To compute lightness, illumination is not estimated, it is held constant

Abstract

The light reaching the eye from a surface does not indicate the black-gray-white shade of a surface (called lightness) because the effects of illumination level are confounded with the reflectance of the surface. Rotating a gray paper relative to a light source alters its luminance (intensity of light reaching the eye) but the lightness of the paper remains relatively constant. Recent publications have argued, as had Helmholtz (1866/1924), that the visual system unconsciously estimates the direction and intensity of the light source. We report experiments in which this theory was pitted against an alternative theory according to which illumination level and surface reflectance are disentangled by comparing only those surfaces that are equally illuminated, in other words, by holding illumination level constant. A 3-dimensional scene was created within which the rotation of a target surface would be expected to become darker gray according to the lighting estimation theory, but lighter gray according to the equi-illumination comparison theory, with results clearly favoring the latter. In a further experiment cues held to indicate light source direction (cast shadows, attached shadows, and glossy highlights) were completely eliminated and yet this had no effect on the results. (PsycINFO Database Record

Figure 1

Katz’s method of anomalous illumination. Observers adjusted the gray of one color wheel to match the gray of another wheel positioned at a different slant relative to the window. The screen with two holes was used only to equate the luminance of the two wheels at the outset. (From Woodworth, 1938).

Figure 2

Left: Simulated display used by Boyaci et al (2003). Seven of the 12 objects were colored. Right: Actual scene used by Ripamonti et al (2004), with two target positions shown above (reprinted with permission from ARVO).

Figure 3

A target square appeared to be almost white when seen in the near plane (left image) but almost black when seen in the far plane. Black is 2.0 on the Munsell scale and white is 9.5. Note: This image illustrates the arrangements but does not capture the effect, due to its limited luminance range and unrealistic depth. Reprinted by permission from Gilchrist (2006).

Figure 4

For monocular observers, each tab appeared coplanar with its background panel and the upper and lower tabs were seen as approximately black and white respectively. Separate binocular observers saw the tabs as extending from the cube (as in the actual arrangement shown at left) with the upper and lower tabs seen as approximately white and black respectively. Luminance values are given in cd/M². Reprinted by permission from Gilchrist (2006).

Figure 5

Plan view of vision tunnel apparatus

Figure 6

Observer’s visual field in Experiment 1. The darker left face of the cube is scarcely visible in this photo but was clearly seen by observers. Seven of the 11 objects were colored.

Figure 7

Data from Experiment 1. As target/background angle increased from zero to 90 degrees, perceive lightness went up, not down. Error-bars indicate +/− 1 standard error of the mean (SEM).

Figure 8

Data from Experiment 2. Articulating the two backgrounds produced the same qualitative pattern with an increased slope. Error-bars indicate +/− 1 SEM.

Figure 9

Data from Experiments 2 and 3 compared. The same pattern of results was obtained even when all of the cues to the direction of illumination were eliminated. Error-bars indicate +/− 1 SEM.