An interesting video discussion of On the Historicity of Jesus took place earlier this year, in which “Kamil Gregor and Joel Pearson discuss the Historicity of Jesus.” It’s well worth a reply. Because Gregor understands the math. So what he gets wrong about it, and about the facts and arguments in OHJ, might be informative to people who don’t understand the math.

Agreements & Differences

Gregor is a historical minimalist about Jesus, saying he agrees with what I define as “minimal historicity” as the most defensible hypothesis (which I test against what I define as “minimal mythicism”). Except, he goes a bit further, and adds that he thinks “the authors” of the Gospels “didn’t know or didn’t care about almost anything about the actual historical Jesus,” and so the Gospels are pretty much wholly myth; but Jesus is not. Other scholars have taken similar positions, including Dennis MacDonald.

I concur. Though of course I take that even one more step, and conclude Jesus himself is also probably mythical. Though only somewhat probably; my upper estimate is a “1 in 3 chance” Jesus nevertheless still existed, which is actually a more respectable probability that he existed than most people realize I grant. Though it should be clear that mathematically, I do not conclude this because the Gospels are wholly myth. I consider the Gospel question a wash: I find it’s at best 50/50 whether the Gospels contain any historical information about Jesus, so they count for nothing as evidence either way.

Yes, personally, I do think it’s worse than that. I’m quite sure the Gospels are not doing real history at all (even when pretending to). But even if we knew for certain that the Gospels were 100% myth, Gregor would be right: that does not, in itself, affect the probability Jesus existed. Except by removing evidence historicists would otherwise appeal to. So in OHJ, I have the Gospels affect that probability not at all, neither increasing nor decreasing it. With one exception Gregor will get to: the connection between the extent of their mythologization of Jesus, and the expectancy that a subject of such mythologization existed. Which is whatever the base rate of that is likely to be. How often do figures that mythologized turn out actually to exist?

Otherwise, Gregor “pretty much agrees” with my “evaluation of the evidence,” in that “by and large the evidence…isn’t really more probable on either hypothesis.” This is indeed what I find for a lot of the evidence, especially two things people often mistakenly think I count as evidence against historicity: the silence of the external record (which I weigh at “no effect,” meaning, it’s equally likely on either minimal historicity or mythicism) and the Gospels (which I weigh similarly at “no effect,” except for that one feature I mention above).

Some of the other external evidence however (like the content of 1 Clement and Hebrews and some strange passages in Ignatius) and the New Testament book of Acts I find slightly harder to explain on historicity (but only slightly; you can see my explanation and estimates in OHJ). I’m not sure Gregor concurs there. Whereas contrary to what most people think, on the upper bound of my error margins I actually count the Epistles of Paul as weighing for the historicity of Jesus—just not as heavily as historicists do, as I correctly allow the content of those Epistles that remains hard to explain on historicity to properly hold back their strength as evidence.

Gregor also agrees that the “method of criteria” employed in Jesus studies to try and extract history from surviving texts is “probably impossible,” or at best, “the level of confidence you can get on that is very low.” In other words, the methods historicists use are very weak; too weak to produce the certainty they declare. But he might go too far in saying, as he eventually does, that no theory of Jesus is even falsifiable, “because” all evidence can be “explained away.” Here Gregor has fundamentally stumbled in his application of Bayesian reasoning—incongruously, since he knows it’s a mistake. So we’ll get into that first.

Gregor also agrees that historians should always be willing to state what the range of their probabilities are for Jesus existing. It’s not really sound to just state a probability—like, say, “I am 90% sure Jesus existed.” Not only because that’s an undisciplined estimate (a historian who says that usually has done no work to actually mathematically justify it; and historians, being professionals, should be disciplined in their assertions), but also because it’s more honest to admit the extent of your uncertainty. What is the highest probability you think Jesus has of having really existed? And what is the lowest probability you think Jesus has of having really existed? Take the questions seriously. How low do you think would be too low? And why?

Finally, Gregor says his own personal estimate is that Jesus has a 70% chance of having really existed, “plus/minus twenty percent.” So he would say, a 50-to-90% chance Jesus existed. Which is probably not high enough to give comfort to historicist scholars. As Gregor says, “on bad days I’m basically agnostic” about the historicity of Jesus. My own range is below 1% to roughly 33%. Which in the grand scheme of things is not very distant from where Gregor is. Most scholars I suspect would assert their range to be above 80% to just shy of 100%. They just won’t be able to explain why that range is justified. Gregor has a better grasp on how to do that. But he’s making some mistakes in getting there that, if corrected, might bring his range closer to mine—or at least to Lataster’s, now published by Brill, which is more centered on 50/50, and thus more assuredly agnostic, as now around eight qualified scholars are, with at least four more concurring that such a conclusion is at least plausible.

The Mathematics of Unfalsifiability

So I’ll start with the last mistake I mentioned: Gregor’s error in the mathematics of unfalsifiability. Which is strange because I know he knows better, whether or not he read Proving History, my guide to Bayesian methods in history, which extensively warns against his mistake and explains how not to make it. If you check its index under “excuse-making,” you’ll see it directs you to the entry for “gerrymandering,” which identifies eleven separate pages discussing the point (to which one should add my discussion of why our not having Pilate’s trial records cannot reduce the probability of historicity, on pp. 219-20).

I’ve discussed this here as well (for example in “The Cost of Making Excuses”). An “unfalsifiable” theory is a theory that no evidence can ever render improbable. The only way that can happen, however, is by adding an excuse for every possible kind of evidence, such that no matter what you observe, it’s always 100% expected on your theory. However, once we frame this in Bayesian terms, such an approach fails. It can be maintained only by violating basic rules of math. Because actually, any excuse you propose itself has a probability. Which necessarily multiplies against the prior probability—and the more excuses you have to invent, the lower that probability goes. This is an inevitable consequence of mathematical laws of probability. There is no way around it.

The only exception are excuses you can independently prove are probable. Which is the function of “background knowledge” (or b) in a Bayesian equation. All probabilities in Bayes’ Theorem are conditioned on b. Thereby being dependent probabilities, the difference between a legitimate excuse and an ad hoc excuse is that the latter is simply made up to get rid of some uncomfortable evidence. As such, that excuse can be no more likely than 50/50 (being, so far as you honestly know, as likely true as not), and could be much less likely (for example, if the excuse depends on rare or bizarre things having happened or being the case for which there is no evidence as actually having happened or being the case). By contrast, a legitimate excuse is one that’s already well-evidenced. Because as such, it has a high enough probability to have no appreciable effect on the prior probability.

An example of a legitimate excuse is the one I outline for defenders of historicity “explaining away” the silence of the extra-biblical record, and with which Gregor concurs: minimal historicity by definition says nothing about Jesus “being famous” and therefore in no way conditions any probability of the evidence on his being so. As a result, the silence of the extra-biblical record is essentially 100% expected even if Jesus existed. This is not an ad hoc excuse; it is logically entailed by the hypothesis already, without any added presumptions, in conjunction with our background knowledge regarding how often total nobodies got mentioned in the historical record in antiquity (which is effectively as close to never as makes all odds). So that approach by historicists is legitimate.

I make this point even in OHJ:

Whatever explanation historicists devise for [any particular] curiosities has to be demonstrably true, and not something they just make up to explain away the evidence. Because such ‘making up of excuses’ would risk the fallacy of gerrymandering, which necessarily lowers your theory’s prior probability since you have to assume facts that aren’t in evidence and that aren’t made probable by any evidence there is.

OHJ, pp. 515-16.

The same goes for mythicists (see OHJ, pp. 246-48, on “The Complexity Objection,” and the same point’s applications on pp. 53-55 and pp. 584-85). Gregor eventually, incorrectly, accuses me of doing this. But in direct contradiction to that, he spends half an hour doing it himself.

Applying the Mathematics of Unfalsifiability

In their discussion, Gregor asks Pearson what kind of evidence would appreciably raise his estimate of the probability Jesus existed, and Pearson replies, in effect, “finding an authenticated letter of Paul” that says something clearly placing Jesus as a recent historical figure, such as saying that Jesus was executed by Pontius Pilate. Gregor replies to this, “Why would that lower the probability of mythicism?” Pearson answers that it reduces the ambiguity of Paul. He’s right. Because I found that that ambiguity weighs down the likelihood of Paul’s Epistles on the hypothesis that Jesus really existed; ergo it is logically necessarily the case that removing that ambiguity must increase the probability of historicity by exactly as much as having that evidence reduced it.

Gregor is outright wrong to say “it doesn’t” have that effect. It absolutely does. In fact it logically necessarily does. What it means for me to say that that ambiguity lowers the probability by X amount is that not having that ambiguity would not lower the probability by X amount, which entails that a conclusion will be more probable without that evidence than with it—exactly as much more probable as I concluded that evidence made it less probable. To deny this is literally as absurd as denying 2-2=0 and insisting, instead, that 2-2=2. Gregor is simply wrong about math here. Yet he does understand the math (as we’ll see eventually). So he has simply failed to apply what he knows at this point, and thus gets everything wrong about the effect any such evidence has on the probability of mythicism.

The mistake Gregor makes is thus not applying the mathematical effect, which he well knows how to calculate, of what I call gerrymandering: adding excuses to change that otherwise inevitable outcome. He hasn’t accounted for that here, where Gregor proposes such an excuse, that “Paul just wasn’t initiated into the cosmic truth,” and therefore mistakenly believed Jesus was executed by Pontius Pilate. But you can’t just assert that as a fact, assuming it’s therefore 100% certain (as Gregor eventually admits). There is no evidence it’s true (even that there was such a fake story covering up the secret then, much less that Paul didn’t know it was fake); Gregor just made that all up. So at best you’re at 50/50, the excuse is as likely true as not (“so far as you know”), which means adding that excuse reduces the probability of mythicism by half. So that probability goes down no matter what you do (a fact Gregor eventually concedes, but never goes back to reapply in this case).

It’s even worse than that, however, because there is abundant evidence against Gregor’s proposed excuse. It would be very unlikely for Paul to have been running around believing and preaching Jesus was actually a recent state-executed criminal, and never have had to argue the point with any of the other apostles or congregations or persecutors who would know better (so the letters as we have them evince he wasn’t doing that); it’s very unlikely for Paul to have not been read in on key secrets like that, and for him yet still to have been fully accepted as an equal with the other apostles running the church, as in fact he was (so the letters as we have them evince he had to have been as informed as they were); and it’s very unlikely for Paul to have been running around preaching such a thing and never found out it was false—as this would be an event within his own lifetime, with living witnesses he himself says he met personally, and he was contending against Jewish leaders persecuting the church (and indeed had been himself one of them)—enough of whom would well know no such person existed to have exploded Paul’s entire mission.

In short, the proposal that Paul could actually have believed or gone around claiming a blatant and refutable falsehood like that, under all those circumstances, is simply so improbable that any theory depending on it would have a vanishingly small prior and consequently a vanishingly small posterior. It therefore could not keep “high” the posterior probability of mythicism as Gregor claims. That would be mathematically and evidentially impossible. No. The net effect would be that such a passage in Paul would simply tank the probability of mythicism. Full stop.

This is why it’s so important that the only passage like this, in 1 Thessalonians 2, has independent evidence of forgery, as mainstream scholars have long established, and as I analyze, indeed even mathematically (contrary to Gregor claiming I didn’t), in OHJ (pp. 566-69). Sure, in Gregor and Pearson’s hypothetical case (for which, as they stipulated, there would not be such evidence of forgery) there would remain a vanishingly small probability Gregor’s proposed explanation of that evidence “was correct.” But no one is warranted in believing things that have vanishingly small probabilities (in case that’s not clear, I’ve made this point quite forcefully elsewhere: see Extraordinary Claims and How Not to Be a Doofus). I’ll say more on this particular item of evidence later. But for now, the point is, there is no ad hoc excuse being deployed here. But for the evidence it’s forged, this passage would tank the probability of mythicism.

The matter factually changes when we are looking at the Gospels because they are written without any evident sources or access to sources, a literal lifetime after the facts, after indeed a war ravaged the whole land pertaining, to audiences in a foreign land in a foreign language, by which time we have no evidence anyone who could gainsay their claims was even alive or aware of their content, nor would we have access to what they said if they even said anything (e.g. conspicuously, no church Epistles survive from the period the Gospels were written, much less outside critiques).

There is a reason myths like this take approximately forty years to arise and spread, as I demonstrate with several examples throughout history in OHJ (from Ned Ludd and John Frum to the Roswell saucer legend; even the empty tomb and the legend of apostles dining for weeks with a resurrected corpse). See OHJ, Ch. 6.7, on the “Rapid Legendary Development” objection. For a further exploration of this whole point of how the religion transitioned from a cosmic to an earthly mythology (using the same transition in respect to the resurrection appearances as a perfect analogy), see How Did Christianity Switch to a Historical Jesus?

Indeed the first Gospel (Mark) might not have even meant his legend to be taken literally—that seems to have been a concern that grew over time: from Mark, who never makes any such claim but nearly the opposite (Mark 4:9-13); to Matthew, the first to at least vaguely claim the events happened in some sense to fulfill prophecy; to Luke, the first to purport to be writing a history at all (yet whose wording in that, like Matthew’s, is coy enough to still satisfy Mark’s notion of insider-vs.-outsider interpretations); to John, the first to actually argue readers should believe him because the things he reports actually happened (which things are conspicuously the least believable things yet claimed by any of them).

So Is Minimal Mythicism Unfalsifiable?

Pearson explains all this to Gregor, and more succinctly; but since Gregor only skeptically admits that “maybe” that makes sense, it’s evident he didn’t grasp the mathematical point behind what Pearson told him. It doesn’t “maybe” make sense; there is literally no other way to do the math. That Gregor doesn’t get it is reflected by his response, which is to erroneously claim my “theory” incorporates an assumption of a conspiracy by the elite in the Christian ranks to “hide” the cosmic truth from the average Christian behind earthly myths they expected those lower ranks to take literally, and therefore they must have “always” done this, therefore even if Paul said Jesus was executed by Pilate we should assume this was the fake, outsider message, and Jesus didn’t really exist.

Gregor gets almost every part of that wrong. Gregor has confused the hypothesis, the five premises of my theory (minimal mythicism as outlined in OHJ on p. 53), with background facts (the elements of which I lay out in later chapters and demonstrate are not merely possible but factually true). He has also gotten the chronology wrong. And he has failed to model anything he is saying in any mathematically correct way—whereas I am quite careful to do so myself.

What Gregor is talking about is what I demonstrate in Element 13 in Ch. 4 of OHJ, that Christianity always had secret teachings (as even attested by Paul), and that (emphasis now added to reveal what Gregor overlooks):

[As the third-century scholar Origen says] ‘mature’ believers are taught one thing, but ‘simpler’ believers are taught another, and in result, he says, many passages in the Gospels are literally false and only allegorically true; as Origen put it, ‘the spiritual truth was often preserved, as one might say, in a material falsehood’. … Thus it is plausible that, like other mystery cults, Christianity also came to be packaged with a set of earthly tales of its savior that were not meant to be taken literally, except by outsiders—and insiders of insufficient rank, who were variously called even by their own leaders ‘babes’ or ‘simpletons’.

OHJ, pp. 120, 124

Note: I am here talking about the Gospels and what came to be the case by the time they were written. I do not here assert this was always the case, much less that it was likely to have been. There were always secret doctrines kept from the lower ranks; that’s true (not, BTW, an assumption: I show that on background evidence its probability is as near to 100% as makes all odds; Gregor incorrectly describes this as something I merely hypothesize). But I certainly never say it was the case, nor even likely the case, that in the time of Paul there were already earthly narratives of Jesus disguising the cosmic one and that Paul would be attempting to fool any reader of the Epistles into believing them (or had been so fooled himself). That would be extremely unlikely. Yes, possible. But the lack of any argument that it was likely renders it not so (and accordingly I never argue this). You must represent this mathematically.

You’ll notice that in my hypothesis as actually defined there are five premises, none of which consists of this assumption. To the contrary, my hypothesis explicitly states (emphasis added to make this clear), “1. At the origin of Christianity, Jesus Christ was thought to be a celestial deity much like any other” who was “2. … originally believed to have endured an ordeal of incarnation, death, burial and resurrection in a supernatural realm” and “4. … an allegorical story of this same Jesus was then composed and told within the sacred community, which placed him on earth” and “5. Subsequent communities of worshipers believed (or at least taught) that this invented sacred story was real.” I am thus proposing exactly the opposite of what Gregor claims: minimal mythicism is the thesis that the earthly narratives were not original, but came later.

Okay. So, what if Gregor rejects my actual hypothesis, and maybe instead wishes to straw-man it, by himself “adding” a “sixth” premise to my five, asserting that these earthly narratives were invented and sold as true even at the origin of Christianity? Well, guess what. I wrote a whole page-and-a-half on why that’s extremely improbable, in fact so much so we can wholly discount that hypothesis. Where did I write that page? Literally immediately after I defined my hypothesis (OHJ, pp. 54-55). My conclusion: “[T]he extremely small prior probability of these ‘alternatives’…means they take up so little of the prior-probability-space that we can safely ignore them.” It’s simply not likely that even Paul could have been preaching “Mark’s Gospel” and it all be false. Even with the truth of Element 13 in b. Therefore, had Paul been doing that, mythicism itself would be extremely improbable. And such would then have been my conclusion in OHJ.

It is therefore the fact that Paul doesn’t do that that so depresses my concluding probability of historicity. It follows mathematically that if Paul did do that it could not depress my concluding probability of historicity. His doing that therefore must necessarily raise the probability of historicity—by exactly as much as I concluded the absence of this evidence depressed it. Gregor completely fails to get this mathematical point here.

Getting the Rules of Evidence Wrong

Gregor goes on to lament that we couldn’t have any evidence that would confirm, for example, my transition theory (premise 5 of my hypothesis; further outlined in How Did Christianity Switch to a Historical Jesus?). But that’s obviously false. If we had, as we have in the case of Mormonism, a complete record of church correspondence and polemics and disputes among its sects from 50 AD to 150 AD, we would have ample evidence in that material to determine how and when any transitions took place. Because we would have people discussing it, and we would have a chronological record of who was coming up with what and when, and who was saying what about it and when. For instance, we’d have discussions of this brand new “Gospel” we call Mark and when and where it was published and by whom and what it was supposed to signify or was taken to signify. We could also have the kind of evidence we do for Nedd Ludd or John Frum (OHJ, Index), which is considerably less, yet still enough for a general proof. Or for Roswell, which has considerably more, allowing us to be even more exact as to the dates and stages of the rising legend.

Just consider three examples in Jesus’s case that we do know existed (but don’t directly have): the original redaction of the Ascension of Isaiah (OHJ, Ch. 31.), the writings of the sect that 2 Peter polemically accuses of heresy for calling our Gospel narratives myths (OHJ, Ch. 8.12), and the source text for Ignatius’s mystical star gospel (OHJ, Ch. 8.6), all of which, if we had them and they turned out to date early enough, would almost conclusively prove minimal mythicism—and if the record were complete, we’d also have tons of correspondence discussing all three, including when and where and why and by whom they were composed, and the arguments each side accusing the other of fabrication was actually attempting to make. With even just that information we’d know a great deal more about the transition, how and when it occurred, and how one side prevailed over the other.

Gregor is correct that many things instead remain unknown on present evidence—for example, we really can’t tell whether the author of Luke believes there actually was an earthly Jesus, or doesn’t but wants others to, or doesn’t want either outcome except for non-Christians. But this has no effect on the probabilities, because not knowing that does not increase or decrease the probability of historicity or mythicism. Which means we don’t have to assume any one of those hypotheses about Luke’s “true beliefs” to get our conclusion. So this being the case does not make either historicity or mythicism unfalsifiable. It just means some things can’t be known about them. Thus, the conclusion of OHJ is that the evidence we do have renders it 67% likely that such a transition occurred; but that same evidence does not tell us precisely when, or through which stages. One does not need to know the latter things to know the former thing. This is therefore not a falsifiability problem.

Gregor is incorrect, however, in his use of “Polycarp” as an example. Given the way we see Polycarp and his sectarian peers behave—which includes Tertullian, Irenaeus, even the author of the Ignatians, which were edited by Polycarp or his successors into a new edition explicitly to push a historicity narrative—it is very unlikely they “secretly” did not believe in the historicity of Jesus. Yes, it’s possible. But it’s not probable. They were in fact arguing very assiduously against anyone who challenged it, declaring them outright heretics. This is essentially the role played by the Ignatians, by the Gospel of John, and by the forged Epistle of 2 Peter being included in the canon favored by Polycarp’s sect. Which would mean they were declaring themselves heretics—which makes no sense. They were trying to purge the very people who would have said Jesus never walked the earth. That is not likely the behavior of someone who agreed with those same people.

Polycarp and his peers and successors no doubt believed, as Origen among them confesses, that much of what is said about Jesus in the Gospels is allegory and not literally true; but like Origen also confesses, they definitely believed some of it was, and in fact were staking the entire truth of their belief system on that being so, right down to fabricating apostolic pedigrees, and arguing for particular dogmas on the basis of the Gospels recording historical facts. (See, for example, how I show the Ignatians were accomplishing this in How Did Christianity Switch to a Historical Jesus?)

This is not the behavior of someone like Mark, who never really says anything he is advocating is literally true, and even includes the open secret that it isn’t (in Mark 4). Contrast that with John, who is going out of his way to insist everything he is advocating is literally true (OHJ, Ch. 10.7). Similarly, we see Paul admitting openly that there were secret teachings; while simultaneously making clear he had no sources for Jesus but scripture and revelation. Polycarp’s camp were behaving in exactly the opposite way. That’s evidence they believed exactly the opposite thing.

Yes, one can say, “Well, maybe they were doing all that while still sharing the same perspective as Mark and Paul in secret.” But that isn’t by that point likely. And we should not be believing unlikely things. Gregor never seems to get this: he keeps assuming we can just “assume” things and then declare them 100% probable; when in fact, that’s not how any of this works.

Is “Born of Woman” Gerrymandered?

All of that occupied about the first third of their hour-and-a-half video. It is only then, finally, that Gregor himself starts explaining the mathematics of gerrymandering a hypothesis and correctly. Yet he never gets how he was screwing it up the last thirty minutes prior. Instead, now Gregor starts to apply this correct mathematics to other premises he claims my hypothesis requires; he never gets back to correcting his previous mistakes by applying it there. And in applying that math to these other premises, Gregor continues to make yet more mathematical mistakes.

Gregor’s first example is the passage in Galatians 4 that Jesus was “born of a woman.” His key mistake here is that at no point does he notice or recall that in fact I scored this evidence against mythicism, finding on the a fortiori side that its likelihood ratio was 2 to 1 in favor of historicity (OHJ, pp. 581-82; and see the table on p. 594). Instead, Gregor falsely reports in this video that I said “this piece of evidence doesn’t increase the probability of historicity.”

It’s weird that critics keep overlooking this. But even weirder a critic who knows how the math actually works and thus should not be overlooking a detail like this. Missing that detail caused Gregor to mix up two different equations (the one for the upper bound of my error margin, and the one for the lower bound of my error margin) and also double-count an impact of an uncertainty measure that was already mathematically accounted for by my upper bound equation. He hosed the entire methodology I set forth in OHJ.

To be fair, if Gregor didn’t read the book carefully, he might have overlooked my demarcation of error margins, and thought my argument for the a judicantiori side was all I argued. Because on that point, at least, I do argue that “born of a woman” has no effect on the probability of historicity. That only gets to my lower probability, though, not the upper end of my error margin which is really the only side that counts. But let’s assume Gregor had correctly said he was only responding to the argument I here made for the lower bound on my probability of historicity. Even making that correction for him, what he then says is still incorrect—because he was looking at the wrong equation!

In OHJ I do at first argue that given the evidence in totality (like the context of Galatians 3-4), this passage is almost certainly allegorical, regardless of whether historicity is true or false (and if you need more convincing, see my more thorough analysis in Yes, Galatians 4 Is Allegorical). Which mathematically means this passage can have no effect on the probability of historicity—either to increase or lower it. This produces a likelihood ratio of 1/1. As I put it in OHJ (emphasis added):

[F]or Paul’s saying that Jesus was ‘made from a woman, made under the law’[,] I showed how even in context that reads as an allegorical statement, not a literal one. And I am personally certain that’s how Paul meant it. So I believe it has a 100% probability on minimal mythicism, given that such allegories are completely expected ([given] Element 14), and given the context of the whole chapter in which he says it (and the preceding chapter as well…). But since all this is not yet commonly accepted (I am looking at the text without the presuppositions of historicity that all previous scholars have done), I will argue a fortiori by saying it has only a 50% chance of being what we’d expect given those facts. And for comparison I’ll assume that this bizarre and inexplicable way of talking about Jesus’ mother is 100% expected on minimal historicity—even though it isn’t. So again, although I doubt it, this passage might also be twice as likely on historicity.

OHJ, pp. 581-82

Note: “I showed” it was an allegory (I did not just “assume” ad hoc that it “could” be) and for that reason I am personally “certain” it is. I thus conditioned the resulting probability not on assumptions but on given facts: the proven tendency of Christian writing in allegory and the proven context of the passage in Paul’s argument and the proven attributes of its being “bizarre and inexplicable” otherwise. Therefore, if all that’s the case, then the probability of that verse’s content is effectively 100% expected on mythicism. I then assumed it’s just as likely on historicity (as nothing in Paul’s allegory requires Jesus not to have had a literal mother). Which means it’s equally likely on either historicity or mythicism. Therefore, 1/1.

And yet, I then point out, that to control for the possibility that I’m somehow mistaken in these facts, I nevertheless assume that maybe this reference is still more expected on historicity than on mythicism, thereby accounting for some appreciable probability that it was meant literally and not allegorically, which would weigh in favor of historicity. This cuts against all the evidence. But that’s why it produces my a fortiori estimate of probability. It’s also why I weigh it so weakly—if this reading didn’t cut against all that evidence, it would weigh a great deal more in favor of historicity than merely 2/1.

Gregor means to argue here that I don’t specify the probability that corresponds to my certainty, and if that probability is low enough (for instance, he enters a mere “90%” in his diagram, which is indeed a rather low level of certainty), it can show up in the math (even if minutely). Of course, I round all concluding probabilities to whole percentiles in OHJ (cf. pp. 248), and I define “virtually certain” in my canon of probabilities (in Proving History, p. 286) as 99.9999% (not 90%), which is what I always mean by default if I call something “certain” in OHJ. So even if Gregor was looking at the correct side of my confidence interval (and he wasn’t), the impact of this consideration on the prior probability would be so far below rounding error it simply would have no observable effect on the math.

Even if I could be persuaded to imagine the odds of Paul’s intent being allegory is only 1000 to 1 rather than a million to one, that’s still 99.9%. Still below visible effect. The difference between 67% and 66.933% (67% x 99.9%) is pointless to even mention. Even in general; the more so when my margin of error (my confidence interval) is over 33 percentiles large! A variance of a fraction of a percent is meaningless at that resolution. Because it is entirely washed out by the margin of error.

And that’s really where Gregor made his biggest mistake. Gregor was incorrectly talking about the lower boundary of my confidence interval, and was here proposing maybe my result should be adjusted toward historicity by the uncertainty in my conclusion about Paul’s use of allegory in Galatians 4. But the upper boundary of my confidence interval already adjusts my result that way—for my prior probability, by over 26 whole percentiles. That means the difference Gregor is talking about is mathematically invisible, washed out beneath my error margin.

Indeed I already accounted for this in my estimates for this specific item of evidence. On the a fortiori side I had already adjusted the 1/1 odds for this evidence to 2/1. If we apply Gregor’s oddly low confidence in allegory here of only 90% then we would get, as he indicates, an updated prior of P(m)(0.9) = (0.67)(0.9) = 0.603, bringing a 67% down to 60%. But in fact, I adjusted the probability already by a factor of fifty percent, 0.5, for (0.67)(0.50) = 0.335, bringing a 67% down to a dismal 34%, a far greater reduction than even his 0.9 produced. Gregor has thus failed to notice that I already accounted for what he’s talking about—and was even more generous to historicity than he was!

Gregor’s claim then that I am “smuggling in additional propositions without declaring them” is therefore false. I openly declare them; I then extensively demonstrate they are factual truths to very high probabilities (and not mere assumptions or hypotheticals); and I already fully account for the mathematical effect of the resulting uncertainty. So Gregor thinks he is schooling me in something that in fact I’ve already done in the actual math, right there visibly in the tables in OHJ.

Gregor’s mistake here is particularly apparent if you notice that he uses as the “prior probability” the value of 67% when he “adjusts” it with his 90%, but that 67% is my upper bound, to which I never apply the 1/1 likelihood ratio for Galatians 4:4 that he is attempting to account for. So it is mathematically incorrect of him to do that. He’s adjusting the wrong equation. Since he was only talking about my 1/1 ratio, he was talking about the likelihood ratio I apply at the lower bound, whose prior probability is in fact ~6% for historicity (OHJ, Ch. 6.8), which means ~94% for mythicism. So even given what he meant to do, his equation should have been P(m)(0.9) = (0.94)(0.90) = ~0.85. But since by “certain” I actually mean not his 0.90 but higher than 0.999, the actual math would be closer to (0.94)(0.999) = ~0.939, a variance from 0.940 so minuscule as to be wholly insignificant. Rounding to the nearest whole percentile, as I do in OHJ, the result is exactly the same: 94%.

However, even that would be incorrect. Because Gregor is double counting. He failed to notice that I already accounted for the uncertainty of understanding Paul as writing allegorically in Galatians 4 in my a fortiori estimate of the likelihood ratio for that evidence. It is therefore mathematically invalid to “double count” that uncertainty by multiplying it against the prior probability again. To the contrary, it was already accounted for in my assignment of a 2/1 likelihood ratio in favor of historicity for this evidence.

These are the kinds of mistakes that are easy to make, especially for people who don’t know math as well as Kamil Gregor does. So it will be valuable, for anyone who really wants to grasp all this, to study his mistakes. You’ll learn a great deal about the method and the math that way.

Continuing the Same Mistakes

Gregor then types up a table (around minute 48) where he claims I add many ad hoc suppositions without accounting for them, and lists them. He commits all the same mistakes here. He confuses my lower bound equation with my upper bound equation, and thus multiplies the wrong priors by the wrong likelihoods; he is double counting uncertainty measures, which have already been counted in the upper bound likelihoods, and thus he is wrongly reducing the prior when those likelihoods have already been raised specifically to address what he’s talking about; and he keeps assuming I “sneak” in and “assume” premises that in fact I actually openly state and defend as highly probable facts.

These include the Cosmic Seed of David, for example, which is identical to my treatment of the allegorical mother: on my upper bound (the only one that uses the prior probability Gregor shows on his table) I weigh 2 to 1 in favor of historicity, not 1 to 1, thus fully accounting for any uncertainty in my reasoning for assigning a ratio of only 1 to 1 on the lower bound. So Gregor is incorrect in saying I don’t account for it, and doubly incorrect in his counting it twice in the same equation! And of course I do not sneak in any hidden premises even to arrive at the 1 to 1 likelihood ratio on the lower bound, but explicitly demonstrate the facts in question that entail that conclusion. Which are not conjectures but data firmly extant in the background knowledge, b, on which all values in Bayes’ Theorem must be conditioned. In this case, the Nathan prophecy, which in conjunction with facts of history well known to Jews of the time, firmly entails any messiah, cosmic or otherwise, had to be made from a seed of David.

Likewise Brothers of the Lord: I already account for Gregor’s uncertainty factor by adjusting the 1/1 to 2/1; and I don’t just assume things to get to the 1/1, either, but demonstrate them in evidence. So Gregor is wrong again. He also weirdly asserts something inexplicable here, claiming that I argue that the brothers passages support historicity even if Paul means they are fictive and not biological brothers. I cannot fathom why he thinks I argue that, or how that argument could even make sense. Obviously if Paul means fictive brothers in both passages, these passages afford no evidence for historicity (OHJ, Ch. 11.10).

Similarly, Gregor wants to separate “made of a woman, under the law” as two separate things to interpret allegorically, but exactly those two things are declared allegorical, together, later in the same passage, in an extended two-chapter argument that is allegorical throughout, so these cannot be independent probabilities. If one of them is allegorical, the other is far more likely to be as well, than it would otherwise be alone. But even apart from Gregor overlooking that, since I already adjust that 1/1 to 2/1, my variance more than exceeds even what his would be if we kept his probabilities as they are. For he would thus assign an uncertainty factor of 0.9 x 0.9 = 0.81; I already assigned it an uncertainty factor of 0.50!

Then Assigning Certainties Too Low

Then Gregor throws in the phrase “was betrayed” as something we can be measurably uncertain about. But here he is simply wrong as a matter of fact. It is indisputable that Paul repeatedly uses the same phrase to explicitly describe God delivering Jesus up and Jesus delivering himself up; there is therefore zero reason to interpret Paul in any other way. This requires no additional premise. It’s simply a fact of the Greek language and the text of Paul. The probability of being wrong about it is literally astronomical.

The result is an insurmountable dependent probability: P(Paul means God/Jesus delivering Jesus up|background knowledge) = P(Paul means God/Jesus delivering Jesus up|Paul repeatedly says so and never the contrary) > 99.9999%. Adding in the fact that Paul demonstrably has no knowledge of the Judas tale (given that there is no missing Apostle in 1 Corinthians 15:5) and the abundant data from the Gospels that further cast that tale into suspicion (Proving History, pp. 151-55) only raises that certainty further off the charts. So there is no measurable uncertainty here that can possibly affect my results in any observable way. The variance would be an invisible microfraction of a percent. Safe to ignore.

The same goes for Paul’s attributing the Eucharist to a vision. Though there I might assign a certainty a few thousand times lower, but still no lower than 99.99%, for much the same reasons (the evidence is detailed in the previous link even more than in OHJ), and this has exactly the same consequence: no visible change in the results, which are rounded to the nearest whole percentile, and thus unaffected by variances in the tenths or hundredths of a percent.

Gregor then inexplicably places the Ascension of Isaiah on the table. I don’t know why. The only effect I estimate that has on the probability of historicity on the a fortiori end (the end that would thus already be including the uncertainty factor Gregor is now trying to double-count again) is minuscule. I assign a likelihood ratio to the combination of the content of that work and a lost text quoted in Ignatius, which reinforce each other, of only 4 to 5 (OHJ, p. 357, cf. pp. 320-23); I think the effect of each by itself would be much less than both happening to be found, but even if we allowed an even split of the weight (which would be over-weighing each), that means a likelihood ratio of barely 9/10. And since all uncertainty has already been included in arriving at that value, Gregor has no basis for adding it in again. Unlike the evidence above, this evidence has no historicist interpretation; it is entirely evidence for mythicism—and merely a question of how strong. As I assign virtually no strength to it, I cannot see any more uncertainty to account for here.

Gregor also does not seem to know that we have manuscripts of the Ascension of Isaiah that lack the inserted Gospel he refers to (more evidence he has not read OHJ carefully?). Gregor seems to think we are merely conjecturing it was an interpolation. The evidence is much more conclusive than that. We have both internal textual and external manuscript evidence. And I provide a Bayesian calculation of their net effect in OHJ:

[E]ven if we imagine that the prior probability of [any] version of the eleventh chapter of the Ascension of Isaiah being [in]authentic is as low as one in a thousand (an a fortiori rate of interpolation in Christian texts: see note in Chapter 11, §8), the evidence is so overwhelmingly improbable on the assumption of authenticity (a million to one at best) that we can be certain what we have in that chapter now is not what was in the original. The original text cannot have been either (a) the elaborate pocket gospel, which fails to correspond with the preceding material in too many ways to be even remotely likely, or (b) the version that lacks that (but has in its place a completely different 11.2 and a revised 11.23), which not only flubs the sequence of events (by deleting Jesus’ expected glorious appearance to the angels of the firmament), and is implausibly brief (given the verbosity of the rest of the text), but also fails to contain any of the events that previous chapters told us would be recounted here.

[8] The defects of the inserted gospel are even greater than the defects of the version that fails to mention what we expect (and which not only fails to explain what happens after 11.2 but flubs 11.23), and thus the one is even less probable on a theory of authenticity than the other. Yet in either case the likelihood of having the text in the given state we have, if what we have is an unaltered text, is in my opinion no better than a million to one against. With a prior probability of a thousand to one the other way, that gets us odds of a thousand to one in favor of my conclusion—that the text in both traditions is missing what it originally contained. Using the odds form of Bayes’s Theorem (see Chapter 12, §1): P(missing|e)/P(¬missing|e) = 1/1000 x 1,000,000/1 = 1000/1.

OHJ, pp. 44-45, incl. n. 8

Again, that’s a certainty factor of 99.9%. Which can have no visible effect on the math, especially upon a 9/10 likelihood ratio. Since the variance from my lower bound to my upper bound is already nearly 0.50, a much larger uncertainty factor than even Gregor’s 0.90, we needn’t add more.

Similarly with respect to 1 Thessalonians 2 containing an infamous interpolated passage, widely-acknowledged as such by mainstream scholars for decades. Gregor falsely claims I “don’t say” what the probability of that is. I most certainly did. I actually produce a Bayesian certainty check on that conclusion in OHJ:

In Bayesian terms, the consequent probability of interpolation is so much higher than authenticity that it would far overwhelm any prior probability based on the known frequency of interpolation. For example, any sound analysis will find the known rate of interpolation in the NT is higher than 1 in 1000 verses per century (counting both interpolated passages and verses with interpolated text within them), and we have at least one whole century of no manuscripts to check by, so at least 1 in 1000 verses in the NT are or contain interpolations undetectable in extant manuscripts (there are nearly 8,000 verses in the NT, so this means at least eight interpolations in the NT will not be detected in extant manuscripts).

But the probability that Paul would write vv. 15-16 on known background evidence is easily millions to one against. In the main text I identified five unlikely features, one of which is extremely unlikely (which I’d estimate can’t be any more likely than 1 in 10,000), and the others very unlikely (no more likely than 1 in 10 apiece, for total odds against of 1 in 10,000), which combined makes the ratio of consequent probabilities 1 in 100,000,000 (one in a hundred million). When this is weighed against a prior of 1 in 1000 against interpolation, the odds that Paul wrote this come out to be less than 1 in 100,000 (odds of authenticity = 1000/1 x 1/100,000,000 = 1/100,000).

Even if the five counts against it have a probability of 1 in 100 (for the least likely) and 1 in 4 (for the remaining four, for 1 in 256 odds), making a ratio of consequents equal to 1 in 25,600 against a prior of 1 in 1000, that leaves us with odds of less than 1 in 25, or about a 4% chance the passage is authentic (odds of authenticity = 1000/1 x 1/25,600 = 1/25.6). And those latter numbers are surely unrealistic.

OHJ, p. 569 n. 73

Again, 1 in 100,000 entails an uncertainty factor of 99.999%, which is again wholly invisible in its mathematical effects under a confidence interval over 33 percentiles wide.

It’s even more incorrect of Gregor to include on his table that Josephus has been twice interpolated. Not only because that is more than a million to one certainty given the extent of the evidence, so its variance won’t even show up under a thirty percentile margin of error, but even more so because that’s irrelevant to the likelihood. The same goes for his inclusion of Tacitus (which he even repeats again incorrectly later in the video): neither passage carries no weight because they are or might be interpolations; rather, neither passage carries weight because neither can be established as independent of the Gospels (and Gospel-dependent Christian legends). Dependent evidence has zero value in Bayesian estimates of likelihood (OHJ, Ch. 7.1). Gregor somehow missed this.

It’s also inexplicable why Gregor would include Pliny on his chart. Pliny never says anything pertaining to the historicity of Jesus (OHJ, Ch. 8.10). Citing Pliny is like citing today’s Pope. “He mentions worshiping Christ, therefore that increases the probability Jesus existed” is a wholly defunct argument with exactly zero weight. In any era. I’ll note only as an aside here that Gregor is also a bit too gullible regarding the dubious stylometric study arguing Pliny’s letter in question was forged; not only because that study’s author (Enrico Tuccinardi, who has published other dubious work, IMO) did not even argue it was forged (he says his own data actually contradict that hypothesis) but that it “has interpolations,” yet I suspect the same data can be explained by that letter being the only one pertaining to religion in Pliny’s state correspondence (see Hurtado for perspective).

On top of all that, around minute 56 Gregor rightly shows that historicists also face some of these uncertainty factors—they have to come up with a lot of excuses for peculiar evidence, too (and a lot more than he mentions: e.g. OHJ, pp. 28, 308, 335-36, 374, 516, 520, 528, 566, 585, 593). And I do think their excuses tend to be much less certain than the ones I establish on a strong basis of evidence. Historicists make up quite a lot of excuses without thinking they have to adduce any evidence for them actually being true, whereas I have assiduously avoided ever doing that. But even if they were equally balanced in how well supported their added assumptions are, that would effectively negate all the uncertainty factors Gregor wants mythicism to undertake. And this is only the more the case if historicists’ excuses tend not to be well supported.

So the point I’ve already made—that all these uncertainty factors are already accounted for by my a fortiori estimates or are so minuscule as to be completely washed out under my enormous margins of error (which, contra Gregor’s assertion, I do in fact explain in OHJ, pp. 54-55, 65 n. 13, 247-48, 513)—is even more strengthened by this additional fact: that they are collectively reduced all the more by the counter-balancing uncertainty factors historicists must incorporate. Which makes it all the more obvious that the net effect of such factors on either side will be so small as to be unobservable at the resolution I’m working with in OHJ. So Gregor is wrong. There is simply no reason to account for them. It’s a massive waste of time—like insisting we count the angels on the head of a pin, when actually the only thing we need measure is the head of the pin.

(I should also add that it almost looked like Gregor wasn’t aware that any reduction in the prior probability of mythicism entails a like increase in the prior for historicity, and vice versa—as they must always sum to 1; I’m sure that was just a misperception, but anyone attempting to replicate his table: make sure your math takes that into account.)

The Rank-Raglan Class…Again

Gregor and Pearson’s final topic of discussion is the Rank-Raglan set as a reference class for Jesus’s historicity. This commonly gets misunderstood; evidently no one actually reads OHJ, Ch. 6, or any of Elements 29, 31, or 45 to 48 in Ch. 5. I only use the Rank-Raglan set as a proxy for mythologization generally, and I only choose it over all the other myth-heavy sets Jesus belongs to because it has the most members and thus gives us the best available estimate of a base rate. See Is Rank-Raglan Indicative? and The Rank-Raglan Class Again for fuller expositions on this.

As I wrote in OHJ:

This is a useful discovery, because with so many matching persons it doesn’t matter what the probability is of scoring more than half on the Rank–Raglan scale by chance coincidence. Because even if it can happen often by chance coincidence, then the percentage of persons who score that high should match the ratio of real persons to mythical persons. In other words, if a real person can have the same elements associated with him, and in particular so many elements (and for this purpose it doesn’t matter whether they actually occurred), then there should be many real persons on the list—as surely there are far more real persons than mythical ones. … But this is not the case. No known historical persons are on the list.

OHJ, p. 231

I then go on to find that it’s wholly unreasonable to imagine the base rate of historicity for persons in that class can be any higher than 1 in 3; and that the data as we have it is most directly compatible with it not being any higher than 1 in 16. Note the enormous size of that confidence interval—it represents a variance of 6% to 33%.

Gregor Confuses Science with History

Gregor’s first objection is methodologically invalid: he says it’s my job to “prove” I’ve presented every member of the Rank-Raglan class we know of before he can agree I have. But that’s impossible. It’s like asking a scientist studying ball lightning to “prove” they found every instance of ball lightning, when they’ve already told you they and several prior researchers (I cite three) found no other instances than their study reports. “Well, I dismiss that, because maybe there is some other instance you missed, so I can ignore your study” is simply not how science works. If Gregor wants to maintain that I’ve missed an example, it is his job to prove that claim by presenting said example. It’s likewise the whole community of historians’ job to do that, if they wish to maintain a contrary assumption to what I found. And if none ever can produce another example (and it’s been years now and none have), that’s the only proof one could ever have that indeed they found every example. I’ve met my burden of proof here. There is no other way to meet it.

Gregor then shifts to an argument about random sampling. How do we know the only cases that survive in the record are a proper random sample of all cases there ever were? As indeed, there are likely to be some lost examples; just as with the scientist studying ball lightning, there will be instances unrecorded that she therefore could not include in her study. Here Gregor confuses historical methods with scientific. We are not polling a population of voters (his example). We are evaluating trends in history using the only data we have. The results will therefore always be less precise—meaning, they will have larger margins of error—because the data will be scanty (particularly for ancient history; in most cases, and certainly this one). Gregor needs to do history here, not science. (See my discussion of the distinction in Proving History, pp. 45-49.)

There is no reason to believe some “magic hand” has selectively suppressed all the Rank-Raglan heroes that actually existed, so as to cause our surviving sample to be entirely comprised of mythical persons. So you cannot reasonably believe that happened. Nor can you claim it did—unless you can present evidence of it. And all we need in history is to discover what is and is not reasonable to believe. What is not reasonable to believe, we ought not believe.

So without any evidence to the contrary, we have to assume the processes of history that destroyed any examples from this set was random with respect to the variable we are discussing, which is the historicity or nonhistoricity of each member in question. Because that’s how historical processes always work in the absence of any known selective force. It’s the principle of indifference: if you do not know of any selective force pushing the set one way, and you do not know of any selective force pushing the set the other way, then the probability of either is the same, and those probabilities cancel out. You are left with the set you have, not a single member of which is historical.

Gregor also has unrealistic expectations about how we do history, by continuing with his bad analogy of voters; there are tens of millions of voters, whereas it is essentially impossible that there were ever more than a few dozen Rank-Raglan heroes in antiquity. So a sample of fourteen is almost certainly a very large part of the whole set. It’s therefore more analogous to having a poll of not 14 voters, but say, 30 million out of 100 million voters—if we’re historians in the future and, let’s say, only 30% of voter records had survived a nuclear war. We have less than a third of the complete set. But no way would Gregor be poopoohing the result that none of those 30 million voters had voted for the Green Party, therefore we can reasonably conclude none or next to none of the other 70 million did. The odds would be astronomically against our having ended up with a biased sample so huge that merely “by accident” none of the “actual” 40 million votes for the Green Party survived. Nuclear bombs don’t selectively target voter records with the word “Green Party” in them.

This analogy is closer to what I’m actually doing than Gregor’s analogy is, although it still falls short in one respect: at the scale of millions, the probability of random bias is going to be much lower than at the scale of dozens. So my margin of error does need to be much larger in the Rank-Raglan case. And lo and behold, that’s exactly what it is. I am already accounting for what Gregor is talking about.

This is again like dismissing a study of a rare phenomenon like ball lightning, on the argument that it didn’t include a thousand examples of it; when in reality, there aren’t but a few dozen recorded. Gregor’s standard is simply impossible to meet, and thus constitutes irrational doubt. In history, it is the norm to have a small number of examples to ascertain trends from. Less than a dozen, indeed less than six, is practically the mode for historical claims about trends and generalizations. And often even if we could have the total set, it won’t be much larger than that. And yet we still have to reach conclusions. It’s absurd to say “we can conclude nothing about a phenomenon in history if there were never more than five repetitions of it.” That’s simply false. Five repetitions is actually startlingly a lot in history, and almost always indicates a trend we need to pay attention to, not ignore. Historians would be delighted to have fourteen!

And here is where we discover Gregor would be a terrible gambler.

Suppose for a moment that the actual base rate is 80% (8 out of 10 persons in the Rank-Raglan class were historical), but we just accidentally, by random chance, only have a surviving record of 14 non-historical ones, thus deceiving us as to the actual base rate of that happening. The probability of that being the case is (1-0.80)^14 = 0.20^14 = 0.00000000016384, which is tens of billions to one against. So we can be pretty sure the base rate was not 80% on this evidence. Suppose, instead, it was 50%; then it’s 0.50^14 = 0.00006103515625, which is tens of thousands to one against. Okay, also not likely. Suppose it was 33%; then it’s (1-0.33)^14 = 0.67^14 = 0.00367322251629, which is roughly 272 to 1 against, a mere fraction of one percent. Still unlikely. But much more likely than the others. You can see where this trend is going.

If we rule out rates too low for only 14 instances to determine, the most probable base rate remaining that’s compatible with this observation turns out to be 1 in 16. This is known as Laplace’s Rule of Succession, and it was derived from Bayes’ Theorem. Indeed, that rate would make the probability of the observed evidence (1-0.0625)^14 = 0.9375^14 = 0.405133232869595, or roughly 41% likely. That’s pretty good. Lower rates would even more likely explain our observation, of course. For example, if the base rate were 1 in 100, then it would be (1-0.01)^14 = 0.99^14 = 0.868745812768978, which is 87% likely. And if the base rate were 0, that we’d observe zero instances in 14 cases is obviously 100% expected. And so on.

You can graph every answer this way, for every possible base rate along the horizontal axis, and each rate’s probability along the vertical axis, and measure the relative area under the resulting graph, to determine how likely it is the base rate falls between any two values when what we observe is 14 instances without a single historical person. What would the probability be, for example, of the true base rate lying between “0” and “1 in 3” given our observation of zero historical members in all of the 14 that are by chance found? There is an equation for that, of course, which I don’t know off hand, but you can already tell the result is going to be high: the entire long tail from 0.3% probability on down covers a minuscule area, whereas the enormous body of the graph that trails the other way, from 0.3% up to 41%, 87%, 100% consumes almost the entire area under the graph.

There can therefore be no sound argument that the base rate of historicity is higher than 0.33, for any population from whom we observe a surviving sample of 14 members and fail to find a single historical member.

Another way to think of this is as a gambler. Suppose you are faced with a slot machine whose odds of a payout you don’t know but have to guess from observing. You pull the lever fourteen times, and never win. How likely is it, do you think, that the odds of a payout were 1 in 4 that whole time, and you were just really unlucky? Well, that’s (1-0.25)^14 = 0.75^14 = 0.017817948013544 = ~1.8%, or roughly 56 to 1. You’d sooner draw an ace of spades straight off the top of a freshly shuffled poker deck. Would you bet much on that? Not really. So if you wouldn’t bet much on that, why would you bet anything on the base rate of historicity being 1 in 4 on the same evidence? And yet for the rate to be 1 in 3 is even less likely, at 272 to 1!

So you cannot possibly say it’s reasonable to think the base rate of historicity among Rank-Raglan heroes in antiquity was any higher than 1 in 3. You’re already pushing hard against common sense to accept that it was even as high as that. Gregor would have to be a truly terrible gambler not to agree. His objections here, therefore, are wholly unfounded and unreasonable.

And this is just from Jesus’s membership in the Rank-Raglan class. He actually belongs to several myth-heavy classes, such that he is actually one of the most mythologized—and indeed most rapidly mythologized—figures in all of history. Which is what makes Jesus less likely to be historical than any other claimed figure. The only question we need answer is “at what base rate.” The Rank-Raglan class gives us a test case of how frequently people that mythologized turn out to be historical. The answer is: almost never. There is no other reasonable conclusion to make.

This is why it is not relevant what period Rank-Raglan heroes were set in. Gregor falsely claims I don’t address that objection; in fact I do. The Rank-Raglan set shows what was typical for heavily mythologized heroes. It thus gives us an applicable base rate to all other as-heavily-mythologized heroes. The most one could attempt to argue is that maybe it’s “harder” to create a mythical hero as rapidly as Jesus was, and therefore the base rate for the Rank-Raglan class wouldn’t apply to him; but that is demonstrably false: such rapid development of mythic persons and events was quite common, even in more adverse conditions, and thus no evident barrier to it existed. And contrary to what Gregor claims, I thoroughly prove this in OHJ (Ch. 6.7, and with other analogs like Ned Ludd, John Frum, and the Roswell saucer, all in the index, and to all of which, as I show, Jesus is highly analogous). So even that objection cannot gain purchase.

I’ll conclude with noting that this base rate only applies to, and can only reference members of the Rank-Raglan set from, the ancient milieu; as I have explained many times before, the content of the Rank-Raglan set in other, significantly different eras of history will probably not represent the base rate in antiquity, so we cannot use them. Thus, the rate of historicity within “Rank-Raglan membership” in modern cinema or industrial-era dictatorships is wholly inapplicable to what it would signify in the Greco-Roman era. I explain the reason we cannot use the base rates of phenomena in different periods of history like this (when it’s a fallacy of anachronism) in Proving History, p. 245; and I discuss the whole method of deriving base rates from small reference classes—because we historians often have to—on pp. 239-45.

However, I also discuss there when we can use base rates from different eras: when we can show the conditions are suitably analogous across eras. Thus, for instance, the Ned Ludd, John Frum, and Roswell cases all prove rapid legendary development is possible even in conditions far more hostile to its success than obtained in the first century; therefore it cannot be argued it was somehow “harder” for similar legends to arise in antiquity at the same pace. Likewise, the anthropological evidence of revolution cults (OHJ, Element 29, Ch. 5) crosses many cultures and eras, and thus establishes an anthropological trend, which must therefore be anchored in universal human psychology and social dynamics, which means when the exact same pattern is observed in Christianity, we can justifiably conclude the same dynamics were operating. Because it would require appealing to extraordinary coincidences to assert otherwise, which is never sound reasoning.

Minor Errors of Fact

That completes my survey of Kamil Gregor’s mathematical mistakes, and major factual mistakes. The following, by contrast, aren’t major points and not mathematical but they require correcting, since if anyone intends to watch the whole video, I don’t want them misled:

At one point Gregor makes the incorrect statement that if Jesus didn’t exist “we don’t know” when the religion began relative to when Paul was writing. That’s false. The Epistles of Paul entail the religion began no more than 20 years before his Epistles were written; they themselves mention sequences of events and spans of years that allow no other interpretation. So that Paul would have lived in the time of Jesus and knew people witness to his life is a fact, no matter when you date the Epistles of Paul (except of course if you are so radical you deem them forgeries, but that’s improbable).

At another point Gregor makes the incorrect statement that the original Apostles rejected Paul because he was “destroying their religion” by “saying you don’t have to follow the Torah.” Both statements are false. Paul did not teach that. Paul taught that Jews were still bound to their promise to God to abide by the Torah (signified by circumcision, establishing that covenant), and it was only Gentiles who could join the community of the saved without following Torah. In other words, Paul merely taught that Gentiles didn’t have to convert to Judaism to be Christians; whereas the first Apostles had been teaching that they did.

And eventually the first Apostles accepted Paul and his teaching (possibly because it was bringing in droves of followers and thus cash, a fact Paul references several times in his Epistles). The Apostles therefore did not condemn Paul as an outsider or an enemy but as an ally and their equal (thus he could claim so, uncontested, in such passages as 1 Corinthians 9 and Galatians 2). He had only been an outsider and an enemy, someone “destroying their religion,” before his conversion, when he was a Pharisee (1 Corinthians 15:8-10, Galatians 1:22-24, and Philippians 3:5-6).

Finally, Gregor concludes by vaguely arguing Jesus is more like the Josephan Christs than the Rank-Raglan heroes (if one looks closely at the Venn diagram he presents near the end), without seeming to be aware that I already addressed and refuted that argument in OHJ (I have a whole section on it: Ch. 6.5). But more importantly, he forgets that other persons were just as rapidly invented in Christian lore and thus Jesus has many mythical companions from that century, including Mother Mary, Mary Magdalene, Judas, Thomas, Barabbas, Nicodemus, Dismas, Gestas, Simon Magus, Elymas, Thecla, and subsequently Longinus and Melchior, Caspar, and Balthazar, as well as countless Medieval Saints who actually never existed, but we’ll relegate that to a Medieval phenomenon, as we have enough ancient examples to make the point: Gregor is simply factually wrong that Jesus is alone in this.

Conclusion

Kamil Gregor gets a lot of facts wrong, and that trips him up a lot. Here I’ve corrected all the errors I noticed. But more unfortunate is how he gets even the math wrong—despite having such a good handle on the mathematics of Bayesian reasoning. And I think others can benefit from studying his mistakes, to produce a stronger understanding of the math and method.

Gregor mistakenly thinks I rely on ad hoc assumptions to “up” the likelihood of evidence on mythicism without correctly paying for it in a lowered prior probability, as one would have to—but in no example he gives did I do that in OHJ. As I just showed, in every single case, the likelihood I figured follows from established background facts, not unevidenced assumptions, and the uncertainties in that are already factored in when I assign my upper bound estimates of likelihood, or are so small as would have no visible effect and thus need not be accounted for. It’s thus ironic that Gregor also mistakenly thinks minimal mythicism is unfalsifiable because you can import ad hoc assumptions to “up” the likelihood of evidence without paying for it in a lowered prior probability. Between the first half hour and the rest of the video he is simply contradicting himself by maintaining both assertions. Neither is correct.

Gregor would do well to study how I’ve proved mythicism is falsifiable a dozen times now, by showing what evidence would make Jesus’s existence probable again, with numerous analogies of how this happens for other figures, from Spartacus to Pontius Pilate. See my article on Hannibal for a full explanation. And in my article on Spartacus, the first paragraph references or links to a dozen other examples I’ve analyzed to demonstrate the point. It is not logically valid to say “Jesus is different from them, because maybe we could explain away any of that evidence if we found it for Jesus.” Because any such explanation would itself be improbable, and thus could not rescue a claim Jesus didn’t exist, any more than it would for them. At no point in OHJ do I rely on any improbable excuse like this. It’s simply a fact that if we had for Jesus the evidence we have for these other figures, the historicity of Jesus would be as assured as theirs.

And that is why we doubt the historicity of Jesus.