In order to test whether the control of eye movement in response to head movement requires an internal model of head orientation or instead can rely on directly sensing information about head orientation and movement, perceived gravity was separated from physical gravity to see which dominated the eye-movement response. Internal model theory suggests that the oculomotor response should be driven by perceived, internalized gravity, whereas the direct sensing theory predicts it should always be driven by vestibularly sensed gravity. Subjects lay on an airbed either supine or on their side and were sinusoidally translated along their dorsoventral body axis. The direction of perceived gravity was separated from physical gravity by performing the experiments in a room built on its side with the direction of its "floor" orthogonal to both physical gravity and the subject's translation. The swinging sum of the imposed sinusoidal acceleration with physical gravity was thus in a plane orthogonal to its sum with perceived gravity. Oculomotor responses to these swinging vectors were looked for and responses were found only to the sum of the acceleration with physical gravity, not perceived gravity. It was concluded that an internal model is not used to drive these compensatory eye movements.