The main vehicle of speaking and reasoning about knowledge in philosophy has recently been epistemic logic.¹ Even though epistemic logic is not the only relevant language-game in town, it offers a useful perspective here, for certain other approaches can be thought of as improvements on epistemic logic. In its axiomatic-deductive forms, epistemic logic is normally considered a branch of modal logic, and its semantics is usually subsumed under the misleading heading of "possible-worlds semantics". I will not attempt here a survey of the existing literature on epistemic logic.² Most of this literature is focused on syntactical (e.g., deductive and axiomatic) methods of dealing with knowledge representation and reasoning about knowledge. This is in my view a serious defect in much of the current work on epistemic logic. For typically the most interesting problems and solutions are found by considering the model-theoretical (semantical) situation. For this reason, I will not attempt here a survey of existing literature, but a review of some of the central conceptual problems arising in epistemic logic.