Changes in interactions between neurons enable diverse computations and flexible behaviors. Such changes can occur very rapidly by rerouting information flow through existing connections, or more slowly by updating connections. The proposed project will study how local and brain-wide dynamics arise during learning of goal-directed behaviors. Experiments will use novel ‘all-optical’ experimental techniques to causally map network interactions at cellular resolution in combination with data-constrained computational models, to follow the learning process in the living brain with unprecedented detail. The investigation will focus on learning mechanisms in several novel memory-guided behavioral tasks, that either do not require learning, or specifically tailored for studying learning within and over days. This will fundamentally advance the understanding of how different learning mechanisms shape brain-dynamics and behavior. Aim 1: Mapping changes in causal interactions (effective connectivity) between neurons in local cortical circuits. Modeling and experiments will allow disentangling contributions of synaptic plasticity and gating to changes in network interactions and representations during learning. Aim 2: Investigating unique properties of cortex-wide neural activity. Preliminary work, based on cellular-resolution mesoscopic imaging of ~1,000,000 neurons, led to the discovery that spatial and temporal scales of brain-wide dynamics follow a power-law. Intriguingly, the most dominant modes of activity are global and fast, differently from any existing network model. The proposed work will uncover biological mechanisms supporting the emergence of these newly discovered cortical states during learning. Aim 3: Investigating functional implications of learned neural network dynamics studied in Aims 1 and 2. To test the hypothesis that such dynamics enable animals to perform flexible memory-guided behaviors, work will focus on modeling the effect of targeted optogenetic perturbations of neural activity on different spatial scales on network dynamics and behavior. Overall, the proposed collaborative study will leverage the PIs complementary expertise, to deepen the understanding of mechanisms and function of neural dynamics on different spatial scales. The experimental and theoretical methods developed as part of this proposal will provide insight for brain-wide control of memory-guided behavior, and will serve as a road-map for future studies of other behaviors controlled by distributed brain networks.