The visual responses of retinal X-cells, recorded as s-potentials in the lateral geniculate nucleus (LGN), and geniculate X-cells, recorded as action potentials in the LGN, were studied in urethane anesthetized cats. We tested two widely held hypotheses: that the receptive field is (1) separable into spatial and temporal factors, and (2) even symmetric in space. The tests were applied to the amplitudes and phases of responses to sinusoidal gratings which drifted across the receptive field. The X-cell responses failed the test for spatiotemporal separability, and cases of spatial asymmetry were observed. A modified "Difference of gaussians" (mod DOG) model proved useful in the interpretation of the amplitude and phase data. Application of the mod DOG model to the amplitude and phase data revealed the existence of three forms of spatiotemporal coupling. Changes in the temporal frequency of the stimulus can change (1) the ratio between center and surround strengths, (2) the difference between center and surround phases, and (3) the spatial extent of the surround mechanism. Our results led us to a new view of receptive field organization in the X-cell: the center mechanism is spatially homogeneous in its dynamics whereas the surround mechanism is spatially inhomogeneous.