A long standing goal in my laboratory has been to understand the neural mechanisms that allow us to represent and identify spatial patterns. Our first line of research concerned retinal adaptation mechanisms, which preserve spatial information despite the enormous range of ambient light levels that occur in the natural environment (e.g., Geisler, 1978a,b Vis. Res.; 1981, J Physio; 1983, Vis. Res.; Hahn & Geisler, 1995; Kortum & Geisler, 1995). This work provided some of the first psychophysical evidence (in detection and discrimination experiments) for multiplicative, subtractive and response-saturation adaptation mechanisms, and has helped to establish them as key components in psychophysical models of light and dark adaptation. Our next line of research concerned sequential ideal observer analysis (Geisler, 1984; Davila & Geisler, 1985; Banks et al., 1987; Geisler, 1989; Arnow & Geisler, 1996). This work showed that a number of interesting effects in human pattern detection and discrimination can be explained by variations in the information available at the level of the photoreceptor responses or at the level of the ganglion cell responses (see also the button labeled Bayesian Ideal Observers).

More recent work (in collaboration with Albrecht’s lab) has been concerned with cortical mechanisms of pattern vision (Albrecht & Geisler, 1991; 1994; Geisler & Albrecht, 1992; 1997; 2001). Much of this work has been directed at developing a functional quantitative description of the responses of single neurons in primary visual cortex, with the aim of understanding how the population of cortical neurons represents spatial and contrast information and how the responses of the population contribute to the discrimination and identification performance of the organism as a whole. The functional description of individual cortical neurons that has emerged from our work and related work in the literature consists of four components a linear filter (summation of excitation and inhibition) that gives a cortical cell its basic stimulus selectivity, a divisive contrast normalization mechanism that preserves selectivity while producing response saturation, an expansive response exponent that enhances the selectivity, and a noise mechanism that makes the variance of the response proportional to the mean response. Although the responses of cortical neurons are well-described by these mechanisms there is a great deal of heterogeneity from cell to cell in the parameters of linear filter and in the parameters of the non-linear contrast normalization and response exponent. To characterize this heterogeneity we have extracted, from our large database of single cell recordings, probability distributions (across cells) for the parameter values describing the linear and non-linear mechanisms. Using these probability distributions we have created “neuron sampling models,” which allow us to simulate the population response of primary visual cortex. We have found that these models, which are based directly on the distributions of cortical cell properties, predict behavioral detection and discrimination performance along a number of fundamental stimulus dimensions.