Emergence of distributed working memory in a human brain network model

Abstract

Working memory is a fundamental cognitive function which allows to transiently store and manipulate relevant information in memory. While it has been traditionally linked to activity in specific prefrontal cortical areas, recent electrophysiological and imaging evidence has shown co-occurrent activities in different brain regions during working memory. To dissect the mechanisms behind the emergence of such distributed working memory activity in the human brain, we built and analyzed a detailed, data-constrained model of the human brain combining multiple sources of large-scale brain data. The model reveals the emergence of distributed working memory patterns which rely on long-range synaptic projections, differing from classical models where working memory emerges due to local recurrent connectivity. When compared to existing evidence regarding areas linked to working memory, our model explains almost 80% of the observed variability across brain areas. Our results predict important differences between the distributed working memory in human and non-human primates, and propose a potential solution for the debate regarding the role of early visual areas in working memory encoding. This work provides a novel approach to simulate the human brain which combines the richness of large-scale brain dynamics with the insight of mechanistic models of cognitive functions.