FROST: a distributed neurocomputational model of working memory maintenance

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Modeling

Background:

• Working memory (WM) maintenance relies on distributed brain networks, including the prefrontal cortex, parietal cortex, thalamus, globus pallidus, and caudate nucleus.
• Sustained neural activation is key to holding information during WM tasks.

Purpose of the Study:

• To propose a computational model (FROST) that explains the neural mechanisms of WM maintenance.
• To elucidate how prefrontal cortex activity is sustained during delay intervals.

Main Methods:

• Development of the FROST (FROntal-Striatal-Thalamic) computational model.
• Simulating WM processes, including item and spatial representation encoding in the lateral prefrontal cortex.
• Modeling sustained activation through prefrontal cortical-thalamic loops involving the caudate nucleus and globus pallidus.

Main Results:

• The FROST model successfully replicates sustained activation in prefrontal cortex during WM delay periods.
• The model demonstrates how prefrontal cortical excitation of the caudate nucleus disinhibits the thalamus via the globus pallidus, sustaining neural activity.
• FROST accounts for various working memory data, including single-cell recordings and human behavioral findings.

Conclusions:

• The FROST model provides a viable computational framework for understanding WM maintenance mechanisms.
• Prefrontal-thalamic loops, modulated by the basal ganglia, are crucial for sustained neural representations in working memory.
• The model's success validates its ability to explain complex working memory phenomena.