A principle of the cognitive neuropsychiatric approach is to go beyond surface phenomenology and to comprehend the symptoms of psychiatric illness in terms of altered functioning within the normal system. This requires us to “take the best cognitive models for a range of normal psychological functions and treat them as if psychiatric phenomena fall within their ambit” (David and Halligan, 1996) bridging the gap between high level experiences or behaviours and the neural systems that underpin them (Halligan and David, 2001). A recent study by Powers and colleagues (Powers et al, 2017) applies computational modelling in the context of an elegant experimental design to achieve this. In doing so, they show how this principled approach may provide mechanistic understandings of how symptoms of psychosis may emerge, evolve and interact.