Human Scene-Selective Areas Represent 3D Configurations of Surfaces

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Scene-selective brain areas are hypothesized to process visual information.
• Two main hypotheses suggest these areas encode either 3D scene structure or low-level 2D features.

Purpose of the Study:

• To determine the extent to which 3D structure versus 2D features explain neural variance in scene-selective areas.
• To develop and test encoding models for 3D scene structure and low-level 2D features.

Main Methods:

• Functional magnetic resonance imaging (fMRI) data were collected while participants viewed visual scenes.
• Encoding models for 3D scene structure and low-level 2D features were developed and fitted to the fMRI data.
• Principal component analysis (PCA) was used to analyze model weights.

Main Results:

• Scene-selective areas represent 3D surface distance and orientation, partly independent of low-level 2D features.
• PCA identified distance and openness as key dimensions of 3D structure represented in these areas.
• Reconstructions demonstrated the model's ability to capture detailed local visual environment information.

Conclusions:

• Scene-selective brain regions encode complex 3D environmental information beyond simple 2D visual cues.
• The findings support the role of these areas in representing the spatial layout of the visual world.