The proposal that the understanding and imitation of observed actions is made possible through the ‘mirror neuron system’ (Rizzolatti, Fogassi & Gallese, 2001) has led to much speculation that a dysfunctional mirror system may be at the root of the social deficits characteristic of autism (e.g. Ramachandran & Oberman, 2006). This chapter will critically examine the hypothesis that those with ASD may be in possession of a 'broken' mirror neuron system (MNS) and propose that the deficits seen in imitation in individuals with ASD reflect not a dysfunctional MNS, but a lack of sensitivity to those cues that would help them identify what to imitate. In doing this, we will also argue that imitation in typically developing children cannot be explained by appealing to a direct-matching mechanism, and that the process by which young children imitate involves a far more complex but effortless analysis of the communication of those who they learn from.