The prevention of viral infection by vaccination relies on stimulating an appropriate immune response in order to reduce the probability with which a virus can establish an infection. Post-vaccination antibody responses have therefore been associated with reducing the probability with which an individual can be infected (i.e., the vaccine's "impact"). Quantifying this relationship is essential in evaluating new vaccines, especially since comparisons between vaccines, and vaccine licensure, may be dependent on antibody responses alone. In this paper two principal questions are identified which need to be addressed in the evaluation of subunit vaccines: i) how do specific antibody levels relate to complete protection from infection or disease and ii) how do antigenic subunits interact in developing protection when combined together in a single vaccine. The aim is to identify explicitly certain assumptions that are frequently made implicitly in the discussion of vaccine action. First, antibody levels are related to levels of protection through a novel statistical analysis of incidence data from a published hepatitis B vaccine trial. The antibody response observed after influenza A virus infection is discussed in relation to the selection of neutralisation escape variants. Finally, by way of example, a theoretical situation is examined and three simple models of subunit vaccine action are constructed in order to describe how antibody levels may be related to population level phenomena such as the elimination of an infection by mass vaccination.