Abstract
Abstract: :
Purpose: Illusory contours are a common phenomenon of visual interpolation, and are thought to play an important role in scene segmentation. Although the perceptual phenomenology of illusory contour processing is well understood, little is known about the underlying neural mechanisms. It has recently been proposed (Stanley and Rubin, Neuron 2003) that neurons in the lateral occipital complex (LOC) first extract a crude representation of an enclosed "salient region" which is then fed back to first–tier visual areas to guide the extraction of illusory contour borders. This model predicts that salient–region (surface–related) activity should precede the extraction illusory contours (border information). Methods: Illusory surfaces/contours were generated by an array of notched circles (inducers) on a gray background. When the notched circles were aligned (0 deg), illusory squares appeared. When the notched circles were mis–aligned (e.g. 20 deg, 50 deg and 70 deg), the illusory squares disappeared. Rounding off the sharp corners of the notched area of the inducers, as in the Stanley and Rubin experiment, greatly reduced the perception of illusory contours, but left "salient regions" intact. VEP responses were measured for inducers that supported illusory contours plus salient regions, or just salient regions. In the test conditions, inducers alternated between aligned (0 deg) and mis–aligned states (50 deg), while in the control conditions, alternation was between two mis–aligned states (20 deg and 70 deg) in which neither illusory contours or salient regions were apparent. Results: Responses to all conditions consisted of a positive peak (P1) near 100 msec and a negative peak (N1) near 170 msec. The N1 response to full–cue illusory contours was larger than in any of the other three stimulus conditions. The P1 and N1 responses in the salient region test condition without illusory contours did not differ from either control condition, but there was a sustained negativity around 200 msec that was absent in the other three conditions. Conclusions: The Stanley and Rubin model predicts that differential responses to the salient region vs control conditions should precede those generated by the full cue stimulus. We find the opposite: the earliest differential activity occurs in the full–cue situation on the leading edge of the N1 response, with salient region responses occurring well after the N1 peak. Our data thus do not support Stanley and Rubin's theory that feedback from the LOC about the location of salient regions may direct resource–intensive contour completion processes in early visual cortex.
Keywords: shape, form, contour, object perception • electrophysiology: non-clinical • visual cortex