What Eyeblinks Reveal: Interpersonal Synchronization in Dyadic Interaction

This study examines eyeblink synchronization in interactions characterized by mutual gaze without task-related or conversational elements that can trigger similarities in visual, auditory, or cognitive processing. We developed a study design capable of isolating the role of gaze in human-human interaction and observed the blinking behavior of dyads with mobile eye tracking glasses under three conditions: face-to-face mutual gaze, mediated mutual gaze through a mirror, and self-directed gaze in a mirror. The results revealed that when the interaction was through direct mutual gaze, eyeblink synchronization increased concurrently with a more structured temporal pattern. Also, the sense of connection between partners mimicked the synchronization. These findings suggest that even minor deviations caused by mediated interaction lead to reduced synchronization and a weakened sense of connection among partners. The paper also discusses the need for methodologies to enhance the efficacy and authenticity of online environments and human-robot interaction.


INTRODUCTION
Blinking is one subtle yet powerful means of gaze-based interaction with another person.In interpersonal exchanges, between 48% to 71% of blinks are reported to be directly associated with communicative aspects [11].Also, even millisecond variations in blink duration can be perceived and impact communicative behavior in face-to-face interaction [16].Hence, blinks seem to be of importance for interactive exchange.However, in eye tracking research, blinks are often treated as erroneous signals and missing values.We developed a study design capable of investigating eyeblinks in mutual gazing in dyads of people.In our study, 20 dyads who did not know their partners either watched themselves in a mirror, another person in a mirror, or another person directly.While dyads exhibited synchronized eyeblinks, the synchronization was the highest in direct mutual gazing and the least when participants sat next to each other gazing at themselves.Astonishingly, the synchronized blinking behavior of dyads differed between direct mutual gazing and mutual gazing through a mirror, suggesting that mediated gaze contact disturbs physiological synchronization.

Mutual Gaze
Eyes, metaphorically the windows to our souls, have been a subject of eye tracking research in the last century.Considering humans' inherent sensitivity to the eyes of others, particularly to eye contact, from birth [10], social and affective aspects of mutual gaze should be investigated comprehensively.Empirical data show that mutual gaze plays an important role in contributing to socially appropriate interactions [1,29] and modulating affective elements such as empathy [23], agreeableness [6], affection and love [19], dominance [12] and emotional bonding [32] .Taken together, mutual gaze embodies a multifaceted and powerful role in social interactions.However, it is still unclear what exactly changes in the gaze, and thus, what can be read from the gaze, when two people directly look into each other's eyes.
It can be argued that maintaining ecological validity in eye tracking studies on mutual gaze is challenging: Some of the reasons behind that argument are that both the moment-to-moment feedback of gaze cues during eye contact and restrictive hardware used by eye trackers pose a challenge in ensuring well-controlled naturalistic experimental interactions [7].But, studying mutual gaze in real-time is now enabled by recent developments in eye tracking hardware which makes trackers hardly distinguishable from glasses, allowing for a relatively natural interaction between two persons.
We report the development of a paradigm allowing for the investigation of mutual gazing.To isolate gaze behavior from other physical exchanges, we compared behavior in three conditions: (A) when both partners were sitting next to each other in front of a mirror looking at their own eyes, (B) when both partners were sitting next to each other in front of the mirror while looking at their partner's eyes, and (C) when both partners were sitting across each other looking into their partner's eyes.

Blink Synchronization
Humans blink spontaneously every couple of seconds.The overall mean rate of spontaneous blinks per minute varies based on the task and is higher during conversation [4,9].Eyeblink synchronization occurs when individuals blink their eyes in a temporally coordinated manner.The subtle act of synchronized spontaneous blinks can be a powerful means of human interaction.In research focusing on conversation, interpersonal blink synchronization was shown to be an important indicator for evaluating cognitive and emotional processes such as interest level [26], excitement [28], engagement [30], and problem-solving performance [14] in collective situations as well as the relationship and interaction between people [24].Nevertheless, one has to bear in mind that the portion of mutual gazing differed within these studies.Taken together, blink rates seem to be a powerful measure to learn about the connection between two persons.
However, there are only a few studies investigating blinking behavior during mutual gazing, that is during times when the eyes of both partners meet.Research on the blinking behavior while watching videos of people speaking revealed that eyeblink synchronization between the listener and the speaker did not occur when only the face of the speaker was presented but not the sound, and synchronized eyeblinks were clustered at the end and during pauses of speech [25].In another study, participants exchanged synchronized blinks with the speaker only when they were interested in the content [26].Nevertheless, it must be taken into consideration that in the aforementioned findings, participants looked at videos.Hence, there was no real-time feedback regarding the other's gaze available for them.Gupta et al. explored respective effects in a more natural setting, where two people had one conversation face-to-face and one conversation back to back [13].They detected eyeblink synchronization in both task types, discussing that participants depended highly on auditory attention, and so did the observed synchronization.Hence, these findings question the impact of eye contact on the synchronization of blinks and suggest that verbal communication underlies the synchronization: bearing in mind that all respective studies used conversational tasks, and that language has a clear rhythm, these findings may not be so surprising.As auditory inputs modulate blinking patterns actively [15,17], the synchronization observed in these studies can plausibly be attributed to the similarities in attentional functions triggered by speech.That is, we cannot exclude that blinks synchronize with the language rather than with the other person.
One study offered a new insight by implementing real-time feedback on eyeblinks during a conversation between human listeners and an android speaker [33].According to the results of the study, humans' eyeblinks were synchronized to the android speaker's eyeblinks only during mutual gaze.While gaze aversion by the android disrupted the synchrony, touching the android's hand enhanced it.Although the study did not involve human dyads, these results suggest that factors that enhance rapport may override the effect of conversational patterns on synchronization.To our knowledge, there is only one study investigating synchronized blink rates in mutual gaze with two real partners, avoiding the videos and conversation [20].This study, however, investigated the blinking behavior of cats and dogs simultaneously with their owners and strangers, counting blinks from certain camera positions, without tracking the eyes.Nevertheless, these data suggest that direct gaze exchange between two familiar subjects increases blink synchronization even without language.

The Present Study
We developed a study design capable of isolating the role of gaze in human-human interaction.Thus, we tracked the eyes of two partners who did not know each other before.They had to look either at their own eyes or at their partner's eyes.To control for the effects of other body parts, both partners were sitting next to each other in front of a mirror.In addition, we realized a third condition in which partners were seated opposite to each other.In all three conditions of looking at their own eyes through a mirror, or looking at their partner's eyes through a mirror, or looking directly at their partner's eyes, their task was to not talk and hold the gaze for one minute.Besides monitoring blinking behavior during this time period, we also asked them to fill out questionnaires after the minute, reporting their own feelings, and how they experienced the partner.The overall aim of the present study was to compare the changes in blinking behavior and synchronization of dyads when the target of gaze changed.If eye contact is important for blink synchronization, then dyads would exhibit increased synchronization of eyeblinks during direct mutual gaze compared to mediated eye contact or self-gazing, with the possibility of this effect being influenced by the sense of connection between partners.If, however, synchronization of behavior including blinks is maintained mainly by other bodily interactions, then no difference between looking at their own or the other's eyes through the mirror is to be expected.

Study Design
To observe eyeblink events during dyadic mutual sustained gaze, we recruited two participants simultaneously to form a dyad.We measured the changes in eyeblink parameters and interpersonal synchronization in eyeblinks within dyads as well as the measures of self-perceived connectedness, dominance, arousal, and valence based on three different target types as experimental conditions.Thus, the independent variable was the target type which featured three levels: face-to-face mutual gaze (referred to as Dyad); mediated mutual gaze through a mirror (termed Dyad-Mirror); and self-gazing in a mirror (Self-Mirror) (see Figure 2).The experimental task for participants was to exchange mutual sustained gaze for 1 minute without any verbal communication in each condition.The experiment was designed as a within-subjects study and we employed a Latin square design to vary the order of conditions in order to avoid the effects of practice, fatigue, and context (see Figure 3).Each condition comprised three trials with a short break following each one.Each trial lasted 1 minute.The first trial served as the training trial to familiarize participants with the experimental procedure, and two more trials were run to ensure the highest data quality.Postcondition, participants filled out two questionnaires.The Self-Assessment Manikin (SAM) Scale [5] is a three-item pictographic scale that measures valence, arousal, and dominance.The Inclusion of Other in the Self (IOS) Scale [2] is a single-item scale that measures how close the respondent feels to another person or group.Fig. 2. Experimental conditions: There were always two participants in the room, sitting either opposite each other (dyad) or next to each other when looking at the mirror.When watching through the mirror, participants either had to look at the partner's eyes (dyad-mirror) or at their own eyes (self-mirror).

Apparatus and Experimental Setup
Eye movement data of pairs were acquired using two mobile eye tracking glasses Neon (Pupil Labs GmbH) simultaneously.The two eye cameras on the glasses record gaze data with a sampling rate of 200 Hz.The scene camera captures scene videos at 30 Hz with a resolution of 1600 x 1200 pixels.With Neon eye tracking glasses, no calibration is required before or during data acquisition.Neon is compliant with the General Data Protection Regulation (EU GDPR).
The experiment took place in an experimental room with no windows (size: 3.42 m x 2.16 m; height: 2.47 m) with constant artificial lighting (145 lx) (see Figure 3, top left).A white desk (size: 150 cm x 80 cm; height: 77,5 cm), a mirror (size: 119 cm x 48 cm), and two identical, movable, and height-adjustable chairs were placed in the room.There was a fixation cross on the wall over the mirror (see Figure 3).The seating height and position of the chairs were only adjustable with the help of the experimenter.The mirror was covered with a white sheet when not used in direct mutual gazing to avoid distraction and changes in luminance.In the dyad condition, participants were seated 100 cm across from each other.In the dyad-mirror and self-mirror conditions, to keep the subjective distance constant, participants were seated 50 cm from the mirror and 100cm from each other.

Procedure
Prior to the experiment, each participant was taken to a different spot in the laboratory where they could not interact with their partner.First, participants were presented with written information about the experimental content and procedure.The experimental procedures and the methods of data storage and use were in accordance with the Declaration of Helsinki and approved by the ethics committee board of Ulm University (263/23).After signing the consent form, participants were familiarized with the experimental procedure.All were naïve to the purpose of the study.They were informed that the purpose of the study was to investigate the movement of their eyes but they did not know that the data would be analyzed in the concept of synchronization of eyeblinks.After the briefing, participants were taken to the experimental room and were seated at the experimental table.The experimenter arranged the distance between the dyads as well as the distance between the mirror and the participants.Then, participants wore eye tracking glasses.The experimenter instructed them about the condition.In dyad and dyad-mirror conditions, the task was to exchange mutual sustained gaze with their partners.In self-mirror condition, the task was to maintain sustained self-gaze.Participants were asked to sit in a neutral manner and to avoid communicating verbally during the trials.Then, the experimenter left the experimental room, and participants attended to the fixation cross over the mirror until they heard an auditory signal from the speaker which counted down to the start of the trial (see Figure 3).When participants heard the start signal, they initiated the task.Each trial lasted one minute and ended with the ending signal from the speaker.After three consecutive trials, one condition was complete, and participants were taken out of the experimental room to different spots in the laboratory where they could not see or interact with their partners to fill out questionnaires.After the questionnaires, the next condition started.After the experiment, a participant folder was kept concerning their experience during the experiment to ensure there were no physical conditions affecting them.

Participants
As this study is the first to investigate synchronicity in eyeblinks during sustained dyadic mutual gazing, a priori power analysis for sample size estimation was not possible.To ensure enough observations, we recruited 20 dyads (all female, M age = 26.6,SD age = 3.97) in the study.All participants had normal or corrected-to-normal vision.As the familiarity between participants is anticipated to have effects on mutual gaze dynamics [6,8], we ensured that participants paired together were not acquainted with each other in any dyad.Given the dynamic and evolving nature of real-time gaze feedback, we anticipated that an individual's blinking behavior would exhibit two distinct patterns when interacting with two different partners.Thus, ensuring that potential confounding effects resulting from familiarity were avoided, participants were allowed to take part in the experiment twice, via partner rotation.Each participant was involved in partner rotation only once.In rotations, participants were paired with a new partner with whom they had not paired before, and each dyad carried out all three conditions.The number of unique individuals as participants was 20.Also, due to observed differences in mutual gaze dynamics between males and females [11,21,22,34], we exclusively recruited participants of a single sex, namely females.All participants volunteered to take part in the study.On-site, they were informed about the goals of the study and the details of the experiment both verbally and in writing.They were also informed that the data would be stored anonymously, and cannot be used to identify participants.It was clearly stated that they had the right to withdraw from the study at any time without giving an excuse.Participants gave their written consent to participate in the study.

Data Analysis
Blink events were automatically detected by the blink detection algorithmic pipeline (event-based recall 0.95; false discovery rate < 0.10) of the eye tracking apparatus.The pipeline removes blink events with a duration of less than 100 ms.Also, the software of the eye-tracking device visualizes the gaze and fixation data on scene camera videos.To ensure that participants exchanged sustained mutual gaze, a post-hoc visual inspection was done on the scene camera videos of each participant during each condition.Based on the inspection results, since it was found that sustained mutual gaze was not impaired in any dyads, there was no data excluded for that reason.To synchronize the data recording onset and ending timings between the partners of each dyad, we used the scene camera videos on the web-based eye tracking platform of the apparatus (namely, Pupil Cloud).We manually timestamped the first fixation after the gaze at the fixation cross as the onset and the end of the following 1 minute as the end of each recording.
Data from 20 dyads were used for the analysis of questionnaires.Data from 4 dyads were excluded from the analysis of eye tracking data because the eye tracker did not record at least one of the partners' eye movements, meaning that 16 dyads were included in the analysis of blink rate, blink duration, and interpersonal blink synchronization.The data from an additional 2 dyads were excluded from the analysis of asynchrony due to the absence of any synchronized blinks in self-mirror condition.Thus, the data from 14 dyads were included in the analysis of asynchrony.
The distribution of continuous variables of blink rate, blink duration, the percentage of synchronized blinks, and blink onset asynchrony for normality was tested with Shapiro-Wilk test.Blink duration, the percentage of synchronized blinks, and blink onset asynchrony data showed normal distribution.Based on this outcome, repeated measures analyses of variance (ANOVA) with Bonferroni correction method were performed to compare the effect of condition type on the means of blink duration, the percentage of synchronized blinks, and blink onset asynchrony.However, the blink rate data collected in self-mirror condition displayed non-normality, as indicated by the results of the Shapiro-Wilk test (p = 0.006), with histograms revealing a right-skewed distribution.
To stabilize variances and improve the symmetry of the distribution, a square root transformation was applied to the eyeblink rate measurements of all conditions to ensure consistency.After the transformation, the Shapiro-Wilk test effectively indicated no significant departure from normality (p> 0.05) for any condition.Subsequent repeated measures ANOVA analysis was conducted using the transformed data.To further examine the relationship between the blink rates and self-reported declarative ratings of dyads, Pearson's correlation analysis was performed.Statistical analyses were conducted on JASP Version 0.17 [18].

Analysis of Blink Synchronization.
To investigate the temporal relationship between the eyeblinks of partners in each dyad, we employed a 3-second analysis window [20,25], centered around the onset of blinks for the participant who blinked less (the "reference") compared to their partner (the "test").This window encompassed 1.50 seconds before and 1.50 seconds after the blink onset of the reference [20].We then compared the onset of blinks in the reference with those in the test.For every eyeblink in the reference, we sought the nearest corresponding eyeblink in the test and calculated the onset time lag (namely, asynchrony) between the two by subtracting the blink onset value of the reference from the test's [27].The intersubject asynchronies within the 3-s analysis window were recorded for each dyad across all experimental conditions.Repeated measures analyses of variance (ANOVA) were performed to compare the effect of condition type on the means of asynchrony values and the proportion of these blink events.For pairwise comparisons, post-hoc tests with Bonferroni correction method were used.

Analysis of Synchronized Blinks across Time Bins.
For each dyad, we constructed a blink frequency histogram with a bin width of 250 ms for each experimental condition by dividing the 3-s window into 12 equal bins [20,25,26].For each condition, we calculated the chance level by dividing the total of the median blink rates observed in each bin by 12 (the number of time windows).One sample t-test (CV = +1.753,t(15) > 1.753) was performed to compare the synchronized blink rates in each bin to the chance level.

Effects of the Type of Dyadic Interaction on Ratings of Connectedness, Valence, Arousal, and Dominance
We collected responses from 20 dyads to obtain data on their affective reaction to three conditions and contrasted the main effect of conditions on participants' ratings of connectedness, valence, arousal, and dominance (see Figure 4).Given that the questionnaires involve ordinal scales, we performed a nonparametric test.
Overall, whether dyads exchanged direct mutual gaze, mediated mutual gaze, or gazed at themselves in a mirror did not affect their perceived valence, arousal, and dominance.However, the ratings of how close they felt to the target was significantly higher in direct mutual gazing than when looking at their own eyes.
We next examined whether blink rates of dyads were correlated (see Figure 5).Results showed that there was a statistically significant positive correlation with a large magnitude between blink rates of dyads in dyad (r(30) = 0.56, p = 0.025) , and a marginally significant correlation in dyadmirror condition (r(30) = 0.49, p = 0.052).In contrast, no correlation was observed for the self-mirror condition (r(30) = 0.123, p = 0.651) .This means that while the blinking frequency of dyads was mutually effective in conditions where the gazing was reciprocal, it did not show the same pattern of mutual effect between dyads when they maintained self-gazing.

Interpersonal Blink Synchronization
To examine synchronized blinks in more detail, we calculated the percentage of blinks of each dyad that fell into a 3-s window to all blinks in each condition following the procedure used in [20,25].The mean percentage of synchronized blinks to all blinks was the highest (M = 50.8%,SD = 15.64,SE = 3.79) in dyad condition.It was 48.71% (SD = 13.67,SE = 3.31) and 40.93% (SD = 19.72,SE = 4.78) in dyad-mirror and self-mirror conditions respectively.The results of repeated measures analysis of variance (ANOVA) yielded a significant main effect of condition on the percentages of synchronized blinks between dyads (F(16,2) = 3.782, p = 0.034,  2 = 0.191).The following post-hoc tests revealed that the proportion of synchronized blinks compared to total blinks in the dyad condition was significantly more than in the self-mirror condition (t = 2.609, p = 0.041).The proportion of synchronized blinks did not differ between either dyad and dyad-mirror (t = 0.55, p = 1.00) or dyad-mirrror and self-mirror (t = 2.06, p = 0.144) conditions (see Figure 6a).These results indicate that while half of the blinks of dyads were synchronized during direct mutual gazing, it was significantly less when they maintained self-gazing.Also, the effect of the mirror decreased the proportion of synchronized eyeblinks; however, when contrasted with direct mutual gaze or self-gazing, the difference was not statistically significant.
Next, we investigated whether the time lag (asynchrony) between the onset time of synchronized blinks differed across conditions.The mean asynchrony between the onset time of synchronized blinks was 497.06 ms (SD = 165.64,SE = 44.27) in dyad condition, 595.446 ms (SD = 178.73,SE = 47.77) in dyad-mirror condition, and 620.882 ms (SD = 203.26,SE = 54.32) in self-mirror condition.Repeated measures analysis of variance (ANOVA) detected that the main effect of conditions did not yield statistically significant differences in the mean blink asynchrony (F(13,2) = 1.514, p = 0.239,  2 = 0.104) (see Figure 6b).A small asynchrony value means a short time lag between two corresponding blink events, and a higher asynchrony value indicates that it took longer for the participant to synchronize their blinks to their partner's.Thus, the result suggests that when participants synchronized their blinks to the blink of their partner, the target they looked at did not alter the observed immediacy between the two sets of blinks significantly.

Temporal Structure of Synchronized Blinks
To examine the onset asynchrony between the synchronized blinks of dyads further, we explored the distribution of the mean rates of synchronized blinks across different time intervals with a bin width of 0.25 seconds within the 3-second window using the procedure employed in [20,25,26].The analysis showed that time lags after the onset of the reference blink varied within the 3-s time window across conditions.Although a synchronization peak appeared to occur in proximity to the reference blink (around 0 seconds) across all conditions, dyad condition revealed both an increase and a structured pattern while the distribution was scattered and random in both dyad-mirror and self-mirror conditions (see Figure 7).Next, we examined whether the variations in the distribution of synchronized blink occurrences were statistically significant across 0.25-s bins.Analysis of comparisons between the blink rates and the chance levels revealed that dyads exhibited higher synchronized blink rates at certain bins and the number of these bins was more in dyad condition.On the other hand, in dyad-mirror condition, the number of bins with a statistically significant increase in synchronized blink rates was less than in the dyad condition.There were statistically significant increases in synchronized blink rates at -0.25-0.00s (t( 15 Similarly, in the case of synchronized blink rates in self-mirror condition, the number of bins with a statistically significant increase in synchronized blink rates was less than dyad condition.There were statistically significant increases at -1.00-0.75s (t( 15 Taken together, these results show that direct mutual gaze displayed a stronger effect on interpersonal blink synchronization than mediated mutual gaze or self-gaze, by both increasing the number of bins with synchronized blinks and structuring them temporally.

Effects of the Type of Dyadic Interaction on Blink Duration
The mean blink duration of participants was 265.337 ms (SD = 57.06,SE = 10.09) in dyad condition, 266.577 ms (SD = 47.97,SE = 8.48) in dyad-mirror condition, and 250.708 ms (SD = 47.48,SE = 8.39) in self-mirror condition per minute.The results of repeated measures analysis of variance (ANOVA) yielded a significant main effect of condition on blink duration (F(31,2) = 4.27, p = 0.018,  2 = 0.12).Post-hoc tests revealed that the blink duration during self-mirror condition was significantly less than dyad-mirror condition (t = 2.63, p = 0.033).There were no significant differences either between dyad and dyad-mirror (t = -0.205,p = 1.00) or dyad and self-mirror (t = 2.421, p = 0.055) conditions.
Taken together with the results of mean blink rate across conditions, the fact that mean blink duration did not differ between direct and mediated mutual gaze validates that the observed eyeblink synchronization was not induced by an increase either in blink rate or blink duration in our study.

DISCUSSION
In the current study, gaze-based interaction between two persons was investigated.Based on the former findings, we concentrated the analyses on blinking behavior.
Importantly, mean blink rates did not exceed the standard spontaneous blink rate in any of the experimental conditions.Also, there were no statistically significant differences in the mean blink rates across the three viewing conditions of direct mutual gaze, mutual gaze through a mirror, and self-gazing in a mirror.Hence, the absolute amount of blinks shall not be regarded as a main factor responsible for any of the observed effects.Rather, our findings indicate that the observed blink synchronization in this study was induced by the temporal adaptation of spontaneous blinks between the dyads.Also of importance is the fact that blink synchronization was observed in a task completely without using language.Hence, the rhythm of the language cannot be regarded as the origin of blink synchronization in the current data set.
Questioning the role of another person sending immediate and direct feedback while exchanging mutual gaze, we compare the two mirror conditions: In both conditions, two partners were seated next to each other at the same distance, looking into a mirror.The only difference was whether each participant looked at their own eyes or at the eyes of the other person.Hence, any other physical interaction among the partners was constant.Both mirror conditions induced similar effects on the affective reactions to the interactions.It can be argued that, even when the gaze was not mutual, there might still be information received from the partner in the periphery.Since physiological factors such as breathing can reveal effects of the the presence of another person, self-gazing can create a similar affective space as gazing at someone.However, there were several differences between these two conditions: While the blink rates of dyads correlated positively when they looked at each other, this effect was not pronounced when they looked at themselves seated next to their partner.Also, the temporal pattern of synchronized eyeblinks appeared to be more scattered and random in self-gazing.Thus, despite the aforementioned presumptive shared affective space, we could still isolate the effects of another person's gaze on the blinking behavior of the partner.An important question to be raised herein concerns what kind of a role a mediator such as a mirror plays in interactions.This question should be investigated further.
Compared to past findings using videos as partners, the mirror condition in our study provided real-time gaze feedback for the partners.But still, mutually gazing at each other through a mirror might be different from direct mutual gaze.In fact, our study revealed several, sometimes minor, differences: The sense of connection to the partner and the eyeblink synchronization was more pronounced when the interaction was face-to-face.Conversely, when dyads looked at each other through the mirror, the synchronization became less structured and more random.
Hence, the question remains of what exactly makes the difference between direct and mediated mutual gaze.Importantly, in both conditions, there was immediate gaze feedback provided.Hence, when thinking about video-based online interaction introducing temporal delays, unfamiliar face sizes, and unfamiliar viewing angles, we might plausibly assume that such differences even increase.For now, we have to conclude that mediated gaze-based interaction differs from direct mutual gazing.How exactly these differences emerge should be a subject for further investigation.
Importantly, the subjective reports of how intense the other person was included in their own experience mimicked blink synchronization in direct mutual gaze.Also, there was a tendency to approach significant differences in the reported level of arousal between looking directly versus looking through a mirror.Hence, we propose that blink synchronization can serve as an indicator of the subjective experience of human-human interaction.We may thus suspect that it can be also used for evaluating the quality of technical interactive devices.To better understand human-human interaction, it would be important to also investigate the effects of pupil mimicry within our setup.But so far, current glasses-like eye trackers do not allow for reliable pupil tracking.
The current findings shed light not only upon human-human interaction, but also give rise to important factors in the interaction of humans with technical devices: It seems that already small deviations from direct interaction, such as using a mirror, disturb the synchronization among partners and thus hamper the inclusion of the other in the self, or perhaps in other words, the feeling of connectedness.
At this juncture, a need for the examination of the efficacy of online interpersonal communication, such as videoconferencing, becomes apparent.The primal issue of interaction in online environments is that, among many other confounds, online conferences lack true eye contact because of the position of the camera.Even though there are some mutual gaze support solutions available [31], their use by the general public is still very rare.
Additionally, as avatars and humanoids are becoming increasingly integrated into our lives, the effectiveness and authenticity of human interaction with avatars and humanoids should be the subject of comprehensive scientific investigation.There are already some valuable steps taken in that area, particularly on the effect of mutual gaze on human-robot interaction [3,11,33].Our study reveals the role of eyeblink synchronization as an indicator of rapport.Thus, we encourage further research to pay attention to blink activity, contributing to a deeper understanding of the relation between blinking behavior and the effectiveness as well as the authenticity of human interaction with avatars and humanoids.

Further Reflections
Building upon the insights gleaned from this research, several pathways for future investigation emerge.First of all, the observed relationship between direct mutual gaze and increased synchronization, as well as the enhanced feeling of connectedness, bears significance across a wide range of collaborative environments.For instance, these findings may inform our understanding of dyadic dynamics within musical ensembles, sports teams, and any scenario where harmonious collaboration and coordination between individuals is paramount and verbal communication is none or limited.Thus, the factors contributing to eyeblink synchronization and its relationship with collaborative performance and coordination warrant deeper exploration.Future research endeavors could explore the transferability of our findings to these diverse domains under varying task complexity and environmental conditions across different settings.Also, further exploration into potential moderating effects such as the role of familiarity and attraction in shaping interpersonal synchronization and rapport could enable a more nuanced understanding of human-human interactions.
However, while our findings provide valuable insights, it is important to address the potential limitations imposed by the sample size in our study for future work.While the results with medium to large effect sizes suggest that the manipulation of gaze targets has a meaningful impact on interpersonal eyeblink synchronization parameters, the small effect size of the eyeblink rate variance suggests that the possibility of Type II errors cannot be ruled out for that analysis.A relatively small sample size of 16 dyads as a result of the need to discard data from some participants might limit the generalizability of our findings.Future research with larger samples may ensure the generalizability, stability, and robustness of the observed effects together with more statistical power.
In addition, the findings of this study heighten the significance of data privacy, human rights, and data sovereignty.The involuntary nature of eyeblink synchronization can reveal intimate and sensitive aspects of interpersonal dynamics, including emotional states, rapport, and engagement.The findings also highlight broader societal risks such as the manipulation and social control of participants by predicting and influencing interpersonal dynamics through eyeblink synchronization.Thus, in such investigations, threats to participants' privacy and autonomy and the risk of exploitation of data must be eliminated by adhering to ethical guidelines and via robust data security measures.

Conclusion
The present paper investigates the effect of interpersonal dynamics on eyeblink synchronization during mutual gaze without having language communication.The findings show that in scenarios where people engage with another person, a temporal adjustment of eyeblinks occurs between them.More importantly, when the interaction is through direct mutual gaze, eyeblink synchronization increases in tandem with a more structured temporal pattern.Also, the perceived connection between partners mimics the synchronization.On the other hand, even minor deviations from direct mutual gazing, such as the introduction of a mediator like a mirror, result in less synchronization and a loss of sense of connection among partners.Mechanisms of respective effects as well as further gaze parameters involved are subject to future studies.

Fig. 5 .
Fig. 5. Correlation graphs representing associations between the mean blink rates of partners across conditions.Note: The shaded areas represent the estimated 95% confidence intervals.

Fig. 6 .
Fig. 6.Proportion of synchronized blinks and time lag (asynchrony) between blink onset of dyads.(a) The percentage of synchronized blinks which fell into 3-s time window.(b) The mean blink asynchrony (ms) between the onset time of blinks of partners.

Fig. 8 .
Fig. 8. Synchronized blink rates across 12 bins within the 3-second window.The red line represents the chance level."+" represents the mean.The asterisk denotes a statistically significant increase in synchronized blink rates above the chance level.*p ≤ 0.05; **p ≤ 0.01