Separate dorsolateral prefrontal cortex regions participate in distinct large-scale networks differentially recruited for social and cognitive control functions.

Abstract

Human prefrontal cortex (PFC) is heterogeneous. In monkeys, side-by-side PFC regions, including within dorsolateral PFC (DLPFC), show distinct long-range anatomical projection profiles, raising the possibility that separate DLPFC regions might specialize as a consequence of the distributed networks in which they are embedded. Recent findings in humans provide evidence for PFC regions preferentially responding to scene (spatial) processing or language processing that are adjacent to distinct domain-flexible regions responding to traditional cognitive control demands. Here, we tested functional specialization of PFC regions linked to another domain-preferential network recruited by certain forms of social processing (theory-of-mind, ToM, tasks). Using within-individual precision neuroimaging approaches, two anatomically distinct DLPFC regions embedded within parallel distributed networks were identified within the idiosyncratic anatomy of each individual (n = 13). Functional responses between these distinct regions revealed a robust functional double dissociation: one DLPFC region was preferentially recruited by ToM tasks and a nearby DLPFC region by working memory task demands. The region responding to social processing demands was small and its position varied slightly from one person to the next, suggesting why it may have been underappreciated in past group-based analyses. These findings add to evidence that PFC features more functional specialization than commonly appreciated and further that the specialization of distinct regions within PFC can be understood by examining the distributed networks within which the regions are embedded.NEW & NOTEWORTHY Dorsolateral prefrontal cortex (DLPFC) is often heuristically associated with domain-flexible cognitive control, rather than specialization for social processing. Here, using precision neuroimaging, we identify a DLPFC region of a domain-preferential network linked to social inference. This region also shows preferential response to theory-of-mind tasks, distinct from nearby regions of a control-relevant network that respond flexibly to working memory task demands. Results add evidence that DLPFC features distinct domain-flexible and domain-specialized regions, with functions aligned to broader networks.