Vestibuloocular reflex performance and adaptation were examined during vestibulocerebellar inactivation by localized lidocaine microdialysis or injection in goldfish. In the light, eye velocity perfectly compensated for head velocity (Vis-VOR) during sinusoidal yaw rotation (1/8 Hz +/- 20 degrees). In the dark, the reflex (VOR) gain was slightly reduced (gain approximately 0.8-0.9). In neither Vis-VOR nor VOR, was gain altered after 1 h of lidocaine microdialysis in the vestibulocerebellum. Before adaptation of reflex gain, the initial suppression or augmentation of Vis-VOR reflex gain produced by in-phase or out-of-phase visual-vestibular stimulation was also unaffected by cerebellar inactivation. Subsequently, 3 h of adaptive reflex training in either the in-phase or out-of-phase paradigm (acquisition phase) respectively decreased (0.30 +/- 0.09) or increased (1.60 +/- 0.08) VOR gain during artificial cerebral spinal fluid (CSF) microdialysis. However, microdialysis of lidocaine completely blocked adaptive gain changes during a 3-4 h period of continuous application. This effect was reversible because VOR gain changes were produced 1 h after lidocaine was replaced with CSF as the dialysate. After adaptive training, bilateral CSF injections (0.25 microl/side) into the vestibulocerebellum did not alter the normal retention or decay of adapted gain changes during a 3 h period in the dark (retention phase). However, injection of lidocaine into the vestibulocerebellum completely blocked retention of the adapted VOR gain returning the gain to values recorded before adaptation. In contrast to either acute or chronic surgical removal, lidocaine inactivation of the cerebellum by microdialysis did not alter either Vis-VOR and VOR behavior or interactive Vis-VOR performance over a wide range of gain extending from 0.3 to 1.4. Thus short-term VOR motor learning is a dynamic process requiring either continuous operation of brain stem cerebellar loops or, alternatively, modifiable sites within or directly influenced by the cerebellum. Our data supports the latter hypothesis, because the direct brain stem VOR pathways appear to be unaltered after cerebellar inactivation, and, hence, independent of the VOR-adapted state.