A while back, Heythrop College professor of philosophy Stephen Law published a peer reviewed paper establishing that the Gospels actually undermine confidence in the historicity of Jesus owing to their inordinately fabulous nature (“Evidence, Miracles, and the Existence of Jesus,” Faith and Philosophy 28.2, April 2011: 129-51). He develops there what he calls the Principle of Contamination:

Where testimony/documents weave together a narrative that combines mundane claims with a significant proportion of extraordinary claims, and there is good reason to be sceptical about those extraordinary claims, then there is good reason to be sceptical about the mundane claims, at least until we possess good independent evidence of their truth.

Law then demonstrates the validity of this principle and its logical consequences for Jesus: because we have no “good independent evidence” for the mundane claim that Jesus existed, the Gospels actually leave us in doubt whether he existed; they do not afford us evidence that he did.

That conclusion hinges on what one concludes from the Epistles of Paul, whether there is anything in them that constitutes good evidence Jesus existed (as opposed to feeble evidence at best), and Law’s paper does not deal with that (see instead Ch. 11 of On the Historicity of Jesus); but it handles all other objections capably. And in comments, Law says he “allow[s] that the existence of Jesus might be a bit more probable than not.” Hence Law does not argue from this that Jesus didn’t exist; only that doubt is justified. That there isn’t good reason to be confident he existed is “not to say we yet possess good reason to suppose he is mythical.” It just means we can’t know whether Jesus existed, from the kind of evidence we have; that such assertions of certainty as we find in the literature, are not logically warranted.

Law has responded, quite capably, to a dishonest and illogical rebuttal from William Lane Craig (Response to Craig’s Crit of My Paper on the Existence of Jesus). So did I (in Craig vs. Law on the Argument from Contamination; note Law’s paper links to mine, but at its old location, so use the link I just provided to find and cite it). Now there is a rebuttal from a team of well-known Bayesian atheists, Robert Greg Cavin and Carlos A. Colombetti: “Evidence, Miracles, and the Existence of Jesus: Comments on Stephen Law,” Faith and Philosophy 31.2, April 2014: 204-16. In which they say some really strange things. Here’s an analysis.

The Principle in Practice

Really, what Law has caught on to is a principle historians have long always taken for granted: absent good evidence otherwise, the more mythified a person is, the less likely it is they existed. The more fabulous the only tales we have of someone are, the more likely we doubt their historicity, unless we have some good mundane corroboration for them. Hence we doubt the existence of Hercules, Dionysus, Romulus, and so on. Historians have always followed this principle. It’s not even controversial. It only becomes so when, oh crap, someone realizes it applies to Jesus! Even secularists start freaking out.

And the fact is, Jesus is one of the most mythified persons in human history. Yes, even from just the four canonical Gospels. Jesus belongs to numerous myth-heavy sets (unlike any other historical person known). He belongs to some sets that contain lots of historical members, too. For example, he fits the Josephan Christs class (Ch. 6.5, OHJ). But the prior probability of historicity lies at the conjunction of sets to which a person belongs. You can’t “ignore” background data to get the prior you want. All the data goes in. So, for example, Jesus lies at the conjunction of the Josephan Christs class and all those myth-heavy classes. None of the members of the Josephan Christ class do. Therefore, the prior probability they exist, does not commute to Jesus.

The best reference class for formulating a prior always lies at the nexus of two properties: (1) it’s specific enough to be certain of a connection other than coincidence; and (2) it has a lot of members. The best conjunction of those two properties, gives you the most authoritative prior probability (which you then must moderate with appropriate margins of error, and those margins might be wide, if even the best available reference class only poorly satisfies (1) and (2), as is often the case for ancient history, where surviving data is scarce).

(1) Too non-specific the properties that gain membership to the set you are using, and the frequency you get will only apply to far too broad a class of things, entailing fatally large margins of error—since you really can’t claim then to know with any confidence that the resulting frequency holds for every subset. For example, the reference class “people claimed to be historical” is far too broad, as it includes such disparate types of people as talking animals, and gods, demons, and angels. Obviously the frequency with which, say, people who are talking animals existed, is not the same as, say, people listed as minor government officials in a mundane history. And this is also true for more subtle distinctions, like “minor government officials in a history written by people who consulted official state documents mentioning them” vs. “legendary kings from eras no documents would have survived from.” So you need a narrower class than just “people claimed to be historical” (see Ch. 6.1-2, OHJ).

(2) Too few members in the set, and the frequency of historicity you get from that set will also leave you with margins of error so wide as to render it useless. Ideally you want more than six members for a set. In ancient history that’s the best you can hope for a lot of the time. In science, you need over twenty, for the very poorest result; over a hundred for a weak result; and at least a thousand for a strong result. But science is, of course, seeking very narrow margins of error at higher confidence levels than we can often expect in ancient history. That’s what makes history different from “the sciences” even though it is also actually a science (just a science with poor and problematic data—data scientists would consider too poor even to use). I discuss that point in Proving History (pp. 45-49).

In Historicity I show Jesus belongs to a lot of sets, whose members are mythical to a much higher frequency than just anyone claimed to exist. In fact at least fourteen (see count). That’s pretty shocking. I only focus on a few of those fourteen sets in detail, the ones with the most specific features (meeting criterion (1)). For example, Jesus very strongly belongs to the counter-cultural hero set (Element 46: OHJ, pp. 222-25), but that set contains only two other definite members. That’s interesting, and reduces the prior for anyone in that set to somewhere around 50/50 (including Socrates; but the evidence for him then simply overwhelms that prior: cf. OHJ, Ch. 8.2), but with so few members, how sure of that can we be? Our margin of error becomes unwieldy.

But there is one set to which Jesus belongs that meets both criteria. The Rank-Raglan hero class has fifteen members (Element 48: OHJ, pp. 229-34). That’s impossible by coincidence. Which gives us a much higher confidence that the resulting frequency of historicity observed for members of that set is no accident. It also has even more properties fixing membership of the set, so the odds of being in that set by accident is also more than sufficiently reduced. But even if it had a high frequency of members being in it by accident, it should therefore also have a high frequency of its members being historical (since picking “persons claimed to be historical” at random, fifteen times, should get you a lot of historical persons; it would be highly improbable all fifteen you picked at random weren’t historical).

Which means the Rank-Raglan set is a good proxy for a broader phenomenon: mythification in general. It’s not that we are proposing being a Rank-Raglan hero causes you to not exist, or not existing causes you to be made into a Rank-Raglan hero. Rather, the correlation between historicity and Rank-Raglan-scale mythification is just a way we can access the frequency of historicity in any heavy mythification condition. In other words, being a Rank-Raglan hero is just one way we can more clearly tell how mythified someone has become, and what the effect of such a scale of mythification is on the prior expectancy of historicity. Persons that mythified, typically don’t exist. And that’s why this reference class entails a low prior for Jesus. But just as when we get a low prior for Socrates, evidence can easily reverse that. We have ample evidence to confirm Socrates existed (evidence we don’t have for Jesus). But if we didn’t, we would indeed be warranted in doubting he existed. Such as in the case of Pythagoras. And Socrates’s fellow set-member, Aesop.

In short, scoring high on those properties for real (or even by legendary accretion) is just so rare, that anyone assigned those properties will likely be mythical. And when we see Jesus fits many other myth-heavy sets, more than any historical person known, our confidence in the value of this proxy increases substantially. Can a historical person get that mythified? Certainly. But the question is: How often does that happen? And therefore how probable is it, usually? The answer, it turns out, is: not very often; therefore, not highly probable. It’s still possible. Indeed, at the edge of my margins of error I give it 1 in 3 odds. Which is a pretty fair chance of Jesus existing. And evidence can easily overwhelm that prior. As it does for Socrates. Indeed, all you need is evidence that’s four times more likely to exist if Jesus existed than if he didn’t. A prior probability that weak, doesn’t take a lot of evidence to reverse. Which is why it’s so weird there isn’t enough evidence to establish Jesus existed.

Consider Alexander the Great. He’s the closest any historical person comes to being in the Rank-Raglan set. He scores, in reality, below half and thus doesn’t make the cut. But imagine he was in it; indeed, imagine he scored a full 22 out of 22. The prior probability he was historical would then still be low—say, the 1 in 3 I end up with on the top end of my error margin for members of that set—because it’s still the case most members of that set aren’t historical. Most persons don’t start historical and become that mythified; most persons that mythified, started out mythical. But then we look at the evidence for Alexander’s existence, and find it’s so vast, it far overwhelms that low prior, and establishes him as historical beyond all reasonable doubt (cf. OHJ, pp. 21-24). And if we didn’t have any of that evidence, then indeed we would have to admit we don’t know Alexander really existed; because most persons like him didn’t. He would then be more like Romulus or Moses. It’s only because the evidence for Alexander (unlike them) actually is so strong, that we can conclude he is indeed one of the exceptions, despite how mythified he became. And looking for that evidence, to see if it overcomes that low prior, is what Chs. 7-11 of OHJ do. Ch. 6 only determines the prior. Which is not the probability Jesus existed. It’s just the probability he existed before we look at the specific evidence he existed.

Law’s Contamination Principle is just another way to proxy the same result. When we look at documents as mythified as the ones Law defines in his set, across all human history, what is the frequency with which the persons in them still nevertheless existed, despite all the absurdities surrounding them? We can’t count people for whom we have good independent evidence they existed. Because Law’s principle by his own definition is excluding those cases. What is the frequency left over, after we rule those persons out? We can’t say it’s any better than 50/50. Because by definition we have no evidence it’s higher. And their being so surrounded by absurdity, argues against it being higher. And that’s Law’s point. Documents like the Gospels, simply can’t add to the probability Jesus was historical. They don’t thereby prove he’s mythical, either. One would need to get to a more refined reference class to get a lower prior than that; which is what I am doing with the Rank-Raglan scale as a proxy for degree of mythification. Law’s principle does not treat of degrees. It’s merely a threshold test. And as far as that goes, it looks perfectly valid to me.

The Argument of Cavin & Colombetti

How do Cavin & Colombetti argue otherwise? One point they make amounts to this: they believe evidence in Paul is “good independent evidence” Jesus existed. This would not invalidate Law’s Contamination Principle of course. It rather argues his principle doesn’t apply. I’ll treat of that argument next, since it has less to do with Law’s paper. Other than to question one premise in it: whether there is evidence outside the Gospels sufficient to be confident Jesus existed. The bulk of Law’s paper is a defense of the Contamination Principle; whether it actually ends up applying to Jesus is a separate question.

As to his principle, Cavin & Colombetti begin by oddly straw manning it in such a way as I hope becomes a legendary joke in the philosophy community: they start by saying his premise begs the question when it asserts “the dubious character of the extraordinary, uncorroborated parts of [an example] testimony does contaminate the non-extraordinary parts,” and proceed to “fix” this by retranslating it to “the dubious character of the extraordinary, uncorroborated parts of their testimony raises some degree of suspicion about the non-extraordinary parts,” which actually begs the question—by presuming that Law meant what they conclude invalidates his premise, rather than considering what he actually in fact meant. Their conclusion thus ends up in their premise! Ladies and gentlemen, that’s what a circular argument looks like. To have ironically made a circular argument to argue someone else made a circular argument, is just precious.

Actually, what Law obviously meant was “the dubious character of the extraordinary, uncorroborated parts of [an example] testimony does render dubious the non-extraordinary parts.” Here we avoid Cavin & Colombetti’s straw man presumption that Law meant merely “some” suspicion. Of course he meant sufficient suspicion. In other words, he is saying we are left in adequate doubt of the matter. And he’s right (just examine the example in question and you’ll agree). This does not mean, as Cavin & Colombetti’s argument absurdly suggests, that Law meant the actual improbability of the absurd details, commutes perfectly to the mundane details, as if claiming that from my story about a man named Jim George landing a spaceship in my backyard, we must conclude that since a spaceship landing in my backyard is a billion to one against, therefore the existence of Jim George is a billion to one against. To the contrary, Law is saying that, in effect, the spaceship’s existence being a billion to one against, leaves us with at best a 50/50 assurance of the existence of Jim George. Still far greater a probability than the spaceship—but by no means high enough to be confident Jim George existed. We would need more evidence for that than just my story of his landing a spaceship in my yard. I cannot fathom how Cavin & Colombetti could object to this conclusion. It’s obviously correct.

Cavin & Colombetti then commit one of the strangest errors I’ve ever seen an atheist bumble into. They say “the assessment of [the Jesus case] is not just a matter of our intuition about [Law’s “Ted and Sarah” case], but also a matter of whether and to what degree the Ted and Sarah case is a strong analogy to the apostles and Jesus.” News to Cavin & Colombetti: No apostles wrote the Gospels. So, no, you are proposing a wholly invalid analogy. Sure, if we heard this story from the apostles, you’d have a point. But then, if we had that, Law probably wouldn’t have written his paper (and the conclusion of my book would be radically different!). It blows me away that atheists are here actually trying to avoid Law’s conclusion by insisting on apostolic authorship of the Gospels! They may as well just convert already and seek positions at the Talbot School of Theology. Worse, these atheists even shill for Catholicism, by affirming the historicity of all the Saints. Um, no—actually, historians doubt the existence of the Saints for whom we lack corroboration of their existence outside wild myths of them (e.g. Saint Sergius and Saint Bacchus, Saint Maurice, and on and on). Because historians follow Law’s principle. As they should.

Cavin & Colombetti then shockingly abuse Bayesian reasoning by trying to criticize Law for rejecting the approach of restoring historicity with a “hoax” hypothesis (that is, they endorse the tactic of proposing that maybe the absurd details of the story were actually mundane). They use this “possibility” to claim he is the one confusing priors with posteriors. Um, no. Yes, “it does not follow that [the] posterior probability [of a hoax hypothesis] on the total evidence is low.” But Law’s principle covers that: it explicitly states it only applies when there isn’t good external evidence of the fact. His principle thus rules out situations like they propose. Moreover, his principle rules out situations where we know the absurd details aren’t absurd. If we had sufficient evidence to believe the absurdities in a story were effected by a hoax, and hence were mundane after all, Law’s principle wouldn’t apply. And he would agree that it wouldn’t apply. So when we lack evidence of such hoaxing, proposing it out of thin air, on no evidence whatever, merely to “rescue” the historicity of a character in the story, is the most craven of apologetics—which Bayesians should well recognize as illegitimate. You can’t rescue the historicity of the talking animals in Aesop’s fables, by proposing Aesop was tricked by people wearing costumes. Obviously the posterior probability remains low that there were any animals, by disguise or otherwise. For Cavin & Colombetti to try and deny this, appalls me.

Cavin & Colombetti then try to prove his principle false with a Bayesian equation…by invalidly presuming the mundane details in a story are true! This is circular reasoning. You can’t conclude the mundane details are true, by starting with a premise that asserts they are true. I’m not kidding. They say every mundane claim made about Jesus starts with a prior probability of 0.99! Where do they get any evidence supporting that? Good luck figuring that out. They give none. But even apart from the fact that no evidence exists that does that, they are arguing in a circle. Obviously if we have already concluded that the mundane details in a story are 99% certain to be true, Law’s Contamination Principle no longer applies. Indeed, Law would fully agree. Law is talking about cases where we don’t already know that.

Starting on no evidence—so we don’t know whether a story is true or false, and we are not presuming Jesus existed—it’s probability is 0.5. It’s as likely true as false. If we want to get that probability higher, we need more data than just “it’s in a story.” Now let’s go through the Gospels: How initially likely is it that someone in it existed? Before introducing reference classes, i.e. starting with no data to derive an expected frequency from, it’s 50/50. How likely is it that Jesus really came from Nazareth, or this was only attributed to him to conform to prophecy as the Gospels report (Matthew 2:23)? We have no evidence either way…so it’s 50/50. Which does not increase the prior. So now we’ve updated our prior of 0.5…to 0.5. How likely is it that Jesus really hung out at a house in Capernaum? Without presuming his existence, well, it’s 50/50. Because we lack any evidence anyone actually saw him there…remember, no apostle wrote the Gospels, nor did any witness, nor anyone who attests they ever even met any witness. So we plug that in and our prior updates to…0.5.

See where this is going?

You can’t get to 0.99 this way. And Cavin & Colombetti don’t even try to get to 0.99. They just “start” there. For no reason whatever.

Worse, Cavin & Colombetti assert there can be no contamination. None. “There is no contamination.” They are literally claiming that when we get evidence that a source is lying on some details, this does not at all reduce the probability of its telling the truth on others. Um, no. Once we know a source is unreliable, we consider it unreliable. Period. How can Cavin & Colombetti explain that in Bayesian terms—without a contamination principle? Obviously, if someone is caught making up half a story, we have no reason left to trust anything in the other half—or to know which details are true even if any are (until we can corroborate anything elsewise). One can say that if it’s 50/50 for every detail, then maybe half the details are true…but which ones? That Jesus existed or that Nazareth did? Nazareth can exist when Jesus does not; and vice versa. So we can’t rescue the historicity of Jesus this way, either.

An apt example is the Historia Augusta, a 4th century collection of biographies of Roman Emperors. Later in that collection, many details begin to be demonstrably fabricated. Historians thus conclude we cannot know anything in those later stories is true—even if any of them are. That is Law’s Contamination Principle. And it is a standard methodological principle in the field of history. Obviously when our evidence includes the datum that “this source is not reliable,” that entails the probability anything they report is lower than if we lacked that datum (or had the datum “this source is reliable”). That’s what saying a source is unreliable means. The notion of rejecting this clearly correct principle is actually a peculiar trick of Christian apologists. Why then are atheists defending it?

Cavin & Colombetti try to drive this home with an appeal to the crucifixion being too embarrassing to invent, but at this point they are confusing those two different arguments: the argument that Law’s Contamination Principle isn’t valid, and the argument that it doesn’t apply (because we have “good independent evidence” Jesus existed after all: this is one of the four pieces of evidence they had already adduced for that conclusion, as I’ll get to next). They can’t make this argument against its validity, because then they’d be forced to affirm countless absurdities, such as that Romulus (Quirinus), one of the three principal gods revered by the state, must actually have existed and murdered his brother, because fratricide was the gravest of sins in the Roman system of theology, so surely no one would have invented that; or that the revered and worshiped savior god Attis must actually have existed and really castrated himself, because emasculation was among the most repulsive and demeaning acts in ancient society that automatically negated social status and respect, so surely no one would have invented that; and so on.

Cavin & Colombetti should be reminded that there are a lot of murdered, desecrated, humiliated, and even crucified gods revered in antiquity (see Chs. 1 & 2 of Not the Impossible Faith): from the crucified Inanna to the dismembered Osiris (see OHJ, index). You don’t get historicity out of that. And if not for them, neither for Jesus. That the Gospels narrate a crucifixion for Jesus, is just as likely on either historicity or myth, because on either theory his being crucified was central to the creed and its soteriology from day one. Whether Jesus existed or not, his theological role was invented so he could replace the temple atonement sacrifice, necessitating that he be sacrificed. Because Christianity was an anti-temple cult, and thus needed a soteriological system to replace its atonement effect. Since that’s already central to the religion, on both theories, the Gospels repeating it is no more likely on one than the other. It’s therefore not evidence against Law’s Contamination Principle’s applicability to the Gospels.

Cavin & Colombetti’s claim that every mundane fact claimed in the Gospels is 99% certain to be true is also contrary to the entirety of mainstream peer reviewed scholarship. For example, as next we’ll see, the mainstream consensus is almost none of the sayings of Jesus in the Gospels go back to him. Not 99%. Not even 50%. And those that they argue do…may have come by revelation, not the man. So their originating at the beginning of the cult is not even evidence for his historicity. Not having any evidence to distinguish which sayings came from him and which from his followers later claiming revelations from him (and we don’t—not any), it’s 50/50. Back to the original problem. The prior never gets updated above 50%, when none of the evidence is better than 50/50. And none is.

Certainly, there is data in the Gospels disconnected from Jesus that’s better than 50/50…for instance, Bethlehem existed. But we don’t presume that. We don’t just assign it a 99% certainty just “because.” All historians reject the absurd method of Cavin & Colombetti. All historians adopt Law’s method: we believe Bethlehem existed, because we have “good independent evidence” it existed. Note: Bethlehem existing, in no way increases the probability Jesus existed. Or at least, by the Lords of Kobol, I hope Cavin & Colombetti aren’t going to stay on trend and start defending the historical truth of the nativity myths, merely to preserve their desperate need to have a Jesus. What about cities in the Gospels we don’t have any good independent evidence for? Like Arimathea? Historians conclude: 50/50. Just as Law’s principle predicts, and justifies. Cavin & Colombetti’s method can make no sense of this. By their reasoning, the probability Joseph of Arimathea existed is automatically 99%. For some reason. Heck if I know what reason.

Do Paul’s Epistles Prove Jesus Existed?

Secondary to the above, I mentioned, was Cavin & Colombetti’s argument for historicity from the Epistles. That doesn’t argue against Law’s principle, but rather with its application to Jesus. Their paper was published in early 2014, and likely it was already in the publication pipe as much as a year earlier. So they could not have consulted the latest peer reviewed literature in biblical studies on the evidence for Jesus in the Epistles: which is Ch. 11 of my book On the Historicity of Jesus (published by Sheffield-Phoenix, a respected biblical studies academic press). But that’s unfortunate. Because they make factually false claims in result.

First, they claim “Paul testifies (Gal. 2:9) that just a few years after the crucifixion he personally met with Peter, James, and other leaders of the church who claimed to be the disciples and family of Jesus.” Actually, no they don’t. None of them, nor Paul, ever call themselves “disciples” of Jesus. The word is unknown to Paul. Only apostles existed, whom Paul understands as those receiving a revelation of the Christ (1 Cor. 9:1 and 15:3-8; Gal. 1:11-18; Rom. 16:25-26). Paul never mentions any of them ever even having seen, much less met, Jesus before his death. That does not entail they didn’t; but we are supposed to be looking for evidence that they did. There is none in Paul.

It’s fairer to say some of them claim (at least as attested by Paul) to have been “family of Jesus.” That’s actually open to interpretation and much more uncertain than Cavin & Colombetti realize. In the final analysis, that that is what Paul meant is at best 50/50. Which renders that evidence likewise useless. But we can’t criticize Cavin & Colombetti for not knowing this, as Law never treats the subject, and they don’t claim to be addressing anyone else, so they won’t have known of the arguments (both good and bad) that mythicists have long used regarding the value of this particular evidence. I do consider it the best evidence historicists have, and in my book I assign it a strength of 2:1 in favor of historicity at the top end of my margins of error. Meaning, that Paul would say this (and it is only Paul who does; the letter of James doesn’t, nor any other source we can trace back to James or any other family of Jesus outside speculation) is twice as likely if Jesus existed than if he didn’t, after considering all the background evidence weighing against that conclusion (e.g. Element 12 in Ch. 4; for a full treatment of the problems with this evidence, see Ch. 11.10 of OHJ).

Second, they claim “one must account for the coherent body of sayings and actions attributed to Jesus in the gospels, which seem to derive from a single creative and sagacious personality. The simplest explanation is that they derive from Jesus himself.” This is actually not even a mainstream conclusion in the field. The mainstream consensus is in fact that most of those sayings do not come from Jesus (likewise most deeds). The mainstream peer reviewed literature establishes not even the Sermon on the Mount does (e.g. the conclusions of Dale Allison, summarized and cited in OHJ, pp. 465-68). Look at The Five Gospels produced by the Jesus Seminar, specifically to assess how many sayings came from Jesus, and weep at how little of it is colored red (or even pink). (Likewise for his deeds: The Acts of Jesus.) IMO, even they are being too generous, but that’s another matter. The point is, Cavin & Colombetti are citing as evidence what is actually rejected as evidence by the field they claim to be speaking for. The argument is also not even logical—as if Jesus were a unique superhuman, such that no other church leaders could ever have thought of saying those things. Nor are his sayings across the Gospels at all consistent enough to have even come from a “single” personality (the fallacy of excluding everything that doesn’t fit and then concluding what remains “miraculously” coheres, I already explained when Bermejo-Rubio did it).