Holistic processing of gaze cues during interocular suppression

Abstract

Direct eye contact is preferentially processed over averted gaze and has been shown to gain privileged access to conscious awareness during interocular suppression. This advantage might be driven by local features associated with direct gaze, such as the amount of visible sclera. Alternatively, a holistic representation of gaze direction, which depends on the integration of head and eye information, might drive the effects. Resolving this question is interesting because it speaks to whether the processing of higher-level social information in the visual system, such as facial characteristics that rely on holistic processing, is dependent on conscious awareness. The Wollaston Illusion is a visual illusion that allows researchers to manipulate perceived gaze direction while keeping local eye features constant. Here we used this illusion to elucidate the driving factor facilitating the direct gaze advantage during interocular suppression. Using continuous flash suppression, we rendered Wollaston faces with direct and averted gaze (initially) invisible. These faces conveyed different gaze directions but contained identical eye regions. Our results showed clear evidence for a direct gaze advantage with Wollaston faces, indicating that holistic representations of gaze direction may drive the direct gaze advantage during interocular suppression.

Figure 1

The Wollaston illusion, wherein identical sets of eyes (e.g., reflected in the iris position and amount of sclera visible) result in a different percept of gaze direction dependent upon the surrounding head orientation. The face images were generated with FaceGen Modeller 3.5., (https://facegen.com/).

Figure 2

Wollaston stimuli. (a) Eyes isolated from a frontal head were paired with heads angled either left or right. (b) Congruent pairings (e.g., leftwards-angled eyes in leftwards-angled head) tended to produce a sense of averted gaze. Incongruent pairings (e.g., leftwards-angled eyes in a rightwards-angled head) tended to produce a sense of gaze directed towards the viewer. When comparing between the congruent and incongruent pairings, the stimuli are comprised of identical base components, but presented in different combinations. Hence, features like iris position, visible sclera and head rotation cues are controlled across this comparison, despite the differences in perceived direction of gaze. The face images were generated with FaceGen Modeller 3.5., (https://facegen.com/).

Figure 3

Example of the Standard stimuli. These stimuli have gaze directed away from the viewer (a) or towards the viewer (b). The eyes in these faces differ in low-level physical cues to gaze direction (e.g., amount and position of visible sclera). The heads and eyes on the top row (a) are deviated at an angle of 20° relative to the viewer, whereas the bottom row contains stimuli with heads rotated at 20° and eyes deviated at an angle of 0° relative to the viewer. The face images were generated with FaceGen Modeller 3.5., (https://facegen.com/).

Figure 4

(a) A schematic of an example CFS trial. A high contrast flashing mask was presented to the participant’s dominant eye, while a face stimulus (here Standard Direct) was presented to the non-dominant eye. (b) A schematic example of a participant’s view during the task. The binocular presentation results in the temporary suppression of the face stimulus from conscious awareness. Participants are instructed to respond as quickly and as accurately as possible when any aspect of the face becomes visible.

Figure 5

Mean suppression times for each condition (a). Error bars denote ± standard error. *(p < 0.05). (b) Sample distributions of the direct effect (i.e., the extent to which faces with direct gaze are detected faster than faces with averted gaze) are reported as latency normalised mean differences in suppression times between averted and direct conditions for Wollaston stimuli and Standard stimuli. Values above zero indicate a direct gaze effect. Note, not all participants show the direct effect with Wollaston or Standard stimuli.