The motion of objects and ourselves along the vertical is affected by gravitational acceleration. However, the visual system is poorly sensitive to accelerations, and the vestibular otoliths do not disassociate gravitational and inertial accelerations of ego-motion. Here, we tested the hypothesis that the brain resolves visual and vestibular ambiguities about vertical motion with internal models of gravity, which predict that downward motions are accelerated and upward motions are decelerated by gravity. In visual sessions, a target moved up or down while participants remained stationary. In vestibular sessions, participants were moved up or down, while they fixated an imaginary target moving along. In visual-vestibular sessions, participants were moved up or down while the visual target remained fixed. We found that downward motions of either the visual target or the participant were systematically perceived as lasting less than upward motions of the same duration, and vice-versa for the opposite direction of motion, consistent with the prior assumption that downward motion is accelerated and upward motion is decelerated by gravity. In visual-vestibular sessions, there was no significant difference in the average estimates of duration of downward and upward motion of the participant. However, there was large inter-subject variability of these estimates.