From John Loftus' Review of Beversluis
The Argument From Reason, as best seen in Lewis’ book, Miracles, “is the philosophical backbone of the whole book,” from which “his case for miracles depends.” (p. 145). Lewis champions the idea that if naturalism is true such a theory “impugns the validity of reason and rational inference,” and as such, naturalists contradict themselves if they use reason to argue their case. If you as a naturalist have ever been troubled by such an argument you need to read Beversluis’ response to it, which is the largest chapter in his book, and something I can’t adequately summarize in a few short sentences. Suffice it to say, he approvingly quotes Keith Parsons who said: “surely Lewis cannot mean that if naturalism is true, then there is no such thing as valid reasoning. If he really thought this, he would have to endorse the hypothetical ‘If naturalism is true, then modus ponens is invalid.’ But since the consequent is necessarily false, then the hypothetical is false if we suppose naturalism is true (which is what the antecedent asserts), and Lewis has no argument.” (p. 174).
In response to Parsons' comment, that's not how the argument from reason goes. If naturalism is true, then no one ever performs a modus ponens inference, and this can be for a number of different reasons.
1) If naturalism is true, then there are no propositional attitudes. Propositional attitudes are necessary for modus ponens inferences, so no one would actually ever perform a modus ponens inference if naturalism is true.
2) If naturalism is true, then there is no mental causation. One mental event cannot cause the occurrence of another mental event in virtue of its content, if naturalism is true.
3) If naturalism is true, then logical laws have no psychological relevance. Only physical laws can be relevant to physical events if naturalism is true; logical laws will be followed only if the physical order to disposes the brain to follow them. There could be arguments in accordance with reason but never from reason, to use Kantian terminology. I'm not saying that if naturalism is true there would be no logical laws, but rather those laws would not and could not have anything to do with what anyone things.
In other words, the argument says that if naturalism is true, then no one reasons validly. Modus ponens would be eternally a valid form of inference, but that fact would be completely irrelevant to any actual reasoning processes, and would be inoperative.