Humans unintentionally copy other people’s gestures, postures and speech patterns. This behaviour has been termed ‘automatic imitation’, and is thought to play a crucial role in the development of social relationships by building affiliation and rapport between interaction partners. The wide-ranging impact imitation may play in social life has stimulated considerable interest among psychologists and neuroscientists, but measuring imitative behaviour with high fidelity poses a problem due to the inherent difficulty of measuring live social interactions. To study imitative processes in a more controlled manner, a reaction time measure has been developed that is argued to index processes associated with imitation. The current paper critically evaluates reaction time measures of automatic imitation and demonstrates that it is currently unclear what reaction time indices measure, due to a lack of research assessing task validity. A consequence of low validity is that a sufficiently solid foundation has not been formed for future research to build upon in a progressive manner. Based on this assessment of the literature, three suggestions are made for future research. First, studies that assess the validity of the paradigm as a measure of automatic imitation are required. Second, theories of imitation need to be developed that move beyond neat divisions between domain-specific and domain-general systems. Third, a greater focus needs to be placed on methodological rigour to enable a cumulative science to develop. In sum, by providing a critical review of the current literature, as well as suggesting new directions for the development of research questions, theories and methods, this paper aims to stimulate discussion regarding optimal approaches to studying automatic imitation. Moreover, the analysis and recommendations presented here generalise to other domains of social cognition, such as gaze perception and perspective taking, which also rely on similar reaction time paradigms.