Rarely categorical, highly separable representations along the cortical hierarchy

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• A central question in neuroscience is whether neuronal populations encode information categorically or through diverse response patterns.
• Understanding this organization is crucial for deciphering neural computation and information processing in the brain.

Purpose of the Study:

• To systematically analyze how cortical neurons encode cognitive, sensory, and movement variables across multiple cortical regions.
• To investigate the scale-dependent nature of neural coding and its organization across the cortical hierarchy.
• To determine whether neural representations are primarily categorical or diverse and the computational implications of these structures.

Main Methods:

• Analysis of over 14,000 neuronal units from the International Brain Laboratory's public Brainwide Map dataset.
• Systematic analysis of neural encoding across 43 cortical regions during a complex behavioral task.
• Application of theoretical arguments and empirical evidence to link neuronal response diversity to representational capacity.

Main Results:

• Neural coding exhibits scale-dependent properties: categorical representations dominate at the whole-cortex scale, reflecting anatomical connectivity.
• Within individual cortical regions, categorical representations are rare, with neuronal responses being highly diverse, especially outside primary sensory areas.
• Neuronal response diversity enables high-dimensional representations and maximal separability, allowing linear decoders to distinguish conditions effectively across all cortical areas.

Conclusions:

• Cortical circuits prioritize neuronal response diversity over strict categorical structure for information processing.
• This diversity supports a computational regime optimized for high-dimensional and highly separable neural representations.
• The findings challenge traditional views of categorical coding and highlight the functional significance of neuronal response variability.