We have previously shown that rod signals influence the balance of both red-green and blue-yellow opponent hue dimensions, but in a pattern that is inconsistent with simple additive combinations of rod and cone signals in opponent-color models (e.g., rods can produce both a red bias and a green bias). The shifts are instead consistent with non-linear models in which rod influence requires non-zero cone signals. Here we challenge this model with new data on the time-course of the rod hue biases. The rod green bias (typically 5-15 nm) and rod blue bias (typically 20-30 nm) observed near unique yellow and unique green, respectively, were not systematically affected by test stimulus duration. Near unique blue, rods exerted a green bias (typically 5-10 nm), for test stimuli less than 50 ms, and a red bias (typically 5-15 nm) for longer test stimulus durations. The rod red bias reached asymptote for all observers within 50 ms. We find that the model accommodates well the group-averaged results but that there are important individual differences among the observers that pose challenges to be addressed in future.