What Happens When You Review Richard Carrier

A lot more has been blogged that relates to the topic of mythicism. I should begin by highlighting the article by Daniel Gullotta which is now available in the Journal for the Study of the Historical Jesus. It provoked the kind of response that one would expect from Richard Carrier, who depicted Gullotta as incompetent, but not as bad as me.

Deane Galbraith tackled the way Carrier responds to critics by using his own tools against him. First, he asked whether Carrier calculated the probability that a negative reviewer was an Evangelical using Bayes’ Theorem. He then followed up by sharing Carrier’s reply, which said – hysterically – that Christina Petterson (the reviewer in question) could be identified as an Evangelical because, among other things, she agrees with James McGrath!

As it happens, Petterson is an atheist. This isn’t just a mistake on Carrier’s part, however. It shows his strong ideological bias, and his inability to evaluate those who disagree with him either clearly or honestly. As Galbraith writes:

Petterson…is not a Christian, and not religious, and never has been. She is an atheist.

This provides a good test, however, of Carrier’s inability to interpret his sources, and his ability to draw inferences from them that are simply not there. Carrier consistently assumes that anyone who disagrees with him must have an evangelical “agenda”. Sadly, this is conspiracy-theory thinking, not scholarly thinking.

Carrier has completely failed to interpret his source, taking inferences from it that simply were not there, and which were quite incorrect.

None of this, of course, is news to those who’ve interacted with Carrier at length in the past.

Of related interest, Larry Hurtado quoted from a recent National Geographic article, which says (pp. 41-42):

Might it be possible that Jesus Christ never even existed, that the whole stained glass story is pure invention?  It’s an assertion that’s championed by some outspoken skeptics–but not, I discovered, by scholars, particularly archaeologists, whose work tends to bring flights of fancy down to earth.

See also his discussion of Galatians 1:11-12, and Philo’s reading of Zechariah, both of which also relate to mythicist claims.


"How would you have responded, if the question had been posed to you?I'm pretty sure ..."

Resurrection and Hope
"When we see God as one, it has a beneficial influence upon our minds, our ..."

Can a Muslim Follow Jesus?
"One thing I'm surprised hasn't been raised in the "Do Christians and Muslims worship the ..."

Can a Muslim Follow Jesus?

Browse Our Archives

Follow Us!

What Are Your Thoughts?leave a comment
  • Phil Ledgerwood

    “It shows his strong ideological bias, and his inability to evaluate those who disagree with him either clearly or honestly.”

    To me, this pretty much sums up the entire enterprise of using Bayes’ Theorem to evaluate ancient historical claims.

    • John MacDonald

      I find Carrier to be a very Black and White thinker: There is only one way to do things, his way. Any other way is illogical. Even someone arguing for historicity has to do it his way. See https://www.richardcarrier.info/archives/13352 . He likes Bayes theorem math, so that is the only way to model and evaluate probability claims. The ironic thing is that contemporary mathematics, like the kind children are now being taught in schools, emphasizes exploring multiple and creative paths to get to the answer.

      • Phil Ledgerwood

        The thing I’ve appreciated about Carrier’s attempts to use a (modified) form of Bayes’ theorem is to bring out into the open that historians have their biases (which we knew) and force a certain amount of articulation around them. If I put a number next to my assumption, you can clearly see how much weight that assumption carries in my conclusions. I appreciate this, actually, as the humanities are notorious fields for being able to say things like, “It is most likely that…” and not show your work. That statement of probability covers up a lot of assumptions and weights that are no longer visible to the reader.

        Obviously, you don’t need Bayes’ theorem for that, but it at least forces the issue. Carrier, to his credit, makes a pretty big deal out of this aspect of using the theorem, which is good, because he uses a modified version and has been critiqued for that, but really, the theorem doesn’t matter for this purpose. He could have just as easily made a table with all of his assumptions and some kind of numeric scale attached and accomplished something similar.

        I think the problems start arising when we start thinking that using Bayes’ theorem as a mechanism for making our biases transparent somehow makes our conclusions more legitimate. I’m a Lean operations consultant, and we have a running joke: “I must be right. I used actual math.” When people read Carrier, they don’t go, “Oh, great, I can now see clearly mapped where Carrier’s biases are and how they effect his reasoning and conclusions,” they go, “See? Bayes’ theorem proves Jesus didn’t exist.”

        No, what Bayes’ theorem proves is, if you start with Carrier’s assumptions and weights, Bayes’ theorem will support Carrier’s conclusions. And this is where things start going off the rails and where your critique starts to kick in. BT is no longer a format for making my biases and weights visible, it is now a simulation for the likelihood of Jesus existing and – surprise, surprise – the simulation says it’s unlikely. Given how the formula was fed, this conclusion is basically tautological, but that’s not how it’s treated – certainly not by the Carrierites and, debatably, not by Richard himself. It’s as if using the theorem makes his conclusions more secure in some way, and historians ignore this “powerful tool” to their peril.

        • John MacDonald

          Carrier also imports traditional deductive reasoning language into his arguments, suggesting “certainty” of his arguments, employing such words like “prove.” For instance, in OHJ, Carrier says “And they often simply mishandle the evidence, such as assuming the Gospels are historical narratives rather then symbolic myths … despite conclusive evidence of the latter, as I’ll prove in chapter 10” (Carrier, OHJ, 12).

          • Phil Ledgerwood

            Oh, epistemic hyperbole is basically the English language as far as Carrier’s concerned.

          • John MacDonald

            It’s only hyperbole if he doesn’t take his extravagant claims literally, lol.

        • arcseconds

          While what you say is basically correct, it kind of gives one the impression that Bayesian formalism is kind of ’empty’, i.e. you just get out what you put in, in the first place.

          In a sense that’s true, but one of the things you ‘put in’ is the evidence, so the system of probabilities being generated actually does create a new result, and this result may well be a surprising one.

          (The price at the till for your groceries might be surprising, too, even though it’s in some sense contained in the prices of the items you put in your shopping bag.)

          Also, the process of making reasoning more transparent itself may end up with someone re-evaluating their position.

          There is a view in epistemology/philosophy of science that Bayesian confirmation basically is inductive logic, this is what it is to be a rational agent learning from evidence. (whether or not it’s practical in any way to use the formalism for reasoning in a real-world case is a different matter). It would be a significant blow to this view if it could be shown that there’s a real sense in which a Bayesian agent can’t discover anything more than they already believe, and in fact there are results that point in the other direction.

          • Phil Ledgerwood

            While what you say is basically correct, it kind of gives one the impression that Bayesian formalism is kind of ’empty’, i.e. you just get out what you put in, in the first place.

            Well, that is basically true. The Bayes’ formula is just that – a formula – it only describes the relationships between data. It may surprise you how the quantities end up being distributed, but Bayes’ theorem cannot produce some result outside your known boundaries. Between possible options A, B, and C, once we plug in the correlative numerical data, it might surprise me that C is more probable than A – an assumption I might have held apart from the data – but the theorem cannot come up with an option D, nor can it generate any distribution apart from the data I supply. Although you are plugging in “the evidence,” that evidence is a quantity. There is nothing else you can plug into BT except numbers, so whatever evidence you have must be converted to numerical values in some way.

            In fact, there is virtually nothing the theorem can produce that you couldn’t have arrived at -without reference to the theorem at all- if you had sufficient iterations to study. Even in very complex scenarios, Bayes’ theorem is more or less a mathematical shortcut that allows us to make projections in the absence of extrapolating data from a very large number of actual trial runs. This is the inductive aspect you mentioned. In essence, the theorem is a stand in for running 800,000 experiments and allocating probabilities that way.

            Even in a Bayesian epistemological view, the establishment of the probability of a belief is still only as good as the numerical evidence can contribute to establishing the probability. This is why, in my opinion and (I believe) the opinion of virtually anyone who uses Bayes’ Theorem in their actual work, BT is not appropriate for scenarios that are wholly unique, or the possible outcomes are unbounded, or the evidence is not easily quantifiable.

            To do this with a historical proposition is intrinsically unwieldy, and you can see Carrier struggling with it in his own books. For example, he has to provide some sort of measurable rate to plug in. So we get things in his writings like “the rate of people who conform to the Raglan hero pattern who don’t really exist” or “the amount of people written about who do not exist.” It’s very hard to find a meaningful quantity to plug in. Why not “the rate of Jewish figures written about in the first century who did not exist” or “the rate of figures to whom miraculous deeds were ascribed who did not exist” or “the rate of men whose names start with ‘J’ who turned out not to exist” and so on? It’s just astounding to think this line of inquiry could produce anything anyone could possibly care about.

            Technically, the worst moments from a statistical standpoint are when Carrier is assigning numbers to his assumptions. Obviously, we do not have a statistical probability on most of what Carrier plugs in, so he just declares those values according to his estimation. The irony comes in for me because I actually think this is perhaps the only value trying to use something like BT can offer historical probability – a quantified list of the author’s assumptions and how much weight they carry toward establishing their conclusions. Although it is presented under a very awkward rubric, I think there’s a lot of value in scholars who work in less-quantitative disciplines or perhaps in any persuasive endeavor at all being able to say, “Here are my assumptions, and here’s how important they are to my trajectory of thought.” What’s more, it also calls us to greater precision than just declaring something to be “very likely” or unlikely. Carrier contends that there’s a lot of assuming and probability by fiat that occurs under the text of much scholarship, and he’s not wrong. I’m just not sure BT is the remedy for that.

          • arcseconds

            Bayesian epistemology is the position that probabilities are essentially the same thing as degrees of belief , and that Bayes’s theorem, and more generally the axioms of probability, are rational constraints on how an agent assigns and updates probabilities (or weights and adjusts their degrees of belief, which is the same thing).

            That we have degrees of beliefs seems obvious and unavoidable. We say we know certain things to be true, but it only takes a bit of reflection to realise that even of the things we say we know, we’re more certain about some things than others. I know I have a job, for example, and I know I’m typing on a keyboard, but I’m not absolutely, incorrigably certain about either. My position might have been disestablished behind my back, or I might be being drugged or deceived by Descartes’s demon or in the matrix or something. But however unlikely it is that I unbeknownst to me that I’ve lost my job, it’s much much more likely than me being mistaken about typing on a keyboard.

            And that’s just the things we take ourselves to know – there’s also ‘being fairly sure’, suspecting, doubting, etc – lots of stuff in between.

            If we were to model this, already it would be tempting to represent degrees of belief with numbers, even if we’re not attempting to do anything normative but just represent relations like ‘much much more likely than’ (with >>).

            Also intuitively one would expect there to be something normative about how we should weight our beliefs, e.g. it seems obviously wrong to think meeting a left-handed stranger tomorrow is more likely than simply meeting a stranger.

            Anyway, you can see where this is going and I’m not going to attempt to motivate the whole apparatus.

            I don’t disagree with anything you’ve said above, but it’s not clear to me that you recognise this as an existing position, but rather seem to be thinking of Bayesian formalism in entirely statistical terms. Bayesian epistemology is a respectable position, held by smart people who have written scholarly, mathematically sound books on the topic.

            It also appears to (mostly?) be Carrier’s position, although it seems that he also says things that make him sound like some kind of closet frequentist. This may be motivating his appeal to simple numerical frequencies of a reference class that you mention, although he also presumably likes the objective nature of a numerical frequency.

            From this perspective, there is nothing inappropriate with translating “I think it’s highly likely that” into a number. Of course, no-one informed about the topic thinks it’s an easy task to reconstruct scientific reasoning / sound epistemology into Bayesian terms. Carrier seems to underestimate the difficulty and overestimate the degree to which he has succeeded. From what I have seen of his work, there’s a lot of informal reasoning that’s often questionable, dressed up with a lot of ‘surely’s and other rhetorical devices, and this is doing most of the lifting.

            One of the insights that Bayesian epistemology gives us is that different people can have different priors and even likelihoods, and therefore reach different conclusions quite rationally on the same evidence. This might not sound like news to you, but it’s certainly news to a lot of people, including apparently Carrier! I’ll come back to that in a bit.

            So even if one could in fact cast one’s entire reasoning process into probability formalism, it doesn’t mean one ‘wins’ and one’s opponents are proven to be irrational. We may just be discovering that your starting point is different.

            There are convergence proofs, so under some reasonable circumstances different starting points eventually lead to the same posterior probability, so it’s not just anything goes, although it is a bit questionable as to how much hope that really gives us.

            On the other hand, one might argue that there are more constraints than simply the axioms of probability. Giving exactly 0 or 1 to any empirical hypothesis seems intuitively unreasonable, for example, as those values are immune to revision. Perhaps too one should not be too far away from currently accepted science and scientific methodology, although this is harder to argue for from purely abstract considerations.

            Anyway, where I’m going with this bit is that Carrier seems to not understand that what he may be discovering when people don’t agree with him is not that they’re deeply irrational, but that their priors and likelihoods are different. I’m not sure how far I’d want to push this really, but it does often seem like he’s not on the same page as any mainstream scholar as far as likely interpretations of the data go. E.g. one might want to say that Carrier implicitly has a high prior to late second-temple Judaism texts (including the NT) as being esoteric works about angels, whereas mainstream historians give a high prior to them being historical in the sense that somewhere in there, some things might not be entirely made up.

            If I was going to take on Carrier’s project, I would first want to show how a Bayesian treatment can recover or represent some uncontentious examples of good historical reasoning. Then I would want to try to cast mainstream arguments in New Testament history in this fashion then see where we get to. If one deviates too far from mainstream reasoning it’s not clear what is being shown any more. Perhaps mainstream methodology is itself flawed, but if that’s the case the goal shouldn’t be so much ‘show Jesus doesn’t exist’ but ‘show how the methdology is flawed’.

            (This is one of my biggest beefs with mythicists. They usually don’t seem to realise (and won’t realise) that their methodological demands for several independent eyewitness accounts preserved in contemporary manuscripts etc. don’t just do the historical Jesus in, but basically all of ancient history. Carrier at least seems to in part think of himself as revolutionizing history, but I’m not sure what we’d end up with if we followed his example across the board.)

    • Jim

      Well possibly you (and Gullotta) just don’t completely understand Bayes’ Theorem. So for a primer, see Godfrey’s December 13, 2017 post on vridar. 🙂 🙂

      But to paraphrase and to economize on space, people just don’t understand Carrier because they don’t understand that BT is based on impressions, and Carriers “impressions” are just as valid as any historical Jesus scholar’s impressions, since we all have impressions. And Godfrey’s impression is that you need to buy both PH and OHJ; just any stats textbook won’t cut it for understanding BT.

      • John MacDonald

        What I find is that for Bayes theorem to work, Carrier still needs to plug in the right numbers based on qualitative assessments. Above I quoted Carrier when he said:

        “And they often simply mishandle the evidence, such as assuming the Gospels are historical narratives rather then symbolic myths … despite conclusive evidence of the latter, as I’ll prove in chapter 10″ (Carrier, OHJ, 12).”

        The problem is that no expert agrees that Carrier “proves” that the Gospels are symbolic myths rather than historical narratives, and so the numbers he plugs in on this point really just reflect his subjective bias.

        • Jim

          Yeah agree. I don’t think that it’s wrong to publish a study that gives the results of a BT analysis of a historical event/figure using a specified list of input parameters that were used to generate the result. But imo, a study author has to be realistic in reporting the significance of the results obtained based on their single study alone … and then tell everyone else that they are all biased.

      • Phil Ledgerwood

        I guess that’s possible, although in all those simulations I ran in the military, nobody told me I misunderstood the theorem.

        I would say that, if you think BT is based on “impressions,” then you probably grossly misunderstand Bayes’ theorem, because it’s not based on impressions; it’s based on knowing the possible ranges of your sets. This is why we can use BT for estimating the number of tanks in a location or the projected scope of damage to Seoul in an invasion of North Korea, but we don’t use BT for determining the probability of a wife loving her husband.

        • John MacDonald

          I think he was joking with you. He put two smiley faces after his comment about you not understanding Bayes, lol.

          • Phil Ledgerwood

            You’re right. I’m an idiot. These conversations breed in me a hermeneutic of suspicion. Thanks for bailing me out!

          • Jim

            A mythicist stole your teddy bear when you were a child? 🙂

          • Phil Ledgerwood

            I’m continually beset by sky demons.

      • Phil Ledgerwood

        OH, sorry, I didn’t realize you were joking at first. 🙂

        • Jim

          🙂 🙂 🙂

      • Matt Cavanaugh

        Here’s all you need to know about Bayes’: GIGO.

  • Jerry Arias

    Carrier liked the Raphael Lataster review.

  • Mark

    The Gullotta is pretty good, I thought. He should have spent more time belaboring the unrelenting evidence that Paul thinks Jesus was a human being – it becomes tedious after a while but in this context can never be tedious enough. Similarly he could have belabored more the total absurdity of thinking that a doctrine like the one Carrier imagines could have had any currency in Jerusalem. The long discussion of MacDonald and Homer, for example, is a waste of time, since it pertains mostly to the credibility of Mark’s stories about Jesus, not the existence of their subject. I had not heard before about Carrier’s counterfeiting the Rank-Raglan scale and his scale-shopping in general; that discussion is pretty devastating. https://drive.google.com/file/d/1ZwMR0IWQSKRAJYJ1m8dPzp09mqRDfZGC

  • John MacDonald

    This is interesting: Carrier, along with Reality Revolutions (whoever they are), have just released a Mobile App for debating the Historicity of Jesus. See https://www.richardcarrier.info/archives/13632

    Carrier describes the App as:

    “A mobile application that would assist in quickly investigating every argument for and against the historical existence of Jesus … By the end of the next year it should contain every argument and item of evidence for and against the historicity of Jesus, easily located and briefed, with web links for further research. It will be the ultimate tool for exploring and debating the subject.”

    • Neko

      “It’s currently in a rough launch version. It has a lot of typos and needs more content and better organization, but those things are all being worked on as we speak. A wonderful update will load by the end of January.”


  • arcseconds

    I have finally managed to get hold of Gullotta’s article. It’s a good read, and I appreciated the thorough references, particularly the ones that summarize mythicism across the years and the look at the early reception history of the crucifixion. The discussion of the use of the Rank-Raglan scale demonstrates even to a greater extent than before that Carrier has chosen (and in part created) a measure that results in his preferred outcome. It would be interesting for someone to further Gullotta’s discussion of the alternatives and see how mythic Jesus looks on those.

    (Once again, I’ll note that the Rank-Raglan business isn’t all that crticial for Carrier’s overall argument, but it is just one of a number of areas where Carrier has been at best sloppy.)

    These points occur to me:

    *) The article doesn’t specifically discuss one fundamental problem that is pervasive with Carrier: whether it’s the Rank-Raglan scale or Homeric parallels or ‘Mark is an esoteric allergory’ these all fall down if mythic elements can be added to a historical core. Obviously this is quite possible, and it’s also obvious that added mythic elements could (intentionally or unintentionally) conform the protagonist to some Hero Paradigm, or parallel Homer, or be an estoric cultic allegory, or whatever else the authors want or are otherwise inclined to do. Slightly less obvious (although still pretty obvious) is the fact that historical events can sometimes happen to fit well within whatever structure is being imposed, or can be interpreted as to do so.

    The mainstream position is that this is exactly what happened (adding mythic elements to a historical core, not the Homeric parallels etc). The consensus Jesus is already fairly minimal, and more minimal understandings are still present in the mainstream, and Carrier’s ‘minimal historicism’ is more minimal again. It seems to me that one could accept all of Carrier’s arguments along these lines and still be a ‘minimal historicist’ — ‘sure, Mark is almost entirely an allegory and/or a Homeric pastiche, but it’s accurred to the biography of a real individual’, and it’s not clear to me how Carrier’s argument rules this out.

    I have seen things that make me think that Carrier is inclined towards a rather simplistic inductivism, e.g. ‘well, I’ve shown that 90% of Mark is allegory, so the other 10% is overwhelmingly likely to be allegory too’, and this is somehow connected with his Bayesianism.

    *) speaking of which, I felt Gullotta dismissed the Bayesian treatment a little glibly. The fact that the same method gives different results shouldn’t be a surprise when the starting points are different — as Matt has already pointed out, garbage in, garbage out, so at least one of Swineburne and Carrier have priors significantly at odds with reality :-). Also, mainstream historical methods also give divergent results, which of course is a common complaint of naysayers. But the divergent results presumably come about at least in part for similar reasons as two idealized Bayesian agents would: different ideas about how likely things are to start with, including what hypotheses are how likely on what evidence — that is to say, how the evidence supports various hypothesis is itself up for grabs.

    (There is a famous Bayesian result that shows that Bayesian agents do converge on the same posteriors eventually no matter what their starting points, if they are given the same evidence, which is comforting in the abstract I suppose, but it’s unclear how this relates to real people, especially given that the amount of evidence required for convergence can be very great. )

    *) The discussion about angel names was good, and what I suspected without knowing much about it: ‘Jesus’ is a name for human males, not angels. The argument would be stronger though if it were established that there was little or no crossover between angel and human names. If half of the time angels are Uriel, Samael, etc. but the other half the time they are James, Tim, etc. then this is less convincing — ‘Jesus’ may not be attested to but if ordinary names are often angel names there’s no particular problem here. I suspect that it’s the case that there is little crossover during this period, but better to prove it.

    It would also be helpful to state more clearly that Philo does not actually assert the existence of an angelic being called Jesus ever.

    *) I liked the explanation about the ‘rulers of this age’. Dismissing this as simply referring to Rome as a mundane political entity has not sat right with me for a while now, as it seems that it was common to see demons everwhere and think they were at least partly in control of things, but it still seemed a stretch to me to think that Paul was referring to a celestial death by demons.

    *) The 2nd and 3rd century evidence that both Christians and their opponents thought Jesus had a mundane existence and was crucified by the Romans is interesting, but it seems to need an additional argument to count against Carrier. Carrier, as quoted earlier in the article, thinks the Jesus story as we know it today came about somewhere in 60-90 AD, so it seems to me that he could just agree ‘yes, by that time Christians had come to think that the crucifixion had actually happened, and so did their opponents who just accepted what the Christians said at face value on this point.’

    (I wonder whether Tacitus should be added to the evidence?)

    No doubt I’ll read this again at some point…