In a prospective study of community-dwelling people 60-90 years of age, we examined the coverage of influenza vaccine during 1992-3 and 1993-4, the efficacy of vaccination in reducing serologically-confirmed clinical episodes of influenza A during 1993, and the effect of cigarette smoking. During 1992 and 1993, influenza vaccine was given to 106/215 (49 %) and 120/204 (59 %) people with risk conditions, and 84/225 (37 %) and 103/235 (44 %) without risk conditions. Influenza vaccination and general practitioner consultations during 1992 were independent predictors of vaccination in 1993, but current smoking was a negative predictor. Of 209 unimmunized people, 8/35 (23 %) smokers had clinical influenza as compared with 11/174 (6 %) non-smokers (OR 4.4, 95 % CI 1.6 to 11.9). Of 371 non-smokers, 1/197 (0.5 %) vaccinees had influenza as compared with 11/174 (6 %) non-vaccinees (OR 0.075, 95 % CI 0.587 to 0.009). No cases of influenza occurred among 21 current smokers who were vaccinated.