Visual Attention
Ronald A. Rensink, Departments of Psychology and Computer Science, University of British Columbia, Vancouver BC, Canada.

In L Nadel (ed.), Encyclopedia of Cognitive Science. (pp. 509-515). London: Nature Publishing Group. 2003.   [pdf]

Visual attention is the factor controlling the selective access and integration of visual information.


Although much of vision appears to be effortless and all-encompassing, there nevertheless exist limits to what it can do. Consider, for example, air traffic control, where it is imperative to keep track of all moving items in a display (corresponding to the airplanes in an airspace). If only a single item is present, it can generally be tracked without problem. It is also possible to track four or five items simultaneously, although some effort is needed. But for twenty or thirty items, even a maximal effort will not suffice, and the task must be shared among several controllers. What appears to be happening in such cases is that visual perception is constrained by a consciously-controlled factor within the observer, a factor that enables certain kinds of processing to take place, but which is limited in the extent to which it can be applied. This factor is referred to as visual attention.

  Interestingly, even though most observers immediately know what to do when asked to "pay attention" to a stimulus, it has been rather difficult to give this an objective characterization. Indeed, until recently there was no general consensus on the basic function of attention--at various times, it was associated with such things as clarity of perception, intensity of perception, consciousness, and selection.

  During the past few decades, however, great progress has been attained by focusing on selection as the basic function of visual attention. Two kinds of selection appear to be particularly important. The first is selective access, i.e., allowing only certain parts or properties of the input to be inputs to later processes. It was initially believed that selective access protected the processors at higher levels from being overwhelmed by too much information. However, more recent work tends to view selective access as a way to delimit control of various actionsÑfor example, focusing on the locations and sizes of nearby items that are relevant to guiding a hand as it reaches for an object.

  The second kind of selection is selective integration, i.e., combining selected parts of the input into new structures that then form the basis of further processing. For example, three adjoining lines could be combined into a complete figure; this figure (and not the lines themselves) might then provide the basis for subsequent control of grasping. It was initially believed that such integration had to be selective in order to make good use of a limited amount of processing "resource". However, the idea of an undifferentiated resource has been largely abandoned. It is also becoming clear that there is little reason to retain most of integrated structures. Consequently, the concept of selective integration is often recast in terms of the selective co-ordination of the outputs of multiple processing streams.

  In this view, then, visual attention is not a unitary faculty. Instead, it is simply the control of information in the visual system, achieved in various ways by various processes. When considered from this perspective, some of the unresolved issues in earlier treatments of visual attention simply vanish. One example is the issue of whether selection is "early" or "late" (i.e., whether it acts on simple, precategorical structures, or more complex ones). Given that perception may be carried out by multiple systems that operate concurrently, there need no longer be a strict processing order, or a single site where selection could act.

  Since a complete understanding of visual attention is still a long way off, this article only surveys some of the behavioral techniques used for its exploration and some of the results obtained, but does not attempt a complete synthesis. It also focusses on the purely visual aspects of the processes involved—i.e., on how the stimuli themselves are handled, rather than how responses to them are generated. Issues such as the sequencing of multiple responses are considered to involve central control at higher levels, and so will not be discussed here.

Back to main publications list.