Time-invariant working memory representations in the presence of code-morphing in the lateral prefrontal cortex

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Working memory maintenance is linked to prefrontal neuronal activity and recurrent connections.
• Distractors can alter prefrontal population codes, posing a challenge to time-invariant memory representation.

Purpose of the Study:

• To investigate how time-invariant working memory information is maintained despite dynamic neural coding in the prefrontal cortex.
• To identify if specific subspaces within neural activity encode memory information robustly.

Main Methods:

• Utilized an optimization algorithm to identify low-dimensional subspaces within prefrontal neural activity.
• Analyzed neural data from a working memory task, including error trials.
• Employed a bump attractor model to simulate and validate findings.

Main Results:

• A low-dimensional subspace containing time-invariant memory information was identified.
• This information was diminished in error trials, correlating with task performance.
• The identified subspace properties were successfully replicated by a bump attractor model.

Conclusions:

• Dynamic prefrontal activity encodes time-invariant memory information within specific neural subspaces.
• This subspace coding mechanism allows for simultaneous encoding of different information types with reduced interference.
• Findings challenge traditional views and offer a new perspective on neural coding in working memory.