In order to detect whether micro-organisms could initiate the autoimmune process in Graves' disease we have studied the temporal and spatial distribution of 857 cases of hyperthyroidism occurring in a community over ten years. Cases were identified through biochemistry laboratory records and following the exclusion of patients with toxic nodular goitre or with insufficient clinical data there were 599 with Graves' disease--an average annual incidence of 15.9 per 100,000. There was a tendency for cases to present in the summer months. The reported onset of symptoms, however, peaked in December and June. There was no evidence of clustering of cases in space and time using two different statistical methods. Incidence rates doubled between 1976 and 1980 and then declined--a trend that could neither be explained by changes in laboratory or clinical diagnosis nor did it correlate with any pattern of microbial disease in the area. We conclude that it is unlikely that infections that behave in an epidemic manner have a causative role in triggering Graves' disease.