Dynamic coordination and segregation mechanisms in higher cortex for parallel task processing

Navigating dynamic environments often requires the brain to process information from multiple tasks in parallel. Yet, how neural resources are allocated across tasks and reorganized with practice remains unclear. Here, we combine an original dual-task paradigm in mice with chronic two-photon imaging, optogenetic manipulations, and recurrent neural network models to dissect the underlying cortical dynamics. We show that interference between tasks arises not only from bottlenecks among task-shared neurons but also from reduced activity in non-shared populations. Intriguingly, this reduced activity is associated with rapid coordination between tasks and supports early dual-task success. With training, performance is optimized by a multi-level reorganization, including the recruitment of specialized neurons and progressive segregation of task representations. Network models on the same paradigm implementing these coordination and segregation schemes accelerate dual-task learning, indicating their functional importance. Together, these findings demonstrate how cortical circuits flexibly redistribute and restructure resources to support parallel task processing.