Project summary The capacity of the visual system is limited: several studies have shown that we can only extract detailed information about a handful of objects at a time. Despite these limitations, people subjectively report rich perceptual experiences and they demonstrate a sophisticated ability to navigate the visual environment. One mechanism that potentially accounts for this discrepancy is ensemble perception: the ability to summarize large amounts of information that exceed the limits of attention. A growing body of research now suggests that people extract the statistical mean from large groups of objects, for example they can report the average size and speed of objects in the visual environment, as well as average expression in a crowd of faces. Although there are several behavioral experiments investigating the speed, efficiency and automaticity of ensemble coding, the neural substrates of this mechanism remains largely unexplored. In particular, the areas involved in ensemble coding, as well as the timing by which ensemble properties are computed remain unclear. The goal of this proposal is to test the hypothesis that ensemble properties are represented in early visual areas (Aim 1), and are computed at early stages of visual processing (Aim 2). One set of experiments (Aim 1) will use functional Magnetic Resonance Imaging (fMRI) and Multivariate Pattern Analysis (MVPA) to decode whether early visual areas contain information about the specific mean value that participants perceive at a given time. Another set of experiments (Aim 2) will use Electroencephalography (EEG) to investigate the timing of ensemble perception, and to test the hypothesis that ensemble perception occurs before the processing of individual object details. Overall, the results will shed light on the neural representation of ensemble perception, and will advance our understanding of the mechanisms by which the visual system extracts large amounts of information from complex scenes without focused attention.