From individual features to full faces: Combining aspects of face information

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computer Vision

Background:

• Human face recognition relies on integrating information from various facial features.
• Understanding how the brain processes facial information is crucial for both human and artificial intelligence.

Purpose of the Study:

• To investigate the integration of information from individual facial features versus feature compounds (external and internal).
• To determine if holistic face processing is efficient and how context affects feature sensitivity.

Main Methods:

• Measured discrimination thresholds for synthetic faces under conditions of full-faces, individual features, and feature compounds.
• Presented features in isolation and embedded within a task-irrelevant face context.
• Utilized a computational model to assess information integration efficiency.

Main Results:

• Sensitivity to internal and external feature compounds was comparable to the most sensitive individual features.
• Embedding individual features in context reduced sensitivity, suggesting holistic processing.
• Sensitivity to feature compounds was unaffected by context, indicating independent processing of external and internal features.

Conclusions:

• Holistic face processing limits access to individual feature information but is not an efficient integration strategy.
• External and internal facial features appear to be processed independently.
• Findings inform models of human face perception and AI development.