Home » Library » Modern Library » Paul Draper Naturalism

Paul Draper Naturalism

In Defense of Sensible Naturalism (2007)

Paul Draper


          Varieties of Naturalism

          Plantinga’s Objective Probabilities

          The Inscrutability of P(R/N&E)

          Plantinga’s Faulty Inference

Should awareness of one’s evolutionary origins make one skeptical about one’s cognitive faculties? In other words, should it lead one to question the reliability of the psychological mechanisms that produce one’s beliefs? Darwin was worried that it should, but he did little more than raise the question. Other thinkers since Darwin have developed this idea in more detail,[1] but none with more rigor and sophistication than Alvin Plantinga. Plantinga holds that naturalists, but not theists, should share “Darwin’s doubt,” not because they believe in evolutionmany theists also believe in evolutionbut because they believe in blind evolution, in evolution that occurs without any prior supernatural planning or concurrent supernatural guidance. According to Plantinga, blind evolution is not likely to lead to reliable cognitive faculties, which means that naturalists who recognize this cannot rationally trust those faculties, and so cannot rationally believe anything at all, including naturalism itself. Plantinga’s recommendation to such naturalists would be to reject, not evolution, but naturalism, replacing it with theism.

Plantinga’s well-known argument, which he reprises in his opening case, can be boiled down to one key premise and two key inferences. Plantinga calls his key premise the “probability thesis”:

(1) P(R/N&E) is low or inscrutable.

This says that the (objective) probability that R is true, given that naturalism (N) and current evolutionary theory (E) are both true, is either low or impossible to assess. R is the proposition that our cognitive faculties are, on the whole, reliable, which Plantinga equates with the claim that the vast majority of the beliefs produced by our cognitive faculties are true. From his probability thesis, Plantinga infers (in my words) that

(2) Informed naturalists cannot rationally believe that R is true.

By “informed naturalists,” I mean naturalists who know (or at least have a true belief about) whether or not the probability thesis is true and who cannot rationally reject evolutionary theory without also rejecting naturalism. From (2) Plantinga infers that

(3) Informed naturalists cannot rationally hold any beliefs at all, including their belief in naturalism.

If this argument is sound, then (informed) naturalism leads irresistibly to its own demiseit leads irresistibly to the conclusion that belief in naturalism is irrational. This is why Plantinga says that naturalism is “self-defeating.” If he is right about this, then naturalism is in serious trouble.

Fortunately for naturalists, he is not right. I agree with Plantinga that the probability thesis (when understood in a “local” sense to be explained later) is true, but my reasons for believing that it is true are not identical to Plantinga’s. While Plantinga seems unsure whether P(R/N&E) is low or inscrutable, I am convinced that it is inscrutable. My objection to Plantinga’s argument is that (2) does not follow from (1): his preliminary conclusion that informed naturalists cannot rationally believe that R is true does not follow from his probability thesis. The precise reason that (2) does not follow from (1) is that (1) says that P(R/N&E) is low or inscrutable, instead of just saying that it is low. If Plantinga replaced his probability thesis with the more specific thesis that P(R/N&E) is low, then (2) would follow, but the argument would still fail because (1) would be false. Before I can develop this critique of Plantinga’s argument, however, some preliminary remarks are needed both about the relationship of naturalism to materialism and about the relationship of beliefs to behavior.

Varieties of Naturalism

Plantinga defines naturalism as the thesis that there are no supernatural beings. I have no objections to this definition. But then Plantinga claims that all or almost all naturalists are materialists. Of course, even if he were right about that, it still wouldn’t be legitimate to equate, as he sometimes does, naturalism with materialism. But he’s not right. An increasingly large number of naturalists reject materialism, if by “materialism” is meant the view that a complete description of what the natural world and human beings in particular are like can be given in terms of their “third-person” properties. Materialism defined in this way conflicts with the fact that human beings and some other animals have properties that are irreducibly “first person,” properties like “being in pain.” When I am in pain, I know about this in a way that no one else can know about it. I know about it from the “inside” (in a sense of “inside” that has nothing to do with any spatial relationship to my skull). Others who know about my pain cannot know about it in the way I do, by feeling it. In this sense my pain is “private.” Others who know about my pain do so by observing my behavior or by detecting certain neuronal activity in my brain or by my testimony or by some other “public” means. This epistemological distinction points directly to what is so special about consciousness, to the fact that there is something it is like (from the “inside”) to be in pain or to see colors or to feel anger or even to think about philosophy. In other words, human beings, in addition to their third-person physiological characteristics also have first-person characteristics.

Materialists claim that human beings and other conscious animals either don’t really possess first-person properties or else they do possess them, but to possess them just is to possess certain third-person properties. Naturalists need not (and in my opinion should not) make either of these claims. Naturalists will hold that the world’s having first-person features depends causally on its having certain third-person features. To put the point in a popular though somewhat misleading way, matter existed long before mind, which didn’t arise until matter was organized in the right sort of way (e.g., in the form of complex nervous systems). There is no reason, however, why naturalists cannot hold that conscious states are real, irreducibly first person, and yet also fully natural biological states of brains. Epiphenomenalists claim that, although the world has irreducibly first-person features, these features lack causal efficacy. Once again, naturalists need not (and should not) agree. Naturalists hold that nature is causally closed. But there is no reason why naturalists must hold that first-person events are not a part of nature and thus no reason why naturalists must hold that the causal structure of the natural world includes only third-person events. Am I suggesting that naturalists should be “dualists”? Perhaps, but it is not a dualism of supernatural souls and natural bodies, nor is it a dualism of the physical and the mental, as if these were mutually exclusive categories. It is a dualism only because it admits, quite sensibly I think, that the natural world has both first-person and third-person features and that no complete description of nature is possible without mentioning both sorts of features.[2]

What does this have to do with beliefs? The existence of beliefs, even nonconscious ones, depends on the existence of first-person states. Our beliefs have content, they are about propositions, and this is possible only because we have first-person features.[3] With this as background, I am now prepared to distinguish what I will call extreme naturalism from what I will call (no doubt to the great annoyance of extreme naturalists) sensible naturalism. “Extreme naturalism” is the conjunction of naturalism with the thesis that

X: Either (i) beliefs don’t exist (eliminativism) or (ii) they exist but they don’t affect behavior at all (epiphenomenalism) or (iii) they exist and they affect behavior but not by virtue of their content (semantic epiphenomenalism) or (iv) they exist and they affect behavior by virtue of their content, but to have a certain content just is to display a certain set of third-person properties (reductive materialism).

Notice that X is compatible with a variety of distinct materialist views about beliefs and also with one nonmaterialist view, namely, “epiphenomenalism,” which does not deny the irreducibly first-person character of conscious beliefs, but does deny their causal efficacy. Sensible naturalism is just naturalism conjoined with the denial of X. In other words, it conjoins naturalism with

S: Beliefs exist, they affect behavior by virtue of their contents, and a belief’s having a particular content is not the same as its displaying a certain set of third-person properties.

It is my view that sensible naturalists are not as vulnerable to the sort of reasoning Plantinga employs as other sorts of naturalists (though for reasons I won’t explain in this essay reductive materialists may be able to withstand Plantinga’s assault as well).

Plantinga’s Objective Probabilities

To see what is wrong with Plantinga’s argument, it is crucial to understand his probability thesis. Recall that the probability thesis asserts that the objective probability of our cognitive faculties being reliable, given naturalism and current evolutionary theory, is low or inscrutable. There is a problem of interpretation concerning the conditional probability[4] in this thesis because, according to Plantinga, R has a high degree of initial warrant and probability, both for theists and for evolutionary naturalists. We all start out with strong noninferential grounds for believing that R is true and thus R is, to use Plantinga’s terminology, “properly basic.” The point of his argument is to show that R becomes improperly basic for evolutionary naturalists once they recognize that the probability thesis is true. All of this is compatible with his view that P(R/N&E) is low or inscrutable as long as it is understood that this probability thesis is not a claim about how likely it is for the naturalist that evolution produced reliable cognitive faculties, given everything we know now. Instead, we are supposed to abstract from the fact that naturalists know how things turned outthey know that whatever produced our cognitive faculties managed to make them reliable. We are supposed to ask instead how likely it was “beforehand” that things would turn out the way we know they did.

I suspect this is why Plantinga thinks he can support the probability thesis with an analogy to an imaginary population of mysterious nonhuman beings, mysterious not just because they inhabit another planet but also because we know very little about them other than that they, like us, have beliefs and belief-producing mechanisms. When we consider this population of alien beings, it forces us to make the right abstractions from what know about ourselves, which helps us to see that the reliability of our cognitive faculties, which we ordinarily take for granted, is actually quite surprising on the assumption that the source of those faculties is naturalistic or “blind” evolution. Of course, this is not to say that in assessing P(R/N&E) we are supposed to abstract from all of our background knowledge, even knowledge of a very general sort. If we did that, then Plantinga couldn’t even get as far as he does in his deliberations about P(R/N&E). But we are supposed to abstract from any specific evidence for R, whether that evidence be inferential or noninferential.[5]

Given this “local” interpretation of the objective probability in Plantinga’s probability thesis, I believe that the thesis is true. As I mentioned earlier, however, my reasons for believing this are not identical to Plantinga’s. Unlike Plantinga, I am confident that P(R/N&E) is inscrutable (and if it is inscrutable then of course it is either low or inscrutable). This is important, because I believe that step (2) of the argument does not follow from Plantinga’s probability thesis as it stands, but would follow from it if it were restated to claim that P(R/N&E) is low. Thus, to show that Plantinga’s attack on naturalism fails, I need to show first that P(R/N&E) is not low but rather inscrutable and second that the informed naturalist can rationally believe that R is true in spite of the inscrutability of P(R/N&E).

The Inscrutability of P(R/N&E)

To calculate P(R/N&E), we will need to use a theorem of mathematical probability. Since S just is the denial of X, we can use the same theorem that Plantinga uses when he calculates P(R/N&E) to derive the following equation:

P(R/N&E) = P(R/S&N&E) x P(S/N&E) + P(R/X&N&E) x P(X/N&E).

Let’s begin with P(S/N&E) and P(X/N&E), keeping in mind our local interpretation of Plantinga’s objective probabilities. Without knowing how things have in fact turned out, it would seem quite impossible to predict or even guess whether naturalistic evolution would lead to the truth of S or to the truth of X. Such objective probabilities are clearly inscrutable. After all, as David Hume warned, we can’t simply examine the nature of physical entities and draw justified conclusions or even justified guesses about what they can or cannot produce. Plantinga seems sympathetic to this position, but he thinks one can also sensibly believe that P(X/N&E) is high. I disagree with this partly because, as I explained earlier, I see no reason to believe that naturalism, which after all just denies the existence of supernatural entities, makes X (the disjunction of materialism and epiphenomenalism) likely. Granted, naturalists do not yet know exactly why irreducibly first-person features arise in nature, but that doesn’t commit them to denying either their existence or their causal efficacy!

Of course, from the fact that P(S/N&E) and P(X/N&E) are inscrutable, it doesn’t follow that P(R/N&E) is inscrutable. According to our equation, it could still be low if both P(R/S&N&E) and P(R/X&N&E) were low. Moreover, I am willing to grant, at least for the sake of argument, that P(R/X&N&E) is low. P(R/S&N&E), however, is not low; it is high. To see why, notice first that, if S is true, then it is not difficult to see how blind evolution could produce reliable cognitive faculties. As Plantinga himself has said,

Now if content of belief did enter the causal chain that leads to behaviorand if true belief caused adaptive behavior (and false belief maladaptive behavior)then natural selection, by rewarding and punishing adaptive and maladaptive behavior respectively, could shape the mechanisms that produce belief in the direction of greater reliability. There could then be selection pressure for true belief and for reliable belief-producing mechanisms.[6]

The difference between me and Plantinga, however, is that while I take this scenario to be highly likely given sensible evolutionary naturalism, Plantinga regards it as merely possible. He says:

That our species has survived and evolved at most guarantees that our behavior is adaptive; it does not guarantee or even suggest that our belief-producing processes are reliable, or that our beliefs are for the most part true. That is because our behavior could be adaptive, but our beliefs mainly false. [my italics]

Plantinga makes a mistake here when he adds the words “or even suggest” to his conclusion. His premise is that our behavior could be adaptive in spite of our beliefs being mainly false. In other words, this scenario is possible. While this premise is true and while it does follow from this premise that our survival and evolution do not guarantee that R is true, it does not follow that they do not even suggest that R is true. In fact, when we assume that sensible naturalism is true, they not only suggest this, they provide strong evidence for it.

To see why, consider the following example. Suppose that when I go to take a bath there is an alligator in my tub. It is certainly possible that I survive these unfortunate circumstances without having true beliefs like “there’s an alligator in my bathtub” and “alligators are dangerous animals.” For example, the beliefs that “there’s a beautiful mermaid in my bathtub” and “mermaids, especially beautiful ones, are dangerous animals” may do just as well (depending on how much I’m willing to risk in order to bathe with a beautiful mermaid). Notice, however, that the vast majority of false beliefs I might have in these circumstances (e.g., there’s nothing in my bathtub, there’s a gentle alligator in my bathtub, there’s a rubber ducky in my bathtub, there’s a dangerous alligator in my bathtub but I can easily overpower it, etc.) will not do just as well, but will lead instead to a, shall we say, “maladaptive” bathing experience. So my survival in these circumstances is much more to be expected if my beliefs about the contents of my bathtub are mostly true than if they are mostly false. More generally, the long term survival of our species is much more to be expected if our cognitive faculties are reliable than if they are unreliable, and that entails that the long term survival of our species is strong evidence for R.[7]

In addition, it is very unlikely that belief-producing mechanisms that do not track the truth would systematically promote survival in a very diverse and often rapidly changing environment. This is why, when Plantinga tries to give examples of how our cognitive faculties could systematically deceive us and still be adaptive, the best he can do is to take essentially reliable faculties, faculties that do track the truth in a diverse and changing environment, and then make them unreliable by tacking on some false “extra.” For example, Plantinga suggests that there might be a primitive tribe, the members of which refer to everything as a “witch,” so that in circumstances in which we would have the true belief that this is an alligator, they would have the false belief that this witch is an alligator. Their beliefs could, however, be just as adaptive as ours, assuming that their behavior is not affected in some maladaptive way by the fact that they believe everything is a witch.

One problem with this example, however, is that while the cognitive faculties of these beings would not be as reliable as ours, they would still be very reliable. For anyone who believes “this witch is a dangerous alligator” would also believe, at least implicitly, “this is dangerous,” “this is an alligator,” “this is an animal,” and so on.[8] Indeed, even if these beings could not express these implicit beliefs in their language, they would still have the crucial concepts and so still have those beliefs.[9] This is not to say that a philosopher as resourceful as Plantinga could not dream up some complicated or bizarre way in which beliefs, including implicit ones, could vary from the truth and yet still allow for survival; but then one has to ask how likely it is that such complicated or bizarre variations would be available for nature to select.[10]

I conclude that P(R/S&N&E) is high. The most that Plantinga can establish is the possibility of cognitive faculties that are both unreliable and adaptive. This does nothing to refute the fact that by far the most likely way for blind evolution to produce adaptive cognitive faculties is to make them reliable. At times Plantinga seems willing to concede this or something close to it. Such a concession is, it seems to me, highly warranted. Therefore, since P(R/S&N&E) is high, since P(S/N&E) and P(X/N&E) are both inscrutable, and since P(R/X&N&E) is (I’m willing to grant) low, it follows from the equation stated at the beginning of this section that P(R/N&E) is inscrutable.[11]

Plantinga’s Faulty Inference

Plantinga might very well grant that P(R/N&E) is inscrutable since he thinks that this is enough to justify an inference to step (2) of the argument, which states that an informed naturalist cannot rationally believe that R is true. This inference, however, is incorrect. In making it, Plantinga seems to forget that informed naturalists start their inquiry into the origins of their cognitive faculties with a rational belief in R. Thus, the purpose of the inquiry is not to move from a state of skepticism about R to forming a belief about its truth or falsity. Rather, the purpose is to determine whether information about the origins of their cognitive faculties will (i) justify their current level of confidence in R, (ii) justify increased or decreased confidence, or (iii) undermine the rationality of their belief in R entirely. Sensible naturalists will be pleased to discover that P(R/S&N&E) is high, thus confirming their initial confidence in R. The fact that deleting S from S&N&E makes it impossible to assess R’s probability should not worry sensible naturalists at all. So long as N&E does not confer a low probability on R, no lowering of their confidence in R is called for.

The two analogies Plantinga uses to suggest otherwise are both seductive but ultimately flawed because they do not make it clear whether or not we begin with a rational belief in the analogue of R. Consider Plantinga’s thermometer example. If, on the one hand, I have no initial reason to trust my thermometer, then discovering that it was produced in factory F and that I cannot assess the objective probability of its being reliable given that it was produced in factory F leaves me in the same position in which I startedI still have no reason to trust the thermometer and so cannot rationally believe that it is reliable. But interpreted this way, the thermometer case fails because it is disanalogous to the case of the informed naturalist, who starts out with a rational belief in R. If, on the other hand, I start out with a rational belief in the reliability of the thermometer, where what makes the belief rational doesn’t depend on any false belief about its origins (e.g., that its origins are relevantly similar to other thermometers most of which are reliable), then discovering completely unhelpful information about its origins won’t render that belief irrational. So whichever way we interpret the thermometer case, it fails to support Plantinga’s inference from the inscrutability of P(R/N&E) to the conclusion that informed naturalists cannot rationally believe that their cognitive faculties are reliable.

One might object here that Plantinga should simply retract his claim that we are initially rational in believing R. But he can’t do that! For then everyone, theist and atheist alike, would be trapped in a skeptical bog, unable to provide good reasons for trusting their cognitive faculties because any reasons they offer will require the use of their cognitive faculties and so will be viciously circular.

If Plantinga’s analogies don’t work, do I have a better one? Try this. Consider evolutionary theists, theists who accept the major claims of contemporary evolutionary biology. Like everyone else, they begin their inquiry into R with a rational belief in its truth. Further, just as sensible naturalists, for example, would be pleased to discover that P(R/S&N&E) is high, so too evolutionary theists, who believe God worked through evolutionary processes to produce our cognitive faculties, will be pleased to discover that P(R/T&E) is high. (“T” stands, of course, for theism.) But now suppose we remove T from this probability and consider only P(R/E). On E alone, it will be impossible to assess the likelihood of R. To make this point clear, consider the following equation, which follows from the same theorem of mathematical probability we used previously:

P(R/E) = P(R/N&E) x P(N/E) + P(R/~N&E) x P(~N/E).

I have already argued that P(R/N&E) is inscrutable. This implies that P(R/E) is inscrutable since P(N/E) is clearly inscrutable. Should this worry the evolutionary theist? No, of course not. Similarly, naturalists shouldn’t worry about P(R/N&E)’s being inscrutable. Subtract enough from the propositions about origins upon which one conditionalizes and inscrutability will inevitably result. This shouldn’t worry either evolutionary theists or naturalists because they both start out with a rational belief in R.[12] Thus, the failure of Plantinga’s argument turns out to be a good thing not just for naturalists but also for evolutionary theists! I should add that whatever is good for evolutionary theists is good for all theists, since not even theists, if they are well-informed, can rationally reject the theory of evolution![13]

Continue the Debate


[1] See, for example: C. S. Lewis, Miracles: A Preliminary Study (London: Geoffrey Bles, 1947, revised 1960); and Richard Taylor, Metaphysics (Englewood Cliffs, NJ: Prentice-Hall, 1963).

[2] Many of my views here echo John Searle’s “biological naturalism,” as explained in, for example, in The Rediscovery of the Mind (Cambridge, MA: MIT Press, 1992). Searle denies he is a property dualist, but he does believe that the world has first-person features and that these features cannot be reduced ontologically to third-person features.

[3] For two distinct defenses of the claim that intentionality depends on subjectivity, see Searle, Ch. 7, and Colin McGinn, “Consciousness and Content,” Proceedings of the British Academy 76 (1988), 219-239.

[4] A “conditional probability” is a statement’s probability “conditional on” or “given that” or “on the assumption that” some other statement is true.

[5] Cf. section II of William Alston’s essay, “Plantinga, Naturalism, and Defeat,” in Naturalism Defeated? Essays on Plantinga’s Evolutionary Argument Against Naturalism, ed. James Beilby (Ithaca, NY: Cornell University Press, 2002), pp. 176-203.

[6] “Reply to Beilby’s Cohorts,” in Beilby, p. 257.

[7] The fact that our cognitive faculties did not evolve simultaneously and yet produce beliefs, the vast majority of which are consistent and a significant portion of which overlap, further enhances P(R/N&S&E).

[8] Jerry Fodor, “Is Science Biologically Possible?” in Beilby, pp. 30-42. Fodor makes this point on p. 34.

[9] Further, even ignoring these implicit beliefs, the cognitive faculties of these beings would still be reliable, not in Plantinga’s sense, but in the sense that they produce beliefs the vast majority of which at least approximate the truth.

[10] Cf. section 1.3 of Branden Fitelson and Elliott Sober’s 1997 essay, “Plantinga’s Probability Arguments Against Evolutionary Naturalism,” <http://philosophy.wisc.edu/sober/fitelsoon%20and%20sober%20on%20 plantinga.pdf>

[11] Plantinga might respond that he could replace R with the statement, call it R*, that the cognitive faculties responsible for scientific beliefs like E and metaphysical beliefs like N are reliable. He could then base his argument on the premise that P(R*/N&E) is low. Of course, this is not the argument he gives in his opening case and so I don’t feel obligated to refute it here. Very briefly, however, my reply would be that the reasoning involved in science is not unique to science. More generally, the psychological mechanisms that produce scientific beliefs like E are in fact the same ones that produce the everyday beliefs that are so important to our survival. Further, as my opening case suggests and as I have argued elsewhere (“God, Science, and Naturalism,” in The Oxford Handbook of Philosophy of Religion, ed. William J. Wainwright [Oxford: Oxford University Press, 2005], pp. 272-303), a case for N can be made that is “scientific” in a broad sense.

[12] Cf. Michael Bergmann, “Commonsense Naturalism,” in Beilby, pp. 61-90.

[13] I am grateful to Sean Allen-Hermanson and John Schellenberg for helpful comments on a preliminary draft of this essay.

Copyright ©2007 Paul Draper. The electronic version is copyright ©2007 by Internet Infidels, Inc. with the written permission of Paul Draper. All rights reserved.

all rights reserved