dabble deepDon't know, don't care, doing it anyway.

BoM Drop: Measuring Book of Mormon Stylometry Claims

A Latter-day Saint with no stylometry training and an AI re-count "it came to pass" from the Hebrew, and what that says about expertise.

The Book of Mormon uses the phrase “it came to pass” about 1,350 times. Believers point to this as a Hebraism, while critics say it is a sad attempt at KJV cosplay. For decades, experts have been using the same tools to argue both sides of this issue, each in their own ways. As a believer myself (I’m a member of the Church of Jesus Christ of Latter-day Saints), I’m not very happy with how that conversation has gone, and I’m getting tired of waiting for “the experts” to correct it. So, here’s me, entering the chat!

“But bro”, you might say, “you know nothing about stylometry! You’re no expert!”

To that, I say what I always say: you’re right, I don’t know, but also, I don’t care, and I’m doing it anyway, with the help of AI. You see, I might not know much about it, but AI does. Also, AI is a very efficient and effective researcher and coder to boot, and all of those skills will come in handy on this project. In fact, when I was first testing this idea, AI almost immediately found things that told me early on that this was going to be a fascinating exercise. In particular: critics are dead wrong on a few key points, especially with regard to “it came to pass” ratios between the Book of Mormon and the KJV Bible, but also: believers have sometimes taken their own claims a little further than I am comfortable with as well.

Let’s dig in!

Where I’m coming from

As I said already, I’m a member of the Church of Jesus Christ of Latter-day Saints. My personal approach to this is obviously going to be somewhat biased. However, I do want to point out something that I consider to be an advantage: I already believe the Book of Mormon is true.

Why is that an advantage?

Consider: if a thing is actually true, then no amount of evidence against it can change that. This means that I can engage directly with evidence that seems to refute my position and learn from it without feeling threatened. However, I’ve seen some people in this particular chat allowing their beliefs to color their arguments a little too much. I have witnessed firsthand the devastating consequences of basing your beliefs on evidence alone, only to find that the evidence you trusted is not even slightly valid. I wish to avoid causing that kind of damage to others, but I also wish to contribute constructively to this conversation.

To that end, I worked with AI to come up with a few guidelines for this project:

  1. Report results the same way whichever way they point.
  2. Compare like with like: narrative with narrative, the Hebrew rather than the KJV’s English, a matched Egyptian text or no verdict, etc.
  3. Write the test down and hash it before running it; never edit it afterward.
  4. Everything reproducible from public texts, by anyone. The repo for this project is on GitHub.
  5. Record every result where the next person (or AI) will look first. To that end, I’m using Requiem on this project as both a spec manager and a results corpus.

It turns out that these rules are extremely helpful! This approach helped me overrule the AI in many cases, and it kept me in line a few times as well. They also helped highlight when claims made by the AI took things too far. Remember: the vast majority of LDS-related content online is very negative towards my religion, and that’s what the AI has been trained on, so keeping it in line is rather important, and I know I’ve still missed a few instances.

The question

“It came to pass” in the KJV is how translators rendered certain instances of the word wayhi. The hypothesis is that a book translated from a Hebrew-ish original text would use the various forms of that word in more or less the same way, with some impacts bleeding in from any other involved linguistics.

Critics make the (accurate but useless) headline claim that The Book of Mormon uses “it came to pass” roughly 8 times as often as the KJV Bible does. That is an English-to-English comparison though, which doesn’t work for the Bible because it wasn’t originally written in English. For the Book of Mormon, the closest thing we have to a source text is the English itself, but it is important to note that it claims to have been written in “reformed Egyptian”, theorized to be somewhat-modified Egyptian characters used to express a form of Hebrew. Taking a hint from Donald Parry (Ensign, 1992; Preserved in Translation, 2020), we will measure Bible usage from the Hebrew, not the KJV’s rendered English, because there are inconsistencies with how wayhi and its forms were translated.

The Book of Mormon has a different problem we need to account for: it contains whole Isaiah chapters and other Old and New Testament passages worded almost exactly as they are in the KJV. To account for this, we will need to exclude those parts from the overall consideration so as to not artificially inflate the similarities between the Book of Mormon and the Bible with the Bible’s own passages.

Finally, we will need to consider the grammatical contexts in which wayhi forms occur separately. I didn’t do this in my first runs because I wasn’t aware of the importance of it, but this turns out to be one of the most important aspects of this research. In particular, biblical Hebrew lacks uses of wayhi that jump straight into the event (“it came to pass that” without “after” or any other time/setting words to decorate it first), but a significant portion of the Book of Mormon’s usages follow exactly that straight-to-event pattern. By separating the different kinds of wayhi usages into their own buckets, we’re able to isolate each type and measure those buckets on both sides independently, and we’re also able to identify where comparison makes sense versus where it is virtually impossible due to the lack of a pre-translation source text for the Book of Mormon.

Lastly, we’ll need to account for genre, which, it turns out, impacts word choice far more than authorship does. If we don’t account for genre, we’re not really comparing like with like, so the numbers will ultimately mean less than we need them to.

What the math reaches, and what it doesn’t

For details on methodologies, decisions, source code, etc., please see the repo and the detailed report it generated for these tests.

English to English Direct Comparison

Critic claim: the Book of Mormon uses “it came to pass” 7.9 times more often than the KJV Bible does.

Believer response from Parry: The Hebrew Bible has roughly 1,204 wayhi, far more than the KJV’s “came to pass”, because the translators didn’t render it consistently.

The critic claim is accurate arithmetic but misleading, and the believer claim doesn’t do the best job of differentiating between the different grammatical contexts for wayhi-related words, so it doesn’t tell the full story. You only get near the 1,204 count (1,224 by my count) if you pool past wayhi with future wehayah and their look-alikes.

If you account for the wayhi counts on the Old Testament using the Hebrew, you get some interesting results. These results get even more interesting when you attempt to isolate like-for-like genres in the Old Testament. The results:

Old Testament side Book of Mormon side all wayhi (count, per 1,000) Book of Mormon per 1,000 ratio
Whole Old Testament Whole text 784, 1.28 5.04 3.9×
17 narrative books Without KJV quotations 629, 1.74 5.26 3.0×
Story books (no Leviticus, Numbers, Deuteronomy) Without KJV quotations 605, 2.19 5.26 2.4×
Story books, narrative only Narrative only 576, 3.43 7.76 2.3×
17 named stories, narrative only Narrative only 365, 4.28 7.76 1.8×

What this means: the more you account for genre and Hebrew sourcing, the closer you get to an equal footing between the two texts. Still, in every case, the Book of Mormon legitimately uses more instances of “it came to pass” than the Bible as a source text does, even when accounting for genre. This gives us some hints about what is actually going on, but we’re missing something in this table that could give us critical insight: more dimensions.

Buckets of Wayhi

In order for something to be a Hebraism, it must follow the same patterns that the Hebrew follows. When we examine the differences between usages in the Book of Mormon vs the Hebrew sources for the Old Testament, we find an interesting pattern: some Hebrew uses are very consistent in terms of context and grammar in ways that only match some of the Book of Mormon uses.

In order to more accurately measure things, we need to first carefully define the grammatical contexts and identify whether they can be compared. Some cannot be because we don’t have the ancient source text for the Book of Mormon, but we do have it for the Old Testament, and can therefore work from there. When comparison can be done in a like-for-like manner, we did so. Also, for simplicity, I show you here the comparison table for only the 17 named stories, narrative only, because including all of the others would make this table more difficult to read. The ordinary verb bucket is included below in order to subtract those instances out of the Old Testament counts because the KJV almost never renders them as “it came to pass” (8 times out of 317), so they don’t match up with anything on the Book of Mormon side. The other two included buckets are the most interesting:

Bucket Example Hebrew wayhi (count, per 1,000) Book of Mormon “it came to pass” (count, per 1,000) Book of Mormon ÷ Hebrew
Setting first “And it came to pass, when they had eaten, that…” 249, 2.92 336, 2.32 0.80× (0.67–0.94)
Straight to the event “And it came to pass that they did march…” 0, 0.00 788, 5.44 — (no Hebrew uses)
Ordinary verb “And there was light” (wayhi with its own subject) 116, 1.36 0, 0.00 — (not rendered “came to pass”)
All past uses 365, 4.28 1,124, 7.76 1.8×

What this means: the Book of Mormon’s rate of “setting first” usages actually measures slightly lower than the Bible’s usage, which is a hard reversal on the critical claims. However, and here’s the key: the Book of Mormon includes a whole class of usages, the “straight to event” bucket, that does not at all appear in the OT Hebrew. To me, THAT looks something like a smoking gun! Definitely worth examining more closely!

Poor Imitations

One test we ran was to pull known period pastiches, books like The Late War and others that critics have claimed the Book of Mormon was based on or straight up copied from. I won’t go into extreme detail here, but I will note: none of the pastiches we tested got beyond 1/3 of the usages of “it came to pass” that the Book of Mormon has. In that sense, either the Book of Mormon was an extremely clumsily executed pastiche (unlikely), falling short by a factor of three or more relative to “comparable” works from that time or earlier, OR it is not a pastiche at all.

Interesting!

Walk Like an Egyptian

We know that the Book of Mormon has an entire category of “it came to pass” usage that does not even appear in the Hebrew for the OT. Where might that be coming from? Perhaps we should consider what the Book of Mormon itself says: there are two distinct linguistic influences on the Book of Mormon text: Hebrew, and Egyptian. We’ve accounted for the Hebrew, but we’ve not even looked at the Egyptian yet. Let’s do that now.

It turns out there are precious few publicly available Egyptian narrative texts, with translations, from before about 500 BC (Lehi leaves Jerusalem for the New World around 600 BC in the book’s own chronology). However, we were able to identify a small handful of these, but they may or may not match in terms of genre all that well. Still, there is also the examination of ancient Egyptian grammar with regards to a wayhi-like character to consider.

First, the Egyptian texts and their results:

Text Stage Composed “Then” particles Per 1,000 est. words
Sinuhe Middle Egyptian c. 1900 BC 7 1.28
Shipwrecked Sailor Middle Egyptian c. 1900 BC 27 12.17
Tales of pWestcar Middle Egyptian c. 1600 BC 126 21.77
Eloquent Peasant Middle Egyptian c. 1850 BC 11 2.04
Kamose Stela II Middle Egyptian c. 1550 BC 1 0.70
Tale of Two Brothers Late Egyptian c. 1200 BC 70 11.02
Contendings of Horus and Seth Late Egyptian c. 1150 BC 157 22.44
The Doomed Prince Late Egyptian c. 1300 BC 34 14.24
The Taking of Joppa Late Egyptian c. 1250 BC 1 1.20
Truth and Falsehood Late Egyptian c. 1250 BC 30 16.85
Astarte and the Sea Late Egyptian c. 1400 BC 9 8.07
Report of Wenamun Late Egyptian c. 1050 BC 0 0.00
Bentresh Stela Late Egyptian c. 400 BC (too late to count on its own) 23 16.15
Piye Victory Stela Late Period c. 727 BC 53 8.26
All 14 named narratives, pooled 549 10.55
Book of Mormon, 1830 narrative 8.01

Next, the results by usage type bucket between event-first vs time/setting first:

Source Form Setting first Straight to the event Share straight to the event
Hebrew narrative (17 stories) wayhi 249 0 0%
Egyptian narratives (PREREG 2) ꜥḥꜥ.n, wn.ꞽn — all 100%
Demotic sentences, 700–300 BC (strict) ḫpr 8 7 47%
pRylands 9 narration (c. 513 BC) ḫpr 1 11 92%
Book of Mormon narrative “it came to pass” 336 788 70%

It is extremely important to note this one thing: the Hebrew side exclusively features the time/setting usage, while the Egyptian narratives exclusively use the straight-to-the-event type. Demotic has a mix between the two in our source texts, but there are too few of these to be conclusive about much. Then we have pRylands, which scores very highly on the straight-to-the-event bucket.

What this means: this does not prove nor rule out the hypothesis that the overages in the Book of Mormon are attributable to the linguistic contribution of Egyptian, because our source pool of truly comparable ancient Egyptian texts is rather shallow, with instance counts in the tens instead of hundreds for each individual document. In order to reach statistical significance, you have to pool the documents together, which reduces their resolution and power in terms of being able to prove anything in particular. However, the fact that these ancient Egyptian documents seem to lean in the direction of the unaccounted-for usage bucket in the Book of Mormon suggests that there might be something here worth further research.

For me, this suggests that the Book of Mormon could actually be carrying the fingerprint of the lost language it describes as “reformed Egyptian”. In other words, the very thing critics point to as evidence it is an obvious fraud may well be some of the best evidence we have of a mixed Hebrew and Egyptian linguistic influence on it. Take that, Richard Dawkins!

A Man of Habits

One interesting observation is that, whatever the original source of that “it came to pass” usage was, Joseph Smith continued to use it in similar ratios in other, later works. This includes his work on the Joseph Smith Translation of the Bible, in which Joseph’s additions were 81% of the straight-to-the-event variety, which doesn’t match the Bible’s own usage. One could argue that Joseph was restoring some previous original state that is not today accurately represented in the KJV, but I personally think that an intensive effort to translate the Book of Mormon in so short a time as Joseph did it would be enough to instill that habit on virtually anyone, so I think either explanation can work without at all damaging my belief that he was in fact a prophet of God.

And, as it turns out, that wouldn’t be the only sign of Joseph’s habitual fingerprints on the Book of Mormon! I’ll cover that in a later post.

How a non-expert did this

So, how did I, a stylometry noob, get here? The same way you can: using AI. This project is something I started in September 2026, and I’ve only devoted a few hours of my time in total to it. I’m not the only person with interest in this discussion who doesn’t have the expertise they think they need to join in on it, so I encourage anyone and everyone who has an interest to get their own hands dirty instead of waiting on the “experts” to do it for you. Chances are, they’ll never get around to it anyway, and if they ever do, it won’t be to your liking, because they don’t have your perspective.

In the past, research like this required a lot of time, and it would also require a lot of effort to digitize records into a format accessible to AI. Others have put in a lot of work to make this kind of research possible, so really, I’m standing on the shoulders of giants here. However, that is not to say that those giants never made mistakes, or that their research turned over every rock that’s actually important.

A few notes though on this particular project: AI has been trained on a lot of material that is very critical of religion in general and very critical of my religion in particular. Because of this, I often caught it hedging claims and results so heavily that the results themselves got buried. Having a set of ground rules helps keep it from going out of bounds and helps keep it focused on what actually matters and on how to talk about it without exposing its own built-in bias. In short: AI is biased by design, and even if you took that away it would still be biased, because it is trained on our own outputs, and people are biased. We have to be careful with that on projects like this. These and other rules in my project caught AI being confidently wrong many, many times, and paying attention myself was the only thing that kept those things from hurting this project. However, I might have missed a few things, so you’ll probably want to pull my work from GitHub and see for yourself.

On the division of labor in this project: I don’t write the code, and I don’t often review it. I simply set the rules, ask the questions, decide which threads to pull further, and correct things when they go off the rails or don’t go far enough. This isn’t just for the AI, it’s also for me: the rules you run the project by largely determine the quality of your results. While I learned a lot about how stylometry works, I am still not a stylometry expert. All I did was hire one in the form of AI and walk it through.

…And that’s what you have to do if you have a project like this that you want to get involved with. As Mark Watney says in The Martian: “You just begin.” Start now, from wherever you are, and figure it out; you have a brain for a reason. Have AI tutor you on the general concepts so you know enough to guide the work in the right direction. Don’t get discouraged when one question turns out to be a dead end. It was cheap to ask, and it’s cheap to come up with other equally important questions. Also, dead ends are not always what they appear to be; sometimes they are just personal or AI bias getting in the way.

Good luck, and happy dabbling!

Sources