Stabilization of images on the fovea during either fore/aft translation of a subject or fore/aft movement of a visual target in front of a stationary observer imposes complex geometrical requirements that depend upon the eccentricity of the object of interest with respect to the eyes. Each eye needs to be rotated independently with varying proportions of conjugate (version) and disconjugate (vergence) eye movements to maintain fixation of the target. Here, we describe binocular coordination in the early response to translational movements of normal subjects along their naso-occipital axis. We recorded the responses evoked by small (about 4 cm), abrupt (about 0.7 g), fore/aft translations in four normal subjects while they viewed a near target. In the forward and backward starting positions the target was 15 or 10.5 cm away, respectively. Each subject was tested with the target centered between the eyes, aligned on the right eye, and placed to the right of the right eye by 3 cm. The three conditions differed only in the lateral eccentricity of the target, yet the geometrical requirements for image stabilization are very different: pure vergence, one eye still, or mostly version. We found that the eye-movement responses closely matched what was needed for visual stabilization of the target, though responses to stimuli calling for divergence were less accurate than those for convergence. The latency of these responses ranged from 40 to 65 ms and achieved about 80% of the ideal response by 90 to 100 ms after the onset of the stimulus. Next, we asked whether these eye movements were generated by the vestibular system or by high-level strategies for image stabilization, such as pursuit. Thus, in a second set of experiments we used the mean profile of fore\aft body motion computed for each subject to drive a small visual target across the same distances and in the same eccentricities used during body translations. We found that visually driven responses had longer latencies (by at least 80 ms, ranging from 144 to 155 ms) and slower dynamics (with significantly lower peak eye velocities), highlighting the different subsystems producing the two types of responses. Saccades were also an important component of the response to both visual and vestibular stimuli, less frequent during the centered-target configuration and more frequent during viewing of eccentric targets. Visual stimuli evoked saccadic corrections more often and at shorter latencies than did vestibular stimuli. Both smooth and saccadic eye movements were appropriately disconjugate and their pattern depended on whether the eyes were converging or diverging.