In place of earlier instrumentalist and phenomenalist interpretations of science both Quine and Sellars have developed highly influential realist positions centering around the doctrine that accepting a theory as explanatory and irreducible rationally entails accepting the entities posited by the theory. A growing reaction against this realism is partially based on perceived inadequacies in the doctrines of Quine and Sellars, but even more on reconstructions of scientific explanations which do not involve such ontic commitments. Three types of anti-realistic positions are (...) considered and criticized. Hesse's neo-Duhemian position couples a statistical theory of inference to a downgrading of purely theoretical statements. Sneed uses the method of axiomatrization through set-theoretic predicates to supply a reconstruction of Kuhn's account of scientific development. Here Ramsey-reduction sentences serve to eliminate purely theoretical terms. The longest section deals with van Fraassen's semantic model based on a state-space representation of scientific theories and the use of this model in interpreting quantum mechanics. It is argued that the model is valuable, but that the anti-realist interpretation accorded it is not justified and entails serious inconsistencies. (shrink)