Fig. 1

Outcome, choice, and learning rate. When outcomes of decisions are witnessed, the prediction for the next choice is updated based on a learning rule where the prediction error (PE) is weighted by the learning rate α. Behrens et al. have shown that average activity in dACC reflects the environment’s volatility and that under high volatility, the options’ values are updated with a high learning rate α. However, at the time of the actual decision on the next trial, volatility no longer exerts a significant effect on average dACC activity. However, the representation of choice value estimates necessary for decision making (the value estimate for option A relative to that of option B) might be represented in some other way such as an anatomically distributed pattern of activity where different value estimates might be calculated with different parameterizations of α, depending on the volatility. Copyright for brain image: Behrens et al., Learning the value of information in an uncertain world, 2007, Nature Neuroscience, all rights reserved