Bayesian epistemology provides formal norms that govern our degrees of belief both at a time and over time. It tells us that our degrees of belief ought to obey the probability axioms. It tells us that, when we get evidence, we ought to revise our beliefs in accordance with our conditional probabilities. Bayesianism can be made compatible with the norms for evidence that traditional epistemology has to offer us. But the question of whether Bayesianism itself implies some norm for evidence has never been addressed. This dissertation considers this question and develops two new Bayesian updating rules as a response to it. These rules refine the requirements that Bayesianism already provides us with by supplementing these requirements with two formal accounts of evidence. In chapter two, I set some of the groundwork for the dissertation by comparing the problem of how evidence, and the experience that gives rise to it, constrain an update to two problems in the Bayesian literature that have similar structures, but that are better understood. In chapters three and four, I look at some places where the literature has foundered on the lack of an account of evidence. I consider two discussions that illustrate how the lack of a constraint on how experience gives rise to an update causes us to see problems where there aren't any and to overlook problems where they do indeed exist. In chapter five, I develop a new account of the structure of Bayesian justification, which I call Bayesian coherentism. Bayesian coherentism is an updating norm that is motivated, both by the problems of the previous chapters and by the desire to provide a unified account of the structure from which updates on certain and uncertain evidence proceed. Finally, in chapter six, I develop a different norm that is guided by considerations similar to those that motivate Bayesian coherentism, but that is also compatible with the recently popular idea that there are no norms of diachronic rationality.