Sensory-memory interactions via modular structure explain errors in visual working memory.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Stimulus estimation errors offer insights into cognitive processes.
• Repulsive bias and minimum variance are known error patterns in visual orientation perception.
• Evolving errors during working memory challenge existing sensory and memory models.

Purpose of the Study:

• To investigate the neural mechanisms underlying evolving errors in working memory.
• To determine if a single-module or multi-module network is necessary to explain these evolving errors.
• To model the interaction between sensory encoding and memory maintenance modules.

Main Methods:

• Developed a computational network model with two distinct modules: sensory and memory.
• Simulated network dynamics under conditions mimicking working memory tasks.
• Analyzed emergent properties like attractor states, drift speed, and diffusion processes.

Main Results:

• A single-module network cannot simultaneously achieve efficient sensory encoding and memory maintenance.
• A two-module network with specific interactions can replicate the observed evolving errors and their characteristic shapes.
• The sensory module shows heterogeneous tuning with inhibitory modulation, while the memory module exhibits continuous attractor dynamics alone.
• The fully connected network forms discrete attractors with moderate drift and nonuniform diffusion.

Conclusions:

• Network interaction between distinct sensory and memory modules is crucial for shaping stimulus representation during working memory.
• The findings highlight the importance of sensory-memory interplay in dynamic cognitive processes.
• This model provides a framework for understanding how the brain integrates sensory information with memory recall.