Natural scenes have become an important research topic in the computational vision community, but have until recently not been used in studies of visual cortical function. We have therefore begun recording from single cells in primate visual cortex during free viewing of natural scenes, and also during controlled experiments that mimic free viewing; we hope to use these procedures to evaluate current theories of processing in primary visual cortex. In one line of research we are evaluating a simple model of V1 processing that incorporates a classical complex center (CRF) as well as divisive contrast gain control and texture processing surrounds. Thus far we have been impressed with the difficulty of constructing a model that adequately accounts for the data. The fit appears to be very sensitive to the accuracy with which both the CRF and the impulse response function are described. The nonlinear components appear to play primarily a modulatory role. In a second line of study we are investigating the possibility of temporal coding in single visual cortical cells during natural vision. As a first step we have characterized spike trains as modulated Poisson processes, and have estimated the time constant of the modulator function. Our preliminary analysis suggests that these cells produce fairly reliable spikes under natural viewing conditions; actual spike trains are often statistically more reliable than artificial trains generated by a modulator smoothed by a Gaussian with a standard deviation of only 10 ms. This reliable spike timing occurs even though the average spike rate is generally lower than would be obtained with conventional controlled stimuli. We also hope to use our methods to reveal novel coding properties of higher visual processing stages, but these studies will inevitably be limited by our ability to characterize cells with highly nonlinear response properties.