It is reasonable to believe that the best (or only) explanation of certain phenomena is true. This is why we are entitled to believe both in successful scientific theories, and in scientific realism, which explains that theories are successful because they are true. This is denied by strict empiricists, according to whom only empirical evidence should determine theory-choice: for them there is no difference between a theory T, and its "surrealist transform" T*, the claim that all phenomena are as if T were true. But there is a fundamental difference, that T explains the phenomena while T* does not. Historical cases of surrealism were Bellarmine's instrumentalism, Berkeley's theological explanation of empirical regularities, and Gosse's creationist explanation of the geological evidence for evolution. Contemporary examples are van Fraassen's constructive empiricism and Stanford's explanation that T is successful because it is "predictively similar" to the true theory. Moreover, the successful predictions made by false theories are not a counterexample to the realist explanation of scientific success, as claimed by Laudan, because either they were not novel, or they can be explained by the partial truth of those theories (to be distinguished from verisimilitude).