If one is not able to come up with a reason to believe anything supernatural, should one “by default” assume naturalism to be true? Or should one remain completely agnostic?

If I see a person hitting a baseball, should I by default assume that that’s all that happened, or should I stay completely agnostic with respect to other hypotheses such as a demon possessing that person and causing him to hit the baseball?

If so, does that also make atheism the null hypothesis? If one is not able to find evidence that god exists, should one by default assume he doesn’t?

  • Watch the news! 🙂
    – Hudjefa
    Commented Aug 28, 2023 at 3:04
  • The better question is whether you should stay completely agnostic to the hypothesis of whether it was hit at all, or whether e.g. an otherwise-undetectable entity moved the ball, and just made it seem like it was hit. Because that would kind of break the idea of observable cause-and-effect, which is the rather far-reaching consequence of accepting the existence of "the supernatural" that is not often mentioned. If you remain completely agnostic, then you presumably need to be largely or entirely agnostic about the reliability of observable cause-and-effect.
    – NotThatGuy
    Commented Aug 28, 2023 at 4:36
  • Meta answer: The key word is usefulness: The point of a brain and of the scientific method is to discover what is most useful in predicting the future, to be able to decide what action to take to get oneself closer to the (arbitrary) own goals. … Belief is useless because it is not based on observed reality, so one will make inferior/harmful decisions. And agnosticism is useless for the same reason, and merely leaves you unable to make predictions at all instead. … Existence is defined as being experiencable for that very reason. … (Absolute) truth is not a useful concept.
    – anon
    Commented Aug 28, 2023 at 10:23
  • 1
    like, no offence but what does it matter if we call it a null hypothesis? are you making any new claim based on reasoning with 'statistics'? if so, i'm sure it's been discussed by someone, and you may want to google it.
    – user67521
    Commented Aug 28, 2023 at 14:52
  • 1
    I don't think the term "null hypothesis" is anywhere used in this way. It's originally a statistical term, but what you call "null hypothesis" here is certainly not a statistical null hypothesis (which would be a probability model). Also in statistics, certainly the null hypothesis isn't something that you should believe "by default". Commented Aug 30, 2023 at 0:23

5 Answers 5


There is no "the" null hypothesis. Every claim that can have a null hypothesis is:

"X perturbations of initial conditions make more likely Y end conditions"

and every null hypothesis is:

"Y end conditions are present without X perturbation of initial conditions."

This is why when people try to find out things about reality, they use measurable or previously measured control groups, rather than the predictions of the very models that they're trying to test.

It's trivial to construct a model for which a rebuttal of the relevant scientific consensus is the null hypothesis - just pack the naturalistic cause into the perturbation of the initial conditions and state the naturalistic outcome as more likely end condition.

"People who conduct the following experiment: Consume potion of flying then jump off of the roof, and who didn't start out with injuries, are more likely to be measured with injuries after a few seconds than people who don't."

The null hypothesis here is "Uninjured people who consume potions of flying and jump off of rooftops are no more likely to be measured with injuries after a few seconds than people who don't do this."

Measurement strongly favors the model over the null hypothesis when you do this, obviously, but that doesn't make the null hypothesis less of a null hypothesis.

  • Note the philosophical implications: the null hypothesis is supposed to deny cause and effect.
    – Joshua
    Commented Aug 28, 2023 at 20:53

I don't think "the null hypothesis" is right analogy here (life is not an experiment), but I would answer the implicit question in the affirmative. All assertions have a burden of proof.

If I see a person hitting a baseball, should I by default assume that that’s all that happened

The burden of proof is on the person claiming otherwise. If there is evidence of demonic possession, that evidence might need to be considered, but if it is not based on any particular evidence then the claim is unreasonable.

If one is not able to find evidence that god exists, should one by default assume he doesn’t?

Here Bertand Russel makes a similar argument. The existence of God is not empirically falsifiable, and the burden of proof should therefore fall on those making such assertions. Opponents of Russel's view done necessarily reject that premise so much as they assert there is in fact some kind of evidence for God's existence.


David Hume wrote: “A wise man apportions his beliefs to the evidence.” but something similar applies to prior beliefs. There is no need to choose between assuming naturalism is true or that the supernatural exists. Instead we can consider how likely both of those things are a-priori. A good start is consilience - we have a good understanding of the capabilities of the human body, and basic physics of objects, so the idea that someone could hit a baseball is consilient with a lot of other things we know. On the other hand, we don't have a coherent theory of demons, or a great deal of unequivocal observations of demons doing things. Instead why not assume it is most likely that the ball was propelled by the baseball player, whilst still holding some non-zero belief in the possibility of it being because of a demon.

Certain knowledge of these things is not available to us from purely observational evidence, so our views will be partially based on our prior beliefs, which we need to state, and which might reasonably differ from one person to another. So there is no right answer to the question. Better to be content with uncertain knowledge - it is the best we can do.

If so, does that also make atheism the null hypothesis? If one is not able to find evidence that god exists, should one by default assume he doesn’t?

AFAICS this sort of discussion is likely just to end up with people reasserting their priors, which doesn't actually tell us anything about whether god exists.

HOWEVER, if you are going to make this argument pseudo-statistical in appearance, then when using a null hypothesis statistical test, the null hypothesis should be the opposite of your research hypothesis, i.e. the thing you don't want to be true. Arguing in favour the null hypothesis is a classical statistical malpractice that is sadly very common in statistics.

  • If 2 people have different prior beliefs, that implies that at least one of them are wrong. Rather than throwing one's hands up and saying there's no right answer, might it not be better to apply scrutiny to one's prior beliefs, rather than assuming what your parents told you while you were growing up is necessarily and unquestionably true? The "prior beliefs" position might also not work that well given the existence of people who either lose or gain religion later in life, because this suggests that there are potentially ways to resolve a difference of prior beliefs.
    – NotThatGuy
    Commented Aug 28, 2023 at 13:28
  • 1
    @NotThatGuy My point is exactly that we need to apply scrutiny to our prior beliefs as that is what largely governs our conclusions in situations where we have little (or weak) evidence. The point is that we shouldn't start with a conclusive either/or prior, but instead to consider the relative strengths of prior positions and propagate that through to our posterior beliefs. It isn't a choice of prior, it is a distribution of belief across priors. My whole point is that nothing is necessarily or unquestionably true either way. Commented Aug 28, 2023 at 13:33
  • 1
    "because this suggests that there are potentially ways to resolve a difference of prior beliefs. " that is a non-sequitur, it just means that in the absence of evidence you have changed your mind about your priors, it doesn't mean your new prior is better than the old one or that anything has been "resolved". The problem is that we don't know which prior position is "right", if we did, there would be no need for evidence. All we can do is set out (and maybe justify) our priors and if possible quantify them and then see how evidence much affect that. Commented Aug 28, 2023 at 13:41
  • 1
    @Galen it is possible to argue for the null in NHSTs, but it requires an analysis of the statistical power of the test, which is generally rather more difficult than the normal quasi-Fisher style test that does not involve the research hypothesis. It rather depends on how self-skeptical the test is required to be. The real problem with NHSTs is that the p-value is not the probability the null hypothesis is true, but many uses of NHSTs rather imply that it is. A common example is climate skeptics saying that there has been no significant warming since [date], there they want to argue that Commented Aug 30, 2023 at 6:54
  • 1
    ... no harm comes from fossil fuel emissions. However the statistical power of the test is always far too low to draw any conclusions from the data (the reason the warming is non-significant is not that the warming trend isn't there, just that there is too little data to draw any conclusion). NHSTs are non-symmetric, if they performed the test with warming as the null hypothesis, that would give a non-significant result as well. Not necessarily malpractice, it would depend on the situation, but the "devils advocate" null hypothesis is less risky. Commented Aug 30, 2023 at 6:56

Taking this as a legitimate question, rather than as an argument in the form of a question:

To address your first question, terms such as "default" and "null hypothesis" apply only when there is a pre-defined, agreed-upon existing framework of inquiry. ("Null hypothesis" in particular is a technical term with a precise meaning that isn't a good fit here, so we'll focus instead on "default.") Something can only be the default in relationship to a shared expectation.

Naturalism is a specific philosophical stance. It is the "default" for many people across the world today, but not for many other people. Asking if something is the default is the equivalent of asking "is this what we all agree on as a baseline expectation?" The answer might be "yes" in some rooms, but not in a room full of philosophers, who tend to not have shared default expectations almost by definition.

The second and third questions outline an argument, and the problem with that argument is "false analogy," which is where a conclusion overreaches the points of similarity in the supposedly analogous situation. In this case, the suggested argument asks us to take the simple, observable physics of a single moment in a baseball game as an appropriate analogy for the experience of life in the universe at large.

  • 1
    The problem is that people have failed to produce even one single verifiable non-natural explanation for anything (also, what it even means for something to be "non-natural" is vague, at best). Unless and until that changes, if you care at all about believing things most likely to be true, naturalism is the most reasonable option. We don't need natural explanations for literally everything to justify naturalism (if ever there were an unreasonable request, that would be it). Rather, naturalism is where you end up when you reject unverifiable things.
    – NotThatGuy
    Commented Aug 29, 2023 at 7:52
  • 1
    For an anti-naturalist, such as myself, it's the equivalent of looking for your keys under the streetlight. Sure, you can see much better there. But you'll never find your keys there, if that's not where you lost them. // To extend the metaphor, I can appreciate the bright light naturalist inquiry shines on many things. But those are not the entirety of the things I care about or find important. Commented Aug 29, 2023 at 12:24
  • 1
    I don't think any prominent naturalist would consider naturalism to be "a universal position". It's hard to deny that the majority of the world's population believes in one deity or another. But that doesn't mean they believe in those for good reason. Verification is one of the few/only reliable ways we have to evaluate whether something is true. If you have another means of evaluating truth, that can demonstrate to be reliable, that could work too. But no-one has produced this for the supernatural. With no reliable method, you can't reasonably have confidence that what you believe is true.
    – NotThatGuy
    Commented Aug 29, 2023 at 12:55
  • Let us continue this discussion in chat. Commented Aug 29, 2023 at 15:01
  • 1
    @NotThatGuy QM is fundamentally different. Really. We may be wrong (in fact, we always are :-) ) BUT see my above statement re Copenhagen definition DEMANDING asking is bad, and many worlds ... agh!!! . Lots has changed in the last 100 years, and since 15 Feb 1988 (Feynman died - but also, nothing has changed. NO other 'theory' demands itself to be unknowable as part of its fundamental structure. ||We accept QM's faster than light aspects without demur. We accept Quantum Entanglements spooky" impossible-using-all-other-Physics without concern. || Commented Aug 30, 2023 at 10:10

Remember the difference between 'concepts' and 'semantics' [paradoxically applies to this statement too].

I prefer the word SUPRAnatural, as in 'above/beyond' the natural.

IF we were to dig deep enough and find that the first subatomic particle in the first neuron that fires in the brain of the individual hitting the ball gets activated by a 'demon'. IF that somehow can be measured, and above all: it OCCURS, as in HAS AN EFFECT in the natural world which is governed by the word 'naturalism', to me, the question then becomes, how can it then be 'supernatural'/'supernatural'?

In most of these situations it's a question of what we perhaps currently know/are able to understand/able to measure. It is less a question of certain phenomena existing beyond EXISTING scientific laws, that we currently cannot grasp. See Schrödinger's book What is Life, chapter 7.

In other words, it's more a question of semantics, and as EVI1M4chine comments above about USEFULNESS. A hypothesis is an EDUCATED guess. Science tries to move deeper and deeper to fundamental causes of effects, like peeling off a layer of onion. The whole, complete answer is impossible to provide at anyone point, then it's possible that the whole universe of all its components would have to be described at the very point in time that the bat hit the ball.

Point being: What would even 'super-/supra-natural' be? Scientific laws evolve with observations and growing knowledge. Anything observable in the natural world by natural minds and sense ought to be 'intra-natural' by default, just perhaps not currently explainable.

You must log in to answer this question.