The cerebellum contains over three quarters of the neurons in the human brain (Herculano-Houzel, 2010), but it has traditionally been studied mainly in sensorimotor contexts. Yet in humans, cerebellar circuits activate during and are required for verbal and spatial processing tasks (eg, Stoodley et al (2012)), and multiple lines of evidence point to cerebellar links to psychoses such as schizophrenia as well as to autism (reviewed in Sokolov et al (2017)). Moreover, the cerebellum is widely connected with multiple cognitive regions of the neocortex (eg, Strick et al (2009)), with lateral cerebellar regions preferentially connected to frontal neocortical areas, both of which are expanded in the primate lineage. Despite this evidence, little is known about the putative cognitive signals that reach the cerebellum at the cellular physiological level.

In particular, the cerebellar input layer that receives external information consists of granule cells. Cerebellar granule cells receive only a few inputs that can arise from a multitude of places throughout the brain and sensory periphery, indicating they may be well positioned to receive and transmit an array of non-sensorimotor signals to the cerebellum. Because of the longstanding technical difficulties arising from their small size and dense packing, however, granule cell responses have not been recorded during cognitive tasks.

We recently performed two-photon calcium imaging in ensembles of individual cerebellar granule cells during conditioning tasks where mice learned to expect upcoming rewards (Wagner et al, 2017). In one such task, thirsty mice used their forepaw to move a small handle forward for delayed receipt of a water reward. Surprisingly, some granule cells preferentially responded to expected or unexpected rewards, or to the omission of expected rewards. Other cells selectively encoded the cognitive state of expectant waiting for an upcoming reward. Multiple experiments and analyses indicated that these reward signals were unexplained by motor signals such as licking or body movement. Granule cell reward signals were present in multiple different reward-related behavioral tasks, and emerged over the course of task learning.

These results raise a host of new questions. Foremost, it will be critical to understand the origin of reward-related signaling in the cerebellum. Anatomical tracing did not reveal direct input to the cerebellar cortex from midbrain dopamine cells like those of the ventral tegmental area that are known to convey reward signals (Cohen et al, 2012). On the other hand, the cerebellar regions we studied received robust input from the pontine nuclei, which relay projections to cerebellum from many regions of the neocortex (Strick et al, 2009), which is a prime candidate for further study. As important for future work will be determining how the cerebellum uses cognitive signals in its own computations, and what role any cerebellar cognitive output has in downstream circuits like those of the cortex. This would help determine whether cerebellar output could be a translational target for schizophrenia or psychosis. Integrating cerebellar circuits into the brain’s broader reward-related computations may help to shed light on the mechanisms of cerebellar involvement in cognitive processes.

Funding and disclosure

This work was supported by Epilepsy Training Grant and Hughes Collaborative Innovation Award. The author declares no conflict of interest.