Abstract
Illusory contours (ICs) are a useful tool to study scene segmentation. In particular, ICs embody two apparently conflicting needs of segmentation: they require integrating information over large distances (‘global’), and concurrently show sensitivity to minute image details like junction structure (‘local’). Understanding this global/local interplay has progressed slowly, partly due to a lack of reliable methods to measure IC completion. We devised a measure of the presence and/or strength of a bounding IC based on performance in a dot localization task. A probe dot was presented briefly near the boundary of a perceptually completed region. Observers reported whether the dot fell inside or outside the region. The slopes of the psychometric functions (fraction ‘out’ responses vs dot position) provide an estimate of the sharpness of the perceived boundary; biases indicate position.
We examined four classes of stimuli: (1) Classical kaniza and other illusory figures (IC stimuli), (2) stimuli where the IC inducers had been altered slightly in ways that reduce the subjective impression of ICs, (3) luminance-defined contours (‘Real Contour’, RC), (4) control stimuli to assess performance based on no-contour alignment information. Results show that ICs can support boundary localization almost as well as RCs (thresholds: 9.2 and 6.0 min-arc respectively). Small manipulations to the local features (e.g. rounding the sharp L-junctions) of the IC stimuli significantly reduced performance to that obtained when only alignment information was provided (thresholds: 19.9 and 16.1 min-arc respectively). Our results are consistent with previous data obtained by subjective report (rating methods), but extend them further and, importantly, provide a more sensitive method for assessing the effect of stimulus manipulations on completion processes.