Live attenuated cold-adapted influenza vaccine is undergoing evaluation in man. Several strains have proven to be safe, immunogenic, nontransmissible, and protective against experimental challenge. In this study of A/Peking/2/79(H3N2), with six internal genes from the cold-adapted (Ca) parent A/Ann Arbor/6/60(H2N2), we encountered at the highest input multiplicity, 28% illness rate among individuals infected with vaccine. Reversion to wild type and excessive viral replication did not occur. Physical characteristics of the vaccine were similar to nonreactogenic vaccine A/Washington/897/80(H3N2). At ten- and 100-fold lower input multiplicities, infection frequency was maintained, but reactions did not occur. We compared the observations in this study with those made in a similar study of A/Scotland/840/74(H3N2), a cold-adapted vaccine with five genes from the Ca parent in which reactogenicity also was noted. The dose of vaccine virus in relation to tissue culture infectious doses required to infect 50% of susceptibles (HID50) was proportionally lower for both A/Peking/2/79(H3N2) and A/Scotland/80(H3N2). Hence, when the vaccine was undiluted the recipients were inoculated with more than 100 HID50. We concluded that the very high input could be avoided if vaccines were screened beginning at 1/1,000 of maximum titers. Ca vaccines must be safe before they undergo field trials.