Abstract In many cognitive processes, information is processed in a parallel manner across many brain regions. This is thought to make our cognitive abilities highly tolerant to perturbations or neuron-loss because disrupted processes are compensated by other redundant neurons coding the same information. Yet, it remains poorly understood how interconnected networks of neurons are organized into redundant representations to produce robustness. We recently discovered that persistent activity in mouse frontal cortex during short-term memory is remarkably robust to perturbations. The two hemispheres of cortex are organized into redundant modules where each one can independently maintain persistent activity. When one suffers a perturbation, signals from the other hemisphere help restore the activity. Modularity and redundancy may be a general organizing principle of information processing and storage in cortical circuits. Understanding the neurobiology of this novel and potentially fundamental organizing principle will have deep implications for design of neural manipulation strategies and understanding behavioral manifestation of chronic neurodegeneration. In the proposed research, we will establish and validate a novel framework that combines computational modeling, population recording, and targeted optogenetic perturbations to identify, probe, and chronically track modular organization of cortical circuits. Our goal is to dissect redundant modular organizations within and across brain areas, obtain a deeper understanding of how multiple redundant modules coordinate information to support robust behavior, and how such modular organization is shaped by learning to manifests in robust behavior.