Visual search, visual streams, and visual architectures

Abstract

Most psychological, physiological, and computational models of early vision suggest that retinal information is divided into a parallel set of feature modules. The dominant theories of visual search assume that these modules form a "blackboard" architecture: a set of independent representations that communicate only through a central processor. A review of research shows that blackboard-based theories, such as feature-integration theory, cannot easily explain the existing data. The experimental evidence is more consistent with a "network" architecture, which stresses that: (1) feature modules are directly connected to one another, (2) features and their locations are represented together, (3) feature detection and integration are not distinct processing stages, and (4) no executive control process, such as focal attention, is needed to integrate features. Attention is not a spotlight that synthesizes objects from raw features. Instead, it is better to conceptualize attention as an aperture which masks irrelevant visual information.Two factors make early vision a difficult computational problem. First, the solution space is combinatorially explosive because images contain a large number of feature dimensions. Second, computation must be rapid, so processing time is limited. Many theories suggest that the visual system solves these problems by means of a divideand-eonquer strategy; the retinal image is decomposed into an array of separate representations that are processed in parallel.

Visual search asymmetry for viewing direction

“…The results of the present experiments may be explained by a built-in mechanism that might function along the same lines as the direction-of-shading mechanism, investigated by Kleffner and Ramachandran (1992) and Aks and Enns (1992); it is difficult to explain the resulting asymmetry by the traditional view (Treisman & Gormican, 1988; see critique by Green, 1991). Treisman and Gormican's (1988) account for search asymmetries exploits the idea ofthe presence rather than the absence of an additional feature that distinguishes the target from the background stimuli.…”

“…Even though this measure might not be indicative for different underlying processing strategies (Green, 1991;Townsend, 1972), it can .be used to summarize the present data. The slopes were That is, this result again shows the existence of the search asymmetry for 3-D stimuli-processing was significantly easier when the target was tilted upward and the distractors downward, as compared with the reversed orientation.…”

Attention in visual search: Multiple search classes

“…can ATTENTION IN VISUAL SEARCH 137 be combined with information about serial versus parallel search. The conclusions in this paper add to the growing body of criticisms of various aspects of present search theories (Bravo & Blake, 1990;Briand & Klein, 1989;Green, 1991;Townsend & Ashby, 1983). These criticisms include, for example, (1) the need to define a priori what constitutes a feature, (2) the need for a dependent measure that reflects spatial attention while minimizing the influence of stimulus characteristics, such as interitem spacing, and (3) the need to consider other measures besides search rate.…”

“…Note that there is some supercapacity for these four target types: on the targetabsent trials for RY and RG and on the target-present trials for SLs and LIs. These negative slopes may be due to distractor grouping when more distractors are present (Bacon & Egeth, 1991;Duncan & Humphreys, 1989), to local comparisons (Green, 1991), or to comparative coding rather than single-target coding (Treisman & Gormican, 1988).…”

“…The early suggestion that features might be those targets for which there were distinct channels in the central nervous system (Julesz, 1982) may not account for some parallel searches, such as particular conjunctions of two or more features (Treisman & Sato, 1990). If parallel processing is mediated by single units, then there would need to be specialized units for conjunctions that are found with parallel search (such as color and form, color and motion, or color and binocular disparity; Green, 1991). However, Treisman and Sato found that no single feature determined the ease of search for conjunctions.…”