How can a Bayesian approach inform neuroscience?

In this review we consider how Bayesian logic can help neuroscientists to understand behaviour and brain function. Firstly, we review some key characteristics of Bayesian systems - they integrate information making rational use of uncertainty, they apply prior knowledge in the interpretation of new observations, and (for several reasons) they are very effective learners. Secondly, we illustrate how some well-known psychological phenomena including visual illusions, categorical perception and attention can be understood in terms of Bayesian inference. We also consider how formal models can clarify our understanding of psychological constructs, by giving a truly computational definition of psychological processes. Finally, we consider how probabilistic representations and hence Bayesian algorithms could be implemented by neural populations. In particular, we explore how different types of population coding may lead to different predictions about activity in both single-unit and imaging studies, and draw a distinction in this context between the representation of parameters and implementation of computations.