The Argument That Convicts Itself

There is an argument that apologists reach for when the conversation turns to evolution, and it goes something like this: You say your brain is the product of natural selection. But natural selection does not select for true beliefs — it selects for beliefs that keep you alive. So on your own account you have no reason to trust anything your brain produces, including your belief in evolution itself. Your position is self-defeating.

In its casual form this is a debating trick, and it is easily answered. But it also has a serious form, and the serious form deserves a serious answer. That version belongs to the philosopher Alvin Plantinga, who called it the Evolutionary Argument Against Naturalism, and it has been in circulation since Warrant and Proper Function in 1993.

I want to take the argument seriously, grant it more than its defenders usually manage to establish, and then show where it leads. It does not lead where they think.

The Argument in Its Strongest Form

Plantinga is careful, and the careless version misrepresents him. He does not claim that evolution gives us no true beliefs at all. He knows perfectly well that a creature that consistently mistook cliffs for meadows would not last long.

His claim is narrower and sharper. Natural selection acts on behavior, not on belief. Belief only matters to selection insofar as it produces action. And a given useful action can be produced by an enormous number of belief-desire combinations, only some of which involve true beliefs. Plantinga's famous example is a hominid named Paul who runs away from tigers — not because he believes tigers are dangerous, but because he believes the tiger is a large friendly cat that he wants to pet, and that the best way to pet it is to run away. Absurd, but it produces exactly the right behavior. Selection sees the running; it never sees the reason.

If most behavior can be produced by wildly false beliefs paired with correspondingly odd desires, then the probability that a mechanism selected for behavior alone also happens to be truth-tracking is low, or at best inscrutable. And if you cannot assign a high probability to your faculties being reliable, then you have a defeater for every belief they produce — including the belief in evolution.

Plantinga then adds the part that matters to him: if God designed our faculties to know truth, this problem does not arise. So theism is epistemically stable and naturalism is not.

That is the argument. It is more interesting than the version you meet on the internet, and it fails anyway.

The Hidden Assumption: Truth or Survival, Pick One

Before getting to the substance, notice the shape of the thing. The argument only works if reliability is all-or-nothing — if a brain is either a truth-detector or a survival device, and cannot be partly one and partly the other. Show that selection optimizes for fitness rather than truth, and the whole apparatus is condemned in a single stroke; every belief it produces falls together.

That is a false dichotomy, and it is the load-bearing error. There is no reason a cognitive system must be uniform in its accuracy. The live possibilities are not two but many: a brain that tracks truth reliably in some domains and not in others; one that tracks it well under some conditions and badly under others; one whose accuracy varies with how quickly and how expensively being wrong is punished; one that is excellent at estimating the trajectory of a thrown rock and hopeless at estimating the trustworthiness of a stranger who shares its accent. Mixtures, gradients, and combinations — all of them are consistent with everything Plantinga establishes, and none of them yields a global defeater.

And a mixture is exactly what we should expect, because selection is not a single pressure applied uniformly to a general-purpose truth organ. It is a large number of separate pressures acting on a large number of separate subsystems, each with its own history, its own cost structure, and its own error tolerance. Where accuracy is cheap and error is fatal, you get accuracy. Where accuracy is expensive and error is survivable — or where error is actively rewarded — you get something else. A patchwork is the predicted outcome, not an awkward concession.

This matters because a global defeater and a local one lead to opposite conclusions. If the argument shows that every belief is equally suspect, then nothing can be sorted from anything and the position is indeed self-defeating — but it takes the apologist's beliefs down with it, since he is reasoning with the same equipment. If instead it shows that reliability varies by domain in ways we can identify, then it is not a defeater at all. It is a map: a way of knowing which of your beliefs to distrust, and how much, and why. The apologist needs the first reading. He is only entitled to the second. And the second is a research programme, not a refutation.

So the honest question is not is the brain reliable? but where is it reliable, where is it not, and what predicts the difference? That question has answers. They come in two parts.

Where the Correlation Holds: The Cost of Being Wrong

The Paul-and-the-tiger example is doing something dishonest, and it is worth being precise about what.

A single false belief can be constructed to produce a single correct action. Fine. But an organism does not perform one action. It performs an unbounded number of them, in novel situations, across an entire life. And Paul's belief system has to generate the right behavior every time. When the tiger is asleep, Paul's theory says approach and pet it — the running was only a means. When Paul is hungry and the tiger is between him and the carcass, his theory gives no useful guidance at all. To keep Paul alive, you must keep bolting on new false beliefs, each one tailored to a new circumstance, each one coincidentally producing the behavior a true belief would have produced directly.

There is a name for a model that requires a fresh epicycle for every new observation, and it is not "equally probable." A false belief system that mimics a true one across all circumstances is a false belief system that has smuggled the true one inside it. The cheap way to get behavior right across an open-ended range of situations is to track the world. That is what perception is for, and it is why perception is expensive: eyes, ears, and the cortical machinery behind them cost a great deal of energy to build and maintain. Evolution does not pay that bill for a system that is indifferent to truth.

So the correlation is real. Not identity — and the difference will matter shortly — but real, and strong, and strongest precisely where being wrong gets you killed quickly.

Where the Correlation Breaks Down

Here is the part where I agree with the apologist, and agree emphatically. The correlation between survival and truth is not perfect, and where it fails, it fails in ways that are entirely predictable.

Selection does not care about truth. It cares about reproduction. Wherever a false belief pays better than a true one, selection will install the false belief without a moment's hesitation. And there are two large domains where false beliefs pay very well indeed.

The social domain. Humans are obligate social animals. For most of our history, expulsion from the group was a death sentence — not metaphorically, but actually. Under those conditions, holding the beliefs your group holds is enormously adaptive, and holding them sincerely is more adaptive still, because sincere believers are harder to detect as defectors. A brain built to arrive at group consensus rather than at truth will outperform a brain built to arrive at truth in any environment where the group punishes dissent. This is not speculation; it is the best available explanation for the strength and stubbornness of conformity effects, in-group favoritism, and motivated reasoning. We have brains that resist changing their minds in proportion to how much the belief marks their membership.

The psychological domain. Beliefs that keep a person functional in the face of grief, terror, and pointless suffering also pay. An organism paralyzed by accurate contemplation of its own mortality is an organism that does not forage. Optimism bias, the sense of personal exceptionalism, the conviction that the dead persist somewhere — these are load-bearing. They keep people upright. Their truth-value is beside the point.

So: yes. Our brains are not reliable truth-detectors. They are reliable survival devices, which makes them very good at truth in some domains and systematically, predictably bad at it in others.

The Two Examples

Consider the group-cohesion case. In 1633 the Roman Inquisition compelled Galileo to kneel and abjure the proposition that the Earth moves. Note that the proposition was, by then, well supported: the phases of Venus, visible through any decent telescope, were flatly incompatible with the Ptolemaic system. The Church's astronomers could look. Some did. What was at stake was not the astronomy but the authority — whether a layman with an instrument could overturn a reading endorsed by the institution. Galileo's crime was disloyalty, and the false belief was maintained because maintaining it was what membership required. It took until 1992 for the Vatican to formally admit the error, three and a half centuries in which the truth was available to anyone willing to look through a tube.

Now the psychological case. Belief in a surviving soul is close to a human universal, and it is not hard to see why. It answers the intolerable: that the people we love are simply gone, that our own extinction is total, that the child in the grave has not merely moved to another room. The belief does real work. It gets the bereaved through the day. It also has no evidence behind it whatsoever, and everything we have learned about the dependence of mind on brain — that damage to tissue reliably destroys memory, personality, judgment, the capacity for love — points the other way. The belief is held not because it is supported but because the alternative is unbearable. That is a perfectly good evolutionary explanation of a belief. It is not a reason to think it true.

The Cost of the Gap

Because the correlation is a correlation and not an identity, the human journey toward accurate belief has been long, slow, and expensive.

It was paid for in people. Giordano Bruno was burned alive in the Campo de' Fiori in 1600. Michael Servetus, who described pulmonary circulation, was burned at Geneva in 1553 with his book chained to his leg. Between roughly 1450 and 1750, tens of thousands of people were executed in Europe for witchcraft — a crime that does not exist, prosecuted with elaborate juridical seriousness, on confessions obtained by torture. Ignaz Semmelweis demonstrated in the 1840s that handwashing cut maternal mortality dramatically, was driven out of the profession for saying so, and died in an asylum while women continued to die of an infection he had shown how to prevent.

None of these were failures of evidence. In each case the evidence was available and the belief was resisted for social reasons — because accepting it meant conceding authority, or admitting fault, or breaking with one's colleagues. Our brains defended the group and let the facts wait. Sometimes for centuries.

The Argument Turns Around

Now put the two halves together.

The apologist's argument concedes — insists, in fact — that human cognition is unreliable wherever truth and fitness come apart. Very well. Where do they come apart? In exactly the two domains just described: beliefs that bind a group together, and beliefs that make existence tolerable.

And what is religion?

It is the paradigm case of both. It is the belief system whose social function is to mark membership and enforce cohesion, and whose psychological function is to promise that death is not the end and that suffering means something. Every feature that makes religion successful is a feature that the evolutionary account predicts for beliefs that persist without regard to their truth: costly signals of commitment, sanctions against apostasy, doctrines held more firmly the less evidence supports them, in-group boundaries drawn along doctrinal lines.

The apologist has picked up a tool that identifies the region of human cognition where beliefs are least likely to be true, and has handed it to us pointed directly at his own position. His argument, run correctly, is not an argument against naturalism. It is the best available explanation of why billions of people believe things for which there is no evidence, and why they believe them with such conviction, and why the conviction is so resistant to argument.

What We Actually Do About It

The last thing to say is the most important, because it is what the apologist's argument entirely misses.

Nobody claiming that our faculties are evolved is thereby claiming that our faculties are trustworthy. That is the whole point. We know our brains are unreliable. We have catalogued the failures in exhausting detail — confirmation bias, motivated reasoning, in-group favoritism, hindsight bias, the availability heuristic, and the rest. The response has not been to trust the brain anyway. It has been to build machinery that does not depend on trusting it.

That machinery is science, and every component of it is an answer to a specific known defect. Controlled experiments exist because we see patterns in noise. Blinding exists because we see what we expect to see. Randomization exists because we select unrepresentatively without noticing. Peer review and replication exist because individual judgment is corruptible and no one can audit their own reasoning. Publishing methods rather than conclusions exists so that a stranger who does not share your loyalties can check your work. Mathematics and formal logic exist because our intuitions about quantity, probability, and inference are demonstrably poor.

The whole apparatus is a prosthesis for a brain we know we cannot trust. It works — not perfectly, but demonstrably, cumulatively, and in a way that no other method of forming beliefs has ever managed. We can predict eclipses to the second and we have eradicated smallpox, and neither achievement came from an individual brain being reliable. They came from a system built by unreliable brains that took their own unreliability as the engineering problem to be solved.

Set that against the alternative on offer. The theist's answer to cognitive unreliability is that God designed our faculties to know truth. But this fixes nothing. Which faculties? Reliable about what? Human beings, using their divinely designed faculties, have believed in an Earth-centered cosmos, in demonic possession as the cause of epilepsy, in witches, in the divine right of kings, in the curse of Ham. They have believed all of this sincerely and in enormous numbers, and they have believed it from within religious traditions that claimed direct access to truth. If the design guarantee were doing any work at all, this record would look different. And there is no independent check available — you would have to use the very faculties in question to establish that they were divinely designed, which is precisely the circularity Plantinga accuses the naturalist of.

The naturalist, by contrast, has an explanation of why the faculties fail, a prediction of where they will fail, and a working method for correcting them. That is not a self-defeating position. It is the only position anyone has ever held that takes the problem seriously.

Conclusion

The apologist is right that evolved brains are not built for truth. He is right that we cannot simply trust our own reasoning. He is right that where fitness and accuracy diverge, our brains follow fitness.

What he needed, and never had, was the extra premise that fitness and accuracy diverge everywhere — that a brain shaped by selection must be uniformly unreliable rather than reliable in some places and not others. Without that premise there is no global defeater, only a patchwork to be surveyed. And what he has not noticed is that he has described his own beliefs with considerable precision, and that the people he is arguing against worked all of this out some time ago and built an institution to compensate for it. The denial of evolution is itself a fine illustration: a belief held against overwhelming evidence, in tight correlation with group membership, defended with a fervor that tracks identity rather than data. It is the argument's own thesis, demonstrated live.

We do not claim our brains are reliable. We claim we have found a way to make progress despite their not being reliable. The name of that way is science, and its record is the only reason anyone is in a position to have this conversation at all.