Abstract
It is essential for object recognition that visual information is appropriately combined. To explain stages of perceptual organization that group elements into contours, the concept of association fields has been invoked. Local elements within the boundaries of an association field are grouped to give rise to the perception of a contour if they are appropriately aligned, reasonably similar and close. However, the size of this spatial window remains unclear, as well as how this changes with visual field location. To address this, we studied the combined influence of eccentricity and inter-element spacing on contour detection. Our findings indicate a clear difference in the processing of contours between the parafovea and the periphery. Contour integration in parafoveal regions is efficient and highly stable across a wide range of inter-element spacings and levels of orientation noise. In the periphery, efficient integration is only observed for elements close enough to fall within adjacent receptive fields, while increased inter-element spacings and orientation changes lead to a failure of contour integration. We conclude that two distinct mechanisms underlie contour integration, each with its own spatial extent and tolerance to noise-with the efficient, association field-like mechanism being a characteristic of central vision.