Models occupy an essential position in the study of infectious disease as a result of the ethical problems of exposing humans to potentially lethal agents. Deliberately induced infections in well-defined animal models provide much useful information about disease processes in an approximation of their natural context. Despite this, animal models are not the natural disease process, and recent experimental advances show, perhaps not unsurprisingly, that there are large differences between natural infections and animal models. Focusing on mouse models of bacterial pathogens, we discuss some of these discrepancies and suggest ways of improving model systems in the future.