Computation of lightness in simple and complex retinal images Lightness refers to the perceived black, gray, or white shade of a surface. The simplest image that creates an impression of lightness consists of two surfaces that fill the entire visual field. This can be achieved by placing an observer's head inside a large hemisphere, the interior of which is painted with two shades of gray. A series of experiments has revealed three rules that appear to completely describe the computation of lightness under such conditions. Most importantly, the highest luminance appears white and serves as the standard for darker shades. The other two rules concern relative area (the larger, the lighter) and luminance range (the perceived range tends to normalize on the full black/white range). These three rules can be applied to frameworks of illumination that are embedded within complex images, given one additional provision. In complex images, lightness is never computed exclusively within a single framework. Rather it is a weighted average of values computed both within its own local framework and one or more other frameworks, either adjacent or superordinate. This is the forgotten principle of co-determination, proposed by Kardos in 1934. A crucial strength of this theoretical approach, called anchoring theory (Gilchirst, et al, 1999) is that it provides an impressive account of just the pattern of lightness errors made by human observers.