Abstract
Fields of illumination exist in the external world, but to what extent are they represented in the visual lightness software? Cartier-Bresson's photograph Trastevere is sliced diagonally into adjacent regions of sunlight and shadow. We pasted identical gray disks onto a digital version of Trastevere, to probe the lightness computation at 12 locations in the scene. Separate groups of fifteen observers viewed the CRT image and made Munsell matches for each disk in 6 different conditions. For identical round disks, the disks in the shadowed region appeared 2.13 Munsell units lighter than those in sunlight, and lightness values were roughly homogeneous within each of these two regions, regardless of distance from the boundary. This effect grew significantly larger (2.67 units) when disk size and shape were made to conform to depicted perspective, and larger still (3.13 units) when disk boundaries were blurred to conform to photograph graininess. Viewed through a pinhole, known to enhance depth perception, the difference grew further to 4.07 units. When gray disks were pasted onto the glass CRT screen, the difference shrank to 1.49 units. This framework difference effect thus varies with the strength to which the disks appear to belong to their surrounding framework. Control conditions ruled out a local contrast explanation. Our manipulations affected only the disks in shadow, not those in sunlight. When the entire sunlit region of the image was blacked out, the disks appeared white. These results are consistent with predictions of anchoring theory, but difficult to reconcile with spatial filtering models.
Supported by National Science Foundation BCS-0236701 and Public Health Service GM 60826-02