate whether changes in pupil size correspond to the luminance of perceptually dominated objects related to the transition of slow-phase direction in OKN following objects in transparent motion. In this experiment, two random dot patterns, each consisting of black and white elements and moving in opposite directions, were superimposed. During the motion stimulus presentation, observers were instructed to immediately press a button to indicate which surface—black or white—they perceived to be dominant. OKN and pupil responses before and after the button press were analyzed. The button press used to report perceptual switches could elicit pupil dilation [11,12] and was presumed to delay from the exact timing of observer’s perception because the observer responded after the perception occurs. Therefore, we performed an objective analysis based on the slow-phase velocity of OKN and pupil response [13,14] because the slow-phase velocity of OKN reflects the perceptually dominant direction of a motion stimulus [6–8]. It was expected that, after OKN transition, the observer’s pupil would respond to the brightness of the random dot pattern corresponding to the direction of the slow phase of OKN. For example, if an observer’s eye-tracking changes from black dots on the left to white dots on the right, the observer’s pupils may then change from dilated to constricted.

2. METHODS

A. Observers

Twenty-eight volunteers (all female with a mean age of 20.0 years and an age range between 18–24 years) naïve to the purpose of the study participated in this experiment. All participants previously had experience in other psychophysical experiments. They had self-reported normal or corrected-to-normal visual acuity, no diagnosed neurological conditions, and provided written informed consent before participating. The study was approved by the Tokyo Institute of Technology Epidemiological Research Ethics Committee and was conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki).

B. Apparatus

Stimuli were presented on a 53.34 cm CRT (Sony GDM F500R, ${{1400}} \times {{1056}}$ pixels, ${35.1} \times {28.9}^\circ$, ${{36}} \times {29.4}\;{\rm{cm}}$) with a refresh rate of 60 Hz. The display was specially designed for precise manipulation of luminance. The luminance of the display was measured with a luminance meter (LS-110, Konica Minolta Holdings, Inc.), while a grayscale circle with a diameter of 2° was presented in increments of 10 RGB values from 0−255 (with the final value of 255 following 250). The luminance of the stimulus was estimated from the lookup table to relate luminance and RGB values based on the measurement. The display’s maximum and minimum luminance were 72.43 and ${0.01}\;{\rm{cd}}/{{\rm{m}}^2}$, respectively. Participants observed the stimuli in a dark room, and a chin rest was used to keep each participant’s head in a fixed position. The viewing distance was 57 cm. Stimuli were produced and presented using a MacBook Pro (Apple) and MATLAB (MathWorks, Inc.), with the Psychophysics Toolbox [29,30] and EyeLink Toolbox [31].

C. Stimuli

The motion stimuli consisted of two groups of randomly positioned black and white dots moving in opposite directions (Fig. 1, right image). The size of each individual dot was 0.25 deg. The luminance of the black and white dots was 0.01 and ${72.43}\;{\rm{cd}}/{{\rm{m}}^2}$, respectively. Dot density for each group was ${3.9}\;{\rm{dot}}/{{\rm{deg}}^2}$. The dots moved with a velocity of 10.5 deg/s (7 pixels/frame). The black dots moved from left (right) and the white dots moved from right (left). When a white dot and a black dot overlapped, the color alternated between white and black for each frame. The size of the motion stimulus was ${35.1} \times {28.9}\;{\deg}$, and the luminance of the background was ${2.36}\;{\rm{cd}}/{{\rm{m}}^2}$. Figure 1 depicts the time course of stimulus presentation in each trial.

 

Fig. 2. Time course of (a) horizontal eye position, (b) horizontal eye velocity, and (c) slow-phase velocity of OKN in one trial for one observer. A positive value on the ordinate in (a) indicates that the eye position is on the right side of the display, and 0 indicates that the eye position is at the center of the display. The red points in (c) indicate objective OKN transitions.

Download Full Size | PDF

D. Procedure

Each trial began when the observer pressed a button. Immediately afterward, a red fixation point (0.37 deg) at the center of the display and stationary black and white dots were presented for 4 s. During this stationary stimulus presentation, the observer was instructed to stare at the fixation point. Then, the fixation point disappeared, and the dots moved for 30 s. During the motion stimulus presentation, observers were instructed to stare at the motion stimulus at the center of the display but not follow any individual element (stare OKN [32]). They were further instructed to report the dominant percept by pressing one of two buttons (left or right) only when the perception was switched and had not continuously pressed the key to report their dominant percept. They were told not to press either button if the dominant percept was ambiguous. The observers could take a break for as long as they wanted between trials, but could not press the button again to launch the next trial for at least 5 s after the motion stimulus ended. The observers performed a practice trial before the actual experiment.

In two consecutive trials, two different motion direction conditions (i.e., black dots to the left and white dots to the right, or vice-versa) were presented in a random order. Each observer completed a total of 10 trials. The 5 min timeframe on the date of recording for each participant is mostly the same as those in the previous studies [7,10,13].

E. Analysis

Each observer’s right eye position and pupil size were measured during the trials with an infrared eye-tracker camera (EyeLink 1000 Desktop Mount, SR Research, Ltd.) with a sampling rate of 1000 Hz. A nine-point eye-tracker calibration was performed at the beginning of the experiment. Periods of blinking were detected using the manufacturer’s standard algorithms with the default settings. For each blink, the data from 100 ms before the beginning of the blink to 100 ms after the end were treated as missing data. Additionally, trials were excluded if more than 35% of the data within the trial were missing. Otherwise, missing data were interpolated with a cubic spline fit. Pupil size was measured as arbitrary units (an area of scaled image pixels). Furthermore, default pupil size and pupil response varied among observers. The pupil size for each trial (${x_i}$) was subtracted by its mean (${\mu _i}$) and divided by its standard deviation (${\sigma _i}$) to normalize to $z$ scores (${z_i}$) so that the mean was 0 and the standard deviation was 1 (${z_i} = ({x_i} - {\mu _i})/{\sigma _i}$) [10,13,14]. Pupil change velocity was calculated by taking the derivative of the normalized pupil size; pupil acceleration was calculated by taking the derivative of velocity. Pupil change velocity and acceleration were smoothed using a Gaussian-weighted moving average filter with a window length of 300 ms. The peak timing of the acceleration in a trace was used to define the onset of pupil constriction or dilatation [33].

Two analysis methods were used: (1) an analysis based on the zero-crossing points of the slow-phase velocity of OKN (objective analysis) [13,14], and (2) an analysis based on the observer’s button press (subjective analysis).

1. Analysis Based on Slow-Phase Velocity of OKN

It was presumed that OKN corresponding to a specific dots pattern did not occur before the button was pressed, since the observers were instructed not to report when the dominant percept was ambiguous. Therefore, the zero-crossing point and magnitude of the slow-phase velocity of OKN were used as the criterion to identify the period of stable OKN and reliably define the point of OKN transition.

Figure 2 depicts a schematic illustration of the objective analysis. The panels of Fig. 2 present: (a) the horizontal eye position, (b) the horizontal eye velocity, and (c) the slow-phase velocity of OKN in one trial of one observer. The ordinate of Figs. 2(a)–2(c) presents horizontal eye position (deg), horizontal eye velocity (deg/s), and slow-phase velocity of OKN (deg/s), respectively. A positive value on the ordinate in Fig. 2(a) indicates an eye position on the right side of the display, and 0 indicates an eye position at the center of the display. The abscissa represents the time (in seconds) from the stimulus onset. Horizontal eye velocity as a function of time was calculated by the derivative of the horizontal eye position [Fig. 2(b)]. To obtain the slow phase of OKN, the fast phase of OKN and saccade components were excluded from the eye velocity trace, based on velocity criterion (${\gt}{{25}}\;{\rm{deg/s}}$). After this exclusion, the data were smoothed with a Gaussian-weighted moving average filter with a time window of 700 ms [Fig. 2(c)]. The zero-crossing points of the slow-phase velocity of OKN were defined as objective OKN transitions [Fig. 2(c), red points]. If the absolute value of the first peak velocity before and after the zero-crossing point was less than 1.0 deg/s, it was excluded from the analysis. If each pair of velocities of the first and second peaks before and after a zero-crossing had the same signs, the zero-crossing was registered as a stable OKN. The data were analyzed using the Benjamini–Hochberg method controlling the false discovery rate (FDR) [34] based on Einhäuser et al. [10], which is not dependent on the sampling rate.

The time interval from one zero-crossing point to the next was calculated based on the above analysis. The results showed that the minimum interval between zero-crossing points was 2.12 s. Therefore, the data on the eye position and pupil size from 2 s before to 2 s after the zero-crossing point were extracted not to include other OKN transition points in the interval.

2. Analysis Based on Observer Button Press

For the subjective analysis, the data of the eye position and pupil size from 2 s before to 2 s after the observer’s button press were extracted. The data including another button press from 2 s before to 2 s after a button press were excluded from the analysis.

3. RESULTS

A. Slow-Phase Velocity of OKN

Figure 3(a) presents the averaged slow-phase velocity of OKN around the timing of the perceptual transition of the motion direction. The ordinate represents the slow-phase velocity of OKN (deg/s). The abscissa represents the time around the perceptual transition, indicated by the observer’s button press, which is zero on the abscissa. The solid yellow line shows the mean value when the dominant perceptual direction of the dots switched from left to right, and the solid green line shows the mean when the dominant perceptual direction of the dots switched from right to left. The color-filled areas show the standard error range among the participants’ mean values. The horizontal black lines indicate time points where both traces are significantly different at an expected FDR of 0.05 ($t$-test, $p \lt {p_{{\rm FDR} = 0.05}}$, a threshold of ${p_{{\rm thresh, FDR} = 0.05}} = {0.044}$); these were all timings in the range of ${-}{{2000}}$ to 2000 ms, except in the range of ${-}{{1725}}$ to ${-}{{1314}}\;{\rm{ms}}$. In Fig. 3, “n” indicates the number of data.

 

Fig. 3. (a) Time course of the slow-phase velocity of the OKN based on the observer’s button press and (b) that relative to the point of OKN transition averaged across observers. A positive value on the ordinate indicates an eye velocity is directed to the right side of the display. The solid lines indicate the mean and the color-filled areas indicate the standard error range among the participants’ mean values. The horizontal black lines indicate the timing at which both traces are significantly different at an expected FDR of 0.05.

Download Full Size | PDF

 

Fig. 4. (a) Time course of average pupil size across observers, aligned to the timing of reported perceptual transition as shown to be 0 on the abscissa, and (b) time course of average pupil size across observers aligned to the timing of objective OKN transition. The solid lines indicate the mean and the color-filled areas indicate the standard error range among the participants’ mean values. The horizontal black lines at the top of each panel indicate the time at which both traces were significantly different, at an expected FDR of 0.05.

Download Full Size | PDF

The slow-phase velocity of OKN changed from negative to positive, or vice-versa, more than 1.3 s before the timing of perceptual transition of the dominant motion direction. The zero-crossing points of the yellow and green lines were ${-}{{636}}$ and ${-}{{1398}}\;{\rm{ms}}$, respectively. These results indicated that the eye movement directional change preceded observers’ perceptual transition.

Figure 3(b) presents the average slow-phase velocity of OKN around the timing of the zero-crossing point, defined as the objective OKN transition, which is zero on the abscissa. The ordinate represents the slow-phase velocity of OKN (deg/s), and the abscissa represents the time around the OKN transition. The solid yellow line shows the mean value when the direction of the slow-phase velocity of OKN switched from left to right, and the solid green line shows the mean value when it switched from right to left. The horizontal black lines indicate time points at which both traces were significantly different at an expected FDR of 0.05 ($t$-test, $p \lt {p_{{\rm FDR} = 0.05}}$, a threshold of ${p_{{\rm thresh,FDR} = 0.05}} = {0.049}$); these were all times in the range of ${-}{{2000}}$ to 2000 ms except around zero, which was in the range of ${-}{{1}}$ to 1 ms.

As shown in Fig. 3(b), the signs of the slow-phase velocity of OKN in both traces were clearly reversed before and after the zero-crossing points. The magnitudes of slow-phase velocity of OKN were almost constant, indicating that OKN was stable before and after OKN transitions. The shapes of Figs. 3(a) and 3(b) are different because they reflect the results based on the data from 2 s before to 2 s after the observer’s button press [Fig. 3(a)] and the results based on the data showing that stable OKN occurred before and after OKN transition [Fig. 3(b)] as the definition.

B. Pupil Response at Perceptual and OKN Transition

Figure 4(a) presents average pupil size around the timing of reported perceptual transition, respectively. These values were normalized to $z$ scores. The abscissa represents the time relative to the observer’s button press, which is zero and indicates the timing of perceptual transition. The solid red line shows the mean data for cases in which perceptual dominance switched from black dots to white dots, and the solid blue line shows data for the cases in which perceptual dominance switched from white dots to black dots. The color-filled areas show the range within the standard error among the participants’ mean values. There were no significant differences in pupil size between the traces. In Fig. 4, “n” indicates the number of data.

As the red and blue lines in Fig. 4(a) show, pupil size started to increase from about 400 ms before perceptual transition (at ${-}{{371}}\;{\rm{ms}}$ for the red line and ${-}{{395}}\;{\rm{ms}}$ for the blue line). As expected, pupil change corresponded to the luminance of the perceptually dominant pattern after—or even before—the observer’s perceptual switch. The pupil size after or about 500 ms before an observer’s perceptual switch from the black dots to the white dots [Fig. 4(a), solid red line] was smaller than after an observer’s perceptual switch from the white dots to the black dots [Fig. 4(a), solid blue line].

Figure 4(b) presents average pupil size around the timing of the zero-crossing point, defined as the objective OKN transition. The horizontal black lines at the top of Fig. 4(b) indicate the time points where both traces were significantly different at an expected FDR of 0.05 ($t$-test, $p \lt {p_{{\rm FDR} = 0.05}}$, a threshold of ${p_{{\rm thresh,FDR} = 0.05}} = {0.013}$); these intervals were from ${-}{{1996}}$ to ${-}{{1792}}$, from ${-}{{127}}$ to 143, from 485 to 826, and from 1772 to 2000 ms.

Unlike the results based on observer button press, the pupil size changed immediately after OKN transitions. The pupil constriction started 209 ms after the OKN transition [Fig. 4(b), solid red line], and the pupil dilatation started 69 ms after the OKN transition [Fig. 4(b), solid blue line]. Then, the pupil constriction and dilatation repeated several times; however, the pupil size remained smaller for the white dots than for the black dots from 269 ms.

4. DISCUSSION

The findings of this study indicate that pupil transition is modulated with OKN transition in transparent motion. The analysis based on the slow phase of OKN showed that the timing of changes in pupil size in response to the luminance of stimuli corresponded to the transition of slow-phase velocity of OKN in response to the motion of one surface of transparent motion stimuli. More precisely, the pupil constriction began 209 ms after the direction of the slow phase of OKN changed from the motion direction of the black dots to that of the white dots [Fig. 4(b), solid red line], and the pupil dilatation started 69 ms after the direction of the slow phase of OKN changed from the white dots to the black dots [Fig. 4(b), solid blue line]. These results suggest a robust connection between the attentional modulation mechanisms of OKN and the pupil response.

Presumably, because the pupil responses related to depth were not observed, object-based attention rather than the depth-based attention [35] to decide the direction of OKN modulated the pupil response for the stimulus with two surfaces different in the element objects and depths used in this study. Previous research has reported that, when observing transparent motion stimulus, individuals perceive transparent surfaces as having different depths [2–5]. Pupil constriction occurs when a gaze position in the depth direction shifts from a distant object to a near object. This is consistent with the well-known connection between pupil, vergence, and accommodation responses, in which pupil constriction is generally accompanied by convergence eye movements and lens accommodation (i.e., the near response triad) [36,37].

We do not know how much depth observers perceived during stimulus presentation, as the present study did not measure vergence eye movements or the magnitude of perceived depth between surfaces. However, we assumed this would not occur for the following reasons. If the perceptual depth switches at the zero-crossing point of the slow-phase velocity of OKN, the pupil response should be either from constriction to dilation or from dilation to constriction regardless of the perceptually dominated objects, before and after OKN transition. However, the actual pupil response observed in the present study was from constriction [Fig. 4(b); solid red line, around ${-}{0.5}\;{\rm{s}}$] to constriction [Fig. 4(b); solid red line, around 0.7 s] or from dilation [Fig. 4(b); solid blue line, around ${-}{0.5}\;{\rm{s}}$] to dilation [Fig. 4(b); solid blue line, around 0.4 s]. Additionally, it has been previously shown that the spontaneous switch occurs even when observers keep attending to the surface at a particular depth [38]. Motion in the same direction as slow-tracking eye movements (smooth pursuit) is perceived as nearer [39,40] because objects that are more distant than the fixation point move on the retina in the same direction as the translating observer, while nearer objects move on the retina in the opposite direction as the translating observer. Therefore, it can be predicted that the perceptually dominant surface was always perceived to be near, and the pupil constricted when stable OKN occurred (i.e., the slow-phase velocity of OKN was high). However, no such pupil responses were observed in the present experiment. Furthermore, the latency of the pupil near response has been reported to be approximately 300–360 ms [41,42] when observers look at star-like targets printed in black on white presented at actual far and near distances longer than those used in the present study (69–209 ms). Additionally, the transient pupil responses in the present study were consistent with those in attentional shifts [23–25]. This also supports our claim that the present study’s findings were not due to the pupil near response. These results indicate that object-based attention modulated the observers’ pupil responses in the present study, not depth-based attention in transparent motion.

Presumably, pupil dilation could occur by pressing a button (motor response) [11,12] after pupil size changed in accordance with the luminance of the perceptually dominant pattern, immediately after OKN transition. The analysis based on observer response showed that the motor response started [Fig. 4(a); ${-}{{395}}$ to ${-}{{371}}\;{\rm{ms}}$] approximately 300 ms before the peak of slow-phase velocity of OKN [Fig. 3(a); ${-}{{105}}$ to ${-}{{70}}\;{\rm{ms}}$]. Therefore, the motor response was presumed to start 300 ms before slow-phase velocity saturation [Fig. 3(b); 878–1111 ms], after OKN transition. In fact, the analysis based on OKN transition showed that pupil dilation started around 578–811 ms [Fig. 4(b); solid red and blue lines]. Furthermore, Hupé et al. demonstrated that motor responses contribute to pupil dilation around ${-}{{300}}\;{\rm{ms}}$ to 1.5 s in relation to button presses. Pupil dilation in the motor response ended at approximately 1.5 s [Fig. 4(b); solid red and blue lines], which is consistent with the previous study. Thus, observers’ pupil were presumed to have changed corresponding to the dark pattern [Fig. 4(b); solid blue lines] and white pattern [Fig. 4(b); solid red lines]. In contrast, both pupil traces in the analysis based on observer response showed a similar change [Fig. 4(a)]. This is presumed to be due to the ambiguity of perception. An observer’s perception tends to be less stable or more ambiguous when OKN is disrupted [38]. Decreases in slow-phase velocity after a button press suggested that an observer’s perception was ambiguous as to which dot pattern was dominant. Therefore, clear pupil response to luminance is presumed not to have been observed in the analysis based on observer response.

Observers’ responses to perceptual transition were delayed by approximately 1.4–1.6 s from OKN transition [Fig. 3(a)]. This was also presumed to be attributable to the ambiguity of the perception around OKN transition. The OKN slow-phase velocity before the zero-crossing point was very low in the present study [Fig. 3(a); before ${-}{1.6}\;{\rm{s}}$]. This suggested that observers’ perceptions were ambiguous as to which dot pattern was dominant before OKN transition. Then, slow-phase velocity of OKN increased after the zero-crossing point, and the observers were presumed to respond after they became confident in their perception around the saturation of slow-phase velocity of OKN. Therefore, this suggests that observers’ responses, not perceptions, were actually delayed, because observers were instructed to be confident in their perceptions before responding. Moreover, it has been shown that the mean reaction time of the subject’s motor report (such as pressing a button with a finger) from a simple light stimulus to the timing of the perceptual establishment is approximately 190 ms [43]. Therefore, the delay of the observers’ response is presumed to include the motor component of the button press.

Presumably, the reason why the pupil changed immediately after OKN transition [Fig. 4(b), 69–209 ms] may be that the pupil was modulated by attentional shifts accompanied by OKN. It has been shown that attentional shifts occur before saccades and pupil responses to the brightness of the background of a saccadic target before saccade initiation [23]. Additionally, the minimum latency of pupil light reflex is approximately 220 ms in general [44]. Therefore, it is presumed that the attentional shift accompanied by OKN occurs about 100 ms (precisely, from 151 to 11 ms) before the OKN transition and modulated the pupil. It has been shown that the pupil light reflex occurs 100 ms earlier when saccade preparation is possible, relative to when it is not [23] and a covert attentional shift precedes saccade by approximately 100 ms [21]. Furthermore, the pupil latency due to attentional shifts to the fast-phase direction accompanied by OKN is 143 ms [24]. These findings are consistent with the results of the present study.

The intermediate layer of the superior colliculus (SCi) is one of the possible neural basis that modulates the pupil due to attentional shifts accompanied by OKN. The SCi encodes both stimulus salience and is relevant to coordinate orienting the shifts in gaze and spatial attention and pupil responses [45–47]. In addition, it has been shown that a monkey’s pupil size changes according to the luminance level at the spatial location corresponding to a microstimulated location in the SCi map in a physiological study of a monkey [25]. These studies suggest that the neural process to modulate pupil response by the attention shift accompanied by OKN may involve the SCi.

Although the latency of pupil constriction for light reflex is shorter than that of pupil dilation for dark reflex in general [44,48], the results of the present study show the opposite feature. This is presumed to be due to high-level processing such as the switching of perception or attentional direction. Previous studies investigating the pupillary response in perceptual switching [13,14] have shown that the response in pupil dilation was large and no clear response was observed in pupil constriction.

The present study suggests that there is a robust connection between the attentional mechanisms of OKN and pupil modulation. Additionally, this connection seems to be related to object-based attention. This finding indicates that attention associated with OKN not only shifts the direction of the motion stimulus but also modulates the observers’ pupil response, to enable efficient detection of future targets. We believe our findings contribute to the understanding of the relationship between the pupillary circuit and reflexive orienting responses toward targets such as OKN.