Sera from 43 persons who developed thyroid cancer on an average 4.8 years after blood sampling were compared with sera from controls. Three controls per case matched for sex, age, place of residence and year of blood sampling, with regard to serum selenium and serum copper. Cases were significantly lower in serum selenium than controls, and the estimated odds ratio of thyroid cancer increased from 1 for levels greater than or equal to 1.65 mumol/l, to 6.1 for levels 1.26-1.64 mumol/l, to 7.7 for levels less than or equal to 1.25 mumol/l. When time from blood sampling to diagnosis of the case was considered, it could be shown that the protective effect of high serum selenium concentrations was restricted to the last (less than 7) years prior to the diagnosis of thyroid cancer. The serum selenium concentration of cases tended to decrease relative to controls the shorter time was from blood sampling to the diagnosis. There was no difference between cases and controls with regard to serum copper.