Ten antisera were produced in rabbits by two or three intravenous injections of inactivated whole influenza type A virions. All contained haemagglutination-inhibition (HI) antibody directed predominantly to an epitope in antigenic site B and, in addition, various amounts of antibodies to an epitope in site A and in site D. The ability of untreated antisera to select neutralization escape mutants was investigated by incubating virus possessing the homologous haemagglutinin with antiserum adjusted to contain anti-B epitope HI titres of 100, 1000 and 10000 HIU/ml. Virus-antiserum mixtures were inoculated into embryonated hen's eggs, and progeny virus examined without further selection. Forty percent of the antisera at a titre of 1000 HIU/ml selected neutralizing antibody escape mutants as defined by their lack of reactivity to Mab HC10 (site B), and unchanged reactivity to other Mabs to site A and site D epitopes. All escape mutant-selecting antisera had a ratio of anti-site B (HC10)-epitope antibody[ratio ]other antibodies of [ges ]2·0[ratio ]1. The antiserum with the highest ratio (7·4[ratio ]1) selected escape mutants in all eggs tested in four different experiments. No antiserum used at a titre of 10000 HIU/ml allowed multiplication of any virus. All antisera used at a titre of 100 HIU/ml permitted virus growth, but this was wild-type (wt) virus. We conclude that a predominant epitope-specific antibody response, a titre of [ges ]1000 HIU/ml, and a low absolute titre of other antibodies ([les ]500 HIU/ml) are three requirements for the selection of escape mutants. None of the antisera in this study could have selected escape mutants without an appropriate dilution factor, so the occurrence of an escape mutant-selecting antiserum in nature is likely to be a rare event.