Abstract

One of the arguments advanced in favor of scientific realism is the 'miracle argument'. It says that for the anti-realist the predictive success of science appears as an utter miracle. This argument indeed has some prima facie plausibility, provided that it is sharpened by construing "predictive success" as prediction of previously unknown laws and the occurrence of a consilience of inductions. Still, the history of science teaches us that it is possible to arrive at predictive success in this sense by employing radically non-referring theoretical mechanisms. The 'miracle argument' is thus unsound. Rather, the capacity of a theory to generate predictive success can be traced back to its "classificativity correspondence."