In a prospective study conducted on the island of Guernsey a cohort of 5162 ostensibly healthy women was enrolled between 1967 and 1976. Blood samples were drawn from each participant, who also completed a questionnaire, which provided information on established risk indicators in human mammary carcinogenesis. Plasma selenium levels were measured in 46 breast cancer cases diagnosed a mean of 11 (S.D. 4) years after entry into the study cohort and in an age-stratified sample of 138 women drawn from the study base. Plasma selenium level in the cases was 109 (28) micrograms/l and in the base sample 103 (22) micrograms/l (95% confidence interval for the overall difference, -2 to 14 micrograms/l). The adjusted relative risk of developing breast cancer in the different quartiles of the selenium distribution was 0.80, 0.79, 0.72 and 1.00, respectively. Thus, in the present study selenium was not a strong indicator of human breast cancer risk.