OBJECTIVE - To analyse the prevalence of Down's syndrome in a specific, geographical area and seek to explain variations with particular reference to ionising radiation. DESIGN--Cases were ascertained by one paediatrician as part of a prospective survey of major congenital malformations in children born to residents of an area of Lancashire between 1957 and 1991. Temporal changes in prevalence rates were detected by a grid search technique using Poisson log linear models. These models were also used to determine the association between prevalence and ionising radiation from atomic fall out. SETTING--The Fylde district of Lancashire in the north west of England. PATIENTS--There were 167 cases, including five stillbirths and eight terminations, among 124,015 total births in a population which increased from about 250,000 to over 300,000 during the study period. MAIN RESULTS--There was significant increase in the prevalence of all cases conceived in 1963 and 1964, and a lesser peak in 1958 which did not quite reach statistical significance. There was no evidence that the increased prevalence in 1963-64 was a result of changes in the maternal age distribution in the population. Babies of mothers aged 35 years and over accounted for more of the variation, especially in 1958 when their increase was significant. There was a highly significant association between prevalence and radiation from fallout produced by atmospheric testing of atomic weapons. The 1963-64 peak coincided with the maximum estimated radiation dose. The lesser peak in 1958 also coincided with increased exposure to radiation from fallout, possibly enhanced by ground deposits after a fire at the Windscale reactor in October 1957. CONCLUSION--This study provides further support for low dose ionising radiation as one aetiological factor in Down's syndrome.