We have studied the manner by which inputs from the two eyes are combined in simple cells of the cat's visual cortex. The stimuli for this study are drifting sinusoidal gratings, shown dichoptically at optimal spatial frequency and orientation. The relative spatial phase (disparity) between the gratings for left and right eyes is varied over 360 degrees. Most simple cells show phase-specific binocular interaction such that response amplitudes and phases vary depending on the relative spatial phase. At one phase, response is greater than either of the monocular responses and often greater than the sum of the two. At the phase 180 degrees away from the optimal, the cell's responses are strongly inhibited and often completely suppressed. Phase-specific binocular interaction disappears when the gratings presented to one eye are made orthogonal to the optimal orientation. The degree of binocular interaction does not depend critically on the ocular dominance of the cells. Simple cells that are nearly equally dominated by each eye always exhibit strong phase-specific interaction. The majority of cells that are strongly dominated by one eye, and even those that appear monocular, show phase-dependent changes in responses. We examined the extent of binocular interaction for cells with preferred orientations near vertical compared with those tuned to other optimal orientations. If these cells are conveying information about depth, one might expect a greater degree of binocular phase-specificity for units preferring nearly vertical orientations, which would then be processing horizontal disparities. We find no evidence for this. Predictions of simple-cell responses are derived from a linear model of binocular convergence in which light-evoked neural signals from each eye are summed linearly to determine cell responses. Data from cells generally follow the prediction of the model for both response amplitude and phase. Deviations from predictions of the linear model are found for a minority of cells. This deviation may be accounted for by a threshold mechanism that comes into play after the linear binocular summation. A small proportion of simple cells that appear monocular by alternate tests of each eye show a purely inhibitory influence from the silent eye. This inhibition is not generally dependent on the relative phase of the gratings. We conclude that most binocular interaction in striate simple cells may be accounted for by linear summation of neural signals from each eye.(ABSTRACT TRUNCATED AT 400 WORDS)