Molecular changes in the haemagglutinin (HA)-coding regions and proteolytic cleavage sites from multiple H5N2 subtype viruses isolated during a recent outbreak of avian influenza (AI) in central Mexico have been characterized. Eighteen isolates, collected during a 15 month period (October 1993 to January 1995) from six central states, were sequenced. None of the 18 predicted HA1 amino acid sequences were identical and changes were not restricted to a specific region of the sequence. Phylogenetic analyses of the HA1 sequences demonstrated two virus lineages, designated Puebla and Jalisco, with sequence variation as high as 10.5 percent for amino acid and 6.2 percent for nucleotide sequences. During the latter months of the surveillance period, highly pathogenic (HP) strains of AI emerged causing lethal disease in commercial poultry flocks. In each of the HP strains isolated, the HA protein was cleaved in chicken embryo fibroblast cells in the absence of trypsin, and two alterations not found in earlier non-HP isolates were detected. In the HA protein, HP strains all had a glutamic acid --> lysine substitution at amino acid position 324 and an insertion of arginine and lysine as new residues 325 and 326. The insertion appears to be due to a duplication of the nucleotide sequence AAAGAA at nucleotide positions 965-970 of the HA1-coding region. Computer-assisted secondary structure analyses place the target for the insertion in a predicted RNA stem-loop structure. A mechanism is suggested by which the polymerase duplicates the sequence.