Perceived postural orientation during whole-body roll-tilt is commonly inferred from settings of a visual line to the perceived gravitational horizontal or vertical. This inference assumes that the change in ocular torsional position (ocular counterrolling) which occurs during roll-tilt has no effect on the perceived orientation of the visual stimulus. We investigated this assumption by measuring, during whole body roll-tilt stimulation, settings of a visual line and a somatosensory bar to the perceived gravitational horizontal and comparing the difference in these settings to the objectively measured ocular torsional position for each subject. Two stimulus paradigms were used: one where the subject was given a roll-tilt stimulus and the eye torted, the other where there was eye torsion without a roll-tilt stimulus. In both paradigms there was a very close relationship in magnitude and direction between the difference in the settings of the two perceptual indicators to gravitational horizontal and the objectively measured ocular torsion. We conclude that change in ocular torsional position alone changes the perceived orientation of a visual line. The corollary is that settings of a visual line cannot be used to infer perceived postural orientation directly.