You are correct that Paul overall is absolutely an imminentist (he thinks the end is due any time no, and lots of passages clearly articulate this, from 1 Cor 15, esp. v. 20, to 1 Thess 4, and beyond). But his information comes from an imaginary Jesus (someone he talks to in his head).

So I don’t think Thompson is arguing against that, per se. You are referring to his early-chapter critique of Crossan (p. 17 in the hardback; cf. pp. 26–28). Thompson is not saying Paul did not think the end was nigh, but that he didn’t need a historical Jesus to get that idea from (1 Cor 7:29 derives from scripture, not Jesus). Thompson may have less literal ideas about what Paul thought “the end” meant, which I wouldn’t agree with; but Thompson is, IMO, never clear enough about that for me to know for sure.

Thank you. Be well.

You have the right idea with “how closely the passages mirror each other.” Two authors can independently come up with an analogy or turn of phrase. But at some point the exact parallel is too improbable by chance. But even model 1 entails the word already exists as vernacular. So model 1 does not mean independent invention of the word (but it includes that).

In mimesis criticism this criterion is called “density,” where the number and precision (the overall density) of parallels is less likely by chance accident than borrowing. The only question then is what kind of borrowing (hence the other models).

When it comes to “how could we begin to really lay out the probabilities” there is a hard way to do it (you can build a Laplacean curve for it) and a soft way (the mental equivalent of “eyeballing it”). On the soft way, “full of idols” (or any extremely rare word) would not suffice to rule out model 1 (your intuition is correct there). It’s the full conjunction that can’t be chance (we don’t need to do the math to tell it will not get a high probability for the null hypothesis): a theô– root participle (seeing) + “full of idols” (rare word) + “the city” in accusative (describing the city as full of idols): three matching details, in the same order, saying the same thing.

Slam dunk if they were both spoken about the same city (Athens), as that fourfold parallel would be near impossible by chance, but that doesn’t survive review even if Bilby had been right about the text of Stratagems. But with the three-word/same-order matrix, we’re already well below 50% by chance and above 50% by mimesis. Which is how we can be confident the medieval redactor used this line from Acts to revise the Strategems (plus the added fact that we know that redactor knew and loved Acts and would be well motivated to play on it here).

Your closing question is right to ask. An example would be archaeologists 2000 years from now working from the ashes of WWIII and only observing less than 1% of the literature of the 20th-21st century and seeing the weird word “muggle” showing up in disparate contexts. You and I know that can’t be a coincidence (it entails some model 2, 3, or 4 back to Harry Potter because we know the first appearance and that it was then invented, and will be available in model 1 as soon as people start using it without knowing where it came from or originally meant). But how could those archaeologists know that?

Best they could say is that the word must have been invented sometime within a century or two before the first appearances they find of it. If they were lucky enough to have the whole Harry Potter series, they could make a decent argument that it must have been invented there, due to how it shows up there and is so specific to that work. Or if they had extant discussions of it originating there. And so on. But odds are they wouldn’t have either. So they’d just be seeing this word show up rarely sometime probably in the 21st century. And seeing it in several disparate places, they might not be able to rule out model 1.

But suppose they saw in disparate places the whole line “Don’t let the muggles get you down” or close equivalents (like “Don’t let those muggles get to you”). That can’t be accidental use of a common word. That’s a quotation or paraphrase, which entails a common source. That rules out model 1.

The only question is then model 2, 3, or 4. Suppose, say, they saw this in a news editorial in a specialist trade newsletter, and then in a private chat between people with no connection at all to that. Then model 2 or 3 is unlikely. Though not impossible, it’s not likely those two chatters were reading that newsletter and riffing on it, or that the author of that article knew their private chat and was riffing on that. More likely they are both riffing on some common source they share, and it might not be possible to know what source that was or its date. Only that it predated both and was likely to be something both would be exposed to despite their social differences. Popular fiction would thus be the most likely hypothesis.

I begin to think of the difference between supernatural romance and literary fiction. There are tropes, characters, phrases, and specific words that will show up all over the place in supernatural romance novels, but maybe only once in literary fiction. If it happens to be that most supernatural romance novels don’t survive 2000 years, but the literary fiction ones do, would finding a choice phrase extremely uncommon in literary fiction actually tell us if that phrase was uncommon overall?

data science methods for authorial identification to identify ‘cascading strata’ in the gospels

I am very skeptical of that.

It’s statistically impossible to do data-based micro-stylometrics. The smallest continuous passage you even have a chance at is around 50 words (it can’t be done with isolated vocabulary changes or single line insertions), and you need a large corpus (tens of thousands of words, ideally a hundred thousand) of a known author to identify. We don’t have that for any Gospel. Even with small chunks, and even more so without a corpus from a specific suspect author, the style needs to be very different (as in, substantially more different than average style variations between authors) to stand out statistically, which depends too much on luck (especially when we are talking about authors trying to sound like each other, or like other revered text types, like the Septuagint). Otherwise the signal dissolves beneath stochastic noise.

I am also aware that he wants to do this with Marcion, but we don’t have Marcion. We have multiple, incomplete, conflicting, modern reconstructions, based almost entirely on Latin, and not just a Latin translation, but a particularly dishonest and unreliable translator, Tertullian. The risk of circular arguments and confoundment is so high here I do not believe this is possible. I strongly suspect reliable results here just aren’t to be had. And the over-confidence I am seeing from Marcion prioritists on this does not cultivate confidence in their results.

There is also a problem with confusing distance grams with chronology or order. I have not seen what Bibly is doing, but I have seen others do this. Distance grams calculate distance in conceptual space, not temporal space. The trees are not actually causal, but “closeness in style,” which is not the same thing. To propose the distance in style is a causal or sequential order (or even to claim you know the direction of the sequence) is to propose a hypothesis that then has to be tested against evidence capable of discerning this (distance grams cannot do this by themselves). And often no such evidence is available and so, again, you have to accept that there are things we just can’t and probably will never know, and we should stop trying.

Also, with respect to your point, distance grams are artificially constructed. The same author can generate a whole tree of distance grams and thus “look like” a bunch of authors and redactors, but that is a statistical artifact of the metrics and settings chosen by the stylometrist. This is easily shown by generating the same trees for known single authors (e.g. do all my books and watch what happens). I discuss this in my Stylometry articles (Savoy gives us a good example of these mistakes, as does Tuccinardi).

Sure, it´s one huge advantage when you´re dealing with English or Spanish, even in the Middle ages, and especially the Renaissance, you have easily enough data to track this.

Unfortunately for us, these kind of works are probably the hardest to date using these kind of linguistic techniques, because of the problem of emulation of older varieties of Greek.

I´ve been browsing linguistics papers on Koine Greek today and there´s been a huge amount of work recently looking at evidence from letters, from contracts, etc, very interesting work is being done tracking pronunciation changes, grammatical changes, etc. It probably doesn´t help someone like Mark Bilby so much, but it´s super interesting in its own right.

Looks like a Shaphan to me.

Thanks for this in-depth reply! Just to clarify, I wasn’t proposing my own idea, this was from Bilby’s podcasts about his prior publication, applying data science methods for authorial identification to identify ‘cascading strata’ in the gospels (including reconstructed Evangelion & Q). Which again, struck me as likely to be at risk of high false positive rate on the details, plus the problems with assuming Q; but just seemed interesting in a more general sense: given unknown authorship, why assume relative authorial separation? – might there be interesting new ideas to explore by considering a scenario where they were in discussion / swapping ideas (in person, in text, or both)? The particulars of Bilby’s waterfall diagram might constitute improbable epicycles; but the idea that in a relatively small educated-Christian community, gospel authors might be trading material back & forth, doesn’t seem obviously less plausible than one-way lines of influence?