Human social interaction depends on non-verbal unconscious behaviour as much as on verbal signals. Mimicry (unconscious copying of actions) is a good example of a social behaviour which is caused by and has consequences for our evaluation of others. However, studying mimicry with traditional methods is hard because of the trade-off between good experimental control and realistic social interaction. INTERACT will (1) establish a new approach to the science of mimicry, bringing together methods from social psychology, cognitive neuroscience and computer science, and (2) use this approach to understand the information processing mechanisms underlying mimicry of hand actions. First, we will develop interactive avatars which can mimic a participant’s hand actions or be mimicked by the participant in the context of a simple drum rhythm task. Using computer-generated avatars allows us to precisely control and measure movement timing and structure during mimicry, and to record how participants interact with avatars with different socially-relevant features (age / attractiveness or even aliens). Thus, the INTERACT system will enable high-resolution, well-controlled studies of how people detect and control mimicry. Second, we will use the interactive avatars to examine mimicry in unprecedented detail, studying how the timing and structure of an action and form of the avatar impact on the control and detection of mimicry in typical adults. Building on this, we will define the brain mechanisms of mimicry and why mimicry might go wrong in adults with autism spectrum condition. The results will test current hypotheses of mimicry and will reveal the information processing mechanisms underlying human mimicry and its relationship to other social processes. Completion of the project will benefit research and practice in social neuroscience, developmental and educational psychology, computer science and robotics, and all researchers interested in human social behaviour.