VISUAL CODING AND THE DEPLOYMENT OF ATTENTION

点击翻译按钮获取中文摘要

DESCRIPTION (Applicant's Abstract): Attention modulates visual perception at many levels, beginning at least as early as extra-striate areas. This project distinguishes several modes of attentional deployment and explores their effects on visual coding. The earliest stages of processing probably operate automatically and in parallel; they are said to be "preattentive". They have only indirect effects on behavior, since all perception depends on some deployment of attention, whether focused on one item or spread globally over the whole scene. Feature integration theory (e.g., Treisman, 1993) proposed that preattentive processing results in a set of feature maps and in a master map of locations. In order to integrate the information in these separate representations, focused attention is required. When attention is spread globally to encompass areas containing more that one element, it makes available global boundaries, and general properties like the direction of illumination, or the slant of the terrain in the scene as a whole. The question arises what information is available, either explicitly or implicitly, about individual elements with each of these modes of attention. Various experimental paradigms will be used to explore the issue. The first study will compare the degree of interdependence, if any, of feature identification and feature localization when tested with global attention and with inattention. A series of experiments will explore whether the asymmetries of feature coding found in search tasks (Treisman & Gormican, 1988) are present only with global attention or whether they appear also in task requiring focused attention. Earlier studies have suggested course coding of feature dimensions at the preattentive level; an experiment will test whether this applies also to unattended stimuli. The dependence of binding errors on the number of instances of a given feature will test whether features are preattentively individuated. Several studies of statistical coding of multi-element display are proposed, to throw light on the information that is automatically extracted when attention is globally deployed. The cumulation of information from changing display will be tested in various search tasks. Implicit processing of high-load display will be tested through experiments of priming from "unseen" elements, on perceptual learning of feature and location contingencies in search, and on the effects of experience on the spontaneous control of attention. The goal in all these experiments is to increase our understanding of the effects of attention on visual processing.