Bas van Fraassen’s reflection principle has come in for a lot of criticism since he first proposed it in 1984. Counterexamples abound, and the counterexamples are both robust and diverse. Everyone agrees that the principle, if defensible at all, requires serious qualification. In this talk I will try spell out, or at least point to, qualifications that are sufficient for a defensible version of the principle, and then consider some applications of it. A notion of endorsement will be the primary concept that I will argue is needed for an appropriate qualification. My main concern is not to defend the principle, but to reflect on the more general issue that the reflection principle helps to bring out: the role in inductive reasoning, belief change, and rational discussion of an agent’s doxastic attitudes (at a particular time) about his or her doxastic state at a different time, or about the doxastic states of other agents. Knowledge and credence are characterized in terms of sets of alternative possibilities (epistemic alternatives), and measures on such sets. Attributing knowledge and credence to others requires characterizing, from one perspective, the epistemic alternatives that are available from a different perspective. Comparing a different epistemic situation with one’s own requires calibrating the possibilities, as seen from one perspective, with the possibilities as seen from a different one, and this kind of calibration is not always straightforward. After ruminating a bit on the abstract notions of endorsement and calibration, I will look at their role in an application of the reflection principle to a notorious example: the problem of Sleeping Beauty.