Mapping attractor fields in face space: the atypicality bias in face recognition

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computer Science

Background:

• Human face recognition involves mapping diverse stimuli to a single memory.
• This mapping is conceptualized as a face memory's 'attractor field'.
• Stimuli within an attractor field activate the corresponding face memory.

Purpose of the Study:

• To investigate if atypical faces have larger attractor fields than typical faces.
• To test the prediction that attractor field size correlates with facial typicality.
• To explore the computational basis of potential atypicality biases in face recognition.

Main Methods:

• Subjects judged likeness of morphed faces (50/50 atypical/typical parent).
• Four experiments were conducted using behavioral likeness judgments.
• A neural network simulation modeled face representations and activation.

Main Results:

• Morph faces were consistently judged as more similar to the atypical parent face.
• Behavioral data indicated a resemblance bias towards atypical faces.
• Neural network simulations showed stronger activation for atypical representations.

Conclusions:

• Atypical faces possess larger attractor fields, spanning a broader region of face space.
• This broader 'attractor field' explains the observed bias towards atypical faces in recognition.
• Evidence supports a computational model where atypicality enhances face memory activation.