Stylometry and the Book of Mormon

What the statistical analysis of authorial fingerprints tells us — and what it does not.

What Stylometry Is and Why It Matters

Stylometry is the statistical analysis of writing style — word frequency, sentence length, function word patterns, vocabulary distribution — used to study authorship. These linguistic patterns are largely unconscious; a writer cannot easily fake or suppress them across hundreds of pages. Stylometry has been used in legal proceedings to authenticate disputed documents, in literary analysis to weigh the authorship of anonymous texts (including disputed Shakespeare plays), and in forensic linguistics to match writings to known authors. The method is powerful, but it is not infallible — results depend heavily on the words measured, the comparison texts chosen, and the statistical model applied.

The naturalistic argument is this: Joseph Smith wrote the Book of Mormon himself, and stylometric analysis should reveal one consistent authorial voice throughout the text. If that claim is correct, then all the named authors within the text — Nephi, Jacob, Alma, Moroni, Mormon — should share a single stylistic fingerprint. The believing counter-claim is that the named authors will instead come back as statistically distinct from one another.

Stylometric analysis tests this claim directly. Either the named voices are statistically distinguishable from one another and from Joseph Smith, or they are not. This is an empirical question — and, as we will see, a genuinely contested one.

The Argument Stylometry Is Supposed to Answer

Critics of the Book of Mormon claim it was written by Joseph Smith — or alternatively by Solomon Spaulding, Oliver Cowdery, or some combination thereof. The claim of sole or small-group human authorship is foundational to every naturalistic explanation of the book’s origin.

If that claim is correct, then all the named “authors” within the text — Nephi, Jacob, Alma, Moroni, Mormon, and the rest — should show the same stylistic fingerprint as their actual author. A single author writing under multiple pseudonyms, the believing argument runs, cannot easily simulate independent prose styles across hundreds of pages.

Stylometric analysis puts that to the test. The results have not been one-sided, and honesty requires presenting both directions.

The 1980 BYU Studies Wordprint Study

The study most often cited in this debate is Wayne A. Larsen, Alvin C. Rencher, and Tim Layton, “Who Wrote the Book of Mormon? An Analysis of Wordprints,” published in BYU Studies 20, no. 3 (1980): 225–251. It is essential to describe what this study actually did, because it is routinely mischaracterized — by defenders and critics alike.

The researchers identified 24 internal Book of Mormon authors who each have roughly 1,000 or more words attributed to them — figures such as Nephi, Jacob, Alma, Mormon, and Moroni. They then applied three separate statistical methods to the wordprints (patterns of noncontextual, high-frequency function words) of those 24 authors. The number 24 is therefore the count of internal authors analyzed; the number three is the count of statistical tests used. Neither number is a count of “voices discovered.”

Their conclusion was that these 24 internal authors are statistically distinguishable from one another, and distinguishable from Oliver Cowdery, Joseph Smith, and Solomon Spaulding — the leading naturalistic candidates of the day. You can read the original study at BYU ScholarsArchive.

The 1990 Hilton Study: The Stronger Pro-Historicity Result

If the 1980 study is the one most often cited, it is not the strongest. In 1990, John L. Hilton worked with a team of researchers — a group based around the University of California, Berkeley that included scientists who were not Latter-day Saints — expressly to test whether wordprinting could withstand a far more demanding design. Their study, “On Verifying Wordprint Studies: Book of Mormon Authorship,” was built to be skeptic-proof: it measured noncontextual word-pattern ratios, drew on large control samples (26 texts by 9 control authors, yielding 325 pairwise comparisons), and imposed a strict minimum sample size of 5,000 words so that no conclusion rested on a thin slice of text.

The conclusion was pointed. Hilton’s team found it statistically indefensible to propose Joseph Smith, Oliver Cowdery, or Solomon Spaulding as the author of the Nephi or Alma material, and they found that the Nephi and Alma blocks measure as statistically independent of one another. Because the method was more conservative than the 1980 study, the control set larger, and several of the researchers were not members of the Church, this is the pro-historicity result a critic can least easily dismiss as motivated reasoning — and methodologically it is the more robust of the two friendly studies. You can read it at BYU ScholarsArchive.

What the Studies Show — and Where They Disagree

A fair summary of the wordprint literature on the Book of Mormon looks like this:

The State of the Evidence

The 1980 study — Larsen, Rencher, and Layton applied three statistical methods to 24 internal authors and concluded those authors are statistically distinct from one another and from Smith, Cowdery, and Spaulding.

The 1990 Hilton study — A more rigorous follow-up, with larger controls and a partly non-LDS research team, found it statistically indefensible to name Smith, Cowdery, or Spaulding as author of the Nephi or Alma text, and found Nephi and Alma independent of each other.

The strongest skeptical study — David I. Holmes (1992), writing in the Journal of the Royal Statistical Society, read the textual variation as a single prophetic register rather than distinct authors.

The field is narrow — Mainstream computational stylometry has barely engaged the Book of Mormon. This is a small, largely insular exchange among a handful of statisticians and a few critics, not a broad scholarly battlefield.

Later studies diverge — A 2008 analysis (Jockers et al.) using closed-set methods attributed portions of the text to nineteenth-century candidates; a peer-reviewed open-set reanalysis (Schaalje et al., 2011) reversed the result. Methodology, and especially corpus selection, drives these differing outcomes.

This matters because the Book of Mormon is not presented as a single-authored text. It claims to be an abridgment of multiple records by multiple historians, later translated through an inspired process. The 1980 and 1990 wordprint data are consistent with that claim. But “consistent with” is not “proven,” and the method itself remains debated.

A useful, balanced overview of the question is available from Scripture Central, “What Can Stylometry Tell Us about Book of Mormon Authorship?”

The Critiques: Holmes (1992), Taves (1984), and Jockers et al. (2008)

Honesty requires acknowledging that the wordprint method has serious critics, and that not every stylometric study points the same way. The strongest of these is not a polemicist at all — it is a mainstream statistician publishing in a top-tier journal.

Holmes (1992). David I. Holmes, a statistician with no stake in the religious question, published “A Stylometric Analysis of Mormon Scripture and Related Texts” in the Journal of the Royal Statistical Society, Series A, 155, no. 1 (1992): 91–120 — a far more prestigious venue than either side’s house journals. Measuring vocabulary richness rather than function-word ratios, Holmes found that the Book of Mormon’s internal “authors” did not separate into distinct voices. He read the variation as a single prophetic register — one writer shifting into an elevated scriptural style, distinct from Joseph Smith’s ordinary prose but not internally differentiated. This is the most serious empirical challenge to the wordprint case, and any honest treatment has to put it on the table. (Holmes 1992)

Taves (1984). Ernest H. Taves published a critique in the skeptical journal Free Inquiry, questioning whether the wordprint technique could reliably separate authors at all. The concerns centered on the choice and stability of the function words measured, the sample sizes available for some internal authors, and the statistical assumptions behind the tests — in short, that the method’s discriminating power was overstated.

Jockers et al. (2008). Matthew L. Jockers, Daniela Witten, and Craig Criddle published an analysis in Literary and Linguistic Computing using two methods — the delta method and nearest shrunken centroid (NSC) classification — that reached an anti-historicity conclusion, attributing large portions of the text to nineteenth-century figures such as Sidney Rigdon and Solomon Spaulding. But the design had a decisive flaw: it was a closed-set test that offered only seven candidate authors — five with known or alleged Book of Mormon connections plus two controls — and was therefore forced to assign every passage to one of them, even where none was the true source. It did not include Joseph Smith himself among the candidates, and it took no account of the book’s claim to be a translation rather than an original composition. Given those constraints, the analysis could not have returned any answer except a nineteenth-century one.

The peer-reviewed reply. The response to Jockers was not confined to blog posts. G. Bruce Schaalje, Paul J. Fields, Matthew Roper, and Gregory L. Snow published “Extended nearest shrunken centroid classification: A new method for open-set authorship attribution of texts of varying sizes” in Literary and Linguistic Computing 26, no. 1 (2011): 71–88 — the very journal that ran Jockers. Their open-set method, which is permitted to answer “none of these candidates,” produced dramatically different results from the closed-set analysis. A rebuttal in the same venue, under peer review, carries far more weight than an informal objection. (Schaalje et al. 2011)

The honest position is this: wordprint evidence is suggestive of internal stylistic variety in the Book of Mormon, but it is not dispositive, and competent researchers have reached opposite conclusions depending on their methods and comparison corpora. Anyone who tells you stylometry has settled the authorship question — in either direction — is overstating the evidence.

The Translation Problem: The Real Hinge

There is a deeper issue underneath every wordprint study, and it cuts in both directions. Whatever stylometry measures, it measures the dictated English of 1829 — not the reformed Egyptian an ancient author is said to have engraved on the plates. The entire function-word method rests on an assumption: that an author’s unconscious, noncontextual word-rate “fingerprint” survives translation into another language intact. That is an assumption, not a demonstrated finding, and it is the genuine pivot of the whole debate.

If the fingerprint does not survive translation, then a finding of distinct internal voices (Larsen, Hilton) is harder to credit as proof of distinct ancient authors — and a finding of nineteenth-century English (the critics) is exactly what one would expect from any translation produced in that era, and so proves nothing about the underlying source. The translation layer simultaneously limits what the believing studies can establish and explains away the critics’ strongest-sounding result. It deserves to be named plainly rather than left implicit.

Why Corpus Selection Drives the Results

A recurring theme in this debate is that the comparison corpus — the set of texts a study measures the Book of Mormon against — has an outsized effect on the outcome. If you test whether a text resembles 1830 American writing by using 1830 American writing as your comparison set, a match tells you the dictated English is of its period, which is expected of any translation produced in that era. It does not, by itself, identify the author.

Likewise, a closed-set classifier that is only permitted to choose among a handful of nineteenth-century candidates will always return one of them, even if the true source is not in the set. This is a known limitation of the method, not a refutation of either side — but it does mean the same data can yield different headlines depending on how the experiment is framed.

The reasonable takeaway is modest: the internal stylistic variety the 1980 and 1990 studies reported is real and interesting, the critiques of the method are also real and substantive, and the question of who the voices belong to is not something stylometry has closed.

Why the Book’s Own Structure Is Relevant

The Book of Mormon does not claim to be the work of a single author. It presents itself as a compilation: an abridgment made by Mormon from a centuries-long archive of records written by distinct individuals across vastly different historical periods. On its own terms, then, Nephi should read differently from Alma, Alma differently from Moroni, and Mormon’s editorial voice should be distinguishable from the voices he abridges.

The 1980 and 1990 wordprint studies reported exactly that kind of internal differentiation. Whether that differentiation is best explained by genuinely distinct ancient authors, by an author skilled at varying register, or by artifacts of the method itself is precisely what remains in dispute — which is why this is presented as evidence to weigh, not proof to wield.

A Researcher’s Summary

One blogger who undertook a personal review of the stylometric evidence — writing for the LDS-themed site Rational Faiths — summarized his own reaction this way:

Personal Conclusion

“One author didn’t write the Book of Mormon. Two didn’t, either. And we don’t have anything else written by the people who did. For me, this is fact. Explain it how you will.”

— from “Book of Mormon Stylometry in Pictures and Tables,” Rational Faiths. This is one researcher’s personal assessment, offered as such — not a scholarly consensus.

Verdict

Stylometry does not prove the Book of Mormon is the word of God, and it does not prove Joseph Smith wrote it alone. What the 1980 BYU Studies wordprint analysis reported is that 24 internal authors, tested by three statistical methods, came back as distinguishable from one another and from the leading nineteenth-century candidates. That is a meaningful and intriguing result.

But the method is genuinely contested. Critics from the 1980s onward have questioned whether wordprints can separate authors as cleanly as claimed, and a 2008 study reached the opposite conclusion. The right posture is intellectual honesty: the wordprint evidence is suggestive of internal authorial variety, the critiques are substantive, and the question is not closed.

Anyone — believer or critic — who claims stylometry has decisively settled Book of Mormon authorship has gone beyond what the data can bear. The evidence is worth examining carefully and reporting accurately. The question is whether you will follow it where it actually leads, rather than where you wish it would.

Further reading: Larsen, Rencher & Layton, “Who Wrote the Book of Mormon? An Analysis of Wordprints” (BYU Studies, 1980) · Hilton, “On Verifying Wordprint Studies” (BYU Studies, 1990) · Schaalje et al., “Extended Nearest Shrunken Centroid Classification” (LLC, 2011) · Holmes, “A Stylometric Analysis of Mormon Scripture” (JRSS, 1992) · Scripture Central, “What Can Stylometry Tell Us about Book of Mormon Authorship?”