The neuronal implementation of representational geometry in primate prefrontal cortex

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Modern neuroscience utilizes population doctrines to represent cognitive variables geometrically in activity space.
• Representational geometry, while powerful, lacks insight into individual neuronal implementation of these representations.

Purpose of the Study:

• To develop a framework that dissects representational geometry into biologically interpretable components linked to single neurons.
• To investigate how individual neurons implement cognitive representations using sparse coding principles.

Main Methods:

• Leveraging the principle of sparse coding.
• Applying a novel framework to extracellular recordings from primate prefrontal cortex.
• Analyzing data from a working memory task with interference.

Main Results:

• Identified components revealed disentangled and sequential memory representations.
• Observed recovery of memory content after distraction, with hidden signals.
• Each component was linked to small neuronal subpopulations with distinct spiking properties and response dynamics.

Conclusions:

• Sparse coding provides a biologically interpretable link between representational geometry and neuronal implementation.
• Recurrently connected circuits, characteristic of the prefrontal cortex, support these sparse implementations.
• This perspective offers mechanistic insights into how neural systems encode and process information at the cellular level.