Illusory-contour formation affected by luminance contrast polarity

We report a new type of illusory contour (Illusory-O) whose formation is contingent upon the contrast polarity of its juxtaposed inducing elements being similar, ie both elements must either be positive or negative in contrast sign. To test the hypothesis that this contingency is primarily dictated by factors that determine amodal surface completion (occlusion) between the inducing elements we conducted a series of experiments employing known spatial properties of the amodal completion mechanism, to show that spatial conditions unfavorable to occlusion lead to a concurrent weakening of the Illusory-O formation. For instance, we found that when the juxtaposed inducing elements (solid rectangles) were spatially misaligned, or when their spatial separation increased, our observers rated the perception of the Illusory-O as reduced. We also showed that, in addition to using solid-form inducing elements, the Illusory-O can be induced by line terminals, as long as these lines respect the requirements of the amodal completion mechanism such as similar contrast polarity and spatial alignment. Then we demonstrated that the role of the amodal completion mechanism is not limited to our particular arrangement of inducing elements by showing that the formation of the illusory Necker cube also relies on similar contrast polarity. Finally, to explain why some illusory contours like the Illusory-O are dependent on contrast polarity while others (eg Kanizsa square) are not, we propose that the key rests upon the visual system's presumption of occlusion. That is, in forming the illusory contour, if the visual system infers that it is a byproduct of the inducing elements being occluded, then having inducing elements of similar contrast polarity becomes a prerequisite. This assumption can be traced to the occurrence in the real world where partially occluded objects usually have visible parts (on both ends) with similar contrast polarity. Along this line of thinking, we suggest a plausible neural circuitry that may be implemented to form both contrast polarity sensitive and insensitive types of illusory contours.