Attractor dynamics with activity-dependent plasticity capture human working memory across time scales

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Working memory is crucial for cognitive functions, involving the brain's ability to retain recent stimuli.
• Neuronal activity configurations, known as attractors, are hypothesized to store working memories.
• Existing models often simplify the temporal dynamics of memory storage.

Purpose of the Study:

• To test if discrete attractor dynamics accurately model working memory across multiple timescales.
• To investigate the limitations of current attractor models in capturing the complexity of memory storage.
• To develop an improved model that accounts for longer delay intervals and intertrial interactions.

Main Methods:

• A human working memory task with varied delay periods was employed.
• Discrete attractor dynamics were initially used to model neuronal activity.
• A novel model combining attractor dynamics with activity-dependent plasticity was developed and tested.

Main Results:

• Discrete attractor dynamics approximated working memory at short timescales but failed to generalize across multiple delays.
• Longer delay intervals revealed more stimulus information than discrete attractors could store.
• The combined model successfully generalized across all timescales and predicted intertrial interactions.

Conclusions:

• Discrete attractor dynamics alone are insufficient for comprehensive working memory modeling.
• Activity-dependent plasticity enhances the durability of information storage within attractor systems.
• The findings suggest a more complex mechanism involving plasticity for robust working memory.