Grouping and detecting vertices in 2-D, 3-D, and quasi-3-D objects.

Abstract

Abstract In two experiments, observers searched for a misaligned vertex in displays depicting one or more twodimensional (2-D), three-dimensional (3-D), and quasi-3-D objects that were drawn with unconnected vertices. Serial search functions were obtained for all three conditions, even after much practice. In addition, search among the 2-D objects was more efficient than search among either 3-D or quasi-3-D objects, which did not differ after practice. This finding suggests that differences in depicted dimensionality cannot explain the difference in performance between the 2-D and 3-D conditions, but differences in stimulus complexity can. Further, after practice with the 2-D displays, observers could detect a misaligned vertex as rapidly among two objects as they could with one-object displays. These findings have implications for models of recognition that assign a critical role to vertices in early vision, as well as the notion that mechanisms underlying preattentive processing differ from those underlying the automaticity of processing that develops with practice. The processes subserving recognition are usually categorized into two broad types: Preattentive processes occur relatively early, in parallel, automatically, and with few resource demands, whereas attentive processes occur later, serially, with focused attention, and with a higher consumption of resources. Further, the analytic function of preattentive processes - to detect simple features throughout the visual field -- differs from the synthetic function of attentive processes, which is to conjoin features from particular locations in the visual field into object files (e.g., Treisman, 1982,1988, 1991, 1993; Treisman & Gelade, 1980; Treisman & Gormican, 1988; Treisman & Sato, 1990; Treisman dc Souther, 1985). Within the context of a visual search task, the original empirical signature of preattentive (parallel) search was a reaction time (Rt function that did not vary with the number of distractors in the display, whereas the signature of attentive (serial) search was a search function that increased with the number of distractors (c.f., Sternberg,1966, 1969). Recent evidence indicates that the type of information accessible in early vision might be more complex than originally supposed (Downing & Pinker, 1985; Nakamaya & Silverman, 1986b; Ramachadran, 1988; Rensink & Enns, 1995). For example, relatively flat search functions have been reported for detecting characteristics of objects that can only be defined by conjunctions of simple image features (Donnelly, Humphreys, & Riddoch, 1991; Elder & Zucker, 1993; Enns & Rensink, 1991; Rensink & Enns,1995). This evidence is inconsistent with the view that focused attention is required to detect conjunctions (e.g., Brown, Weisstein, & May, 1992; Cohen, 1993; Egeth, Virzi, & Garbart, 1984; Epstein & Babler, 1990; Humphreys, Quinlan, & Riddoch, 1989; McLeod, Driver, & Crisp, 1988; Treisman & Sato, 1990; Wolfe, Cave, & Franzel, 1989). Evidence that feature conjunctions can be detected early in processing has important implications for models of recognition. In the present paper, we focus on one type of feature conjunction-vertices. Vertices represent an interesting case because it has been proposed that a conjunction of vertices into simple forms may play a critical role in the formation of descriptions derived from nonaccidental properties, such as colinearity, closure, and (Donnelly et al., 1991, p. 562). Thus, our main goal was to examine further the role of vertices in the rapid extraction of an object's two-dimensional (2-D) or three-dimensional (3-D) structure. To do so, we blended the stimuli and methodology used in two studies that found evidence for rapid detection of conjoined features (Donnelly et al., 1991; Enns & Rensink, 1991). The first study, by Donnelly et aL (1991), exploited Gestalt principles of closure and good continuation to show that observers could rapidly detect misaligned vertices in line drawings of 2-D shapes made from just unconnected vertices. …