Statistically efficient neural encoding of natural object variability shapes the temporal dynamics of visual...

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Object perception is a dynamic process occurring over milliseconds.
• Understanding the organizational principles of neural responses in object perception remains a challenge.
• Traditional methods may limit interpretations by focusing on pre-selected object features.

Purpose of the Study:

• To apply a data-driven framework to uncover organizational principles of neural object representations.
• To model time-resolved neural activity using behaviorally relevant object dimensions.
• To investigate how the brain represents real-world objects dynamically.

Main Methods:

• Utilized the THINGS initiative dataset with systematic sampling of real-world objects.
• Derived behaviorally relevant stimulus dimensions from large-scale similarity judgments.
• Applied Partial Least Squares Regression (PLSR) to create neural encoding models for EEG and MEG data.

Main Results:

• A small set of latent components reliably captured temporal dynamics across EEG and MEG.
• These components showed consistency with prior MRI findings.
• Identified components encoded diverse visual and semantic object features, not aligning with canonical models.

Conclusions:

• Object representations in the brain are structured by principles of statistical efficiency.
• The brain captures feature co-occurrence within natural object variability for efficient processing.
• Findings suggest a dynamic, statistically driven organization for visual object processing.