Differences in the Judged Direction of Gaze From Heads Imaged in 3-D versus 2-D

Perception of direct gaze in a video-conference setting: the effects of position and size

A common problem in video conferences is gaze direction. In face-to-face communication, it is common that speaker and listener intermittently look at each other. In a video-conference setting, where multiple participants are on the screen, things are complicated and not necessarily optimal. If the listener feels looked at when the speaker looks into the camera, how tolerant is the listener for slight deviations? And does this depend on the position of the speaker’s tile on the screen, or the size of the tile? In a first experiment, participants from a student population judged whether they are looked at, while vertical gaze direction of the looker was varied. Furthermore, the position of the tile on the screen varied. The results showed that a slightly upward directed gaze was optimal for the direct gaze judgment, with a width of ± 4 degrees. Optimal gaze direction was somewhat higher for tiles at the bottom of the screen. A second experiment tested the effect of size on the perception of horizontal gaze directions. Size was found to increase the gaze cone. The paper concludes with some recommendations for a setup of video conference systems, optimized for perceived gaze contact.

Is there a 'zone of eye contact' within the borders of the face?

Eye contact is a salient feature of everyday interactions, yet it is not obvious what the physical conditions are under which we feel that we have eye contact with another person. Here we measure the range of locations that gaze can fall on a person's face to elicit a sense of eye contact. Participants made judgements about eye contact while viewing rendered images of faces with finely-varying gaze direction at a close interpersonal distance (50 cm). The 'zone of eye contact' tends to peak between the two eyes and is often surprisingly narrower than the observer's actual eye region. Indeed, the zone tends to extend further across the face in height than in width. This shares an interesting parallel with the 'cyclopean eye' of visual perspective - our sense of looking out from a single point in space despite the physical separation of our two eyes. The distribution of eye-contact strength across the face can be modelled at the individual-subject level as a 2D Gaussian function. Perception of eye contact is more precise than the sense of having one's face looked at, which captures a wider range of gaze locations in both the horizontal and vertical dimensions, at least at the close viewing distance used in the present study. These features of eye-contact perception are very similar cross-culturally, tested here in Australian and Japanese university students. However, the shape and position of the zone of eye contact does vary depending on recent sensory experience: adaptation to faces with averted gaze causes a pronounced shift and widening of the zone across the face, and judgements about eye contact also show a positive serial dependence. Together, these results provide insight into the conditions under which eye contact is felt, with respect to face morphology, culture, and sensory context.

Inverting the Wollaston Illusion: Gaze Direction Attracts Perceived Head Orientation

In the early 19th century, William H. Wollaston impressed the Royal Society of London with engravings of portraits. He manipulated facial features, such as the nose, and thereby dramatically changed the perceived gaze direction, although the eye region with iris and eye socket had remained unaltered. This Wollaston illusion can be thought of as head orientation attracting perceived gaze direction when the eye region is unchanged. In naturalistic viewing, the eye region changes with head orientation and typically produces a repulsion effect. Here we explore if there is a flip side to the illusion. Does the gaze direction also alter the perceived direction of the head? We used copies of the original drawings and a computer-rendered avatar as stimuli. Gaze direction does indeed alter perceived head orientation. Perceived head orientation is biased toward the direction of gaze.

Examining Gaze Cone Shape and Size

Another person's looking behavior is used by observers to judge gaze direction and fixation points. An important task in this context is the judgement of direct gaze, that is, the perception of being looked at. The cone of gaze can be defined as the range of fixation points that support direct gaze. The cone concept implies that this range lawfully increases with distance, but that the cone angle is constant. The present experiment tested the concept with a larger number and a more extended range of distances than previously done, and took care of possible directional errors. The gaze cone was found to be roughly linear, and stable between 1.6 m and 7.9 m - an almost perfect cone. The mean cone size subtended 5.2° in diameter when averaged over ascending and descending series. Measures differed, however, in ascending and descending series, consistent with a conservative bias. Also, the variability of judgements increased slightly with distance. Results are discussed considering whether cone size is actually smaller than often reported in the literature.

Facing the Spectator

We investigated the familiar phenomenon of the uncanny feeling that represented people in frontal pose invariably appear to “face you” from wherever you stand. We deploy two different methods. The stimuli include the conventional one—a flat portrait rocking back and forth about a vertical axis—augmented with two novel variations. In one alternative, the portrait frame rotates whereas the actual portrait stays motionless and fronto-parallel; in the other, we replace the (flat!) portrait with a volumetric object. These variations yield exactly the same optical stimulation in frontal view, but become grossly different in very oblique views. We also let participants sample their momentary awareness through “gauge object” settings in static displays. From our results, we conclude that the psychogenesis of visual awareness maintains a number—at least two, but most likely more—of distinct spatial frameworks simultaneously involving “cue–scission.” Cues may be effective in one of these spatial frameworks but ineffective or functionally different in other ones.