Macaques show an uncanny valley in body perception

Poster Presentation 23.339: Saturday, May 18, 2024, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Face and Body Perception: Bodies

Lucas Martini1 (), Anna Bognár2, Rufin Vogels2, Martin Giese1; 1University of Tuebingen, 2KU Leuven

Highly realistic virtual stimuli from avatars have been essential for the investigation of face processing and will likely be similarly important for the study of body perception. In face avatars, humans as well as macaques show an uncanny valley, i.e. when such stimuli are made increasingly realistic the emotional preference for them does not linearly increase with realism, but is drastically reduced for almost realistic presentations. An uncanny valley has also been found for humans for the observation of robots. Whether such an uncanny valley exists for body stimuli in non-human primates is not known. METHODS: We developed a novel computer animation pipeline that allows the animation of a commercial monkey body avatar from markerless multi-camera motion capture recordings of real monkeys. Based on a minimal number of hand-labeled keyframes, our method generates highly realistic dynamic body animations. In a behavioral experiment with eight male rhesus monkeys, we validated our macaque model by monitoring the subjects’ eye movements on renderings of a submissive turning and a neutral walking movement. These stimuli were shown from different viewpoints, incorporated realistic background, and degraded variants of the most realistic avatar. RESULTS: The analysis of the eye movement data revealed an uncanny valley effect in coverage of the stimulus, number of image fixations, and mean fixation duration (p=0.002, p=0.02, and p<0.001), where the gazing behavior for the non-degraded avatar did not significantly deviate from the real video. CONCLUSION: Monkeys exhibit an uncanny valley for the observation of monkey bodies. This shows the universality of this effect across different social stimuli and primate species.

Acknowledgements: European Research Council (2019-SyG-RELEVANCE-856495). International Max Planck Research School for Intelligent Systems (IMPRS-IS).