But how does the brain do this? This is the topic of a recent publication in Nature Neuroscience by Julius Koppen, who performed this work as part of his PhD research, and Devika Narain, associate professor at DCN and also affiliated with Erasmus MC.
Relying on past experiences in an uncertain world
“The world around us is full of uncertainties,” Narain explains. In uncertain situations, we rely on our past experiences. For example, on a foggy day, you usually have an idea of speed limits on local roads.
Many researchers, including those from DCC, such as Prof. Floris De Lange and Prof. Pieter Medendorp, study how past knowledge shapes behaviour. They often use mathematical models, like Bayesian theories, to describe this. Until now, there was little direct evidence that brain cells can actually store such prior knowledge. This study shows that they do, even for simple movement behaviours, like predictive eyeblinks.
Blink and predict
In their research, they studied a well-known behaviour in mice, called Pavlovian learning. A mouse predictively blinks at a set time after a light signal because it learns to anticipate an airpuff. In the experiment, the researchers changed the timing of the airpuff as a probability distribution, creating uncertainty.
The result? The mouse adapted its blinking to the changing probabilities. “The eye doesn’t close at one exact moment,” Narain says. “It follows a complex pattern showing how often a blink is expected at different times.”
Narain uses a familiar example: “If you often play tennis with the same person, you learn where their serve usually lands.” It’s not always exactly the same, but you learn the probabilities involved and adjust your game based on this prior experience. In the same way, the eyeblink system learns the patterns of the airpuff and adjusts the eyelid movement.
Prior knowledge in the brain
The researchers went a step further and recorded where this knowledge is stored in the brain. They found that Purkinje cells in the cerebellum hold these patterns. The cerebellum helps with movement and coordination, and Purkinje cells make movements smooth and well-timed, including blinking.
“We show that these cells store memories of probability patterns,” Narain explains. “The brain learns from past experiences and adjusts movements to reduce uncertainty.” This is remarkable because it shows that individual brain cells can actually store patterns of probability.
A brain that predicts
They also discovered something unexpected. When uncertainty was high, the Purkinje cells generated an extra signal. “It was as if the brain found a way to reduce the uncertainty,” Koppen explains. “It was making its own prediction.” This predictive signal has never been observed before and shows how the brain can make reliable guesses in a world full of surprises.
Reigniting an old view on learning
This work may renew interest in brain theories that use probabilities, like Bayesian theories. “Even simple systems are sensitive to probabilities in the world, and that brain circuits store them directly,” says Narain. “Hopefully, this could help us understand complex behaviors and treat disorders affecting movement or mental health”.
The next step is to see if this mechanism also works in more complex decisions. Narain concludes: “The brain is always finding ways to make uncertainty predictable, and we are just beginning to understand how it works.”