Abstract

The No Miracles Argument (NMA) is perhaps the most prominent argument in the debate about scientific realism. It contends that the truth of our best scientific theories is the only hypothesis that does not make the astonishing predictive and explanatory success of science a mystery. However, the argument has been criticized from a Bayesian point of view as committing the base rate fallacy. We provide two Bayesian models (one related to the individual-theory-based NMA and one related to the frequency-based NMA) that respond to that objection. The first model takes into account the observed stability of mature scientific theories, the second the success frequency of theories within a scientific discipline. We conclude that the NMA can be used to defend the realist thesis and that its validity is a highly context-sensitive matter.