Dynamic network coding of working-memory domains and working-memory processes

Eyal Soreq,Adam Hampshire,Robert Leech

doi:10.1038/s41467-019-08840-8

Eyal Soreq, Adam Hampshire + Show 1 more

Open Access

https://doi.org/10.1038/s41467-019-08840-8

Copy DOI

Abstract

The classic mapping of distinct aspects of working memory (WM) to mutually exclusive brain areas is at odds with the distributed processing mechanisms proposed by contemporary network science theory. Here, we use machine-learning to determine how aspects of WM are dynamically coded in the human brain. Using cross-validation across independent fMRI studies, we demonstrate that stimulus domains (spatial, number and fractal) and WM processes (encode, maintain, probe) are classifiable with high accuracy from the patterns of network activity and connectivity that they evoke. This is the case even when focusing on ‘multiple demands’ brain regions, which are active across all WM conditions. Contrary to early neuropsychological perspectives, these aspects of WM do not map exclusively to brain areas or processing streams; however, the mappings from that literature form salient features within the corresponding multivariate connectivity patterns. Furthermore, connectivity patterns provide the most precise basis for classification and become fine-tuned as maintenance load increases. These results accord with a network-coding mechanism, where the same brain regions support diverse WM demands by adopting different connectivity states.

Highlights

The classic mapping of distinct aspects of working memory (WM) to mutually exclusive brain areas is at odds with the distributed processing mechanisms proposed by contemporary network science theory
During Study 1, participants were presented with arrays containing numbers and fractals displayed at a subset of spatial locations within a 4 × 4 grid
There were no significant effects of manipulation and no other significant interactions. These results show that participants were able to perform both tasks with high accuracy and that there were the expected costs of WM load on performance

Summary

Introduction

The classic mapping of distinct aspects of working memory (WM) to mutually exclusive brain areas is at odds with the distributed processing mechanisms proposed by contemporary network science theory. There is a need to reconcile the network perspective with the robust evidence of mutually exclusive localist mappings of WM functions, as provided by the early-brain imagining literature[5,6,7,8,9,10,11,12,13,14,15,16,17] We addressed these issues by developing a multivariate machine-learning pipeline to examine how patterns of brain activity and connectivity dynamically changed during the performance of two novel fMRI tasks that were designed to probe distinct aspects of WM function[9,28,29,30,31]. Our results strongly support the network-coding perspective, provide new insights into the underlying mechanisms of WM and demonstrate how models of the early localist literature may be reconciled

Objectives

Methods

Conclusion