The deduced amino acid sequences of the haemagglutinins of avian influenza viruses, isolated from an outbreak in turkeys in Norfolk, England in 1991/92, were determined by PCR amplification and cycle sequencing. Both the highly pathogenic and avirulent isolates had the same cleavage site sequence with multiple-basic amino acids, which normally would be expected only for the former. Clones derived by plaque picking from the highly pathogenic isolate ranged from low to very high pathogenicity in vivo and these, and the original isolates, showed nucleotide and amino acid variation at one or more of five possible sites, none of which were at the cleavage site. None of these site variations correlated with pathogenicity, suggesting that the factor responsible for the suppression of the expected effects of the multiple-basic amino acid haemagglutinin cleavage site in the avirulent isolate may not have been part of the haemagglutinin amino acid sequence.