The measurement of patterns of health and illness is far from easy, even when definitions have been agreed. Researchers employ two main sources when measuring ill-health—official statistics and community surveys. Official statistics provide data on persons who have had some contact with the health services—so-called ‘treated’ cases. This means that, whilst the data are more readily accessible, they are contaminated by illness behaviour; that is, by people's willingness to use health services, their access to services, their perception of their illness, and so forth. Community surveys circumvent this problem by screening general populations independently of health service contact. However, they usually rely on various self-report scales for measuring ill-health, and the relation between these measures and clinically defined sickness is problematic. Not surprisingly perhaps, mortality statistics are often used as a substitute measure of morbidity statistics on the grounds that in developed societies where many people die from degenerative conditions, the age at which an individual dies offers some measure of their life-time health. Given the limitations of these different measures it is necessary, where possible, to examine a range of data in analysing the social distribution of sickness.