Natural Selection and the Problem of Evil (2007)
By the “problem of evil,” philosophers mean, roughly, the question of whether the suffering, immorality, ignorance, and other evils in our world are strong evidence against theism. My opening case in this debate will address this question and also the question of what relevance, if any, natural selection has to the problem. Darwin himself suggests two radically different answers to this second question. On the one hand, in a passage sympathetic to theism that he wrote in 1842, he points out that, according to his theory, “death, famine, rapine, and the concealed war of nature,” led directly to “the highest good, which we can conceive, the creation of the higher animals.” On the other hand, in a letter about religion that he wrote in 1879, he says that the argument from suffering against theism “is a strong one; whereas . . . the presence of much suffering agrees well with the view that all organic beings have been developed through variation and natural selection.” Darwin’s earlier remark is unhelpful to the theist. A good theistic explanation of natural evils like death and famine must do more than show that such evils lead to a greater good like the existence of higher animals. It must show that a creator, despite being all-powerful, could not obtain that good without those evils. His later remark is more interesting, especially if it is intended to suggest that his theory provides a “naturalistic” explanation of various facts about suffering that is superior to the best theistic explanations.
Still, that is just a suggestion and not an argument. What is needed, in order to show that evolutionary biology makes a theistic solution to the problem of evil more difficult, is a serious argument from evil against theism in which evolutionary theory plays a significant role. My goal here is to construct just such an argument. A key component of my strategy for accomplishing this goal will be to compare theism to a specific alternative hypothesis, which I will call “naturalism.” I should mention that in referring to theism and naturalism as “hypotheses,” I mean nothing more than that they are statements that are neither certainly true nor certainly false. I will begin by explaining these two hypotheses. Then I will show that naturalism has both smaller scope and greater simplicity than theism and for that reason is more plausible than theism. Next I will use certain known facts about good and evil in order to test the two hypotheses. Specifically, I will show that the “predictive power” of naturalism with respect to these facts is much greater than that of theism. (It is here that evolutionary biology will play an important supporting role in the argument.) I will close by explaining why my argument provides a very strong reason to reject theism.
1. Theism and Naturalism
Successfully comparing hypotheses requires clarity about the hypotheses being compared. So let’s begin with some definitions. First, by “theism” I mean not just the vague claim that “God exists,” but the more specific hypothesis that an omnipotent, omniscient, and morally perfect entity—I will call such an entity a “perfect God” for short—created the natural world. To say that God is “omnipotent” (all-powerful) and “omniscient” (all-knowing) is to say that it is logically impossible for something to have more productive power or more propositional knowledge than God has. It is more difficult to define “moral perfection,” but God’s being morally perfect entails at a minimum that God never performs a morally wrong action. Second, “naturalism” is the hypothesis that the natural world is a closed system, which means that nothing that is not a part of the natural world affects it. Naturalism is logically incompatible with theism because theism implies that the natural world was created (and so affected) by a supernatural entity (namely, God), while naturalism implies that there are no supernatural entities, or at least none that actually exercises its power to affect the natural world. By the “natural world,” I mean the collection of all existing physical entities (past, present, and future) together with any entities whose existence depends (either causally or ontologically) on the existence of those entities. “Natural” entities are entities that are part of the natural world so defined, and a “supernatural” entity, if there is such a thing, is simply an entity that can affect the natural world despite not being a part of it.
Now that we know what our two hypotheses assert, one might be tempted to ask which is better. This is a bad question, however, because it is doubtful that it has a single answer. We use hypotheses to pursue a variety of different practical and epistemic goals, and a characteristic of a hypothesis that furthers one of our goals may not help us or may even hinder us in our pursuit of some other goal. My own goal, as an old-fashioned philosopher, is truth. I want to know which of the two hypotheses, naturalism and theism, is more likely to be true. Thus, I will restrict my attention to theoretical virtues and vices that are virtuous or vicious because they affect the probability of a hypothesis’ being true. Specifically, I will compare the scope, simplicity, and predictive power of naturalism and theism. The first two of these factors, scope and simplicity, affect the probability of a hypothesis by affecting its intrinsic probability—its probability independent of what anyone knows or believes or perceives or remembers. The other factor, predictive power, affects probability only relative to the specific features of one’s “epistemic situation.” I do not deny that there are other factors besides these three that affect how probable a hypothesis is, but one can only do so much in the limited space allowed by my stingy editor.
Let’s start with scope. Roughly speaking, scope is a measure of how much a hypothesis purports to tell us about the contingent features of the world. Relative to certain practical goals, the larger the scope of a hypothesis, the better; but relative to the goal of truth, large scope is a vice rather than a virtue. For the more that a hypothesis says that might be false, the more likely it is to say something that is false, and hence the less likely it is to be true. For example, the statement that there is an animal behind the door says much less than the statement that there is a dog behind the door which in turn says much less than the statement that there is a collie wearing a red scarf behind the door. Thus, the first of these statements is intrinsically much more probable (though perhaps less useful) than the second and the second is intrinsically much more probable than the third. Similarly, the statement that there is no collie wearing a red scarf behind the door says much less than the statement that there is no dog (of any kind) behind the door which in turn says much less than the statement that there is no animal behind the door, not even an ant or a spider. Thus, of these three statements, the statement that there is no collie wearing a red scarf behind the door is the most probable intrinsically, while the statement that there is no animal of any kind behind the door is the least probable intrinsically.
It is not as easy to compare the scope of hypotheses like theism and naturalism, because neither entails the other and they are asymmetrical in more than one important respect. They both make a claim about the natural world, but the claims they make about the natural world are not symmetrical. Naturalism claims in effect that natural entities all lack supernatural causes, which is very different from saying, as theism does, that they all possess (proximate or remote) supernatural causes. Of course, theism says much more than this. It says that all natural entities share a single ultimate supernatural (necessary) cause, and while it doesn’t say that this common supernatural cause is wearing a red scarf, it does say that it is omnipotent, omniscient, and morally perfect. Because of the great specificity of these claims, I conclude, somewhat tentatively, that theism has much greater scope than naturalism does. I also conclude, quite confidently, that naturalism does not have greater scope than theism.
A hypothesis can be simple in more than one way, and simplicity can make a hypothesis better just by making it easier to use and understand. When, however, the simplicity of a hypothesis is understood to be a measure of the degree of (objective) uniformity that the hypothesis attributes to the world, then it is more than a merely pragmatic theoretical virtue. It is a sign of truth. Two examples will, I hope, help to make this point clear. First, compare the hypothesis that emeralds will remain green in the future to the hypothesis that they will sooner or later change from green to blue or from green to some other color. The former hypothesis is more probable than the latter, not because (or not just because) we have evidence that color changes of this sort never occur. Rather, it is intrinsically more probable because it attributes objective uniformity over time to the world while the latter hypothesis attributes objective change. A second example concerns Aristotle’s theory that physical objects are of two fundamentally different sorts: terrestrial and celestial. Unlike terrestrial objects, celestial objects are not composed of earth, water, air, or fire; and the laws that govern their behavior are not the same as the laws that govern the behavior of terrestrial objects. Even in the ancient world, it was recognized that attributing such ontological variety to nature was a weakness in Aristotle’s physics. Alternative theories that postulated greater uniformity were intrinsically more probable than Aristotle’s theory. Of course, Aristotelian physics was widely accepted for centuries, but only because it appeared to have much greater predictive power than its simpler competitors. In the end, of course, it proved to be false.
This position on simplicity is controversial. I am convinced that it is correct because our reliance on inductive reasoning is clearly justified and this is possible only if objective uniformity, either over time or at a single time, is intrinsically more probable than change or variety. This is not to deny, of course, the obvious fact that many hypotheses postulating change or variety are highly probable. The evidence supporting them can more than make up for their low intrinsic probability. Notice, though, that when we do discover change or variety we favor hypotheses that claim that this change or variety is itself uniform! For example, having observed that some freely falling bodies change their speed by accelerating, the Galilean hypothesis that they all accelerate at the same rate is taken to be more likely than other hypotheses that yield the observed data equally well.
Considerations of simplicity understood in this way provide another reason to believe that naturalism is intrinsically more probable than theism. On the one hand, theism postulates that one sort of entity—a perfect God—is the ultimate cause of other entities of a fundamentally different sort—natural entities. Naturalism, on the other hand, attributes one kind of ontological uniformity to the world: all entities that affect the natural world are themselves natural. Since naturalism attributes greater uniformity to reality than theism, it is simpler than theism in the crucial sense of the word. And since, as we have seen, it is also no larger (and probably much smaller) in scope, it is safe to conclude that, prior to testing our two hypotheses, naturalism has a significantly higher probability than theism. It is, one might say, a more plausible hypothesis.
4. Predictive Power: Preliminary Points
Although theism is less plausible than naturalism, I believe that it is more plausible than other forms of supernaturalism. More to the point, I do not believe that it is so implausible that it should not be tested. Thus, a comparison of the predictive power of our two hypotheses is called for. When I speak of the “predictive power” of a hypothesis, I am referring to the ability of that hypothesis to “yield the data” either deductively or inductively. In other words, the predictive power of a hypothesis with respect to some fact is the degree to which that fact should be expected to obtain on the assumption that the hypothesis is true. For example, suppose that money is missing from a safe and a reliable police detective tells us that he found Smith’s fingerprints on the safe. Does the hypothesis that Smith stole the missing money have a high degree of predictive power with respect to the fact that his fingerprints are on the safe? The answer will be “yes” if, independent of the detective’s testimony, the probability of Smith’s fingerprints being on the safe on the assumption that he stole the money is high. In other words, the answer is “yes” if a reasonable person who has not spoken to the detective and who assumes that Smith stole the money would expect his fingerprints to be there.
It is crucial to recognize, however, that what is important is not a hypothesis’ predictive power per se, but rather the ratio of its predictive power to the predictive power of the hypothesis or hypotheses to which it is being compared. For example, if Smith is the manager of the store and so his fingerprints are likely to be on the safe whether or not he stole the money, then those fingerprints will not provide strong evidence of his guilt because both the hypothesis that he is guilty and the hypothesis that he is innocent predict the presence of his fingerprints equally well. By contrast, suppose that Smith is not the manager of the store. Instead, he is a thief, but a thief who we know usually wears gloves. In this case, it is unlikely that his fingerprints would be on the safe if he stole the money. In other words, the predictive power of the hypothesis that he stole the money is low with respect to the fact in question. Nevertheless, if it is even more unlikely (or much more unlikely) that his fingerprints would be on the safe if he didn’t steal the money, then those prints are still evidence (or strong evidence) favoring the hypothesis that he stole the money over the hypothesis that he didn’t.
One last preliminary point. A hypothesis can have a much higher degree of predictive power with respect to some fact than a competing hypothesis even if that fact obtained and was known to obtain prior to the formulation of either theory. Some scientists and philosophers of science (mistakenly) give “extra credit” to hypotheses that predict facts that are not known to obtain at the time the prediction is made, as opposed to just “retrodicting” previously known facts. Since, however, this distinction is utterly irrelevant to evidential strength, I will use a single term, “predictive power,” to refer to the ability of hypotheses to do either.
5. Predictive Power: Main Points
My task, then, is to compare the predictive power of naturalism to the predictive power of theism. But with respect to which facts? I cannot examine all of the relevant facts in a short essay. (That is one of the reasons this book contains several sections, each dealing with a different area of evidence.) Here I will focus on the facts reported by the following statement, which I will call “E”:
E: For a variety of biological and ecological reasons, organisms compete for survival, with some having an advantage in the struggle for survival over others; as a result, many organisms, including many sentient beings, never flourish because they die before maturity, many others barely survive, but languish for most or all of their lives, and those that reach maturity and flourish for much of their lives usually languish in old age; in the case of human beings and some nonhuman animals as well, languishing often involves intense or prolonged suffering.
It may seem intuitively obvious that naturalism has greater predictive power with respect to E. However, a surprisingly large number of philosophers are skeptical of claims like this. Further, the claim that naturalism has greater predictive power with respect to E is not by itself all that significant. What needs to be established is that naturalism has much greater predictive power with respect to E. Given all this, appeals to what seems intuitively obvious will simply not suffice. Supporting arguments are needed, and I will provide two. First, I will show that, on the assumption that theism is true, we have good reasons to be surprised by E, reasons that we do not have when we assume that naturalism is true. Second, I will show that, on the assumption that naturalism is true, there are good reasons to expect E, reasons that we do not have when we assume that theism is true.
I am willing to grant, at least for the sake of argument, that suffering, when it contributes in a biologically appropriate way to the flourishing of a human being or animal, is not all that surprising on theism. From a theistic perspective, it makes sense to say that we are supposed to suffer in this way. That’s just the sort of organisms we are. But from a theistic perspective, it also makes perfectly good sense to say that sentient organisms are supposed to flourish. (Or at least this would make perfectly good sense if we did not already know that countless sentient beings do not flourish.) The argument for this has three premises. First, almost all sentient organisms are capable of flourishing in biologically realistic circumstances. This is proven by the fact that many do flourish and by the fact that the differences between those that do flourish and those that do not are in almost all cases relatively small. Second, sentient organisms have a good—they certainly can be benefited or harmed—and the failure to flourish is incompatible with achieving that good. Third, a perfect God, being perfect in moral goodness, could not care more deeply about sentient beings achieving their good, and being perfect in power and knowledge, could not be better positioned to ensure that sentient beings achieve their good. Therefore, the fact, reported by E, that countless living organisms, including sentient beings, never flourish at all and countless others flourish only briefly is extremely surprising given theism. It is not what one would expect to find in a living world created by a perfect God.
Of course, this argument assumes that a perfect God would be concerned with the good of individual sentient beings and not just with the good of one or more larger “holistic entities” like populations of organisms, ecosystems, or the biosphere. Few, however, will challenge this assumption, and so few will deny that, other moral considerations held equal, a morally perfect God would strongly prefer that every sentient being flourish for a significant portion of their lives. Granted, it is possible that an omniscient God would have good moral reasons unknown to us to permit sentient organisms to languish. This is why claims about what a perfect God would prefer must be prefaced with “other moral considerations held equal.” But it is also possible, and no less likely, that such a God would have good moral reasons unknown to us to prevent sentient organisms from languishing—reasons in addition to the reasons that are known to us. Thus, the probability of E given theism will depend largely on the moral reasons concerning E that we know about, not on the ones we don’t know about, and the reasons we know about all lower the probability of E given theism, which is to say that they all lower the predictive power of theism with respect to E. Since no parallel reasons lower the predictive power of naturalism with respect to E, it follows that E creates a serious intellectual problem for theists quite apart from evolutionary biology.
Evolution, however, makes this problem much worse for the theist, because it provides reasons to expect E given naturalism that we do not have given theism, and thus increases the ratio of the predictive power of naturalism to the predictive power of theism many-fold. The key point here is that naturalism, together with relevant background knowledge, is not neutral with respect to Darwinian evolution: given naturalism, it is very likely, not only that common descent is true, but also that what I will call “Darwinism” is true: natural selection accounts for all or almost all of the fantastic complexity we find in the living world.
Granted, when we assess the predictive power of naturalism with respect to E, we must abstract from the observations and testimony upon which our knowledge of E is based, and that involves abstracting from much of the evidence we have for the operation of natural selection. But even apart from that evidence, Darwinism is almost certainly true given naturalism because no other viable naturalistic explanation of biological complexity is available. Evolutionary change may have many causes besides natural selection (e.g., genetic drift), but that’s because not all evolutionary change involves increased complexity. The gradual development of highly complex organic systems requires something that can coordinate or give direction to a series of small evolutionary changes across many generations. And if naturalism is true, then what could possibly do that besides natural selection, broadly understood? Thus, when we assess the predictive power of naturalism with respect to E, we can equate it with the predictive power of Darwinian naturalism—that is, of naturalism conjoined with Darwinism. And E is far from surprising given Darwinian naturalism, for natural selection (and in particular “survival selection”) cannot operate unless there are winners and losers in the struggle to survive and reproduce. In the absence of supernatural assistance, a Darwinian world containing sentient organisms is almost inevitably cruel.
6. Objections and Replies
I will now consider three objections to my arguments in the previous section. First, one might object that what works for the naturalistic goose will also work for the theistic gander. In other (less fowl) words, if the naturalist can use Darwinism to help predict E, then why can’t the theist do the same? This objection misses the mark for two reasons. The first is that theism undercuts many of the predictions that Darwinian evolution makes about good and evil. For example, while survival selection is not at all surprising on Darwinian naturalism, other less brutal forms of natural selection would be more likely on theism. Thus, E is not as likely with respect to Darwinian theism as it is with respect to Darwinian naturalism.
The main reason, however, that this first objection fails is that one would not expect Darwinism to be true given theism. Darwinism differs from common descent in this respect. We have overwhelming evidence for a single tree of life (or at least for the thesis that the all life forms currently well known to science are related by descent). This evidence makes common descent extremely likely both on theism and on naturalism. Darwinism, however, though perhaps compatible with theism, is not particularly likely on theism (especially when one abstracts from the observations and testimony upon which our knowledge of E is based, which we must do in order to assess the predictive power of any hypothesis with respect to E). For Darwinism is the hypothesis, not just that natural selection can cause some evolutionary change (which has been proven experimentally), but that it accounts for all or almost all biological complexity. Given naturalism, we are justly confident in the truth of Darwinism, not because we know in any historical detail exactly how natural selection led to biological complexity, but rather because natural selection provides a way of explaining such complexity without having to appeal to the purposes of a supernatural designer. If theism is true, however, then natural selection is not needed to solve the problem of apparent teleological order in the living world. Theistic evolution could be Darwinian, but it could also proceed in a variety of other non-Darwinian ways. As long as a perfect God is guiding evolutionary change, natural selection is not crucial for the development of biological complexity. Thus, given theism, it would not be surprising at all if natural selection played no significant role in the development of such complexity. This means that, if E is to be expected on Darwinism, then that is a predictive success for naturalism, but not for theism.
This leads, however, to a second objection. Even if naturalists have Darwinian resources not available to the theist, don’t theists likewise have resources not available to naturalists? Specifically, theists have for centuries tried to explain various evils in terms of theism by appealing to morally significant libertarian free will. By “morally significant libertarian free will” I mean the ability to make choices for moral reasons, and to be morally responsible for those choices, without being causally determined to make them. The view that human beings have such freedom—I will call that view “libertarianism”—is very controversial. Among philosophers, most though not all naturalists reject libertarianism while most though not all theists accept it.
Let’s suppose, for the sake of argument, that the reason for this particular pattern of disagreement is that libertarianism is likely on theism but not on naturalism, in the same way that Darwinism is likely on naturalism but not on theism. Can libertarianism enhance the predictive power of theism with respect to E as much as Darwinism enhances the predictive power of naturalism with respect to E? The answer is clearly “no.” For while the free choices of human beings do have some impact on which sentient beings flourish and which languish, it is simply not true that, if there were no human beings making bad free choices, then most other sentient beings would flourish for all or most of their lives. One might be tempted to make an appeal to the doctrine of the Fall here, but that doctrine, if taken in the literal way necessary to make free will responsible for the failure of living organisms to flourish, is incompatible with the fossil record, which reveals that sentient beings existed and often failed to flourish long before human beings existed.
Finally, a third objection to my arguments concerning predictive power is that E doesn’t report everything we know about the kinds, amounts, and distribution of goods and evils in the world. This is true, but then the crucial question is whether we have any reason to think that this other knowledge helps the theist. I don’t have the space here to deal adequately with this complex issue, but my position is that this other knowledge just makes matters worse for the theist. Consider, for example, what we know about the good of knowledge (and the corresponding evil of ignorance). Human beings know a lot about their immediate environment and about other matters upon which their survival directly depends. Our cognitive faculties are, however, much less reliable when it comes to moral and religious matters. Surely this is much more surprising on theism than on (Darwinian) naturalism. Or consider the moral qualities of human beings. Humans are as a rule very strongly disposed—I’m tempted to say “hard-wired”—to act selfishly. They are instinctively much more concerned about their own interests than about the interests of others. They do possess some altruistic tendencies, but these are typically very limited. This combination of a deeply ingrained selfishness and limited altruism can be given a plausible Darwinian explanation, but is very hard to understand if, for example, God wants human beings, through the exercise of their free wills, to make substantial moral progress in their short time on earth.
Generally speaking, the pattern of good and evil in the world appears quite random from a moral point of view. It does not systematically promote or reflect any discernible moral ends. This fact is further evidence for naturalism, because, while it is compatible with theism, it is exactly what one would expect on naturalism. As David Hume wrote, “The whole presents nothing but the idea of a blind nature, impregnated by a great vivifying principle, and pouring forth from her lap, without discernment or parental care, her maimed and abortive children.” Hume (being dead) will not object if I replace his colorful appeal to a “blind nature impregnated by a great vivifying principle” with the more specific thesis of Darwinian naturalism.
I have shown in this essay that naturalism is, by virtue of its smaller scope and greater simplicity, a more plausible hypothesis than theism and also that naturalism has much greater predictive power than theism with respect to the facts reported by E. Of course, there may be other facts about the world that theism predicts better than naturalism, and there may be other factors besides scope, simplicity, and predictive power that affect the probabilities of these two hypotheses. Still, it follows from what I have shown that naturalism is much more probable than theism, all else held equal. And that entails that, all else held equal, theism is very probably false.
 Philosophers typically distinguish the “logical” problem of evil from the “evidential” problem. The logical problem of evil is the question of whether any known facts about evil are logically incompatible with theism and so conclusively disprove theism. Few contemporary philosophers defend an affirmative answer to this question. The “evidential” problem of evil is the question of whether any known facts about evil bear some negative evidential relation to theism other than the relation of logical incompatibility. Philosophers of religion are divided on the correct answer to this question. This essay defends an affirmative answer.
 The Foundations of the Origin of Species: Two Essays Written in 1842 and 1844, ed. Francis Darwin (Cambridge: Cambridge University Press, 1909), pp. 51-52.
 The Autobiography of Charles Darwin and Selected Letters, ed. Francis Darwin (New York: Dover Publications, Inc., 1958), p. 64.
 To say that a hypothesis is a “statement” is to say that it makes an assertion and so is either true or false.
 Although the word “created” is past tense, I do not intend to assume that the creator is, according to theism, a temporal being rather than a timeless one, nor do I assume that the natural world is not infinitely old. I only assume that, if theism is true, then the natural world depends for its existence on a perfect God. This implies that the natural world is “affected” by God in the broad sense that God makes a difference to whether it exists and what it is like, but the exact nature of the dependence relation between God and the world is left open.
 This sentence employs a number of technical terms. First, to say that something is “logically” impossible is to say that it is impossible by virtue of entailing a contradiction. For example, it is logically impossible for a five sided triangle to exist because that would entail the existence of something that both has and does not have three sides. “Productive” power is the power to create or produce or bring about events, objects, etc. This is distinct from the power to be affected by other things and the power to perform actions. “Propositional” knowledge is knowledge that a proposition (statement) is true or that it is false. It is distinct from acquaintance knowledge, which is knowledge one has of something (e.g., boredom or the taste of chocolate) by experiencing it and thus being directly acquainted with it.
 A feature of the world is “contingent” if it is logically possible that the world not have that feature. For example, the world has the feature of containing me. Since it is possible that the world not have contained me, it follows that this is a contingent feature of the world. By contrast, the world also has the feature of either containing me or not containing me, which is not a contingent but a necessary feature of the world.
 Those familiar with Goodman’s paradox may object that uniformity is language relative. This is one of the reasons I use the adjective “objective” in front of uniformity. The claim that emeralds change from “grue” to “bleen” attributes objective uniformity to the world while the claim that emeralds change from green to blue attributes objective change to the world.
 There is an implicit premise in my argument here, namely, that theism is not significantly more “coherent” than naturalism, where the degree of “coherence” of a hypothesis depends on the evidential relations between its parts.
 Notice that this assumption does not deny that holistic entities like ecosystems are just as “real” as individual organisms. It also does not explicitly deny that holistic entities have moral standing, although I don’t believe they do because I don’t believe that entities lacking any sort of subjective awareness can be literally harmed or benefited.
 William Dembski makes this sort of point in more than one place. See, for example, The Design Revolution: Answering the Toughest Questions about Intelligent Design (Downers Grove, Illinois: InterVarsity Press, 2004), chap. 36. See especially pp. 261-263.
 I use the term “natural selection” to refer, not just to cases in which a characteristic is selected because it enables an organism to survive long enough to reproduce (“survival selection”), but also to cases in which a characteristic is selected either because it makes an organism that survives long enough to reproduce more likely to find a willing mate (sexual selection) or because it makes an organism that finds a willing mate likely to have a greater than average number of surviving fertile offspring (fecundity selection). Other variations involving group selection are also possible.
 William Hasker has helped me to appreciate this objection.
 See Stewart Goetz and Charles Taliaferro’s opening case (“An Argument from Consciousness and Free Will”) in section one of this e-book for more on the tension between naturalism and libertarianism.
 Dialogues Concerning Natural Religion, Part XI.
 I am grateful to Glenn Branch, Jeff Jordan, Jeff Lowder, and John Schellenberg for helpful comments on earlier drafts of this paper.
Copyright ©2007 Paul Draper. The electronic version is copyright ©2007 by Internet Infidels, Inc. with the written permission of Paul Draper. All rights reserved.