We examined the relation between consumption of 22 dietary items and subsequent bladder-cancer incidence in a cohort of atomic-bomb survivors in Japan. Subjects were 38,540 people (14,873 men and 23,667 women) who responded to a mail survey carried out between 1979 and 1981 and who had no known cancer diagnosed before the start of follow-up (1 January 1980 for men, 1 February 1981 for women). Consumption frequencies for 22 dietary items were ascertained with pre-coded answers. As of the end of 1993, there were 114 (83 men and 31 women) incident cases of bladder cancer among 450,326 person-years at risk. Statistical analysis was done using Poisson regression for grouped survival data Consumption of green-yellow vegetables and fruit were protectively associated with risk. Adjusted for gender, age, radiation exposure, smoking status, educational level, body-mass index and calendar time, the relative risk (RR) for those consuming green-yellow vegetables 2-4 times per week and almost everyday was 0.62 [95% confidence interval (CI) 0.39-0.98] and 0.54 (95% CI 0.30-0.94) respectively, as compared with those consuming once per week or less. The corresponding RR for fruit consumption was 0.50 (0.30-0.81) and 0.62 (0.39-0.99) respectively. Chicken consumption was unexpectedly associated with decreased risk, but additional adjustment for consumption did not change the relation of green-yellow vegetables or of fruit to risk. The consumption of the other dietary items, including meat and green tea, was not related to risk. The findings add to evidence that high consumption of vegetables and fruit are protective against bladder cancer.