Interpolation processes in the perception of real and illusory contours.

The spatial and temporal characteristics of mechanisms that bridge gaps between line segments were determined. The presentation time that was necessary for localisation and identification of a triangular shape made up of pacmen, pacmen with lines, lines, line segments (corners), or pacmen with circles (amodal completion) was measured. The triangle was embedded in a field of distractors made up of the same components but at random orientations. Subjects had to indicate whether the triangle was on the left or on the right of the display (localisation) and whether it was pointing upward or downward (identification). Poststimulus masks consisted of pinwheels for the pacmen stimuli or wheels defined by lines. Stimuli were presented on a grey background and defined by luminance or isoluminant contrast. Thresholds were fastest when the triangle was defined by real contours, as for the pacmen with lines (105 ms) and the lines only (92 ms), slightly slower for corners (118 ms) and pacmen (136 ms), and much slower for the amodally completed pacmen (285 ms). For all inducer types localisation was about 20 ms faster than identification. In a second experiment the relative length of the gap between inducers was varied. Thresholds increased as a function of gap length, indicating that the gaps between the inducers need to be interpolated. There was no significant difference in the speed of this interpolation process between the pacman stimuli and the line-segment stimuli. About 40 ms were required to interpolate 1 deg of visual angle, corresponding to about one third of the distance between inducers. In a third experiment, it was found that processing of isoluminant stimuli was as fast as for low-contrast luminance stimuli, when targets were defined by real contours (lines), but much slower for illusory contours (pacmen). The conclusion is that the time necessary to interpolate a contour depends greatly on the spatial configuration of the stimulus. Since interpolation is faster for the line-segment stimuli, which do not elicit the percept of an illusory contour, the interpolation process seems to be independent of the formation of illusory contours.