Working memory cells' behavior may be explained by cross-regional networks with synaptic facilitation.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• Cortical neurons display various firing patterns during working memory tasks.
• Persistent elevated firing is thought to represent working memory retention.
• Existing models with fixed synaptic weights struggle to replicate observed neural variability.

Purpose of the Study:

• To investigate how dynamic synapses and multi-area network architectures influence working memory network states and dynamics.
• To determine if dynamic synapses can inherently produce the diverse firing patterns observed in cortical neurons.
• To assess if distributed network architectures with dynamic synapses match real-world cortical variability and firing statistics.

Main Methods:

• Simulated working memory networks with dynamic synapses.
• Analysis of network architectures incorporating multiple cortical areas.
• Examination of neural firing rate patterns and statistics across simulated trials.

Main Results:

• Networks with dynamic synapses naturally exhibit multiple distinct firing patterns during working memory delays.
• The variability and firing statistics in these dynamic, distributed networks align with empirical observations in the cortex.
• Dynamic synapses are crucial for generating the range of neural activity seen in working memory.

Conclusions:

• Dynamic synapses and distributed network architectures are essential for explaining the complex neural dynamics of working memory.
• Computational models incorporating these features provide a more accurate representation of cortical function during working memory tasks.
• This work reconciles theoretical models with experimental data on neural correlates of working memory.