One of the most fundamental problems in visual perception and cognition is information overload. Given limited processing capacity, there's simply too much perceptual information competing for the control of thought and action. Hence, observers must use selective attention to rapidly prioritize which aspects of a complex scene are most relevant to behavior. For instance, drivers must be able to efficiently locate and identify traffic signals and stop signs amidst a dizzying mosaic of visual information. People are highly adept at this, while artificial systems are not. This is because human observers employ powerful top-down knowledge to guide their perceptual interactions with the complex environment. Top-down knowledge is useful because the visual world is highly structured. For example, spatial layout of landmarks are relatively stable, certain objects tend to covary with others (toasters with ovens, desks with lamps, etc.), and even moving objects such as cars tend to move around in systematic, predictable ways. Sensitivity to this structure, presented to observers in the form of global visual context, provides useful constraints on visual processing. Importantly, this contextual information must be "learned" by perceivers. Thus, the goal of this project is to establish the importance of contextual information and learning mechanisms in object recognition and visual search. The first set of studies examines how visual contexts are defined, how contextual information is learned and represented, and how learned information is applied to new instances. The second set of studies examines how temporal context information influences visual processing, and whether learning affects other visual processes such as the ability to segregate figure from ground. Successful completion of these studies will provide important insights into top-down processing and learning mechanisms in human vision. This understanding will be applicable to the design of more effective artificial systems, enabling these to benefit from perceptual experience as people do.
