Perception can be considered a process of inference where bottom-up sensory information is integrated with previously learnt information about the world. Imbalances in this process can contribute to the development of hallucinations. My talk will cover work in Parkinson’s disease visual hallucinations, where we have used a variety of methodologies (i.e., cognitive tasks, MRI, and computational modelling) to explore sensory evidence accumulation processes and top-down influences over vision related to the default network. These studies will be discussed with reference to the evolution of visual hallucinations in Parkinson’s disease and the potential role for a predictive processing account.