In this paper the relation between protein allergenicity (the capacity to induce IgE antibody production or the capacity to activate mast cells sensitized with IgE antibodies induced by a cross-reactive allergen) and protein structure is discussed. While cross-reactivity is to a large degree predictable from primary sequence comparisons, the IgE-inducing capacity is mostly determined by factors other than the primary sequence. Two routes to IgE are discussed: (1) the atopic route (used by allergens from pollen and mites) in which a direct switch from mu to epsilon is common and (2) the 'modified Th2' route (used by allergens from pets) in which the class switch to IgE is often preceded by a switch to IgG4. According to this working hypothesis, the choice between these two routes is determined at the level of the germinal center activity.