Differential effect of contrast polarity reversals in closed squares and open L-junctions.

Mark M Schira,Branka Spehar

doi:10.3389/fpsyg.2011.00047

Abstract

Scene segmentation depends on interaction between geometrical and photometric factors. It has been shown that reversals in contrast polarity at points of highest orientation discontinuity along closed contours significantly impair shape discrimination performance, while changes in contrast polarity at straight(er) contour segments do not have such deleterious effects (Spehar, 2002). Here we employ (semi) high resolution fMRI (1.5 mm × 1.5 mm × 1.5 mm) to investigate the neuronal substrate underlying these perception effects. Stimuli consisted of simple elements (a) squares with contrast reversals along straight segments; (b) squares with contrast reversals in the corner (highest orientation discontinuity); (c) L-Junctions with contrast reversals along the straight ends; (d) L-Junctions with contrast reversals in the corner. Element with contrast polarity reversals are easy to distinguish though appear geometrically equivalent. For squares with contrast polarity reversals only along straight lines we find significantly lower BOLD modulation compared to any of the control conditions, which show similar responses to each other. In the light of previous psychophysical work (Elder and Zucker, 1993; Spehar, 2002) we speculate that this effect is due to closure perception. We observe this across a wide range of areas on occipital cortex.

Highlights

The visual system exhibits a remarkable ability to recover considerable portions of objects’ boundaries that are routinely missing in complex natural scenes due to occlusion, camouflage, or low visibility
We generally find moderate increase in BOLD responses to the experimental condition relative to baseline; each condition resulted in an average signal modulation off 0.35% relative to the Fixation condition
In summary we find a significant difference for shape and an interaction in V1, V2, and V3, a significant effect of Contrast Reversal in V3A/B and no significant effects in hV4 and occipital lateral fovea (OLF)/LO1

Summary

Introduction

The visual system exhibits a remarkable ability to recover considerable portions of objects’ boundaries that are routinely missing in complex natural scenes due to occlusion, camouflage, or low visibility. A central feature of these models is that the contour linking mechanisms operate on outputs of oriented spatiotemporal filters; a notion that is consistent with preferential linking of collinear or nearly collinear neighboring elements (Hess et al, 1998; Hess and Field, 1999). These assertions have been well supported at the neurophysiological level, where the same geometrical features have been shown to influence contextual interactions observed at earliest stages of cortical processing of contour fragments (Li and Gilbert, 2002)

Objectives

Methods

Results

Discussion

Conclusion