Parasites present a threat for free-living species and affect several ecological and evolutionary processes. Immune defence is the main physiological barrier against infections, and understanding its evolution is central for predicting disease dynamics. I review theoretical predictions and empirical data on natural selection on quantitative immune defence traits in the wild. Evolutionary theory predicts immune traits to be under stabilizing selection owing to trade-offs between immune function and life-history traits. Empirical data, however, support mainly positive directional selection, but also show variation in the form of selection among study systems, immune traits and fitness components. I argue that the differences between theory and empirical data may at least partly arise from methodological difficulties in testing stabilizing selection as well as measuring fitness. I also argue that the commonness of positive directional selection and the variation in selection may be caused by several biological factors. First, selection on immune function may show spatial and temporal variation as epidemics are often local/seasonal. Second, factors affecting the range of phenotypic variation in immune traits could alter potential for selection. Third, different parasites may impose different selective pressures depending on their characteristics. Fourth, condition dependence of immune defence can obscure trade-offs related to it, thus possibly modifying observed selection gradients. Fifth, nonimmunological defences could affect the form of selection by reducing the benefits of strong immune function. To comprehensively understand the evolution of immune defence, the role of above factors should be considered in future studies.