One of the most widely used experimental models for the study of learning processes in mammals has been the classical conditioning of nictitating membrane/eyelid responses, using both trace and delay paradigms. Mainly on the basis of permanent or transitory lesions of putatively-involved structures, and using other stimulation and recording techniques, it has been proposed that cerebellar cortex and/or nuclei could be the place/s where this elemental form of associative learning is acquired and stored. We have used here an output-to-input approach to review recent evidence regarding the involvement of the cerebellar interpositus nucleus in the acquisition of these conditioned responses (CRs). Eyelid CRs appear to be different in profile, duration, and peak velocity from reflexively-evoked blinks. In addition, CRs are generated in a quantum manner across conditioning sessions, suggesting a gradual neural process for their proper acquisition. Accessory abducens and orbicularis oculi motoneurons have different membrane properties and contribute differently to the generation of CRs, with significant species differences. In particular, facial motoneurons seem to encode eyelid velocity during reflexively-evoked blinks and eyelid position during CRs, two facts suggestive of a differential somatic versus dendritic arrival of specific motor commands for each type of movement. Identified interpositus neurons recorded in alert cats during classical conditioning of eyelid responses show firing properties suggestive of an enhancing role for CR performance. However, as their firing started after CR onset, and because they do not seem to encode eyelid position during the CR, the interpositus nucleus cannot be conclusively considered as the place where this acquired motor response is generated. More information is needed regarding neural signal transformations taking place in each involved neural center, and it its proposed that more attention should be paid to functional states (as opposed to neural sites) able to generate motor learning in mammals. The contribution of feedforward mechanisms normally involved in the processing activities of related centers and circuits, and the possible functional interactions within neural systems subserving the associative strength between the conditioned and unconditioned stimuli, are also considered.