Individual differences in internal models explain idiosyncrasies in scene perception.

Area of Science:

• Cognitive Science
• Neuroscience
• Computer Vision

Background:

• Predictive processing theories suggest vision relies on internal models of the world.
• The specific content and individual variability of these internal models remain largely unknown.
• Understanding these models is crucial for explaining individual differences in visual perception.

Purpose of the Study:

• To investigate the contents of individual internal models of scene categories.
• To determine how these idiosyncratic models influence scene perception and categorization.
• To introduce drawing as a novel behavioral method for accessing internal predictive models.

Main Methods:

• Participants drew typical scene categories to represent their internal models.
• Drawings were converted into standardized 3D renders used as experimental stimuli.
• Scene categorization accuracy was compared using personalized renders, others' renders, and photographs, analyzed with deep neural networks.

Main Results:

• Scene categorization was significantly more accurate using 3D renders based on participants' own drawings.
• Performance correlated with the graded similarity between candidate scenes and the personalized internal model render.
• This suggests scene perception is guided by a match with individual, idiosyncratic internal models.

Conclusions:

• Drawing provides a valuable behavioral readout of internal world models.
• Scene perception is demonstrably influenced by personalized, internally generated expectations.
• This study introduces a novel methodology for exploring individual differences in visual cognition.