Abstract
A new kind of perceptual grouping is proposed. Without any reference to light source location or intensity, the visual system can effectively take the illumination into account by perceptually grouping together patches of the retinal image that have the same illumination level. This proposed grouping process is orthogonal to the more conventional grouping in which parts of an object lying in different fields of illumination are grouped together. Adjacent, coplanar retinal patches sharing a sharp border produce the strongest illumination group. The strength of this group is reduced as the patches are separated laterally by a gap (while still coplanar), separated in depth by a gap (while still parallel) or rotated out of the same plane (while still adjacent). We varied the proximity between a target patch and the patch with the highest luminance in the same plane. (1) When these patches were separated by a gap, lightness was strongly dependent upon proximity. As the gap between coplanar patches increases, the likelihood increases that an illumination boundary falls within the gap, and thus their grouping by illumination weakens. (2) When the patches were connected by other coplanar surfaces, proximity made no difference. Connected patches remain strongly grouped for illumination regardless of intervening distance because any illumination boundary will be revealed by the connecting patches. (3) When the highest luminance surrounded the target, separating them with a gap produced only half the proximity effect, presumably due to the reduced likelihood that an illumination boundary falls just within the circular gap. Grouping by illumination explains these results and many others.