Evidence that smoking tobacco harms health has accumulated over 200 years, but was largely ignored before 1950, when five case-control studies associated smoking with the development of lung cancer. The idea that it might cause the disease was greeted with scepticism, and it was nearly 10 years before it became generally accepted. By then there had been additional evidence from cohort studies, and known carcinogens had been identified in tobacco tars. Cigarette smoking has now been positively associated with some 40 causes of death and negatively associated with eight or nine. A few of the associations are due to confounding, but the great majority reflect causality. In several instances cigarette smoking increases the risk of death ten-fold, and altogether it doubles the annual risk of death at all ages combined, in both sexes. Tobacco smoke in the environment also has a small effect on the health of non-smokers, particularly in infancy and childhood, but also to some extent later in life. Nearly a quarter of all deaths in men and a tenth in women in industrialised countries in 1990 were attributed to smoking, giving a total of 1.8 million a year. In 20-30 years' time the total is estimated to rise to 10 million a year, with 7 million in low income countries, if smoking habits persist unchanged.