This thesis discusses self-referential probabilities in some gory detail. We discuss a number of semantics models and initial work on how rationality considerations should apply in such cases.