Alvin Plantinga's formulation of the argument is here. I'll try to summarize it as I understand it.

  1. Naturalistic evolution selects for traits that tend to lead to survival.
  2. Some true beliefs about the natural world lead to survival.
  3. Some false beliefs can also lead to survival if those beliefs lead to behavior that promotes survival.
  4. The probability of a given survival-promoting belief being true is low or inscrutable.

Conclusion: If naturalism is true then our cognitive faculties are unreliable, then we can't know whether naturalism is true or false. Plantinga states this is

an argument for the irrationality of N, not for its falsehood

Example 1:

A person holds the true belief that there is water to the south. That belief drives the person south to where the water is.

Another person in the same location, however, believes that there are monsters to the north, west, and east, and is driven south towards the water by his false belief.

Yet another person believes that to the south there is an attractive mate and a beanstalk that reaches to the heavens. He's driven south towards the water by his false beliefs.

There is no reason that natural selection should favor true beliefs over false ones when the false ones lead to greater survival.

Example 2:

Person A is thirsty and believes there is water to the south (true). Person A believes thirst is a sign that his body is dehydrated and needs water (true). Once he drinks, he feels energized and believes the water quenched his thirst (true).

Person B is thirsty and believes thirst is a sign from a distant alien that the alien wants to communicate (false). Person B also believes that water is the means by which he can talk back to the alien (false). When person B gets to the water and drinks, he hears kind and encouraging voices in his head that he believes are the alien (false). After drinking, he feels energized and believes the alien gave him power (false).

I know that's a ridiculous example, but it makes the point. Why should natural selection favor Person A's true beliefs over person B's false beliefs?

I suppose you could say that if Persons A and B were to build worldviews based around their experience, Person A's worldview would more closely match up with reality. But their worldviews wouldn't necessarily be mutually exclusive. Perhaps Person C could believe that thirst is a sign of possible dehydration (true) and that it's a sign from an alien (false).

Natural selection may favor the true belief inasmuch as it affects survival, but it does not always select against the false belief.

  • Is it correct to identify their false beliefs as the cause of their survival (points 3 and 4). Surely luck and chance are the true cause of their survival since it is down to luck and chance that their false beliefs just happen to coincide with survival-promoting behaviour.
    – nwr
    Commented Jan 9, 2016 at 19:54
  • 1
    how is this different from saying the probability that Achilles will win a race against a tortus is 1/2 since there are two participants in the race? Sure, it's 1/2 if you throw out all the relevant information about the question. The fact that our species is a product of Darwinian evolution isn't necessarily relevant to the question of whether a particular belief is true. Please tell me I'm missing something. Commented Jan 10, 2016 at 21:11
  • @Timkinsella Are you saying his argument is trivial?
    – user18800
    Commented Jan 10, 2016 at 21:16
  • What do you mean by trivial? Commented Jan 10, 2016 at 21:18
  • Sorry for the snark, btw Commented Jan 10, 2016 at 21:25

6 Answers 6


I have seen the version of Plantinga's argument that includes your premises 1–3 above, doesn't adopt premise 4, and concludes that since beliefs held by organisms produced by organic evolution aren't necessarily true, naturalism isn't necessarily true. (Note that I don't think any epistemologist would say that beliefs are “generated by natural selection,” as you put it.)

A response to that argument, offered by Michael Ruse among others (e.g. in Ruse's Darwinism and Its Discontents) is that of course it's true that evolved beings are imperfect at truth-acquisition. We are epistemically limited, as philosophers say, and roughly-speaking all philosophers and biologists accept that. However, to the extent that that observation justifies doubt about any one belief like naturalism, it to that same extent can justify doubt about other human beliefs, including alternatives to naturalism. If the observation casts doubt on naturalism, it by the same token casts doubt on supernaturalism and the existence of God. In other words, premises 1–3 are clearly true, and the conclusion that we can't be certain about our beliefs follows from them, but that is unsurprising. And that argument does not help support any less-obviously-true claims Plantinga would like it to help support. Indeed, it undermines them to the same extent.

But you've asked about a version which includes the stronger Premise 4, that “There are more false beliefs that lead to survival-promoting behavior than there are true beliefs that do the same.” It's hard to imagine how Premise 4 could be true, and your example certainly does not support it. The person who goes south and finds water because of the false belief that there are monsters in other directions could just have easily have gone east because of the false belief that there are monsters in the south. Indeed, other things being equal on this simple model, that way of using beliefs to guide action would fail to lead someone to water 75% of the time! Biologically, organisms confused about their environments don't last very long.

So far, that is just to argue, quickly, that Premise 4 is dubious. But we might also want an argument for the view that natural selection generally has, when it has produced belief-forming apparatuses, favored their reliability. That is just to say that organisms with more reliable belief-forming apparatuses have survived at a higher rate than similar organisms with less reliable belief-forming apparatuses. Philosopher Christopher Stephens offers such an argument in the paper “When is it selectively advantageous to have true beliefs?”. He concludes:

The model developed here indicates that natural selection will not always favor true beliefs. However, non-reliable belief formation policies will get the organism into trouble in two basic kinds of cases: ones in which the relevant propositions are systematically important, and cases in which there are several possible actions that the organism must consider, each of which has important consequences. Although not all problems are like this, many (perhaps most), are.

His argument is developed more fully at that link.

  • 1
    The revised version of your question (substituting "no more than 50%" for "most") represents more closely the argument rebutted by Ruse, as above. I would add, about your addition, that natural selection doesn't operate on beliefs, as they aren't heritable variation. Yet as Stephens argues, natural selection favors the reliability of knowledge acquisition mechanisms under some circumstances, contra your revised P4. And we're back to the point that of course having some false beliefs is consistent with survival, but that that doesn't help Anti-naturalism. Commented Jan 9, 2016 at 21:34
  • 1
    Sorry, that's not the case; natural selection works on heritable variation. Unless you believe that beliefs are genetically determined (and even if any were, certainly the beliefs in question for Plantinga are not), beliefs are not heritable variation. Commented Jan 9, 2016 at 22:23
  • 2
    No, @IsThatTrue, it's not the case that"P(B(SNT)/SN) is left intact" by the EAAN, if by "intact" you mean "on a better epistemic footing." The EAAN only reaches its conclusion by invoking skepticism which the supernaturalist has to answer to, too. So, if it points out that the naturalist has a tough time defeating that skepticism, that's true, but the same skepticism equally damages beliefs under supernaturalism. Plantinga argues the first part (so yes, I understand that's his point), but doesn't acknowledge the second. Supernaturalism per se isn't in better shape to answer skepticism. Commented Jun 4, 2016 at 13:39
  • 2
    No, I know all that, and that Plantinga would like to leave the argument there. But notice that your interpretation makes the EAAN trivial. EAAN is only epistemically significant if it shows that under Naturalism one is on worse epistemic footing than if one doesn't assume Naturalism. And the EAAN, by implicitly restricting its ambitions in the way you describe, fails to make that case. Beyond that limited scope of the EAAN, the idea that “ A supernaturalist … does not have to answer skepticism," while a naturalist does is uh ... optimistic. Commented Jun 7, 2016 at 18:28
  • 1
    @IsThatTrue I think you may be under the misconception that evolutionary biology is an argument. Commented Apr 24, 2018 at 20:21

There are two formulations of the Evolutionary Argument Against Naturalism (EAAN). The one outlined in the cited notes (NOTES) uses the earlier, pre-2008, version where naturalism is described through the relationship between belief and behavior. The more recent formulation in Where the Conflict Really Lies (WTCFL) describes naturalism with materialism "with respect to human beings" (WTCRL, 318). Here belief or the content of cognitive processes either supervenes on or can be reduced to neurophysiological properties. See DicePower's question for further discussion of this distinction.

One problem with the EAAN is that it isn’t clear if rejecting naturalism implies that one should accept some type of theism. Clearly, “Jewish, Moslem, or Christian” theists have an explanation for why our cognitive faculties actually are reliable: “God is the premier knower and has created us human beings in his image, an important part of which involves his giving them what is needed to have knowledge, just as he does” (NOTES, 7). Is “traditional theism” the only way out?

The evolutionary theory Plantinga describes is phyletic gradualism: “natural selection, genetic drift, and other blind processes working on such sources of genetic variation as random genetic mutation” (NOTES, 2). Might an atheist supporting the punctuated equilibria evolutionary theory where species have ontological status avoid the falsification he builds for naturalism?

Does a form of panpsychism offer an atheist an out? Consider Thomas Nagel’s 2012 book, Mind and Cosmos. The subtitle of this book says it all: “Why the Materialist Neo-Darwinian Conception of Nature Is Almost Certainly False”. Nagel is a proponent of panpsychism. See his “What is it like to be a bat?” and “Panpsychism” in Mortal Questions.

Plantinga also writes (WTCRL, 316) that the problem of reliable cognitive faculties under naturalistic evolution has appeared troubling to atheists as well:

Nietzsche, Nagel, Stroud, Churchland, and Darwin, nontheists all, seem to concur: (naturalistic) evolution gives one a reason to doubt that human cognitive faculties produce for the most part true beliefs.

So, one problem with the EAAN is Plantinga doesn’t make it clear how to avoid naturalism without accepting traditional theism. Perhaps that is up to those taking non-theistic or even non-traditional theistic positions to articulate.


Plantinga is playing a semantic trick in his last step by switching R (reliability) with N (naturalism) which destroys his reliance on Bayes' theorem. He concludes that our belief forming mechanisms are unreliable and therefore our belief in naturalism is unreliable, so it is unlikely to be true under naturalism, if we assume naturalism. This is false. If we assume naturalism, naturalism is, by definition, true.

Put mathematically, P(N|N) = 1 by definitions*

Even if something is improbable under some arguments, if it is true, it is true.

If Plantinga was employing no trickery he could not arrive at any conclusion other than this.

Plantinga's semantic trick covers up that his argument is actually an excellent one for showing great evidence for evolution and naturalism. As Platinga correctly points out, we'd expect both people with false beliefs and true beliefs under naturalism and evolution, since false beliefs could still lead to survival. And this is exactly what we see: people with both true beliefs and false beliefs (e.g. people who believe in a flat earth).

If, instead, our formation of beliefs was rooted in a perfect being who guaranteed our beliefs were true, we'd expect everyone to agree on what is true, and that is exactly what we do not see.

It seems to me that Plantinga has come extremely close to the correct conclusion, but he's unfortunately fallen at the last hurdle.

There have been philosophers who have also made the observation that Plantinga has a less than firm grasp of the mathematics, Elliot Sober:

Not only is a low value for Pr( X | Y ) not sufficient for Y ’s defeating X; it also is not necessary, if defeaterhood is to ground the idea of self-defeat.The reason is that Pr( Y | Y ) = 1, for all Y .And as difficult as it is to connect low probability to defeaterhood, it seems even harder to see why the inscrutability of Pr( X | Y ) should help establish that Y defeats X

In his last section, Sober also makes similar observations concerning that Plantinga's argument, rather than being a serious problem for evolution, which expects bad belief forming mechanisms, it is rather a very serious problem for theists, who, Plantinga argues, should expect good ones.

*Here, for instance, one can read about probability mathematics. From the definitions given there, we can do a bit of math (Using arbitrary arguments A and B)

P(A|B) = P(A ∩ B)/P(B)

If we let A = B, we get

P(A|A) = P(A ∩ A)/P(A)

we also have that

P(A ∩ A) = P(A) + P(A) - P(A ∪ A)

P(A ∪ A) is the probability that A occurs OR A occurs, it can be seen from the diagram on page 2 from the link above that it would be true that P(A ∪ A) = P(A) (this is formally provable using set theory, which gives you A ∪ A = A ((and also A ∩ A = A)) but it is perhaps more illustrative to those not versed in mathematics when shown this way)

so we have P(A ∩ A) = P(A) + P(A) - P(A)

therefore P(A ∩ A) = P(A)

and therefore P(A|A) = P(A)/P(A) = 1

  • Would you have references to people who share your view? This would give the reader a place to go for more information and strengthen your answer. Welcome to this SE! Commented Nov 16, 2018 at 3:25
  • 1
    I've included a link on probability mathematics and expanded on why P(N|N) = 1. I don't know of any philosophers who support my views, they come just from seeing the fallacy in the mathematics being presented. Commented Nov 16, 2018 at 4:29
  • I've updated with reference to Elliot Sober, who presents a variety of arguments, including Plantinga's failure to grasp mathematics Commented Dec 5, 2018 at 4:49
  • 1
    The post is kind of difficult to follow because it starts with formulas, which moreover have undefined letters in them. It might help if you started by stating what you are trying to show ("Plantinga is playing a semantic trick..."), then explained it informally, and moved the formulas to the end of the post into a kind of appendix. It would also help readability if you used the standard operation symbols, like ∩ and ∪.
    – Conifold
    Commented Dec 5, 2018 at 7:48
  • 1
    Thanks for the advice. Perhaps this is a better attempt? Commented Dec 5, 2018 at 16:50

The EAAN is not even wrong.

The first problem is that beliefs are not inherited traits. Natural selection may well favour individuals whose beliefs lead them to greater reproductive success, and these beliefs may well be passed down culturally to their offspring, but the theory of evolution has nothing to do with this. Prof Plantinga clearly has no idea about the theory of evolution.

The second obvious problem is that the EEAN is that it is based on a dualistic notion that beliefs have a neurophysiological structure and associated semantic content. The example of a hominoid running away from a tiger whilst thinking he is trying to pet it shows how Prof Plantinga thinks about how the mind works. However, the modern way in which the brain is thought to work is that part of the brain controls our flight from a danger, such as tiger, with no semantic content whatsoever whilst a separate part of the brain is responsible for language and the semantic content of thinking. Creatures with no areas of the brain capable of language still have a part of the brain capable of recognizing and fleeing from danger, but will have no beliefs about this. Humans have inherited these brain structures, but have an additional part of the brain capable of forming semantic content and beliefs, but which are not inheritable. Prof Plantinga clearly has no idea of theories of how the brain works.

  • Beliefs are an inherited trait to a degree. we as a species have a an informal logic that is intrinsic to us, and was tuned evolutionarily. Applying that logic to itself, to correct some of our intuitive failings, leads to Classical logic.
    – Dcleve
    Commented Jan 7, 2023 at 20:32

If the Evolutionary Argument Against Naturalism by Alvin Plantinga provides the naturalist with a defeater for all of her beliefs, then an extension of it appears to provide God with a defeater for all of his beliefs. After bringing out this puzzle, we suggest several way sin which the proponent of the EAAN might solve it, but also show some potential weaknesses in these purported solutions. Whether the solutions to the puzzle that we consider ultimately succeed is unclear to us.

Suppose that omniscience entails R (as is no doubt true). It follows from this that R will be necessarily true of God. But everything entails necessary truths, so everything willentail R. Thus, it is impossible for there to be any defeater-deflector for God if propositions entailing R are precluded from acting as defeater-deflectors. But this must be wrong: it is wholly implausible to suppose that God is not capable of having a defeater-deflector. And hence, it appears that we cannot rule out O as a candidate for being a defeater-deflector on account of it entailing R.We can circumvent this issue by noting that it is only in a section). This is because if God is omnipotent, then he is capable of ensuring the reliability of his cognitive faculties. And, as a rational being, he would no doubt seek to do this.So, when confronted with the claim that P(R/D*) is low, God could happily grant it but hold that he has no reason to accept D* he has no reason to think that he did not intervene to ensure that R holds for himself (since he is a rational being and is omnipotent). Of course, this does not commit us to saying that God has actually intervened to ensure that R holds for him; indeed, since his omniscience entails R, we may positively say that he did not intervene to ensureR. But, again, we have set O aside here as a possible defeater-deflector, andhence, he must look for a different candidate. (In other words, if O is inadmis-sible defeater-deflectors, the rejection of D* needs to be considered independent-ly of it.) But there is a serious problem with this reply: God, since he is actually omniscient, actually accepts D*. That is, the fundamental problem with this reply is that it assumes that God does not affirm something that he does, in fact, affirm(namely, D*). So, even if this strategy were to work, God would not, and could not, use it.Another problem with this reply is that one might think that is impossible for one to ensure R for oneself. Omnipotence does not include the ability to do impossible things;e.g., we can rule out the idea that God designed his own cognitive faculties, for he would need to have them causally prior to his designing them. And, if God can ensureR holds for him, could not the naturalist make a similar suggestion?

There is an interesting consequence that arises from this section. Suppose God does stave off defeat while the demigod does not. Is there another defeater-deflector that can save the demigod? If the reasoning behindthe original EAAN is correct, it is hard to see how. Without a designing agent, natural selection is the only game in town, and he has thrust back upon the original EAAN. So, if the strategy in this section is right, it seems that to show how it is possible for God to ensure R for himself in a way that is not available to the naturalist. What is it precisely about omnipotence that allows D* to be rejected by God that does not enable a naturalist to reject D?Consider the use of epistemically circular arguments of the sort endorsed by Alston(1986). Supposing epistemically circular arguments are legitimate, if God can use them, then so can the naturalist. Omnipotence plays no essential role in the legitimacy of epistemically circular argu-ments. What is needed is a symmetry-breaker such that there is a way for God to ensureR, but not for the naturalist. Saying God used his omnipotence to ensure R is not anymore informative than saying the naturalist used her finite power to ensure R. In both cases, we would like to know how.One potential move on behalf of the EAAN proponent is to say that while we have sufficient access to the ways the naturalist could ensure R, such that we know she cannot ensure R, and so know she cannot escape defeat, we have limited knowledge of God, and he may have ways to ensure R that are unknown to us. Since God is significantly different from humans, we are not in a position to say what the conditions for warrant are for him. Similarly, the defender of the EAAN could suggest that we are not in a position to say what the conditions for defeat are for God.(Perhaps God knows things innately, in which case he does not come to know things by reasoning through inferences.) Consequently, it could be argued that we should be agnostic about whether our revised version of the EAAN is successful against God; that is, we should be agnostic about whether it provides God with a defeater for R since we do not know the conditions of defeat for God. However, it follows from the same line of reasoning that we also should be agnostic about whether any purported defeater-deflector is, for God, actually a defeater-deflector. For, if we are truly in the dark about the conditions of defeat for God’s knowledge, then by the same token, it seems that we are in the dark about the conditions for defeater-deflectors. Thus, if the medieval theory of analogy undercuts our ability to say that God’s knowledge is defeated, it undercuts our ability to say that it is not defeated, and this means that the puzzle discussed in this paper remains unsolved. Conclusion In this article, we have brought forth a puzzle about the EAAN: if it defeats the naturalist’s beliefs, then a very similar line of reasoning appears to defeat God’s beliefs.We have considered four possible solutions to this puzzle, but whether any of the multimately succeed is not an issue the authors can come to agreement on. Thus, the reader may decide for herself whether this puzzle leaves the EAAN in peril, or whether the solutions offered above are viable. Either way, this is a significant worry that all proponents (and opponents) of the EAAN must deal with. https://medium.com/@godlessvideos/does-the-evolutionary-argument-against-naturalism-defeat-gods-beliefs-bad509fcd76c

  • What's the argument for the claim "it is wholly implausible to suppose that God is not capable of having a defeater-deflector"? Is the idea that if God is omnipotent, he should have the power to make R (the reliability of his own faculties) false for himself? Couldn't R be one of God's necessary properties that he cannot change, any more than he can change other necessary truths like 1+1=2? Most theistic philosophers think omnipotence just means the power to do anything logically possible, not to change facts which are necessarily true.
    – Hypnosifl
    Commented Sep 12, 2020 at 22:37
  • If you define omnipotent as the ability to do anything that is logically possibly then he should have the power to make R false for himself. It is logically possible for you and I to make R false for ourselves but omnipotent god cannot? So you would be limiting him by his nature. Then by that definition any being who could do what it was in there nature to do would be omnipotent even if the only thing in their nature allowed for was to touch their ear. @Hypnosifl Commented Sep 12, 2020 at 23:50
  • I think a lot of theists would limit omnipotence so that it doesn't include the ability to change logically necessary truths, and also doesn't include the ability to change truths that they see as "necessary" in some broader metaphysical sense, like the existence of God or the fact of his omnipotence and omniscience. I don't know if any have offered any precise definition of this sort of "metaphysical necessity" if it isn't the same as logical necessity, but I get the impression it's implicit in some of their arguments, especially when they talk of God as a "necessary being".
    – Hypnosifl
    Commented Sep 13, 2020 at 0:37

Plantinga's EEAN relies upon applying the absolutist Classical Logic truth categories (all claims are either True of False to the products of evolution, and correctly notes that evolution cannot produce something "true".

But this is a common problem for all pragmatic processes. It is explicitly true of empiricism as well. Empiricism is based on four categories: Supported, counter-supported, currently indeterminate, and incoherent/unevaluable. All four categories violate the Law of the Excluded Middle, LEM. Three involve explicit judgment calls, and the evidence that one uses to do empiricism is both subjective in nature (observations) and is generally treated as more reliable only if it is confirmed by a consensus of experts (expert is another subjective category). Intersubjective consensus of experts relative to data, analysis, and conclusions is how all science is done (peer review).

Analytically oriented critiques of empiricism point out that consensus of experts is difficult to distinguish from the bandwagon fallacy. And for outsiders accepting a consensus of experts, the justification of accepting science is an Argument from Authority. Also, the credibility arguments for accepting science PRESUME the empirical metric of utility, and are therefore circular. And finally, even the “accept as supported” is FAR short of the “true” that science claims as its goal. Karl Popper tried to show how empiricism approaches truth as a limit case, but his effort to prove the Verisimilitude convergence of science failed.

The solution to the EEAN argument, is to agree with Plantinga that neither empiricism, nor evolution can achieve the "truth" of classical logic, but that is because they rely upon pragmatic logic, and the pragmatic truth metric of utility.

Our reasoning process is fallible, and can be deluded as to whether it has correctly identified or rejected "truth". But it is usually "good enough" that we can rely upon it in most cases. And we have found ways to improve it, via self critique (developing classical logic buy applying our intuitive logic to itself in self critique), and further methods like consensus of experts to additionally boost its utility and reliability.

So -- EAAN is based on some correct premises, but its proposed truth metric of classical logic, is the inappropriate one to apply to our word. Applying the pragmatic metric instead, leads to evolution supporting naturalism (or at least empirical science) as the two use the same pragmatic fallibilist process.

You must log in to answer this question.