The reliability of the determination of antibody avidity in polyclonal sera by indirect sandwich ELISA was studied. Binding of IgM and IgG (sub)classes in unpurified serum to Streptococcus pneumoniae type 3 capsular polysaccharide, which was coated onto ELISA plates, was inhibited with different inhibitors. The inhibitor concn at which 50% inhibition of antibody binding to the ELISA coat was achieved, was used as a measure for antibody avidity. As this 50% inhibition value is dependent upon the dilution of the serum and thus upon the initial amount of free antibody, it is necessary to define (a narrow range of) final ELISA absorbance values to which the dilutions of non-inhibited sera have to be adjusted. The shapes of the serum dilution curves have a good correlation with the numerical 50% inhibition values of the antibody avidity. The inhibition ELISA is suitable to compare the avidity values of the different antibody isotypes, but two remarks should be made: (1) antibody heterogeneity should be considered to influence the results and prevent the accurate measurement of absolute numerical avidity values. Because in the ELISA system merely antibody "activity" is measured, comparison of the efficacy of vaccines by means of the 50% inhibition (avidity) value of various antibody (sub)classes can still be performed in a reliable way; (2) results of the determination of the 50% inhibition values of the different antibody (sub)classes showed them to be dependent on the molecular ratio between antibody (sub)class levels. More aspects of the determination should be taken into account, like shapes of simple dilution curves, influences of various inhibitor concns in the diluent and whole (extended) inhibition curves.