During the 7th Complement Genetics Workshop, Mainz, Germany, May 1998, a complement component C4 typing exercise took place with the aim of applying present technologies to the definition of reference C4 alleles/phenotypes and the recognition of nonexpressed (Q0) C4 alleles within expressed haplotypes. Eleven samples were submitted from 3 laboratories and tested by 14 participating laboratories with basic protein-typing technologies; in addition, each laboratory contributed data from local expertise. The samples were introduced to the reference typing for one or more characteristic allotype or for partial or total nonexpression of one isotype. The blinded samples were centrally evaluated and the results discussed among the participants at a plenum meeting. From the results, the samples could be classified into a group of common, easy to diagnose pheno-/allotypes, less common but still unanimously recognised variants, and a third group with difficult pheno-/allotypes. Within the latter group, the allotypes were either new (C4A '92'; C4B '93') and/or showed partial or total reversed antigenicity and unusual Rodgers/Chido (Rg/Ch) PCR subtypes (C4A '92'; C4A 12; C4B '35'; C4B '13'). Semiquantitative C4-alpha-chain estimates of relative isotype levels correlated well with the number of alleles seen at each locus by agarose gel electrophoresis, and were superior to other isotype quantitation methods. From the evaluation of the reference typing it was concluded that the recognition of rare, aberrant or hybrid C4 alleles with partial or total reversed Rg/Ch antigenicity or monoclonal reactivity is still difficult in most instances; besides isotype-dependent lysis, relative migration values, immunoblots with Rg- and Ch-specific monoclonal antibodies, Rg/Ch PCR typing, side-by-side comparison with already described allotypes will ultimately be required. The recognition of nonexpressed alleles within C4A and C4B expressed phenotypes remains the major obstacle in C4 genetic typing. Finally, a conclusive interpretation of DNA typing results will be achieved only in the context of complete allotyping results at the protein level, and at present cannot replace conventional protein allotyping.