News

Identification of brain regions important in behavioural adaption to reward

Image: Diederen et al., 2016

Every day we make many decisions: where to go for lunch or which road to take home to avoid traffic. In order to pick the best, we have to learn what to expect from each option. Learning what to expect is difficult since the outcomes of our choices tend to vary from time to time. The quality of lunch served in a restaurant may for example vary slightly from day to day depending on the freshness of the ingredients. However, a big change in food quality could result from the restaurant having hired a new chef. Therefore, we need to learn what kind of variability is small enough to ignore in order that we may notice when the food in a café suddenly becomes unacceptably bad or exceptionally good. In other words, we need to be able to update our predictions when the outcome changes much more than expected.

Neuroscience has long recognised that we learn from our mistakes. The brain is sensitive to surprising outcomes - so-called prediction errors - and uses these to drive the learning process. But the world is an unreliable place and, if we changed our beliefs in response to every unpredicted outcome we would probably function very poorly. One of the real challenges facing us is knowing when we should take notice of surprises ("prediction errors") and when we should just ignore them and put them down to chance. Indeed, it has been suggested that certain symptoms of mental illness may arise because of a difficulty in successfully achieving this balance - either failing to adjust our learning when we really should or else adjusting it too readily and so developing an unrealistic view of the world.

Recently, work done by Cambridge Neuroscientists has shown that we are very sensitive to just how variable the outcomes in the world may be and that we can adjust our learning to achieve this balance. When there is a great deal of variability, we seem to dampen down our learning and become relatively unmoved by prediction errors.

In this current study, published in the journal Neuron and carried out in the Departments of Physiology, Development and Neuroscience and Psychiatry, the investigators led by Dr Kelly Diederen took this insight further and identified brain regions that are associated with this subtle adjustment in our tendency to learn under different environmental conditions. By carefully changing just how variable a financial reward was, they were able to show that healthy people can put the brakes on learning as the variability becomes greater. They found that two regions in the brain (the midbrain and ventral striatum) were sensitive to how variable the rewards were. Specifically, these regions kept track of predictions errors relative to the exact variability. They also observed that people who were more sensitive to reward variability showed better learning on our task, suggesting that individual differences in such sensitivity might distinguish ‘good’ from ‘poorer’ learners.

Reference: Adaptive Prediction Error Coding in the Human Midbrain and Striatum Facilitates Behavioral Adaptation and Learning Efficiency Diederen, Kelly M.J. et al. 2016 Neuron

Posted on 13/05/2016