PROJECT SUMMARY In the mouse retina, about 40 types of retinal ganglion cells (RGCs) communicate visual information to the rest of the brain. A great deal of processing takes place before RGCs send their output downstream. Some RGCs respond selectively to a narrow range of shapes, contrasts, and directions of motion or prefer localized stimuli that move differentially from their surroundings. These computations are supported by interactions between more than a hundred interneurons whose interactions give rise to the receptive fields (RFs) that describe the relationship between the stimulus to the response of the RGC. However, despite significant recent advances in the field, we still do not know what visual features are detected by the majority of RGC types. One obstacle to progress is current techniques to study RF composition, which either require prolonged recording sessions, challenging experimental techniques, or fail to detect crucial RF components. We are also limited in the conceptual understanding of how neural circuit organization translates to function and what RF motifs give rise to specific visual computations. In this proposal, we will take an innovative approach that combines machine learning techniques, biophysically realistic modeling, electrophysiology, and glutamate / calcium imaging to develop a comprehensive description of the visual abilities of multiple RGC types in complex visual scenes that is grounded in empirical data. The proposed research will substantially advance our understanding of basic and advanced response characteristics of visually active cells, opening new horizons in the examination of neuronal function in and beyond the retina.