Primate head-free saccade generator implements a desired (post-VOR) eye position command by anticipating intended head motion. J. Neurophysiol. 78: 2811-2816, 1997. When we glance between objects, the brain ultimately controls gaze direction in space. However, it is currently unclear how this is allocated into separate commands for eye and head movement. To determine the role of desired final eye position commands, and their coordination with intended head movement, we trained three monkeys to make large gaze shifts while wearing opaque goggles with a monocular 8 degrees aperture. Animals eventually developed a new set of context-dependent eye-head coordination strategies, in particular expanding the head range and compressing the eye-in-head range toward the aperture (while wearing the goggles). However, when we shifted the location of the aperture to a different subsection of the normal head-free oculomotor range (by covering the original aperture and creating a new one), eye-head saccades failed to acquire visual targets, because they continued to drive the eye ultimately toward the now occluded original aperture. Even when a head-stationary saccade acquired the new aperture, subsequent head-free saccades drove the eye eccentrically toward a point that anticipated the intended head movement, such that the subsequent vestibuloocular reflex slow phase brought the eye onto the location of the original aperture. Animals could only acquire the new aperture consistently after several days of retraining. These results suggest that 1) eye-head coordination is achieved by a plastic, context-dependent neural operator that uses information about initial eye/head position and intended movement to compute desired combinations of final eye/head position and 2) acquisition of these positions involves sophisticated anticipatory compensations for subsequent movement components, akin to those observed previously in complex oral and manual behaviors.