A Model for Epistemic Games Tanja Hötte Abstract: In this thesis, I introduce and investigate a new framework for modeling epistemic information in games that allows to express more general types of uncertainty than, for instance, imperfect information. In particular, we liberate the view on epistemic alternatives: not just other states of the same tree are considered, but completely independent other game trees. In doing this, we open a wider perspective on modeling games in epistemic logic, beyond existing work on standard imperfect information games. We give a basic language and a minimal logic for epistemic game models and proved completeness. Furthermore we discuss several other modal formulas one might take as axioms for special subclasses of epistemic game models, such as almost bending back. By extending the language with inverse action relations and some iteration operators, we increase its expressiveness, e.g. it now becomes possible to define the property of being a single tree model by a modal formula. In comparing our static approach to the dynamic update systems of Baltag, Moss and Solecki we find several means of extending their account to include the treatment of games.