We investigated visual-vestibular interactions in normal humans, where a constant speed of optokinetic stimulation was combined with whole body oscillation of lateral linear acceleration (10 m stroke). The acceleration mode was not sinusoidal, but rectangular (step). The pure optokinetic reflex (reference OKR) and the OKR under combined stimulation (combined OKR) were induced by a random-dot pattern projected onto a hemispherical dome-screen affixed to a chair on a linear accelerator. The translational vestibulo-ocular reflex (tVOR) was determined separately in the dark during acceleration-step oscillation. Since the tVOR was masked by the OKR under combined stimulation, the interaction was assessed as changes in combined-OKR velocity at two segments of opposing acceleration; in other words, tVOR directions identical to (agonistic) and opposite to (antagonistic) the OKR direction. When a moderate optokinetic stimulus-speed of 40 deg/s was combined with a moderate acceleration of 0.3 G (3.0 m/s2) as in Experiment 1 (N=10), the combined-OKR velocity always increased during the agonistic condition, and the motion of the visual pattern was perceived as slow and clear in this segment. On the other hand, during the antagonistic condition, the combined-OKR velocity either remained unchanged or increased moderately, and the motion of the visual pattern was sensed as fast and unclear. Notably, in most subjects, the velocity difference in combined-OKR between the agonistic and antagonistic conditions was around the value of the tVOR velocity. In five of the ten subjects who completed an additional test session with the acceleration level increased from 0.3 to 0.5 G (4.9 m/s2), similar findings were maintained individually, suggesting independent behavior of tVOR. Therefore, we hypothesized that the interaction could be direction-selective; in other words, both tVOR and OKR are additive during the agonistic condition, but tVOR is suppressed during the antagonistic condition. To extend this hypothesis further, another group of subjects was exposed to three different optokinetic-stimulus speeds of 20, 40 and 60 deg/s combined with an acceleration of 0.3 G (Experiment 2, N=15). Combined stimulation tended to optimize the combined-OKR velocity around the given optokinetic stimulus-speed, especially in those cases where the reference-OKR velocity deviated significantly from the stimulus speed. Changes in combined-OKR velocity were small at 20 deg/s, and were likely to be linear across the agonistic and antagonistic conditions. With increasing optokinetic stimulus-speeds, the direction-selective asymmetry hypothesized above was maintained in more than half of the subjects, while in the other subjects the combined-OKR velocity difference increased remarkably, probably due to an enhancement of the OKR gain. We conclude that tVOR suppression during the antagonistic stimulus-condition and non-linearity in the tVOR-OKR interaction are characteristic of the otolith system, even under moderate-stimulus environments, in contrast to the linear eye-movement interaction in the semicircular canal system.