This talk offers an overview of active inference – a set of methods in computational neuroscience – with a focus upon how it might be used to understand pathological neuronal computation in neurological disorders. The key themes of the talk include (1) the relationship between the brain’s internal model of its environment and the (synaptic) message passing architectures that support perceptual inference; (2) planning as inference, and changes in behaviour with changes in prior beliefs; and (3) the role of neuromodulatory transmitters in setting the precision (or confidence) of our internal models.