Duke Study Identifies Different Model of Cerebellar Learning

Humans often learn motor skills through a process of trial and error, with mistakes leading to negative consequences that teach us to avoid future errors. But what if NOT making a mistake in the first place also helps us learn those skills?

A new Duke study presents alternate theories on how motor learning takes place in the brain’s cerebellum. The study, “Coordinated Cerebellar Climbing Fiber Activity Signals Learned Sensorimotor Predictions,” was published recently in Nature Neuroscience. The lead author, Jake Heffley, and Court Hull, the corresponding author, are, respectively, a graduate student and assistant professor, Department of Neurobiology, in the Duke School of Medicine. Hull is also a member of the Duke Institute for Brain Sciences (DIBS) Faculty Research Network.

When you take a walk, your cerebellum, a small region at the back of the brain, takes in information from the spinal cord and other parts of the brain, processes it, and helps your body refine what actions to take, Hull explained. For example, if you’re walking a familiar path, you have learned to sidestep a big puddle in the middle of the path, thanks to information processing in the cerebellum.

Neuroscientists have long known that the cerebellum can make “predictive associations” between sensory inputs and motor, or action, outputs. This is an important part of what allows humans to generate coordinated responses to environmental stimuli, e.g., sidestepping a predictable puddle. But how does the cerebellum learn these predictive associations?    

It was previously thought the cerebellum could learn only from erroneous outcomes, or making mistakes. “The primary model of cerebellar learning suggests that neural inputs to this area called climbing fibers are triggered by movement errors, and act to correct erroneous movements,” Hull said. That is, you learn to avoid the puddle because in the past, you have walked through it and gotten your shoes wet. Cerebellar theory would suggest that this mistake is corrected because climbing fibers in the cerebellum signaled there was an error, and instructed new learning to help avoid the puddle in the future. However, what if the cerebellum could learn from a positive outcome that didn’t require stepping in the puddle in the first place?

Hull and his co-authors developed a sensorimotor task where mice learned to associate a reward with a correctly timed action (pressing a lever). Results suggest that climbing fibers within the cerebellum can carry instructional signals that are not driven simply by movement errors. Instead, these signals can both predict and evaluate the expected outcome of movements, even if that outcome is a positive one.

The research suggests that motor learning is not linked directly to a stimulus and response, but can instead use behavioral context and abstract associations. “In other words, the cerebellum can learn not just by making mistakes, but by associating new cues in the environment with expected outcomes, regardless of whether those outcomes are positive or negative,” Hull added.   

This is important because, in many cases, learning from a correct outcome can have great advantages. For example, you can learn to avoid a puddle by associating cues in the environment, such as a particular bend in the path that predicts the upcoming puddle, but you don’t need to make the mistake of stepping in it in the first place to learn such associations.

“This extension of cerebellar learning models will help us understand how learning occurs in different situations and what experiences can allow learning to happen,” Hull said. “This has the potential to inform how motor skills are taught throughout a lifespan.”

Citation:

Heffley W., Song, E.Y., Xu Z., Taylor, B.N., Hughes, M.A., McKinney, A., Joshua M, Hull, C. Coordinated cerebellar climbing fiber activity signals learned sensorimotor predictions, Nat Neurosci. 2018 Oct;21(10):1431-1441. doi: 10.1038/s41593-018-0228-8. Epub 2018 Sep 17.