Are you looking at me? About the perception of direct gaze and gaze direction

This thesis provides innovative learnings on the area of direct gaze (i.e., the range of gazes within which an observer feels looked at). It discusses valid methods to assess the area of direct gaze, the nature of its shape, and whether it is comparable to other social gestures. Previous work focused on the meaning of horizontal and vertical eye movements for the perception of direct gaze. The present dissertation extends this view on the role of vergence eye movements and the angle kappa, i.e., the deviation of the pupillary from the foveal axis, since they influence the visible pupil position as well. <br />Our first study revealed that humans accept a large range of vergence angles as direct gaze. This way, we added a third dimension to the gaze cone model – the depth dimension. <br />The second study tested the depth dimension of direct gaze for different distances. The results indicate that observers do not accept a convergence angle larger than approximately 13° as being directed at them, relatively independent of the observer’s position. Therefore, the proximal border appears to be allocated 15-30 cm in front of the looker independent of the observer’s distance to the looker. Both studies revealed an infinite distal border behind the observer since 10° of divergence was accepted as being directed to the observer. Taken together, the results suggest to model the area of direct gaze as a truncated cone in the depth dimension with the starting point being fixed 15-30 cm in front of the looker, independently of the observer’s position, and an infinite end.<br /> In our third study, we asked how the information of both eyes is combined to one single estimation of gaze direction and directness of gaze. On account of vergence, the absolute rotation of both eyes is different from each other in near distances. Therefore, both eyes do not have the exact same rotation level. As observers’ accuracy in estimating the looker-observer distance only by the vergence cue is limited, it is likely that observers cannot precisely trace back the different rotation levels to the presentation distance. It has been discussed in previous literature whether the perception of gaze direction is based on the externally rotated eye (rotated away from the body midline) or the internally rotated eye (rotated towards the body midline). The present study found the decision of whether a gaze is direct to be based on the internally rotated eye, which was the more rotated eye (extremer rotation towards the side). Differences between the present and previous studies may be conciliated by the new generalization that the decision is based on the more rotated eye - independent of whether it is internally or externally rotated. Which eye is more rotated depends on the presentation distance (and consequently the vergence angle) and on the angle kappa of the looker model. Although angle kappa varies greatly in the population, it has often been neglected in previous research and could therefore explain the large variations in the gaze cone measure (from 3° to 15°). The fourth study compared the use of different tasks to measure the gaze cone (method of constant stimuli, method of adjustment and triadic measurement) and critically discussed how the used measurement method influences the measurement of the gaze cone. Results of two experiments indicate that thresholds are larger when measured by the method of adjustment than when measured by the method of constant stimuli. Furthermore, in our first experiment we found a non-linear overestimation factor, indicating that gaze directions near 0° are less overestimated than larger eccentricities. Correcting the thresholds with individually obtained overestimation factors widens the gaze cone but does not eliminate the average difference between the methods of adjustment and constant stimuli. Finding these differences between methods within participants, the fourth study calls for caution to interpret the exact size of the area of direct gaze, on the one hand, and comparisons made between studies using different methods, on the other hand.<br /> In our fifth study, we were interested in the perception of a direct social gesture. We compared the perception of direct looking and direct pointing using 2D stimuli (photographed arm model/ avatar) and 3D stimuli (mannequin and real-life pointer). All experiments provided evidence for a similar area of direct looking and direct pointing, suggesting shared mechanisms in the perception of both gestures. This result favors a geometrical approach in gaze and pointing perception. However, only the first experiment with the 2D stimuli suggested a perfect cone shape for looking and pointing (i.e., the same angle in different distances). However, with the 3D stimuli for looking and for pointing, we found a significant influence of the distance on the cone size, partly rejecting the cone hypothesis. Instead, by using a linear regression as an alternative analysis, we could provide more detailed insights about the shape, revealing a truncated cone with an area of origin rather than the cone originating from the interpupillary point. Overall, the findings of this dissertation add several new dimensions to our understanding of gaze perception, including: a) the precise role of vergence in natural settings, b) the principles underlying the estimation of gaze direction and direct gaze, c) the shared mechanisms between social gestures, particularly looking and pointing, and d) insights into the impact of different measurement techniques.

Level thesis:degree_level

thesis.doctoral

Grantor dc:publisher

Universität Bielefeld

Year

2025

Author dc:creator

• Linke, Linda

Repository record source_url

https://pub.uni-bielefeld.de/record/3003476

OAI identifier oai:identifier

oai:pub.uni-bielefeld.de:3003476