Natural infections with influenza A viruses have been reported in a variety of animal species including humans, pigs, horses, sea mammals, mustelids and birds. Occasionally devastating pandemics occur in humans. Although viruses of relatively few HA and NA subtype combinations have been isolated from mammalian species, all 15 HA subtypes and all 9 NA subtypes, in most combinations, have been isolated from birds. In the 20th century the sudden emergence of antigenically different strains transmissible in humans, termed antigenic shift, has occurred on four occasions, 1918 (H1N1), 1957 (H2N2), 1968 (H3N2) and 1977 (H1N1), each time resulting in a pandemic. Genetic analysis of the isolates demonstrated that 'new' strains most certainly emerged after reassortment of genes of viruses of avian and human origin in a permissive host. The leading theory is that the pig represents the 'mixing vessel' where this genetic reassortment may occur. In 1996, an H7N7 influenza virus of avian origin was isolated from a woman with a self-limiting conjunctivitis. During 1997 in Hong Kong, an H5N1 avian influenza virus was recognised as the cause of death of 6 of 18 infected patients. Genetic analysis revealed these human isolates of H5N1 subtype to be indistinguishable from a highly pathogenic avian influenza virus that was endemic in the local poultry population. More recently, in March 1999, two independent isolations of influenza virus subtype H9N2 were made from girls aged one to four who recovered from flu-like illnesses in Hong Kong. Subsequently, five isolations of H9N2 virus from humans on mainland China in August 1998 were reported. H9N2 viruses were known to be widespread in poultry in China and other Asian countries. In all these cases there was no evidence of human to human spread except with the H5N1 infections where there was evidence of very limited spread. This is in keeping with the finding that all these viruses possessed all eight genes of avian origin. It may well be that infection of humans with avian influenza viruses occurs much more frequently than originally assumed, but due to their limited effect go unrecognised. For the human population as a whole the main danger of direct infection with avian influenza viruses appears to be if people infected with an 'avian' virus are infected simultaneously with a 'human' influenza virus. In such circumstances reassortment could occur with the potential emergence of a virus fully capable of spread in the human population, but with antigenic characteristics for which the human population was immunologically naive. Presumably this represents a very rare coincidence, but one which could result in a true influenza pandemic.