The 'broken mirror' theory of autism has received considerable attention far beyond the scientific community. This theory proposes that the varied social-cognitive difficulties characteristic of autism could be explained by dysfunction of the mirror neuron system, thought to play a role in imitation. We examine this theory and argue that explaining typical imitation behavior, and the failure to imitate in autism, requires much more than the mirror neuron system. Furthermore, evidence for the role of the mirror neuron system in autism is weak. We suggest the broken mirror theory of autism is premature and that better cognitive models of social behavior within and beyond the mirror neuron system are required to understand the causes of poor social interaction in autism.