Defining the neural correlates of ensemble representation in humans are sensitive to the summary statistics of complex scenes. The visual system readily extracts summary statistical representations across a host of visual domains, including orientation, direction of motion, size, position, and expression, and can do so efficiently and implicitly. Summary statistical extraction, or ensemble coding, operates over both space and time, and shows sensitivity to the average, the range, and even the variance of sets of objects. Despite the seeming ubiquitous and fundamental nature of ensemble representations, the precise level at which ensembles are extracted remains unknown, and there has been no examination of the neural correlates of this process. These issues will be explored using low-level behavioral psychophysics and functional neuroimaging techniques. The proposed experiments will delineate elements key to understanding the role summary statistics play in vision, including: a) does the visual system adapt to ensemble stimuli (e.g., sets of oriented gabor patches, sets of faces) in the same way that it adapts to singletons (e.g., a single oriented gabor patch, a single face)? b) What parts of the brain play a role in calculating/extracting summary statistical information? The broader objective of the proposed studies is to develop expertise in functional magnetic resonance imaging techniques, ranging from design (e.g., block, fast event-related) to analysis (e.g., deconvolution, pattern classification).
Understanding ensemble representations is tantamount to understanding object recognition. Given the importance ensembles seem to play in visual perception (e.g., texture segregation, scene perception), knowing their implementation may lead to greater understanding of visual deficits, such as object agnosias. If ensemble representation were preserved in the presence of visual deficits (such as object agnosia), ensembles may be used to support alternative clinical treatments.