Where is the Sun?

Abstract

It is well-known that the human visual system uses a light-from-above assumption when it interprets shaded stimuli as 3-D shapes. In our present experiments, we explore in more detail this lighting-from above assumption. We used a 2afc SOA paradigm with masking involving stimulus screens composed of 24 smoothly shaded circles arranged on a circular grid. Such shaded circles are typically interpreted as 3-D convexities or indentations that are lit from a direction that is parallel to the shading gradient. Background distractors were shaded in one orientation, thus simulating a consistent lighting direction, while the target pattern was shaded to simulate illumination fro the opposite direction. One target pattern was present during 50defined by the shading gradient orientation of the distractors, is held constant within a block and varied systematically between blocks. Data was collected from 12 naive subjects on a total of 12 lighting directions, spanning the 360-degree space evenly, using a staircase method in blocks of about 80 trials. Our results show that observers perform best when the distractor patterns are shaded to suggest a direction of lighting that is oblique rather than directly over-head. In particular, oblique lighting from the left, i.e. top-left lighting at around 20 degrees from the vertical, is preferred. This lighting-from-left preference is especially obvious when the light-from-left condition {90 degrees left of the vertical) is compared with the light-from-right condition. We find that both right and left handers show a top-left lighting preference, with right-handers preferring a significantly larger angle left of the vertical than do left-handers. Our experimental results are consistent with the observation that 3-D scenes are predominantly depicted with light from the top-left in classical paintings as well as in computer graphic 3-D scenes.