Ideal discriminators in spatial vision: two-point stimuli.

An earlier paper [J. Opt. Soc. Am. A 1, 775 (1984)] described a general ideal discriminator (a Stimuli-Defined-Exactly observer) whose sensitivity is limited only by the initial sequence of mechanisms in the visual system--the optics of the eye, the receptor lattice, receptor optics, and photopigment spectral sensitivities. In the present paper, further properties of the model are derived, and a similar model (a Stimuli-Defined-Statistically observer) is developed for conditions with stimulus uncertainty. To test the predictions of the models, two-point intensity discrimination, resolution, and separation discrimination were measured as a function of point-source energy and background luminance. Three important differences between resolution and separation discrimination were found: (1) At moderate to high intensity levels, separation threshold is much smaller than resolution threshold. (2) At low intensity levels, resolution threshold is smaller than separation threshold. (3) Separation threshold decreases more rapidly than resolution threshold as a function of intensity. All three properties were predicted by the models. However, as expected, the quantitative fits are not accurate. The value of the models is that they allow one to determine, for almost arbitrary visual-discrimination tasks, what aspects of performance are accounted for by preneural factors. Furthermore, the models provide a precise metric of the discrimination information available at the receptors. This permits meaningful comparison of human performance across different tasks.