1. To investigate the type of vestibular signals that neurons in the caudal parts of the vestibular nuclei transmit to the cerebellum and spinal cord, we studied their responses to natural vestibular stimulation in vertical planes in decerebrate cats with the caudal cerebellum removed. Most neurons were in the caudal half of the descending vestibular nucleus, the remainder at corresponding levels of the medial nucleus or the medial-descending border. 2. Dynamics of the responses of spontaneously firing neurons were studied with sinusoidal tilts delivered at 0.05-1 Hz near the plane of body rotation that produced maximal modulation of the neuron's activity (response vector orientation). For most neurons the predominant vestibular input could be identified as coming from otolith organs (46%) or vertical semicircular canals (37%). Some neurons had otolith+canal convergence (9%) and others either had such converging input or some other form of central processing (8%). 3. Gain and phase of the responses of otolith neurons were comparable with values obtained in earlier studies on Deiters' nucleus and the rostral descending nucleus. Many canal neurons had a steeper gain slope and more advanced phase than observed previously for more rostral neurons. This may be due to more irregular afferent input to many neurons or to the absence of the vestibulocerebellum. 4. Response vector orientations of canal neurons were closely bunched near the planes of the ipsilateral vertical canals. The small number of contralaterally projecting vectors showed evidence of convergence between the two contralateral vertical canals. As is the case elsewhere in the vestibular nuclei, there was no evidence of convergence from bilateral vertical canals. Response vector orientations of otolith neurons were restricted to the roll quadrants; the majority pointed ipsilaterally. 5. Antidromic stimulation with an electrode in the restiform body or with several electrodes in the dorsal half of the white matter of the upper cervical cord was used to identify neurons projecting to the cerebellum and spinal cord, respectively. A substantial number of spontaneously firing neurons projected to the cerebellum, but there were few spontaneously active vestibulospinal neurons. The properties of the vestibular input to cerebellar-projecting neurons were the same as those of the population as a whole, but the effect of tilt on vestibulospinal neurons appeared weak or absent. 6. Many neurons were inhibited by stimulation of the restiform body. We suggest that this is mainly due to stimulation of the axons of vestibulocerebellar Purkinje cells. 7. Our results demonstrate a robust vertical vestibular input to the caudal parts of the vestibular nuclei.(ABSTRACT TRUNCATED AT 400 WORDS)