Richard Carrier, Rank-Raglan Heroes, and Bayes Theorem

I recently prepared this as a comment on a small part of Richard Carrier's video critique of Bart Ehrman's claims in his debate with Robert Price.  The comment, however, appears to be to large for YouTube to digest, and I would like to have it available for reference, so I am posting it here.  It will be of little interest unless you are interested either in the historical Jesus (or Mythicism with respect to Jesus), or Bayes Theorem.  If you fall into the first camp, but are not familiar with Bayes Theorem, here is a brief introduction (which, however, does make a mistake in assuming the second test [5:12], as described by him, would be independent).




1) Richard Carrier says [19:22] "Here what [Bart Ehrman] has committed is the logical fallacy of affirming the consequent, and it's a well known, very basic fallacy.  He's basically saying, 'Some historical figures have been heavily mythologized, therefore all mythologized figures were historical figures."

In point of fact the fallacy here is that Richard Carrier is raising a ridiculous straw man.  Ehrman is pointing out why he finds the argument from the mythologizing of the story to the mythical status of the person in the story unconvincing.  That is, he is showing why he finds the argument - the story about X has mythological elements, so therefore x was a mythological figure unconvincing.  What is more, he does so in the most time honored method of demonstrating an argument is invalid - by providing examples where the premises are true, but the conclusion is false.  In the core example he provides, Augustus is not a mythological figure, so the conclusion is false, yet his story is told (or was told in the past) with mythological elements, so the premises of the invalid argument that Ehrman rejects are true.  Hence Ehrman's case that the argument is invalid is correct.

Nor is Carrier's misunderstanding very understandable.  Ehrman concludes his argument [18:38] "The fact that stories are molded to a model has no bearing on whether the person actually existed or not."  Carrier parses this as Ehrman saying, "The fact that stories are molded to a mythological model proves the person about whom the story is told to have been historical."  He needs this blatant misrepresentation to portray Ehrman's argument as a fallacy. The fact is, however, that the difference between "has no bearing" and "proves their historicity" could not be any starker, and Carrier's misrepresentation is so gross it is difficult to believe it was accidental.

2) Richard Carrier says [19:50], "What we really want to know is, if we put into a basket... if we collect together ....  If we put into a hat all the heavily mythologized people, and throw them all in there, and I think Ehrman would have to admit that some of them are not historical; and we would also admit, with him, that some of them are historical.  So the question is, when you throw them all into a hat (all of these heavily mythologized people), how many of them are historical. In other words, what is the frequency - how commonly is it the case that when somebody is heavily mythologized, they turn out to be an historical person, rather than an actually mythical person?   And this is the analysis that [Ehrman] is dodging here, because he doesn't understand this is the actual logic we are looking at.  We want to look at the frequency -how common. ... Now when you do this analysis - when you throw all the heavily mythologized people into a hat, it turns out most of them are mythical.  In fact, I come up with the conclusion that at best, one in three are historical."


There are several problems with this.  Firstly, while Richard Carrier frames the argument this way in his books and lectures, most mythicists do not (or did not before Carrier came along).  Ehrman was not responding specifically to Carrier’s argument, and so did not need to structure the argument the way Carrier does.

Secondly, despite Carrier’s pontifications, the fact is that nobody has demonstrated the validity of inductive arguments, including Bayes Theorem.  Hume’s problem of induction has not been resolved; and indeed, if anything has been confirmed by the No Free Lunch Theorem. Further problems have been found by Goodman, and formal difficulties with inductive reading shown by Popper in Chapter 8 of “The Logic of Scientific Discovery”.  Despite these difficulties, I believe that induction is justifiable, and that Bayes Theorem will be at the heart of any successful theory of induction. However, that successful theory of induction does not yet exist. Consequently we cannot rationally castigate the use of more intuitive reasoning in history. https://en.wikipedia.org/wiki/New_riddle_of_induction https://en.wikipedia.org/wiki/Problem_of_induction#Karl_Popper

Thirdly, there are major problems with using Rank-Raglan heroes as a reference class to determine the prior probability that Jesus existed.  Most directly, this creates a problem because Carrier fudges the criteria in order to force Jesus into the class. Applying the Raglan criteria directly, Jesus only has a Rank-Raglan score of 7.5 as shown by Johann Ronnblom.  Therefore, by Carrier’s criteria that a score of at least 12 is required to be a Rank-Raglan hero, Jesus is not one. That contrasts with the score of 20 given by Carrier. Carrier accomplishes this score both by altering the criteria to better fit known accounts of Jesus life (eg, by changing the criteria that the heroes mother is a royal virgin to that of being a virgin simpliciter); and by scoring very loosely (eg, by scoring Jesus a point for being the heir of a king because of the presumed descent from David, even though the majority of descendants of David are not heirs to his throne).  In sharp contrast, Carrier scores Alexander very strictly, coming up with a Rank-Raglan score of 10 - significantly less than the score of 13 by Ronnblom. This is striking. The fact that Carrier is stricter in applying his (modified) criteria to Alexander than to Jesus shows he is motivated by his prior beliefs to ensure Jesus does, and Alexander does not meet the cut of point in applying his method. Regardless of that, the accurate scoring shows Jesus not to be a Rank-Raglan hero, and therefore Carrier’s attempt to establish a prior falls at the first hurdle. http://ronnblom.net/is-jesus-a-rank-raglan-hero/#fnref-36-64 http://www.bibleinterp.com/articles/2014/12/mcg388023.shtml

The problems of Carrier’s use of the Rank-Raglan hero classification do not end there.  If a criteria is to be used to establish a prior probability for use in Bayes Theorem, it is important that it be robust.  That is:

1. The prior probability determined by the criteria should not vary greatly based on random selections of subsets of the available data;
2. The prior probability determined by the criteria should not vary significantly based on different subsets selected on principled criteria, and if they do, those principled criteria must either be used to condition the determination of the prior, or as evidence used to determine the posterior probability; and
3. The prior probability should not be alterable arbitrarily by clearly disconnected events.

A criteria that does not meet these standards will generate priors likely to be shown incorrect by accessing new data, or by more detailed analysis; or that will be potentially gamable.  Further, criteria not meeting these standards will leave open too much scope for researchers to bias the results, either deliberately or unconsciously.

It is clearly the case that the Rank-Raglan hero classification fails to meet all three criteria.  It fails to meet the first criteria because the reference class is too small. It fails to meet the second criteria because there are large differences in the proportion of known historical and legendary Rank-Raglan heroes based on time.  The more recent the time period, the greater the number of Rank-Raglan heroes known to be historical, although that trend has probably now reversed. That, in turn is based both on the fact that the more recently the person lived, the more secure information we have establishing their existence (if they did exist); and most recently, in a change in the type of tropes with which we fictionalize the lives of living heroes.  Finally, the third standard is violated because generation of deliberately fictional characters (such as Gandalf or Aragorn) arbitrarily inflates the number of Rank-Raglan heroes that are fictional. I myself am an unpublished author of fantasy stories. It is absurd that because (at least) one of my invented characters qualifies as a Rank-Raglan hero, the probability that Jesus is a historical figure has decreased, yet Carrier’s use of the type would theoretically make it so.

It is clear that Carrier could largely avoid these problems by restricting his criteria to Rank-Raglan heroes purported to have lived since the fifth century BC, or better, Rank Raglan heroes of the first century Roman Empire.  Doing so, however, would not allow him to retain so low a prior probability. Instead he deflates the prior by allowing in pre fifth century BC figures, almost all of whom are ahistorical.

Finally, there is a problem with Carrier’s interpretation of his prior.  Specifically, for several characters that are listed as Rank-Raglan heroes (King Arthur, Robin Hood) there are historical figures who are plausibly the person tales of whom were embellished to a point which is almost unrecognisable.  For others, or which less is known, it has been speculated that such a figure existed in the past, but that no independent record has come down to us to allow us to identify such figures. These suggestions are speculative, but they are not known to be false.  Therefore it is not known that Carrier’s list of ahistorical Rank-Raglan heroes is also a list of people whose legend grew around a person who actually existed, or whether they were entirely invented.

This is a minor distinction if the question is, how much can we know on the basis of history about Jesus.  If the Jesus legend was seeded from an actual person, then we can know very little about that person (although, not perhaps, nothing at all).  But the existence of such a person contradicts Carrier’s mythicist account of the rise of the Jesus legend. Indeed, it almost precisely aligns with the account given by Ehrman.  So Carrier’s core argument cannot be brought to rebut an account such as Ehrman’s unless he first excluded all Rank-Raglan heroes whose legends may have accumulated around an historical seed - something I doubt he could know.