In order to study the effects of filtering on the serum cholesterol-elevating effect of boiled coffee, 20 healthy volunteers consumed, in random order, 6-10 dl d-1 of strong boiled coffee (BC) and similarly boiled coffee that had been passed through a conventional paper filter (BFC), for periods of 4 weeks in a crossover design. During periods of BC consumption serum total cholesterol and LDL-cholesterol levels (P less than 0.05), as well as serum triglyceride and apoprotein B concentrations and the LDL/HDL ratio (P less than 0.01), were significantly higher than during BFC periods. Serum HDL-cholesterol and apoprotein A-I levels remained unchanged. Filtering removed more than 80% of the lipid-soluble substance that was present in boiled coffee. The results indicate that the hypercholesterolaemic factor in boiled coffee, which is presumably lipid-soluble, is retained by the paper filter. They also suggest that boiling is not essential for the previously observed difference between the effects on serum lipoproteins of boiled coffee and filtered coffee.