Abstract
We characterized the rich temporal dynamics of color processing using a continuous tracking paradigm. Specifically, we estimated temporal impulse response functions associated with tracking chromatic Gabor patches, and measured how the lag of these functions changes as a function of chromatic direction and contrast. Three observers tracked, with a cursor, the position of a chromatic Gabor conducting a horizontal random walk with a mean velocity of ~4 deg/sec. The Gabor was composed of two vertically-oriented 1 cpd sinewave modulations (one L-cone-, one S-cone-directed) windowed by a 2° Gaussian. The contrast ratio of these two cone-directed components defines a polar angle (i.e. a color direction) in cone contrast space. We measured tracking performance for stimuli in twelve color directions, with six log-spaced contrasts in each direction (72 conditions). To determine how tracking performance varied with color direction and contrast, we first computed the cross-correlation between target and tracking velocities. This yields an estimate of the impulse response function associated with the signals that drive tracking. From parametrized fits to the empirical cross-correlation functions, we estimated the temporal lag in the tracking signal for each stimulus condition. For all subjects, we found that i) temporal lag decreases as contrast increases, for all color directions and ii) nominally L-cone isolating stimuli are associated with smaller lags than nominally S-cone isolating stimuli, when contrast is equated. A model based on a single underlying chromatic mechanism accounts for the data well. In this model, lag is determined by a weighted sum of nominal L- and S-cone contrasts passed through a decaying exponential. Weights on L-cone contrast were 30-60x higher than weights on S-cone contrast. The model predicts that there is a color direction, near the nominal S-cone isolating direction, for which tracking is not possible. This prediction remains to be tested.