A dynamic neural resource model bridges sensory and working memory

Ivan Tomić,Paul M. Bays

DOI: https://doi.org/10.1101/2023.03.27.534406

2024-02-05

Abstract:Probing memory of a complex visual image within a few hundred milliseconds after its disappearance reveals significantly greater fidelity of recall than if the probe is delayed by as little as a second. Classically interpreted, the former taps into a detailed but rapidly decaying visual sensory or “iconic” memory (IM), while the latter relies on capacity-limited but comparatively stable visual working memory (VWM). While iconic decay and VWM capacity have been extensively studied independently, currently no single framework quantitatively accounts for the dynamics of memory fidelity over these timescales. Here we extend a stationary neural population model of VWM with a temporal dimension, incorporating rapid sensory-driven accumulation of activity encoding each visual feature in memory, and a slower accumulation of internal error that causes memorized features to randomly drift over time. Instead of facilitating read-out from an independent sensory store, an early cue benefits recall by lifting the effective limit on VWM signal strength imposed when multiple items compete for representation, allowing memory for the cued item to be supplemented with information from the decaying sensory trace. Empirical measurements of human recall dynamics validate these predictions while excluding alternative model architectures. A key conclusion is that differences in capacity classically thought to distinguish IM and VWM are in fact contingent upon a single resource-limited WM store.

Neuroscience

What problem does this paper attempt to address?

The problem this paper attempts to address is the dynamic changes and interrelationship between visual working memory (VWM) and visual sensory memory (IM) across different time scales. Specifically, researchers have found that detecting memory of a complex visual image within a few hundred milliseconds after it disappears results in significantly higher recall fidelity compared to detecting it after a delay of 1 second. The classic explanation suggests that the former relies on detailed but rapidly decaying visual sensory or "iconic" memory (IM), while the latter depends on the limited capacity but relatively stable visual working memory (VWM). However, there is currently no unified framework that can quantitatively explain the dynamic changes in memory fidelity across these time scales.

To fill this gap, the paper extends a static neural population model by introducing a time dimension to describe the rapid sensory-driven accumulation of each visual feature in memory and the slow accumulation of internal errors, leading to random drift of memory features over time. The researchers demonstrate through experiments that early cues can improve recall by enhancing the ability to effectively limit the strength of the VWM signal, rather than directly reading information from an independent sensory store. This suggests that the capacity difference between IM and VWM is actually a conditional difference based on the limited working memory storage of a single resource.

Overall, the paper aims to establish a comprehensive model that can explain the dynamic changes in visual memory across different time scales, thereby better understanding the relationship between visual working memory and sensory memory.