Multiplicative intrinsic noise and the limits to visual performance

Intrinsic noise sources in the visual system may conveniently be divided into those which are additive and those which are multiplicative. Additive noise is independent of the strength of the light signal, while the magnitude of multiplicative noise is proportional to the magnitude of the light signal. Additive noise has been thought to exist in the visual system since Fechner (1860) first described it with the term “Augenschwarz”. It has since been termed “dark light” by Barlow who has described both its effect upon visual threshold performance and some physiological correlates of its presence (e.g. 1957, 1965). Multiplicative noise, however, has been neglected as a potential source of intrinsic noise (as pointed out by Lillywhite and Laughlin, 1978) possibly because there was no known example to point to. This report discusses the effects of multiplicative intrinsic noise upon visual signals and the consequent limiting effects upon visual performance. There are four reasons for considering this topic now. Firstly, the presence of multiplicative intrinsic noise was recently demonstrated in an insect visual system (Lillywhite and Laughlin, 1979). Secondly, recent measurements of current responses in single vertebrate photoreceptors suggest that each effectively absorbed photon produces a variable or noisy response (Yau et al.. 1977). The noise is not a constant factor, independent of the light stimulus, but accompanies the response to each photon and hence multiplicati,ve intrinsic noise is also present in the vertebrate visual system. Thirdly, the traditional tests for whether performance is limited only by the random, Poisson nature of most light signals, such as the Rose-De Vries law and, more recently, Receiver Operating Characteristic (ROC) analysis. will be shown in this paper to be inadequate in the case of a visual system with multiplicative intrinsic noise, although the usefulness of ROC analysis as a measure of underlying signal distributions and performance is undisputed. Lastly, the presence of multiplicative intrinsic noise in the human visual system might explain why psychophysical measurements of human detective quantum efficiency are at least two (scotopic) to ten (photopic) times lower than the estimated capture efficiency of the photoreceptors, while performance is reportedly limited only by the random, Poisson nature of the light stimulus.