Miracles and Hume

Nobody in this dispute denies that the reports exist. Every side agrees that people have said, in large numbers and often at cost to themselves, that something happened which the ordinary course of things does not allow.

The question is what a report can do against a regularity that has held every other time it has been checked. That is a question about evidence rather than about theology, and most of what follows would apply unchanged to a claim about anything else that has never been observed to happen.

Laws and their violation

The second half of that is what separates a law from an accidental regularity. It may be true that every coin in my pocket is copper, and that fact tells you nothing about what would have happened had I put a different coin in. A law tells you what would have happened, which is why it can be used to predict and to explain.

There is a long dispute about what makes the difference, and it bears directly on the present question. On a regularity view, of the kind Hume himself held, a law is nothing more than a universal pattern in what actually occurs. On richer views, a law states a genuine necessity in nature.

The regularity view creates an immediate problem for the idea of a violation. If a law just records what always happens, then an exception does not break the law; it shows that the generalisation was never true and therefore never a law. On that account "violation of a law of nature" describes nothing, and this objection has been pressed as showing the concept of a miracle to be incoherent rather than merely improbable.

Richard Swinburne's answer, in The Concept of Miracle (1970), is to treat a miracle as a non-repeatable counterinstance to a law. Laws are established by what can be repeated. If an event conflicts with a well-supported law but cannot be reproduced, and no amended law covering it does better at everything else the original law covered, then the rational course is to keep the law and treat the event as an exception rather than to abandon a generalisation that works everywhere else.

That is a serious reply and it has a price. It makes the identification of a miracle depend on a judgement that no better law is available, which is a judgement about future science.

What a miracle is

The characterisation is Hume's, and choosing it is not neutral, which is worth registering before his argument is assessed. Defining a miracle as a violation puts the event in maximal conflict with the best-supported thing we know, and an argument that such events are hard to establish is then partly an argument from the definition.

Aquinas defined it differently, as an event exceeding the productive power of nature, where nature covers what created things can do. That formulation does not require anything to be broken. It requires only that the cause lie outside the created order, which is a claim about what produced the event rather than about a regularity being contravened.

The difference matters for what a case would have to show. On Hume's definition, establishing a miracle requires establishing an exception to a law. On Aquinas's, it requires establishing that nothing in nature could have done it, which is harder in one way, since it quantifies over everything nature can do, and easier in another, since it does not require the event to be inconsistent with anything.

Hume's argument

The Enquiry was published in 1748, and the section has two parts that do different jobs and are frequently run together.

Part I is the general argument. A wise person's confidence tracks the evidence. The evidence for a law of nature is uniform experience, which is the strongest evidence available for anything. The reliability of human testimony is itself something we know from experience, and experience tells us that testimony fails often: people are mistaken, people lie, people repeat things they did not check. So testimony to a violation of a law sets a weaker body of evidence against a stronger one.

From this Hume derives the maxim that has carried the argument ever since: testimony suffices to establish a miracle only if its being false would itself be more extraordinary than the event it reports, and in weighing the two, the greater always destroys the lesser.

Two misreadings are worth heading off, because both are common and both make Hume look easier to answer than he is.

Hume does not argue that miracles are impossible. The argument is entirely about what it is reasonable to believe on testimony, and it would be unaffected if miracles occurred regularly, so long as our evidence for them came only through reports.

And on the most defensible reading he does not claim testimony could never in principle suffice. The maxim states a condition, and a condition can be met. Part II is what converts the in-principle point into a practical verdict.

Part II offers four empirical considerations. No miracle report, Hume claims, comes with enough witnesses of sufficient education, judgement, and reputation, attesting to something done publicly in a well-known place. The passions of surprise and wonder give people a positive appetite for such stories and a motive to pass them on. Reports of this kind are commonest among populations with the least means of checking them, and thin out as those means improve. And the different religions each have their own miracles, supported by evidence of much the same quality, so that each set of reports counts against the others, since the events are offered as credentials for incompatible systems.

The fourth is the strongest and it is aimed at a specific target. Hume's readers rejected the miracles of other traditions as a matter of course, and his point is that they had no principle for doing so which did not also dispose of their own.

The four considerations are empirical, and that is their weakness as well as their force. An empirical claim about the quality of the available reports can be answered by producing a better report, and that is precisely what the modern defence attempts.

The Bayesian turn

The formal treatment of testimony, developed since Hume, does not favour him.

Charles Babbage put the central point in 1837. If several witnesses report the same event and their errors are independent, then the probability that all of them are wrong is the product of the individual probabilities of error, and a product of small numbers gets small very quickly. Given enough independent witnesses of even moderate reliability, testimony can outweigh any prior improbability short of impossibility. Modern reconstructions using the probability calculus reach the same result, and the arithmetic is not in dispute.

John Earman pressed the case hardest in Hume's Abject Failure: The Argument Against Miracles, Oxford University Press (2000), arguing that Part I does not establish what Hume claimed and that the reasoning behind the maxim is confused. Earman is not defending any miracle report; his target is the quality of Hume's argument, and the criticism has been influential. Robert Fogelin replied in A Defense of Hume on Miracles, Princeton University Press (2003), and Peter Millican has argued for readings of Hume that survive the formal objections.

The place where Hume's position is best defended is exactly where the arithmetic is weakest, which is independence. The multiplication works only if the witnesses' errors are unconnected. Witnesses to religious events are rarely unconnected: they typically share a community, expectations, and an interest in the outcome, and the reports usually reach us through a chain of transmission rather than from the witnesses directly. Enthusiasm is contagious in a way that produces correlated error, and correlated error does not multiply down. Hume's second and third considerations are, in modern terms, arguments that the independence condition fails, and that is a much better argument than the maxim it was attached to.

What a miracle would show

Suppose all of this were settled and some extraordinary event were established beyond reasonable doubt. There is a second question, and the theistic side of this debate has generally been weaker on it than on the first.

An unexplained event is not by itself a divine act. To get from one to the other you need an attribution, and the attribution is not supplied by the event. The reasons offered are typically the event's religious setting, its timing, what was said before it, and its fitting a purpose one would expect God to have. Every one of those depends on a prior view about what God is like and what God would do, which was the thing the miracle was introduced to support.

This does not make the inference worthless, and the fine-tuning article set out the structure it would need. It does mean that a miracle cannot do the job it is usually assigned in popular apologetics, which is to establish a religious claim from a standing start. At most it can confirm a hypothesis that is already on the table for other reasons, by being much more expected on that hypothesis than on the alternatives. That is a real evidential role and a considerably more modest one.

Next

The reports considered here concern events within a life. The next article turns to the claim that something survives the end of one, and to the question of what would have to be true for the thing that survives to be the same person.