A recent model of cerebellar learning in eyeblink conditioning predicts two sites of plasticity, the cerebellar cortex and cerebellar nuclei, which store information relating to timing and driving the movement, respectively. Consistent with this idea, lesions of the cortex or reversible "disconnections" of Purkinje cell output to the nuclei have been shown to disrupt response timing to produce short-latency conditioned eyeblinks. To better characterize potential cortical and nuclear plasticities, we analyzed the effects upon nictitating membrane (NM) and eyeblink conditioned responses (CRs) of different drugs administered to the cortex and to the nuclei. When either excitatory or inhibitory inputs to the cerebellar cortical lobule HVI were blocked by infusions of the AMPA receptor antagonist CNQX or the GABA-A receptor antagonists picrotoxin or SR95531, CRs were abolished. Similarly GABA-A receptor antagonists in the cerebellar nuclei abolished CRs. CR latencies were never shortened. However, blockade of AMPA/kainate receptor-mediated excitatory transmission to the nuclei had no effect upon CR frequencies or latencies. These results suggest that normal cortical and nuclear function is required for performance of NM and eyeblink CRs. We saw no evidence that CRs can be driven by AMPA/kainate receptor-mediated transmission from mossy fiber afferents to the cerebellar nuclei. So, although plasticity in the cerebellar nuclei is not ruled out, it is unlikely that a long-term change in AMPA receptor-mediated transmission from mossy fiber inputs to the nuclei is an essential mechanism in eyeblink conditioning. Our findings indicate that a fully functional olivo-cortico-nuclear loop is required to express all characteristics of associatively conditioned responses.