Are Face and Object Recognition Independent? A Neurocomputational Modeling Exploration.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Face and object recognition are often considered independent abilities.
• Recent research suggests a correlation between face and object recognition that increases with experience.

Purpose of the Study:

• To explain why a shared resource for visual abilities does not lead to competition.
• To model how experience influences the correlation between face and object recognition.

Main Methods:

• Utilized a neurocomputational model of face and object processing.
• Modeled domain-general ability (v) as computational resources and experience as exemplar frequency.
• Trained a network on face and object recognition tasks at different levels of categorization.

Main Results:

• The model replicated behavioral findings: correlation between subordinate-level face and object recognition increased with experience.
• No increased correlation was observed for basic-level categorization tasks.
• The model suggests shared features explain the lack of competition between domains.

Conclusions:

• Experience drives a "spreading transform" that generalizes across domains, enhancing recognition.
• The fusiform face area (FFA) is implicated as the source of this correlation, not basic categorization areas.
• The type of experience is crucial, particularly for subordinate-level individuation.