A combined analysis of hydrophilicity, accessibility and flexibility parameters of the deduced amino acid sequence of measles virus (MV) haemagglutinin (H) was used to select 10 regions for synthesis of 10- or 11-amino acid-long peptides. Nine of these sites are probably exposed on the surface of the protein, as polyclonal sera against either purified MV or purified H bound to these peptides as tested by enzyme immunoassay (EIA). Nevertheless, human sera from acute or chronic MV infection did not bind significantly to any peptide, indicating that the selected sites do not function as natural complete epitopes. All antisera raised in rabbits against keyhole limpet haemocyanin-conjugated peptides had a high titre to the homologous peptide and nine of them bound to MV lysate antigen, purified MV and/or purified H as tested in EIA. None of the sera had haemagglutination-inhibiting antibodies and only one antiserum (against peptide 185-195) had a neutralizing antibody titre of 1/160. Only a minority of the antisera were positive in Western blot (four of 10), radioimmunoprecipitation (two of 10) or immunofluorescence (three of 10). The results indicate that the computer program used in this analysis can predict surface-exposed areas of MV H but that the small peptides synthesized have little resemblance to natural antigenic sites.