Abstract. This paper makes a foundational contribution to the discussions on the very nature of belief change operations. Belief revision and pdate are investigated within an abstract framework of epistemic states and (qualitative or quantitative) conditionals. Moreover, we distinguish between background knowledge and contextual information in order to analyse belief change more appropriately. The rich epistemic representation framework allows us to make a clear conceptual distinction between revision and update on the one side, while revealing structural similarities on the other side. We propose generic postulates for revision and update that also apply to iterated change. Furthermore, we complete the unifying picture by introducing universal inference operations as a proper counterpart in nonmonotonic reasoning to iterated belief change.