That’s all true. But IMO, as I show in OHJ (citing the study by Hamerton-Kelly), the story was invented symbolically, not because it was a prophecy Jesus had to fulfill. It’s simply fiction illustrating what the author thinks is why God allowed the Romans to destroy the temple and stay in power. It crucially connects intentionally with the “Jesus hates figs” story and the “it’s all about prayer now” denouement. Mark did not care whether the story was plausible, as he never cared about that (the fig story isn’t even plausible for a supernaturalist worldview, much less a naturalist one; so Mark was not even trying for plausibility).

We know the temple story is a fabrication because of the physical side of it and the fact there were guards present to stop such things. It would’ve taken a small army to do what Jesus was portrayed as doing. It’s telling that the story is connected to prophecy. The writers of the gospels were then compelled to enter the story because of it being foretold in Jeremiah 7

There is one “sort of” exception that ends in the same place I do in the lecture but someone might want to see that walked out to understand it.

I treat the vocabulary and grammar/storytelling evidence separately. But technically these will have some dependency relation, since a forger who fails at one of those tasks is more likely to fail at the other, while a success at one of those tasks could cast a different light on the failure of the other.

The following explains why I skipped that in the lecture:

⌘

The relation is not total and would only reflect a bottom-margin effect. At the a fortiori margin (the one I calculated), this small dependency effect is already accounted for; but you’d have to actually work it out at the a judicantiori margin.

On this demarcation of the error margins see Proving History pp. 85–88, On the Historicity of Jesus, p. 599, and Obsolete Paradigm, pp. 154, 224, and Appendix.

But to illustrate how in this case any residual dependency relation is already factored in at the top and would factor in at the bottom, I provide the following:

⌘

If all we had were the vocabulary incongruity, someone might say that’s weak evidence because “it could be a coincidence.”

That actually entails a low probability, which is what we have to mathematically account for, i.e. the 1/20 margin I assign there is already assuming this is correct, that “it’s just a coincidence.” That’s simply the frequency with which we estimate Josephus would himself fail at his own vocabulary tendencies, relative to how often a forger would.

So the 1/20 I assigned is literally the likelihood ratio given that hypothesis—in comparison to forgery, which all but entails this effect and thus this observation is far more “likely” on “forgery” than on “coincidence.” We are simply measuring that fact, by comparison to how often we think Josephus naturally conjoins a density of incongruous vocabulary like this.

But if we then showed someone the grammar/storytelling incongruities, they’d agree that makes the vocabulary argument for forgery stronger. Not only because two coincidences is less likely than one, but also because it’s harder for an author to muck their own grammar and storytelling tendencies (and harder for a forger to get those right) than for vocabulary. And this has to be reflected mathematically. So I reflect it in the 1/20 for this observation, making the combined observation much stronger, at a 1/400 likelihood.

This reflects the fact that Josephus might rarely conjoin so much incongruous vocabulary but retain his characteristic grammar and storytelling, while (conversely) a forger might nail his characteristic grammar and storytelling (in trying to sound like him, or in choosing a way of telling the story that just happens to resemble his) but muck up the vocabulary. So seeing both does amplify the unlikelihood.

And of course the grammar/storytelling might be ambiguous as well (neither clearly Josephan or un-Josephan) while the vocabulary remains incongruous or vice versa. So to have both failures together is much stronger evidence than having only one of them clearly observed. While having one failure and not the other leaves us in a weaker evidential position (where we are positing only one rather than two coincidences).

So these two categories of evidence are independent enough to get our a fortiori result. Because the one observation is not automatically caused by the other, and they are not always correlated.

But, for example, one might think this vocabulary incongruence is really 1/1000 unlikely and the grammar/storytelling incongruence 1/2000 unlikely (at the a judicantiori margin), but think that any forger who mucked the vocabulary would also be more likely to muck the grammar and storytelling as well, or vice versa.

Suppose that this correlation means that both combined will be found a thousand times more often than chance alone would predict. Chance alone predicts it will happen 1/000 × 1/2000 = 1/2,000,000 times. But if the conjunction is a thousand times more common than that, then it will happen 1/2,000 times.

That means every g-fail comes with a v-fail while only half of v-fails come with g-fails. But since g-fails can surely happen without v-fails (indeed it’s easier to fake the vocabulary), their conjunction cannot be a thousand times more common than chance.

Let’s say it’s 500 times more often. Then conjunctions occur 1/4,000 times. That means g-fails come with a v-fails half the time, while v-fails come with g-fails a quarter of the time. Which is as expected (an author is more likely to fail their own v than their own g).

But as these effects all fall below our a fortiori margin, they don’t affect it, as the a fortiori estimates have already accounted for this effect. But at the a judicantiori side we’d have instead of 1/1000 and 1/2000, either 1/1000 and 1/4 (accounting for the dependency effect from v to g) or 1/2000 and 1/2 (accounting for the dependency effect from g to v), either way getting 1/4000, not 1/2,000,000.

And these are just hypothetical numbers to illustrate the point. The actual correlation size and thus its effect on likelihoods would have to be actually calculated from a detailed study of how often the observed scale of v-fail and g-fail occurs in authentic Josephus material, and how often these two fails are found together. And the result of such a study will not likely be anywhere near our a fortiori estimate and thus won’t alter it (other than to demonstrate that it should be worse than even we assigned).

Yes. See other comment. This one may be up in a few days or weeks.

Are the zooms libraried somewhere please?

Indeed. I am hoping to get that done next month. It’s not a high priority. The case I summarized in the zoom already decisively refutes him. So writing an article would only be showing that. But it’s worth showing that.

Thank you!

These happen only because they are funded. So definitely throw thanks to the host, via their YouTube channel housing all the lectures I do with them (including, soon, this one).

Schmidt does specifically criticize one (and only one) argument of mine, but only in an appendix. So I will get to that in a separate article to come.

I take it this is still to come?!

“By the 90s at least Mark & Matthew were known” – what is the evidence? Do you mean as complete texts? What version of Mark?