PROJECT SUMMARY/ABSTRACT The ability to quickly learn novel cognitive procedures from instructions – rapid instructed task learning (RITL; “rittle”) – is substantially better in humans than other species and state-of-the-art artificial intelligence algorithms. Despite how essential RITL is for life success (e.g., education, using new technology), and despite the fact that it is impaired in a variety of brain disorders (e.g., schizophrenia), we lack basic understanding of how brain network processes generate RITL abilities. We recently found that RITL is implemented via cognitive control networks (CCNs), brain network communities organized around lateral prefrontal cortex, given that they rapidly and systematically shift their activity patterns according to novel task demands. What remains lacking, however, is knowledge of how these activity and connectivity reconfigurations support both the generalization and task- specific coordination of cognitive representations necessary for rapid transfer of prior learning to novel tasks. There is, therefore, a critical need to determine how RITL is made possible by CCN-based representation interactions. Without such information, the promise of cognitive and network neuroscience for understanding the neural basis of RITL will likely remain limited, blocking future treatments for RITL-related cognitive control deficit. The long-term goal is to understand how RITL and related forms of cognitive control are implemented in the human brain and develop interventions to reduce RITL deficits. The overall objective of this proposal, which is the next step toward the long-term goal, is to determine how CCN-linked representation interactions enable RITL abilities. The central hypothesis is that CCNs enable RITL via compositional-conjunctive hierarchies – splitting tasks into compositional representations that can later be reused and recombined hierarchically via conjunctive representations, allowing immediate transfer of prior learning to novel instructed tasks. Tractability here has been enhanced by new cognitive paradigms and network neuroscience tools such as activity flow modeling that now enable investigations of how cognitive representations interact via neural connections. Each aim focuses on a RITL processing stage: encoding (Aim 1), maintenance (Aim 2), and implementation (Aim 3). Each aim will be pursued using distinct cognitive manipulations but a common set of analysis tools: representational similarity analysis and activity flow modeling applied to both fMRI and high-density EEG data. At the completion of the proposed research, expected outcomes are to have determined how distinct network processes during encoding, maintenance, and implementation of novel instructions dynamically shift activity flowing from stimulus to response to enable novel task performance. These results are expected to have a positive impact as they will provide a new network-based understanding of the dynamic working memory processes underlying especially flexible cognitive processes during RITL, with applicability to understanding the neural basis of cognitive control processes impaired in a variety of brain disorders.

Project Narrative We aim to utilize the tools of network science to understand how instructed learning (e.g., during psychotherapy) is implemented in human brain networks, from initial learning to expertise after practice. Understanding the brain network and information-transformation processes underlying instructed learning (and subsequent practice) will provide vital clues to the brain’s dynamic transformation from injured to rehabilitated in a wide variety of psychiatric diseases.