Cognitive processes are disentangled at cortex-wide scales

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Decision-making relies on multiple cognitive processes.
• These processes are distributed across the cortex but implemented via orthogonal population-activity patterns within regions.
• The integration of local neural geometries across the cortex remains unclear.

Purpose of the Study:

• To investigate how cortex-wide neural dynamics support distinct cognitive processes during decision-making.
• To determine the geometric organization of neural activity representing sensory evidence accumulation, short-term memory, and choice.

Main Methods:

• Developed a virtual-navigation task for mice to dissociate evidence accumulation, memory, and choice.
• Utilized cortex-wide widefield calcium (Ca2+) imaging.
• Applied dimensionality reduction and decoding models to analyze neural population activity.

Main Results:

• Observed distributed but near-orthogonal coding subspaces for distinct cognitive processes (evidence accumulation, memory, choice).
• This neural geometry disintegrates during erroneous choices.
• The memory subspace, but not others, aligned with a spontaneous activity-timescale hierarchy, suggesting co-option of intrinsic circuit properties.

Conclusions:

• Cortex-wide neural dynamics supporting distinct cognitive processes are disentangled.
• This disentanglement is crucial for accurate decision-making and breaks down during errors.
• The memory subspace's alignment with intrinsic circuit properties suggests a mechanism for its integration into decision computations.