Directional biases for blink adaptation in voluntary and reflexive eye blinks

Abstract

The oculomotor system is subject to noise, and adaptive processes compensate for consistent errors in gaze targeting. Recent evidence suggests that positional errors induced by eye blinks are also corrected by an adaptive process: When a fixation target is displaced during repeated blinks, subsequent blinks are accompanied by an automatic compensating eye movement anticipating the updated target location after the blink. Here, we further tested the extent of this “blink adaptation.” Participants were tasked to look at a white target dot on a black screen and encouraged to blink voluntarily, or air puffs were used to elicit reflexive blinks. In separate runs, the target was displaced by 0.7° in either of the four cardinal directions during blinks. Participants adapted to positional changes during blinks, i.e., the postblink gaze position was biased in the direction of the dot displacement. Adaptation occurred for both voluntary and reflexive blinks. However, adaptation was unequal across different adaptation directions: Horizontally, temporal displacements experienced larger adaptation than nasal displacements; vertically, downward displacements led to larger adaptation than upward displacements. Results paralleled anisotropies commonly found for saccade amplitudes, and thus it is likely that gaze corrections across eye blinks share general constraints of the oculomotor system with saccades.

Introduction

The oculomotor system is an error-prone system (Deubel, Wolf, & Hauske, 1984). Our saccade landings are often not accurate. Therefore, error correction in the form of corrective saccades helps to shift our gaze onto the target (Deubel, Wolf, & Hauske, 1982; Henson, 1978; Kapoula & Robinson, 1986). Other kinds of corrective eye movements, such as drifts and microsaccades also function to reorient our gaze during retinal slips (Collewijn & Kowler, 2008; Rolfs, 2009). 

Eye blinks also introduce instabilities to gaze maintenance that need to be corrected. Eye blinks are therefore often followed by (micro)saccades to correct for gaze instabilities introduced by the blink (Costela et al., 2014). Maus et al. (2017) showed that the oculomotor system also corrects for positional changes that occur during a blink predictively. In the experiment, a white dot was repeatedly displaced by either 0.5° or 1° rightward during a blink. Initially, the gaze positions after a blink landed closer to the dot's original position. As the experiment went on, observers became adapted to the positional changes, even when they could not see the change because it occurred when the eyes were closed. Maus et al. (2017) observed that the gaze positions after the blink gradually landed nearer to the dot's displaced position. This adaptation started to plateau after ∼35 blinks. When the observers were maximally adapted to the positional change, and the dot was not displaced, the gaze position after the blink showed an aftereffect, i.e., gaze position went to the anticipated dot position after the blink. This implied that the oculomotor system could predict positional changes occurring during eye blinks. Adaptation to the positional changes were extinguished quickly. Observers reported being unaware of the displacements throughout the experiment even though their oculomotor system clearly compensated for the displacements during blinks. 

This “blink adaptation” generally mirrors the phenomenon of saccadic adaptation: Saccades to targets that are displaced during a saccade become adapted (McLaughlin, 1967; Pélisson, Alahyane, Panouilleres, & Tilikete, 2010). Saccadic adaptation occurs for successive saccades of different sizes and in different directions (Rolfs, Knapen, & Cavanagh, 2010; Watanabe, Ogino, Nakamura, & Koizuka, 2003). Adaptation is direction-specific as adaptation to one direction does not transfer to another direction (Albano, 1996). The size of the displacement for inducing adaptation affects the amount of adaptation acquired (Watanabe, Noto, & Fuchs, 2000). In addition, backward displacements, where the target position jumps against the direction of the saccade, are better adapted than forward displacements (Panouillères et al., 2009). 

The superior colliculus issues commands to make saccades (Scudder, Kaneko, & Fuchs, 2002). These commands are transmitted to the burst, tonic, and omnipause neurons in the pons, medulla, and the reticular formation. These neurons innervate the motoneurons, which activate the ocular muscles (Fuchs, Kaneko, & Scudder, 1985), and influence saccade size and velocity (Scudder et al., 2002). Saccades take flight when ocular muscles apply a burst of mechanical force. A separate sustaining force, lasting up to 200 ms, is then applied to hold the eyes in place when the saccade ends (Robinson, 1964). Eye blinks and blink-related eye movements may rely on similar circuitry as saccades. Eyelid muscles receive bursts and tonic signals from the orbicularis oculi and levator palpebrae superioris motoneurons to initiate a blink, hold the eyelid shut, or raise the eyelid (Evinger, 1995; Fuchs, Becker, Ling, Langer, & Kaneko, 1992). Burst neurons in the superior colliculus, which issues saccade commands, are momentarily suppressed by reflexive blinks (Goossens & Van Opstal, 2000). Microstimulating the superior colliculus decreases blink amplitude (Basso, Powers, & Evinger, 1996). Activation of the omnipause neurons inhibits eye blinks (Mays & Morrisse, 1995). 

Anisotropies in saccade magnitude and velocities occur depending on the trajectory. When making saccades along a horizontal plane, the velocity and position of the abducting (temporally moving) eye is greater than the adducting (nasally moving) eye (Collewijn, Erkelens, & Steinman, 1988a). When saccades are made along a vertical axis, upward saccades tend to undershoot the target while downward saccades tend to overshoot the target (Collewijn, Erkelens, & Steinman, 1988b). Centripetal saccades (toward the primary orbital position) are faster than centrifugal saccades even if the velocities do not interfere with saccade accuracy (Pelisson & Prablanc, 1988). Since anisotropies occur in saccadic control, and the blink circuitry may share some of the same neural circuitry as the saccadic system, the constraints observed in saccades and saccadic adaptation may apply to blink-related eye movements and blink adaptation as well. 

There are three types of eye blinks: spontaneous, reflexive, and voluntary blinks. Spontaneous blinks occur at an average of 12–19 times per minute (Bentivoglio et al., 1997; Griepentrog & Lucarelli, 2011; Karson, 1983; Karson et al., 1981). These blinks are thought to be controlled by basal ganglia activity (Ponder & Kennedy, 1927) and are related to dopamine levels in the brain (Karson, 1983). Patients with Parkinson's disease showed decreased spontaneous blink rates (Agostino et al., 2008; Korošec, Zidar, Reits, Evinger, & Vanderwerf, 2006). Reflexive blinks are elicited by startling stimuli such as loud beeps (Silverstein, Graham, & Bohlin, 1981) or external stimulation (Evinger, Manning, & Sibony, 1991; Kaneko & Sakamoto, 1999). Reflexive blinks are triggered by the basal ganglia inhibition of the brainstem, although the exact circuity is not well-understood (Valls-Sole, 2012). Voluntary blinks are executed at will, by command, or when cued. Although different blink types have specific physiological characteristics (Kaneko & Sakamoto, 1999), the blink motion is suggested to be controlled by the same premotor neuron (Korošec et al., 2006). 

Maus et al. (2017) found adaptation for spontaneous blinks but did not test for other blink types. If reflexive blinks cannot reproduce the adaptation, then it is possible the adaptation is part of a unique component found in the spontaneous blink circuitry but not the reflexive blink circuitry. 

According to Hering's law, both the left and right eyes are yoked together (Hering, 1977). This arises because of the agonistic and antagonistic properties of the extraocular muscles. Thus, uniocular movements of one eye approximate those of the other eye. The yoked nature of both eyes is clear when making saccades; both eyes are moving in the same direction toward a target. However, eye movements during a blink are drastically different when compared to saccades. At the start of a blink, both eyes turn nasally inward while moving (Collewijn, Van Der Steen, & Steinman, 1985; Rottach, Das, Wohlgemuth, Zivotofsky, & Leigh, 1998). Other blink-related eye movements help to reset torsional eye movements (Khazali, Pomper, Smilgin, Bunjes, & Thier, 2016) and correct for fixational errors (Khazali, Pomper, & Thier, 2017). 

Since both eyes experience asymmetrical movements during a blink, we might expect anisotropies in blink adaptation in each eye due to the differences in blink-related eye movements. 

In the current study, we explore whether blink-related eye movements for dot displacements during a blink become equally adapted in all directions. We conducted two experiments in which a dot was consistently displaced during blinks in one of four directions during each run. In Experiment 1, we recorded eye movements of the right eye only (viewing remained binocular), while in Experiment 2, we recorded binocular eye movement data. If blinks share similar brain circuitry as other eye movements, we would expect the constraints observed in saccades to apply to the blink-related eye movements carrying blink adaptation as well. Therefore, we should expect similar anisotropies in adaptation magnitude for blink adaptation. 

We also investigated whether the type of blinks would affect adaptation. Air puffs were used to elicit reflexive blinks. We compared adaptation magnitudes between reflexive and voluntary blinks within the same participants. If blink adaptation is controlled by the general circuitry, which produces both kinds of blinks, we would expect adaptation to occur similarly for reflexive and voluntary blinks. If blink adaptation was observed in only one type of blinks, then this would suggest that adaptation depends on the circuitry specific to this type of blinks. 

We also examined whether blink-related eye movements would exhibit consistent blink adaptation in both eyes by recording binocular eye tracking data. Since blink-related eye movements are nonconjugate, we expect to find anisotropies in left and right eyes. 

Methods

Participants

Participants were recruited from Nanyang Technological University. Twenty participants (M = 21.50 years old, SD = 2.60, 15 females) were recruited for Experiment 1. All participants, except for one, were naïve to the purpose of the experiment. We avoided recruiting participants with corrective contact lenses to not exacerbate potential dry eye problems in lens wearers with the application of air puffs. A separate group of 12 participants (M = 21.25 years old, SD = 1.48, nine females) were recruited for Experiment 2. Seven participants had normal visual acuity; five wore corrective contact lenses. All participants gave informed consent prior to the start of the experiment. Upon completion, participants were either rewarded with course credits or up to S$10/hour. The experiment was approved by Nanyang Technological University's institutional review board (IRB). 

Apparatus and stimuli

The experiment was programmed using MATLAB (MathWorks, Natick, MA) and Psychtoolbox-3 (Kleiner et al., 2007). Visual stimuli were presented on a 20-in. CRT Monitor (Sun Microsystems Corp., Mountain View, CA) at a resolution of 1152 × 864 and refresh rate of 100 Hz. A chinrest was positioned 75 cm away from the monitor. The XY-coordinate gaze positions and pupil area were tracked using an EyeLink 1000+ eye tracker (SR Research, Ottawa, Canada). 

The stimulus was a white dot (diameter = 0.2°) displayed on a black background (black = 0.75 cd/m², white = 132 cd/m²) in both experiments (Figure 1). The dot was confined to an area 15.79° × 11.84° about the center of the screen. Whenever a blink was detected during the experiment, the dot was displaced in one of four directions (left, right, up, or down) by 0.7°, consistently throughout the duration of one block. A blink was defined as the eyelid fully covering the eye (pupil area = 0). In addition to displacements during blinks, the dot was randomly assigned to new positions within the confined area randomly after 2–6 blinks. This prevented the dot from moving out of the central area of the screen. 

Figure 1

 

Trial sequence for Experiment 1. The dot and displacement step are enlarged for illustration. Each blink was considered a trial. The dot is displaced by 0.7° in one of four directions during a spontaneous/voluntary blink, or a reflexive blink in response to an air puff in each block. In this example, the dot was displaced rightward. The dot was displaced during adaptation and top-up trials. The trial sequence for Experiment 2 is identical, but both eyes were recorded, and no air puffs were used.

In two separate conditions of Experiment 1, eye blinks were either spontaneous/voluntary, or reflexive in response to an air puff. Because participants were aware that frequent blinking was necessary to complete the experiment, and participants were encouraged to blink once every 3–4 s, most blinks might more accurately be characterized as voluntary eye blinks. However, spontaneous blinks might also still have occurred during the run. The dot was displaced for both spontaneous and voluntary blinks, and we can therefore not strictly distinguish between spontaneous and voluntary blinks. 

Reflexive eye blinks were elicited by an air puff delivered to the subject's left eye. The puffs were released using an electronically controlled valve connected to the stimulus PC using an Arduino Micro board (Scarmagno, Italy) and exited a compressor tank at 10–20 psi (Bolbecker et al., 2009; Gandhi & Bonadonna, 2005). Each puff lasted 50 ms and was delivered along a plastic tube positioned a centimeter away from the eye. The tube was directed to the area below the eye brow and above the eyelid. A pressure regulator was used to adjust the output air pressure to a comfortable level. Based on feedback from a pilot experiment, participants reported having the least number of adverse reactions (such as dry eyes) when using this location to trigger the reflexive blinks. Occasionally, a manual puff was triggered via a keypress by the experimenter, when participants did not reliably blink following air puffs. In blocks with reflexive blinks, only blinks triggered by air puffs were accompanied by displacements of the target. 

The apparatus and stimuli were identical for Experiment 2, but participants made only voluntary/spontaneous blinks, and the eye tracker recorded movements of both eyes. 

Procedure

In Experiment 1, participants were randomly assigned to a pair of displacement directions: Up & Right (N = 10, M = 21.60 years old, SD = 3.07, six females) or Down & Left (N = 10, M = 21.40 years old, SD = 2.01, nine females). They were told the task was to look at the dot on the monitor throughout the experiment. They were also briefed that the dot would move to a new location frequently (referring to the obvious location changes in-between blinks) but were not told about the position changes during blinks. After the briefing, the eye tracker was calibrated on the right eye using 9-point calibration and a validation procedure to assure tracker accuracy. The instructions were shown on the screen reminding the participant to follow the dot throughout the experiment. Participants pressed a button on the keyboard to start the experiment. 

There were two conditions in Experiment 1, and one condition in Experiment 2. In Experiment 1, we tested for spontaneous/voluntary blinks in one condition and reflexive blinks in the other condition. These conditions were run in separate blocks of the experiment. In Experiment 2, we only tested for voluntary blinks. Each blink was considered a trial. There were 115 trials per run. The dot was displaced in a single direction in each run. For Experiment 1, participants completed two runs of orthogonal displacement directions (i.e., up & right, or down & left) per condition, for a total of four runs. The pair of orthogonal directions were randomly assigned, and participants started with a random direction. The displacement directions for voluntary and reflexive blink conditions were the same within each participant. For Experiment 2, participants completed four runs, one for each of the four directions. 

Each run for Experiment 1 consisted of the following components: 10 baseline trials, 40 adaptation trials, and three sets of five post-adaptation trials and 15 top-up trials each, and a final set of five post-adaptation trials, for a total of 115 trials. In Experiment 2, the order was similar, consisting of 10 baseline trials, 60 adaptation trials and two sets of five post-adaptation and 15 top-up trials each, and a final set of five post-adaptation trials, for a total of 115 trials. The larger number of adaptation trials for Experiment 2 was to elicit stronger adaptation and allow for longer saturation of adaptation. 

Participants in Experiment 1 also completed a resting-state and training phase of the experiment. In the resting-state blink phase, participants were instructed to look at a black screen in a relaxed manner for three minutes. Blinks were detected, and blink intervals computed. A gamma distribution was fitted to the histogram of blink intervals and used to generate random interblink intervals for the reflexive blink condition (as in Experiment 2 in Maus et al., 2017). The random intervals were additionally limited to be between 1.5 and 6 s to deliver the air puffs. There were two reasons in limiting the air puff intervals: first, to reduce the runtime of the reflexive blinks condition such that participants made as little spontaneous blinks as possible while completing the experiment mostly with reflexive blinks, and secondly, to avoid participants not blinking reflexively to a second air puff after too short intervals (i.e., air puffs at < 1.5 s intervals). 

The purpose of the training was to adjust the precise location and strength of the air puffs and accustom participants to the puffs, so that they would not startle or blink excessively for multiple times. The training took an average of 35 minutes to complete including breaks. Training first started with air puffs delivered to the left eye once every three seconds for 90 seconds, while the experimenter adjusted the strength and location of the puff. Following that, we delivered air puffs once every 3 seconds for 3 minutes. Finally, participants were administered air puffs as in the actual experimental conditions: 50 ms air puffs delivered randomly at intervals between 1.5–6 seconds. The air puff interval followed the fitted gamma distribution. 

Upon training completion, participants proceeded with calibration, tracker validation, and the experiment. In the reflexive blink condition of Experiment 1, the dot was never displaced for voluntary/spontaneous blinks. Only blinks occurring up to 150 ms after the air puff were considered reflexive blinks. Otherwise, the blink was treated as a voluntary/spontaneous blink. When participants did not react to the air puff, we manually triggered the air puff several seconds later using keypress with the same intensity and puff duration, which tended to lead to a reflexive blink in most cases. 

Analysis

The analysis in both experiments mimicked those from a previous study (Maus et al., 2017). Blink onsets were defined as the first eye tracker sample where no pupil area was detected (area = 0). The average horizontal and vertical gaze positions -200 ms to -100 ms before each detected blink (pregaze) were defined as 0°. The variance of this window was also computed. The blink end was determined when the pupil area exceeded 80% of the pregaze pupil area. Tracker noise produces unstable estimates of gaze position immediately after a blink. We identified the first stable gaze position after the blink, when the standard deviation of gaze position values within a 10 ms-sliding window fell below 1.5 times the pregaze variance. The postblink position was computed by averaging gaze positions in a window from the start of the stable gaze position lasting 100 ms, or until the first saccadic eye movement after the blink, whichever was earlier. The postblink positions were manually inspected to ensure there were no false detections of saccades. 

We then calculated the adapted gaze position (postgaze) by subtracting the postblink position from the pregaze position. Single-trial data of one participant from Experiment 1 was visualized to demonstrate how the pregaze and postblink values were derived from the analysis (Figure 2). This participant completed upward and rightward displacements in both voluntary (Figure 2a–2d) and reflexive (Figure 2e–2h) conditions. 

Figure 2

 

Time courses of gaze positions across single blinks (thin blue lines) and means for the whole experiment (thick, dark blue lines) from one participant in Experiment 1. This participant completed the upward (a, b, e, f) and rightward (c, d, g, h) displacements in voluntary (a–d) and reflexive (e–h) blink conditions. Only blinks from the baseline and late adaptation phases are shown. Up and leftward are plotted down on the abscissa. Pupil area (normalized) is plotted in red.

Blinks were considered reflexive if the duration of the blink was below 100 ms. In addition, the latency between the puff and the blink must be shorter than 150 ms. This interval was derived from a pilot experiment. Outliers from each participant were discarded when the mean postgaze position ≥ 2° about the dot position and removed from the data. This ensured that the postgaze positions analyzed did not reflect trials where participants were accidentally not looking at the dot. 

To investigate whether adaptation occurred, we compared the post-gaze positions for the phases in a run. In each experiment, we compared the early adaptation blinks (blinks 11–20, the first 10 blinks with displacements), the late adaptation blinks (the last 10 blinks of the adaptation trials, e.g., blinks 41–50 in Experiment 1), and the combined post-adaptation blinks (e.g., blinks 51–55, 71–75, etc., in Experiment 1; 20 postadaptation trials in total from each experiment) to the baseline (blinks 1–10). 

We conducted a 4 × 4 analysis of variance (ANOVA) to determine whether there were significant differences for the various phases of a run for both experiments. The between-group variable was directions (i.e., up, down, left, and right), and the within-group variables were the phases (i.e., baseline, early, late, and post blinks). In all ANOVAs, we used the Greenhouse-Geisser correction for subsequent within-subjects analysis, if Mauchly's test of sphericity was violated (Mauchly, 1940). We then ran post hoc analyses using Bonferroni-corrected α to examine whether the postgaze position across the phase of each run was equivalent within each subject. 

In Experiment 1, we also ran a 2 × 4 repeated-measures ANOVA (blink type × directions). We analyzed the main effect of direction, blink types, and the interaction between direction and blink types. In Experiment 2, we performed a repeated-measures ANOVA to determine if the adaptation magnitudes were different for each eye across the same participant. We then performed planned comparison paired t tests to evaluate whether the phases were different from the baseline within each direction for blink types (Experiment 1), and for each eye (Experiment 2). The t tests were considered significant if the p value fell below the Bonferroni-corrected α of 0.05/3 = 0.017. We further calculated the adaptation gains for each direction by dividing the average gaze magnitude of the late blinks by the displacement magnitude of the dot (0.7°). 

Results

Seventeen (85%) participants in Experiment 1 and six (50%) participants in Experiment 2 reported being unaware of the dot displacements during blinks. In the voluntary blink condition in Experiment 1, the baseline phase showed no change of postblink gaze from the preblink gaze, as expected. Introducing the fixation step during a blink, however, introduced positive biases in the postblink gaze position toward the new fixation position, moderately in the early adaptation phase, and more prominently in the later adaptation phase (see Figure 3). This bias persisted after the adaptation phase. The time-course of postgaze positions after voluntary and reflexive blinks over the course of a run is shown in Figures 4 and 5. 

Figure 3

 

Average postblink gaze positions for each phase of Experiment 1 for (a) voluntary/spontaneous blinks, and (b) reflexive blinks. The colors of the bars represent the different displacement directions of the fixation dot during each blink in the adaptation phase (e.g. red = upward displacement). The gray dotted line shows the actual displacement of the target dot (0.7°). Error bars are standard errors of the mean. *** p < 0.001, ** p < 0.01, * p < 0.05. Blink adaptation occurs for all four directions in voluntary/spontaneous blinks; however, with widely differing adaptation magnitudes across different displacement directions. In reflexive blinks, adaptation was only significant for rightward displacements.

Figure 4

 

Time course of gaze positions for voluntary blinks in Experiment 1. Baseline: blinks #1–20. Black arrows depict the direction of the dot displacement: (a) Upward, (b) Downward, (c) Leftward, and (d) Rightward. Dot was displaced by 0.7°. Error bars represent SEM. Positive values in the figure reflect gaze position in the same direction as the displaced dot. Gaze position after a blink gradually shifted toward the dot position, showing adaptation. This effect was evident in most displacement directions.

Figure 5

 

Time course positions for reflexive blinks in Experiment 1. Blink adaptation in the reflexive blinks follows the same trend as in voluntary blinks (Figure 4), but is less pronounced.

Experimental phases were significantly different within each direction, F(3, 144) = 9.71, p = 7.06 × 10⁻⁶, η_p² = 0.20 (Figure 3a). We then conducted a planned comparison t test to determine whether there was blink adaptation for each displacement direction. Blink adaptation was observed if there were significant differences between late blinks and baseline blinks using Bonferroni-corrected α = 0.017. T-tests showed evidence of blink adaptation across all directions; Upward: t(9) = −3.70, SD = 0.28, p = 0.005, Gain = 21.01%, Downward: t(9) = −6.44, SD = 0.31, p = 0.0001, Gain = 122.94%, Leftward: t(9) = −3.07, SD = 0.29, p = 0.013, Gain = 28.85% and Rightward: t(9) = −8.03, SD = 0.19, p = 2.16 × 10⁻⁵, Gain = 82.96%. 

We found significant differences in the postblink gaze positions across the four directions, F(3, 144) = 8.17, p = 4.66 × 10⁻⁵, η_p² = 0.017. Post hoc tests showed that downward displacements were significantly larger than upward (p = 0.006), and rightward were significantly larger than leftward displacements (p_right = 0.003). In the repeated-measures ANOVA, there were no significant main effect between voluntary and reflexive blinks, F(1, 36) = 1.02, p = 0.32, η_p² = 0.03. There was also no interaction effect between directions and blink types. 

For Experiment 2, the baseline again showed no change of postgaze from the pre-gaze for both eyes. When the fixation stepped during a blink, there were positive biases in the postblink gaze position toward the new fixation position (see Figure 6). This bias persisted after the adaptation phase in each eye. For the left eye, there were significant differences for the phases of each run, F(3, 176) = 11.07, p = 1.07 × 10⁻⁶, η_p² = 0.19 (Figure 6a). The planned comparisons t tests showed blink adaptation for all directions; Upward: t(11) = −2.57, SD = 0.39, p = 0.03, Gain = 59.59%; Downward: t(11) = −6.07, SD = 0.54, p = 8.04 × 10⁻⁵, Gain = 106.29%; Leftward: t(11) = −5.23, SD = 0.16, p = 2.83 × 10⁻⁴, Gain = 48.76%; and Rightward: t(11) = −11.03, SD = 0.11, p = 2.77 × 10⁻⁷, Gain = 54.55%; however, the upward displacement direction did not survive Bonferroni-corrected α = 0.017. There were no significant differences for the direction of the dot displacement, F(3, 176) = 0.26, p = 0.86, η_p² = 0.004. The interaction between direction and the phase of each run was not significant. 

Figure 6

 

Mean gaze positions after a blink in Experiment 2 for (a) Left eye, and (b) Right eye. Both eyes show blink adaptation in all directions (Late vs. Base).

In the right eye (Figure 6b), the 4 × 4 ANOVA illustrated significant differences for the phase of each run, F(3, 176) = 9.61, p = 6.53 × 10⁻⁶, η_p² = 0.16. Planned comparisons t tests on the right eye data showed significant blink adaptation for all directions (Figure 6b); Upward: t(11) = −4.18, SD = 0.40, p = 0.002, Gain = 66.79%, Downward: t(11) = −3.53, SD = 0.50, p = 0.005, Gain = 80.47%, Leftward: t(11) = −9.37, SD = 0.12, p = 1.42 × 10⁻⁶, Gain = 34.51%, and Rightward: t(11) = −3.85, SD = 0.31, p = 0.003, Gain = 76.62%. The postgaze position was also significantly different for displacement directions, F(3, 176) = 3.36, p = 0.02, η_p² = 0.06. Post hoc tests with Bonferroni-corrected α illustrated that the leftward displacement direction was significantly different than the rightward displacement direction (p = 0.02). There was no difference between the upward and downward displacements. There was no significant interaction effect between the direction and the phase of each run. 

We then conducted a repeated-measures ANOVA to examine whether the adaptation magnitude differed across each eye. Mauchly's test of sphericity was violated; χ²(5) = 19.87, p = 0.001. Thus, the following analysis utilized Greenhouse-Geisser corrections. The ANOVA showed significant differences for the phase of the run across both eyes F(2.59, 227.57) = 71.58, p < 0.001, η² = 0.45. The interaction between eye and direction was not significant. 

Post hoc analyses with Bonferroni-corrected α indicated that the early (p < 0.0001), late (p < 0.0001), and post (p < 0.0001) blinks were significantly different from the baseline blinks in each eye. The postgaze positions of the left eye did not differ significantly from the postgaze positions of the right eye (p = 0.89). 

Discussion

When a dot steps repeatedly and consistently from its original position during a blink, the oculomotor system learns about this change. The gaze position after the end of each blink then gradually adapts over repeated blinks and is shifted toward the displaced position of the dot, even when we cannot see the displacement. We found evidence of this blink adaptation for all displacement directions tested. We further tested whether blink adaptation is elicited by reflexive as well as voluntary/spontaneous blinks. Although our data showed significant adaptation in reflexive blinks only when the dot stepped toward the right, the repeated-measures ANOVA did not show significant differences between voluntary and reflexive blinks. Finally, we measured blink adaptation in both eyes for four displacement directions. We found that adaptation occurred in both eyes, with some interesting anisotropies for temporal and nasal directions of adaptation. 

Adaptation magnitude is unequal across directions

Our findings add to the initial report on blink adaptation by Maus et al. (2017). We show here that blink adaptation occurs for dot displacements in any of the directions tested. However, adaptation magnitude is not isotropic; we found larger adaptation for rightward and downward displacements. 

The larger adaptation for downward directions may be explained by the nature of blink-related eye movements. During a blink, the eyes move inward nasally and downward concurrently, while the eyelids are closing (Bour, Aramideh, & De Visser, 2000; Collewijn et al., 1985). These eye movements may be different for blinks occurring at eccentric gaze positions, when the eyes move toward the primary gaze position and then back (Bour et al., 2000; Riggs, Kelly, Manning, & Moore, 1987). However, on average, participants looked straight ahead in our experiments. Most blinks were likely accompanied by this downward component, regardless of the kind of blinks (i.e., spontaneous, voluntary, or reflexive blinks). As the blink ends and the eyelids begin to open, the eyes return from the downward position to their previous position. Based on this blink-related eye movement, it is possible that the eyes do not fully return to the previous position for a downward displaced dot. Instead, the eyes are more likely to stop at the predicted dot position resulting in a larger adaptation gain observed. 

To measure reliable blink adaptation for upward displacements, more adaptation trials might be required. Maus et al. (2017) reported that adaptation for rightward displacements began to plateau after ∼35 blinks. We replicated a similar pattern in our rightward adaptation. However, the blink quantity required for adaptation to plateau was different across the four directions. In the time-course plots for upward displacements in voluntary blinks (Figure 4a), the gaze position after a blink never truly plateaued. This meant that the small adaptation magnitude for the upward displacement might be attributed to two factors, namely, (a) adaptation in the upward direction appears slower, i.e., it requires more adapting blinks to reach a plateau, or (b) the vertical component of blink-related eye movements reduces the effects of adaptation. In Experiment 2, where participants spent 60 blinks to induce adaptation compared to only 40 blinks in Experiment 1, we already see somewhat more robust blink adaptation in the upward direction. 

Our present eye gaze recording method relies on video-based eye trackers. Therefore, we cannot ascertain which phase of blink-related eye movements is altered by adaptation, since the tracker does not provide information about the orbit's rotation while the lid is closed. Further research using search coils (as, e.g., by Khazali et al., 2016; Khazali et al., 2017) might better elucidate how the eye is moving during an ongoing blink. 

Regardless of the explanations for the differences in upward and downward adaptation magnitude, it is essential to point out that similar differences are also observed for vertical saccades. Collewijn et al. (1988b) showed that when participants made vertical saccades between 10° and 70°, upward saccades tended to undershoot the target, whereas downward saccades tended to overshoot the target. We observed similar anisotropies for blink adaptation in the absence of saccades. These anisotropies seem to be general for many different cases of eye movements, hinting at similar neural circuitries controlling blinks and eye movements in different directions. 

Possible adaptation in reflexive blinks

The lack of significant blink adaptation for reflexive blinks does not suggest that blink adaptation is unique for voluntary/spontaneous blinks. The adaptation magnitude for both the voluntary and reflexive blinks in Experiment 1 showed a similar pattern throughout the phases of the experiment (Figure 3b). Furthermore, the repeated-measures ANOVA showed no significant differences between the two blink types, although adaptation in the reflexive blink condition was weaker. 

Weaker adaptation in the reflexive blink condition might be due to contamination by spontaneous blinks during the adaptation phase. Participants might have blinked spontaneously, not only when an air puff was delivered. Thus, reflexive blinks were interspersed with spontaneous blinks. However, only reflexive blinks triggered the adapting displacement step of the fixation target. Furthermore, participants may sometimes have blinked several times following a puff. Since these additional blinks were not paired with a target step in the same way as the reflexive blinks triggered by an air puff, adaptation may have been diluted by these additional blinks. Blink adaptation in Maus et al. (2017) was rather short-lived; the effect was extinguished after just four to five blinks without a target step. These factors help explain the lack of significant findings in the reflexive blink condition. 

Since adaptation in reflexive and spontaneous/voluntary blinks overall followed a similar pattern, it is likely that the adaptation occurs in the general neural circuitry related to the production of blinks, and not in mechanisms exclusive to either blink type. To further examine how blink adaptation differs between the types of eye blinks, a different method for eliciting reflexive blinks could be used. For instance, the orbicularis oculi nerve located above the eyebrows could be stimulated electrically to induce reflexive blinks (Kimura, Powers, & Van Allen, 1969; Shahani, 1970). 

Binocular differences in blink adaptation

The stark contrast between adaptation magnitudes in the horizontal direction in Experiment 1 suggests that blink adaptation favors changes of eye gaze in the temporal and not nasal direction. In Experiment 1 we recorded only the right eye and found a strong bias for rightward adaptation to reach larger magnitude than leftward adaptation. These results are consistent with past findings in which eye movements toward the temporal direction tend to be greater than movements in the nasal direction (Collewijn et al., 1988a). To further investigate this, we recorded binocular eye movements in Experiment 2. If blink adaptation were stronger for temporal directions, we would expect to find the opposite pattern in the left eye. This was, however, only partially supported. We found that the adaptation magnitude for both leftward and rightward displacements in the left eye were equivalent, whereas we replicated the anisotropy for the right eye. Nonetheless, left displacements led to larger adaptation magnitude in the left eye than in the right eye, and conversely right displacements led to larger adaptation magnitude in the right eye than in the left eye. 

Saccadic and blink adaptation

Our results shed light on possible parallels between blink adaptation and saccadic adaptation. Despite the lack of fast ballistic movements as in saccadic adaptation with large saccades, the oculomotor system can adapt to position changes during a blink. We find anisotropies for downward and upward steps during eye blinks, which may be related to well-known anisotropies for inward and outward steps during saccadic adaptation (Panouillères et al., 2008). Since blink-related eye movements rotate the eyeball down and then up, a downward target step may be equivalent to a shortening and an upward target step to a lengthening of the returning eye movement. However, blink-related eye movements are not saccades (Khazali et al., 2016; Khazali et al., 2017), and adaptation of saccades does not directly transfer to blinks or vice versa (Maus et al., 2017, experiment 6). Also, it is unclear whether a backward step relative to the eye movement returning to the preblink fixation, or a backward step relative to the initial eye movement should lead to “easier” adaptation. 

Some questions remain unanswered. For instance, saccadic adaptation occurs under “random-walk” sequences of saccades with different directions, as long as the target is displaced in the direction of the saccade (Rolfs et al., 2010). One could test whether random-walk positional changes during blinks would also elicit blink adaptation. Since adaptation can be transferred to other saccades within a certain spatial field (Collins, Doré-Mazars, & Lappe, 2007), another potential experiment would be to investigate whether such an adaptation field is also present for positional changes across eye blinks. 

Conclusion

The oculomotor system is error prone and corrections are made to help reorient our gaze. We investigated whether the oculomotor system also corrects for positional changes happening during blinks. Regardless of whether blinks were spontaneous, voluntary, or reflexive, we found evidence of anticipatory corrective eye movements during blinks when a dot was displaced in the cardinal directions. This suggested that the type of blink is irrelevant if an eye blink is made. Anisotropies in blink adaptation are consistent with well-known directional biases in saccades, suggesting that oculomotor constraints also apply to adapted blink-related eye movements. 

Acknowledgments

This work was supported by Nanyang Technological University Research Scholarship. 

Commercial relationships: none. 

Corresponding author: Wee K. Lau. 

Email: . 

Address: School of Social Sciences, Psychology Programme, Nanyang Technological University, Singapore.