Brothers (Brothers L. Concepts in Neuroscience 1990;1:27-51) proposed a network of neural regions that comprise the "social brain", which includes the amygdala. Since the childhood psychiatric condition of autism involves deficits in "social intelligence", it is plausible that autism may be caused by an amygdala abnormality. In this paper we review the evidence for a social function of the amygdala. This includes reference to the Kluver-Bucy syndrome (which Hetzler and Griffin suggested may serve as an animal model of autism). We then review evidence for an amygdala deficit in people with autism, who are well known to have deficits in social behaviour. This includes a detailed summary of our recent functional magnetic resonance imaging (fMRI) study involving judging from the expressions of another person's eyes what that other person might be thinking or feeling. In this study, patients with autism or AS did not activate the amygdala when making mentalistic inferences from the eyes, whilst people without autism did show amygdala activity. The amygdala is therefore proposed to be one of several neural regions that are abnormal in autism. We conclude that the amygdala theory of autism contains promise and suggest some new lines of research.