The data from the Oxford Vegetarian Study consists of the observation of around 11,000 individuals followed since the early 1980s. There are around 6000 subjects who do not eat meat and 5000 roughly matched individuals who do. An assiduous questionnaire survey was conducted at recruitment which inquired about health status, diet, drinking and other habits. We have tested the hypothesis that the U-shaped curve relating quantity of alcohol consumed and health is an artefact of selection of some individuals with high consumption and high risk migrating to the no-consumption group but retaining a high risk. The Oxford Vegetarian Study consists of a high proportion of lifelong teetotallers and ex-drinkers, and hence is particularly suitable for testing this hypothesis. We have examined the standardized rates of cardiovascular risk factors among the different dietary and drinking groups separately for men and women, as the bulk of the observed relationship of alcohol with health is mediated through cardiovascular mechanisms. We were unable to find a difference in the prevalence of risk factors between ex-drinkers and teetotallers, but we did find differences associated with dietary practices, particularly among females. These data cast some doubt on the hypothesis that selection may explain the apparent protective effect of moderate drinking when compared with groups currently not drinking (for whatever reason) and heavy drinking.