As a powerful social stimulus, eye contact communicates emotions and social intentions, such as approach-oriented motivations and general interests in the observer (Macrae et al., 2002; Adams and Kleck, 2005; Adams et al., 2006; George and Conty, 2008). Eye contact has been shown to capture attention more readily compared to averted gazes (Senju and Hasegawa, 2005; Vuilleumier et al., 2005; Conty et al., 2006), evoke positive affective responses (Bindemann et al., 2008; Willis et al., 2011; Chen et al., 2017), facilitate face recognition (Vuilleumier et al., 2005), positively affect attractiveness evaluation (Ewing et al., 2010), bias face preference (Jones et al., 2006), trigger self-referential processes (Conty et al., 2016; Hietanen and Hietanen, 2017), and eventually enhance interpersonal synchronization (Patterson, 1982). In most above-mentioned and many other laboratory studies, researchers often operationalize eye contact by presenting a direct gaze from a central face image to an observer. Although eye contact often functions as a signal facilitating social interaction, it can be also perceived as a threatening signal (Ellsworth et al., 1972; Emery, 2000) or has negative connotations. For instance, angry faces were rated less approachable and trustworthy when displaying direct gazes compared to averted gazes (Willis et al., 2011), and angry faces with direct gazes were categorized quicker and their emotion status judged to be more intense than those with averted gazes (Graham and LaBar, 2007; Sander et al., 2007; Bindemann et al., 2008).

Given the importance of eye contact, it is not surprising that eye contact can be processed at a subliminal level to modulate basic cognitive processes. For instance, subliminally presented eye contact facilitates visual awareness compared to averted gaze in interocular suppression, and is more capable of capturing attention as well as attracting eye movement than averted gaze (Stein et al., 2011; Chen and Yeh, 2012; Madipakkam et al., 2015). However, there is less investigation regarding whether and how eye contact modulates social cognition when the eye contact was presented without visual awareness. Given the functional diversity of eye contact, an intriguing question is whether the role of a subliminally presented eye contact would be different from that of a supraliminal eye contact. There has been evidence on different functions of eye contact with and without awareness in social decision making (Luo et al., 2016), but it is not clear whether such a dissociation is specific to that task, or can also be observed in other aspects of social cognition.

We explore this question by focusing on a fundamental aspect of social cognition: gaze-induced joint attention. Gaze-induced joint attention refers to the phenomenon that people automatically follow the averted gaze of others to align their attention (Driver et al., 1999; Nuku and Bekkering, 2008; Xu et al., 2011). As an essential non-verbal channel of interpersonal coordination, gaze-induced joint attention has been shown to emerge as early as in the first 3 months in infancy (Charman et al., 2000), and occurs automatically even if the averted gazes are task-irrelevant (Driver et al., 1999; Friesen et al., 2004; Frischen et al., 2007; Chen et al., 2017) or subliminally presented (e.g., Sato et al., 2007; Xu et al., 2011; Bailey et al., 2014; Mitsuda and Masaki, 2017; but see Al-Janabi and Finkbeiner, 2012 for evidence of the involvment of volitional control). Further, gaze-induced joint attention has been demonstrated to be specifically linked to the social relevance of gaze. It shows features distinct from that of non-social central attentional cues (Friesen et al., 2004), and it is modulated by social context such as the perceived intentionality of the gazer (Wiese et al., 2012, 2014; Wykowska et al., 2014). Also, gaze-induced joint attention recruits brain areas putatively involved in “Theory of Mind” (George and Conty, 2008), such as superior temporal sulcus and intraparietal sulcus (Vuilleumier et al., 2002; Calder et al., 2007) in addition to brain regions typically involved in attention-based orientation.

In the present study, we are interested in how this specific type of social attention might be affected if eye contact was presented prior to the averted gaze cues, and whether awareness of eye contact plays a moderating role. Supraliminal eye contact has been shown facilitating gaze-induced joint attention. Eye contact between two individuals can subsequently generate quicker attention orienting of one person following the gaze direction of the other, compared to when there is no such eye contact (Bristow et al., 2007). Also, gaze-induced joint attention following eye contact is accompanied by increased neural activity in regions critical in understanding communicative intent, such as medial prefrontal cortex (Schilbach et al., 2010; Caruana et al., 2015). Enhanced gaze-following behavior following eye contact is even found among 4–6 month-old infants (Farroni et al., 2003; Senju and Csibra, 2008). Given these findings, one might wonder whether such a potential impact of direct gaze may unconsciously occur, as eye-contact effects are thought to be partially triggered by subcortical regions (Senju and Johnson, 2009), which is less dependent on visual awareness.

To answer this question, we adopted a gaze-cueing paradigm, in which an averted gaze was used as an uninformative cue for a joint-attention task. Prior to the onset of an averted-gaze cue, either supraliminal (Experiment 1) or subliminal (Experiment 2) eye contact may occur. Following the literature (e.g., Kampe et al., 2003; Conty et al., 2006; Hietanen et al., 2008; Stein et al., 2011), we operationalized eye contact as a central face gazing directly at the observer. Experiment 1 extended previous research about eye contact’s facilitation effect on joint attention in that it examined the effect of task-irrelevant eye contact on subsequent joint attention. When Bristow et al. (2007) reported a facilitation effect of eye contact, their task was to judge the congruency of subsequent gaze shift in relation to a target. This task may have encouraged the participants to purposefully process gaze direction. In contrast, the present study rendered both eye contact and averted gaze cue task-irrelevant and therefore uninformative to participants, so as to examine whether and how eye contact affects subsequent joint attention automatically, without being intentionally processed. We expected to find the gaze-cueing effect, i.e., a shorter reaction time on detecting the location of a target following the congruent compared to the incongruent gaze cues. Furthermore, this gaze-cueing effect would be enhanced by the preceding supraliminal eye contact, despite of its task irrelevancy. Experiment 2 investigated the effect of subliminal eye contact on gaze-induced joint attention by rendering eye contact subjectively invisible with the technique of interocular suppression. We hypothesized that joint attention can also be modulated by subliminal eye contact.

In Experiment 1, participants viewed face stimuli containing averted gaze cues that were task-irrelevant, and reported the location of subsequent target stimuli. Half of the trials presented a face with direct gaze, which established an eye contact with the participant (the eye-contact condition); no such eye contact was established in the other half of the trials (the no-contact condition). Then the face displayed averted gaze, and participants were asked to report as quickly and accurately as possible the appearance of a target and its location, either left or right to the face. The direction of the averted gaze was either toward (in the congruent trials) or away from (in the incongruent trials) the target. The gaze direction was therefore independent of the location of the target and uninformative for participants to complete the task. Thus Experiment 1 adopted a 2 (Prior Eye Contact: Eye-contact vs. No-contact) × 2 (Gaze Congruency: Congruent vs. Incongruent) within-subject factorial design.

Prior research has demonstrated existence of subliminal processing of eye contact (Stein et al., 2011; Chen and Yeh, 2012; Luo et al., 2016; Chen et al., 2017). However, it is not clear whether and how subliminal eye contact affects subsequent joint attention. Using an interocular suppression technique termed Continuous Flash Suppression (CFS, Tsuchiya and Koch, 2005), Experiment 2 presented the first endeavor to investigate in particular the effect of subliminal eye contact on gaze-induced joint attention. In CFS, awareness of a stimulus presented to one eye is suppressed by a stream of noise images presented to the other eye. Compared to other paradigms such as masking and binocular rivalry, a remarkable advantage of CFS is that it induces robust and prolonged suppression, thus allows the suppressed stimuli to remain invisible for up to several seconds.

Using the same 2 × 2 within-subject factorial design (Gaze Congruency: congruent vs. incongruent; Prior Eye Contact: Eye-contact vs. No-contact) as in Experiment 1, Experiment 2 employed the same spatial-cueing paradigm, in which the presence of eye contact varies prior to a non-informative averted gaze cue. Unlike Experiment 1, the eye-contact face and the no-contact face were suppressed from participants’ awareness. The dependent variable was still the reaction time of target detection.

The present study reported dissociable effects of supraliminal and subliminal eye contact on gaze-induced joint attention: supraliminal eye contact increased subsequent gaze-cueing effect, but subliminal eye contact reduced it, compared with the corresponding no-contact conditions. The enhancement of the gaze-cueing effect following supraliminal eye contact is in line with previous literature (Bristow et al., 2007; Senju and Csibra, 2008), demonstrating the facilitation role of supraliminal eye contact in the alignment of social attention and interpersonal coordination (Patterson, 1982).

More importantly, Experiment 2 found reduced gaze-cueing effect following subliminally presented eye contact. Notably, the impact of subliminal eye contact was not a mere weakened version of that of supraliminal eye contact, neither did awareness simply amplify the effect of eye contact. Instead, the dissociation is in the directions of subliminal and supraliminal eye-contact effects. One speculation of ours is that subliminal eye contact may have social implications different from supraliminal eye contact. Existing research revealed that eye contact affects behavior in various ways, social facilitation being only one of it. For instance, eye contact can also be perceived as a threatening signal (Ellsworth et al., 1972; Emery, 2000). Subliminal eye contact evoked a larger negative potential at 200 – 300 ms in the partial-frontal area (Yokoyama et al., 2013), similar to subliminally presented fearful faces (Kiss and Eimer, 2008), suggesting that it probably conveys a negative rather than congenial social message. In line with this explanation, Luo et al. (2016) found that subliminally presented eye contact drove the individuals with a proself social value orientation to make less cooperative decisions compared to the no-contact condition, probably due to the ancient alarm reflex being evoked by the threatening implication of subliminal eye contact. Taking these findings, we speculated that the reduced gaze-cueing effect following subliminal eye contact might come from a mechanism similar to delayed attention disengagement induced by various threatening visual stimuli, including threatening words, emotionally threatening pictures, as well as supraliminal and subliminal fearful and angry faces (Fox et al., 2001; Yiend and Mathews, 2001; Carlson and Mujica-Parodi, 2015). In our study, subliminal eye contact might have functioned as a salient warning signal, thus similarly held attention and interfered with subsequent processes (such as the processing of averted gaze cues), and, consequently, reduced the following joint attention effect (the relative advantage of the congruent trials compared to the incongruent trials). If this speculation stands, our results would join existing evidence (Luo et al., 2016) in suggesting that the social meaning of eye contact may be contingent on visual awareness, and that the social-facilitation impact of eye contact may change remarkably in the absence of awareness. From this perspective, such a qualitative difference in the impact of eye contact at different levels of visual awareness may reflect the complexity of social cognition and the rich social implication of stimuli such as eye contact.

How can such awareness contingency of eye-contact effect exist? We speculated that visual awareness modulates the neural correlates of eye-contact processing. According to the fast-track modulator model (Senju and Johnson, 2009), eye contact is jointly processed by a subcortical pathway and a slow, deliberate cortical social brain network that is further subject to factors such as task demands and social contexts. Many brain regions in the cortical network are impacted by visual awareness. For instance, subliminal face stimuli elicited weaker neural activities in the prefrontal regions and the fusiform face area compared to supraliminal stimuli (Leopold and Logothetis, 1996; Tong et al., 1998; Jiang and He, 2006). In contrast, regions in the subcortical pathway have been demonstrated to respond to eye contact even in the absence of visual awareness (Madipakkam et al., 2015; Rothkirch et al., 2015). Thus, the effect of subliminal eye contact may mainly result from basic, reflective processing via the subcortical pathway, while the modulation from sociocultural norms may be weakened due to the limited access to certain cortical regions, as found in implicit social attitudes tasks (Greenwald et al., 1998; Nosek et al., 2002; Kiefer and Sekaquaptewa, 2007). On the other hand, the processing of supraliminal eye contact is subject to the modulation from top-down, context-specific inputs via the cortical network. As a result, different from the prosocial impact of supraliminal eye contact, a more reflexive warning function may manifest when the eye contact was subliminal, leading to a delay of attention disengagement or other types of fear response. In the present study this may lead to the reduction of subsequent gaze following response. This speculation can be further examined by directly testing the affective impact of subliminal eye contact or investigating the neural correlates of subliminal eye contact in the attention networks, the limbic system and other brain regions critically involved in affective processing.

The social account of the subliminal eye contact effect is but one potential explanation. The design of our study does not exclude alternative non-social explanations of our findings. For instance, in Experiment 1, it is possible that the gaze shift from the baseline face to the no-contact face might have distracted attention away from the central face, hence resulted in a smaller gaze cueing effect in this condition. Alternatively, the transitions from the baseline face to the eye-contact face or the no-contact face might be perceived as upward or downward biological motion, respectively, and the downward biological motion might be perceived as relatively subtle and may not facilitate the processing of subsequent averted gaze cue as much as the upward biological motion in the eye-contact condition. The subsequent gaze-cueing effect may have consequently reduced in the no-contact condition compared to the eye-contact condition. Further, in Experiment 2, the absence of visual awareness may compromise the perception of upward biological motion and affect the subsequent gaze-cueing effect. It may also be possible that the dissociations between upward and downward gaze shifts or biological motion are contingent on visual awareness or other aspects of the experimental setting. Exactly whether and how these factors may lead to the change of the direction of eye contact effect requires future research, for instance to investigate the neural activities in the attention networks and the cortical regions critically involved in gaze processing or the processing of biological motion while orthogonally manipulating the presence of gaze cues and visual awareness. Also, it is possible that inhibition of return (IOR) may have started to affect the gaze-cueing effect with the longer SOA in Experiment 2 (529.4 ms) than that in Experiment 1 (300 ms). Since IOR of centrally presented gaze cues was typically reported with much longer SOAs, (e.g., longer than 1 s, Frischen and Tipper, 2004; Frischen et al., 2007; McKee et al., 2007; Marotta et al., 2013; Hudson and Skarratt, 2016), future research is needed to explore whether IOR also affects responses as early as ∼500 ms, and further whether such an IOR effect can be moderated by eye contact and result in the reduced gaze-cueing effect compared to the no-contact condition.

Although there is also a study that failed to find evidence of the dissociation between subliminal and supraliminal eye contact (Chen et al., 2017), we speculate that it may be due to the specific research design with an affective priming task. The functional dissociation between supraliminal and subliminal eye contact may exist in certain, but not all, domains of social cognition. Taken together, research thus far indicates intricacy in the relation between visual awareness and social cognition that involves eye contact, and calls for future investigation to examine the functions of subliminal and supraliminal eye contact in specific contexts before generalizing findings from any single paradigm. Research in this direction might even shed light on the relation between awareness and sociality in general.

Our study extends previous research investigating the potential dissociation between explicit and implicit social cognition into the research of eye-contact processing (Ewing et al., 2010; Luo et al., 2016). Utilizing one of the most basic, yet essential non-verbal form of social cognition, joint attention, we illustrated how the impact of eye contact may be contingent on visual awareness. We speculated that this contingency is because subliminal and supraliminal eye contact has different affective or social implications, a direction future research should continue to explore.

This study was carried out in accordance with the recommendations of the Ethics Review Committee of Peking University with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethics review committee at the Peking University.

SX and HG proposed the ideas and designed the experiments. SX performed the experiments and analyzed the data. SX, SZ, and HG wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.