What computation does the Cerebellum perform? The Universal Cerebellar Transform

The cerebellum has a very stereotypical neural circuit that is repeated throughout the entire brain region. Moreover, despite the diverse type of sensory and motor input into the cerebellum, only the Purkinje cell acts as the single output pathway from the cerebellar cortex. Moreover, over half of the neurons in the human brain are granule cells whose axons innervate the Purkinje cells. Thus, with so many neurons and such an impressive bandwidth of processing, it is quite likely that the cerebellum is performing a critical computation. Furthermore, given its highly conserved and modularity, some have hypothesized it is doing one unique computation, called the Universal Cerebellar Transform. 

The Universal Cerebellar Transform is not an easy computation to even define. Classically the cerebellum has been thought to be responsible for fine motor control, but in reality the cerebellum has been involved in social cognition and learning. It may even have a role in autism. Thus, whatever the Universal Cerebellar Transform is, it highly abstract and very versatile. That being said, one of the best ways to get a handle on what the Universal Cerebellar Transform could be doing is to examine movement, and movement like computation. 

Specifically some new research out of Dr. Samuel McDougle's lab shows that the cerebellum is critical in performing mental rotations, and manipulations of concepts. This idea is born directly out of the cerebellum's role in movement. Our classical understanding of cerebellum is that it creates a mental representation of our bodies location is space, and uses this model to predict (and correct) as we move our bodies. Thus if the Universal Cerebellar Transformation hypothesis was true, we would expect mental models to also use this sort of internal manipulation of a model.  Take for example, mental rotation. The task is to determine if two images of an object are the same object, only rotated in space. 

The larger the rotational angle between the two objects, the longer it takes for you to "rotate" it in your minds eye. This is exactly analogous to physically rotating the object in your hand. Thus if the cerebellum is necessary for rotating in your hands perhaps it required for rotating in your minds eye. This is exactly what Dr. McDougal tested in his paper. They show that in patients with cerebellar damage (and only cerebellar damage), patients take significantly longer to rotate the object in their minds eye than healthy controls. This is highly suggestive that  the cerebellum is indeed relevant in manipulating objects in the minds eye.

Perhaps more convincingly is that the cerebellum is also involved in performing mental addition. There is copious amount of evidence that addition is performed on a mental number line (even if you are not actively visualizing said number line). In fact the larger the distance between the two numbers you are adding, the longer it takes you to perform the mental addition. This is indicative of you requiring more time to slide the number along the number line. What is mind blowing is patients with cerebellar damage require more time perform this mental manipulation as well. Again showing evidence that cerebellum is involved whenever there is a mental representation being manipulated. (A fun aside, multiplication is considered a pure memory task and is unaffected by cerebellum damage, you can thank your elementary school teacher for the years of rout memorization for this.)

It is important to note, while this work does show excellent evidence for the Universal Cerebellar Transform, it does not necessitate that the hypothesis is true. It is very hard for social cognition and reward processing to be mapped onto a Universal Cerebellar Transform. The leading hypothesis is that the cerebellum is involved in predicting the outcome of social interaction much like its involved in predicting motor outcomes, but there is a lot more work required to show this is indeed the case. Of course the cerebellum may not be performing one single operation, but a large swath of different computations. A lot more work is required to suss out all of these different hypotheses, and is an active and exciting frontier of neuroscience.

Author: Alexander White