We know that both the cerebellum and sleep are implicated in learning new procedural skills, yet the sleeping cerebellum has been largely overlooked by neuroscience. Using wireless recording in unrestrained non-human primates, we are investigating cerebellar activity and cerebro-cerebellar communication during natural sleep. We observe many of the neocortical hallmarks of sleep also evident in the cerebellum and, surprisingly, communication at spindle frequencies directed from the cerebellum to the neocortex. I will interpret these data in the context of a hypothesis of how internal models in the cerebellum might play a role in sleep-dependent motor learning.