A dual neural system supporting multi-tasking cognitive flexibility

Abstract One ubiquitous feature of human intelligence is the ability to flexibly switch between multiple tasks. Abstract task representations provide a basis for task learning, switching, and generalization, yet how the brain coordinates multiple task representations to support multitasking and task-switching remains poorly understood. We recorded functional MRI activity in human participants while they concurrently held multiple task rules in working memory and prioritized different rules for stimulus processing across trials. Our results reveal two distinct coding schemes for task prioritization. First, an active coding scheme supports the spatial separation of prioritized and unprioritized task rules: prioritized rules are represented across a distributed cortical network, whereas unprioritized rules are only found in the posterior cortex. Second, besides this active scheme, subregions of the default mode network, including the medial prefrontal cortex and hippocampus, contribute to offloading task representations into an unprioritized state and subsequently maintain sustained representations of these rules across trials via latent neural codes. Behavioral predictions using on-task, trial-wise and sustained, block-wise representations further support the neural dissociation. These findings unveil a dual neural system with distinct coding schemes that jointly enable cognitive flexibility for task implementation and switching.