Project Summary/Abstract Imagine a baseball outfielder tracking a fly ball. To catch it successfully, the player must not get distracted by a passing bird, even if it has a similar size and apparent motion similar to the ball. This is a challenging example of stimulus selection and maintenance in the presence of distractors, but we all do this sort of task every moment. Yet how brains internally represent multiple objects and coordinate activities across diverse neural populations to perform the selection process has remained elusive. In mammalian brains, the sheer number, complex projections and distribution of neurons involved in stimulus selection processes (e.g., attention) make it challenging to visualize all their connections to one another and decipher the dynamics of the network. The fruit fly Drosophila melanogaster will naturally orient to (select) and track (maintain selection of) an object among many in a virtual reality arena, a behavior influenced by the fly’s navigation system. This unambiguous readout in a numerically simple brain (~150,000 neurons) provides an unparalleled opportunity to investigate how multiple distinct populations of neurons mediate stimulus selection and maintenance. Further, Drosophila’s rich genetic tools make it feasible and straightforward to reproducibly label and manipulate the same population of neurons from animal to animal in physiological experiments. The proposed studies build a quantitative understanding, at the microcircuit level and ultimately across multiple neural populations, of: (1) how an animal’s selection of a stimulus among many manifests across multiple levels of visual processing, (2) how the stimulus selection decision impinges on information processing in the upstream visual neurons, and, finally, (3) how internal brain states such as hunger and locomotion modulate stimulus selection. This proposal describes a novel and powerful experimental approach. We will use multi-color two-photon calcium imaging to record the activity of distinct neuronal populations that constitute the fly's navigation system while monitoring the fly’s flight behavior in a virtual reality arena. The technique we will develop also incorporates simultaneous, high resolution (spatial and temporal) optogenetic stimulation, which allows us to probe multi-population neural dynamics and interactions with single-cell resolution. We will develop single-cell level computational models and test how the population activity of visual (ring) neurons and feedback from landmark-selective (compass) neurons interact with each other and the fly’s internal states. This project's successful execution will provide the first concrete understanding of multi-population dynamics underlying stimulus selection.