AbstractThe computation performed by a neural circuit depends on how it integrates its input signals into an output of its own. In the retina, ganglion cells integrate visual information over time, space, and chromatic channels. Unlike the former two, chromatic integration in the retina is largely unexplored. Analogous to classical studies of spatial integration, we here studied chromatic integration in mouse retina by identifying chromatic stimuli for which activation from one cone type is maximally balanced by deactivation in the other cone type. This revealed nonlinear chromatic integration in subsets of On, Off, and On-Off ganglion cells. Nonlinear On cells differed from the other two classes by displaying response suppression rather than activation under balanced chromatic stimulation. Furthermore, nonlinear chromatic integration occurs independently of nonlinear spatial integration, depends on inhibitory signals from the receptive field surround, and may provide information about chromatic boundaries, such as the skyline in natural scenes.