An outbreak of equine influenza H3N8 in a riding school is described retrospectively with emphasis on diagnosis and putative vaccine failure. In March 1995 an outbreak of equine influenza occurred among 11 horses in a riding school, where most horses had received basic primary immunizations and several booster vaccinations against influenza. Six of the 11 diseased horses had received their last booster vaccination within 5 months of the outbreak. Nevertheless, the influenza infection spread rapidly and clinical manifestations were prominent with frequent, harsh, dry coughing often accompanied by high fever. Nasal swabs were taken from 11 diseased horses. Influenza A virus of the equine H3N8 (equi-2) subtype was isolated from five nasal swab extracts. Stored nasal swab extracts were also retrospectively investigated in two different enzyme immunoassays designed to detect the type-specific conserved nucleoprotein of influenza A viruses, and in a single-tube reverse transcription-PCR (RT-PCR) using a set of primers based on highly conserved regions of the matrix gene of influenza A viruses. Five nasal swab extracts were found positive in a DAS-ELISA and seven in the Directigen((R)) Flu A (DFA) assay, respectively. Two nasal swab extracts from which virus was isolated did not give a positive result in the DAS-ELISA, and one of these also did not give a positive result in the DFA assay. Nine nasal swab extracts were found positive by RT-PCR. Moreover, all virus isolation and/or ELISA positive nasal swab extracts were confirmed by RT-PCR. Three nasal swab extracts were negative by virus isolation, PCR and ELISA. A significant rise in HI titre against influenza A/eq/Miami/63 (H3N8) virus was detected in seven of the nine paired sera available. In acute phase serum samples from 10 horses, SRH antibody levels varied widely. However, some horses with high, or at least putatively clinically protective SRH antibody levels, showed clinical signs and infection was confirmed. Antigenic analysis of two isolates showed that A/eq/Holland/1/95 (H3N8) and A/eq/Holland/2/95 (H3N8) cluster with the UK isolate Osgodsby/92, the Swedish isolate Borlänge/91 and some other European isolates, with H/2/95 identical in reactivity to Borlänge/91 and H/1/95 more similar in reactivity to Osgodsby/92 than H/2/95. Nucleotide and deduced amino-acid sequences showed large differences of both isolates as compared with Miami/63, Fontainebleau/79 and Kentucky/81, the influenza A H3N8 subtype strains incorporated in the vaccines used in this riding school. The role of antigenic drift in vaccine breakdown is discussed in the light of evidence for vaccine breakdown in the UK in 1989, Sweden in 1991 and in the USA since 1991.