Subjects with diabetes who attended rural surveys in Western Australia from 1978 to 1982 were followed up to ascertain death rates and the causes of death recorded on death certificates. Cardiovascular disease was assigned as the direct cause of death in 63% of deaths, with equal rates in male and female subjects, and renal disease in 8% of deaths with the proportion in women (12%) being greater than that in men (4%). The diagnosis of diabetes was stated on only 65% of the death certificates, and in only 24% was diabetes recorded as a direct or antecedent cause. In the same cohort the Australian Bureau of Statistics coded diabetes as the underlying cause of death in 24%, while attributing deaths to cardiovascular disease in 50% of the cases. This study suggests that diabetes is considerably underreported both on doctors' death certificates and in the mortality figures of the Australian Bureau of Statistics.