Ten families with 82 members were investigated for C4A- and B polymorphism in a blind trial. Phenotyping was done on neuraminidase treated sera by immunofixation and simultaneously by hemolytic overlay electrophoresis. In addition Rg, Ch, BF, C2, HLA-A, B, C, DR, and GLO were determined. After decoding the samples the reliability of blind typing was found to be 84.4% according to segregation patterns. Inconsistencies occurred mostly when A4, A2, or A92 were present. The detection of silent A*Q0 and B*Q0 alleles was more critical than that of "difficult" allotypes. The quantitation of the C4A/B ratio by densitometry of stained gels or by conventional immunochemical measurements of serum C4 level could not substantially improve the identification of A*Q0 or B*Q0. C4 dependent activity in radial diffusion hemolysis showed satisfactory correspondence with the number of expressed C4B alleles. At least three haplotypes with two C4A genes (duplicated A genes) were observed as ascertained from offspring analysis in accordance with the MHC segregation pattern. Individuals with the duplicated C4A gene (C4A*3, A*2, in the absence of any other expressed A allele or together with C4A*92) showed only partial inhibition of Rodgers antisera. Partial inhibition of Chido antisera was seen in individuals with C4B 2 (in the absence of other B allotypes). The findings support the hypothesis of at least two structural C4 loci. They also demonstrate the inconsistency of quantitative data in the recognition of silent alleles.