Vision: Cognitive Influences

To be useful in guiding behavior, perceptual systems must derive representations of the world that allow for fast, accurate responses to the world. Perceptual systems draw on various processes to optimize behavior. For example, early feature detection processes in vision can detect cooccurrences between features, and detecting these co-occurrences permits rapid grouping of features into longer contours. But, cognitive influences [Page 1049]also exist in vision, as when our expectancies or prior experience influence vision. These cognitive influences also can optimize behavior by altering visual representations.

There are many empirical demonstrations of cognitive influences in vision, and several examples are discussed in this entry. However, more important than demonstrations of cognitive influences are how these cognitive influences operate and affect vision. Does cognition influence vision only after visual perceptual processes have completed, as a way to “clean up” perception? Or, does cognition influence these perceptual processes as they operate, altering the operation and outcome of perception directly?

The foregoing questions point out that a still-standing issue in visual perception is the manner in which cognition influences perception. There are two dominant theoretical approaches to understanding the role of cognition in vision. One view, the bottom-up account, assumes that vision proceeds in a bottom-up or feedforward manner in which higher-order percepts are created from combinations of lower-level features. Strong versions of this bottom-up account propose that early visual processes are unaffected by—that is, encapsulated from—cognition and other later-level processes, as Zenon Pylyshyn, among others, has argued. In short, observers' expectations should not affect what they perceive. Under such a view, cognitive influences in vision are postperceptual and occur after visual processes have completed.

In contrast, the other view, the interactive account, proposes that vision operates in an interactive manner, in which both bottom-up information from the external world and top-down information (e.g., an observer's goals, expectancies, or prior experience) combine to determine visual perception. According to this interactive account, top-down information can be used to guide perception in the face of incomplete bottom-up information, and this top-down influence operates on perceptual processes themselves. In this way, vision can be viewed as performing what Hermann von Helmholtz termed “unconscious inferences” or unconscious problem solving.

The literature is full of arguments for both bottom-up and interactive accounts of various visual processes. The persuasiveness of these arguments rests on a number of methodological issues that arise when distinguishing bottom-up and interactive accounts. In short, it can be difficult to understand the mechanism that produces a cognitive influence in vision with psychophysical data because many experiments can be interpreted in terms of both bottom-up and interactive accounts. Given this, an instructive approach is to focus on how these cognitive influences shape vision. To understand how cognition affects vision—whether in a bottom-up or an interactive manner—requires an understanding of the methods used to investigate cognitive influences and how the various mechanisms can be disentangled. Many of the relevant methodological issues appear in early studies of cognitive influences in vision, including some of the earliest studies on figure-ground perception.