For many years, my lab has been actively involved in developing and applying ideal observer theory in the study of perceptual systems. Ideal observer theory uses the concepts of Bayesian statistical decision theory to determine optimal performance in a task, given the physical properties of the stimuli and some biological constraints. Organisms generally do not perform optimally in any given task, and thus the aim of ideal observer theory is often not to model the performance of the organism per se, but instead to provide a precise measure of the stimulus information available to perform the task, to provide a computational theory of how to perform the task, and to serve as an appropriate benchmark against which to evaluate the performance of the organism. Bayesian ideal observer theory is a very powerful tool and provides the theoretical foundation for much of the work in my lab.

Our early work introduced the concept of sequential ideal observers and applied them in the domain of pattern detection and discrimination (Geisler, 1984; Geisler & Davila, 1985; Banks, Geisler & Bennett, 1987; Geisler, 1989). The central idea of sequential ideal observers is to allow inclusion of known anatomical and physiological constraints into the ideal observer analysis. Comparing optimal performance with and without the constraints provides a precise measure (and understanding) of the information processing limitations imposed by those constraints. Our early ideal observer work also uncovered some interesting physical laws; for example, the physical limit for spatially resolving two features decreases with the fourth root of stimulus intensity, and the physical limit for detecting changes in the relative location of two spatially separated features decreases with the square root of intensity. These laws go some way toward explaining, for example, “hyperacuity”—the ability to discriminate changes in spatial configuration that are far smaller than the spacing between receptors. More recent work has applied Bayesian ideal observer theory in analyzing neural processing in the retina (Arnow & Geisler, 1996) and in the primary visual cortex (Geisler, et al., 1991; Geisler & Albrecht, 1995; 1997; 2000). Most recently, we have been applying ideal observer theory in the study of perceptual tasks that involve complex natural stimuli (Klarquist, et al., 1994; Geisler et al., 2001) and in the study of the evolution of perceptual systems (Geisler & Diehl, 2003).