Abstract
An important first step in vision is the perceptual grouping of image elements that belong to an object and their segregation from elements that belong to different objects and the background. Previous research on how image elements are grouped together has yielded conflicting results. On the one hand, so-called “pathfinder” studies where the elements of a single curve have to be grouped on the basis of their colinearity concluded that this is accomplished in parallel across the visual field. In contrast, in so-called “curve-tracing” tasks where subjects have to judge whether items are located on a target or a distractor curve, a serial and time-consuming process seems to be involved. Here we attempt to resolve this apparent conflict. Subjects had to indicate which of two cues was located on a target curve. We varied the number of curves (the target curve alone or together with a distractor) and the presence or absence of random background elements, to test the processes involved in grouping and segregation. We show that when a curve has to be segregated from the background (figure-ground segregation), this can be accomplished in parallel. In contrast, when a target curve has to be segregated from another curve that is equally colinear (figure-figure segregation), a serial and time-consuming process is involved. This suggests that figure-ground segregation is solved by local operators that are sensitive to the degree of colinearity between elements. In the case of figure-figure segregation, however, an additional process is required that combines the local groupings to form a globally coherent representation of an entire curve. Moreover, we show that background elements reduce the speed of figure-figure segregation, which implies that figure-ground and figure-figure segregation interact.