Abstract
Adaptation is often equated with sensitivity regulation (as in a multi-scale meter) and/or resetting of a baseline (AC coupling). A more sophisticated form of adaptation might deform the input-output functions of the visual system in more complex ways, so as to maximize the average precision with which environmental inputs are represented. But the simple operations of zero adjustment and sensitivity scaling are able to achieve this goal fairly well when applied both to intensity, and (at later, but still intermediate stages of processing) to contrast. The formal simplicity of these operations conceals interesting and useful complexity: contrast gain as well as intensity gain are regulated through changes in temporal and spatial integration. The disruptive effects of photon noise at low luminance, and of fluctuations in nerve impulse counts at low contrast, can be reduced by spatial and temporal integration, but at high luminance or high contrast more rapid local analysis becomes possible, at the affordable price of reduced gain for luminance or contrast.
Dark adaptation can be slow because it doesn't, in nature, need to be fast; it is less obvious why light adaptation is so fast, but such fast adaptation may help create the representation of contrast during fixational eye movements.
In addition to its benefits for discrimination and detection, spatial contrast adaptation can help compensate for prior optical or neural losses present on short or long time scales. But deprivation can alternatively lead to added central losses. Perhaps inputs that are attenuated but still provide systematic signals may be boosted, while inputs so weak as to be noise-dominated are lost.