This study investigated whether the tilt of the subjective vertical induced by galvanic vestibular stimulation, demonstrated by asking subjects to set a rod to the vertical, was specific to the visual modality or could be found in two tasks relying on proprioceptive and somatosensory cues. In all cases, settings were significantly deviated in the direction of the anode, but errors were smaller in the somatosensory tasks than in the visual task. We propose that the effects observed in the somatosensory modality reflects only a modification of the central representation of gravity, whereas visual effects are also in part the consequence of unregistered ocular torsion.