Abstract

I argue that it is possible to acquire evidence that has no probability, not even zero, prior to its acquisition. If I am right then, contrary to certain Bayesian models of confirmation, conditionalization is not the only possible basis upon which a rational agent will alter her credence in some hypothesis in response to new evidence. My conclusion follows from certain analyses of the Sleeping Beauty problem. Because those analyses are controversial, however, I alter the Sleeping Beauty scenario to generate an obvious example of evidence that has no prior probability.