Synaptic augmentation in a cortical circuit model reproduces serial dependence in visual working memory

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Visual working memory (VWM) exhibits serial dependence, where current information integrates with recent past information.
• This temporal smoothing enhances memory stability against noise and occlusion.
• The underlying neural mechanisms of VWM serial dependence remain largely unknown.

Purpose of the Study:

• To investigate the circuit-level neural mechanisms of serial dependence in VWM using a biophysical model of the cortex.
• To explore the roles of persistent neural activity and activity-silent synaptic plasticity in VWM serial dependence.

Main Methods:

• Development of a hybrid biophysical neural network model of the cortex.
• Simulation of neural network dynamics incorporating persistent activity and dynamic synaptic plasticity.
• Comparison of model predictions with behavioral data on VWM serial dependence.

Main Results:

• The model successfully reproduced behavioral serial dependence.
• Both strong reverberation for persistent activity and dynamic synaptic plasticity were crucial for replicating the phenomenon.
• Elevated neural activity led to synaptic augmentation, biasing subsequent trial activity and causing population response shifts.

Conclusions:

• A combination of persistent neural activity and dynamic synaptic plasticity provides a plausible neural basis for VWM serial dependence.
• The findings offer testable hypotheses for future physiological research.
• This biologically informed model offers a quantitative explanation for observed human behavior in VWM.