Abstract
Studies of the effects of noise have been useful in understanding pattern perception, but the form of TvC functions in noise is not established, and models do not account for all the facts (Pelli, 1985; Legge, Kersten & Burgess, 1987). We measured TvC functions for a small Gaussian spot and a narrow 4 c/deg Gabor pattern in 0, 0.44, and 1.77 microdeg2sec static noise. There were three observers. TvC functions decrease and then increase as pedestal contrast increases. Noise both increases thresholds and changes the form of the TvC function. As noise level increases the pedestal contrast at which the threshold is minimum increases causing the functions to come together and often cross at high contrast. So there is a range of high contrasts over which noise has essentially no effect or a facilitatory effect. These data are well described by an extension of a model designed to account for the effect of dissimilar pattern masks (Foley, 1994). In this model the noise not only adds variance to the visual signal, it also produces a divisive inhibitory input to the detecting mechanism resulting in a change in the form of the response function which steepens it and shifts the most sensitive part of its range to higher contrast, thus overcoming the masking effect of the noise. To further examine the effects of noise, we determined classification images for the Gaussian spot for Cped = 0, threshold, and 0.25. The classification images had a center -surround form and were essentially the same except for a difference between the stimulus present and stimulus absent images in the 0 pedestal case. These images are consistent with our model and inconsistent with models in which the pedestal produces an equivalent noise sufficient to make the stimulus noise negligible.