Is light in pictures presumed to come from the left side?

For good ecological reasons humans assume that the illumination in pictures comes from overhead rather than from below. Recent work has also raised the possibility that the preferred angle of illumination is between 20 degrees and 30 degrees to the left of vertical, although the ecological basis for that is obscure. We describe two studies of this question. In study 1, twenty subjects looked in free, unrestricted vision, at a picture of a single 'bubble' which appeared either convex if the illumination was from above, or concave if the illumination was from below. There was no evidence in study 1 that illumination from the left produced a different appearance from illumination from the right, the preferred angle of illumination being 0.5 degrees to the right of vertical. In study 2, on forty subjects, there was fast presentation (200 ms) of an array of 16 bubbles, one of which, the target, was illuminated from the opposite direction to the others, and hence 'popped out' from the display, appearing concave if the others were convex, or vice-versa. The preferred angle of illumination for detection of the target was about 14 degrees to the left of vertical in study 2. The estimates from studies 1 and 2 did not differ significantly, and when combined gave a preferred angle of 9 degrees to the left of vertical. The reasons for the seeming discrepancy between studies 1 and 2, or the smaller overall effect than that found in previous studies are not clear since the methodologies seem equivalent. In both studies 1 and 2 there was a highly significant association between the direction of spontaneous head tilt and preferred direction of illumination, and this factor has not previously been taken into account in studies of the phenomenon. Contrary to the study of Sun and Perona (1998, Nature Neuroscience 1 183-184), there was no association with handedness or other measures of functional lateralisation.