Case-control and cohort studies may be employed to assess the protective efficacy of vaccines. The appropriate measure of vaccine efficacy is shown to depend upon the mode of action of the vaccination. Two models of vaccine action are considered. In the first, vaccination is assumed to reduce the instantaneous disease-rate in the total vaccinated population by a constant proportion and, in the second, vaccination is assumed to render a constant proportion of individuals totally immune from the disease. The implications of these two models on the behaviour of different measures of vaccine efficacy in cohort studies is explored. It is shown that the design of case-control studies to measure vaccine efficacy is dependent upon which model is considered appropriate. In particular, under the second model, individuals who have already had the disease under study should not be excluded from the control group.