1. We measured the response of macaque ganglion cells to sinusoidally modulated red and green lights as the relative phase, theta, of the lights was varied. 2. At low frequencies, red-green ganglion cells of the parvocellular (PC-) pathway with opponent inputs from middle-wavelength sensitive (M-) and long-wavelength sensitive (L-) cones were minimally sensitive to luminance modulation (theta = 0 deg) and maximally sensitive to chromatic modulation (theta = 180 deg). With increasing frequency, the phase, theta, of minimal amplitude gradually changed, in opposite directions for cells with M- and L-cone centres. 3. At high frequencies (at and above 20 Hz), phasic cells of the magnocellular (MC-) pathway were maximally responsive when theta approximately 0 deg and minimally responsive when theta approximately 180 deg, as expected from an achromatic mechanism. At lower frequencies, the phase of minimal response shifted, for both on- and off-centre cells, to values of theta intermediate between 0 and 180 deg. This phase asymmetry was absent if the centre alone was stimulated with a small field. 4. For PC-pathway cells, it was possible to provide an account of response phase as a function of theta, using a model involving three parameters; phases of the L- and M-cone mechanisms and a L/M cone weighting term. For red-green cells, the phase parameters were monotonically related to temporal frequency and revealed a centre-surround phase difference. The phase difference was linear with a slope of 1-3 deg Hz-1. If this represents a latency difference, it would be 3-8 ms. Otherwise, temporal properties of the M- and L-cones appeared similar if not identical. By addition of a scaling term, the model could be extended to give an adequate account of the amplitude of responses. 5. We were able to activate selectively the surrounds of cells with short-wavelength (S-) cone input to their centres, and so were able to assess L/M cone weighting to the surround. M- and L-cone inputs added linearly for most cells. On average, the weighting corresponded to the Judd modification of the luminosity function although there was considerable inter-cell variability. 6. To account for results from MC-pathway cells, it was necessary to postulate a cone-opponent, chromatic input to their surrounds. We developed a receptive field model with linear summation of M- and L-cones to centre and surround, and with an additional M,L-cone opponent input to the surround.(ABSTRACT TRUNCATED AT 400 WORDS)