When we look at the world, how do we know which parts of the visual input belong to the same object and which do not? The process known as perceptual grouping takes elements of the visual input and combines them into what we experience as a visual scene that contains objects, people, plants, shadows, and so on. Most of the time perceptual grouping is involuntary but it can come under voluntary control. For this reason, the study of perceptual grouping is a part of the larger effort toward understanding consciousness. Although phenomena of perceptual grouping are an essential foundation of perception, they are often described using a list of qualitative "principles," such as proximity, similarity, and good continuation, that are vague and unquantified. Michael Kubovy at the University of Virginia and Sergei Gepshtein at the Salk Institute for Biological Studies are proposing to clarify some of the fundamental processes of perceptual grouping, using rigorous methods of measurement and modeling. The researchers start with simple visual patterns that allow them to study one force of perceptual grouping at a time. The individual forces of grouping will then be combined, using more complex visual patterns, with the goal to derive general quantitative laws of perceptual grouping. The researchers will study the interaction of geometric factors (such as proximity between elements of visual displays) and intensive factors (such as the luminance and contrast of the elements) in perceptual grouping. The laws of combination of grouping factors will be compared with the laws of combination of other sensory cues, which have been intensively studied in the perception of visual depth and in multisensory integration.
Software to be developed for this research program will be made accessible to the public. In addition to addressing fundamental issues in visual perception, the work has the potential to influence developments in visual media such as art, animation, and film.
Research funded by this grant was a combination of theoretical and experimental studies of visual perception. A central thread of this work was a new theoretical perspective in which aspects of perception traditionally studied under the rubrics of 'early vision' and 'mid-level vision' are governed by the same basic principles. The two key principles identified are computational and economic. - The computational principle is the tradeoff of precision of measurement known as ‘the uncertainty principle’ (Gabor, 1946). Capacities of individual visual cells are limited such that every cell cannot be optimized for measurement of location and content of stimuli. - The economic principle concerns selective allocation of limited neural resources (a limited number of such cells) to enable efficient perception of the great number of potentially important visual stimuli in the variable visual world. In this project we investigated how tracing interactions of these principles across levels of the visual process helps one to understand some visual phenomena that previously appeared puzzling and unrelated. An example of such phenomena in the 'mid-level vision' is the finding that parts of dynamic visual displays sometimes are more likely to from perceptual groups when they are closer to one another (grouping by proximity) and sometimes when they are farther apart (failure of grouping by proximity). An example in the 'early vision' is the previously puzzling finding that visual adaptation sometimes improves and sometimes impairs visual sensitivity to subsequent stimuli.
