Smoking is strongly associated with age-specific death rates for a number of diseases. Increased age-specific death rates for a disease may imply either more deaths from the disease with increased absolute lifetime risk, or earlier deaths, without increased absolute lifetime risk. The British doctor smoking data are re-analysed using lifetable methods for survival, cumulative mortality and the disease-specific cumulative mortality. The most significant effect of smoking is on survival: life shortening amounts to three years for light smokers, five for moderate smokers, and eight for heavy smokers, compared with those who never smoked. Smoking increases the absolute number of deaths from some causes, including lung cancer; for other causes, including ischaemic heart disease, it brings forward death without increasing the absolute number of deaths. The smoking associations with more or earlier death have implications for research into the mechanisms of disease causation, for health promotion, for rational health-care planning, and for social policy.