From the highest upvoted answer on Is any aspect of the supernatural testable? What level of proof is possible for the supernatural?:

However, you are probably wasting your time on the various hobgoblins and eerie powers you list. We do not see such phenomena, werewolves, resurrections, or mind controls, requiring scientific explanation. One must first demonstrate them publicly and repeatedly, which in most cases defies their very definition.

As Hume pointed out, what we mean by a "miracle" is precisely that which cannot be predicted, controlled, or experimentally repeated. We now fly, cure plague, transmit voices over miles, and perform other formerly "miraculous" deeds. But they are no longer "miracles," except metaphorically.

(emphasis mine)

The requirement that to demonstrate that something may happen or has happened (especially a miracle) requires public and repetitive demonstration of it happening strikes me as odd.

There are lots of things that are accepted to have happened even though they cannot be demonstrated repeatedly. It would seem to me that not even rationalists consistently stick to the requirement of repeated demonstration:

  • Court room evidence. It would be absurd to require that a person accused of murder commits this act consistently, repeatedly and publicly in circumstances resembling the circumstances of the alleged already committed crime before declaring him guilty. Instead even rationalists accept the validity of abductive reasoning in such cases in place of experimentation. We collect observations of the current state of affairs and through examination of such observations we conclude that the most likely reason of the presence of these observations is that the suspect has indeed committed murder. We do NOT require public, repeated demonstration of the suspect committing murder.
  • This also seems the only available way to practice history. In most cases we cannot conduct experiments that would verify historical claims. For example, we cannot go back to year 814 and produce a living Charlemagne to demonstrate the possibility that Charlemagne died in 814.
  • Less absurdly, it would seem to me that we cannot experimentally prove evolution either. Evolution happens too slowly, we do not have that much time. We may have plenty of evidence that evolution has been and is going on, but NOT repeated demonstration of it happening. Yet, it is irrational to disbelief the theory of evolution, not the other way around.

It would seem to me that rationalists dropped the requirement of repeated demonstration in all cases BUT in the case of miracles, where they still require repeated, experimental demonstration, even though - as the answer I quoted points out - the very definition of a miracle says we cannot experimentally test it.

It would seem to me that perhaps this sort of argument conflates 'has happened' with 'does happen'. Courtroom evidence may strongly suggest that the suspect HAS committed murder, while repeated demonstration in experiments could show that the suspect has an ongoing inclination to commit ever more murders - these are two different things. Same for Charlemagne's death in 814. (Evolution is even trickier, because we accept from existing evidence other than repeated demonstration that it both has been happening and is happening.)

Based on the above, may I ask why is it required for a miracle to be repeatedly demonstrated through experimentation in order to accept the possibility of it happening? We cannot provoke a miracle to happen, but we may ABDUCTIVELY, rather than experimentally, conclude that it may have happened in the history.

Resurrections for example, since the answer I quoted mentioned them. As I understand, those like N.T. Wright or Gary Habermas attempt to employ historical, abductive reasoning to conclude that Jesus has likely risen from the dead. It is not my intention to ask here whether the conclusion is warranted or not. Rather, I'd like to ask if this methodology is the correct one to approach such a problem. If I understand the answer I quoted above, it is not - we cannot accept resurrections until we repeatedly, experimentally demonstrate people rising from the dead. And yet even rationalists are happy to employ such reasoning to demonstrate the existence of Jesus.

Are miracles somehow different? When must we demand repeated, public demonstration, and when is abductive reasoning satisfactory?


3 Answers 3


When it comes to inductive reasoning (and based on how I define induction it includes abduction), I’m a convinced Bayesian. I'll offer an answer from that perspective.

Given a piece of evidence, E, my credence in an explanation, H, will be a function of its prior probability, P(H), and its explanatory power relative to alternative explanations (that is, the ratio between P(E|H) to P(E|not-H)). We plug these values (not usually exact values, but “feelings” that can perhaps be mapped to intervals, or upper/lower bounds on values) into Bayes’ theorem and derive the posterior probability of H given E, P(H|E). You’re quite right to say that repeatability isn’t necessary to reason from a piece of evidence to a particular conclusion. We do it quite well all the time, even in science. But in the case of a very surprising occurrence such as a purported miracle, it is really really nice if we can repeat it. We want to be sure that something miraculous is really happening, as opposed to something more mundane such as events being exaggerated or misreported. And, miracles aren't the type of things that we can reproduce on demand.

This isn't a problem per se, but it means that your willingness to accept a supernatural explanation will depend on your prior credence in the existence of the supernatural. This brings us to a second difficulty. Supernatural explanations usually have a very low prior probability (for various reasons, such as the entities involved in such explanations often have no prior evidence for their existence). However, if you already have a reasonably high prior credence in the belief of supernatural entities (maybe you’ve encountered one, or maybe some philosophical arguments convince you that God exists) then it may be rational to accept a supernatural explanation if it's explanatory power is great enough. Then, as you say, it’s a matter of discrediting alternative naturalistic explanations.

Since you mentioned the resurrection of Jesus, there is a debate on YouTube between William Lane Craig and Bart Ehrman on just this topic where the Bayesian method comes up. Craig makes the point (similar to the one above) that your whether you accept supernatural explanations will depend crucially on the prior likelihood you have in believing in the supernatural in the first place. Ehrman makes the point that as a historian he is committed to methodological naturalism and so the prior probability of any supernatural explanation must be 0 (on methodological naturalism).

  • Thank you for your answer and for your link to the debate. Since it's very late now I'll check it out later BUT there's one thing that surprises me... "Ehrman makes the point that as a historian he is committed to methodological naturalism and so the prior probability of any supernatural explanation must be 0 (on methodological naturalism)." If we put this into Bayes' theory than this will necessarily mean that NO evidence could convince a methodological naturalist that something supernatural may have happened.
    – gaazkam
    Commented Feb 19, 2020 at 0:22
  • 2
    God may literally speak to a methodological naturalist in person and raise his grandfather from the dead in front of his very eyes and a methodological naturalist will still say 'I can't see anything supernatural here?' Because even 100% * 0% = 0%. I can accept that 'extraordinary claims require extraordinary evidence' but not that NO evidence is sufficient for extraordinary claims. I find it dubious to presuppose the LACK of existence of anything supernatural, if for no other reason then because it seems unwarranted to claim that only this exists what we can observe.
    – gaazkam
    Commented Feb 19, 2020 at 0:32
  • @gaazkam It's been a while since I watched the debate and I don't want to misrepresent Ehrman, but I believe that this is what he says. Craig rebuts by saying something like: okay, maybe as a historian committed to methodological naturalism, you can't accept the resurrection at your day job. But according to more general norms of reasoning that don't presuppose naturalism, you can and should. Commented Feb 19, 2020 at 0:36
  • 1
    @gaazkam Yes I'd agree with what you say. I don't know what to think about the resurrection itself, since I'm not knowledgeable enough about the history. But with regards to this point about method, I believe Craig's argument is more convincing. Commented Feb 19, 2020 at 0:43
  • More convincing than Ehrman's point, I mean. (Just wanted to clarify, since I reread my previous comment and realized it was a little unclear.) Commented Feb 19, 2020 at 3:24

This question conflates two separate and distinct issues:

  • The existence of an event (a singular phenomenon): something that someone encounters as an experience and records in the stream of intellectual and social memory.
  • The existence of a class of events (a normative category): a pattern or type of event that occurs often enough to give it its own conceptual structure above and beyond the specific events.

For instance, if we hold a ball in the air and let it go, we see an event that we call 'falling'. In fact, we see enough things 'fall' — a pattern of behavior — to create an separate concept called 'falling' that we can talk about without regard to any particular thing that might fall. Further, we can analyze that concept and pattern to produce ideas like Newton's law of gravity, which serves as a tool to analyze that perceptual class we call 'falling'. Now imagine if that were not the case: that nothing in the world ever fell except for maybe that one time when an apple hit Newton on the head. No class of events, no pattern of behavior, no general law; just some guy named Isaac that we roll our eyes because he keeps talking about this bizarre 'falling' idea (as though anyone could really believes he saw an apple 'fall').

So the difficulty here is that when we talk about 'miracles' or any other metaphysical event, we generally fail to make that distinction. If someone says "I believe in miracles" they almost never clarify whether they mean:

  • "I believe that the specific events recorded in collective knowledge actually occurred", or
  • "I believe that the specific events recorded in collective knowledge represent a class of behavior that always occurs"

And why would anyone make that clarification? That is so much over-thinking that it hurts my head laying it out.

This gets us to the demand for repeated public demonstration. In the scientific model it is not sufficient merely to say: "Hey, that happened." Even if one has clear irrefutable proof that that did indeed happen, something happening once is not part of a phenomenal pattern, and very few conclusions can be drawn from it.

In this sense we can define a 'miracle' as an event that some person or group experienced, which could not be explained by any (current) phenomenal theory, and cannot be experienced with enough consistency or regularity to create a class of events, but which nonetheless created some collective effect.

The problem — the point of contention — is actually a matter of dueling syllogistic fallacies. People who believe in miracles generally want to extend their belief in singular historical events to a belief in a class of such events (the illicit minor fallacy); people who pooh-pooh miracles generally try to extend their rejection of any class of such events to a rejection of any singular historical events (the illicit major fallacy). The whole conversation is so politically charged that these fallacies are (for the most part) irresolvable.

  • It is like Alice in Wonderland: on a space station, falling would be the miracle. But from earth, being on a space station is the miracle.
    – Scott Rowe
    Commented Oct 20, 2023 at 1:41

The point is that most of the examples you quote are not in themselves hard to believe.

Charlemagne died in 814- well, he was mortal, and died around that time, and there is more than one source that said he died in 814, and anyway, what's the problem if it was 815 instead?

Maybe the defendant committed murder- sure, why not, people commit murders, why should the defendant be any different? What's hard to believe about it?

Evolution- sure, that's plausible. Some organisms fare better than others, and therefore are more likely to produce.

All those examples are inherently plausible because they are specific instances of commonplace happenings that are not in fundamental conflict with our understanding of the laws of physics. The fact that murders happen all the time makes the possibility of another murder get an automatic tick in the plausible box. Miracles, on the other hand, are by definition extraordinary to the greatest degree, so they can't be given the automatic plausibility tick that we give to specific instances of commonplace events.

Rational humans have a tendency to be more inclined to credit something if there is a reliable and valid explanation for it. The traditional way to develop reliable and valid explanations is the scientific method- you create a hypothesis about how a class of events happens and you put it to the test, usually through repeated experiments. We went through that process with special relativity, general relativity, quantum theory, the Higgs Boson, black holes, and a host of other theories, so why should miracles get a free pass?

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .