(Selections from Contemporary Debates in Philosophy of Science ed. by Christoper Hitchcock)

Many theories posit entities that cannot be directly observed: atoms, quarks, magnetic fields, genes, mental
representations, and so on. By their very nature, it is never possible to confirm the existence of such entities by
direct observation. (There is, however, a question about what counts as a "direct" observation - is looking
through a microscope allowed?) Nonetheless, theories that posit such entities often do make predictions that can
be tested by observation, and some theories are highly successful in this enterprise. Does this sort of predictive
success give us reason to believe that the posited entities really exist? One argument, often dubbed the "miracle
argument," claims that the empirical success of a theory would be a miracle, or at least a coincidence of cosmic
proportions, if the theory were wrong about the basic entities underlying the phenomena in question. Jarrett
Leplin and also Andre Kukla and Joel Walmsley, reject the miracle argument in its broadest form. Nonetheless,
Leplin argues for the realist position that we are sometimes warranted in believing in the existence of the
unobservable entities postulated by science. First, there is a certain presumption in favor of belief in these entities,
and none of the standard antirealist arguments are successful in dislodging this belief, once admitted. Secondly,
one specific kind of predictive success, the prediction of novel phenomena, does provide particularly strong
warrant for the belief in unobservables. Kukla and Walmsley criticize this argument, and other refinements of the
miracle argument, on the grounds that they presuppose a certain conception of what it would be to explain the
predictive success of a scientific theory. They arrive at the strong conclusion that no version of the miracle
argument could possibly give us grounds for believing in unobservable entities.

The Realist Side of the Debate:

The Anti-Realist Side of the Debate