What Would It Take To Get You To Read Richard Carrier’s Book?

What Would It Take To Get You To Read Richard Carrier’s Book? September 21, 2015

A commenter on this blog mentioned how useful it would be if scholars in Classics, ancient Jewish history, or Roman history were to read and give their impressions of Richard Carrier’s book, On the Historicity of Jesus: Why We Might Have Reason for Doubt. The truth is that, just as I am perhaps going to read mainstream scholarship in fields other than mine but related to mine, on which I thus need to keep up to date, but am not going to read every monograph in those fields since they may or may not prove to be persuasive, likewise those in fields which neighbor my own are unlikely to read and comment on a book like Carrier’s, but only on ones which present the consensus, or which are seen to make a big splash among scholars working in this precise area.

But since the mythicist crowd doesn’t trust New Testament scholars and historians of ancient Christianity, considering us to be importing Christian assumptions into our work, it might help if scholars who cannot be so accused were to chime in. Of course, Bart Ehrman and Maurice Casey have done precisely that. But mythicists then claim that New Testament scholars are not using the same methods that other historians are. And so it would be useful if historians in other areas were to do likewise – although it probably will not change anything.

I am quite sure that, if any completely secular Classicist or historian writes a critical review of Carrier’s book, they will simply become the focus of attacks by the online mythicist crowd. I’m not sure what the mythicists will appeal to in their attempt to discredit such individuals, but they will certainly try something. It is what mythicists do. Perhaps they would even find themselves on the receiving end of invective from Carrier himself (see here for what they might have in store for them).

If this doesn’t seem like a very compelling invitation, well of course it isn’t. I’m asking someone who is busy and finds it hard to keep up with the abundant literature in their own field to read something in another, and more than that, to read 700 pages of unpersuasive claims, in order to try to persuade people who are unlikely to listen, and who are on the other hand very likely to cause them headaches for their efforts.

And so I have the feeling that simply offering to buy someone a beer is not going to be sufficient recompense.

And so what, if anything, would convince you to take on this unpleasant, unappealing, and unrewarding task?

"Thanks James for the post. Shame I missed the deadline for the conference looked quite ..."

Cognitive Science, Memory, Oral Tradition, and ..."
"Oh wait, is that what happened after this? : https://www.youtube.com/wat..."

The Definition of Nerdiness
"Emotional immaturity is rampant, especially on the internet. It's almost as if the Nerds took ..."

The Definition of Nerdiness

Browse Our Archives

Follow Us!

TRENDING AT PATHEOS Progressive Christian
What Are Your Thoughts?leave a comment
  • Gary

    “And so what, if anything, would convince you to take on this unpleasant, unappealing, and unrewarding task?”

    If it is available free in a public library, I would read it.
    I am selective on what I pay my own money for.

    Besides, the selection process of getting into a public library usually pre-sorts and eliminates stuff that is fluff (except for kid’s books).

    • Erp

      I think you’ll find best sellers in the public library even if they are fluff. Admittedly my check of two local public library systems didn’t find it. The university library does have it, and, it is currently checked out by faculty/staff so maybe someone has already taken the offer.

      • Gary

        Public library, with limited funds, get popular books. From what I read on this blog, it sounds like Carrier’s publishers are off-the-wall. Since the publisher is not widely known, and his books aren’t popular, I doubt if he gets into a public library. It’s not in my library. If it gets there, I’ll read it for free, just for curiosity sake. But not on my $. Now, if Ehrman says it is a must-read, I’ll buy it (if it is in the $20 range). But I think hell might freeze over if Ehrman recommends it. I already have some of Ehrman’s books. And hell would freeze over before I would buy any academic books in the $100 or more range. I leave that to academics.

        • Erp

          Carrier’s most recent book is from a legit academic publisher, Sheffield, Phoenix Press, which is why it is gaining some attention. Though it also has the academic price tag, $35 for paperback.

          • Gary

            Well, maybe it will make it in public libraries. If it does, I’ll give it a try. Better than most tv. But if I buy a book, it means I want to re-read it, and reference it in the future. His, I would think, would not qualify.

          • Erp

            What I’ve read and heard so far does not impress me so I may be waiting for the University library myself (certain advantages to being even low level staff at a university).

            However I’m wondering whether what we want is a combo of an expert in Bayesian statistics with a philosopher of history.

          • ncovington89

            I think a philosopher of history (like say, Robert Greg Cavin) and an expert in ancient Greek or GrecoRoman history who has at least some reading in the new testament would be a good idea.

    • Geoff B

      I checked out Jay Raskin’s Evolution of Christianities from the public library.

      • Gary

        Zero books by Jay Raskin in my public library. Over 30 books with Bart Ehrman’s name attached in my public library. Based upon empirical data, I’d say that the probability that Jay Raskin exists is less than 0.033 of the probability that Bart Ehrman exists. All other things being equal. And assuming a perfect vacuum.

  • John MacDonald

    Tell them that if they read Carrier’s book and find it persuasive, they can be considered “cool” by amateur internet atheist bible enthusiasts.

  • spinkham

    If history/NT studies is like most other disciplines, there’s probably a small number of philosophers of history or similar titles who would be the appropriate people to read and address the book.

    (edit: Sorry, I was acting under the assumption this post addressed both books, the methodology one and the application one. I think they former might be useful even if the latter is fatally flawed.)

  • Samphire

    My great, great, great, great grandfather refused to read On The Origin Of Species for the same reason.

  • Cecil Bagpuss

    The obvious incentive is that they would have the chance to join the club of those who have been called liars or lunatics by Richard Carrier.

  • Neko

    Geza Vermes explained “Christian Beginnings” (Yale University Press, 2014) in 244 pages. Richard Carrier explained Christian Beginnings in 700 pages (Sheffield Phoenix Press, 2014). Geza Vermes was an internationally renowned and prolific scholar of Jewish and Christian history and an expert in the Dead Sea Scrolls. Richard Carrier is a Ph.D. of ancient history and independent scholar internationally renowned for his rude and obnoxious attacks on prominent experts in New Testament studies, and who serves as high priest of a cult that supports his mission of spreading the revealed truth of Jesus mythicism. Who is Professor Stressed-Out-and-Overworked likely to read for current thinking in this related field?

    However, since Carrier’s two-volume set proposes a superior, scientific methodology for historians, I’d think that would interest scholars of ancient history.

    • John MacDonald

      I fail to see the value of scholars in Classics, ancient Jewish history, or Roman history, who are not experts in the historicity of Jesus, being recruited to read Carrier’s book because we value their opinion so much. I wonder if there are any Classics professors out there who are looking to recruit New Testament scholars to help them evaluate a book arguing that Socrates never existed.

      • Neko

        I understand. I thought the proposed methodology, not the dispute over historicity, might have interest for historians.

        • John MacDonald

          It is of questionable veracity to say someone is an expert in form, but not in content. And we need to keep in mind that few if any scholars of the ancient world would recognize Carrier’s methodology as useful. Besides, we have scholars readily available that are experts in the content of Christianity as well as the methods for investigating that content: They are called New Testament scholars. Dr. McGrath is one of these. I think confusion arises because people think Carrier is a secular historian using tried and tested historical method (while NT scholars use fallacious methods). Carrier’s methods would be idiosyncratic at best among mainstream historical scholars.

          • Neko

            My layman’s impression is that Carrier is a self-styled insurgent who considers NT historical criteria unsatisfactory and is committed to developing a more scientific approach. I’m aware his application of Bayes’s Theorem has been debunked, but my point is that an ambitious critique of methodology might interest historians. Apparently I’m wrong about that.

          • John MacDonald

            The irony is that Carrier’s hyper-mathematical method in which everything becomes demonstrable emerged in a postmodern age, where, since the 1960’s in France, every claim to objectivity has come under suspicion.

          • Kris Rhodes

            But Carrier doesn’t claim objectivity.

          • John MacDonald

            Carrier claims that he uses a deductive (not inductive) mathematical method to determine that the highest probability for Jesus existing is 1/3. Sounds like he’s claiming “objectivity” to me. How do you define “objectivity?”

          • Kris Rhodes

            He is very explicit, and very clear, in OHJ and elsewhere, in explaining that his Bayesian method is subject to subjectivity in the assignment of probabilities to evidence. The advantage he sees in his method is that it shows you what your assumptions about probabilities actually imply.

          • John MacDonald

            And yet he is “certain” that the limit of the probability that Jesus existed is objectively 1/3.

          • Kris Rhodes

            Can you quote the passage you have in mind here from OHJ?

          • John MacDonald

            It’s near the end of the book where he draws his mathematical conclusions from all the arguments he has made throughout the book.

          • Kris Rhodes

            I’m not seeing it, and I’m surprised if he said that. It would contradict the main thrust of his argument.

          • Geoff B

            The main thrust of his argument?

          • Kris Rhodes

            That was poorly put–I meant it would contradict one of the most important aspects of his overall argument, namely, it’s aspect of using formal tools that make for an explicit method by which people can zero in on the truth together even if none of them (including Carrier) has the truth from the outset.

          • ncovington89

            Bayes theorem is inductive reasoning. Or at least, it’s a deduction based on inductive premises.

          • John MacDonald

            Hi Nick. Carrier argues, for instance, in OHJ that he has objectively proven through deductive methods that the probability limit that Jesus existed is 1/3. He claims (see for instance, his comments on the “Richard Carrier’s Dishonesty” blog post on this forum) his method is deductive. This is because he doesn’t understand the difference between (1) a “deductive conclusion” and (2) an “inductive inference.” Inductive reasoning (as opposed to Deductive or Abductive reasoning) is reasoning in which the premises seek to supply strong evidence for (not absolute proof of) the truth of the conclusion. While the conclusion of a deductive argument is certain, the truth of the conclusion of an inductive argument is probable, based upon the evidence given. Hermeneutic presentations are examples of inductive reasoning, not deductive reasoning.This “category mistake” is why Carrier will often claim he has “proved” this or that in a hermeneutic presentation, while in fact he has only “argued for it.”

          • ncovington89

            Induction and Deduction are not mutually exclusive things that never touch each other. For example:
            1. Socrates was a man.
            2. All men are mortal.
            Conclusion: Socrates was mortal.
            Premise 2 is arrived at inductively, and here this inductive premise is used in a deductive argument to generate a conclusion.
            Bayes theorem, as I view it, is just acknowledging that inductive conclusions are not certain, so the deductions from inductive premises is never certain. So when we figure out roughly how much uncertainty we have, and then do the same for every known argument for and against some position, Bayes allows us a way to sort out what is probably right.

          • John MacDonald

            Carrier’s claims to have “proved” this and “proved” that shows that he misunderstands what is possible through hermeneutics.

          • ncovington89

            I suspect you have never read his books? Because he tends to use “prove” in the sense of “show beyond reasonable doubt.” Do biblical scholars not acknowledge that some conclusions in there field have been shown to be very probably true, to the point where doubting is unreasonable? Sure they do.
            And besides this, at least for the historicity of Jesus Carrier is pretty modest and says that his conclusions could easily be reversed, if someone shows that there is relevant that has not been included in his calculation or if his weighing of the probabilities is wrong.

          • John MacDonald

            Sure, but Carrier claims he has proven things when he has been nowhere near to meeting his burden of proof. And Dr. McGrath has raised this point a number of times. Just read through McGrath’s comments in the “Carrier’s Dishonesty” blog post.

          • ncovington89

            Let me put it this way: what assertions do you find in “On the Historicity of Jesus” that you would say have not met the burden proof? I’d be eager to see them, especially starting with the premises that are crucial to Carrier’s case for a mythic Jesus.

          • Kris Rhodes

            I don’t see any confusion on Carrier’s part on this point. You can make deductive arguments about probabilities. For example:

            1. If you are probably going to the store, then I am probably going to the movies.
            2. You are probably going to the store.
            3. Therefore, I am probably going to the movies.

            The conclusion apparently has a probability attached to it, yet the argument is, for all that, deductively valid.

            An even more relevant example:

            1. It is improbable that X happened.
            2. It is less probable that Y happened than that X happened.
            3. Therefore, it is improbable that Y happened.

            This is a deductively valid argument. Even despite the fact that the conclusion has a probability attached to it.

            (It may make you more comfortable to add in an unstated premise: “Any probability less than an improbable probability is itself improbable.” This is really just to explicitly unpack the interpretation we surely all are applying to the terms in the argument as originally stated so is not strictly necessary to make the thing valid.)

            It is deductive because the conclusion is presented as following with certainty from the premises. In other words, the conclusion can be paraphrased as “It is therefore made certain by these premises that it is improbable that Y happened.”

            The conclusion of an inductive argument can’t be paraphrased that way.

          • Geoff B

            He claims to give a generous reading to the historicist case to guard against his own bias.

          • Kris Rhodes

            That’s not the same thing as claiming objectivity. It is, in fact, to own up to a lack of it.

          • Neko

            Ha. Yes. How can we ever know objective truth when we are prisoners of our subject position? Bayes’s Theorem to the rescue!

          • ncovington89

            There are lots of times when we know, intuitively and/or logically, that one thing is more probable than another. Sometimes we don’t have hard data to know exactly how much more probable. In which case we are stuck with making a guess. Which is still subjective, but Carrier aims to make his guesses on the conservative side. So you could say that there is minimal subjectivity in his approach: most of his estimates come out of objective data, and a few estimates are supposed to be overly conservative subjective estimates.

          • Neko

            Thank you for your sober response to my somewhat glib comment.

            I downloaded a sample of Carrier’s OHJ and read the following in the preface:

            I demonstrated that the most recent method of using ‘historicity criteria’ in the study of Jesus has been either logically invalid or factually incorrect, and that only arguments structured according to Bayes’s Theorem have any chance of being valid and sound.

            You’ll forgive me if I find this claim bombastic and un-conservative.

          • ncovington89

            That’s rather confused, I was talking about Carrier’s estimates of probability being conservative on certain points.
            I don’t know what about Carrier’s statement you find ‘unconservative’ or in what sense you are using that word, but it is definitely true that historicists don’t seem to be able to make an argument that conforms to logic and is built on facts and not speculations. If you think you can make an argument for the historicity of any gospel event that does not violate logic or facts, I’m all ears.

          • Neko

            I understand what you were specifying. It projected an image of conservative reserve which is comically undermined by Carrier’s claim, in defiance of an entire discipline, to have identified the exclusive formula by which valid conclusions may be drawn about the existence of Jesus.

            I don’t read ancient languages, nor do I have the skill set, to make informed arguments about the the historicity of any gospel event. I’d suggest you consult the experts, but apparently you don’t consider them up to snuff. Oh well.

          • ncovington89

            “I understand what you were specifying. It projected an image of conservative reserve which is comically undermined by Carrier’s claim, in defiance of an entire discipline, to have identified the exclusive formula by which valid conclusions may be drawn about the existence of Jesus.”

            Your comment was very strange, because it did not address my refutation of your claim.

            “I’d suggest you consult the experts, but apparently you don’t consider them up to snuff.”
            You were given an opportunity to prove something in the gospels and chose to respond with a snarky remark instead. And the experts don’t fare any better, I’ve read ’em. And this how dialogues between mythicists and historicists go: The historicists get challenged, and instead of making a defensible case based on evidence, they resort to snark, baseless accusations (as when McGrath says mythicists are “proof-texting” but fails to demonstrate that the text in question was quoted out of context).
            You know, when you asked me (on another blog here at Patheos) why I thought the gospels were symbolic, I did not scoff and say, “Go consult Robert M Price and Richard Carrier.” I gave you an honest answer, and even cited some evidence for that answer. However, when I ask you a question, I get snark and no answer with any substance behind it. Why don’t you treat others the way you want to be treated?

          • Neko

            Wait, what? What claim did I make again that you refuted?

            There is quite a difference between asking why you think the gospels are symbolic and you asking me to “prove” the historicity of an event in the gospels. Indeed I would never make such a disingenuous request, because, as you know, there is no “proof” one way or the other.

            Further, why would you ask me to prove anything about an ancient text? Who do you think I am? And if you’ve “read ’em,” that is, historical Jesus scholars, and “the experts don’t fare any better” than a black cat on the internet, then clearly what I said was true: you don’t think the experts are up to snuff.

            You did give a considered response to my question on that other blog, and I had meant to respond but never did. Sorry about that!

          • ncovington89

            “Wait, what? What claim did I make again that you refuted?”
            You apparently thought Bayes’ theorem was sheer subjectivity, and I answered by saying in essence that Bayes’ theorem constrains and controls the subjectivity involved in weighing evidence.

            “Indeed I would never make such a disingenuous request, because, as you know, there is no ‘proof’ one way or the other.”

            There are two ways I could interpret this. If you mean you think we cannot show the gospels are historical versus completely symbolic, that neither one is favored (rendered more probable) by the evidence we have, then that is admitting that four favorite historical evidences for Jesus are worthless (which is an admission I would welcome). On the other hand, if you only mean that you don’t think any historical claim can be “proved” as in “absolutely proved” then you badly misunderstand what I wrote: I use the word “prove” as a shorthand for “demonstrate beyond reasonable doubt.” And New Testament scholars, knowledgable and clever as they are, have failed to secure any gospel story at all as true beyond a reasonable doubt. Does that make them “not up to snuff”? I don’t hold that dismissive an attitude, I think much NT work is valid (i.e. that that shows mimesis between the OT and new) I just disagree on the issue of whether we can show some minimal list of facts in the life of Jesus actually occurred.

          • Neko

            What on earth gave you the idea that I thought Bayes’ theorem was complete subjectivity? You completely misunderstood my comment. It was an ironic reference to a purportedly objective panacea to the distortions of subject position.

            Update: I see you added an entire paragraph to your post. No, I don’t think the gospels are worthless as evidence. I understand historical analysis of NT texts to be a matter of establishing probabilities and plausibility, rather than aspiring to meet a legalistic standard of “true beyond a reasonable doubt.”

            You wrote:

            I think much NT work is valid (i.e. that that shows mimesis between the OT and new) I just disagree on the issue of whether we can show some minimal list of facts in the life of Jesus actually occurred.

            I will take your verdict as an internet amateur under advisement.

          • ncovington89

            I apologize if I misread your comment earlier. Did your earlier comment mean that Bayes’ theorem would not really dissolve subjectivity inherent in history? If so, then I agree that it does not, but I think it does promise to minimize subjective gut judgements, and to reduce the gut judgements we do make to such little weight that they would be acceptable to the majority of people.

          • Neko

            Not a problem. All I meant to express was amusement at Carrier’s confidence in the scientific rigor of his methodology. You might call my skepticism a “subjective gut judgment.”

          • ncovington89

            Well, although Carrier’s basic method (Bayes theorem) is by my lights the right way to go, it is still important to remember that his Bayesian calculations could have to be revised if the background knowledge that went into his equation is suspect, or if there is evidence (for either mythicism or for historicity) that has not been given proper weight.
            This might be true with his handling of the passages in Paul/Hebrews/Clement/Revelation that discuss Jesus’ ancestry. Though the rest of Carrier’s book was rock solid, on this one point I didn’t think he was very convincing. So I am stuck in the position of thinking there are several good arguments for mythicism and one very good argument for historicism.

          • Neko

            Based on no more than antipathy and a hunch I initially suspected Carrier of scientism. But subsequently I read mathematicians argue that Carrier doesn’t know what the heck he’s doing with Bayes’ theorem. I’d be interested to read one review by a probabilities person who approves Carrier’s method for historical research.

            You are way down in the weeds with this stuff. I only occasionally drop by to check out the latest in the Jesus wars, and last week I had the good fortune to arrive just as a marathon slugfest took off, inspired, just as sure as the sun rises in the east, by Neil Godfrey accusing James McGrath of being “dishonest.” One of these days Godfrey’s going to get eaten by a wolf.

          • ncovington89

            Cavin does agree with using Bayes theorem for at least some historical claims: https://m.youtube.com/watch?v=MBARhmMV5dU

            Timothy McGrew has published an article on the resurrection and Bayes theorem, so he too agrees with using it for at least some historical claims ( if you go to lydiamcgrew.com she has posted the article she wrote with Timothy).

            William lane Craig brought up the use of Bayes theorem in his debate with Bart Ehrman.

          • Neko

            For some reason this video has no sound (it’s not a technical issue on my end). Anyway, I don’t have two and a half hours to devote to Dr. Cavin’s position. Is it summarized in text somewhere? Thanks.

            I couldn’t find the McGrew article you mentioned, but for our purposes what is the point of trying to establish the probability of a supernatural event? I’m satisfied to limit historical inquiry to the non-supernatural sphere. Same goes for William Lane Craig. I’ll further note that none of your experts are “probabilities people.”

            As an aside, the following book review is a useful summary of Bayes’s theorem for the layman (I surrender to the ugly form of the plural s since the Times so decrees):


          • ncovington89

            McGrew and Cavin are definitely probabilities people, they both have the academic background and have published peer-reviewed papers discussing Bayesian arguments. I don’t how to help you on the Cavin debate, try searching youtube for “Licona Cavin debate”

          • Neko

            My bad, you’re right about McGrew, and I’ll take your word for it on Cavin. At some point I’ll investigate further, thank you.

            The video was purposely muted because of copyright violations.

          • ncovington89

            Hmmm, well if you search youtube for “Cavin Licona debate” I am sure it is there somewhere. But yes, Bayes’ theorem is often used even in philosophy to model the logic of the arguments presented, if we can do it in philosophy I don’t why it would fare worse in history. Historians often have access to much better frequency data than philosophers do anyway, simple because of the nature of philosophy.

          • Neko

            You wrote:

            Hmmm, well if you search youtube for “Cavin Licona debate” I am sure it is there somewhere.

            That occurred to me. I mentioned it so you’d consider not wasting people’s time in future with a video lecture with no sound.

            I’m aware that BT is used in a variety of applications, but of course, the interest here is with Carrier’s particular application, and from what I’ve read, historians (and, of course, mathematicians) find Carrier’s application problematic. But…like I said, I’ll investigate your enthusiasts at some point (unhappily, since they are concerned with a supernatural event).

          • John MacDonald

            There is a reason that Ehrman scoffs at Craig’s attempt to use Math to prove the resurrection.

          • ncovington89

            Well I disagree with Ehrman and instead side with philosophers like Cavin that Bayes theorem can and should be used to evaluate such claims. I just disagree with the numbers evangelicals use to make their case.

          • John MacDonald

            If you like Cavin, what do you think of Licona’s reply to Cavin regarding Bayes Theorem? :

            “[H]istorians don’t use prior probabilities in historical inquiry. One cannot calculate the prior probability that the U.S. would drop nuclear bombs on Japan during WWII, since in all of human history no nation had dropped a nuclear bomb on another before or since WWII. Moreover, I’ll be 51 in two weeks. That’s a lot of days in my life. Yet Sunday was the first day I had ever spent in Temecula, California. Given my “tendency” not to go to Temecula, one should conclude that I wasn’t there that evening. Historians examine a historical report then look at the evidence for the event occurring. Thus, prior probabilities are the wrong tool for historical inquiry. It’s like using a calculator for an archaeological dig.”
            My question would be if Bayes Theorem can accomplish what Carrier thinks it can accomplish, why is it not popular among philosophers and historians?

          • ncovington89

            “One cannot calculate the prior probability that the U.S. would drop nuclear bombs on Japan during WWII”

            Coming up with a prior wouldn’t be difficult. Given that the United States had develop nuclear bombs and was in the midst of a war with Japan at the time, and given that Nations often use the weapons they develop against the enemies they war with, we can deduce a fairly high prior probability. And even if you assumed for argument’s sake that the prior was small (one in ten thousand, let’s say) the evidence we have would still lead us to the conclusion that it happened, because we have such a huge wealth of evidence for that conclusion.

            “Given my “tendency” not to go to Temecula, one should conclude that I wasn’t there that evening.”

            This was rather silly of Licona, since you shouldn’t work from prior probabilities alone. Bayes theorem is prior probability AND consequent probability. Licona’s testimony could be used as a consequent probability that would easily overcome the small prior probability that he went to Temecula.

            “Thus, prior probabilities are the wrong tool for historical inquiry. It’s like using a calculator for an archaeological dig.”
            Very ironic: go to amazon and search “Bayesian approach to interpreting Archaeological Data.”

          • arcseconds

            For this to be anything other than a guess, though, would require an enormous amount of research into how often secret weapons under development were actually used.

            And if the nuclear bomb was really the subject of doubt, we’d have to take the argument back one step further and ask ‘what was the probability of it being developed in the first place?’, because presumably if someone thought it didn’t happen they might well think it was never developed.

            That seems an awful lot of work to go to when you can just note it’s all been extremely thoroughly documented.

            You think it’s ‘the way to go’. Are you familiar with any examples where it has been shown to be able to capture historical reasoning that isn’t disputed? If so, I’d love to read them. If not, why are you so confident that it’s a good technique when it’s unproven?

            (The bayesian archeaological data book you refer to is obviously using it for data that’s ammenable to statistical analysis, and hopefully no-one thinks it’s inapplicable there.)

          • ncovington89

            For this to be anything other than a guess, though, would require an enormous amount of research into how often secret weapons under development were actually used.”

            I don’t think that’d be necessary. If you agree that happens with even a small frequency, which would have to be true based just on common observation about human nature, then the conclusion would follow.

            “And if the nuclear bomb was really the subject of doubt, we’d have to take the argument back one step further and ask ‘what was the probability of it being developed in the first place?”
            Okay, you could do that. Where you would end up would be based on the testimony of so many people who were there who claimed that it was indeed developed. Analyzing that testimony through the well-established generalization that most people tell the truth most of the time would lead directly to the conclusion that it probably happened. And the sheer number of converging testimonies that would support it would leave it beyond any reasonable doubt.

          • arcseconds

            Yes, but we already do all of that without the use of Bayes’s theorem. The last paragraph is just what anyone used to considering historical evidence would write, no knowledge of Bayesian epistemology required.

            So it seems it’s absolutely of no use here, and the calculation of a prior, which is what a formal, quantitative use would require, is a difficult and pointless exercise.

          • ncovington89

            Bayes theorem is really a much more precise version of the common sense rules of reasoning we already use. Does that make it pointless? No. We need a more precise method in those cases where people disagree, or in very complex cases where there is a lot of evidence and some of that evidence points towards different theories.

            I also don’t think getting a prior is nearly as hard as you seem to believe it is. You can look at frequency data, you can deduce from General facts, or, if both of those completely fail you, just treat all possibilities as of equal likelihood.

          • arcseconds

            Frequency data for nuclear bomb attacks? There have only been 2 in all of history!

            My point is that getting a prior for an event such as the nuclear attacks that’s anything more than an somewhat informed guess is actually quite hard. We could follow your suggestion and look at all secret weapon projects and see how many result in actual deployment of the weapon. That’s a lot of work, and it’s unclear that the data would actually be available… secret weapons are, well, secret, and we know from the experience of Bletchley Park that a lot of things were destroyed after the war.

            I don’t even know what information we’d be putting into this, to be honest. I mean, my probability for the attacks having happened is nearly 1, but then I know that they’re well documented, so presumably the prior is for someone who’s ignorant of the evidence. (Yet somehow well informed about secret weapons projects? )

            Are we trying to model someone receiving the reports in 1945? Someone doubting this now? How much do they know about physics?

            (The physical possibility of a nuclear explosion was not generally known before 1945, so someone informed about physics but not absolutely up with the latest developments might well think this kind of explosion a physical impossibility, so would be giving a much lower prior than someone who was ignorant… )

            I’m struggling to see how we can get a reasonable prior for this that isn’t a number plucked out of the air to reflect ‘pretty low but not impossible’.

            And that’s with a well-documented historical case. With someone like Jesus I really don’t know where to start. People mentioned in the literature who’s name begins with ‘J’ (well, iota)? Carrier’s Rank-Raglan treatment is nearly as ridiculous if you ask me. But if we group him with 1st-century messianic claimants.. are there any of them that are considered to be non-historical?

            This is the whole reference class problem that Ian mentioned before. And there is already a debate between Carrier’s followers and, shall we say, McGrath’s followers about the appropriate reference class here, a problem that’s been created by Carrier’s Bayesian treatment, which doesn’t look like it’s going to be easily resolved by more Bayes.

            I’m struggling to see how any of this helps us much. We seem to have traded informally reasoning about various kinds of documentary evidence into informally reasoning about priors (and likelihoods) which, unless they fall into the rare situation where we already have a statistical handle on them, seems to be trading an area which is often relatively clear and which most of us can reason about in an OK fashion (e.g. the large documentary evidence for nuclear attacks in Japan) to areas which are far murkier and filled with pitfalls (calculating priors for nuclear attacks).

            When I first started paying any attention to Carrier, while I always recognised Bayesian treatments may just push the debate into debating priors and likelihoods rather than what is currently debated, I was cautiously optimistic that it might nevertheless cast some light in some areas. It might make people get clearer about what their assumptions are. Another possibility is that small differences in priors and likelihoods can lead to very different posterior probabilities on the same evidence, and if we could identify such a case that would be useful (the people could stop arguing about it and accept there are two different but equally rational conclusions on the current data, maybe).

            But every time Ian talks about this, I get a lot less sure.

            In simple cases where people disagree, I think you just end up pushing the disagreement off into the priors, as I mentioned.

            In complex cases, it’s really unclear that Bayesian treatment will be tractable or worthwhile. Errors will tend to propagate and grow the more applications of BT you use, and Ian’s feeling is that often what you end up measuring is the mere fact you have a new piece of data, which tends to reduce the probabilities. If that’s true, that’s really worrying.

            While it might be nice if there was a kind of calculus that could be deployed to clear up these circumstances, I’m still wondering what basis you have for your confidence that this is possible.

          • ncovington89

            “Frequency data for nuclear bomb attacks? There have only been 2 in all of history!”

            I said that I think we can either (a) use frequency data OR (b) deduce a probability from more general facts OR (c) if we can’t do either of those, treat all possibilities as equally likely. In the case of nuclear weapons, the frequency was zero but only for the reason that they were not developed. Which is why we should not use frequency data, we should an alternative method. The best alternative to me looks like (b). So, without the evidence of nuclear attacks, what likelihood would you give to the following propositions:

            (A) the US was at war with Japan in the 1940s.
            (B) the U.S. Had developed nuclear weapons by the 1940s.
            (C) a nation at war uses its deadliest weapons against its opponents (remember we need not believe this is the case 100% of the time, only some percentage of the time)

            The probability of each of these statements combined is the prior probability that the U.S. Dropped atom bombs on Japan. Remember, you don’t need exact frequency data on any of this, a guess is fine by me, just as long as the guess is a believable one. Now I know what your going to say: a guess, really?!? Yep, a guess. And guessing is perfectly okay. Here’s why: our intuitions are molded by thousands of real life experience, so they are perfectly valid to use. Especially, I think, if we “err on the side of caution” with our guesses. For example, I might think that (C) holds true in 90% of cases. But realistically I could not be sure that it wasn’t only 89 or 88%, and realistically it could be way lower, maybe even just a half-percent. But regardless of what it really is, any remotely believable number would still just the conclusion that the atom bomb was used on Japan because the evidence would always bring it up. And here’s the clincher: there’s no method that can beat what I have just laid out. You may be uncomfortable with using some intuition, but you would be using that no matter what. We all know, intuitively, that every proposition has some burden of proof. The proposition “Nick owns a TV” has a small burden of proof, you would believe it just based on my word that I did. The proposition “Nick owns a Time machine” has a heavier burden of proof; I’m sure you would not believe me if I said it, you would want extensive corroboration. Bayesian prior probability is really just the “burden of proof” in mathematical terms. If you trust yourself to be able to figure out, or come to some rough sketch of, the burden of proof for “Nick owns a TV” or “Nick owns a time machine” then there’s no problem at all putting that burden of proof into its mathematical form.

            “And that’s with a well-documented historical case. With someone like Jesus I really don’t know where to start. People mentioned in the literature who’s name begins with ‘J’ (well, iota)? Carrier’s Rank-Raglan treatment is nearly as ridiculous if you ask me. But if we group him with 1st-century messianic claimants.. are there any of them that are considered to be non-historical?”

            I don’t know that I agree with carrier’s rank-raglan argument either. That said, I think carriers more abstract proposals for the methods we should for finding a reference class (that he discusses in proving history) are pretty reasonable, even if his particular selection is one that I am not so sure of. But here is the key thing: history is all about what is probably true (as Bart Ehrman and many others have put it) and probability ALWAYS involves comparison with some group of similar things or events. So when John Dominic Crossan says that many crucified Jews were not buried and that he cannot help but think Jesus was one of them, he is placing Jesus in the reference class “crucified Jews,” noting that most members of the class do not receive burials, and deducing that therefore Jesus probably wasn’t. The problem of reference classes is not INTRODUCED by Bayes theorem, it is within all historical reasoning, sometimes hiding under a rug but still there.

            “This is the whole reference class problem that Ian mentioned before. And there is already a debate between Carrier’s followers and, shall we say, McGrath’s followers about the appropriate reference class here, a problem that’s been created by Carrier’s Bayesian treatment, which doesn’t look like it’s going to be easily resolved by more Bayes.”

            I haven’t followed that debate, so I don’t know what the particulars are of that debate. But I do think there are some general strategies you could use to dissolve it, and those strategies receive some attention in both Carrier’s books.

          • arcseconds

            What I’m looking for here is what, exactly, a formal deployment of Bayes’s Theorem gives us over informal reasoning. We can already do the informal reasoning, so when you talk to me about guesses and wanting evidence for time machines, well, everyone knows all that already, no-one needs Bayes’s theorem to work that out. If you’re just using it on guesses, then I don’t see what it gives you over informal reasoning. Essentially it is just informal reasoning dressed up as though it’s formal at that point, as far as I can see.

            So far, it just looks like it involves introducing more problems. One problem you might want to consider is that our informal reasoning does happen largely subconsciously, and what is happening there appears to be very complex. By introspecting for a ‘feeling’ as to the probabilty of a nuclear bomb attack, and translating it into a number, and then putting it in to Bayes’s theorem there are new possibilities of error . Is that feeling really all that’s going on in you when you’re thinking about a complex historical problem? Is the number really representing your feeling? Is there really just one Bayesian update going on, or are there several?

            In the case of the nuclear attacks, I wonder whether anyone ever considered reference classes and priors for nuclear attacks before Licona’s quip. Reference classes and priors aren’t a problem for most issues in history unless you’re trying to be all Bayesian. I don’t think we even need to think about it: all we need to know is that it was probably pretty low but not nigh-impossible, and there’s a wealth of data that says it did happen. Bayes’s theorem gives us nothing here.

            In Crossan’s case, seeing as he’s already sorted out a reference class, again there’s nothing more to be done here.

            You have of course been following the debate between Carrier and his followers on the one hand and people supporting the mainstream on the other: you’ve been participating in it.

            The particular problem that I was talking about was the problem of the reference class for Jesus. Carrier thinks the Rank-Raglan high scorers is an appropriate class, mainstream historians think the messianic claimant class is the one to go for. Once again, this is a problem that comes about because of the necessity of coming up with a prior, a problem that never existed until Carrier published his book. And we seem to have irreconcilable differences on this. New irreconcilable differences, that Bayes’s theorem is unable to help us with. It’s creating problems, not solving them.

            If you are serious about the use of Bayes’s theorem to solve problems in history, you really ought to read Ian’s posts. Ian is the only one here who is really qualified to talk about the matter: he has both advanced training and practical experience of using Bayesian treatments to solve real-world problems.

            The most recent episode starts here, and you should also read his blog posts on the matter.

            I am still wondering where your confidence in this is coming from. You’ve just ignored the problems I raised with propagating errors in repeated applications, for example, and you appear to be handwaving at the existence of ‘general strategies’ to solve problems you’re not clear on.

          • ncovington89

            “What I’m looking for here is what, exactly, a formal deployment of Bayes’s Theorem gives us over informal reasoning. ”

            That’s been explained earlier: a more precise method is better. Think about why philosophers spell out there arguments explicitly in the form of a syllogism. We should use Bayes’ for the same reason.

            “We can already do the informal reasoning, ”

            You haven’t grasped my point: if you’re comfortable with informal reasoning you cannot have any objection at all to the exact same sentiments being placed in their corresponding mathematical language.

            “Is that feeling really all that’s going on in you when you’re thinking about a complex historical problem?”

            But you are also relying on intuitions when you reason informally. Think about it: you decided, intuitively, that the amount of evidence we have for the atom bombs dropping was good enough to meet the burden of proof which you presumably also arrived at intuitively. So if this is a good objection against Bayesianism, it is a good objection against every known form of reasoning. It should also be pointed out, once again, that your intuitions about how people behave are the result of your entire life’s experience and thus come from an enormous amount of empirical evidence. So using intuition as a guide isn’t like reading tea leaves.

            “Is the number really representing your feeling?”

            Sure. I think that independently of the evidence the prior probability of the atom bomb dropping is not vanishingly small.

            “Is there really just one Bayesian update going on, or are there several?”

            I don’t know what you mean here.

            “Reference classes and priors aren’t a problem for most issues in history unless you’re trying to be all Bayesian.”

            Recall John Dominic Crossan on the non-burial of Jesus. He is comparing Jesus to other ancient crucified Jews. A reference class is a comparison of one member against a group of similar members. So therefore, Crossan is placing Jesus in the reference class “ancient crucified Jews.” Crossan infers that since most members of the class “ancient crucified Jews” were not buried, so probably Jesus wasn’t either. Which is the same thing as a Bayesian assigning the burial of Jesus a low prior probability after placing him in the reference class “ancient crucified Jews.” Six of one, half a dozen of the other. Interestingly, the problem of choosing the right reference class emerges here. Just as some are worried about what we should be comparing in Bayesianism (reference class), so have some worried that Crossan is

            comparing Jesus to the wrong group. The dissenters point out that Romans often allowed burial during times of peace, and so Jesus should be compared with (in other words: “placed in the reference class of”) “Ancient crucified Jews executed during times of peace.”

            I don’t think we even need to think about it: all we need to know is that it was probably pretty low but not nigh-impossible, and there’s a wealth of data that says it did happen. Bayes’s theorem gives us nothing here.

            “You have of course been following the debate between Carrier and his followers on the one hand and people supporting the mainstream on the other: you’ve been participating in it.”

            You referred me to a discussion between “McGrath’s followers” and “Carrier’s followers” about the reference class Jesus should be placed in, and I don’t think I know what you mean, I don’t know where these discussions are or what points have been brought up in it.

            “The particular problem that I was talking about was the problem of the reference class for Jesus. Carrier thinks the Rank-Raglan high scorers is an appropriate class, mainstream historians think the messianic claimant class is the one to go for.”

            If by “mainstream historians” you mean one guy, James McGrath, then OK.

            “Once again, this is a problem that comes about because of the necessity of coming up with a prior, a problem that never existed until Carrier published his book.”

            No, it’s always existed: History is about what probably happened, so considerations about probability are a necessity. Probability NECESSARILY involves comparison. “Reference class” is just a name for what you are COMPARING TO, So every historian is always using reference classes for everything, even if they don’t call it that or realize it. They are talking about what probably happened, which they cannot do without having an idea of what they are comparing with, and “reference class” is just a name for what you are comparing your subject with.

            “Ian is the only one here who is really qualified to talk about the matter:”

            I wouldn’t lean on his authority too heavy there. You do realize that some very well-trained philosophers (Robert Greg Cavin, for example) AGREE with Carrier that Bayes’ theorem can be used to analyze historical events. In fact, much of what Cavin says here in this slideshow corroborates Carrier’s own philosophy of historical method:


            In fact, Bayes theorem is used within philosophy routinely to model the logic of various arguments put forth. Any objection you could possibly have to using Bayes in History would apply AT LEAST AS MUCH to its use in philosophy. Oddly, all of these philosophers who are trained in formal logical and probability theory don’t see a problem with it.

            “you appear to be handwaving at the existence of ‘general strategies’ to solve problems you’re not clear on.”
            I think you’ve picked your opinions on all this by reading Carrier’s critics without having read Carrier’s work on the subject. I always recommend reading both sides of the argument: I mean if you only read what creationists write about evolution you probably would think evolution was just crazy… Until you read a few books on evolutionary biology. Carrier does, in fact, discuss some strategies for solving disputes over reference class, for example. So read “Proving History.” I mean, I could spend a lot more time than I already have explaining all this to you, but I think it’s better for you if you actually engage the material you’re critical of. Not to sound like I’m downtalking or anything, but it is always a must to read someone before thinking they are obviously wrong.

          • arcseconds

            There’s a lot here, most of which I disagree with, but this perhaps is the core of the problem:

            You haven’t grasped my point: if you’re comfortable with informal reasoning you cannot have any objection at all to the exact same sentiments being placed in their corresponding mathematical language.

            This is a piece of informal reasoning. Please put it into mathematical language.

          • ncovington89

            Ok, here you go. My response is easily derived from the cubic formula: http://www.math.vanderbilt.edu/~schectex/courses/cubic/

            In all seriousness though, what you are saying is that you are allowed to assert that such and such is probably true about history, but nobody is allowed to put such a statement into mathematical language, even though part of what mathematics does is to encode probability in very precise terms. Bayes theorem is derived from self evident principles like the law of identity, “a equals a.” So I’ll refute you that way: if you say “an atom bomb was probably dropped on Japan in 1945” you are asserting that the probability of an atom bomb being dropped on Japan is something over 50%. That’s what “probably” means: at least a little over fifty percent. A equals A, so that must be what you mean.

          • arcseconds

            I’m sorry… are you objecting to putting your argument into mathematical language?

          • ncovington89

            Well, here’s a more serious answer, the most serious I can think of: Some arguments can be translated into math, some can’t. Arguments that can be modeled with math should be, or at least should fall apart when modeled with math, because if it does that can mean the argument is fallacious, or not as good as your gut tells you. Now I think all I have said so far falls under the “can’t be modeled with math” category I mentioned earlier. The historians statements about probability though, ALL of them, can be modeled with math (that’s what math DOES) and should be, for a wide variety of reasons. One of them being that you have a method, and every known method, like inference to the best explanation, has no formal justification outside Bayes theorem (in other words, you can’t prove the method actually tracks the truth, but you prove Bayes tracks the truth) and Bayes theorem solves and corrects a lot of stuff not addressed with Inference to the best explanation. Read the slides from robert Greg cavin, or the book from carrier that corroborates his conclusion.

          • arcseconds

            Also, what’s this about the law of identity? Are you actually familiar with how Bayes’s theorem is usually derived?

            Because it’s not normally derived from a = a, and the principles might not be as self-evident as you think they are.

          • ncovington89

            I think you can get Bayes theorem in a self evident way, from visual examples: http://www.skepticink.com/humesapprentice/2012/10/03/proving-darwin-math-and-tomatoes/

            By my lights, all math is self-evident, since one plus one equals two, ‘two’ being another way of expressing ‘one and one,’ thus boiling down to a disguised case of a = a.

          • arcseconds

            Regarding philosophy:

            Very little work in philosophy is done in the form of deductive logic (a quaint term that’s been out of fashion for about a century… have you been learning philosophy from Jesuits or something?) outside logic itself.

            Yes, you can find plenty of papers (but by no means all) that have a deductive argument presented semi-formally (complete symbolism is rare) somewhere at the core. But virtually all of the work in these cases goes in to justifying the premises, and that work is not normally deductive. Instead, the work typically involves such diverse elements as thought experiments, conceptual analysis, definition of terms, appeals to intuition, and sometimes outright rhetoric and purple prose.

            (The habit of making papers look all logical by labelling statements S and S’, etc, the use of logical connectives, and so forth are a feature of anglo-american philosophy of the last 100 years, and you shouldn’t mistake the quasi-logical conventions with the use of actual logic. )

            I think it’s hard to argue that the relatively minor use of formal deductive logic in philosophy has really had the clarificatory impact you appear to be suggesting Bayesian treatments will in history. Philosophers still disagree just as much as they ever did, and it’s not as if philosophers prior to the publication of Principa Mathematica in 1910 are somehow hampered by the lack of modern logic and logic-like conventions: look at the continued interest in philosophers like Kant, for example.

            (That’s not to say that formal logic hasn’t had a huge impact on analytic philosophy, but the impact is indirect rather than in its direct application, and is largely what could be called cultural in nature, and is difficult to separate out from other matters.)

            I suppose a clear framing of a deductive argument and even the use of the logic-like conventions can add clarity in some instances, and I’m prepared to entertain that casting arguments in a Bayesian format might do so too, but I’m not at all certain about this. I actually think the logic-like conventions in philosophy are often more a distraction than anything else, and that the writers would actually be better off really working at the clarity of their English (or other natural language) presentation than trying to shoe-horn a small part of their argument into a formal logical form, especially as it’s still the natural language arguments that are doing >95% of the work.

            And Bayes’s theorem is used far less than deductive logic in philosophy. As far as I know, the main interest in Bayes’s theorem is in philosophy of science, where Bayesian epistemology is felt by some to be the proper ‘logic’ of science, in the sense that all scientific reasoning is really Bayesian reasoning ‘under the hood’, so to speak. But thinking that is different from thinking that explicit use of Bayes’s theorem is really what scientists need right now to get their area some much -needed rigour. The idea that philosophy of science is there to tell scientists how to do science went out of fashion a long time ago.

            Also, last time I checked, philosophers of science were struggling to do ‘rational reconstructions’ of famous, uncontested, and well-documented discoveries in science. If it’s this difficult to capture the reasoning of a scientist, capturing the reasoning of a historian will be about as difficult, and of course capturing the reasoning in a formalism after the fact is easier than doing the reasoning in the formalism up front.

            Outside of numerical statistical work, I don’t think reasoning using Bayes’s theorem or explicity casting things in formal probablistic terms is very common at all in the sciences. If it had utility, one would expect scientists to be using it by now, as many of them understand probability and statistics to a considerable extent already. So if it’s as useful as you say it is, why don’t scientists always or even frequently present their arguments in Bayesian terms?

            Still, I suppose if I squint a bit and imagined historians getting clued up about probability and Bayesian epistemology, maybe I could imagine Bayes’s theorem appearing in history papers about as much as formally-framed deductive arguments do in analytic philosophy papers, which is to say a small amount of the burden might be bourne by them. And maybe it could give some clarity to certain matters. But the analogy with philosophy doesn’t suggest to me it’ll be the dawn of a new rigour you appear to think it will be.

          • ncovington89

            Bayes theorem is very commonly used in both science AND philosophy. There are literally hundreds of books and papers that utilize Bayes in each field. Go to google scholar and look up Bayesian phylogeny to see its use in studying how organisms relate to each other. And that’s only one of many places it is used in the sciences. For example: http://onlinelibrary.wiley.com/doi/10.1002/sim.2672/abstract;jsessionid=B3BAD4298C5E6AAA2D0A4738AEF604A5.f04t03

            Bayes theorem is used in philosophy, for example, in Timothy mcgrew’s paper on induction: http://homepages.wmich.edu/~mcgrew/kyburg7d.htm

            And just go to philpapers.com to see the innumerable philosophy papers that formulate Bayesian arguments! In fact, the Blackwell companion to natural theology made Bayesian arguments in nearly every chapter! And as previously stated there are people using Bayes in archaeology. So what are you waiting for? Join the Bayesian revolution!

          • arcseconds

            Of course it’s used in philosophy: Bayesian epistemology is a major area in philosophy of science, as I mentioned (that covers its appearance in a paper about induction). However, outside that area philosophers rarely construct Bayesian arguments.

            Actually, even within Bayesian epistemology the use is usually not what you’re expecting historians to do. They don’t usually argue for particular empirical propositions, but rather show how empirical propositions in general could be established using Bayesian epistemology. So the arguments they actually use are often deductive, mathematical arguments to prove something about the Bayesian framework, rather than using the Bayesian framework directly to make a probabilistic argument for something, a bit like how in logic the interest is often not proving things using a system but proving things about the system.

            I’ve taken your advice and gone and looked at philpapers. I used the advanced search to find free published papers in the last month with the words ‘a’, ‘the’, or ‘and’ in the title (had to put something in the text field). I ignored ones where the full text wasn’t actually available.

            I searched for ‘Bayes’ (which would also match ‘Bayesian’), ‘prior’ and ‘probability’, and found about what I expected. Only one paper mentions ‘Bayesian’, where it is talking about Bayesian epistemology (somewhat briefly, on cursory inspection), in the manner I describe above, i.e. it doesn’t make a direct Bayesian probabilistic argument for some empirical proposition, but rather discusses features of the system. One paper mentions ‘prior probability’. The rest don’t contain the term ‘Bayes’, and ‘prior’ and ‘probability’ don’t show up in contexts giving anything that looks like a Bayesian argument.

            So we could tentatively conclude from this that about one in 10 papers talks about Bayesian epistemology directly, another one in 10 refers to something from it rather obliquely and in passing, and less than one in 10 (actually probably less than one in 20) give Bayesian arguments of the sort you’re expecting historians to give.

            On the basis of this, I’m sticking with my assertion that Bayesian arguments are uncommon outside Bayesian epistemology itself.

            The Contribution of A.V. Kuznetsov to the Theory of Modal Systems and Structures.

            Stephen Biggs & Jessica M. Wilson (forthcoming). Carnap, the Necessary a Priori, and Metaphysical Anti-Realism.

            A Golden Opportunity: Religious Pluralism and American Muslims Strategies of Integration in the US After 9/11, 2001.

            A Philosophical Examination of the Traditional Yoruba Notion of Education and its Relevance to the Contemporary African Quest for Development.

            How Deep is the Distinction Between A Priori and A Posteriori Knowledge?

            (does discuss Bayesian epistemology at a ‘meta’ level, doesn’t formulate a Bayesian argument)

            Damiano Canale & Giovanni Tuzet (2008). On the Contrary: Inferential Analysis and Ontological Assumptions of the A Contrario Argument.

            The Normative Turn: Counterfactuals and a Philosophical Historiography of Science.

            (mentions ‘prior probability of a scientific revolution’ once. No formal Bayesian arguments as far as I can see)

            The Phenomenal Concept Strategy and a Master Argument.

            What’s in a Name? Modest Considerations on the Situatedness of Language and Meaning.

            Female Under-Representation Among Philosophy Majors: 
A Map of the Hypotheses and a Survey of the Evidence
            (could have even expected to mention Bayes if they were doing any sophisticated statistics.)

            (EDIT: corrected incompetent URL transformation)

          • ncovington89

            You can stick to that all you want, but the fact is philosophers do often use Bayes even outside Bayesian epistemology, and I cited plenty of evidence in direct and clear contradiction of your opinion on that point. If you can’t handle being wrong on something, you are not worth my time for discussion. Good day.

          • John MacDonald

            So you’re saying the formal logic of Bayes Theorem offers veracity of argumentation beyond what we can produce via informal logic?

          • arcseconds

            Well, one of us apparently can’t be wrong about something at least. But are you sure it’s me?

            Let’s just look at this ‘plenty of evidence’ you gave about philosophy.

            You mentioned a single paper on induction. Nothing can be concluded about frequencies from a single paper, and in any case this falls under Bayesian epistemology.

            You mentioned the Blackwell companion to natural theology. This is hardly representative of academic philosophy, and most analytic philosophers would snort at the idea that it even counts as being philosophy.

            And you told me to look at phil papers. That’s an excellent suggestion, so I did that, and I found no papers using Bayesian epistemology in the manner you’re expecting history to do in my sample of 10.

            From that I conclude I was right and it’s uncommon at best for them to be using it.

            So sorry, but you didn’t provide very convincing proof for your assertion that they often use Bayes, and the best evidence we’ve got so far is my selection of 10 papers off the top of a search that doesn’t specify anything about Bayes.

            (I’m sure it’s used often enough for a search on Bayes’s theorem to turn up lots of stuff, but that obviously doesn’t show it’s common. You can show that philosophy papers ‘often’ talk about God that way, but in fact mentions of God in the general literature is really quite rare. )

            (Also, my background in philosophy is pretty strong, so if it were at all common I reckon I would have noticed this by now. Perhaps there’s been a sea change recently or something, but I can’t see any evidence for it.)

            Regarding science, yes, Bayesian statistics is a thing, and there are some areas like phylogeny that make use of it, but this is still a highly specialised, technical, quantitative field. It has not replaced, or even made significant incursions on, the use of natural language arguments in science. That’s based on my experience reading science papers, but I’m sure looking at an arbitrary selection of journal articles will back this up.

          • ncovington89

            Gee, should I bother spoonfeeding you more examples of Bayesian reasoning in philosophy and science so that you can keep making up reasons that all of it “doesn’t count”? Theistic philosophers aren’t serious ones, so they don’t count (what about agnostic or atheist philosophers who use it in philosophy of religion like Stephen Law, Evan Fales, Paul Draper, Robert Greg Cavin, are they not serious philosophers?!?), Bayesian epistemology doesn’t count (even when Bayes theorem is explicitly USED to establish something), Cladistics is a very specialized field so it doesn’t count, Archaeology (didn’t see you respond to that one, but I’m sure you can cook some excuse or other).
            Here’s what I want: I want you to tell me what you’d have to see you to change your mind on this issue and I have a caveat of my own: whatever your request is it has to be something that could be shown to you in the space of a blog comment. Because Lord knows if I cited three dozen papers defeating your claim, you’d just ‘three dozen papers prove nothing, they’re probably just a small minority.’ My Bayesian prior is strongly against the hypothesis that you will offer a reasonable and precise answer to this.
            And moreover, I have read an enormous amount of philosophical literature over the past 8 years or so, and I don’t know how in the world you could have read much of any of it without running into lots of Bayes.

          • arcseconds

            You are wanting historians to replace natural language arguments with formal bayesian reasoning.

            What I would like you to do most of all is to show this has been successful in some other area. Or even successful in history, for that matter. If you can find somewhere where someone has successfully recast a piece of reasoning that is accepted by most academic historians in Bayesian grounds that would advance your argument considerably. It would advance it still more if this was able to elucidate things that were otherwise hidden.

            If you can’t show even by one example how this can be useful, then your argument is really just a promissory note, isn’t it? It amounts to “I think it will be useful on the basis of armchair reasoning”. It would be much more convincing to all concerned if you could show that it is more useful.

            Technical and quantitative areas clearly don’t count. You aren’t arguing ‘well, Bayesian treatment is used in medical statistics all the time! so historians should use it too!’, are you? Of course, when historians actually have stuff that it makes sense to do quantitative statistical analysis on, then Bayesian treatments may well be appropriate. But that doesn’t show anything about the applicability to arguments on the basis of primary texts, which aren’t quantitative or statistical in nature.

            Note that I have never been arguing that Bayesian treatments are of no use anywhere. What I have been arguing is (1) that it’s unclear at best what benefit they’d bring to history, and as secondary matters (2) outside of technical, quantitative, statistical treatments Bayesian arguments are rare in scientific papers, and (3) outside of Bayesian epistemology itself, they’re rare in philosophical papers.

            (2) and (3) are relevant because they show that explicit Bayesian reasoning have not replaced natural language arguments in these other areas, (2) being the more interesting claim here as it could be argued that philosophy doesn’t make too many empirical arguments.

            I have always been excluding quantitative, technical areas, including Bayesian epistemology itself, because they don’t show what you need to show: that it’s a viable replacement for informal reasoning. So you can stop pretending that this is a new restriction that I’m introducing on an ad hoc basis now.

            By bringing up technical, quantitative and statistical areas you are either trying to swap the hard question (can informal arguments be usefully cast in Bayesian form in history?) with an easy one (are Bayesian treatments ever used anywhere?) or are trying to argue against a straw man. I am not, and nor have ever been saying that Bayesian treatments are not useful anywhere.

            As far as philosophy goes, I’m not very familiar with philosophy of religion. So maybe it is in widespread use there. But then you may be cherry-picking the few examples where it is used… what happens if I pick ten philosophy of religion papers at random? am I going to find Bayesian arguments widespread there? I don’t have time to go through this exercise again right now, but I may do so later.

            Note that I was never arguing that it was never used. So searching for ‘Bayes’ doesn’t show anything at all: that’s clearly cherry picking.

            What I expect if someone says something is ‘very common’ is to find it at a high rate in a random sample. How else could a statement like this be proven?

            And I looked at the best random sample I could, given short notice, using a source that you suggested, and I did not find it was very common. In fact, not one of the papers argued for a proposition to a certain probability on Bayesian grounds.

            Why don’t you do what I suggested, eat you own dogfood, and put your own argument that this reasoning is very common in philosophy into Bayesian terms, and show me why I must update my probability of a given paper containing explicit Bayesian reasoning to a probable? It’s a probablistic argument, so you can’t say Bayesian treatments aren’t applicable. If you can’t do this, or won’t do this, for this simple matter, how can you demand that historians do so for much more complex ones?

            (Note that my exercise of a random(ish) sample can be easily put into Bayesian form. Maybe I need to lead by example here? )

          • arcseconds

            Also, I’ve clearly done more work here to establish my side of the argument than you have, so who is spoonfeeding who here, again?

          • ncovington89

            I asked an easy question and haven’t yet gotten an answer. So I guess my consequent probability is in: you don’t have a precise and reasonable answer to my challenge. I think we’re done here, I’m not wasting any more time arguing with you.

          • arcseconds

            If you’re referring to what proof I would require, I have in fact answered that:

            What I expect if someone says something is ‘very common’ is to find it at a high rate in a random sample. How else could a statement like this be proven?

            So find a random sample (that can’t pre-select for Bayesianism in any way) that shows a high frequency of Bayesian arguments. To be worth anything towards your thesis that historians should make use of them, they need to be Bayesian arguments of the sort you expect from historians, i.e. a formalization of intuitive reasoning, using Bayes’s theorem (or probability calculus more generally) to derive a probability for a particular contingent statement. Abstract proofs in Bayesian epistemology (like the ‘wash-out’ proofs) aren’t of any use to historians directly: it’s not a historian’s job to prove formal features of the Bayesian system.

            Let me turn the tables on you: what would convince you that I’m right, and Bayesian arguments aren’t very common outside Bayesian epistemology itself?

            You seem to be trying really hard to treat me as some kind of bad child here who’s intent on playing games, and painting yourself as the person who’s doing all the work here. But that’s entirely unjustified. You accused me of ad hoc rejection of your examples, but that’s not the case. Now you’re trying to say I’m not cooperating: also not the case.

            I understand you may be frustrated with not convincing me that Bayesian arguments are just everywhere, but I have yet to see any convincing proof of that.

            Flounce out if you want, but don’t mistake your pique for a problem on my side.

          • arcseconds

            Also, surely these are probabilistic arguments you’re making. Would you mind formulating them in terms of Bayes’s theorem? Maybe then you’ll see why they don’t add up to much.

          • Ian

            I wouldn’t lean on his authority too heavy there.

            I agree, wholeheartedly.

            You do realize that some very well-trained philosophers (Robert Greg Cavin, for example) AGREE with Carrier that Bayes’ theorem can be used to analyze historical events.

            The problem of quote mining is ever present. I don’t have a lot of time, atm, but a quick flip through some of your citations doesn’t show a lot of understanding of what they’re actually doing.

            1. Bayesianism is not Bayes’s Theorem – they are different things. Don’t cite the former as evidence of the use of the latter.

            2. I am not complaining about the use of Bayes’s Theorem in general, I regularly use it to derive information about past events or unknown properties.

            3. Bayes’s Theorem is provably valid but has a much smaller domain of applicability than the internet Bayes true believers seem to think. This is not a personal failing, the confusion over point 1 is widespread in popular treatments, and the occasional academic paper from a non probability specialist.

            4. Bayesian interpretation of probability can model any knowledge. Bayesian probability is therefore a totally valid tool in history, or any other field.

            5. Bayesian probability is not the only, or necessarily the best tool for knowledge representation. It depends on the situation, the kinds of inference being carried out. Bayesian probability is good at generalising certain kinds of logical inference, but terrible at others. If you limit yourself to only certain classes of logical argument, it is tractable for specific questions. But even then, the calculations don’t scale well for most complex situations (there are some exceptions: it is very successful on large problems, with independent subproblems, for example).

            6. It is not sensible to use numerical estimates in places that contribute only an insignificant quantity to a result. If you want to figure out how fast my car travels, then it is foolish to do a calculation to demonstrate the speed that the exhaust fumes leave the tailpipe when the car is at rest. Though, of course, the exhaust velocity does contribute to the car’s speed, to some extent. Doing a calculation is not impressive: you have to show why it is the correct calculation. Otherwise you’re doing mathematical masturbation. This is roughly my judgement of OHJ, he wants to find the speed of the car, but the calculation he does measures the tail-pipe velocity.

            7. There’s all together too much self-congratulatory back-patting over any calculation on the ‘mythicist’ couches, afaict. There’s also way too much math phobia in humanities generally, I think. I think the former is largely due to the latter: Carrier, like WLC, capitalises on the math-ignorance for a rhetorical impact that the actual math they’re doing doesn’t warrant. My initial review of PH was largely a criticism of this: I didn’t feel Carrier made a genuinely honest attempt to explain and impart an intuition for what the probability calculations were doing under the hood. But he sure told us it was proven enough!

            8. I’ve suggested that Carrier actually try to do some inference with actual Bayesian math, on a data set of reasonable complexity, in any way where he can check his results quantitatively. I think the actual limitations of a number kicked out of running P(A|B) = P(B|A) . P(A)/P(B) will be very clear.

            I apologise generally for replying and running from the conversation. I wish I had more time. It’s frustrating that I’m seemingly to be used as an ‘authority’ in this discussion when a) I’m not much of one, and b) I don’t have time to really explain myself, and clearly I’ve been pretty crap at doing so to date. So, hope that helps to at least say where I’m coming from, even if I don’t have the resources to argue for it, in anything more than a cursory way.

          • ncovington89

            “1. Bayesianism is not Bayes’s Theorem – they are different things. Don’t cite the former as evidence of the use of the latter.”

            I cited Cavin as a philosopher who believed Bayes theorem could be used to analyze historical events. I am open to learning here, but I don’t see how much of anything here in Cavins slides can be understood if this is not, in fact, what cavin is doing:

            See especially his slides 401-403 in which he says that inference to the best explanation criteria of adequacy are beautifully accommodated by Bayes theorem.

            “Doing a calculation is not impressive: you have to show why it is the correct calculation. Otherwise you’re doing mathematical masturbation. This is roughly my judgement of OHJ, he wants to find the speed of the car, but the calculation he does measures the tail-pipe velocity.”

            I thought Carrier spent most of the book arguing his estimates on various issues were correct? I dunno, I reckon maybe you’re referring to some of his more subjective estimates. I don’t see those as being such a big problem, though. I think we know that the continents fitting together is more probable under the theory of plate tectonics than under the hypothesis that the land masses have stayed still since earths formation. We may not know exactly how much more probable, but it certainly more probable to some degree. I don’t see why we can’t, or shouldn’t, model that probability judgement with Bayes, especially if we try to argue a fortiori as carrier suggests.

            “Carrier, like WLC, capitalises on the math-ignorance for a rhetorical impact that the actual math they’re doing doesn’t warrant. My initial review of PH was largely a criticism of this: I didn’t feel Carrier made a genuinely honest attempt to explain and impart an intuition for what the probability calculations were doing under the hood. But he sure told us it was proven enough!”

            I really don’t understand where all these suspicions of deception are coming from. As a good ole southern boy myself, I don’t think it’s okay to accuse someone of dishonesty based on vague, subjective, and ultimately undemonstrable suspicions. Carrier has a bit of a background in philosophy, and Bayes theorem is commonly used to model the logic of arguments within academic philosophy. In fact, the volume “the empty tomb” which carrier contributed to had a couple of chapters in which philosopher Michael Martin made Bayesian arguments against the resurrection.

            As for 8, if you’ll give me a real life example to which the answer is known (or just make one up) I’ll try my hand at it.

          • Ian

            1. Bayes and BT – Some of your citations suggested to me that you confused the two. I wasn’t suggesting none of the citations featured BT.

            Apropos of nothing, I couldn’t figure out, on Cavin particularly, if he was confused or being confusing. Slides are difficult, because I don’t know what he said, and I confess I didn’t read all 400+ slides (glad I didn’t sit through that presentation!), but prima facie, the slides look a mess, and the same kind of bombastic pseudo-math for rhetorical effect I see in Carrier. The article you cited on Bayesianism, I thought was very good though. But my opinion on both is irrelevant to your point, which was that other people refer to BT in thinking about historical topics. I think citations outside debates, in papers or monographs might make your point stronger, but it isn’t my argument, either way.

            2. I think you missed the point I was making. Sorry if I was confusing. The estimations aren’t as significant as what estimations you choose to estimate. By putting numbers to them, it makes the end-of-the-process seem like the significant thing. Even if it turned out to be the right thing to focus on, the error involved in the problem-selection dwarfs the error in your result, making the result largely irrelevant.

            3. Suspicions of deception, is too strong. And dishonesty? Not quite. Sorry if saying he didn’t make an honest attempt implied ‘dishonesty’ – if it did, feel free to mentally excise that word, I’m happy to withdraw it. I don’t think Carrier is dishonest, personally (seems that word is rather too freely used at the moment). I felt PH was tendentious in its treatment of probability. It used math as a way of insinuating a rigour for his theses that aren’t justified, rather than stepping through the probabilistic representation of knowledge in a way that I thought might help Historians understand when it might be useful, and how. The example I gave is the fact that he doesn’t describe what BT is derived from and why and how. It appears in PH like some standalone incantation of probabilistic reasoning, rather than one of a countless number of probabilistic identities that follow trivially from the meaning of its constituent parts.

            8. How about taking the Netflix prize data and test cases?

            Not sure I’m going to have chance to reply later, I’m afraid.

          • arcseconds

            Am I leaning too heavily on your authority?

            Surely as the only person here who has a provably solid mathematical background and uses Bayes’s theorem in their professional life, you are the person in this forum whom anyone genuinely interested in the applicability of Bayes’s theorem to historical problems (or empirical problems more generally) should most be listening to.

            And the fact that you’re pretty sceptical of its applicability should give enthusiasts pause. It’s certainly given me pause, and I was way less enthusiastic than nconvington89 to start with!

            Note that I didn’t say ‘Ian says it, therefore it’s true’, I just encouraged nconvington89 to read your stuff. I’m afraid the only way to get me to not recommend your stuff is to not write any of it! Or write arrant nonsense. Or a convincing counter-argument under another name.

          • Ian

            Sorry, wasn’t ‘frustrated’ at you, but at mostly at my lack of time at the moment. My opinions being discussed without me having the time to contribute.

            I don’t think you implied ‘Ian said it, so it’s true’, but I’m not sure it is sensible to refer to expertise with folks who’s entire thesis is based on opposing a mainstream academic position. Carrier told me he had shown his work to a tenured mathematician who’d said it was correct, so that’s my authority refuted anyway! 🙂

            I find tracking these arguments like nailing jelly to a wall, to be honest, terms are used so vaguely, methods are so fluid, and citations are so tenuous, that I’m just unable to respond clearly. At least I read what I’ve been writing at 3am the last week or so, and I don’t understand what I was getting at.

          • arcseconds


            Well, mythicists aren’t officially committed to undermining all expertise, just biblical scholars. And one hopes that some strong bias towards the non-existence of Jesus isn’t also accompanied by an equally strong bias toward the Bayesianization of history.

            And I guess expertise ends up being a double-edged sword. Unfortunately, mentioning a philosopher giving Bayesian arguments for something does nothing for me: I’ve too much experience with second-rate philosophy. Some will give any kind of argument for anything so long as they can get it published somewhere, and we all know Bayesian-style arguments lend themselves well to justifying nonsense. Plus unfortunately many of them do not really understand what they’re talking about. I’ve definitely seen philosophers commenting on relativity in a way that indicates they really don’t understand it, and I’ve every expectation this applies to Bayesian stuff too.

            The very best philosophers of science do have a good understanding of what they are talking about though: some of them have strong backgrounds in physics and mathematics, and they weren’t necessarily second-raters in those areas, either. This doesn’t automatically confer correctness of course, but they aren’t people who are fudging along with an early undergraduate understanding at best, and pretending they’re an expert.

            Unfortunately, in this circumstance it ends up being a judgement call as to whom to trust, and I’m obviously going to prefer my judgement to nconvington89’s.

            I must admit I’m not expressing myself as clearly as I could hope, either. One thing I’ve been trying to get across to nconvington89 is that the sorts of treatment one finds in Bayesian epistemology, while of course they’re Bayesian arguments in a perfectly good sense, do not entail or even really suggest that Bayesian approaches are directly applicable to empirical reasoning.

            Even assuming Bayesian epistemology is the one true epistemology, just because everything in principle can be justified on Bayesian grounds doesn’t mean it’s sensible to always do your reasoning with Bayes’s theorem. Maybe one could liken it to the reduction to set theory in mathematics: even if one accepts that everything is really sets under the hood, it doesn’t mean it’s sensible to get out the ZFC axioms to add up your weekly accounts. But even that analogy doesn’t help much: if you did it correctly of course you could do your accounting in set theory, whereas the gap between ‘in principle’ and a systematic Bayesian treatment of a historical argument seems much larger.

            But my attempts to make this point has been seemingly interpreted as a dirty trick to avoid discussing Bayesian epistemology…

            It’s an interesting question as to who would be the best person to review Carrier’s book. A mathematician can check that he hasn’t made any mathematical mistakes, which I imagine is what was done, but isn’t necessarily the best person to point out that most of the argument isn’t actually being carried by the mathematics. And even if the reviewing mathematician did remark on this, is Carrier really likely to admit it, especially as he likes to style himself as a philosopher? It seems easy for him to say “well, your expertise is in maths, and you’ve checked that, and there’s nothing wrong with it, let me worry about the rest of the argument”.

            At some point I’d like to come back to what you were saying a few days ago about having to drop the assumption that probabilities have to add up to 1… once I think of something coherent to ask about it…

          • Kris Rhodes

            What did you read by mathematicians arguing that Carrier doesn’t know what he’s doing? The only piece of writing I know about right now along those lines is a series of blog posts by Ian Somebody (I forget) which, deep in the subsequent comment threads, seemed to become much less devastating as the two of them talked it out. His criticisms seemed ultimately to be more about clarity and pedagogical value IIRC, and this became much clearer in the comments.

            Are there other mathematicians you have in mind?

            Having said that, though I’m sympathetic to Carrier’s arguments, I can attest to the fact that he mishandles some technical logical vocab in Proving History. As it turns out, on my reading, despite his misuse of terms, what he is saying turns out to be right (or right enough for his purposes) in the relevant passages of that book.

            From a logical standpoint, what he says about Bayes seems to check out as far as I can tell, (and what I think “I can tell” based on my own training is whether or not a person’s logic checks out,) even IF it turns out his use of it is somehow at odds with common practice.

          • Neko

            “Ian Somebody.” That is funny. Ian is a scientist. He has a Ph.D. in the mathematics of evolution. He also has an undergraduate degree in theology. He is an atheist.

            The following is his review of Richard Carrier’s Proving History:


            Here is another critique, by cosmologist Luke Barnes, that was pointed out to me the other day:


          • Kris Rhodes

            Yes, I’m aware of Ian’s status as a scientist with a PhD in a field of mathematics, that is why I mentioned him in relation to this topic. I don’t remember his last name. But I figured you probably knew who I was talking about. Not sure what you’re taking issue with.

            I read his review, as I indicated, and as I further indicated, it becomes clear especially in the comment thread to the three posts comprising that review, and in a comment thread in a relevant post on Carrier’s own blog, that Ian’s criticisms are more about Carrier’s clarity and the pedagogical value of his approach. I’ve mentioned this to Ian and he didn’t agree, but it’s all laid out there in the comments. I can’t see where Ian actually takes on Carrier on the substance of the question of whether Carrier is engaging in a sound application of BT, even if Ian takes issue of which (mathematically equivalent) form of BT Carrier uses.

            I’ll take a look at the Barnes one, thanks for pointing it out.

          • Neko

            I beg your pardon, I didn’t think we could’ve been talking about the same posts.

            You wrote:

            I can’t see where Ian actually takes on Carrier on the substance of the question of whether Carrier is engaging in a sound application of BT, even if Ian takes issue of which (mathematically equivalent) form of BT Carrier uses.

            This is a red herring. There is a digression about the long v. short form of the theorem but ultimately Ian argues that Carrier does not engage in a sound application of BT. But…Ian’s around, and he can speak for himself.

          • Ian


            There are a few problems with Carrier’s use of Bayes, that I can see:

            1. He deals with probabilities rather than probability distributions, and so he can hide places where BT is not well behaved in his choice of numbers. BT gives you garbage results in many situations, and for any practical use, you have to do error and sensitivity analyses.

            2. He does not properly deal with the considerable literature on the Bayesian/frequentist problem, the problem of priors, or the effect of such problems on the results. Instead, he claims to have solved this problem: a fact that would be of considerable use to those wrestling with the issue on real data, were it true.

            3. He does not properly deal with the considerable literature on the problem of reference classes, or their impact on the results of Bayesian reasoning. Again he claims to avoid these problems, but doesn’t display any indication he understands them. Again, a dependable heuristic to solve this problem would have huge impacts (probably about 90% of the Bayesian prediction I do involves trying to find reference classes that are at all useful).

            4. He does not consider or explain the effects of non-independent data on BT. Assuming that, because BT can be used to combine non-independent data, that such dependencies are somehow automatically accounted for by the method.

            and, on ‘tone’ or ‘clarity’:

            5. He uses BT rhetorically as if it is the only reliable basis for making judgements about truth claims, thus implying that any scholar not doing BT is at best hiding their biases. Yet, by not addressing the actual practical problems of doing Bayesian reasoning on real world data, he is being less honest about his reasoning, not more. He’s happy to make very small assumptions explicit, but only by at the same time hiding the vast majority of his assumptions behind the same kind of narrative arguments.

            6. The fact that he doesn’t attempt to present the probability theory in a way that communicates what is happening and why, makes it seem like BT is primarily an exercise in post-hoc rationalisation. Very much as those who claim it demonstrates miracles. This is typified in the way he fixates on a form of BT that makes it much more complex than it needs to be. For Carrier, BT is a rhetorical tool, not a mathematical one.

            (there may be more, I’ve not gone through his work in detail for a couple of years).

            The take home message of my three part discussion of probability on my old site is this: Bayes Theorem applied to Bayesian probabilities gives you out results that is less reliable than the values you put in. Carrier’s results are just less accurate versions of his own guesses dressed in mathematical clothes for rhetorical effect.

            He has, of course, narrative explanations of why he is right on these fronts. But it is no coincidence that at no point does he ever use his method to derive quantitative results on complex data that can be checked.

            That’s not surprising, it is hard. So instead he relies on toy examples (always frequentist, or frequentist priors with a hand waved Bayesian conclusion), and appeals to the fact that BT is mathematically proven.

            I get the sense he thinks that the way we teach probability theory in 101 classes, or found in popular books like ‘The Theory that Would Not Die’, is the ‘real’ math, rather than just a simplified pedagogic introduction. Thus to solve problems in Bayesian theory he can simply deal at that level and not engage with the mathematical literature.

            When I have tried to engage with him, he’s referred me off to various discussions of these problems, where he addressed them (narratively, of course) and asked me to respond with specific counter arguments. At the same time insisting his work has been peer reviewed by a mathematician who was happy with his conclusions. Thus taking an actual mathematical problem, and reducing it to a rhetorical challenge, reversing the burden of proof, and arguing to an unknown authority.

            He is undoubtedly a skilled rhetorician, he makes his living from being a writer, and defending his thesis is his full time job. He has the time to atomise any argument, respond atom by atom and weave a convincing story. He would — undoubtedly — hand me my ass if I tried to step through these problems adversarially. My debating skills don’t get much beyond telling people I’m right and behaving self-righteously: I get too emotional and I don’t have the tactics. But at the same time he is not willing to submit himself to learning humbly, because he views his thesis as being correct and demonstrated.

            That I don’t take up that rhetorical challenge (both from lack of will and lack of time) he has interpreted as evidence I am either wrong, or have withdrawn my objections.

            If he can demonstrate his method in any way that can be checked, then it would be interesting to a wide range of people who use statistical and probabilistic reasoning to determine what can be known and with what confidence. This is a big constituency: every Fortune 500 company has full-time employees (some of them have dozens, financial institutions have thousands) wrestling with this. It isn’t my main mathematical speciality, but I could fill my time with contracts doing it, if i wanted. When I do, I deal with big data, often very messy, often self-contradictory, occasionally financial, but more often behavioral. I’d love it if something as simple as P(A|B) = P(A)P(B|A)/P(B) were able to used to give useful results, in the optimistic blanket way Carrier portrays.

            This is a bit annoying because I’m actually quite amenable to a lot of his conclusions. I do, for example, think that things we think we know are less firmly rooted that we believe (something I see a lot among experts in a set of data, compared to quantitative results), I do think that strict ‘criteria of authenticity’ in the vain of the Jesus seminar is dubious. Though I am on the historicist ‘side’, my (amateur) intuition is for a more minimal Jesus than I think most scholars would propose. But all I get from Carrier is bombast, rhetoric, arrogance and derision. And other than the odd sniping blog comment, it’s not my job, or worth any great effort to ‘oppose’ him. I’ve too much real math in my inbox 🙂

            [FWIW I don’t use my surname, not out of a desire for anonymity, but because I work for myself, and with such a niche set of skills, the popularity of either this or Carrier’s blog would easily swamp any search results for my name, and I’d rather keep those searches going to stuff where clients can find me. I’m always happy to correspond from my own email address with anyone who wants to.]

          • Ian


            7. He excludes the middle, so explanations are anti-correlated. This is one of those cases where it makes a lot of rhetorical sense, but if you actually do it on real data, you get nonsense results.

          • arcseconds

            What middle do you see there being between historicity and myth?

            I suppose options where one of the stool-legs of the ‘minimal historical Jesus’ is missing might count, for example a Jesus that was an apocalyptic preacher and was crucified, but this happened in 10 BC or something, might count?

          • Ian

            When you try to do a series of calculations on a P(A) or P(B) pair, where you’ve assumed P(A) + P(B) = 1, one of the horns tends to go down (to unknown) and the other goes up. So you may end up with P(A) = 0.01, P(B) = 0.99. This can give you the impression that you’re demonstrating which one is true, overwhelmingly, when actually it is an artefact of the calculations you’ve performed.

            You get better results assuming an (unknown) middle, and ending up with, say P(A) = 0.02, P(B) = 0.01, and *then* excluding the middle, to conclude that A is (say) twice as likely as B.

            The reason is that a bayesian probability is an encoding of a lot more than it first appears. You’re never actually dealing with the correct numbers, you basically need complete knowledge of the state of the whole system at each calculation to do that. [which is why it is tractable in frequentist cases where you can define that system]. In reality you’re always dealing with some proxy you can measure, and you can’t control for all dependencies. And by assuming P(A) + P(B) = 1, you’re inadvertently biasing all your numbers, in a way that is hard to detect.

            It’s exactly the kind of problem that a narrative argument will resolve simply: of course P(A)+P(B)=1!, but if you do it on real data, you’ll often end up with results that don’t match what you can measure.

            On a side note, excluding the middle isn’t just a problem with missing a third option, but also missing cases where both are correct. That is particularly insidious when you have categories that are loosely and wordily defined.

          • arcseconds

            Yes, I understand the problem with assuming that P(A) + P(B) = 1 when in fact it doesn’t. If there’s a significant other possibility C, then by arguing against A you increase P(B ∪ C), which is not the same thing as increasing P(B), and is particularly a problem if you’re trying to argue that B is nearly certain.

            However, prima facie either Jesus existed or he didn’t: “Jesus half existed” or “Jesus sort-of existed but really more subsisted” or “Jesus isn’t the kind of thing that can either exist or not exist” aren’t statements that make a lot of sense.

            So at the very least thinking that history and myth exhaust the possibility space has some superficial appeal.

            So I’m wondering what exactly you see as being neglected in this particular case.

            It also seems to me that if P(A) + P(B) ≃ 1, so that P(C) is small but not zero, you can’t decrease P(B) significantly without coming at the expense of P(A).

            And if Bayesian reasoning is going to be at all useful in any way, it kind of has to be able to be somewhat robust in the face of the approximation P(C) = 0, because there’s an awful lot of possibilities that in principle we shouldn’t assign a probability of 0 to, because they are still possibilities, however unlikely (Jesus was a space-alien, Jesus was a time-traveller, brownian motion and quantum fluctuations reassembled Jesus after death, etc.) but nevertheless aren’t sensible to consider.

            Naturally this means that we’re always at risk of ‘something else’ that we hadn’t considered being the case, but as far as I know that’s totally unavoidable.

            In other words, I think there’s some basis for Carrier arguing that he can in fact argue against A and thereby shore up B in this particular case.

          • Ian

            I think you missed my point, sorry if I was unclear – I’ll respond when I’ve got more time tomorrow.


            Yes, I understand the problem with assuming that P(A) + P(B) = 1 when in fact it doesn’t.

            wasn’t the problem I raised.

          • arcseconds

            I suppose I have been assuming that this is about “I’ve argued against the historicity of Jesus, so I’ve therefore argued for him being a myth”. If Carrier does this in other cases, then he’s on his own 🙂

          • arcseconds

            ah, well, it really seemed to me that this was exactly the problem that you raised 🙂

            I wait the clarification with bated breath…

          • Ian

            Now I’m not writing at 3am, I’m not entirely sure what I was meandering towards. Thanks for the clarification with Kris too: no respect was due.

            Okay, um. There is a practical issue I was fixing on:

            You tend to get lower probabilities for those you can update more easily (probabilities tend to decrease with information, as a very loose approximation). When you exclude the middle initially, you tend to get lower probabilities for the cases you have more data on, and this inflates the other cases which are locked to summing to one. A practically better solution is to not assume they are exclusive. I agree there are cases where theoretically you should be able to assume exclusivity: but even in those cases it often isn’t good in practice. Here’s an example of something I’ve worked on. Given the browsing history of a visitor to an ecommerce site, what is the probability they are male or female? The result of the calculations had a higher accuracy if I didn’t assume people were either male or female up front, but only imposed that dichotomy at the end of the process. If we did do it first, one of the genders ended up acting as a kind of ‘default’ in a way that the data didn’t warrant: in a way that was a function of what data we ran, rather than the content of the data. I think, if you could do the calculations Carrier suggests (and I think that is very unlikely), you’d have this problem.


            I don’t know. But I suspect it is a problem when you get beyond toy conditionals. So you’re dealing with P(X|a,b,c,…,m), say, which means to update the probabilities you need to be able to determine P(N|a,b,c,…,m), but that’s very unlikely. Most often you’re actually updating with P(N|a,e), or some such: something the data actually provides. You might think that P(N|a,e) is effectively P(N|a,b,c,…,m), but even if it is close, those errors compound. So the effect is that your probabilities include cases you should have excluded. Over multiple instances of this, you end up with bias that is purely down to the number of updates, not the content of the update. That’s my intuition, though. All I’m confident is that the bias exists in practice.

            A theoretical problem:

            Carrier’s case is not “Jesus existed” vs “It is not the case that Jesus existed”. It is “Myth->Man” vs “Man->Myth”. He’s aware that is a continuum, but one he feels he can put a dividing line somewhere. This seems to me to be an excellent place to put a rabbit in the hat. By not being consistent about that dividing line, or by treating bits of data as supporting one end of the continuum, without guaranteeing that there is no data in the other end, he’s in danger of having something that is, in fact, non-exclusive.

          • arcseconds

            That’s very interesting. Thanks.

          • Jim

            I’m one of those people who assumes that stats are for
            finding where you last left your car keys, so talking to me about anything in this field is like talking to a soap dish. Nonetheless, I do have a strange question that if/when you get some time, I’d appreciate your thoughts on.

            I realize that applying BT to something like history is different from applying stats to hard numerical data where you can carry out linear regression analysis, generate standard error bars etc. So what I’m wondering is, is there any way to come up with a some type of estimate similar to a confidence level or maybe a maximum likelihood estimate for probability outcomes from a BT analysis of say the historicity of Jesus.

            Or is there just no practical way to get a handle on a sense
            of (relative) accuracy on outcomes obtained for this type BT application? Or maybe this is just a totally meaningless question for historical analysis applications?

            I’m just wondering about this from the perspective of how to
            appropriately present the data, and in a way useful for comparison purposes. I mean if the obtained probability for an HJ is for example 32% plus/minus 200%, well …

          • Ian

            You can work through BT by hand with a set of values, to see how it responds to errors in your data, yes. And if you input a range, you’ll get a range (there are some fiddly bits, because you have to determine whether the errors are correlated, but it is possible). I’m not aware of anyone who’s done it, though.

            It wouldn’t be needed with, say, Carrier’s Rank Raglan numbers, because that’s just a description of some data. Once he then uses that as some kind of proxy for a historical conclusions, it moves into the realm of hopelessness, imho. I just don’t see how you’d do it.

          • Jim

            Ty for your insights.

          • Kris Rhodes

            Wow thank you for this post, I appreciate it!

            What’s the reference class problem exactly? It seems like Carrier thinks it can be taken care of by noting that if you’re “doing Bayes stuff” (my phrase) on a set of probabilities, you can just go ahead and pick any one of them out and call it the prior probability (and IIRC prior probability is a function of the reference class). The math “works out” the same way no matter which of your “inputs” you use as the prior. What is the problem with this way of addressing the reference class problem? (Or am I remembering wrongly that the reference class problem is basically the same thing as the problem of how to choose your prior probability?)

          • Ian

            The problem isn’t just w.r.t. priors, it can appear at any point.

            The idea is that the situation you’re considering is simultaneously a member of many (possibly infinite) classes, by virtue of its properties. When you treat some measurable data as a proxy for a likelihood, you inevitably have to find that data based on a subset of those properties. But there are basically unlimited ways to do that, even assuming those properties are crisp and unambiguous. It isn’t enough, in practice, to pick the ‘closest’ class, based on some narrative argument or justification.

            The math definitely doesn’t work out the same. In fact, the choice of reference class dominates the result.

            Even in cases where we can check the result and work out which the best classes are, that often turns out to be coincidental, and prediction gets worse on successive data.

            You can combine statistics based on multiple reference classes, in theory. In practice it is very difficult, because the statistics on those classes are rarely independent, and correcting for that dependence is very hard. And if you don’t, you’re just counting your biases multiple times.

            Kahnemann won the Nobel prize in economics for solving a very constrained version of this problem, as part of his ‘Prospect Theory’. It can give good results, but isn’t applicable to Carrier’s concerns.

          • Kris Rhodes

            I think I understand what you’re saying.

            I was going to reply something along the lines of “Can’t you do the Bayes thing to the evidence you’ve currently got in mind, including all the biases contained in your choice of reference classes etc etc, and thereby come up with an answer to the question ‘how certain should I be given my own assumptions,’ then update as your interlocutors come to the table with different reference class choices and different probability sets based on those, just do a Bayesian update based on all that — assuming their probabilities are accurate given their reference classes — and thereby come up with a new, updated answer to the question of how certain you should be, and so on as the discussion continues?”

            But on re-reading your post I think something like this is what you’re alluding to when you mention statistical approaches aggregating information across multiple reference classes, and what you’re saying is that something like that may be valid in theory (?) but in practice is basically impossible.

            Does that sound about right to you?

            Though I know Carrier argues that he’s a frequentist, and I know people take him to task for seeming to elide between frequentist and other interpretations of probability as it suits him, I myself have never been sure why it’s particularly important whether he’s a frequentist or what, because it seems like the main idea of his argument only requires that we interpret probabilities as degrees of certainty. Whether these correctly correspond to frequencies or not seems like a side issue to me.

            So for example, I think I’m thinking along roughly Carrieran lines when I think the following:

            The K-axioms are modeled by certain rational constraints on degrees of certainty, and Bayes follows logically (given a definition of conditional probability) from the K-axioms, hence these rational constraints on degrees of certainty are also a model for Bayes’ Theorem. This means that if I interpret Bayes’ Theorem as being “about” degrees of certainty, then when I “input” my set of degrees of certainty into Bayes, the result is guaranteed to be the degree of certainty I should have concerning whatever question it is I was trying to answer using Bayes. This is not a factual claim about how probable something objectively is, it’s a narrower claim about how probable I should think it is.

            What in that reasoning seems flawed to you, and does it seem to you to be more or less than what Carrier is claiming for his use of Bayes? It seems to me he positively hammers on the idea that he’s using Bayes something like we might use a set of logical rules like modus ponens etc, as a way for us to talk to each other about what _really follows_ from what. Sometimes people accuse him of claiming a certain kind of objective certainty that I’m just not seeing as I read him. To the contrary, he ends his book (albeit somewhat combatively sounding) with a clear invitation for others to add their own evidence and probability estimates into the mix–the dialogue as a whole being intended to help correct for the biases of the individuals taking part in it.

            I want to say (or more properly in most cases, ask) things about a lot more of what you said in your long post above, but right at this particular moment it feels like a ‘missing forest for trees’ moment because it feels like (“feels like” because I recognize there’s likely to be a lot I don’t know that I don’t know here) you’re expecting Carrier to be using Bayes for purposes Carrier doesn’t claim to be using it for. The basic problem is, I’m not sure right now how to translate a lot of your objections into terms that have to do with discussions in a field so much broader and more imprecise than one like evolutionary mathematics. And right now, I’m not sure the objections actually _apply_, exactly because they would only apply to the more precise kind of work you seem to have in mind.

            I don’t suppose you would be willing to do me the favor of giving a kind of toy illustration of the way you think results of the use of Bayes in something like the way Carrier does it, in the context of an imprecise soft science like History, will yield useless results?

            This post was a bit of a meander, I wrote a lot more than I intended, and I now have to get to my office for work, so my apologies for that.

          • Ian

            Long posts are interesting, thanks!

            1. On multiple reference classes: It is easy to say ‘just do the updates with Bayes’s theorem’ – but what that actually entails is not simple. One example. If you have P(X|A,B), you can’t update with P(C|X), you need P(C|X,A,B) – in other words, every update needs to take account of all previous data included.

            2. On ‘confidence’: This is yet another interpretation of probability. And, to be honest, is one I’m very happy with. Happier than Bayesianism, at least. But you have to be careful. Probabilities as confidences have slightly different math, you have to be clear what you’re doing, and stick to it. So, for example A or ~A does not imply P(A) + P(~A) = 1. I.e. Just because I am not confident it will rain tomorrow, doesn’t mean I am confident it will not rain tomorrow. In fact, there are good arguments why P(A and B), when A and B are independent, is not necessarily P(A) x P(B). It may be that P(A and B) = min(P(A), P(B)), in that case, which in turn changes the form of Bayes’s theorem (I’m not advocating that view, just pointing out that it is made seriously).

            Carrier, if this is his interpretation, doesn’t make that interpretation clear, and certainly doesn’t use that interpretation in his math. I didn’t get that at all from Proving History, it seems at odds with his ‘cheeky’ solution to the Bayesian/frequentist problem. If he occasionally seems to imply that he’s interpreting it that way, then that’s further evidence he’s got some major methodological problems.

            I didn’t know Carrier considers himself a frequentist. To be honest that just makes me totally despair of his understanding. Given that his entire thesis is Bayesian.

            3. On Carrier’s use of probability theory. My specific math problems do miss the forest for the trees, yes. Here’s my ‘forest’ problem. Or an analogy. Imagine devising a Bayesian process for interpreting an image. Is this an animal or a person? You find a set of reference classes, build a probability and give a result. Human’s are involved: they have to find the reference classes, fill in gaps with estimations, and so on. But then we crank the math and get a number, which should be a representation of our inputs. We do this. But the results aren’t very good, even with the most careful inputs. You’re very likely to end up with a system that interprets an african American guy as a gorilla (an example from Google’s system).

            When a person makes a judgement on an image, they actually deal with a lot of very complex interactions between information, experience and reasoning. Interactions that aren’t input->algorithm->output, but use judgement all the way through. Trying to encode how an expert thinks in Bayesian probability is mindbogglingly difficult. It was thought we could do this, back in the 80s, to build expert systems. Outside a few high profile wins, they didn’t turn out to be very successful; mostly because it proved intractable to model expert decision making in that way.

            Now Carrier isn’t concerned with mindboggling complex formulae. He is trying to pretend that simple Bayesian reasoning gives you useful information about historical questions. Further, he wants to suggest that anyone who wants to think about the complex process should reduce them to simple Bayesian formulae. Rather than complex accumulations of intuition, knowledge, reasoning all mixed together.

            Carrier is an insurgent: he wants to undermine expert intuition, because he wants to argue that all the people with the most experience and knowledge are wrong, as he is right. So he’s setting up a pseudomathematical scheme, in an attempt to get the methodological high ground. But the extent his approach is valid (and it is valid, to a certain extent), it is not relevant (or, at least he goes no way to demonstrating it is). If he wants it to be relevant, he needs to make some significant progress on the theory: theory he seems unaware of.

            If he was able or willing to use probability theory on a complex data set, in ways he could check, I think he’d understand immediately how little he knows, and how irrelevant the kinds of problems he wants to focus on.

          • Kris Rhodes

            Quick question about one thing you said, while I’m thinking about the rest of what you said.

            It concerns interpreting probabilities as degrees of certainty, and the axioms as certain rational constraints on degrees of certainty. You said

            “So, for example A or ~A does not imply P(A) + P(~A) = 1. I.e. Just because I am not confident it will rain tomorrow, doesn’t mean I am confident it will not rain tomorrow.”

            I don’t understand this. It’s true that me not being confident that it will rain tomorrow does not mean I am confident that it will not rain tomorrow. But not being confident it will rain tomorrow, I would think, means that for some X, my degree of certainty that it will rain tomorrow is less than X. (Where X denotes some “threshold of certainty” appropriate to the context.) Call my degree of certainty Y. My degree of certainty that it will rain is Y (which is X). ?

            In other words, however confident I am of some A, my degree of confidence that ~A should be one minus my confidence that A. I can imagine maybe this being seen as _debatable_ for some people, but outright clearly false? I’ve either completely misunderstood you or I just plain disagree I guess…

          • Ian

            I don’t understand what role X was playing in your explanation. Can you explain that.

            Normally, the degree of certainty in something you have no knowledge about is close to zero. If I ask ‘how confident are you that it will rain, here, where I am, tomorrow?’ if you answer 0.5, then you can rescue P(A) + P(~A) = 1. You’re not talking about confidences in the way we normally think of them, it will be interesting to unpack your view a bit more. I wonder what difference in practice such a scheme would give to Bayesianism, though.

          • Kris Rhodes

            X didn’t really play any necessary role, now that I re-read it, but to explain what I meant:

            You know how sometimes in everyday conversation, if you ask someone “are you certain” they’re okay to reply “yes” if they’re, say, 90% sure, while in other contexts, they should not say “yes” unless they’re 99% sure?

            Those two percentages (90% and 99% respectively) would have been values of X in my explanation. That’s what I meant by calling X a “threshhold of certainty.”

            I know that was confusing because probably the word “certainty” more properly brings to mind absolute certainty, i.e. a degree of confidence of 100%. So let me re-say what I was saying, without this confusing talk of threshholds of certainty.

            Basically: Isn’t it true that if my degree of confidence in P is V, then my degree of confidence that not-P should be 1 – V?

            If that’s true, then I don’t know why you say that A v ~A fails to imply that P(A) + P(~A) = 1.

            But if that’s not true, I don’t know why it’s not true. You said “Just because I am not confident it will rain tomorrow, doesn’t mean I am confident it will not rain tomorrow,” but I’m having trouble seeing how that would show that Av~A fails to imply that P(A) + P(~A) = 1.

            I take “I am not confident it will rain tomorrow” to mean “My degree of certainty that it will rain tomorrow is less than the value which would indicate confidence.” You’re right that this doesn’t mean I am confident it will not rain tomorrow, but that’s not surprising to me–subtract my degree of certainty that it WILL rain tomorrow from the number 1, and the result can still be less than the value which would indicate confidence.

            Using variables:

            I take “I am not confident it will rain tomorrow” to mean there is an X such that a degree of certainty equal to or greater than X would indicate confidence, and my degree of certainty that it will rain tomorrow is less than X.

            If I assert the statement as just defined, this assertion is compatible with the following second statement, (even if, as I said above, one’s degree of certainty for V must equal one minus the degree of certainty for not-V): I am not confident it will NOT rain tomorrow. That is just to say that one minus my degree of certainty that it WILL rain tomorrow is ALSO less than X.

            (So for example, perhaps X = 75%, and I am only 55% certain it will rain. That means I am 45% certain it will not rain. Both of these values are less than X, i.e., the value that would denote confidence. I’m not confident either way.)

            Well the above was clear as mud. I’ll offer the following very coarse grained summary as a way of helping the reader (and myself) articulate what I’m trying to say and what strategy I’m using to say it:

            You said “Just because I am not confident it will rain tomorrow, doesn’t mean I am confident it will not rain tomorrow,” which I’ll now abbreviate as

            K: ~[c](R) does not imply [c]~(R)

            And you’re saying that K, in turn, means that:

            L: A v ~A does not imply P(A) + P(~A) = 1

            But what I am saying is that K is compatible with the negation of L (and hence that, contrary to what you seemed to be saying, that K is true does not mean that L is true). I’ve argued for this by giving a scenario (the one with the 55% degree of certainty that it will rain outside) which satisfies both K and ~L. (In that scenario, “~[c](R)” is true and “~[c]~(R)” is true, hence ~[c](R) does not imply [c]~R. Moreover, in that scenario, degrees of certainty have been treated such that A v ~A _does_ imply P(A) + P(~A) = 1.)

          • Jim

            Even though I know squat about BT, I’ve found your dialogue with Ian very informative. Maybe I’m weird?

            So to help me in trying to catch up: you are arguing for normalization P(A) + P(~A) = 1 (as is typically applied to hard numerical data/means/etc.), while Ian is saying that for applications of BT for predicting historical probabilities, normalization is not as useful/relevant of a strategy as are the ratios (because of the nature of the input), or something like that? Or am I on the wrong planet?

          • Kris Rhodes

            I think in this particular sub-topic of our conversation, it’s that I am arguing that normalization applies to plausible rational constraints on degrees of certainty, while Ian thinks it doesn’t. Does that answer your question or did I misunderstand?

            The relevance for the larger discussion is that I was trying to figure out whether we can at least use Bayes, even if “sloppily,” to accurately-enough check ourselves on the inferences we make, by thinking of it as a way to determine how certain we should be of various propositions. Ian was saying (I think) that that won’t work because it’s not clear that the probability axioms (from which Bayes is provable) are accurately modeled by degrees of certainty (in other words, it’s not clear that normalization applies to the domain of degrees of certainty).

            That probably confused things even more! 😉

          • Jim

            Sure my head is still spinning, but thanks for your summary. It has helped me as I am trying to understand the nuances of each of your arguments. These insights will help me for when I get around to going through OHJ. So thanks again.

          • Ian

            Right, you’ve got two separate things.

            1. A numeric function, P, representing the degree of ‘certainty’ (in your phrasing).

            def: A v ~A -> P(A) + P(~A) = 1 [as for normal probabilities]

            2. A boolean function representing ‘confidence’, C, which is a threshold on the degree of certainty.

            C(A) -> P(A) >= X [for some ‘minimal confidence’, X]

            Then you are correct. Under your scheme, you can be both ‘not confident in A, and not confident in ~A’ at the same time.

            The difference to my original claim was that I didn’t use a threshold. function for ‘confidence’. The ‘confidence’ I referred to was your ‘certainty’ metric.

            Your version is fine, you get to define stuff however you like (one thing I get frustrated with Carrier in PH, for example, is that he seems to suggest that probability has a specific and platonic meaning: this is what probability *is*, rather than it being like a ‘number’ – capable of encoding and representing a large number of epistemic ideas). My only concern with your version is that, I’m not sure what your P value, ‘certainty’, does, that a Bayesian interpretation doesn’t.

            Given that’s your definition, I should now retrace the rhetorical steps and look at the argument you were making with your definition.

            The advantage of using ‘confidence’ or ‘knowledge’ interpretations of probability (in the way I meant those words) is that you effectively get an extra dimension to capture information about the world. You are capturing knowledge of the likelihood in a way that is orthogonal to the likelihood itself. That is very useful, and is why I tend to use this approach.

          • Kris Rhodes

            I was really hoping to hear back from you about my last post, explaining what role X played in my explanation and attempting to re-argue my point more precisely. It may have been a bit of a mess, but I think the symbolized version of the argument’s overall structure at the _end_ of my post may have helped clarify what I was trying to say throughout the post.

            Basically, independently of Carrier and Bayes and so on, I’m really interested in general in the idea of the interpretation of probability (esp. as degrees of certainty). If it is clear that normalization doesn’t apply to rational constraints on degrees of certainty, that’s definitely something I want to learn about. Hence my last post…

          • Ian

            I just sat down to respond 🙂 Sorry I’m so busy. Hadn’t forgotten though.
            Will reread and respond now.

          • arcseconds

            With due respect to Ian, I’m not sure his answer will clarify this for you if you don’t know what a reference class is.

            Even if you do, I’m sure others don’t.

            The reference class is the class that you’re using to calculate a probability. Do we treat you as a commenter on James’s blog, a commenter on James’s blog who’s a mythicist, a human being, or an object with a mass of between 10 and 400 kg? There are countless possibilities, as you have lots of properties, and there’s endless ways of slicing and dicing the properties to form reference classes.

            So if we’re trying to work out what the probability is of, say, you voting conservative in the next election, the different reference classes give different probabilities. I imagine most of James’s regular commenters don’t vote conservative, but the proportion will be a lot higher than that of all objects between 10 and 400kg, as most of those objects aren’t eligible to vote in any elections.

            That’s a bit of a silly example, a more concrete case might be served by Carrier choosing to treat Jesus as a member of the reference class of high Rank-Raglan score figures. Several regulars here (Mark especially likes this point) have argued that he’s more appropriately treated as a messianic claimant. That’s an argument over reference classes, and they yield markedly different probabilities of their members being historical.

            (But maybe we should be treating him as a member of the reference class of messianic figures said to have lived in the first century. Or of figures that appear in letters written in greek and have supernatural powers attributed to them. Or of cult founders. Or ancient cult founders. Or post-Augustus but pre-Constantine cult founders.)

          • jekylldoc

            ncovington89 –
            “There are lots of times when we know, intuitively and/or logically, that
            one thing is more probable than another. Sometimes we don’t have hard
            data to know exactly how much more probable. In which case we are stuck
            with making a guess”
            Although Bayes’ Theorem gets used that way, including in my own field of economics, it is not really what it is useful for. The explanation in Kahneman’s “Thinking, Fast and Slow” is much better – drawing inferences about particulars when we have a small sample, it is helpful to use the population probabilities as priors. If you don’t know the population probabilities, then you should not be using Bayes’ Theorem.
            Using subjectively assigned probabilities as priors is GIGO. Carrier is only using the Theorem in that way, and doesn’t recognize the problem. Furthermore, his subjective assignments of prior probabilities are so bad (abusing probability theory) as to be almost arbitrary, and do not add precision or useful structure except in a few cases of comparison with apologetics. Thus he is unpersuasive.

          • Kris Rhodes

            //Using subjectively assigned probabilities as priors is GIGO. Carrier is only using the Theorem in that way, and doesn’t recognize the problem.//

            He explicitly acknowledges the problem, more than once, and discusses its implications for his views and for the dialogue as a whole.

            Edited to add: Half, maybe more than half, of the point is to use the theorem to figure out what one’s assumptions actually imply. A problem Carrier sees in others’ work is that they don’t understand what their assumptions actually imply. Another half (maybe less than half) of the point is to avoid wishy-washiness by actually putting a value on your assumptions so that you can respond to direct criticisms of them as necessary. And of course there are other halves besides these two 😉 I’m just mentioning these as major aspects of what’s going on.

          • jekylldoc

            Kris –

            You may be right. As I say, I have only read about 1/3, and some of that was skimmed. But his atrocious assignment of prior probabilities for different historical versions (1/10 if there are 10 of them, 1/6 if there are 6, etc) came off without a single acknowledgement of the problems with making such an assumption, and indeed he based a rather strong conclusion on that silliness. If he understands the problem of subjective priors, he showed no evidence of it there, or in the other passages on BT that I read.

            Incidentally, I like his view that one can spell out the implications of ones assumptions using probability calculations. But Carrier seems to thing Bayes’ Theorem is some sort of magic key for getting it right, then uses it as a smokescreen to hide his tendentious jumping to conclusions.

          • Kris Rhodes

            I think Carrier has engaged in some really, really counter-productive rhetoric, that’s for sure. I recognize what leads him to do so–he has no real hope of being taken seriously, not really, so why hold back?–but from a “posterity” viewpoint, assuming he’s right, I think he might have had a better chance at getting, you know, citations and stuff, twenty years down the line if he’d taken a more sober approach.

            But reading through the rhetoric, to the logic, I think he wins pretty handily.

          • Neko

            You wrote:

            I recognize what leads [Carrier] to [engage in counterproductive rhetoric] so–he has no real hope of being taken seriously, not really, so why hold back?–

            Lame excuse for acting like a colossal jerk.

          • arcseconds

            I haven’t read Carrier’s book, but every concrete example of the logic he is employing looks at best dubious, and at worst complete hokum to me.

            That tends to make me think it would be a waste of time reading his book, of course.

          • ncovington89

            But Carrier doesn’t use *only* subjective probabilities.

            “Furthermore, his subjective assignments of prior probabilities are so bad”
            Got any examples?

          • jekylldoc

            ncovington –
            As I wrote to Kris Rhodes just below this,
            “his atrocious assignment of prior probabilities for different
            historical versions (1/10 if there are 10 of them, 1/6 if there are 6,
            etc) came off without a single acknowledgement of the problems with
            making such an assumption, and indeed he based a rather strong
            conclusion on that silliness”

            The “theoretical” discussion of this is in “Hypothesis Formation and Prior Probability” (at loc 1220 of 46662 in my Kindle version) where he asserts “if there is no evidence for or against a claim” independently of the oddity you are inventing it to explain, he clarifies just before this, “then its probability is 50%”. Remember that he is using prior in the sense of population or “true” probabilities, which means it gets a lot of weight unless you have quite a bit of specific data to weigh against it.

            He is much better when he discusses an actual basis – if we know Christians interpreted the OT allegorically 90% of the time in a given context, then the prior probability for a given passage in that context is 90%. The problem is that Bayes’ Theorem treats the two cases the same (nothing known, and a population rate known), but they are not at all equal.

            But at the beginning of that section he argues that if scholars of equal competence defend from the same evidence 10 different Jesuses, the initial odds that any one of them is correct can be no better than 1 in 10. What he has done is pull out one piece of evidence (a scholar) and treat it as “priors”, which has the status of population probabilities. No mention of degree of overlap between the different versions, or degree of agreement on which evidence should be given greater weight. No mention of any of the myriads of factors which go into the judgment of scholars. This is sloppy to the point of travesty – he is assigning a prior probability in a respected probabilistic procedure to an arbitrarily selected bit of evidence.

            What he does with that is well-intended – to make an argument for including only the intersection of the many sets as a starting hypothesis. But the insertion of the math is so arbitrary that he undermines his own credibility by using the math at all.

            That is the general case with Carrier’s use of BT. While his intentions are, I would say, honest, he has no discussion of the complexities that he sweeps under the rug with it, and plunges ahead as if his effort to be conservative was enough to make his procedure appropriate. It isn’t.

          • arcseconds

            I had been wondering whether Carrier was reasoning like this on the basis of one or two other remarks that people have made.

            Assigning probabilities equally between a number of cases when you have nothing else to go on is called the ‘principle of indifference’ or the ‘principle of insufficient reason’, and it is known to be problematic.

            There’s a standard case to illustrate this involving probabilities of box sizes, which I can go into if anyone’s interested.

            There are of course involved arguments about how to determine priors in this case, and if Carrier thinks he can just baldly say ‘two case? 50%!’ without even naming the principle let alone admitting that it’s disputable, then he’s either not competent enough to be writing the book, or he is being deeply irresponsible.

          • jekylldoc

            arcseconds –

            I don’t dispute the principle of insufficient reason, although I can well believe someone has shown that it does not take much to abuse it. My contention is that, despite some effort to phrase the explanation in a way that would make the assumption reasonable, he then proceeds to apply it in a way that does not. If it was indeed all we had to go on, I would not have been so put off. But of course that is not even close to being true.

            We have, for example, much more evidence that Jesus was a healer and exorcist than we have that he was a military leader. So why would anyone start out treating the two as equally probable?

            Once again, to re-state an important point, prior probabilities represent what you have to infer with before incorporating data observations. That incorporation is “updating”, for which statistical reasoning can be used. But “what you have to infer with” is supposed to be P(B¦A), and P(B) and P(A). B can be thought of as the data you are using to update your beliefs, A as the implication you want to assess, and you are looking for P(A|B), for example, the probability that “A:Jesus really lived” given “B:we observe the earliest source, the epistles, saying next to nothing about his historical life.”

            Most of the time Carrier uses this with at least an effort to be correct. But given that he has no P(A) to work with, the idea that we are updating something with our information must be treated very delicately, or, better still, set aside to avoid the false pretense of rigor.

          • ncovington89

            “But at the beginning of that section he argues that if scholars of equal competence defend from the same evidence 10 different Jesuses, the initial odds that any one of them is correct can be no better than 1 in 10.”
            Well I wouldn’t treat this as a rigorous Bayesian argument that he is making here. He’s really just making the commonsense point that everybody can’t be right at once, so therefore we have to take a more careful look at the evidence and develop a more rigorous way of dealing with that evidence. Of course, one of the portraits of Jesus might hold up to such a rigorous treatment, and we can’t rule that out a priori. However, to even get to the point of working out a portrait or life story for Jesus we have to first establish that he existed.

          • jekylldoc

            Right. It is the faux rigor that bothers me. He could have made a sensible argument without the rigmarole, does in fact make a sensible argument with the rigmarole, but managed to put me off in the process.

          • Kris Rhodes

            I have to admit I have come more and more intensely to wish Carrier had never breathed a word about Bayes.

            But having said that, I still don’t get why people insist on the GIGO objection. You can give a parallel GIGO objection to Modus Ponens but that, of course, would be silly. The deal with modus ponens is that your conclusion is only as good as your premises. No one, Carrier included, would say otherwise about Bayes. But just as this doesn’t mean there’s anything wrong with using modus ponens using your own subjective impressions of which propositions are true and which false, I also don’t see what would be wrong with using Bayes on your own subjective impressions of what is probable and what is not probable.

          • arcseconds

            I don’t think anyone’s saying Bayes’s theorem is bogus. McGrath grumps about it from time to time, but even he acknowledges that there might be some value in taking a Bayesian approach.

            I think it might be useful to bring out certain assumptions.

            Plus training in formal logic does (I believe) help with informal reasoning, even though no-one (so far as I know) actually tries to formalize real-world arguments, and I think training in probability also helps.

            But it’s actually very difficult to formalize reasoning about real-world problems (Ian’s already alluded to some of the difficulties). Philosophers of science have attempted to cast famous scientific discoveries in Bayesian terms, and it’s a difficult and contentious issue. And these are cases where the data and actual history are usually well-known!

            One large complicating factor is that any real-world problem is going to have lots of pieces of evidence. You can model this by repeated application of Bayes’s Theorem, but then errors accumulate, and as Ian’s previous review demonstrates, Bayes’s theorem is quite sensitive to small differences at low probabilities.

            Also, Bayes’s Theorem has us update a single probability, that of the hypothesis, on the basis of a single piece of evidence. But actually of course in general hypotheses are not free-floating. Let’s say we’re somewhat impressed by ncovington89’s argument that there’s textual evidence that some Christians thought Jesus existed about 100 years before he is normally thought to have existed, and that mythicism can explain this better than historicity. We can plug that into Bayes’s Theorem… but if we increase the probability of mythicism, we also need to decrease the probability of the phrase ‘the brother of the lord’ referring to an ordinary sibling relationship, which wasn’t even at issue in nconvington89’s account.

            So what are we going to do then? Run some sort of iterative process over every single belief updating them all every time a new piece of evidence comes in? Seems we’re back to multiplying errors in this case…

            Basically, real-world arguments are super-complicated. You can try to model the whole thing formally, which is very difficult (and if you can really do justice to an actual important piece of reasoning you should be able to publish it in a top journal). Or I suppose you take a small number of considerations (maybe just one) and formalize that… but then it’s unclear what the merit of doing this is, because you’ve left so much unformalized.

          • jekylldoc

            Kris, I am sorry if I have overstated the matter. Carrier’s use of Bayes’ Theorem is not totally off base, or irresponsible. He does in spots go out of his way to acknowledge the problems of subjective assumptions, and I think he does make an effort to estimate conservatively, at least by his own lights.

            If the Bayesian pretentiousness was the main problem I had with his work, I would not bother to criticise it (except to clarify for people who are scratching their heads and asking what is this Bayes thing). I have serious problems with two other crucial arguments he makes, the interpretation of the Testament of Isaiah ( or whatever the name is – I will edit tomorrow but am in bed already) and the interpretation of the “born of a woman” passage. It is not so much his conclusion that gets on my nerves, it is his “there is no other way to read this” approach to things that obviously do have other ways to read them.

            And for me his BT material is cut from the same cloth. Sure he gives lip service to possible problems, but it is clearly beyond him to actually look at the details of the matter through eyes that are not already convinced. He has seen his gestalt, and as a result he seems incapable of paying attention to matters that are problematic for it. One word for that is “committed” but another is “tendentious”.

          • Geoff B

            I would call it fuzzy math, not hyper-math.

          • Pseudonym


            Mathematicians would beg to differ. Not to mention statisticians.

          • Geoff B

            Yes, because as you say, Carrier proposes a new methodology for weighing the probabilities of particular hypotheses. Most mainstream historical scholars have no training in Bayesian analysis and not that much in statistics. So you are correct that “mainstream” historians would view Carrier’s methods as idiosyncratic.

            None of that takes away from the fact that NT scholars have build end arounds to the historical method to get around the fact that the evidence does not support the level of certainty they claim.

          • ncovington89

            That being said, there are qualified philosophers who have some background in history who think Bayes’ theorem is a good method for history, or at least for some historical claims, like Timothy McGrew, Robert Greg Cavin, William Lane Craig.

  • I am by no means a scholar. I bought Carrier’s previous work on Jesus Mythicism because it supposedly presented evidence of a pre-existing belief in a celestial Jesus. The evidence was incredibly scant. In order for me to buy and read Carrier’s present work, someone trustworthy would need to tell me that he had presented much stronger evidence for the pre-existing belief in a celestial Jesus.

    • Geoff B

      Carrier did not present his thesis of “pre-existing belief in a celestial Jesus” until OHJ. This comment appears to be disingenuous.

      • ncovington89

        I think Carrier did discuss it in Not the Impossible Faith, but that book was not “about Jesus mythicism” like Bilbo said, and it also very wrong to say that the evidence for a pre-Christian Jesus is “scant.” On my blog, “Hume’s Apprentice” the posts “On the Historicity, Part 12” and “The Pre-Christian Jesus in Philo” document even more than Carrier did.

        • Geoff B

          Ok, read that. Very informative. So I guess the historicist explanation is that either Jesus modeled himself as an earthly incarnation of the pre-Christian Jesus or his followers made a fictive link to Jesus, modeling his legacy after the ideas such as those found in the Wisdom of Solomon. Could be. But, then, how do they determine what was true history and what was made up? Jesus actually crucified was interpreted as the shameful death of the righteous man. Again, could be. Which of Jesus’ supposedly illiterate followers recast Jesus as the Jesus. Was it coincidence that his name was Jesus (and yes, I know, it was a common name).

          • Incarnation of the pre-Christian Joshua/Jesus? What is that even supposed to mean? Are you suggesting that the generic depiction of a righteous man in Wisdom of Solomon was turned into a narrative about a crucified and thus disqualified Davidic anointed one? To what end, and by whom?

          • ncovington89

            He is saying that a believer in the historical existence of Jesus might explain the similarity between Philo’s Jesus and the Gospel Jesus in the following ways: Maybe a historical Jesus taught that he was the archangel Jesus of Jewish lore, or maybe Christian story tellers mixed the real memories of a historical Jewish with previous myths about an angelic Jesus. I think it’s plausible, but it may not be *the* best explanation on the table.

          • By “Philo’s Jesus,” you mean Joshua the high priest from the time of Zechariah, to whom he gives an allegorical treatment of his characteristic sort? I suppose living during the Babylonian exile is “pre-existence” in relation to the time of Paul, but that figure does not fit what Paul says about his own Joshua being someone of Davidic descent. I suppose one could posit some sort of reincarnation scenario, but that would not really provide a means of sidestepping the historicity of two Joshuas, never mind one of them.

          • Paul E.

            I have seen this argument alluded to before, and I suppose I should read Carrier’s book (tl;don’t have time), but is this the idea that Philo’s Logos was somehow “named” Jesus? If so, it seems, to put it charitably, highly problematic.

          • Yes, if one is not inclined to be charitable, one would call it an attempt to dupe people who don’t bother fact-checking the claim into believing that Philo says something he doesn’t.

          • Geoff B

            Here is a great opportunity to demonstrate your intellectual honesty and scholarly credibility:

            Quote Carrier making this claim, related to Philo and Zechariah 6:12, so that we can see that, yes, indeed, he is duping people.

          • Geoff B

            No, James, I asked you to prove your worth with a direct quote, not one of your blogs characterizing Carrier as saying something. Just a quote, should be easy, right?

            Can you do it or not?

            Does Philo or does Philo not refer to a character mentioned in Zechariah whose name is Jesus?

          • That wasn’t the topic – it is whether Carrier claims that Philo said the Logos is called “Jesus.” I gave you a link to a video in which Carrier speaks about this very topic, and a blog post in which I respond to and comment on that video. I have no intention of subjecting myself to his drivel again in order to spare you the inconvenience of doing so. I have better things to do with my time than that, particularly tonight.

          • Geoff B

            What does Carrier say in the video? What is the direct quote? You made the claim, I am challenging you to put forth the evidence. That burden is on you, not me to go watch some video.

          • I will provide the reference to the video – you can see the claim on the screen at the 16:20 mark. But I am doing that mostly for the benefit of others who may be interested. As for you, you are being banned. I have been addressing mythicism for many years, and it is thoroughly inappropriate to ask, never mind demand, that I revisit topics that I have covered in the past, when you could easily look up that earlier treatment – especially when I have made it so easy and provided you the link. In the same way, more generally, demanding that I look things up for you that you could easily look up yourself – especially at a time when I have indicated publicly on this blog that I am getting ready to go to a conference to give a paper – is in no way acceptable behavior.

          • Grog

            Ouch. That’s not very hospitable, is it? Well, what I see Carrier saying in the video and what you won’t quote is that Philo refers to a “celestial being named Jesus.”

            Is it not true that Philo refers to a celestial being named Jesus? What is the name of the celestial being Philo refers to? [If you follow the reference, you will see that it is Joshua, right? So how is Carrier duping anyone?

            And is this anywhere near as misleading as Ehrman’s claim that “we have” Aramaic sources written with a few years of the crucifixion in the Huffington Post?

            You can do better than this, James. Banning people asking you to put up or shut up? Come now. Get a grip. Clearly you lost that round.

            You know the only other forum that has ever banned me? Stormfront.org. You’re keeping some grand company there, James.

          • Philo does not refer to a celestial being named Jesus or Joshua. He uses a story about Joshua, just as he uses so many other stories, to talk about the Logos.

            I know that internet trolls prefer to treat anyone who bans a troll as somehow the loser in an interaction. Personally I think that if I allowed such individuals to waste my time so that I was less well prepared for a conference I would be attending, or allowed this kind of stuff to distract from getting real scholarly work done, then I would be the loser. Different people clearly have different perspectives on what really matters.

            Internet trolls, in my view, also automatically lose when they create a new fake account to comment after being banned…

          • Jim

            I disagree with you a bit on the second part. From your YCAS2015 pic (of you, Goodacre and Ehrman), isn’t that like three scholars versus one troll? Now how can anyone expect a troll to win under those stacked circumstances … other than maybe in their own mind?

            Secondly, how do we know that you didn’t write your posts at a Jays game instead of the lecture hall? 🙂

          • John MacDonald

            What would motivate a full professor in theological research to be anything less than intellectually honest on his own religion-blog?

          • Geoff B

            What Carrier says is that Philo refers to a character in Zechariah whose name is Jesus:

            “I have also heard of one of the companions of Moses having uttered such a speech as this: “Behold, a man whose name is the East!”{18}{#zec 6:12.} A very novel appellation indeed, if you consider it as spoken of a man who is compounded of body and soul; but if you look upon it as applied to that incorporeal being who in no respect differs from the divine image, you will then agree that the name of the east has been given to him with great felicity. (63) For the Father of the universe has caused him to spring up as the eldest son, whom, in another passage, he calls the firstborn; and he who is thus born, imitating the ways of his father, has formed such and such species, looking to his archetypal patterns.”

            Zechariah 6:11-12:Take the silver and gold and make a crown, and set it on the head of the high priest, Joshua son of Jozadak.[d] 12 Tell him this is what the Lord Almighty says: ‘Here is the man whose name is the Branch, and he will branch out from his place and build the temple of the Lord.13 It is he who will build the temple of the Lord, and he will be clothed with majesty and will sit and rule on his throne. And he[e] will be a priest on his throne. And there will be harmony between the two.’

            Carrier observes that Philo referred to a figure named “Joshua.” This is a correct observation, not “highly problematic.”

            The counterargument is that perhaps the name Joshua here is just a coincidence, being a common name and that Philo had no intention of linking the Logos to the name Joshua.

            That could be. But that wouldn’t stop a well-educated mystic/preacher or community of cultists from reading Philo and themselves making the connection. Philo says the figure in Zechariah, whose name is Joshua, is the Logos. It does not seem too farfetched for “Joshua Savior” to become the name of the Jesus sect.

          • Saying “it is plausible that this person named Josh was invented, simply because he was not the first person named Josh” isn’t especially persuasive. Indeed, why not just say that Joshua the high priest was himself invented based on an earlier Joshua?

          • Geoff B

            1. No one is saying that.

            2. Why not, indeed. In fact, we could just go back to invented Joshuas based on invented Joshuas. Does this strengthen the case? Was the first Joshua a real historical figure?

            Didn’t you complain about Godfrey using quotes around a paraphrase of your argument?

          • ncovington89

            Set aside the passage about Joshua. Philo definitely believed in a logos who was God’s firstborn, the second Adam, and many other things that Paul also says about Jesus. They had in mind the same thing. So either Paul believed that the logos flew down into the body of a human baby or he believed that in some way he could speak of the logos as a Son of David (as Philo seems to, I think there is a passage in which Philo calls the Logos the Shoot, a messianic title).

          • I think you are getting things backwards, and wrong, on several points. Philo takes texts about the first Adam (not the last, which is what Paul talks about), Joshua the high priest, the messianic branch, and many other people and topics, and uses them to talk about God, the Logos, and mystical realities. That is a far cry from these things being ‘names’ or ‘titles’ of the Logos in the sense that one would normally use that term.

            If one assumes that Paul wrote Colossians, as you see, to but roughly half or more NT scholars do not, why would you view Paul’s language as more like Philo’s and less like the application of Wisdom language to Torah in Ben Sira and Baruch, and what do you think the significance is of relating Paul more to Philo than to those parallels?

            You seem to be excluding a wide array of interpretative options, as mythicists are prone to do, jumping immediately to one which is a variant of later Christian incarnational theology, which mythicists join conservative Christians in reading into early Christian sources uncritically.

          • ncovington89

            “Philo takes texts about the first Adam (not the last, which is what Paul talks about),”
            Philo talks about there being two Adams, one on Earth and the other in heaven, and in corroboration the language used to describe the heavenly Adam is a dead-ringer for many things Paul says about Jesus.
            “Joshua the high priest, the messianic branch, and many other people and topics, and uses them to talk about God, the Logos, and mystical realities. That is a far cry from these things being ‘names’ or ‘titles’ of the Logos in the sense that one would normally use that term.”
            I’m not sure I understand what you mean here. Philo is clear that ‘the branch’ was indeed something meant to be applied to the heavenly man. Seems fair to me to say that’s a title.

            “If one assumes that Paul wrote Colossians,”
            Who assumes that? Not this guy. Where are you getting this from?
            “why would you view Paul’s language as more like Philo’s and less like the application of Wisdom language to Torah in Ben Sira and Baruch, and what do you think the significance is of relating Paul more to Philo than to those parallels?”
            I have not studied Baruch or Ben Sira, so I can’t speak to whether they influenced Paul or not. And concerning Philo I don’t know that Paul ever read him, I just think the abundance of parallels suggest that there was a ‘common tradition’ behind both, and that tradition (or some variation of it) may be in other sources like Ben Sira.

            “You seem to be excluding a wide array of interpretative options, as mythicists are prone to do, jumping immediately to one which is a variant of later Christian incarnational theology, which mythicists join conservative Christians in reading into early Christian sources uncritically.”
            Do you mean high Christology? We’ve had this discussion before and I just don’t how we can understand certain passages without it. And it isn’t just mythicists and conservative Christians who believe high Christology is the early Christology, which something I know you’re aware of.

          • Geoff B

            It really does not appear that James has read Carrier’s book. Carrier goes into great detail on this point:

            “Both [Philo and Paul] also call the earthly Adam the ‘first man’. Paul then calls Jesus the ‘last Adam’, but describes him in terms identical to Philo’s ‘second’ man (who in order of creation is actually first).”

            Carrier quotes Philo’s distinction between the first man and the second man:

            “‘the races of men are twofold: for one is the heavenly man, and the other the earthly man’ and ”the heavenly man, as being born in the image of God, has no participation in any corruptible or earthlike essence.'” [Philo, Allegorical Interpretations 1.31]

            Paul, 1 Cor 15:40

            There are also heavenly bodies and there are earthly bodies; but the splendor of the heavenly bodies is one kind, and the splendor of the earthly bodies is another.

            1 Cor 15:45

            So it is written: “The first man Adam became a living being”[f]; the last Adam, a life-giving spirit.46 The spiritual did not come first, but the natural, and after that the spiritual.47 The first man was of the dust of the earth; the second man is of heaven.48 As was the earthly man, so are those who are of the earth; and as is the heavenly man, so also are those who are of heaven. 49 And just as we have borne the image of the earthly man, so shall we[g] bear the image of the heavenly man.

            I think your fans would do well to read what Carrier actually says, rather than take your word for it.

          • Better still, they could read any of the many other works of scholarship that has examined Paul’s comparisons of Adam and Christ, which discuss not only Philo but also Rabbinic and Gnostic interest in Adam Kadmon, and discuss not only what is similar but also what is different between them.

            I do not take kindly to people claiming that I have not read a book about which I have taken the time to write three articles as well as a number of blog posts.

          • ncovington89

            Of course, I lean towards mythicism, so I suspect this means trouble for historicism. Historicists could explain it the way you have, but… When we read about an unhistorical demigod and later see the same character inserted into a historical narrative, we would usually take that at face value as a myth later used to construct a fictitious history (like King Arthur). It’s just more suited to Ockham’s razor.

      • The book I bought, because I was told that Carrier gave evidence for a pre-existent belief in a celestial Jesus in it, was Not the Impossible FaithM. I forget the exact page it was on, but it wasn’t very impressive. However, if Carrier has presented stronger evidence for his claim somewhere else, I would be curious to know what it is.

        • Kris Rhodes

          There’s this book On the Historicity of Jesus which he published several years later and in which he gives his case for that idea. It’s stuff from that book that most of us are talking about in threads like this.

  • arcseconds

    While I’d certainly be interested in the reaction of a classicist or a historian of Roman or Jewish history to Carrier’s books, I don’t think it would be the most amazingly useful thing ever, so as the commenter in question I feel like qualifying ‘how useful it would be’.

    The phrase implies ‘very useful’, and I’m not sure I’d go that far. I’m prepared to go as far as ‘somewhat useful’ 🙂

    As I think it’s rather unlikely they’ll agree with Carrier’s assessment of the matter, the usefulness would mostly be rhetorical.

    On reflection, I actually think it would be more useful to get such a person (or a few such persons) to weigh in the debate more generally, and confirm (if in fact it is true, of course, but I believe it is) that NT studies is not doing history in some radically different way to people studying related areas, and their reaction to the argument for the historical Jesus.

    • I think that a look at various points of overlap could tell you that now. How many historians also participate in some way in Biblical studies? How many Biblical studies people are housed in a department of Classics? How many have contributed to the same edited volume or presented at the same conference? It would be a little work to assemble the data, but I know that we intersect often enough for there to be at least some record. I remember a Classics professor giving a presentation when I was at Durham on the proposal that Jesus was a Cynic. Interestingly, the only two sayings that seemed to him to fit that scenario were ones that I would place against a Jewish background. There are also permeable lines between us and Jewish history, the study of Gnosticism, and the study of Late Antiquity, with conferences and symposia where it would have been very noticeable if we were all using radically different methods.

      • arcseconds

        It’s hardly surprising of course, and indeed what I believed to be the case. Of course biblical studies papers cite Roman history sources from time to time, and one would expect it to happen the other way around, too. And of course academics in similar areas are going to pay attention to one another at least to some extent. It would be not very surprising to find families with one person in biblical studies and one in a related area who proof-read each other’s papers!

        But evidence of cross-citation and ‘co-habitation’ I don’t think will have the same impact on mythicists and the mythicist-curious as clear statements from such people that they don’t see a problem here. The cross-citation data still requires an extra step: “surely if there were a problem, they would notice and say something!” which such people would simply not take, and even bystanders would notice some weakness here.

        Of course die-hard mythicists will attack anyone who supports historicism, and of course it’s always open for them to say “all of history is bunk!”. But rhetorically biblical studies is an easy target, because of the insinuation of Christian bias and it’s easy for people to imagine it’s a small and insular area. It’s a lot less palatable to try to suppose that this is the case for the entire academic area of ancient history. Some will no doubt take this step too, but it starts to look a lot more ridiculous.

  • ncovington89

    “And so what, if anything, would convince you to take on this unpleasant, unappealing, and unrewarding task?”
    This *assumes* that the historian or classicist in question would disagree with Carrier. But how could you possibly know this would be the outcome in advance? James, with all due respect I think this one of many illustrations that you view mythicism with pure prejudice. Every idea deserves a fair hearing, at least.

    • I obviously cannot know what any particular Classicist or historian would say. But it is a simple Bayesian calculation to figure out the probability that most will find it pure hokum and wonder why they allowed a blog post to persuade them to waste their time on it…

      • Andrew Schefe

        Please show your working

      • ncovington89

        Well I don’t think we should work from prior probability alone. We also have to take consequent probability into account, and the only way to get a consequent probability is to see what the expert says.

      • jekylldoc

        James –

        I don’t think you should be so dismissive of Carrier. While he is tendentious and sloppy, and I would say (disclosure, I have read only the first third of OHJ) wrong, the issues he raises are not non-issues.

        Scholars have to wrestle with the Pauline silence on Jesus’ life all the time. There is no definitive answer on the subject. Jesus was clearly obscure, if not mytholoical. There clearly were strands of mythological thinking which fed into early Christianity. Euhemerization did occur at some places in the Hellenic world. Mark and Luke/Acts probably borrowed from literary sources to give juicy details to not-very-historical stories.

        It is unlikely that some smoking gun is going to be found which definitively shows either that all of the above add up to mythicism or that all of the above add up to historicism, (with syncretism at the earliest stratum). Why not just treat it as another problem to be worked over?

        There are lots of interesting questions involved in assessing whether the specific parts of the argument imply what advocates claim they imply.

        In my view what is missing is not a definitive assessment by some properly qualified scholar, but just a sufficiently balanced weighing of each of the different parts of the mythicist argument.

        • Paul is at least as silent about the details of a supposed Jesus myth as about the details of the life of the man Jesus, and so it is not more of an issue for one than the other, and in either case is explicable in terms of the genre of epistles, coupled with Paul’s lack of direct witness of at least most is not all of what Jesus said and did.

          I do not think that writing review articles engaging Carrier’s claims is being dismissive.

          • jekylldoc

            Indeed, I agree, and I agree with most of what I have seen you post on the subject. A few comments do come across as dismissive (“why should anyone take this seriously?” type remarks).

            I don’t think the mythicists have zero of interest to say. Some of the things I have learned from them about literary borrowings may help to illuminate the reason for the Marcan secret, for example. If there were widespread discussions of a defeated Messiah before Jesus, that influences both how we see Jesus and how we see the early church.

            All grist for the (academic) mill, is my point.

  • You wrote, “…what, if anything, would convince you to take on this unpleasant, unappealing, and unrewarding task?”

    Well, I’m not a professional historian. But as a literature teacher and avid reader of textural criticism, etc., I don’t think I would tackle the arduous task for most anything.

    I’ve read a lot of the pros and cons regarding the hypothesis, but my effort to get through a mythicist book on why Jesus never existed was worse than hiking through a midwestern blizzard;-)

    I personally still don’t see why bright educated thinkers have adopted this view. Their reasoning seems weak, their proposals are very convoluted (that Jesus was a fictional character created living up in the heavens based on Hellenism, etc.), they don’t seem to have a good grasp of textural criticism, and worst of all they make basic errors regarding the NT text itself!

    • Geoff B

      What do you mean “Jesus was a fictional character created living up in the heavens?”

      I am not sure you understand the strongest mythicist hypothesis. Also, you don’t seem to be aware that mythicism is about exploring the uncertainty that exists in the Jesus historicist position.

      Do you have any examples of how “they” don’t seem to have a good grasp of textual criticism (examples from Carrier, Price, or Brodie would be must valuable here) or “basic errors?”

      • See James’ “Did Jesus Die in Outer Space? Evaluating a Key Claim in Richard Carrier’s The Historicity of Jesus.

        Also see strangenotions.com, An Atheist Historian Examines the Evidence for Jesus (Part 1 and Part 2)
        (quoted section below)

        As I wrote above I am mostly an onlooker, not a professional historian, not an expert in antiquity. But I was a literature teacher for many years, have read a lot of books by scholars on textural criticism (including most of The Historical Jesus: The Life of a Mediterranean Jewish Peasant by John Dominic Crossan).

        In my judgment, Crossan is too much of a historical minimalist, and I don’t agree with his Christian worldview, but I greatly respect him as an historian. Last year I went to a series of lectures by him. Brilliant! (Much less dry than his books).

        Maybe I am expecting too much from mythicists but after reading great well written tomes by Bart Ehrman (8 so far) and in biology, Richard Dawkins’ The Ancestor’s Tale, I do expect serious books to be well written.

        Someone back in 2012 suggested I read The Jesus Puzzle by Earl Doherty; it was a real letdown.

        Doherty is smart, knows a lot of things, but his understanding of ancient Hellenism and Second Temple Judaism seems weak and he made textural errors claiming that the NT didn’t say what it does say.

        I even got out my marked up copy of the Jesus Puzzle and leafed through it, trying to decide whether or not to take the time to type out all the weak stuff for you.

        But you know what, I decided to save time and just quote some of the same points as mine by O’Neill for you. Go to his site for more. He’s an historian; I’m not, but he makes similar points.

        I also recommend, How Jesus Became God, which I finished only a few months ago. Brilliant writing, well thought out, plenty of evidence. But all of Ehrman’s books are very good, even the ones with which I strongly disagree.

        I’ve also read some writing by Price including one of his more popular books, but not his book on mythicism. He is a brilliant scholar, but the critical reviews I read led me not to read that book by him because it sounds like he only covers ground I’ve already been over many times.

        I haven’t Carrier, except articles on the Internet, not read Brodie.

        Maybe I should, but right now I’m in the middle of Reconstruction by historian Eric Foner (another brilliant thinker), etc.

        I do think I understand the “strongest mythicist hypothesis.

        Tim O’Neill agrees with my own evaluation of Doherty, so he must be correct;-)

        from O’Neill’s site:

        “Jesus was a celestial being who existed in a realm just below the lunar sphere and was not considered an earthly being at all until later.”

        This is the theory presented by another self-published Mythicist author, Earl Doherty, first in The Jesus Puzzle (2005) and then in Jesus: Neither God nor Man (2009). Doherty’s theory has several main flaws. Firstly, he claims that this mythic/celestial Jesus was based on a Middle Platonic view of the cosmos that held that there was a “fleshly sub-lunar realm” in the heavens where gods and celestial beings lived and acted out mythic events. This is the realm, Doherty claims, in which it was believed that Mithras slew the cosmic bull, where Attis lived and died and where Jesus was crucified and rose again. The problem here is Doherty does very little to back up this claim and, while non-specialist readers may not realise this from the way he presents this idea, it is not something accepted by historians of ancient thought but actually a hypothesis developed entirely by Doherty himself. He makes it seem like this idea is common knowledge amongst specialists in Middle Platonic philosophy, while never quite spelling out that it’s something he’s made up. The atheist Biblical scholar Jeffrey Gibson has concluded:

        “… the plausibility of D[oherty]’s hypothesis depends on not having good knowledge of ancient philosophy, specifically Middle Platonism. Indeed, it becomes less and less plausible the more one knows of ancient philosophy and, especially, Middle Platonism.”

        Secondly, Doherty’s thesis requires the earliest Christian writings about Jesus, the letters of Paul, to be about this “celestial/mythic Jesus” and not a historical, earthly one. Except, as has been pointed out above, Paul’s letters do contain a great many references to an earthly Jesus that don’t fit with Doherty’s hypothesis at all. Doherty has devoted a vast number of words in both his books explaining ways that these references can be read so that his thesis does not collapse, but these are contrived and in places quite fanciful.

        Finally, Doherty’s explanations as to how this “celestial/mythic Jesus” sect gave rise to a “historical/earthly Jesus” sect and then promptly disappeared without trace strain credulity. Despite being the original form of Christianity and despite surviving, according to Doherty, well into the second century, this celestial Jesus sect vanished without leaving any evidence of its existence behind and was undreamt of until Doherty came along and deduced that it had once existed. This is very difficult to believe. Early Christianity was a diverse, divided, and quarrelsome faith, with a wide variety of sub-sects, offshoots, and “heresies”, all arguing with each other and battling for supremacy. What eventually emerged from this riot of Christianities was a form of “orthodoxy” that had all the elements of Christianity today: the Trinity, Jesus as the divine incarnate, a physical resurrection etc. But we know of many of the other rivals to this orthodoxy largely thanks to orthodox writings attacking them and refuting their claims and doctrines. Doherty expects us to believe that despite all this apologetic literature condemning and refuting a wide range of “heresies” there is not one that bothers to even mention this original Christianity that taught Jesus was never on earth at all.”
        An Atheist Historian Examines the Evidence for Jesus (Part 1 of 2) by Tim O’Neill

        Now, I need to stop and wash the dishes for my wife, very intellectual exercise:-)

        • Geoff B

          I stopped reading this at “O’Neill.” Sorry, but he has zero credibility to me.

          • Scott Scheule

            Mind saying why?

          • Geoff B

            My own personal experience with Tim is that he isn’t intellectually honest. I will leave it at that.

          • Scott Scheule

            Ok. I admit Tim sometimes can’t admit he’s wrong about something, but that applies to most people I know. Carrier certainly included. See his interactions with Thom Stark or Luke Barnes or Tim. At any rate, lack of intellectual honesty aside, Tim seems to have a damned good handle on the subject and makes some damned powerful arguments. I have no idea why one would ignore him in toto.

          • Geoff B

            Some maybe and a lot not. Carrier actually was fairly responsive to Thom Stark and admitted that Stark’s argument had caused him to amend his position.

          • Scott Scheule

            We disagree. Carrier admitted very few of the mistakes Stark pointed out, and simply modified his original post. I mean, read the exchange. Stark accuses him multiple times, rightfully, of ignoring what he says. But the interaction with Barnes is much more telling, imo.

            Regardless, both are intellectually dishonest in some sense, but Carrier is the bigger offender so far as I can tell. So I remain puzzled why you’d write O’Neill off.

        • Jan Steen

          I second your recommendation of Tim O’Neill’s articles.

          His hilarious takedown of David Fitzgerald’s atrocious book Nailed: Ten Christian Myths That Show That Jesus Never Existed At All is also well worth reading:


          Follow up after response by Fitzgerald:


  • Scott Scheule

    What would convince one? Intellectual honesty. Intellectual fortitude. That’s right, the kind of honesty and maturity to look past the fact that, yes, Richard Carrier is an unrepentant asshole, an unemployed blogger who acts like a spoiled child when people disgreee with him, who NEVERTHELESS argues eloquently and persuasively in support of his thesis and thus deserves a response to his case for a mythical Christ. We, the Richard Carrier-haters, are tired of our only support being people who complain about rudeness and unprofessionalism. We know that–distasteful as Carrier may be–he makes a rational and prima facie valid argument, and we want the academy to support its consensus powerfully and conclusively. Put Mr. Ziff in his place, but please, do it honestly, with evidence and argument. Don’t lie about his arguments. Don’t distort his arguments. Take them at their best and shred them or admit that this young upstart has embarrassed us all.

    • Kris Rhodes

      Taken literally I like what you said to a degree but I’m not sure if you’re seriously saying you yourself are a “Richard Carrier-hater” and genuinely intend to include yourself in the “us” in the last sentence. My sarcasm meter is failing me.

      • Scott Scheule

        Yeah, fair enough. My view on Carrier is this. He acts like a prick online. He has trouble admitting he’s wrong and he mouths off in fields he has little expertise in. That being said, he can make a good argument in fields he’s studied and he deserves to be answered there. I suppose the answer is this: I am quite disappointed in Richard Carrier in his interactions with Thom Stark and Luke Barnes. But I don’t know anyone who hasn’t made similarly egregious faux-pas, so it would be hypocritical to hold Richard to such an unreachable standard.

  • I read the whole book and I made many (26) comments on Carrier’s OHJ . First get http://historical-jesus.info/blog.html , then use the browser FIND function with {Carrier’s OHJ} .
    Blog posts 91 to 106 & 110 are all about OHJ. Blog posts 16, 17, 19, 20 37, 40, 60, 68 & 70 are in part about OHJ.

    Cordially, Bernard

  • Cecil Bagpuss

    Is anyone going to be taking Carrier’s online course on the New Testament? Apparently, the course will address a number of important issues, including the following:

    How screwed up are the manuscripts of the New Testament?

    • John MacDonald

      I took Carrier’s course about “On The Historicity Of Jesus” last year. I enjoyed it, but I should have asked more questions.

      • Cecil Bagpuss

        John, I’m impressed. You have read the book and taken the course! So you have no excuse for disagreeing with any of Carrier’s arguments 🙂

        • John MacDonald

          lmao !

  • Hail King Jesus(LORD Over All)

    I agree to what Arseconds is saying to Nick. Except I don’t understand why he would say that Blackwell Companion To Natural Theology is not a credible book? Wiley is a very good publisher and is peer-reviewed.

    • Cecil Bagpuss

      We need to be consistent. If we say that Richard Carrier’s use of Bayes’ theorem is designed to lend a spurious air of scientific rigour to what is essentially a fraudulent enterprise, then we should be equally critical when others do something similar.

      We can’t condemn Carrier on the one hand, but endorse Tim McGrew on the other. If someone has an argument to make, then let him state it in plain English, without resorting to some pseudo-mathematical formulation of it.

      Carrier’s argument should be seen for what it is: a piece of epistemological necromancy by which he “proves” that the early Christians believed something that they never actually state. This is the kind of “scholarship” that one might expect from someone who charges people for the privilege of having dinner with him.

      • John MacDonald

        The problem with Carrier’s use of Bayes’ theorem is that at some point the assignment of probabilities becomes arbitrary and falls in line with Carrier’s “gut feeling” that Jesus began as a myth.