Abstract
The ability to detect a contour among randomly aligned elements has been studied extensively and led to the proposal of the association field: a spatial window in which elements can be linked into contours. Human vision is presumed to depend on two other distinct visual fields: receptive fields which constrain the low-level processing of contrast, and integration fields which determine the spatial windows in which elements can be integrated into a common texture and lose their individual features. Theories of mid-level vision would benefit from having a common framework that encompasses neural processing across these different fields. However, it is notoriously hard to study whether association fields are modulated by spatial properties similarly to perceptive fields and integration fields. The difficultly does not only stem from the longer extent over which elements can be combined, but also from the confoundability between effects of element eccentricity, inter-element spacing, and display density. In the current study, these factors were manipulated in a way that allows assessment of their respective influence on contour integration. Participants were presented with a spatial 2AFC “snake in the grass” task with three eccentricities (7°, 10.5° and 15.75°), five inter-element spacings (0.7°,1.05°,1.58°,2.36° and 3.54°), and three orientation “wiggles” (5°, 15° and 25°). “Snakes” consisted of seven elements which were equally distant from fixation. We found that at moderate eccentricities (7° and 10.5°), contour integration was successful over the full range of inter-element spacings. Yet, at a farther eccentricity (15.75°) elements needed to be closer, especially for the strongest “wiggle”. This implies that unlike integration field sizes, association field sizes do not scale with eccentricity, with smaller association fields in the periphery.