Ensemble perception of faces with naturalistic occlusions

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computer Vision

Background:

• Ensemble perception allows the visual system to extract summary statistics from redundant information.
• Previous studies show intact ensemble representations for faces with simple occlusions (e.g., bars).
• The impact of naturalistic occlusions (masks, sunglasses) on facial ensemble perception is not well understood.

Purpose of the Study:

• To investigate how naturalistic occlusions affect ensemble perception of faces.
• To determine if the visual system can form robust ensemble representations despite partial occlusion by masks or sunglasses.
• To explore the role of feature learning in maintaining performance under occlusion.

Main Methods:

• Continuous report tasks were used to assess participants' ability to report the average face.
• Faces varying in identity or expression were partially occluded by surgical masks or sunglasses.
• A "face wheel" interface was employed for reporting the perceived average face.

Main Results:

• Participants accurately extracted average facial features even when faces were significantly occluded by masks or sunglasses.
• Performance decreased when the face wheel varied across trials, suggesting a reliance on learned features.
• Robustness of ensemble perception under occlusion is partly supported by learning specific facial features.

Conclusions:

• The visual system demonstrates robust ensemble representations for faces with naturalistic occlusions.
• This robustness is partially attributed to the visual system's ability to learn and utilize specific features within a given set of faces.
• Understanding these mechanisms is crucial for fields like human-computer interaction and artificial intelligence.