Combined eye-head movements are routinely used to orient the visual axis (gaze) rapidly in space. The gaze control system can be modeled using a feedback system in which an internally created instantaneous gaze position error signal equivalent to the distance between the target and the current gaze position is used to drive brainstem eye and head motor circuits. The visual axis is driven until this gaze position error (GPE) is zero. The neural structure of the feedback system is discussed here. The midbrain's superior colliculus (SC) is implicated in gaze control but its 'location' in the feedback circuitry is debated. Our moving hill hypothesis proposed that the SC is within the feedback loop and that GPE is encoded topographically by a moving locus of activity on the motor map. In cat, fixation neurons of the superior colliculus encode GPE, which supports this model. Our preliminary evidence in both monkey and cat shows that neurons on the motor map respond to and encode, at very short latency, gaze shift perturbations. This further supports the hypothesis that the SC is within the gaze feedback loop.