In a prior study we reported that the human translational vestibulo-ocular reflex (tVOR) in response to vertical (bob) 2 Hz oscillations generated eye rotations of only 60% of those required to keep the eyes pointed at a stationary visual target, whether located at near (approximately 17 cm) or far (2 m). Best responses occurred in ambient illumination and we concluded that relative image motion between the target and background was an important determinant of tVOR behaviour. To investigate further how visual conditions influenced tVOR, we measured responses as subjects binocularly viewed the bridge of their own nose in a mirror at approximately 8.5 cm, a visual condition that required similar convergence to viewing the near target, but cancellation of tVOR. Median tVOR cancellation gain [(near-viewing response-mirror viewing response)/near-viewing response] was 0.81 (range 0.55-0.97), which was substantially greater than the gain of smooth visual tracking of a large visual display moving at 2 Hz (median gain 0.27, range 0.09-0.42). Thus, visual inputs other than smooth tracking must contribute to tVOR cancellation. We then compared tVOR response to 2 Hz bob as subjects fixed upon a visual target at 17 cm and viewed a large textured background at 1.5 m that was either stationary or moving at 2.1 Hz. Vertical eye rotations waxed and waned as a function of the difference between platform and background oscillations. These findings support our hypothesis that tVOR evolved not to stabilize the image of the target on the fovea, but rather to minimize retinal image motion between objects lying in different planes, in order to optimize motion parallax information. A geometrically based optimization function is proposed to account for tVOR responses at different target distances.