Naturalism undoes itself in the sense here that asks: Why should we trust human reason anyway?
And why should that reason be the ultimate arbitor or reality?
The argument is structurally identical to the one he would make a couple of years later in his book Miracles, in a chapter called “The Cardinal Difficulty of Naturalism.” There he argues that if human reason is a product of the blind, undirected forces of natural selection, and unrelated in any necessary or intrinsic way to the way things actually are, then we cannot trust the claims made by the use of human reason, including the claim that human reason is a product of the blind undirected forces of natural selection. Naturalism, then, says Lewis, undoes itself.
This reminds me of a lecture I heard in gradschool about the nature of reality.
It was, of course, obsurd – but the point of the talk was that life would not look much different if we were simply brains in an electrical vat living a simulation.
In a world of 100% naturalism, why trust human perception?