No abstract

This article presents a theory of selective attention that is intended to account for the identification of a visual shape in a cluttered display. The selected area of attention is assumed to be controlled by a filter that operates on the location information in a display. The location information selected by the filter in turn determines the feature information that is to be identified. Changes in location of the selected area are assumed to be governed by a gradient of processing resources. Data from three new experiments are fit more parsimoniously by a gradient model than by a moving-spotlight model. The theory is applied to experiments in the recent literature concerned with precuing locations in the visual field, and to the issue of attentional and automatic processing in the identification of words. Finally, data from neuroanatomical experiments are reviewed to suggest ways that the theory might be realized in the primate brain. The identification of shapes and objects in the environment plays an important adaptive role in our daily, moment-to-moment activities. A typical visual scene contains many objects, but there is a limit on the number of objects that we can process at one time. This limitation implies that, at some stage or stages in the information flow through the system, the information arising from some objects must be momentarily excluded from processing. This exclusion from processing may occur by operations that either enhance the information from a target object, or by operations that suppress the information from the distractor objects, or by operations that do both. Described in this manner, exclusion from processing closely resembles what traditionally has been termed selective attention. The notion of selective attention gained a foothold in the mainstream of psychology in the late 1950s, particularly through the influence of Broadbent's (1958) filter theory and Guthrie's (1959) revision of his learning theory ("what is being noticed becomes the signal for what is being done"). Soon aRerward, a controversy arose concerning whether the selection process occurs early (Broadbent, 1958) or late (Deutsch & Deutsch, 1963; Norman, 1968) in the flow of information. The two contrasting views of the locus of selection can be seen in theoretical issues raised two decades later (for reviews, cf. Broadbent, 1982; Johnston & Dark, 1986; Kahneman & Treisman, 1984; Shiffrin, 1988). A strong form of the late-selection theories assumes that all objects in a visual display (falling on an effective area of the receptor surface) are identified and that the selection process chooses the identified object that will be processed fur-We thank E. G. Jones for his helpful comments concerning neuroanatomy, and we thank Ryozo Yoshino for his suggestions in mathematical matters. We also thank Marc Carter, Jan I.awry, and Dale McNulty for reading the manuscript and making valuable comments and criticisms.

The spatial extent of attention to visually presented letters and words was investigated using a probe technique. The primary task required subjects to categorize (a) five-letter words, or to categorize the middle letter of (b) five-letter words or (c) five-letter nonwords. The probe task required the subjects to respond when the digit 7 appeared in one of the five letter positions. Probe trials were inserted at the onset of letter and word processing in Experiment 1 and 500 msec after letter and word processing in Experiment 2. In both experiments, probe trials produced a V-shaped function of reaction times across probe positions for the letter-categorization task for word and nonword stimulus conditions. In contrast, a relatively flat reaction time function was found for the word-categorization tasks. An analysis of the data based on a quantitative model of attentional spotlight distributions suggests that the spotlight width in the letter tasks is one letter space, and the spotlight width in the word task is typically five spaces.

Positron emission tomographic scans were recorded from human subjects following an object-identification task, one version of which required attentional selection and the other version of which did not. In one experimental session, the attention-demanding displays were presented in the left visual field and the nonattention displays were presented in the right visual field. In a second session, the sides of the displays were reversed. Analysis of the scans indicated that, averaged across the 2 sessions, the pulvinar showed greater glucose uptake when it was contralateral to the display of the selective attention task than when it was contralateral to the display of the nonattention task. The pattern of the data indicated that the degree of the attention task effect on pulvinar glucose uptake may differ between the hemispheres. In view of known connections between the pulvinar and cortical areas that mediate object identification, the present finding suggests that the pulvinar operates interactively with these cortical structures when an identification process demands selective attention.

No abstract