Accept no substitutes:  Against best-system theories without naturalness

According to the best-system theory, a law of nature is nothing more than a certain kind of pattern in what actually happens. Out of all the possible "systems'' (sets of sentences), there is one that does a better job than all of the others in summarizing a lot of what actually happens in a simple way; the sentences in this best system, according to the best-system theory, are the laws of nature. David Lewis's version of this idea appealed to a distinction between natural and non-natural properties. But various people have objected that this opens up an epistemic gap between science and metaphysics, and have gone on to propose versions of the best-system theory that do not appeal to naturalness. I defend Lewis's version of the best-system theory against the objection, and discuss various problems confronting naturalness-free versions.