Larmer considers an argument as follows:
1) If one is a metaphysical naturalist then one should be a methodological naturalist. i. e. refuse ever to postulate nonphysical entities as the cause of a physical event.
2) One should not believe in nonphysical entities without good evidence.
3) There is no good evidence for nonnatural entities.
4) Therefore one should accept metaphysical naturalism, and by logical extension, methodological naturalism.
He then develops a dialogue between the metaphysical naturalist and his opponent over premise 3.
NN: I disagree that there is no good evidence for nonnatural entities.
MN: Such evidence cannot exist.
NN: Why not?
MN: Because any investigation of the causes of physical events must employ methdological naturalism, i. e., it must assume that it is never, in principle, legitimate to posit a nonnatural cause for a physical event.
NN: Remind me once more of your good reason for thinking that metaphysical naturalism is true.
MN: The good reason for thinking that metaphysical naturalism is true is that there is no
good evidence that nonnatural entities exist.
NN: Would methodological naturalism ever permit one to posit a nonnatural entity as the
cause of a physical event?
MN: No. I have already made that clear.
NN: Let me get this right. Your acceptance of metaphysical naturalism is based on the fact
that there exists no evidence that nonnatural entities ever cause physical events?
NN: And your endorsement of methodological naturalism follows from your acceptance of
NN: This seems question-begging. You endorse metaphysical naturalism on the basis that
there exists no evidence that nonnatural entities ever cause physical events, yet adopt a methodology
that rules out the possibility of ever recognizing evidence of nonnatural causes. You are
using your metaphysic to justify your acceptance of methodological naturalism, but your acceptance
of methodological naturalism serves to guarantee that even if evidence for the existence
of nonphysical causes exists it can never be recognized as such.