Proponents of an accuracy-centered epistemology have argued that proper scoring rules can be used to assess the accuracy of degrees of belief (or “credences”), and have suggested that certain core epistemic norms for credences can be understood and justified with the help of such rules.  Many who go this route are attracted to the idea that revising credences in light of new evidence should proceed by a process that S. Zabell and P. Diaconis once called “mechanical” updating.  The idea is that, upon receiving a new item of data, an agent should always move to the credal state, among those consistent with that data, which maximizes expected accuracy.  Mechanical updating provides a rationale for Bayesian conditioning in contexts where the new data involves learning some proposition with certainty.  However, for less conclusive experiences it can be shown that each proper score has its own characteristic mechanical update rule.  For example, the so-called logarithmic score has Jeffrey conditioning as its “mechanical” update, while H. Leitgeb and R. Pettigrew have shown that the Brier (quadratic loss) score has a completely different “mechanical” update.  Since there are overwhelming epistemic reasons to prefer Jeffrey conditioning to any other update rule, it looks as we must either jettison mechanical updating or embrace the log score as the one true measure of epistemic accuracy.  Neither option is appealing: the accuracy-centered approach seems deeply committed to mechanical updating, but the approach’s appeal diminishes quite drastically if it is forced to single out some particular score as correct.  Moreover, as Joyce will show, the log score has some serious drawbacks when it comes to updating (in certain contexts);  specifically, it classifies far too many things as learning experiences (in the sense of Skyrms), and fails to recognize that certain sorts of belief changes, even reliable ones, can decrease overall accuracy.  Fortunately, there is no real dilemma here.  As Joyce will show, a proper application of the mechanical updating process will mandate Jeffrey conditioning as the uniquely correct belief revision rule in all contexts where it can be applied.  In the course of making the case for this conclusion some suggestions (and speculations) will be made about how to relate scoring rules to learning experiences and the various “value of learning” results that one can find in the literature.