The effectiveness of immunization against influenza in elderly persons is uncertain. A retrospective cohort study in a New York City nursing home examined the occurrence of pneumonia and its related mortality over three consecutive influenza seasons (Nov 1 through April 30, 1979 to 1980, 1980 to 1981, and 1981 to 1982). Nearly one half of approximately 450 residents (mean age, 84 years) accepted immunization each year. The vaccinated and unvaccinated groups were similar. The attack rate of pneumonia did not differ significantly between the vaccinated and unvaccinated groups in any of the three influenza seasons. When influenza was occurring in the community (1979 to 1980 and 1980 to 1981), however, the risk of death from pneumonia in the unvaccinated group was three-fold higher than in the vaccinated group (60% vs 18% and 73% vs 25%, respectively). In a year when influenza was specifically sought and not found in the facility (1981 to 1982), however, vaccination did not affect pneumonia-related mortality. This study also suggests that estimates of mortality due to pneumonia should include deaths that occur up to 60 days after onset of pneumonia; shorter follow-up may overestimate the protective effect of vaccination.