Atheism, Agnosticism, and Burden of Proof

“How can you be an atheist if you cannot disprove the existence of God?” This all-too-common question is often related to a misunderstanding of the concept of burden of proof, of how that concept relates to belief, and of how both of these ideas relate to the definitions of agnosticism and atheism.

Burden of Proof

You have probably heard the term “burden of proof” used in courtroom settings, often in the context of a criminal trial where the accused is innocent until proven guilty. What this means, of course, is that the accused’s innocence is assumed to be true, unless someone can actually prove otherwise. In other words, the accused’s innocence is the default position. As a result, it is absolutely not required for the accused to prove his innocence; he has only to show that, based on the prosecution’s case, there is no good reason to believe in his guilt; that the arguments and evidence presented by the prosecution are either unreliable, or do not make his guilt any more likely than some alternative explanation. Simply put, the burden of proof is on the prosecution.

Importantly, when a jury returns a finding of “not guilty” they are not saying that they believe the suspect is innocent beyond a reasonable doubt. They may have a truckload of doubt about his innocence. Their finding means only that reasonable doubt exists as to the suspect’s guilt. If there is such reasonable doubt and the burden is on the prosecution, then the jury is ethically and rationally required to acquit.

Why do we put the burden of proof on the prosecution? Because otherwise the prosecution’s job would be much too easy. For example, imagine you had to prove your innocence against the charge that you have the supernatural ability to cause cancer in humans anywhere in the world, and that you actually use this ability for your own sadistic pleasure, which is behind all the world’s cases of previously unexplained cancers. To make their case the prosecution puts you on the stand and asks, “Well then, if you aren’t guilty, how do you explain all the cases of mysterious cancers?” Helplessly, you admit that you can’t, to which the prosecution replies with an accusing finger, “Ah ha!” With your fear and frustration mounting you ask, “What makes you think it’s me!?” The prosecution immediately points out that Analogyland is not a country like the US: here you are guilty until proven innocent, and so the prosecution does not have to explain or prove anything. On the contrary, it is you that has to do the explaining. Nonetheless, the prosecution is feeling a bit generous and they volunteer that they are charging you based on their own psychic powers, powers that give them direct knowledge of evil people like you. When you ask what evidence they have that such a psychic sense is even reliable, they angrily warn you that your attempts to confuse the court will not be tolerated. The prosecution then reminds you once again that the only party who must present evidence is you, and that if you cannot prove the prosecution wrong, you are guilty by default.

Needless to say, you would be doomed in such a situation. In fact, in any land where your guilt is the default assumption, you would be doomed to a guilty verdict whenever the charge against you was unprovable, and the list of unprovable charges is limited only by one’s imagination (e.g., you are a witch who magically eliminates all evidence against her, or the reincarnation of Hitler pretending to be a good person, or an evil deity whose pretense at being a normal person is just what one would expect, etc., etc.). As you can imagine, in such a land you could easily send anyone you didn’t like to prison by dreaming up some unprovable claim; and, of course, they could do the same to you. One doesn’t have to look too far into history to find examples of this, such as the famous Salem Witch Trials.

To clarify the issue even more let’s leave the courtroom setting all together and just look at some off-the-wall factual claims that someone who comes knocking at your door might make. Let’s say someone, who calls himself Contactee Bob, knocks on your door and claims that leprechauns are real, but they are so smart and powerful they know how to avoid most human contact. However, Bob explains, leprechauns do make direct telepathic contact with a few deserving humans. When these leprechauns reveal their existence through such telepathic communication, the humans on the receiving end experience it as direct, revealed knowledge of the leprechauns’ existence. Bob claims to be such a contactee and hopes that one day you will be too. “Think positive leprechaun thoughts,” says Bob, “and you will be contacted too!”

At the end of his speech you ask Bob how he knows there aren’t alternative explanations of his “revealed knowledge” experience that are at least as probable as his leprechaun explanation (such as an obvious psychological one). Bob has heard this before, of course, and he’s ready: “Okay, Mr. Skeptic, prove there are no leprechauns.” Bob has even studied the contactee apologetic literature and takes his objection a step further: “If you claim that there are no leprechauns, then that implies you are omniscient since you would have to have knowledge of all parts of the universe to know that there are no leprechauns anywhere in it.”

This last statement of Bob’s would actually be true if the burden of proof were on you to disprove his claim, rather than on him to prove it. Bob has shifted the burden of proof onto you, the innocent bystander being subjected to his claims. He is arguing that his claim must be considered true by default unless you can disprove it, while all he has to do is watch you flounder in the attempt. What Bob probably doesn’t realize, however, is that in order for him to consistently hold such an interpretation of burden of proof, he would have to believe everything that he couldn’t disprove. While his interpretation of burden of proof lets in his leprechauns, it does so at the price of letting in an unimaginably large army of other bizarre creatures from the depths of human imagination and mythology. Consider the implications of this. Just to amuse yourself you could dream up claims of your own and present them back to Bob, claims like, “there are undetectable trans-dimensional hyper-intelligent fish all named, curiously enough, Wanda,” “There is an invisible, 3-headed dragon named Morris Minor who is a Douglas Adams fan and is responsible for manipulating the weather in such a way that it appears to be almost, but not quite, entirely unpredictable,” etc., etc., and Bob, by his understanding of burden of proof, would have to believe them all simply because he could not disprove any of them. This is what follows when the concept of burden of proof is misapplied or ignored. Clearly, this way madness lies.

What is the common element in the courtroom and Contactee Bob examples? It is that in both cases the burden of proof is not on the person making the positive claim. Now, by “positive” claim I mean any truth claim, such as “X is true” or “X is false.” Notice that saying something is false is also a positive truth claim: you are claiming that the assertion “X is false” is a true statement (i.e., you positively disbelieve X). For example, the claim “leprechauns do exist” is just as positive a claim as is “leprechauns do not exist.” Each is a claim to truth, and the burden of proof properly lies with the person making either claim.

But wait a minute. Doesn’t this take us right back to where we started? Doesn’t this mean that if I lack belief in Bob’s claims, then I have a burden of proof on me to prove him wrong just like he argued? Well, that depends on what you mean by “lack belief.”

Absence of Belief vs. Disbelief

A new born baby lacks belief in quite a bit, including the concept of God. Obviously, however, this is not the same thing as saying the baby disbelieves these claims (i.e., believes that these claims are false); the baby is merely absent belief—due in this case to its lack of awareness of the very concepts in question. Similarly, an adult who has grown up deep in an Amazonian jungle, and who has never even heard of people existing outside the jungle (let alone of such outsiders’ religions) is certainly absent belief in the Judeo-Christian God; however, once again, one could not say that this Amazonian regards God with positive disbelief, only that he is absent belief—due in the case to his never having been introduced to the idea.

In each of these examples, the individuals lack awareness of the concept of the Judeo-Christian God. Clearly, it makes no sense to say that they believe the claim “God exists” is false. On the other hand, it also makes no sense to say that they believe the claim is true. As a result, we cannot call them theists, and, depending on how one defines the term, we may not be able to call them atheists either (we’ll explore this a bit more later). But we can say that they are absent belief.

Note that in the preceding examples it also makes no sense to say that these individuals believe that the probability that God exists is around 50%. This leads us to a third sense in which one could be said to be absent belief in a claim. This occurs when there is at least some evidence for a claim, but this evidence is offset by equally strong evidence for the falsity of the claim or for alternative explanations. In other words, on balance, the claim is seen as somewhere around 50% probable. In such a case one would be rationally justified in “suspending judgment,” in making no commitment at all. If we’re talking about the claim that the Judeo-Christian God exists, then, as in the prior cases, we cannot call a person holding this view a theist.

One thing should be clear from the preceding discussion. We all start out at some point in our lives from the position of absence of belief due to absence of information. Clearly there is no burden of proof on individuals in this “position” (really a non-position). They are, as it were, simply waiting for input. It would be irrational for such people to claim to believe one way or the other. However, while you start out in life by justifiably saying, “the burden of proof is on you Mr. Positive Atheist and Ms. Theist,” as soon as attempts are made to meet that burden, the ball is right back in your court. You will need to assess the arguments and then assess the probabilities that the claim is true or false. Importantly, as the last of the earlier examples demonstrates, absence of belief may be the result of just such a reassessment: it may indicate not the default starting position of ignorance, but a reasoned, defensible conclusion that the odds of the claim’s being true is around 50% (not a view that I happen to share, by the way). In this special case, one is making a positive claim that the arguments and evidence support even odds.

Believing in Proportion to the Evidence

At this point all we have said is that one believes, disbelieves, or is absent belief in some claim. We’ve also suggested that this latter state is due either to ignorance of the claim (no burden of proof) or to an assessment that the probability of its being true is around 50% (carries a burden of proof).

This notion of probabilities of truth is much closer to the way science actually works than it is to the simplistic notions of absolute proof and absolute disproof, with everything in between treated as “unknown,” or “speculative.” Looking at truth values in this probabilistic fashion suggests a continuum; and assuming we believe only when we are rationally justified to do so, our “beliefs” cannot simply be “on” or “off.” Instead, our beliefs are a function of the probability that they correspond to factually true statements: the higher the probability, the more we believe; the lower the probability, the less we believe; and the lower still, the more we disbelieve.

So, what affects the probability that a claim is true? Certainly direct evidence, both for and against, directly affects it, including such things as the claim’s predictive success and explanatory power; but it is important to recognize that background knowledge is also an extremely relevant form of evidence. By “background knowledge” we mean the entire database of well-confirmed human knowledge.

Some claims, if true, would require that we throw out some of this well-confirmed background knowledge. To be sure, having to throw out cherished parts of our background knowledge is not unknown in history, but it occurred only when the evidence behind such a revolution was truly overwhelming. For example, if some theory is extremely well supported (say, at better than 99.9% as in the case of the Theory of the Atom), then any new claim that entails throwing out this part of our background knowledge needs to be supported by even better evidence than that supporting the existing theory. If weak or no evidence is offered, then should we remain absent belief in this new claim, as if the probability were around 50%? Definitely not. Why? Because the evidence against the claim is the evidence we already have for the background knowledge, which means the probability that the claim is true, in the absence of strong evidence, is extremely low.

There are also claims that, while not contradicting background knowledge, are completely unconnected to it, or go well beyond it in some radical way. But here again, the fact that a claim is “out of nowhere” does not make it false, though it does raise the evidentiary bar. Examples of this include certain notions of “supernatural” (specifically those that do not entail a logical contradiction as does the Judeo-Christian God). For example, while it doesn’t directly contradict our background knowledge, a claim that an intelligent, extraordinarily powerful alien somehow “started” the world 5 billion years ago is certainly a claim that is not in any sense a plausible extension of our background knowledge; it resolves no mysteries (other than replacing current mysteries with even bigger ones); and it lacks even suggestive support either directly, or indirectly from our background knowledge.

Now this latter situation is particularly interesting for a number of reasons: the claim is logically possible, it has no negative evidence against it, and it conflicts with no part of our background knowledge. On the other hand, it is has no evidence for it, not even the mild suggestive support that comes from its being plausibly related to, or extended from, our background knowledge. The question then is should we consider such an “out of nowhere” claim as having a 50% probability of being true? Should we say that it is as likely to be true as not? Definitely not. Why? Because the only thing such a claim has going for it is mere logical possibility. However, the list of the logically possible is effectively infinite. Since there is good reason to believe that the set of all factually true statements is vanishingly small compared with the set of all logically possible statements, then the odds are vanishingly small that some “out of the blue” logically possible statement would actually correspond to a fact of reality. Without some evidence supporting a claim there is no reason to believe that the odds favoring its factual truth are different than the odds favoring the factual truth of a randomly constructed logically possible statement, and the odds of that are effectively zero. Therefore, we have to conclude that an entirely unfounded claim—even in the absence of negative evidence—deserves not simply suspension of belief, but actual disbelief.


Of course, if a claim is not even logically possible, then all this talk about probabilities becomes moot. In such a case I can be absolutely certain (100% probability) that a claim is false—without any evidence at all. A claim is logically impossible when it is self-contradictory (also called incoherent). For example, if you claim that there exists a married bachelor or a triangular square, then the probability that you are wrong is 100%. Of course, these examples are obviously incoherent; however, many claims are no less incoherent, though this fact is far less obvious. That’s where careful analysis comes in. When such analysis shows that some claim entails a logical contradiction, then one can be sure that the claim is false—again, without introducing evidence at all. This kind of evidence-independent argument is referred to as an a priori argument: it is based on logic alone.

Belief vs. Knowledge

So far, we have been talking about rationally justified beliefs. To be sure, beliefs are often held without rational justification—even without the pretense of rational justification, such as when people talk about believing something “on faith.” This subject is dealt with elsewhere1, so I won’t repeat it here beyond noting that faith is not a justification, but an admission that there is no justification. If there are rational reasons to believe something, then “faith” is superfluous. If there are no rational reasons to believe something, then continued belief is, by definition, irrational. Clothing irrationalism with the more comfortable term “faith” changes nothing. Indeed, free of the restrictions of rational justification, I could have faith that the tooth fairy is real, or any of the world’s sundry gods, goddesses, wood nymphs, and other human imaginings. The problem is that there are no independently validated rules that govern the use of faith, which makes it a kind of blank check. As a result, faith allows in God—along with any and everything else. Reason is the only validated means we have to qualify a belief as actual knowledge, the only means we have to separate the “wheat from the chaff.”


Some so-called claims are not really claims at all. Claims made with meaningless terms amount to nothing more than a kind of noise. Examples include, “xonipboo loves you” and “myrssla exists.” There not only is no actual evidence for or against such statements, there is no conceivable evidence that would count for or against them; consequently, they are factually meaningless. Since such a claim asserts nothing, then it means nothing to say one believes that a factually meaningless claim is either true or false—to any degree at all. The subject of such statements does not even have a referent. If I say I believe in “it” (or don’t believe in “it”), then I literally cannot know what I am talking about—I literally have no idea just what it is that I think I believe (or disbelieve) in.

Defining Atheism and Agnosticism

Pulling all these ideas together, we can now make what I hope are more useful and more accurate definitions of atheism and agnosticism.

At the most literal level, atheism can be analyzed as follows: “a” means without, and “theism” means belief in God. So, literally “atheism” means without belief in God. “Theism” of course, means with belief in God. Now, one either believes in the Judeo-Christian God or one does not. If one does, then one is a theist; if one does not, then one is an atheist. Technically, then, the terms “theism,” and “atheism” are mutually exclusive and jointly exhaustive—there is no third alternative. Even the case of absence of belief falls under atheism, since absence of belief means absence of belief in God—that is, without belief in God. Since such persons are clearly not theists, they are (literally speaking) a-theists.

Of course, while literally true this definition does not really capture the typical modern usage of the term “atheism.” Indeed, under this definition all newborns would be considered atheists, which seems inconsistent with the way people usually understand the term. The most simplistic understanding of the terms “theism” and “atheism” in common usage is that atheists are absolutely sure that there is no God, while theists are absolutely sure that there is. Everyone else would be considered some flavor of agnostic. Interestingly, “agnosticism” once referred only to the view that some particular claim is unknowable. Under this definition many believers would be, in fact, agnostics, or more specifically, agnostic theists; they agree that God’s existence is unknowable, but they believe anyway on faith, while agnostic atheists reject faith as invalid, and so remain absent belief. So, under this view “agnosticism” refers only to one’s views about knowability, and not to one’s beliefs. However, this usage of the term “agnosticism” also doesn’t seem to reflect typical modern usage. Currently, the term seems to refer to the view that one neither accepts nor rejects God’s existence because God’s existence is unknowable. There is also the more general term, “negative atheism,” which refers only to the first part of that definition: that one can neither affirm nor deny God’s existence—for whatever reason. However, “agnosticism” is sometimes used in that more general sense as well, in which case it is synonymous with “negative atheism” (I use “agnosticism” in this latter sense—as synonymous with negative atheism). Negative atheism is the position that one should not believe, not that one should disbelieve. Positive atheism, on the other hand, argues that one should positively disbelieve—that is, one should believe that the claim, God exists, is false.

These more popular definitions leave a few gaps; for example, they all seem to implicitly assume that everyone is familiar with the subject matter and the arguments. If so, then what about our newborns, and those who have never been exposed to Judeo-Christian theism? Well, if they lack belief, which they do, then they are not theists; however, they do not disbelieve in God, so they are not positive atheists either. Therefore, they are a kind of negative atheist—but it’s the uncritical, uninformed kind and this is usually made clear just by qualifying one’s usage of the term with just such modifiers. Additionally, these popular definitions do not address whether the views are justified by appeal to reason. To be sure, people can be found in each of these categories whose beliefs have nothing to do with reason. For example, some ardent positive atheists have no rational justification for their views at all—perhaps because they are postmodernists or social constructionists, or because of political dogma, or because it’s part of the doctrine of some earth-spirit religion that they accept on faith, etc. Similarly, many theists openly reject reason, and even seem to find delight in saying that they believe despite flagrant logical contradictions—contradictions that they are all too happy to admit to. Where this distinction must be made—between appealing to reason and flouting it—the words “critical” or “rational” are often introduced, such as “critical positive atheist.”

If we take these definitions, and the caveats just mentioned, and couple them with our earlier discussion about believing in proportion to the evidence—that is, viewing “truth” as a probabilistic function of all relevant evidence and knowledge—then we can summarize everything into the following chart2:


Primary Term Other Common Terms Attitude Evidence Description
Positive Atheism Atheism Rejection Absolutely disprovable Logical impossibility
Positive Atheism Atheism Rejection Completely unfounded No direct or indirect support at all, even though there is no negative evidence
Positive Atheism Atheism Rejection Empirically disprovable Logically possible but odds are very close to 0
Negative Atheism Agnosticism Skepticism Empirically unlikely Odds are low
Negative Atheism Agnosticism Uncommitted Empirically indeterminate Substantial evidence, but about equally substantial evidence for alternatives
Negative Atheism Agnosticism None No conceivable evidence The claim in question is factually meaningless
Negative Atheism Agnosticism None Unaware of any relevant evidence Has never heard of the claim in question, or has never given it any thought.
Positive Theism Agnosticism / Pragmatic Theism Inclined Empirically likely Odds are high
Positive Theism Theism Acceptance Empirically provable Logically possible that’s it wrong, but odds are very close to 100
Positive Theism Theism Acceptance Absolutely provable Logical necessity



To tie our whole discussion together, let’s apply these ideas to the earlier Contactee Bob example. Contactee Bob doesn’t appear to be claiming something that is logically impossible, but he certainly hasn’t given us any reason to think that his claims are true; that is, he hasn’t given us any reason to suppose that his claims are more probably true than not; consequently, he has failed to meet the burden of proof. So, do I believe that Contactee Bob’s claim is true? No. Do I believe that Contactee Bob’s claim is definitely false? Technically, no, but I am not saying it’s a coin toss either, that it is as likely to be true as false. In the Contactee Bob example, we do have knowledge of background facts that bear on the likelihood of his claim being true. So, while there is no evidence supporting Bob’s claim, there is evidence supporting the alternative claim that Bob has deluded himself. We know, for example, that there are many supposed contactees for many types of strange beings, such as angels, aliens, various gods and goddesses, etc. As a group, these contactees report mutually contradictory experiences—even when the contactees are talking about the very same being. While this doesn’t conclusively prove that Bob is wrong, it is consistent with the alternative claim that all contactees of all of these “beings” have deluded themselves regarding the nature of their experiences. This background evidence, then, makes the competing self-delusion claim far more likely than Bob’s claim.

Bob also argued that it is self-contradictory to deny the existence of leprechauns because such denial implies omniscience; that is, we would need to have knowledge of all parts of the universe to know that there are no leprechauns anywhere in it. Our discussion up to this point should now make clear the fatal flaw in this argument: Bob assumes that absolute disproof is necessary to justify both absence of belief and disbelief. But, as we have seen, one can be absent belief or even disbelieve in the existence of leprechauns without at the same time claiming absolute knowledge that leprechauns do not exist. One should be at least absent belief based on Bob’s having failed to meet his burden of proof. But, when one also considers how this claim fits with our background knowledge and other known facts about claims of this type, then one can go further and have very good reason to suppose that Bob’s claims are extremely improbable, and therefore, to positively disbelieve them.

So, whether people claim that Santa Clause, telepathic space aliens, or God exists, the burden of proof is on them. None of these claims becomes true by default; that is, they do not become true on the grounds that they cannot be disproved. If those making a claim do so in a logically consistent way but cannot meet their burden of proof, then absence of belief is the best that can be rationally achieved. However, depending on other relevant facts (such as background knowledge), we may have good reason to suppose that the claim, while not absolutely false, is probably false in the extreme, thereby making positive disbelief the only rational option. Moreover, if they make a claim in a self-contradictory way, then we can be 100% sure that they are mistaken even without any evidence at all. Finally, there is also another way in which people may make their claims: they can use factually meaningless terms.

It is interesting to note that if one successfully demonstrates that religious terms like “God” are factually meaningless, then one cannot be a positive atheist, since to positively claim that God does not exist implies that the word “God” has meaning. I take the approach, along with Michael Martin3, that one can argue that religious terms are meaningless, and then go on to argue that if we assume, for the sake of argument, that the terms are meaningful, then contradictions immediately follow. So, to the extent I am successful with the first part, I am a negative atheist, and to the extent I am wrong in the first part but right in the second, I am a positive atheist.

1 P. Wesley Edwards, “Reason vs. Faith,” FreethoughtDebater, <>, 2004.

2 This chart is influenced by something similar in Michael Scriven, “God and Reason,” in Critiques of God: Making the Case Against Belief in God (Amherst, NY: Prometheus Books, 1997), p.109.

3 An excellent, if somewhat advanced, overview of these issues is Michael Martin, Atheism: A Philosophical Justification, (Philadelphia: Temple University Press, 1990). A classic, and very easy to read work that hits the same points is George H. Smith’s, Atheism: The Case Against God, (Amherst: Prometheus Books, 1989).