A dynamic neural resource model bridges sensory and working memory.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychology

Background:

• Visual sensory memory (iconic memory, IM) and visual working memory (VWM) are traditionally viewed as distinct memory systems.
• IM is characterized by high detail but rapid decay, while VWM has limited capacity but greater stability.
• Existing models lack a quantitative framework to explain memory fidelity dynamics across these timescales.

Purpose of the Study:

• To develop a unified computational model that accounts for memory recall dynamics across different time scales.
• To investigate the relationship between sensory-driven memory accumulation and internal memory drift.
• To test whether a single memory store can explain phenomena attributed to both IM and VWM.

Main Methods:

• Extended a stationary neural population model of VWM by incorporating a temporal dimension.
• Modeled rapid sensory-driven activity accumulation and slower internal error accumulation leading to memory drift.
• Compared model predictions with empirical measurements of human visual recall dynamics.

Main Results:

• The extended model quantitatively accounts for memory recall fidelity over time, integrating sensory input and internal decay.
• Early cues enhance recall not by accessing a separate store, but by increasing VWM signal strength and utilizing decaying sensory traces.
• Model predictions align with human behavioral data, supporting a unified memory framework.

Conclusions:

• The distinction between iconic memory and visual working memory capacity arises from a single, resource-limited working memory store.
• Memory recall dynamics are explained by the interplay of sensory input accumulation and internal memory drift within a unified system.
• This framework provides a parsimonious explanation for memory performance differences previously attributed to separate memory stores.