The Marvelous and Miraculous: A Defense of Hume

1. Hume's argument against miracles is often paraphrased as follows: in evaluating stories of miracles, we should balance the testimony for the event against the evidence for the law the event violates. Since the evidence for the law is "a uniform experience," it always outweighs the evidence for the testimony. Hence, there is always "a direct and full proof" against the miracle. The only condition under which it would be reasonable to accept the testimony would be when its falsity would be "more miraculous than the fact it endeavors to establish."1

Many critics have found this argument inconsistent with our practice in evaluating testimony. The first was George Campbell who argued in 1762 that we routinely accept testimony that is less probable than the events themselves. According to Campbell, if someone mentions "that at such an hour, of such a day, in such a part of the heavens, a comet will appear," the chances "would not be as millions, but as infinite to one, that the proposition is false" since comets occur so infrequently. Yet if "you have the testimony of but one man of integrity, who is skill'd in astronomy," that the comet appeared at that time and place, "you will not hesitate one moment to give him credit."2 If we take his claim that the odds are infinite as hyperbole, Campbell's point is that a comet is considerably less likely than the falsity of the testimony, yet we accept the testimony without hesitation and without any implication that this is irrational. No matter how great the presumption against the comet, one reliable witness can overturn it. And, if we accept testimony to comets in this way, there is no reason why we shouldn't also accept testimony to miracles.

Campbell's objection has had a long history. Richard Whately used it as the basis of his satire, Historic Doubts Relative to Napoleon Bounaparte in 1819.3 Whately argued that, on Hume's principles, we should believe that Napoleon's career was an elaborate hoax. In addition to being unique in recorded history, it resembles epic poetry and reminds "the sober-thinking few of the Arabian Nights." "All the events are great and splendid, and marvellous: great armies, great victories, great frosts, great reverses, 'hair-breath 'scapes,' empires subverted in a few days." Every event has "that roundness and completeness which is so characteristic of fiction": complete victories, total overthrows, subversion of entire empires.4 Taken together, they add up to "a tissue of absurdities" on Humean principles. Whately's point is that Napoleon's victories are no less incredible than the divinity of Jesus; each is unprecedented; thus one who believes what is said about Napoleon but not Jesus is "unduly prejudiced against whatever relates to religion."5 The objection can also be found in the recent literature. One contemporary critic argues that, if Hume were correct, we should not believe newspaper stories about lottery winners, since the probability that the reported winner has won is considerably less than the probability that the newspaper is mistaken.6

What are we to make of this? I think the criticism is valid against the argument I sketched in the beginning. But it is also not clear that this is the argument Hume had in mind. There are two reasons for thinking this. First, when asked to comment on Campbell's book by Rev. Hugh Blair, Hume was unmoved, writing "I find no difficulty to explain my meaning and yet shall not probably do it in any future edition."7 And in the brief comments he sent to Blair, Hume's main point seems to be that Campbell has systematically misunderstood him. Second, Campbell's claim that miracles are just unusual events like comets or accidents also appears in Bishop Butler's defense of miracles which predates Hume's essay. Given Hume's knowledge and admiration of Butler, it would itself be extraordinary for him to have overlooked the objection, if he were just claiming that miracles are untenable based on a balancing of the probabilities on both sides.8

2. This still leaves the question of what Hume's criticism of miracles is, if not this. My interpretation is that he holds that testimony to a miracle automatically disqualifies itself; hence, miracle stories are not evaluated by rules governing other testimony, but are unreasonable on their face. Hume expresses this by distinguishing between the miraculous and the marvelous or, as he also calls it, the extraordinary. Miraculous events are violations of laws of nature and so are unacceptable on testimony, no matter how good the witnesses, but marvelous events are acceptable on good testimony, since they are consistent with laws. Let us first look at this distinction, then turn to Hume's claim that we can know that testimony to a miracle is unacceptable on its face.

This remark occurs in the context of his discussion of Joan of Arc. Among the stories about her are these: that she recognized the king in a crowd of courtiers even though she had never seen him before and he was dressed just like them, that she proved that she was on a divine mission by revealing a secret to the king "which nothing but a heavenly inspiration could have discovered to her," and that she described in detail the special sword she wanted to carry into battle and where it could be found even though she had never seen it. Hume considers these to be miraculous and holds that "they were spread abroad, in order to captivate the vulgar."10 But he still accepts the story of her career: that she was a peasant woman, that she led the battle to raise the English siege of Orleans, and that she escorted the king to Rheims to be coronated. These events, he thinks, are not miraculous and so can be believed on good testimony.

Hume accepts the story of Joan because there is nothing miraculous in a peasant leading an army to a great victory. Even if it has never happened in the course of history, the event violates no laws. We know that leaders sometimes rise from low positions to command armies; we also know how it raises the morale of troops to think their leader has been sent by God, so even her victories can be explained. This part of the story of Joan is acceptable, given the evidence. But it is not reasonable to believe that Joan could recognize the king in disguise without ever having met him or that she had special knowledge of a sword she had never heard of or seen before. These events violate known laws and so are unacceptable even though they are based on the same testimony as the others.

The distinction also appears in the Treatise (III i 2). Hume considers whether virtue is natural and distinguishes three senses of 'nature'. "If nature be opposed to miracles, not only the distinction between vice and virtue is natural, but also every event which has ever happened in the world, excepting those miracles on which our religion is founded"(T 474; Hume's italics). In the second sense, "nature may be opposed to rare and unusual." He takes this to be the common use of the word and observes that we do not have "any very precise standard" for judging what is natural in this sense. "Frequent and rare depend upon the number of examples we have observed; and as this number may gradually increase or diminish, it will be impossible to fix any exact boundaries betwixt them." Virtue is also natural in this sense, since every nation distinguishes between virtue and vice. In the third sense, nature is "opposed to artifice." In this sense, Hume says, "tis certain, that both vice and virtue are equally artificial, and out of nature." (T 475). Hume does not elaborate on the distinction between the miraculous and the unusual in this passage, but it is clear that he distinguishes them. In one sense, the natural contrasts with the miraculous, and in another with the marvelous.11

A fuller statement of the distinction occurs toward the end of the miracles essay. Hume considers two imaginary stories about January 1, 1600. He says that it would be reasonable to believe on good testimony that the earth was dark for eight days starting on January 1, 1600, but not that Queen Elizabeth died on that day, then appeared again after being interred for a month and ruled England for another three years. The difference between the cases is that, according to Hume, eight days of darkness does not violate any law since "the decay, corruption, and dissolution of nature" is rendered probable by many analogies. Thus, an event exhibiting it "comes within the reach of human testimony if that testimony be very extensive and uniform." But he says he would not have "the least inclination to believe" that the queen could be resurrected since it is "so signal a violation of the laws of nature" (I 138).

Campbell thinks Hume is confused about this. Unlike Hume, he thinks it would be as miraculous for the world to be dark for eight days as for someone to rise from the dead, and concludes that there is no distinction for Hume to draw.12 But this is a mistake. Hume clearly thinks eight days of darkness is possible, given the laws of nature. It is thus extraordinary, not miraculous, and hence is believable on testimony. Campbell disagrees on where to classify the event, but this does not mean that the distinction itself is unacceptable. Hume's point would have been clearer if he had taken a less controversial example of the marvelous. For instance, he could have contrasted the story of Elizabeth's resurrection with one of her leading an English army in 1600 at the age of 67 to the conquest of both France and Spain. This would be extraordinary, but given our knowledge of the exploits of Joan of Arc, Caesar and other great leaders, it is not impossible. Hence, according to Hume's principles, it would have been acceptable on strong evidence, even though the queen's resurrection would not have been.

Given this, it is clear why Campbell's criticism did not impress Hume. Since comets occur, they do not violate any law. Hence it is reasonable to accept reliable testimony to them. The fact that they are statistically extraordinary does not matter. Hume would also not have been impressed by Whately's example. Napoleon's career is not miraculous even if we admit its uniqueness in history. Nor is it a miracle for my friend to have won the lottery even though the odds were millions to one against him. In each case, the evidence against the events can be overcome, since we all agree that analogous cases have and can occur. That someone could be raised from the dead is quite a different matter. Since this violates a law, we do not recognize exceptions; hence according to Hume, testimony cannot reasonably establish that it has happened.

In the Inquiry, Hume makes a threefold distinction between demonstration, proof and probability (I 69n; see also I 120, T 124). He does not explain demonstration, but refers to Locke and seems to mean by it what he means by 'knowledge' in the Treatise: "assurance arising from the comparison of ideas" (T 124). Proofs, he says are "such arguments from experience as leave no room for doubt or opposition" (I 69n). He thinks that we have proofs in this sense for the propositions that all men die and that the sun will rise tomorrow. Since all men have died in the past and the sun has always arisen, the evidence for these is complete, even though they are matters of fact and could be false. In the case of probabilities, there are counter cases which cause doubt, leading us to weigh the evidence on both sides. But in a proof it is not necessary to balance the evidence in this way.

In the case of miracles, we have a law supported by "a uniform experience." According to Hume, this means that testimony to its violation is unsupported by other experience; if the witnesses are right, the event they attest to would be the first case in which the law has failed. Thus we do not have to weigh the testimony against the evidence for the law; it is reasonable to reject it regardless of the credibility of the witnesses. But if there are known violations or known cases that are analogous to them, we know that exceptions are possible. In these cases, we have to weigh the testimony along with the evidence that the event did not happen. These events are more or less marvelous depending on how unusual they are, but they are not miraculous. As a result, it can be reasonable for us to accept them.

4. Hume's argument here may seem to rule out miracles by fiat, but he thinks it is supported by our practice in evaluating evidence. In the Inquiry, he praises the Cardinal De Retz, for instance, for thinking it is unnecessary to trace the falsity of testimony for a miracle "through all the circumstances of knavery and credulity which produced it." Rather, Hume says, the cardinal concluded "like a just reasoner, that such an evidence carried falsehood upon the very face of it, and that a miracle, supported by any human testimony, was more properly a subject of derision than of argument" (I 132). Similar passages appear throughout the essay. They are often taken to have only a rhetorical function: Hume is donning a mantle of piety by citing churchmen who have been as skeptical as he is. But I think he has another aim. He wants to show that "just reasoners" follow the principles of testimony he is attempting to defend in the essay, Thus, his criticism is supported by our practice in evaluating testimony and does not simply impose an arbitrary standard.

In the letter to Blair, Hume makes an even stronger claim. He asks rhetorically: "Does a man of sense run after every silly tale of witches or hobgoblins or fairies, and canvass particularly the evidence?" He then adds sarcastically that he never knew anyone who "examined and deliberated about nonsense" who did not end up beieving it.14 People sometimes think Hume is just being dogmatic here, but his view is actually quite reasonable. If a witness testifies at a trial that the defendant got past the bank guards by turning himself into a roach, the court would not balance the chances on either side, but would consider the testimony unreliable on the face of it. In fact, the witness' story would undermine any further evidence he might give as well. This is not dogmatic, but reasonable. It is unreasonable on the face of it for a human to turn into an insect, since the event violates known laws. The witness is thus lying or deluded, no matter how honest and trustworthy he has been in the past. In the language of the law, it is beyond reasonable doubt that such events do not occur.

In Campbell's defense, it must be said that Hume does not always express the point clearly in "Of Miracles." He claims, for instance, that we should weigh the evidence on both sides and accept the stronger. He also says that we should accept testimony to a miracle only if its falsehood would be more miraculous that the falsehood of the fact it claims to establish (I 123-124). In these passages, he seems to equate 'miraculous' and 'improbable' despite his view that the miraculous is beyond the improbable, and so leaves himself open to Campbell's objection. But these passages are not decisive. I will concentrate on his remak that the falsity of the testimony must be more miraculous than the falsity of the law, since this is the crux of the argument in part I of the essay.

In this part of the essay, Hume says that he will assume that the grounds for and against the supposed miracle are both proofs. This means that the law has always been confirmed, i.e. supported by "uniform experience," whereas the witnesses to its violation have always been reliable observers, truthful and had good memories. On this supposition, then, "there is proof against proof, of which the strongest must prevail" (I, 122).15 Hume's argument is that we should accept the law and not the miracle, since, no matter how reliable the witnesses, the relation between their testimony and reality is never a law. We know that witnesses make mistakes and lie, so it is possible that these witnesses are mistaken or lying. Although their testimony is a proof "considered apart and in itself" (ibid.) it is always defective, since there is always a reasonable doubt about it. Read in this way, Hume does not hold that here are degrees of miracles or that miracles are just extraordinary events. We reject the testimony because its falsity is not miraculous at all. If witnesses were never deceived, never misremembered and never lied, we might conclude that there is a law relating testimony and reality, and hence have to weigh one miracle against another, namely, the miracle that the testimony failed and the miracle that the law failed. But this is not the case in our world. Here the falsity of the testimony is never "more miraculous than the fact which it endeavors to establish" (I, 123), for it is never a miracle for testimony to be mistaken.16

5. As I have interpreted him, Hume takes stories of the miraculous to be something more than just highly improbable. This is reflected in his definition of miracles. A miracle is a violation of a law of nature and so is not relative to evidence. More fully, he defines a miracle as "a transgression of a law of nature by a particular volition of the Deity, or by the interposition of some invisible agent" (I 123n). Campbell complains that this definition suggests that God is engaged in wrong-doing, i.e. he is a transgressor of the laws, but of course Hume has no such implication in mind.18 His point is that a miracle is inconsistent with a law. The concept of miracle is thus a metaphysical notion, if you will, and not an epistemological one. This is squarely in the theological tradition. Aquinas, for instance, holds that "A miracle properly so called takes place when something is done outside the order of natute."19 He does not elaborate, but it seems clear that by "the order of nature" he means natural law. His example of a miracle is the curing of a fever "without the operation of nature." Presumably his view is that, if a child is cured because the fever has run its course or because it was given a cool bath, i.e. because of some natural phenomenon, the event is not a miracle; only if God intervenes with the order of nature, say because he is answering the parent's prayers, is it a miracle.

These theses follow from Hume's understanding of miracles. First, a miracle can occur even if we have no evidence it has. Hume says that the raising of a feather a fraction of an inch when there is no wind is as much a miracle as the raising of a house, even though we might have no suspicion that it occurred. Such miracles are invisible while those that affect our senses are visible.20 Second, a miracle must transgress an actual law and not just a believed one. If we mistakenly believe that L is a law, we might believe that a violation of L is a miracle, but this would be a mistake. Events are not miracles just because they violate believed laws; they must be inconsistent with actual laws. Finally, on Hume's account, miracles do not admit of degrees. They either violate laws and are miracles, or they do not and are not miracles; there is no middle ground.

6. This throws light on a possible misunderstanding about the story in the Inquiry of the Indian prince who refused to believe the European explorer's accounts of the freezing of water. Hume thinks the prince "reasoned justly"; since he had never experienced frost or ice, it would have taken "very strong testimony" to convince him (I 121). But Hume makes it quite clear that frost was not miraculous, but just extraordinary (I 122n). Yet, presumably the prince believed that frost was miraculous; hence, if pressed, he would have rejected the story as contrary to law. But this is not relevant. It is not a law that water is always liquid just because someone believes it is. Had the prince taken it to be a law, he would have been mistaken, as a trip to "Moscovy" (or Ottawa) in winter would have shown him. Hume would admit that it is always difficult to tell when a regularity is a genuine law and when it is just a local phenomenon. Indeed, he is the father of such doubts. But this does not mean that laws are just experienced regularities. To be a law, a regularity must be universal with respect to time and place, and although violations of laws are logically possible, they are not physically possible. As a result, the prince would have been mistaken, no matter how strongly he believed that frost violated a law.

7. In closing, I would like to consider two objections that are often made against Hume: a. The first is that he is mistaken in thinking that we should always prefer the law over an alleged violation, since the law itself is inductive. Suppose L is a law and e an event that violates it. Since L is supported to degree p there is no reason in principle why the evidence for e could not be greater than p, and hence no reason why we should not think that a miracle has occurred.

The problem with this is that it confuses the issue of whether the law statement is true with the question of whether a miracle has occurred. The defender of miracles does not just claim that beliefs about laws may be mistaken; he is committed to holding that they are true even though events that contradict them have occurred. We might believe that L is a law, then reject this belief because we have strong evidence that e has occurred , but this does not show that e is a miracle. We have taken e to show that L is not a law and never was, despite our previous evidence. To be a miracle, L must be an actual law.

In the theological tradition, a miracle occurs when God temporarily suspends the law and causes the event to occur. This is what it means for a miracle to be inconsistent with the course of nature. To use the previous example, suppose a child recovers from a disease that always causes death. This would be miraculous because the law is actual; it is just that it did not operate in this instance because God intervened directly to allow the child to recover. Hume holds that we can never reasonably believe that the law holds and that the child survived the disease. It is always more reasonable to believe that the law held even in this case than that the child had a disease and lived. If the defender of miracles wants to claim that there is no law, well and good, but this will then imply that the child's living was not a miracle. The fact that our evidence for the law is imperfect and that our acceptance of it might be a mistake is thus beside the point.

b. This explains another objection that is often made to Hume. It is often argued that, if Hume is right, we cannot explain how laws come to be rejected in science. On Hume's principle, it is never reasonable to believe that a law has been violated; thus, it is never reasonable to accept testimony that there is experimental evidence against the law. One answer to this is that, in fact evidence against accepted laws is not readily accepted, but has to be confirmed by other tests. But there is also a more basic problem. A scientist who claims to have refuted a law is not claiming that a miracle has occurred, since he does not believe that the disconfirming experiment violates a law. His view is that there never was a law; we just believed mistakenly, as it happens, that there was. We might say that the experiment violates the law, but this is only a way of saying it is inconsistent with what we have accepted as a law. Neither we nor the scientist want to say (i) that there is a law of nature as opposed to a believed law and (ii) that the experiment violates this law.

What would be needed to show that a miracle has occurred? First, we would need a clear law and, second, good evidence that a violation has occurred even though nature continues to be governed by the law, i.e. evidence that a negative instance of the law has occurred. If both these conditions are satisfied, we have three possibilities. (1) We could accept the event and reject the law. This would be to deny that the event is a miracle, for without the law there is no reason to think it required some special intervention by God and was not produced by natural causes. (2) We could accept the law and reject the event. This would be to follow Hume's advice and reject the testimony for the miracle. (3) We could accept both the law and the negative instance. As we saw earlier, this is not a contradiction, for the religionist can argue that God merely suspended the law momentarily while causing the event. This avoids the contradiction, but leaves the problem of showing that God intervened in this way. Since we have no direct knowledge of his intentions, the best we can do is show that the law held before and after the event, but this, Hume argues is evidence that the event did not occur.21 It is more plausible to think the witnesses lied than that the law failed in just this case; for, as he says, "it is nothing strange, I hope, that men should lie in allages" (I 127). On the one hand, we have evidence for the law that puts it past doubt, i.e. we have a proof; on the other hand, we have the testimony that a negative instance occurred. No matter how strong the testimony, there is always a reasonable doubt about it. Since we know that men lie, often for noble reasons to further causes they think everyone should accept, it is possible that they are lying now. We might check their character and quiz them further, but this does not eliminate the doubt, for even saints lie for the sake of our souls, if not for their own. The result is that it is more reasonable to believe that the law has continued to hold through the time when the witnesses said it failed. And since this argument can be generalized, it follows that it is never reasonable to believe in miracles.

It might be thought that this is just the original balancing-of-probabilities arguement all over again, but this is a mistake. No matter how good the evidence supporting the witnesses, we cannot eliminate the reasonable doubt that they are lying or mistaken, whereas no one on either side has any doubt about the law being actual. There is thus always a doubt about the testimony, but none about the law. This argument is based on rules of evidence Hume thinks have been shown to be reliable guides in the past. They are supported by experience in a broad sense, but are not just inductive generalizations from instances. They are deontological, if you will. They characterize the reasoning of the best and most careful thinkers, and so help define what it is to be a reasonable thinker. These principles might lead us to make mistakes in specific instances, but overall, Hume thinks, they will lead us to avoid more mistakes that we would make otherwise; they also help us to discover the truth more often than not. Looked at in this way, Hume's position is much more powerful and interesting than Campbell and Whately take it to be. It also seems to me to be the correct way of looking at the status of principles of evidence in non-deductive logic and the theory of knowledge.22

1. David Hume, An Inquiry Concerning Human Understanding. ed. Charles W. Hendel (Indianapolis, Ind, 1955), p. 123. References to this edition will be given in the text and indicated by I followed by the page number. References to the Treatise are to A Treatise of Human Nature, ed. I. A. Selby-Bigge (Oxford, 1888) and will be given in the text as T followed by the page number.

3. Richard Whately, "Historic Doubts Relative to Napoleon Buonaparte," in Essays in Philosophy, ed. Houston Peterson (New York, 1958), pp, 143-171.

6. Roberet Hambourger, "Belief in Miracles and Hume's Essay," Nous, 14 (1980), 591-593. The criticism even predates Hume's essay. See footnote 8.

8. In defending miracles in The Analogy of Religion, pt. II, ch. II, in 1736, Bishop Butler argued that the prior improbability of an event is no reason for doubting its occurrence. See The Works of Joseph Butler, ed. W. E. Gladstone (Oxford, 1896), II, 216-220. As Butler put it: "There is a presumption of millions to one, against the story of Caesar, or of any other man" (216-217).

9. David Hume: Philosophical Historian, ed. D. F. Norton and R. H. Popkin (Indianapolis, Ind., 1965), p 128.

11. There is also a suggestion of the argument against miracles (as I interpret it) in the Treatise. In book II, iii, 1, Hume says that if a traveler claimed he had seen fruits ripen in the winter in a climate of 50 degrees north latitude, "he would find few so credulous as to believe him" (T 402-403). He then claims that the traveler would find the same reaction, if he said that he had visited a country where the inhabitants have the same character as those in Plato's Republic or in Hobbes' Leviathan. His point is that "There is a general course of nature in human actions, as well as in the operations of the sun and the climate." Hume does not mention miracles in this passage, but in the comparable passage in the Inquiry he says that, if a traveler reported a people "entirely divested of avarice, ambition, or revenge," we would "prove him a liar with the same certainty as if he had stuffed his narration with stories of centaurs and dragons, miracles and prodigies" (I 94).

14. Ibid. Compare the more muted passage on I 128: "Fools are industrious in propagating the imposture, while the wise and learned are contented, in general, to deride its absurdity, without informing themselves of the particular facts by which it may be distinctly refuted."

15. In his remarks about the notion of proof in the letter to Blair, Hume says that "there are degrees of this species, and when a weaker proof is opposed by a stronger, it is overcome." Letters. I, 350.

16. The argument in part I of "Of Miracles" is often taken to be a priori, while the arguments in part II are a posteriori. See R. M. Burns, The Great Debate on Miracles: From Joseph Glanvill to David Hume (Lewisburg, Pa., 1981), pp. 142 ff., esp. 142-143, 153-154, 158-159. If my interpretation is correct, this is mistaken. An a priori argument does not rest on any empirical premisses or assumptions, but Hume's argument rests on the premiss that what witnesses say is sometimes false. The confusion arises, I think, because Hume holds that the argument in part I is sound regardless of the specific testimony for the miracle. Hence we do not have to investigate this evidence, but know beforehand that it is weaker than the evidence for the law. This makes the argument appear a priori, even though it is not.

17. Leibniz thinks that everything happens according to the divine plan, but distinguishes between "subordinate maxims" and "the most general of God's laws." The former are natural laws because they govern "the nature of things"; the latter are God's general intentions. Miracles, he says, violate the subordinate maxims but not the most general laws. G. W. Leibnez, Discourse on Metaphysics section 7, in Philosophical Essays, tr. Roger Ariew and Daniel Garber (Indianapolis, 1989), p. 40. When Hume says that miracles violate laws of nature, he has in mind these subordinate maxims. As he makes clear in section XI of the first Inquiry, he thinks we have no evidence of God's overall intentions.

19. Summa Theologica, I, qu. 110, art.4, The Basic Writings of St. Thomas Aquinas, ed. Anton C. Pegis (New York, 1945, 1, 1022.

20. The distinction between visible and invisible miracles is found in Butler. He holds that the incarnation of Jesus and the revelation of truths in the Bible are invisible miracles, since they are not "a proof of a divine mission," Analogy of Religion, pt. II ch. ii, see Works, II, 213. A visible miracle, on the other hand, is one that falls within our experience and so can act as a sign to guide the faithful, in particular, to prove the invisible ones. Both types are outside the course of nature in Aquinas' sense, since they result from God's direct intervention rather than from natural law.

21. That we have no evidence of God's intentions is the topic of section XI, "Of a Particular Providence and of a Future State."

22. An earlier version of this paper was read at the College of St. Rose, Albany. I would like to thank the audience for comments that helped form the final version. I would also like to thank Michael Smithurst who commented on the paper at the 20th Hume Conference in Ottawa (1994) and members of the audience who encouraged me to develop the argument further.