« Ancient Myths vs. the Bible | Main | Study with Greg in France »

March 21, 2013


There are millions of people murdered every year, and most of them by family or someone they knew, so I don't really see how this would be in anyway extraordinary, so the analogy seems somewhat flawed. If we had proof of millions of worlds created by millions of gods, no one would probably question that this is the case for our world too.

"most of them by family or someone they knew"

Can you show what data you have to back up this claim?

Darth Dutch


The UN did a global study of murder and published the data last year. The total number of murders worldwide was about 500,000. So no, millions of people are not apparently murdered every year. In addition, a Department of Justice report published in 1994 used murders that were prosecuted in 1988 as their sample. They report that 16% of those murders were alleged to be conducted by family members. So again, no, most murders are not conducted by family members as you proposed.

So, it appears as if Mr. Wallace is correct, that murder itself is extremely rare, and those conducted by family members even more rare. So the description of being "extraordinary" seems apropos.

Okay, so I botched the number on actual homicides (about 500.000 seems to be the estimate, not counting deaths in war and conflict). As for the perpetrator, unfortunately the link has died but U.S Department of Justice estimates that women are at a significantly higher risk to be killed by a family member.


So while my numbers were off, I still stand by my original point that there is nothing extraordinary on being killed by member of the family, especially for women, as hundreds of thousands of women have shared the same fate.

Unfortunately this data is not very recent, but DOJ estimate is that 33% of women were killed by intimate partner.


@ Erkki
An "intimate partner" is not a husband and is not family. Marriage changes things and hence the rareness or extraordinary as JWW speaks to...


I still don't see the "extraordinary". No matter how you interpret the numbers, there are thousands of confirmed cases of husband killing the wife. I don't know the particulars of the example case, but I assume that they had some reason to believe to husband did it. Maybe a conflicting story. And as there are thousands of murders, I don't find it particularly unbelievable that there exists a case where there is no material evidence, but the guy still did it.


Let's set aside the fact that you woefully overestimated the number of murders from millions to half a million, and changed your statement of "most" to 33% - still a significant number and all tragic, I agree. But you still try to dismiss it even when your numbers were way off base.

I think you're missing the main thrust of the article. It wasn't simply that a wife was killed by her husband. It was that he was convicted without the usual things present in a murder trial - a body, for example. It was that fact that there was no physical evidence which tied him to the murder. It was that he vehemently denied it up until the point that he then confessed. This was, as JWW states, a highly unique case for those reasons.

Darth Dutch


If a realtor told you that murder was a customary (i.e. not extraordinary) event in the neighborhood where you were house shopping, would you simply reply, “Yes, murder is not an extraordinary event at all. How much for the house?”

If not, what would be the ideal word you would want to hear from the realtor about the frequency of murder?

I have no doubt that this was an very unique case. I just don't see it relevance as an analogy. There are thousands of murders happening and thousands of cases, so to have one where there is very little physical evidence, not even a body, and the accused still confesses and is found guilty, does not seem anyway unlikely or impossible. In other cases, there could be mountain of evidence and the accused was still not guilty. Rarity itself does not equal "extraordinary". There is extremely low chance of winning in lottery, but there is nothing extraordinary in lottery winners, as there are millions of lottery tickets.

It seems to me that Erkki's basic argument stands even if the numbers were off.

The rub seems to come in how Wallace used the word extraordinary. Does extraordinary mean "rare" or "totally unique"? Could Wallace define exactly what he means by "extraordinary" for me please?

Murder, while rare compared to many other events, is still a well-attested and documented phenomenon, although the details of each instance may be unique. However, when speaking of the origin/source of the universe, we have nothing to compare it to because this universe is the only one in our experience. Even if there is a multiverse, we have no experience of that and cannot directly measure it. Wallace needs to defend his jump from cases of murder to the origin of the universe in his discussion of "extraordinary."

These models of cosmology are based on data. True the data requires interpretation. But the data does constrain those interpretations. Some make predictions, which have or can be tested. If these predictions come true, this provides powerful evidence for that model.

In our everyday lives most do follow the statement "extraordinary claims require extraordinary evidence." Perhaps this statement does not apply in every case, but Wallace has not provided a convincing argument for disregarding it in the case of the origin of the universe.
As an aside, I don't think cosmology will ever provide proof for God. This is not to say there is no God. It only means that it is the wrong tool to look for God.

Erkki and especially Caleb seem to have it pretty well. I will add that the phrase "extraordinary claims require extraordinary evidence" is extremely imprecise at best, and so I don't recommend using it. Nevertheless, it does communicate that there is something intuitively different about the existence of God or the occurrence of miracles on one hand, and the occurrence of natural phenomena (even relatively rare natural phenomena) on the other. We might not be able to nail down exactly what the difference is, but I think it's fair to say that we all understand (or should) that there is a difference, and moreover that this difference is relevant to how we must prove that one versus the other sort of thing has occurred.

Can someone please answer these questions?

1. What makes, for example, the claim that Christ is raised extraordinary?

2. What is the evidence that it requires?

3. What makes that evidence extraordinary?

I just don't buy the "extraordinary claims require extraordinary evidence" statement. Extraordinary claims require adequate evidence, just as does everything else.

BTW-I am not surprised that our friend J would botch an analogy. He's done it before. His love of CSI analogies derails him every time.


Yeah, I think that the "extraordinary claims require extraordinary evidence" is just another one of those slogans.

A charitable interpretation is that if a proposition has low prior probability, then it's going to take a lot of evidence to raise the probability to the point that it's something it makes sense to believe. In practice, I think atheists use it to reject all evidence.

    I just don't buy the "extraordinary claims require extraordinary evidence" statement. Extraordinary claims require adequate evidence, just as does everything else.

I think the whole point is precisely that on extraordinary claims, "adequate evidence" would have to pretty extraordinary. Basically, how I would describe it is that more fantastic and out of ordinary the claim, the more solid evidence it requires, because otherwise Occam's Razor just kicks in and more mundane explanation becomes likely.

For example: we find a murdered person. We can assume he was killed by A) human B) demon C) extra-terrestrial. We have zero evidence who has killed him. Are we just as rational in picking any choice from A, B, and C as the likely murderer? No, because we know from experience that humans have killed people before and we know the mechanics of how and why humans kills humans, as well as methods to prove it. We have zero experience from people being murdered by demons or extra-terrestrials. It does not prove that human killed him, but it creates a rational assumption.

We have zero evidence who has killed him.
Isn't this evidence?
we know from experience that humans have killed people before and we know the mechanics of how and why humans kills humans
We might add that compared to the evidence we have that humans exist, our evidence for demons and ETs is thin. If we really had zero evidence, I don't think the assumption that it was a human murderer would be any more likely than the demon or ET. Of course, if we really had zero evidence, I don't think we'd even be able to say that a murder occurred.


"A charitable interpretation is that if a proposition has low prior probability, then it's going to take a lot of evidence to raise the probability to the point that it's something it makes sense to believe. In practice, I think atheists use it to reject all evidence."

I'm on board with you and your take. I think we can refine this a bit though. If you have evidence of sufficient quality, it is not necessary to have a whole boatload of it. It is just a tricky call to know what the right balance is between quality and quantity and if it is adequate for most reasonable folks to agree that the balance has been met.

Erkki S.

I can appreciate your line of reasoning, but since so much hinges on the nature and quality, as well as quantity, of evidence in a specific case of murder, making broad generalizations are nothing more than a starting point of educating yourself on how to go about an investigation. When the boots hit the ground and you are at the murder scene, there will be plenty of evidence to examine and consider and then discover the cause and effect that resulted in the existence of that evidence of that type in that location. The nature of the evidence itself will lead you to the most likely source of it, while the outline you describe will be only a very loose guideline in evidence discovery and analysis.

I think that evidence is always there, but that does not mean that it might not be missed or somehow compromised. I'm not sure I buy a crime without any evidence. That might speak more to the quality of the investigators in the case.

Look, even a stab wound is evidence.

Let's all keep in mind that a lot of what J. Warner was working on was old cold cases. They come in certain very limited flavors and the field of murder investigation is much broader than that.

Well, naturally my example was a very crude demonstration of the basic principle. Obviously any real life scenario would be a lot more complicated and would have some real evidence for the perpetrator. I think the whole underlying problem with Mr. Wallaces "stick" of treating the divinity of Jesus (if I have understood this correctly, I have not read his book) as a some form of cold case investigation is that, while interesting thought exam, it seems like proving who has killed who is very different sort of endeavour from proving whether someone is actually the Messiah, or more closely to the topic of OP, who created the universe, or if the universe is even created at all.

Erkki S.

When it comes to the cosmological argument, I think that you have a point that it is not definitive that the universe was actually created. But then, it does not claim that for itself to start with. It merely makes the claim that the universe having a beginning must have had a cause and that rests on a constellation of evidence that things coming into existence have always, in our shared experience, had a cause. The only thing that the cosmological argument does is point in a kind of general direction and god is certainly within that general direction. So, the quality of the evidence in this case, is not the best. As a result, we must rely on more evidence from other sources or maybe also look to better quality evidence in those other sources. I think that there is enough evidence there and of a type of quality that makes a compelling case for a creator behind what we see. But that is taking multiple sources of evidence into account.

The OP hints at the crux of the matter but misses it. Choosing a Californian at random, apparently, you'd have a .000004% chance of having the murderer at this case.

That number, 0.00000004, could be taken as your Bayesian prior probability of guilt. It's probably a bad choice. But we'll come back to that.

What does it mean to say the verdict was rational?

Let's take 'beyond a reasonable doubt' to mean 0.99 probability of guilt.

Bayes's theorem tells us how to evaluate a piece of evidence: multiply the prior probability by a factor.

That factor is a ratio.

The denominator, in this case, is the probability of the evidence presented in court regardless of the guilt of the defendant.

The numerator is the probability of the evidence presented in the trial given the guilt of the defendant.

The evidence, then, was sufficient to update the jury's estimate of guilt by a factor of 0.99/0.00000004 = 24,750,000.

The jury, were they following Bayes's theorem, would have to say they felt that the evidence presented in the case was much Much MUCH more likely to be in place if the defendant was guilty than if he were not. 25 million times more likely.

The video doesn't claim to present the evidence exhaustively. The defendant told multiple stories about the night of his wife's disappearance. These were not compatible with each other. He also made what seemed to be at least one 'Freudian slip'. Maybe I missed some factors. Maybe the jury saw something important the video left out.

But, given what the video presents, the evidence doesn't seem to justify a factor of 25 million.

Getting back to the prior, if you take into account that spouses account for a disproportionate number of murders, you can't justify starting with 0.00000004 (the random Californian) as a prior.

Suppose you choose something closer to maybe 0.01. That would be saying, in this case, that 1% of missing wives that have been missing for 30 years turn out to have been murdered by their husbands. Seems reasonable.

With a prior like that, the evidence only needs to justify a multiplier of 99. In other words the jury would be saying

The evidence presented seems 99 times more likely if we assume the guilt of the defendent than if we don't.

This factor, 99, seems more within the reach of the kind of evidence the video presents than does the factor 24 million.


Just as a general comment on the OP:

I don’t even know how one would categorize “extraordinary evidence” unless you were talking amounts or in generalities. In other words, “He’s toast. We have an extraordinary amount of evidence that points to his guilt.” Even in that example, extraordinary doesn’t mean a whole lot to the careful observer.

Evidence is unique in most cases as it points specifically to a unique conclusion. For example, there is a unique palm print that points to a unique murder of a unique person. Extraordinary evidence, is almost like saying, “extraordinary math problem”. Does the math problem work? Does the math problem use established mathematical procedures?


Bayes suggests a way to define extraordinary evidence.

If the probability of a hypothesis, H, prior to the presentation of E was very low - say one in a million - then let's call H is an extraordinary claim.

If the evidence, E, is a million times more likely given the hypothesis, H, than it is given not-H, then E is just the kind of evidence you need to make belief in H rational.

In that case, it makes perfect sense to me to call E 'extraordinary'.



I don’t necessarily have a problem with Bayes' definition as far as it goes. I just don’t know exactly what “extraordinary evidence” would look like and Bayesian inference doesn’t help me.

Let’s say for example, I state that my blood type is different from all known blood types. Most would call that an extraordinary claim. The evidence to prove this would be a simple blood test. It turns out, my claim is 'true'. My blood type is unrecognizable. I make that example because it would be at odds with current scientific medical theory. Now is it more likely that I have a blood type that is different from all humans or that the testing equipment was wrong the first 50 times my blood was tested? Bayesian inference would lead us to…what exactly here?

Normal and reliable evidence (e.g. a blood test) is sufficient to convince me of things. If, by extension, evidence is labeled “extraordinary” simply because of the initial hypothesis, then it really has no tangible real world use.


If I'm not mistaken, what you seem to be saying is that regardless of the data-based estimates, you choose to continue to consider murder by a family member not extraordinary. Is this because you know the estimates to be based on poor data? Are the actual estimates incorrect? Or is it that it just "feels" like a lot of people are murdered by family members?

I would agree that these type of murder cases seem to appear often in the local and national news. However, I remember growing up in Colorado a lot of what I heard about winter in New England was the awful nor'easter storms that struck Boston bringing Ice and no electricity for days. Then I moved to Boston expecting winters to be filled with these awful snow storms. In my 6 years living in Boston I dealt with 1 significant nor'easter. Snow storms tended to be much like those on the Eastern slope of Colorado. The national news outlets didn't report the typical storms that occurred in Boston because that was commonplace - unremarkable. So it is with the news regarding murders. We here about the remarkable, the out of the ordinary, the unusually tragic. Many of these might involve a family member because that makes it seem all the more tragic.

Think about your statistic - 33% of all women murdered are killed by an intimate partner. Let's forget about family member versus intimate partner. A study done using FBI crime records reports that 23% of murder victims are women. This amounts to 8% of all murder victims being murdered by an intimate partner. This does not include men who are victims of murder by an intimate parter, but I think we would all agree that number will be far less than for women. Again, given the relatively small number of murders that occur each year in a population of 300,000,000, this still sounds far, far from typical.



The blood test is extraordinary evidence by the objective measure I gave which measures its power to distinguish H from not-H.

The blood test is common, 'simple', 'normal', and extraordinary.


So extraordinary evidence essentially means that the evidence shows that the claim is probably true (and without it, the claim would probably be false). The evidence is otherwise common, simple and normal.

A claim that is probably false based on its 'prior' probability also seems to get the tag "extraordinary".

So Carl Sagan's dramatic and high-flown slogan, "Extraordinary claims require extraordinary evidence" turns out to mean this

Claims that, prior to considering the evidence are probably false require common, simple, normal evidence that shows that they are probably true before it is reasonable to suppose that they are true.
Viva Carl Sagan!

If that's all that atheists are up to with this slogan, then it's really quite prosaic. It hardly needs to be mentioned. And the debate with theists becomes one where both sides have to grapple with evidence from the other side, and at least one side is often trying to build a cumulative case with all sorts of disparate evidence.

Maybe it's just me, but I think when atheists invoke the "extraordinary evidence" catchphrase, they are usually trying to do something more. It is as if they are trying to cast a magic spell that will strike theists dumb.


Imagine KWM's blood example with a twist: there's a cheap test that is only right. say, 95% of the time and the normal test that has ALWAYS proved right.

When the cheap test says KWM's blood is unique in history what shall we do?


I suppose we will argue a bit and then try the normal test. And just one run probably wouldn't be good enough. We'd probably need several runs conducted by independent labs before we accepted KWM's anomalous blood type.

My point was not to deny the use of evidence, probability and Bayes' theorem and all that.

No reputable apologist I know of does deny that.

What they do refuse to do is be cowed or bluffed by things like the 'extraordinary evidence' catchphrase.

I suppose...

I suppose so! That would give us extraordinary evidence!

Anyway, when I say 'extraordinary claims require extraordinary evidence' what I mean is I'm a Bayesian. Nothing more or less.

When others say it, I suggest you and other reputable apologists ask them what they mean.

When you say "extraordinary claims require extraordinary evidence" you are saying that you are a Bayesian. I suppose, that you expect your interlocutors to make their probability inferences in compliance with Bayes' theorem.


I think I would have put the same point this way: "All claims require ordinary evidence", but to each his own.

I still think that the claim is usually used, though not by you Ron, to bluster, not argue.

All you have to do is ask. If the purpose is bluster, asking will uncover that.

"That’s not unusual for my cases (all of them have been built on circumstantial evidence)... There wasn’t a single piece of physical evidence. In fact, we didn’t even have the victim’s body. There wasn’t even a crime scene; the case was worked as a “missing person” investigation back in 1981 and no one examined the home where the victim was killed."

"After weeks of testimony, the jury deliberated for only 4 ½ hours. They found the defendant guilty."

Would someone here explain to me, as if I was a twelve year old, how on Earth this case can possible be solid? A court case based on testimony several years after the facts, with zero physical evidence of any sort.

I'm a thorough going Bayesian when it comes to evaluating epistemic claims. If call the a priori probability that the man is guilty P(G), the posterior probability that the man is guilty given the evidence P(G|E), the probability of getting the evidence given that he's guilty P(E|G), and the probability of getting the evidence regardless of whether he is guilty or not P(E). Then the correct way to update ones beliefs is.

P(G|E) = P(E|G)*P(G)/P(E)

Now if P(G) is sufficiently (i.e the claim is sufficiently 'extraordinary') then P(E|G)/P(E) has to be sufficiently high. Now we know what P(G) is from the story 0.013%, so in order to raise P(G|E) to 0.999 (at this point you'd on average only put one innocent man behind bars per thousand guilty), we need a bayes factor of roughly P(E|G)/P(E) = 7679.

Now lets be very generous and say that P(E|G) = 1, so that all the evidence are very likely given the truth of him being guilty. Then we require that the likelihood of the witness testimony turning out like it did to be less than 1 out of 7679. This could be done by examining historical cases of how likely witness testimonies allign with a person's guilt.

If its less than that probability, then the evidence would be extraordinary in the sense that.. the evidence is sufficient to justify warrant in believing the person to be guilty.

How this can be achieved with simple witness testimony, I can't claim to know. The story could have done with putting emphasis on what kind of testimony was delivered.

Otherwise it seems like Stand to Reason is encouraging careless interpretation with the evidence.

My mistake I accidentally turned P(G) into the probability of putting 'a murderer' behind bars rather than 'the murderer'. Mea culpa.

And of course the probabilities can be given a lot further updates with more background information. For example I agree with the suggestion that given that we know that she's been missing, then we simple use the posterior probability of her being murdered, given that we know that she's a missing person, which might 0.01. However they're running a trial on whether she was murdered, but whether this particular man was the murderer. So you'd need factor in the probability of her being murdered by her husband. This background information would make it significantly more likely that he was the killer, and perhaps witness testimony would suffice for the rest.


You have to consider how likely the E is on not-G. As I said above, this is a big problem: Even if we watch the video, we don't have all the evidence the jury saw.

We know - as far as I recall from the video, E consists of:

1) The defendant's story changed multiple times.
2) He seemed to make a 'Freudian slip' on tape. He said something like "It wasn't that" - where 'it' seemed to refer to his motive for killing his wife.

So what's the ratio P(E|G)/(P(E|G+ P(E|not-G) ?

I don't think it's in the millions as the OP seems to imply (whether intentionally or not). Is it in the thousands? Hard to say. I have know people who seem to tell lies that seem, well, crazy. Is the defendant one of them? Who knows?

You might say his confession settles it.

But, he continues to lie (about the location of the body) after the confession.

To what end?



P(E|G)/( P(E|G)+ P(E|not-G) )

    If I'm not mistaken, what you seem to be saying is that regardless of the data-based estimates, you choose to continue to consider murder by a family member not extraordinary. Is this because you know the estimates to be based on poor data? Are the actual estimates incorrect? Or is it that it just "feels" like a lot of people are murdered by family members?

"Rare" or "atypical" is not the same as "extra-ordinary". Obviously any specific type of murder is gonna be in the minority, as there are numerous types of violent killing and numerous reasons to kill someone, and the statistics clearly seem to show that women are very unequally represented in the category "killed by spouse". And obviously most people are not murdered at all. Not everybody gets lung cancer either, but that does not make lung cancer extra-ordinary.

The comments to this entry are closed.