My research focuses on human social interaction, specifically upon our ability to infer beliefs and intentions in other minds. Little is known about the neural basis of our ability to read the intentions of others, but a very likely candidate is the mirror neuron system. Mirror neurons discharge not only during action execution but also during action observation. It is this property that makes these neurons a possible neural substrate for action understanding. The MNS has been the focus of much interest since its discovery because it has the properties that may enable us to infer the goal of an observed action. However, how the visual information from an observed action maps onto the observer's motor system and how the goal of that action is inferred remain a matter of debate. My research uses non-invasive measures of human brain activity, fMRI, EEG and MEG, as well as behavioural studies to test the hypothesis that the MNS is best explained within a predictive coding framework that appeals to hierarchical generative models and empirical Bayes. This approach provides a mechanistic account of how evoked responses in the visual and motor systems are organised and explains how the causes of an action can be inferred from its observation.