The deep dive explains the argument. This track makes you able to wield it — rebuild it from memory, meet its hardest objections, and keep its conditional shape clear when challenged.
The argument from reason, developed by C.S. Lewis and sharpened by philosopher Alvin Plantinga, asks a probing question: if our minds are nothing but the product of blind physical forces, why should we trust them? On strict naturalism, every belief you hold is the end result of physics, chemistry, and natural selection — processes that have no interest in truth as such, only in survival and reproduction. But a process selecting for survival is not the same as a process aimed at producing true beliefs; a mind could be well-adapted yet systematically mistaken.
Here is the sting in the tail. If our reasoning is fully explained by non-rational causes, then the naturalist cannot, without strain, both explain reason as a byproduct of non-rational causes and rely on it to establish naturalism itself. The argument is a problem for naturalism, not for reason: it does not say reason is unreliable. The reasonable move is to reject naturalism, not reason. Our reliable rational faculties make better sense if they derive from a rational Mind than from mindless atoms.
Naturalists reply that true beliefs generally aid survival. Plantinga’s rejoinder: on naturalism a belief drives behaviour through its neurophysiology, not its meaning — and naturalism has not yet shown how content-blind selection yields truth-tracking minds. Critics such as Fitelson and Sober argue his probability claims rest on a flawed treatment of conditional probability, Plantinga replies, and the exchange continues without a knockout. Our verdict, chastened to its proper size: on this score theism explains the trustworthiness of reason better than naturalism does. Note the scope: this argument reaches a rational Ground for our reasoning — not yet the Father, Son and Holy Spirit. Identifying that being with the God revealed in Christ is the work of the other arguments, of revelation and, above all, of the historical evidence for the resurrection (see The Minimal Facts Argument).
Not more prose — the load-bearing structure, defended bone by bone, then proven from memory. Work top to bottom; the checkpoint at the end asks you to rebuild it with nothing in front of you.
Learn these first — above all ground vs cause and the defeater, and the Anscombe distinction that keeps the argument honest.
If naturalism is true, our beliefs are wholly the products of causes that are not themselves rational — physics, chemistry, and selection for survival, not truth.
Then the naturalist has a defeater for trusting his own faculties — and so for every belief they produce, naturalism included.
So naturalism is self-defeating to hold; reason is better explained as deriving from a rational Mind than from mindless causes.
The argument is a problem for naturalism, not for reason: it does not say reason is unreliable, but that the naturalist cannot, without strain, both explain reason as a byproduct of non-rational causes and rely on it. So the conclusion is to reject naturalism, not reason. Keep that straight and the it-is-self-refuting objection dissolves.
Turn the naturalist’s own act of arguing back on his account of the mind, then reveal — and note exactly how far it reaches, and how far it does not.
Honestly stated: even at its most successful this is not a direct proof of God — it is a problem for naturalism. What it points toward is that reason is more plausibly grounded in a rational source than a mindless one — a reason to prefer theism over naturalism on this score, not a demonstration of the God of Abraham; it says nothing of incarnation, resurrection, or covenant, and does not by itself rule out every non-naturalist alternative. It works best as a defeater-style consideration and one strand in a cumulative case, held with confidence rather than certainty, resting above all on the historical evidence for the resurrection.
Precision check: P1 is naturalism’s own account of the mind. Naturalism holds that reality is exhaustively physical, and that the human mind is the product of causes that are not themselves rational — the blind churning of physics and an evolutionary process that selected our ancestors for survival, not for truth. The sharpest objection to it is answered below.
This is not something the theist imposes; it is what naturalism says. Reality is exhaustively physical, and the mind is the terminal link in chains of physics, chemistry and unguided selection — a process that rewards survival-conducive behaviour, not true belief.
Natural selection rewards adaptive behavior; it does not directly “see” the content of beliefs. What selection cares about is that an organism behaves in survival-conducive ways, and — on a naturalist picture where beliefs are neurophysiological states — behavior is caused by the neurophysiological properties of those states, not by their semantic content. So the truth or falsity of the belief is, evolutionarily speaking, along for the ride.
The British statesman and philosopher Arthur James Balfour gave it an early modern formulation in The Foundations of Belief (1895), arguing that naturalism is self-undermining because it can provide no grounds for trusting the very cognitive processes it presumes to be reliable; he pressed the point against Darwinism specifically, noting that natural selection optimizes faculties for survival, not truth.
The most famous one-sentence version belongs to the geneticist J. B. S. Haldane — an atheist (and later a Marxist), which makes the admission striking. In Possible Worlds and Other Essays (1927) he wrote: “If my mental processes are determined wholly by the motions of atoms in my brain, I have no reason to suppose that my beliefs are true. They may be sound chemically, but that does not make them sound logically. And hence I have no reason for supposing my brain to be composed of atoms.”
Honest limit: Haldane later retracted the argument; but the difficulty he had named outlived his second thoughts.
Surely true beliefs help you survive. An animal that correctly perceives the cliff edge, the predator, the ripe fruit, outbreeds the one that does not. Selection, the objection runs, would therefore tend to track truth, because acting on accurate beliefs is precisely what aids survival. This is the line taken by several contributors to Naturalism Defeated? (ed. James Beilby, 2002), the major collection of critical essays on the EAAN — including responses pressing that the connection between true belief and fitness is far tighter than Plantinga allows.
Plantinga’s reply is the crux, and it is genuinely subtle. Grant that behavior is adaptive. The question is whether the content of the beliefs causing that behavior must be true. On a naturalist account, he argues, a belief causes behavior in virtue of its neurophysiological properties — its place in the physical causal web — not in virtue of what it means. The semantic content rides on top of the neurophysiology. And for any given adaptive behavior, indefinitely many belief-contents are compatible with it. So even granting that our behavior is well-tuned to the world, the probability that the contents attached to it are largely true remains low or inscrutable. Fitelson and Sober contest those probabilities and the exchange continues — but naturalism has not yet shown how content-blind selection yields truth-tracking minds.
Precision check: P2 is the defeater that follows — Lewis’s self-defeat and Plantinga’s probabilistic EAAN. Plantinga’s move is probabilistic rather than a claim of flat self-contradiction. The conclusion is not “naturalism is false” but something subtler: the naturalist who accepts evolution acquires a defeater for the reliability of his faculties — and therefore a defeater for every belief those faculties produce, naturalism included. The position is, in Plantinga’s diagnosis, not self-contradictory but self-defeating to hold. Even at its most successful, the Argument from Reason is not a direct proof of God. It is a problem for naturalism; it is not a demonstration of the God of Abraham: it says nothing of incarnation, resurrection, or covenant, and on its own it does not rule out every non-naturalist alternative.
Line 1 — Lewis: ground, not merely cause
When I conclude B because I see that A entails it, my believing B must be governed by the logical relationship between the propositions — by the ground, not merely pushed along by prior physical causes that know nothing of logic. One belief can be the cause of another (being hit on the head causes me to believe I am surrounded by green fog); but reasoning requires that one belief be the logical ground of another. Knowledge, Lewis argued, occurs only when the ground-consequent relation and the cause-effect relation coincide in the right way — when I believe a conclusion because I grasp that the premises entail it. If naturalism reduces all believing to mere event-causation, that coincidence is never secured, and the whole edifice of reasoning — including the reasoning that produced naturalism — is left without validity.
If our beliefs are simply the terminal links in chains of non-rational causation, the worry runs, then no belief is held because it is true or well-supported; each is held because the atoms did what atoms do. And that conclusion is self-undermining, because it applies to the belief “naturalism is true” as much as to any other.
Half a century later the philosopher Victor Reppert gave Lewis’s intuition a rigorous reconstruction in C. S. Lewis’s Dangerous Idea: In Defense of the Argument from Reason (2003), distinguishing several distinct “arguments from reason” — from intentionality, from truth, from mental causation, from the psychological relevance of logical laws — and defending each against contemporary objections.
Line 2 — Plantinga’s EAAN
Consider the conjunction of naturalism and unguided evolution (call it N&E), and ask: what is the probability that our cognitive faculties are reliable — that they generally produce true beliefs — given N&E? Plantinga argues this probability is low or inscrutable.
His memorable illustration is Paul, a hominid who needs to avoid tigers. Paul might run from the tiger because he believes tigers are dangerous and running away is the way to escape. But he might equally run because he believes the tiger is a large, friendly cat he wants to pet — and that the best way to pet it is to run away from it. The behavior is identical; the beliefs are wildly false. Endless such combinations of false belief plus compensating false belief yield adaptive behavior. If content is invisible to selection, the probability that the contents are largely true looks low or impossible to estimate.
Honest narrowing: this is genuinely contested. Elizabeth Anscombe’s 1948 reply — that a merely non-rational cause is not an irrational one — forced Lewis to substantially revise the chapter (1960); the revised cause/ground version is the one in view. And the EAAN is disputed: Fitelson and Sober press Plantinga’s conditional-probability treatment, and teleosemantics aims to make belief-content causally relevant after all. Plantinga replies; the exchange continues without a knockout.
On 2 February 1948 the philosopher G. E. M. (Elizabeth) Anscombe — a brilliant Catholic logician and student of Wittgenstein — read a paper to the Oxford Socratic Club. Her central objection was a careful distinction Lewis had blurred. There is a difference between an irrational cause of a belief and a merely non-rational one. Giving a full causal history of how a belief arose does not, by itself, impugn its validity. As Anscombe put it, if a man has reasons, and they are good reasons, and they are genuinely his reasons for thinking something, then his thinking is rational — whatever further causal statements we might truly make about the neurons underneath. Naturalistic causation and rational justification might operate at different levels of description, both true at once, the way a sentence can be both a sequence of ink-marks and a valid argument. Lewis’s word “valid,” she noted, was being made to do too many jobs.
This critique had teeth, and Lewis felt it. He substantially revised the chapter for the 1960 Fontana edition, changing its title from “The Self-Contradiction of the Naturalist” to “The Cardinal Difficulty of Naturalism,” and recasting his case in terms of the cause/ground distinction and the conditions under which reasoning can yield knowledge. The revision is widely regarded as an improvement, Anscombe herself among those who thought so. Her objection told against the first formulation; the revised argument, carried forward by Reppert and Plantinga, is the one in view here.
Critics such as Branden Fitelson and Elliott Sober argued that Plantinga’s probability claims rest on a flawed treatment of conditional probability. Others contend that content and neurophysiology are not as detachable as Plantinga needs — that on the right theory of mind, semantic content just is causally efficacious. Let the critic have the last objection plainly: if a fully naturalistic account of mental content can be given on which content is causally relevant and tied to the world, then the inference from “selection sees only behavior” to “content is probably false” fails, and the defeater dissolves. That is the possibility naturalistic theories of content — teleosemantics among them — aim to make good.
Plantinga has replied in turn, and the exchange continues without a knockout on either side. The theist’s reply is not to deny the difficulty of the project but to point at its cost: the more one insists that semantic content is robustly real and genuinely causal, the harder it becomes to see how a purely physical, content-blind process could have produced and sustained it — which is, in a different key, the original worry returning. The argument is not refuted, but neither is it a proof. It leaves naturalism owing an account of how a content-blind process produced minds that track truth, and that debt has not been paid.
Each of these in its strongest form first, then the reply — and the hardest objection of all is flagged below. Watch how often the conditional shape of the argument does the work.
The one that matters most: Anscombe’s 1948 reply, which made Lewis rewrite the chapter — and Fitelson and Sober on Plantinga’s probabilities. The deep dive answers it at length — take it with you, not just the replies below.
Strongest form
Elizabeth Anscombe’s 1948 reply to Lewis: there is a difference between an irrational cause of a belief and a merely non-rational one. Giving a full causal history of how a belief arose does not, by itself, impugn its validity. As Anscombe put it, if a man has reasons, and they are good reasons, and they are genuinely his reasons for thinking something, then his thinking is rational — whatever further causal statements we might truly make about the neurons underneath. Naturalistic causation and rational justification might operate at different levels of description, both true at once, the way a sentence can be both a sequence of ink-marks and a valid argument.
Reply
The critique had real teeth; Lewis substantially revised the chapter (1960) around the cause/ground distinction, a revision widely regarded as an improvement, Anscombe herself among those who thought so. The revised argument, carried forward by Reppert and Plantinga, asks how a content-blind process produced minds whose beliefs are governed by their grounds — and that debt has not been paid.
Strongest form
Accurate perception and reasoning obviously aid survival — creatures with false beliefs get eaten — so selection favours reliable cognition.
Reply
Selection acts on behaviour, and on a naturalist picture a belief causes behaviour through its neurophysiological properties, not its meaning. Plantinga’s Paul thinks the tiger is a friendly cat he wants to pet and believes the way to pet it is to run away: adaptive behaviour, false beliefs. So the probability that the contents are largely true stays low or inscrutable. Fitelson and Sober contest those probabilities and the exchange continues — but naturalism has not yet shown how content-blind selection yields truth-tracking minds.
Strongest form
You used reasoning to argue that reasoning is unreliable — so the argument defeats itself.
Reply
It does not say reason is unreliable. It says the naturalist cannot, without strain, both explain reason as a byproduct of non-rational causes and rely on it. The conclusion is to reject naturalism, precisely so we can keep trusting the reason we cannot help relying on. It targets naturalism, not rationality.
Strongest form
Mental states are brain states; a reason can be a cause. Because-it-is-true and because-of-physics describe the same event at different levels — no conflict.
Reply
But if the physical level does all the causal work, the rational level is epiphenomenal — along for the ride, doing nothing. Then you hold a conclusion because of neural causes. Truth drops out of the explanation, which is exactly the problem.
Strongest form
Machines perform valid inference physically, which shows reason needs no immaterial mind.
Reply
Computers manipulate symbols whose meaning and validity we assign — their reasoning is borrowed from rational designers. They shuffle tokens without grasping truth (Searle’s Chinese Room). Derived, observer-relative rationality does not explain original, truth-grasping rationality.
Strongest form
Our faculties could simply be reliable as a brute fact, warranting our beliefs without any theistic story.
Reply
The argument grants they might be reliable — but holds that naturalism plus evolution gives no reason to expect it and, Plantinga argues, makes the probability of reliability low or inscrutable: a defeater the naturalist must answer. Theism supplies the missing reason to expect reliability; bare naturalism has not yet shown how content-blind selection yields truth-tracking minds.
Strongest form
Appealing to God to ground reason is just filling an explanatory gap.
Reply
It is not a gap in data; it is a structural problem — it leaves naturalism owing an account of how a content-blind process produced minds that track truth, and that debt has not been paid. It is an inference to the best explanation of a feature naturalism struggles to account for in principle, not a placeholder for future science.
It would be a failure of gentleness and respect to pretend the matter is settled. It is not. The EAAN is contested, and many capable philosophers — theist and naturalist alike — reject it. Branden Fitelson and Elliott Sober argue that Plantinga’s probability claims rest on a flawed treatment of conditional probability; others contend that, on the right theory of mind, semantic content just is causally efficacious. Plantinga replies in turn, and the exchange continues without a knockout on either side.
The real fork: whether a fully naturalistic account of mental content can be given on which content is causally relevant and tied to the world (dissolving the defeater), or whether a content-blind physical process simply cannot produce truth-tracking minds. That is where the live debate sits. (A footnote worth knowing: the legend that Anscombe’s reply so crushed Lewis that he gave up apologetics for children’s fiction is largely debunked — resist a good story the facts do not bear.)
The bounded verdict: even at its most successful the argument is not a direct proof of God — it is a problem for naturalism, a defeater-style consideration and one strand in a cumulative case. What it points toward is that reason is more plausibly grounded in a rational source than a mindless one. Our verdict, chastened to its proper size: on this score theism explains the trustworthiness of reason better than naturalism does, and the argument retains its power to provoke — it asks the most committed skeptic to turn his skepticism on the instrument of skepticism itself. Held with confidence rather than certainty, resting above all on the historical evidence for the resurrection.
Close the sections above. This is the retrieval that turns reading into knowing — and it is how this argument reaches Bronze · Understand on your shelf. Opening accordions no longer counts; showing you can rebuild the structure does.
Knowledge you can't retrieve under pressure isn't yours yet. Four drills that take you from recognising this argument to defending it cold.
Scroll the sections away and generate it from memory — recalling is what builds durable memory; re-reading only feels like it does. Keep the conditional shape (a problem for naturalism, never a claim that reason is unreliable or that naturalists are irrational) and the scope honest (a rational ground of mind — not a direct proof of God). Pick your level; the bar rises as you go. Scored out of 10, accuracy first. An 8+ is one half of Silver.
Twelve cards drawn from the ground/cause distinction, the EAAN, the objections, and the honest limits above. Answer out loud before you flip — the effort of recall is the whole point; a card you merely recognise hasn't stuck. Each card is graded: type the key idea, flip, mark it. Three correct recalls and a card rests; it comes back on Today when it is due.
Start here: draft your reply, then reveal a model answer to compare. This is the worked example — Drill 4 then makes you do it cold across the objection set.
"That is the natural reply, but look at what selection can actually see. It selects behaviour — does the creature flee the tiger? — and behaviour comes from beliefs and desires working together. You could flee a tiger because you believe it is dangerous, or because you believe it is a friendly cat you want to pet, and believe the way to pet it is to run away. Same life-saving behaviour, wildly different belief. Selection rewards the running, not the accuracy, so survival-tracks-truth is not guaranteed — they can float apart. And here is the bind: on your own account of where your mind came from, is your trust in the conclusion that naturalism is true justified? Naturalism has not yet shown how content-blind selection yields truth-tracking minds. The view quietly undercuts the very reasoning that got you to it."
The other half of Silver. Draft a reply from memory, reveal the certified answer, and rate yourself honestly. Clear all four with a “nailed it” and you can field them in the wild — not just recognise them on the page.
Mastery isn't a feeling — it's three levels, each earned by showing you can do the thing, then proven again over time. Don't trust the warm glow of having read it; prove it cold.
Gold can slip back to Silver if you stop turning up for the spaced reviews — real mastery is perishable, which is why it's worth something. The dashboard brings this one back when it's due.
Rate yourself — then see whether your Explain It Back score backs it up. Felt-confidence and shown-confidence are not the same thing, and the gap is worth knowing before a real conversation.
Everything above rehearsed objections the page showed you. Here is the real test: the deepest contemporary naturalist reply — a theory of meaning designed to close the gap the argument exploits. Reason it out cold, then reveal how the argument handles it. An honest “nailed it” is the first half of Gold.
This is the real frontier, and the honest theist takes it seriously: if a fully naturalistic account of content can be given on which content is causally relevant and tied to the world, then the inference from “selection sees only behaviour” to “content is probably false” fails, and the defeater dissolves. That is exactly what teleosemantics aims to make good. The theist’s reply is not to deny the project’s difficulty but to point at its cost: the more you insist that semantic content is robustly real and genuinely causal, the harder it becomes to see how a purely physical, content-blind process could have produced and sustained it — which is the original worry returning in a different key. The argument is not refuted; but naturalism still owes an account of how a content-blind process produced minds that track truth, and that debt has not been paid.
Five more cases for God remain in this set after this one — each with the same path from understanding to defence. The Ontological Argument is next.
Apologia Daily — defending the faith with gentleness and respect.