A geometric shape regularity effect in the human brain

Area of Science:

• Neuroscience
• Cognitive Science
• Computer Vision

Background:

• Human cultures have long recognized and produced regular geometric shapes, yet the underlying neural mechanisms remain unclear.
• Behavioral evidence indicates humans are sensitive to geometric regularities like symmetry and parallelism, using them for efficient shape encoding.

Purpose of the Study:

• To investigate the neural systems and dynamics involved in the perception of regular geometric shapes.
• To compare brain activity during geometric shape perception with computational models.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used in adults and 6-year-olds.
• Magnetoencephalography (MEG) data were collected from adults.
• Participants viewed simple geometric shapes (e.g., hexagons, triangles, quadrilaterals).

Main Results:

• Geometric shapes, compared to other visual categories, showed hypoactivation in ventral visual areas.
• Increased activation was observed in the intraparietal and inferior temporal regions, areas linked to mathematical processing.
• Convolutional neural networks (CNNs) modeled early visual activity but not later dorsal and prefrontal signals.

Conclusions:

• The perception of abstract geometric regularities involves distinct neural pathways beyond early visual processing.
• These findings suggest that geometric shape perception engages an additional symbolic mode of visual perception.
• Advanced deep-learning models incorporating discrete geometric features are needed to fully capture these neural signals.