Neural mechanisms underlying dynamic manipulation of sequential information in working memory

Maintaining and manipulating sequential information in working memory is crucial for complex cognitive functions, such as language processing and action planning. While prior research has extensively characterized the neural substrates of passive sequence maintenance, the mechanisms underlying dynamic sequence manipulation remain poorly understood. This fMRI study employed univariate analyses, multivariate pattern analysis, and functional connectivity methods to dissociate neural mechanisms supporting active sequence reorganization from passive maintenance. Thirty-one healthy adults performed a digit-ordering task with two conditions: pure recall (maintaining preordered sequences) and reorder & recall (dynamically organizing randomized sequences). Univariate analyses revealed that manipulation, compared with maintenance, robustly activated a frontoparietal network (FPN) and deactivated the default mode network (DMN). Multivariate pattern analysis further demonstrated that manipulation-specific neural representations were decodable within frontoparietal regions, indicating their role in encoding the transformation of information. Critically, functional connectivity analysis revealed a dynamic network reconfiguration during manipulation, characterized by strengthened intra-network connectivity within the FPN and significant decoupling between the FPN and DMN. Within this reorganized architecture, the anterior cingulate cortex emerged as an integrative hub within the FPN, while the middle frontal gyrus mediated the interaction between the FPN and DMN. These findings provide a comprehensive, multilevel account of working memory manipulation, highlighting that it relies not merely on localized activation but on a large-scale, demand-driven reconfiguration of brain networks, involving a precise balance of network segregation and integration orchestrated by key cortical hubs.