Horizontal and vertical smooth pursuit was compared with otolith-ocular responses in 11 patients with cerebellar ataxia and 21 normal subjects using three-dimensional magnetic search coil eye movement recordings. Otolith-ocular responses were investigated during off-vertical axis rotation. This stimulus induces nystagmus consisting of the exponentially decaying canalicular response, and an eye-velocity modulation and offset which arise from the excitation of the otoliths by the gravity vector, which lasts as long as the rotation continues. Otolith-ocular reflexes are intimately interrelated with visual tracking when real targets are viewed during linear motion. The responses of both the translational vestibule-ocular reflex and the pursuit system have been shown to be linearly dependent on the inverse of the viewing distance, so that a common central pathway for the two systems has been suggested, probably travelling through the cerebellum. Thus, the aim of the study was to evaluate to what extent these reflexes are disturbed in cerebellar disease. The results confirm the earlier notion that in normal subjects pursuit performance is better for horizontal than for vertical tracking, and that it is better for upward than for downward tracking. This pattern is also found in patients. In addition, smooth pursuit performance is clearly degraded in patients, but the modulation of eye-velocity during off-vertical axis rotation is enhanced. Since the amount of this enhancement does not correlate with the amount of pursuit impairment, degradation of smooth pursuit and pathological enhancement of otolith-ocular responses seem to be independent effects of cerebellar degeneration. Thus, the increase in the otolith-ocular response in patients cannot be attributed to adaptational mechanisms trying to overcome the smooth pursuit deficiency; it is more likely to represent pathological disinhibition of otolith derived responses. The absence of compensatory eye-velocity offset during off-vertical axis rotation may reflect the fact that in patients the otolith signals are not utilized in computations thought to be important for spatial orientation mechanisms arising from the interaction of vestibular, visual and somatosensory signals.