The neuraminidase (NA) genes of influenza B viruses B/Maryland/59, B/Hong Kong/8/73, B/Singapore/222/79, B/Oregon/5/80, B/USSR/100/83, B/Victoria/3/85, B/Leningrad/179/86, B/Memphis/6/86, and B/Memphis/3/89 have been sequenced. The deduced amino acid sequences show high variability in the stalk domain of the NA, but a surprising degree of sequence conservation in the head region which carries all the antigenic and enzyme activity. The variable region coding for the neuraminidase stalk also translates into a variable section in the overlapping NB polypeptide, which is coded from a second reading frame that overlaps the first 100 amino acids of NA. The influenza B NAs are antigenically distinguishable with monoclonal antibodies in neuraminidase-inhibition tests, even when there is only one amino acid sequence difference. However, seven of nine escape mutants selected with monoclonal antibodies were distinguished only by the antibody used for selection. When NA heads of influenza B viruses are crystallized, there are remarkable differences in crystal morphology between neuraminidases which have very few sequence changes.