Our cognitive faculties include memory, perception and rational intuition. In science, as in every day life, these work together to produce beliefs. It is natural to assume that our cognitive faculties produce beliefs that are mostly true. But Alvin Plantinga has given a forceful argument that, on naturalism,  this assumption is unsafe.
Consider: The naturalist believes the mind “just is” the brain and so takes a belief to be something like a long-standing structure in the nervous system. The problem is that neurology can produce behaviours that increase fitness whether or not the beliefs annexed to that neurology are true. Survival, to be sure, does require cognitive devices that track crucial features of the environment and are appropriately connected to intention and muscular reflexes. That is not disputed. What is disputed is the necessary annexation between those cognitive devices and true beliefs. In fact, adaptive behaviour does not require true belief—or belief at all.
Think of an organism fleeing from a predator. Undoubtedly, its cognitive devices are tracking the predator and producing a useful response. But “tracking” itself is not belief and, so long as the neurology of the organism causes it to flee, the belief annexed to its neurology need not even contain a predator and it certainly need not be true. “It could be true,” says Plantinga, “it could be false; it doesn’t matter.”
Darwin himself was troubled by this. “With me the horrid doubt always arises whether the convictions of man’s mind, which has been developed from the mind of the lower animals, are of any value or at all trustworthy,” he wrote in a private correspondence. “Would any one trust in the convictions of a monkey’s mind, if there are any convictions in such a mind? The problem was also noticed by C. S. Lewis, the chemist J. B. S. Haldane  and atheist philosopher John Gray. “Modern humanism,” Gray writes, “is the faith that through science humankind can know the truth. But if Darwin’s theory of natural selection is true, this is impossible. The human mind serves evolutionary success, not truth.”
Plantinga’s argument applies to all beliefs but with a force that increases as beliefs become irrelevant to survival. Perception, for example, is especially relevant to feeding, fleeing, fighting and reproduction and so beliefs directly informed by perception may be taken to be more reliable. Beliefs about physics, aesthetics and philosophy, on the other hand, are irrelevant to survival. These must be regarded as far less reliable. Metaphysical beliefs, including both naturalism and theism, fall into this second category.
What then is the likelihood, on naturalism, that some belief p instantiated in an organism is true? Plantinga suggests that, since the alternatives seem about equiprobable, we should give it a probability of about a half. And what, in that case, is the probability that its cognitive faculties are generally reliable? Plantinga suggests we consider his cognitive faculties reliable if they generate true beliefs 45 percent of the time. He writes,
If I have one thousand independent beliefs, for example, the probability that three quarters or more of these beliefs are true will be less than 10–58. And even if I am running a modest epistemic establishment of only one hundred beliefs, the probability that three-quarters of them are true is very low—something like .000001
The rest of the argument follows by tautology: If I cannot trust my cognitive faculties, I cannot trust any belief they produce and especially not any metaphysical belief; but naturalism itself is a metaphysical belief produced by my cognitive faculties; therefore, I cannot trust naturalism. Plantinga concludes by saying that naturalism is self-referentially incoherent and cannot be rationally affirmed.
I think it is worth dwelling for a moment on the inescapable circularity of every possible objection to this argument: Any theory p which purports to prove the reliability of your cognitive faculties is itself a product of the cognitive faculties whose reliability it seeks to prove. Thomas Reid memorably analogised this problem by observing that, “If a man’s honesty were called into question, it would be ridiculous to refer to that man’s own word whether he be honest or not.” In a like case, Reid said, it is absurd to try and, “prove by reasoning that reason is not fallacious.”
 Naturalism is a philosophical viewpoint entailed by atheism according to which everything arises from natural properties and causes and supernatural or spiritual explanations are excluded or discounted.
 Haldane complained that if the thoughts in his mind were just the motions of atoms in his brain (a physical object that has arisen by motiveless and unguided mechanisms) why should he believe anything his brain tells him—including the idea that his brain is made of atoms? Lewis, for his part, wrote,
If all that exists is Nature, the great mindless interlocking event, if our own deepest convictions are merely the by-products of an irrational process, then clearly there is not the slightest ground for supposing that our sense of fitness and our consequent faith in uniformity tell us anything about a reality external to ourselves. Our convictions are simply a fact about us—like the colour of our hair.