Electrophysiological tests of neural models: evidence for nonlinear binocular interactions in humans.

Evoked cortical potentials were used to test neural models of binocular interactions in humans. The critical feature that distinguished between basic models was whether the two monocular signals combined linearly or nonlinearly. Stimuli were presented separately and simultaneously to two eyes with the relative temporal phase parametrically varied. Fourier analysis was applied to the recorded electroencephalogram to measure the frequency response of the visual cortex to this stimulation. Comparisons of the models' predictions with the experimental results indicate that the greatest contribution to the cortical response originates in a neural pathway that contains nonlinear binocular combination, and a smaller contribution originates in a pathway with linear binocular combination. Current models, based on psychophysical studies of humans and on single-cell neurophysiological investigations of cats, are inadequate to explain these results. A model of binocular interaction is proposed that does account for the present findings and is also supported by neurophysiological studies on monkeys. This model, which contains both nonlinear and linear binocular interactions, has profound implications for perceptual processing.