Representation of regular and irregular shapes in macaque inferotemporal cortex

Area of Science:

• Neuroscience
• Cognitive Science
• Computer Vision

Background:

• The inferior temporal (IT) cortex is crucial for visual object recognition.
• Understanding how IT neurons represent shape is key to deciphering visual perception.
• Previous models often focus on physical image properties, not perceptual salience.

Purpose of the Study:

• To investigate whether macaque IT neuronal responses reflect perceptual or physical shape similarities.
• To compare neural representations with human sorting behavior and computational models.

Main Methods:

• Testing IT neuronal responses to four distinct shape categories varying in symmetry and contour properties.
• Using multidimensional scaling to analyze population neural codes.
• Comparing neural data with human sorting tasks and image-based computational models (pixel energies, Gabor-jet filtering, HMAX).

Main Results:

• Neural modulations were stronger for shapes differing in nonaccidental properties (NAPs) than in contour configurations.
• Population codes of neural activity could effectively distinguish the four shape groups.
• Human sorting behavior mirrored the neural modulation patterns, but current image-based models failed to predict these findings.

Conclusions:

• IT neuronal representations of shape emphasize perceptually salient features, not just physical reality.
• The brain prioritizes features relevant for essential categorization over a literal physical depiction.
• Current computational models do not fully capture the principles of shape representation in IT cortex.