We studied the risk of thyroid neoplasia in Marshall Islanders exposed to radioiodines in nuclear fallout from the 1954 BRAVO thermonuclear test. We screened 7266 Marshall Islanders for thyroid nodules; the islanders were from 14 atolls, including several southern atolls, which were the source of the best available unexposed comparison group. Using a retrospective cohort design, we determined the prevalence of thyroid nodularity in a subgroup of 2273 persons who were alive in 1954 and who therefore were potentially exposed to fallout from the BRAVO test. For those 12 atolls previously thought to be unexposed to fallout, the prevalence of thyroid nodules ranged from 0.9% to 10.6%. Using the distance of each atoll from the test site as a proxy for the radiation dose to the thyroid gland, a weighted linear regression showed an inverse linear relationship between distance and the age-adjusted prevalence of thyroid nodules. Distance was the strongest single predictor in logistic regression analysis. A new absolute risk estimate was calculated to be 1100 excess cases/Gy/y/1 X 10(6) persons (11.0 excess cases/rad/y/1 million persons), 33% higher than previous estimates. We conclude that an excess of thyroid nodules was not limited only to the two northern atolls but extended throughout the northern atolls; this suggests a linear dose-response relationship.