To assess the risk of mortality associated with different health behaviours in a Mediterranean elderly population, we have analysed the mortality experience of a cohort of 1219 non-institutionalized men and women aged 65 years or over, who had participated in the 1986 Health Interview Survey of Barcelona.At baseline, self-reported information on smoking, alcohol consumption, physical activity, and hours of daily sleep was collected through face-to-face home interviews. After 5 years of follow-up, a total of 224 (18.4%) participants had died.After adjusting for age, level of education and perceived health status, current smokers at baseline had a relative risk of dying during the follow-up of 3.0 [95% confidence interval (CI) 1.78, 5.04] compared with non-smokers. Those reporting moderate alcohol consumption had a relative risk of dying of 0.38 (95% CI 0.17, 0.86) compared with abstainers. Finally, individuals who reported a sedentary life-style had a risk of dying of 1.53 (95% CI 1.08, 2.15) compared with the more active.Although limited for interpreting associations as causal relationships, our results add to the evidence that the adverse effects on survival of smoking and of being sedentary extend later into life. Because of the high prevalence of a sedentary life-style among elderly people, results suggest that promoting physical activity may have an important role in enhancing survival in this age group. © 1995 British Geriatrics Society.