Theileria parva is the causative agent of East Coast fever (ECF), an important cattle disease in East and Central Africa. One of the methods for control of ECF is 'infection and treatment', a procedure in which an animal is infected with the live parasite and at the same time treated with a long-acting oxytetracycline formulation, restraining the infection and allowing a protective cellular immune response to develop. Optimal immunizing doses were estimated using models of trichotomous response: dysimmunization (death or severe reaction during immunization), immunization failure (death or severe reaction during lethal challenge) and successful immunization (neither dysimmunization nor immunization failure). In this paper we present methods of interpreting immunization trials and apply these methods to previously unpublished data from two such trials: one with a mixture of three T parva stocks and one with a single T parva stock. We explain why titration trials conducted with a cocktail of antigens could predict a suboptimal immunization dose. Indeed it is possible for a combination of three individually efficient stocks to result in a mixture with which optimal immunization response might be difficult to achieve, because of averaging effects. The corresponding interpretation provides insights into why standard immunization trials for T parva have not yielded the results that might be expected of them. The results of this work may also have implications for the use of antigen cocktails in cancer, HIV and malaria vaccine trials. (c) 2007 Elsevier Inc. All rights reserved.