Journal of Experimental Psychology: Human Perception and Performance 1989, Vol. 15, No. 4, 771-784 Copyright 1989 by the American PsychologicalAssociation, Inc. 0096-1523/89/$00.75 Visual Gap and Offset Discrimination and Its Relation to Categorical Identification in Brief Line-Element Displays David H. Foster Department of Communication and Neuroscience University of Keele, Keele, Staffordshire, England Mario Ferraro Istituto di Fisica Superiore Universit~ di Torino, Turin, Italy Visual processing was investigated in judgments of relative line position. Stimulus continua were generated by bisecting a straight line and displacing the segments. Experiment l measured discrimination of pairs of longitudinally displaced segments at equal steps along the continuum. At long (2 s) durations discrimination fell smoothly, but at short ( 100 ms) durations it was sharp- peaked. In Experiment 2 the short-duration stimuli were labeled with subsets of the labels no gap, just a gap, and more than just a gap. Theoretical discrimination performances were computed and the one based on no gap and just a gap closely fitted observed performance. Experiments 3 and 4 were similar to 1 and 2, with lateral replacing longitudinal displacement. Similar "categor- ical" performance was obtained. It was concluded that there are discrete mechanisms for early detection of relative line position and that 2 labels can be used to characterize performance in each direction. Judgments about the relative positions of objects and parts of objects play a fundamental role in the analysis of visual scenes and the generation of responses to those scenes. Deci- sions about the relative positions of edges and lines are particularly important, both in the early stages of visual processing, when they may help to define the boundaries of objects and how they should be separated from each other and their backgrounds, and in the later stages, when they may form the basis of more complex object descriptions. How well can judgments about relative position be made? The investigation of the limits on this kind of visual sensitivity has a long history (Andrews, Butcher, & Buckley, 1973; Baker & Bryan, 1912; Berry, 1948; Fechner, 1860; Volkmann, 1863; Westheimer & McKee, 1977; WOlfing, 1892). Traditional measurements of performance have been of the vernier-acuity type: A straight line is bisected and the task is to detect the lateral displacement of one of the segments relative to the other (e.g., Baker & Bryan, 1912; Berry, 1948; Sullivan, Oatley, & Sutherland, 1972). Under suitable conditions, dis- placement thresholds of a few seconds of arc visual angle are usually achieved, an order of magnitude finer than the angular separation of adjacent foveal cones. Performance of this order is not, however, peculiar to judgments of lateral offset, and similarly fine hyperacuity performance (Westheimer, 1975) has been obtained in a variety of judgments of relative posi- tion, including, for example, the alignment of three points (Ludvigh, 1953), the separation of two parallel lines (West- heimer & McKee, 1977), and the spacing of three parallel lines (Baker & Bryan, 1912; Klein & Levi, 1985). We thank James E. Cutting, Peter C. Dodwell, Bruno Repp, Arthur G. Samuel, and anonymous reviewers for helpful comments on an earlier version of this article. This work was assisted by awards from the Medical Research Council, UK, and the Consiglio Nazionale delle Ricerche, Italy. Correspondence concerning this article should be addressed to David H. Foster, Department of Communication and Neuroscience, University of Keele, Keele, Staffordshire ST5 5BG, England. 771 Hyperacuity levels of performance are usually achieved with displays of long duration and focused attention. In natural viewing, these conditions may not be obtained: Scenes may be fixated only briefly as the eye pauses between one saccade and the next, and there may be a large number of objects in the scene competing for attention during a fixation. Although there are few data currently available on the effects of reduced exposure duration on judgments of relative posi- tion (e.g., Burbeck, 1986), results from experiments on target detection in brief multielement displays suggest that there may be profound changes in the organization of spatial sen- sitivity during what has been referred to as preattentive vision. Beck and Ambler (1972) measured the discriminability of abutting line segments forming L and T figures in peripherally presented arrays and showed that when displays were brief, and followed by a poststimulus mask (which controlled the time available to attend selectively to individual figures), subjects became less sensitive to relative line position (i.e., whether a figure was L or T) than to line orientation (vertical, horizontal, or oblique). A similar result was obtained when subjects had to discriminate texture displays made up of these elements (Beck, 1966, 1982). This result was also obtained in measurements of pattern-specific adaptation effects with sim- ilar kinds of displays (Foster & Mason, 1980). An exception to the finding of reduced sensitivity to line arrangements in preattentive vision was reported by Bergen and Julesz (1983), who demonstrated that although L and T figures were indeed poorly segregated in texture displays, L and intersecting line segments forming plus sign (+) figures were well segregated. Yet carefully controlled search-time measurements by Treisman and Gormican (1988) failed to obtain evidence that the intersection property (and the prop- erties of juncture and convergence or parallelism) was preat- tentively coded. They suggested (Tresiman & Gormican, 1988) that "no functional feature detectors exist that respond uniquely to properties of line arrangements, at least at the parallel preattentive levels of early vision" (p. 34). Studies on such early visual processing have traditionally used stimuli drawn from small classes, often comprising just