Modeling human social vision with cinematic stimuli: An integrative approach

Area of Science:

• Cognitive Neuroscience
• Visual Perception
• Social Cognition

Background:

• Sociability is fundamental to human interaction.
• Visual processing encompasses low-level physical, mid-level semantic, and high-level social information.
• Understanding how these visual levels guide social perception is crucial.

Purpose of the Study:

• To map how different levels of visual information simultaneously guide social vision.
• To investigate the association between perceptual features and eye-tracking parameters during naturalistic viewing.

Main Methods:

• Utilized large-scale eye tracking with 166 participants watching 193 minutes of film.
• Extracted 39 stimulus features (low-level, mid-level, high-level) from movie stimuli.
• Employed integrative analysis with cross-validation to link perceptual features and gaze behavior.

Main Results:

• Gaze position was best predicted by faces, motion, and entropy.
• Faces and eyes received prioritized attention.
• Pupil size correlated with luminance, scene cuts, and emotional arousal.
• Blinking decreased during periods of high attentional engagement.

Conclusions:

• Human social vision is predominantly guided by low-level physical and mid-level semantic visual features.
• High-level social information, like emotional arousal, primarily affects pupillary responses.
• This study provides insights into the multi-faceted visual processing underlying social navigation.