Bible Textual Criticism

---
name: bible-textual-criticism
description: "Use when the question is what the biblical text actually says and how anyone knows — lower vs higher criticism, how copying errors happen (haplography, homoioteleuton, itacism, harmonization), external and internal evidence and the harder and shorter reading canons after Royse, NT textual criticism from Erasmus and the Textus Receptus through Westcott and Hort to NA28, NA29 and the Editio Critica Maior, how many Greek manuscripts there really are, the key witnesses, text-types, the Coherence-Based Genealogical Method and the initial text, how much is actually uncertain, the famous variant passages (Mark 16:9–20, John 7:53–8:11, the Comma Johanneum, John 1:18), positions in the field from reasoned eclecticism to Byzantine priority and the TR, Hebrew Bible pluriformity with the Dead Sea Scrolls, Samaritan Pentateuch and Septuagint, diplomatic vs eclectic editions (BHQ, HUBP, HBCE), and the digital tools including Codex H and the Enoch dating model. Skill 4 of 9 of the From Autograph to Your Bible reference."
---

# How Textual Criticism Works

> **Skill 4 of 9** of the *From Autograph to Your Bible: A Deep Dive into the Biblical Languages, Translation, Textual Criticism, Archaeology, and Historical Method* reference
> (plugin `biblical-languages-texts-archaeology-and-history`), carrying source Part III, *How Textual Criticism Works*: §8–§13 — copying errors, the reasoning, NT and Hebrew Bible textual criticism, the CBGM and the digital tools. Sibling skills:
> `bible-hebrew-aramaic-and-greek` (§1–§2 — Biblical Hebrew and Aramaic, Koine Greek and the Septuagint),
> `bible-latin-german-and-english` (§3–§5 — the Latin, German and English Bibles and the registers they created),
> `bible-how-translation-is-done` (§6–§7 — major-language committee translation and minority-language field translation),
> `bible-archaeology` (§14–§22 — fieldwork, stratigraphy, dating, the science toolbox, epigraphy, the big debates and the 2026 finds),
> `bible-history-and-historical-method` (§23–§28 — the historian's method, ancient history, the historical-critical toolkit, the historical Jesus and reading a consensus),
> `bible-forgeries-looting-and-provenance` (§29–§31 — why provenance matters, the forgery and looting case files, and archaeology and politics),
> `bible-three-verses-through-the-pipeline` (§32–§34 — Isaiah 7:14, Deuteronomy 32:8 and Luke 2:1–2 through every discipline),
> `bible-reference` (front matter, §35–§38 and sources — how to use the reference, learning paths, free tools, the glossary, the open questions and the sources consulted).
>
> Section numbers are **the source's own and shared across the whole set**: a reference written
> as §N → `skill` points into that sibling skill.

> **Confidence markers are the source's own**, and this skill keeps every one where the source
> put it: **\[settled\]** broad consensus across confessional and secular scholarship ·
> **\[majority\]** the mainstream position, with a live minority dissent worth knowing ·
> **\[contested\]** genuinely disputed among competent specialists; no consensus ·
> **\[as of 2026\]** a status fact likely to change (editions in progress, statistics, new
> finds). The source was compiled on September 30, 2026, and its \[as of 2026\] facts were
> checked against live sources that day. They have not been re-checked for this pack — treat
> each as dated, and verify it before relying on it.

> **Pack orientation** (the pack's words, not the source's): Part III of the source, whole: the discipline that decides which words a translator has to
> translate. It covers the New Testament and the Hebrew Bible separately, because, as §12.1
> puts it, the Hebrew Bible is a different problem with different witnesses.

## §8 The basic problem

We have **no autographs** (original documents) of any biblical book — nor of virtually any ancient literary work. We have copies of copies, made by hand for well over a thousand years before printing. Every hand copy introduces changes. **Textual criticism** is the discipline of collecting the surviving witnesses, classifying their differences (**variants**), and reasoning back toward the earliest recoverable text — while also studying the history of the text as a subject in its own right.

It is also called **"lower criticism"** (establishing the text), as distinct from **"higher criticism"** (questions of authorship, sources, date, and history). "Lower" means foundational, not inferior.

## §9 How copying errors happen

### §9.1 Unintentional changes (the great majority)

| Error | What happens | Example pattern |
| --- | --- | --- |
| **Haplography** | Writing once what should be written twice | Skipping a repeated letter or word |
| **Dittography** | Writing twice what should be written once | Duplicated syllable or line |
| **Homoioteleuton / homoioarcton** | The eye jumps from one ending (or beginning) to an identical one further down, omitting everything between | Very common cause of omitted lines |
| **Parablepsis** | General "looking aside" error |  |
| **Letter confusion** | Similar shapes: Hebrew ד/ר (d/r), ו/י (w/y), ב/כ; Greek uncial Ε/Θ/Ο/Σ | 2 Samuel vs 1 Chronicles name differences |
| **Itacism** | In later Greek, several vowels and diphthongs came to sound alike (η, ι, ει, οι, υ) | "we" (ἡμεῖς) vs "you" (ὑμεῖς) — dozens of variants |
| **Metathesis** | Transposed letters or words |  |
| **Mishearing** | When copying by dictation |  |
| **Wrong word division** | Early Greek manuscripts were written without spaces (*scriptio continua*) |  |

### §9.2 Intentional changes

Scribes also changed texts on purpose, usually thinking they were *correcting* them: **harmonization** (making Mark match Matthew, or a quotation match the Septuagint), **clarification** (adding "Jesus" where the text says "he"), **grammatical smoothing**, **conflation** (combining two known readings), liturgical additions (the doxology at the end of the Lord's Prayer in Matthew 6:13), and occasionally **theologically motivated** changes. Bart Ehrman's *The Orthodox Corruption of Scripture* (1993) argued that proto-orthodox scribes altered texts in christological controversies; critics (e.g., Daniel Wallace, Michael Holmes) accept that some such changes happened but dispute how many and how significant. **\[contested in degree\]**

## §10 The reasoning: external and internal evidence

Textual critics weigh two kinds of evidence at every **variation unit** (a place where manuscripts differ):

### §10.1 External evidence (the witnesses)

- **Date and quality** of the manuscripts (though an early manuscript can carry a bad text, and a late manuscript can preserve an early one).
- **Geographical distribution** — independent support from Egypt, the West, and the East is stronger than support from one region.
- **Genealogical relationships** — witnesses are **weighed, not counted**. A thousand copies of one exemplar are really one witness.
- In NT criticism: **Greek manuscripts**, **early versions** (Latin, Syriac, Coptic, and others), and **patristic quotations**.

### §10.2 Internal evidence (the reasoning about the text)

**Transcriptional probability** (what would a scribe likely have done?):

- *Lectio difficilior potior* — the **harder reading is preferable**, because scribes tended to smooth difficulties, not create them. (The rule has limits: a reading can be "hard" because it's nonsense produced by error.)
- *Lectio brevior potior* — the **shorter reading is preferable**, because scribes tended to add. This old canon (Griesbach) was significantly qualified by **James Royse's** study of early papyri (2008), which found early scribes more often **omitted** than added. **\[majority, post-Royse\]**
- **The reading that best explains the others** — the master principle. If reading A explains how B and C arose, but not vice versa, A is likely earlier.

**Intrinsic probability** (what would the *author* likely have written, given style, vocabulary, theology, and context?).

## §11 New Testament textual criticism: the story

| Date | Milestone |
| --- | --- |
| 1516 | **Erasmus'** Greek NT (*Novum Instrumentum*), based on a handful of late manuscripts. For the last six verses of Revelation his manuscript was defective, so he back-translated from Latin. |
| 1550–51 | **Robert Estienne (Stephanus)** prints an edition with variant notes (1550), then introduces the **verse numbers** we still use (1551). |
| 1633 | The Elzevir edition's preface advertises a text "now received by all" — the origin of the label **Textus Receptus** (TR). |
| 1707 | **John Mill** lists roughly 30,000 variants — shocking contemporaries. |
| 18th c. | Bengel (weigh manuscripts by families), Wettstein (manuscript sigla), Griesbach (canons of criticism) |
| 1831 | **Karl Lachmann** publishes the first major edition that abandons the TR for a text built from the oldest witnesses. |
| 1844/1859 | **Constantin von Tischendorf** acquires leaves (1844) and then the main body (1859) of **Codex Sinaiticus** from St. Catherine's Monastery, Sinai — a controversial story still disputed between the monastery and the modern holders. |
| 1881 | **Westcott & Hort**, *The New Testament in the Original Greek*: the "text-type" theory (Neutral/Alexandrian, Western, Syrian/Byzantine), strong preference for Vaticanus and Sinaiticus; the basis of the Revised Version. |
| 1898 | **Eberhard Nestle** publishes the *Novum Testamentum Graece*, originally a majority text of the three leading 19th-century editions. |
| 20th c. | Papyri revolution: **Chester Beatty** papyri (P45, P46, P47) in the 1930s; **Bodmer** papyri (P66, P72, P75) in the 1950s. |
| 1959 | **Kurt Aland** founds the Institute for New Testament Textual Research (**INTF**) in Münster. |
| 1966 | United Bible Societies' *Greek New Testament* (UBS1) — designed for **translators**, with a letter-grade {A}–{D} confidence rating per variant. |
| 1979/1993 | NA26 / NA27 (same text as UBS3/UBS4). |
| 1997– | **Editio Critica Maior** (ECM) begins (Catholic Letters first), using the **CBGM**. |
| 2012 | **NA28**: revised apparatus; Catholic Letters text changed in just over 30 places to follow the ECM. |
| 2017 / 2021 / 2024 | ECM volumes for Acts (2017), Mark (2021), and Revelation (2024). |
| **In preparation** | **NA29** — see below. |

### §11.1 Where things stand in 2026 **\[as of 2026\]**

- **NA28 (2012)** remains the current standard critical text.
- **NA29** is being prepared by an editorial committee appointed by the United Bible Societies. The Deutsche Bibelgesellschaft says it will adopt all of the ECM's textual changes in Mark (ECM 2021), Acts (2017), and Revelation (2024), thoroughly revise the apparatus in those books, adjust the witnesses cited in Paul based on ECM preparatory work, and add newly accessible papyri. No release date has been announced in the sources checked.
- The INTF states its aim is to have the complete ECM published **by 2030**.

### §11.2 How many manuscripts? (The honest answer)

The official register is the **Kurzgefasste Liste** maintained by INTF. Each witness gets a **Gregory-Aland (GA)** number in four categories:

- **Papyri** (𝔓1, 𝔓2...) — written on papyrus; the earliest.
- **Majuscules** (01, 02...; also known by letters: 01 = א Sinaiticus, 02 = A Alexandrinus, 03 = B Vaticanus, 05 = D Bezae) — capital letters on parchment.
- **Minuscules** (1, 2, 3...) — lowercase cursive script, from the 9th century on; the bulk of manuscripts.
- **Lectionaries** (ℓ1, ℓ2...) — arranged by liturgical readings.

In a September 2023 INTF post, Katie Leggett and Greg Paulson gave the then-current count of Liste entries as **about 5,700**, while explaining why "how many manuscripts" is a trickier question than it seems: one GA number can be a scrap with a few verses or a codex of 300+ leaves; one physical book can carry several GA numbers; scattered leaves share one number; palimpsests put two witnesses on the same page; some manuscripts are destroyed, missing, or in untraceable private hands; and the Liste is regularly purged of duplicates. They recommended asking "how many witnesses have been catalogued" instead, and showed that removing destroyed and missing items leaves about **5,541**. **Rounded popular figures like "5,800+" come from older counts or different counting rules.**

Add to that roughly 10,000+ Latin manuscripts, thousands in Syriac, Coptic, Armenian, Georgian, Ethiopic, Slavonic, and others, plus a vast body of patristic quotations.

### §11.3 The key witnesses

| Witness | Date (approx.) | Notes |
| --- | --- | --- |
| **𝔓52** (John 18) | Traditionally c. 125; many now argue for a broader 2nd–early 3rd century range (Brent Nongbri, 2005) | Popularly "the earliest NT fragment"; palaeographic dating ranges are wide **\[contested\]** |
| **𝔓46** | c. 200 | Earliest substantial Pauline collection (includes Hebrews among Paul's letters) |
| **𝔓66**, **𝔓75** | c. 200 / 3rd c. (some argue later) | John and Luke–John; 𝔓75 is remarkably close to Vaticanus, supporting the antiquity of that text form |
| **Codex Vaticanus (B, 03)** | c. 325–350 | Greek Bible; generally the most valued single NT witness |
| **Codex Sinaiticus (א, 01)** | mid-4th c. | Oldest complete NT; also includes Barnabas and Hermas |
| **Codex Alexandrinus (A, 02)** | 5th c. | Mixed text; Byzantine in the Gospels |
| **Codex Bezae (D, 05)** | c. 400 | Greek–Latin bilingual; the chief "Western" witness, with a dramatically different Acts (\~8–10% longer) |
| **Codex Washingtonianus (W, 032)** | 4th–5th c. | Contains the "Freer Logion" addition in Mark 16 |

### §11.4 Text-types and their critics

The older model grouped witnesses into **text-types**: **Alexandrian** (Egyptian; generally prized), **Western** (paraphrastic tendencies), **Byzantine** (the majority of later manuscripts; smooth, full), and sometimes **Caesarean** (now largely abandoned). Modern specialists increasingly treat these as **clusters on a spectrum** rather than discrete recensions, and the CBGM (below) was designed to work without them.

### §11.5 The Coherence-Based Genealogical Method (CBGM) — the method reshaping the text

Developed by **Gerd Mink** at Münster, the CBGM tackles the core problem of NT transmission: **contamination**. Scribes corrected manuscripts against multiple exemplars, so a simple family tree (a stemma) of manuscripts is impossible — any witness can have readings both older and younger than another witness.

The CBGM's key moves:

1. **Treat texts, not manuscripts.** The unit is the *state of the text* a manuscript carries (a "witness"), which may be far older than the physical manuscript.
2. **Collate everything** in full transcriptions and record every variation unit.
3. **Pre-genealogical coherence.** Compute the percentage agreement between every pair of witnesses across all variation units. This is fairly objective: it's just counting agreement.
4. **Local stemmata.** At each variation unit, editors decide — using traditional internal and external criteria — which reading gave rise to which (A → B, A → C, etc.). This is where human judgment enters.
5. **Genealogical coherence.** Aggregate thousands of local decisions to find, for each witness, which others are its **potential ancestors** (witnesses that more often have the prior reading where the two differ).
6. **Textual flow diagrams** show how a given reading moved through the tradition. If a reading's supporters are incoherent (scattered across unrelated witnesses), it probably arose more than once independently — a clue that it's a natural scribal change.
7. **Iterate.** Where a local stemma implies implausible genealogy, editors revisit it. The process loops.
8. **Global stemma.** A simplified overall picture of relationships.

**What the CBGM is not:** a machine that spits out the text. Its developer and INTF describe it as a tool for testing editorial hypotheses; human editors still define readings, build local stemmata, weigh internal evidence, and decide what to print. **Critics** have raised methodological objections (e.g., Stephen Carlson on potential bias; debates over pre-genealogical coherence; the incomplete integration of versional and patristic evidence), and INTF scholars have published replies. **\[method majority-adopted; details contested\]**

**Two important consequences:**

- The ECM marks some places with a **split guiding line** (a diamond ◆ in the ECM/NA), where editors could not decide between two readings — an explicit admission of uncertainty. This increased the number of acknowledged "undecided" places.
- The goal was reframed from "the original text" to the **initial text** (*Ausgangstext*): the hypothetical text from which the extant tradition descends. Whether the initial text should be identified with the authorial text — and whether "original text" is even a coherent target, given ancient publication practices — is a live debate (Eldon Epp, Michael Holmes, David Parker's "living text"). **\[contested\]**

### §11.6 How much is actually uncertain?

- Estimates for the number of NT variants run around **400,000–500,000+** across all Greek manuscripts — more variants than words in the NT (\~138,000). The number is huge **because** the manuscript base is huge; more copies mean more copying differences.
- The overwhelming majority are spelling, word order, the Greek *movable nu*, the article with names, and obvious slips — meaningless in translation. A small minority are both meaningful and viable.
- Popular claims like "99% certainty" are advocacy summaries, not measured statistics; what can fairly be said is that the wording of the NT is known with **unusually high confidence by ancient standards**, and that the uncertain places are **mapped and footnoted**, not hidden. The public Ehrman–Wallace debates are the best-known clash over how to characterize this.

### §11.7 The famous variant passages

| Passage | Issue | Where it stands |
| --- | --- | --- |
| **Mark 16:9–20** ("longer ending") | Absent from Sinaiticus and Vaticanus; other endings exist ("shorter ending"; Freer Logion) | Most scholars think Mark originally ended at 16:8 or that its ending is lost; the longer ending is ancient (2nd c.) but likely secondary **\[majority\]**. Translations print it with brackets or notes. |
| **John 7:53–8:11** (woman caught in adultery) | Absent from the earliest manuscripts; floats in others (after Luke 21:38, after John 21) | Almost universally judged not original to John, though possibly an authentic early tradition **\[settled that it's not Johannine\]** |
| **1 John 5:7–8** (Comma Johanneum) | Trinitarian phrase found in Latin tradition; appears in only a handful of late Greek manuscripts, some of them clearly back-translated | Omitted by virtually all critical texts **\[settled\]**; Erasmus added it in his 3rd edition (1522) under pressure, reportedly after a manuscript (Codex Montfortianus) surfaced containing it |
| **Luke 22:43–44** (sweat like drops of blood) | Present in some early witnesses, absent from others (including 𝔓75, B) | Genuinely disputed **\[contested\]** |
| **John 1:18** | "the only God" (μονογενὴς θεός, 𝔓66, 𝔓75, א, B) vs "the only Son" (μονογενὴς υἱός) | Most critical texts print *theos*; translations divide, often with a note |
| **Mark 1:1** | "Son of God" present or absent | Bracketed in NA28 |
| **Hebrews 2:9** | "by the grace of God" (χάριτι θεοῦ) vs "apart from God" (χωρὶς θεοῦ) | Classic *lectio difficilior* debate |
| **Acts (the "Western" text)** | Codex Bezae's significantly longer and different Acts | A textbook case of two text-forms of one book |

### §11.8 Positions in the field

- **Reasoned eclecticism** — weigh external and internal evidence case by case (the mainstream; NA/UBS/ECM). **\[majority\]**
- **Thoroughgoing eclecticism** — rely almost entirely on internal evidence (G. D. Kilpatrick, J. K. Elliott).
- **Byzantine priority** — the majority text form is closest to the original (Maurice Robinson & William Pierpont's edition). **\[minority\]**
- **Textus Receptus / Confessional Bibliology** — the TR (as received by the Reformation churches) is providentially preserved; associated with KJV-preference and "KJV-only" movements. **\[minority; theological rather than purely text-critical\]**
- **Conjectural emendation** — adopting a reading found in no manuscript. Rare in NT criticism (the Nestle-Aland text prints a conjecture at Acts 16:12), much more common in Hebrew Bible criticism. INTF hosts an online database of NT conjectures.

## §12 Hebrew Bible textual criticism

### §12.1 Different problem, different witnesses

The Hebrew Bible situation differs sharply from the NT: the medieval Masoretic manuscripts are remarkably uniform (the Masoretes' counting systems worked), but they are late (10th–11th century), and earlier evidence shows the text was once **plural**.

The witnesses:

- **Masoretic Text (MT)** — the Tiberian manuscripts (Aleppo, Leningrad, etc.).
- **Dead Sea Scrolls** — biblical manuscripts from Qumran and other Judean Desert sites (3rd c. BCE – 2nd c. CE). Some closely match the later MT (proto-MT); others align with the Septuagint's Vorlage or the Samaritan Pentateuch; others are independent. **Emanuel Tov** and **Eugene Ulrich** have been central to describing this pluriformity; Ulrich describes "multiple literary editions" of some books.
- **Samaritan Pentateuch** — a separate Hebrew tradition of the Torah in Samaritan script; around 6,000 differences from MT, most trivial, but it agrees with the LXX against MT in a large number of places and contains some deliberate sectarian changes (e.g., Mount Gerizim).
- **Ancient translations:** Septuagint (the most important), Aramaic Targums, Syriac Peshitta, Latin Vulgate.

### §12.2 Striking Hebrew Bible variants

| Passage | Variants | Significance |
| --- | --- | --- |
| **Deuteronomy 32:8** | MT: "according to the number of the **sons of Israel**"; 4QDeut^j and LXX: "**sons of God**" | Most scholars see the scroll reading as earlier, reflecting a divine-council worldview; ESV and NRSVue follow it **\[majority\]** |
| **1 Samuel 10:27–11:1** | 4QSam^a contains a paragraph about Nahash the Ammonite gouging out right eyes, absent from MT; Josephus knew it | The NRSV (1989) added it to its main text |
| **1 Samuel 17:4** | Goliath's height: MT "six cubits and a span" (\~9'9"); 4QSam^a and LXX "four cubits and a span" (\~6'9") | A reminder that "the Bible says" depends on which text |
| **Jeremiah** | LXX (and 4QJer^b,d) about one-eighth shorter than MT, different chapter order | Evidence of two literary editions |
| **Psalm 145** | An acrostic missing its *nun* line in MT; present in 11QPs^a and LXX | A textbook example of a lost line recovered |
| **Psalm 22:16 \[17\]** | MT "like a lion (*ka'ari*) my hands and feet"; LXX and some Hebrew witnesses "they dug/pierced" | Famous for its Jewish–Christian polemical history |

### §12.3 Three editorial philosophies for the Hebrew Bible **\[as of 2026\]**

| Project | Base | Method |
| --- | --- | --- |
| **Biblia Hebraica Quinta (BHQ)** | Leningrad Codex | **Diplomatic** — print one real manuscript; put variants and editorial judgments in the apparatus and commentary |
| **Hebrew University Bible Project (HUBP)** | Aleppo Codex | **Diplomatic**, with a far fuller multi-apparatus documentation |
| **The Hebrew Bible: A Critical Edition (HBCE)** (formerly the Oxford Hebrew Bible), led by Ronald Hendel | Reconstructed | **Eclectic** — print the reconstructed earliest inferable text with each choice justified; defended in Hendel's *Steps to a New Edition of the Hebrew Bible* (2016) |

The critique of the eclectic approach is that it prints a text that "never existed"; the critique of the diplomatic approach is that it privileges one medieval manuscript. Some propose a third option (synoptic editions presenting parallel text forms). **\[contested\]**

## §13 Digital tools and new science

- **NTVMR** (New Testament Virtual Manuscript Room, Münster): images, transcriptions, the Liste, the ECM online.
- **Codex Sinaiticus Project**: the whole manuscript digitized and reunited online across its four holding institutions.
- **Leon Levy Dead Sea Scrolls Digital Library** (Israel Antiquities Authority): high-resolution multispectral images of the scrolls.
- **Multispectral imaging** recovers erased undertexts in palimpsests. In spring 2026 a University of Glasgow–led team announced recovery of 42 lost pages of **Codex H**, a sixth-century manuscript of Paul's letters taken apart on Mount Athos in the 13th century and recycled as binding material, now scattered across libraries in Italy, Greece, Russia, Ukraine, and France; the recovered text preserves early chapter divisions of Paul's letters that differ from ours. **\[as of 2026\]**
- **AI and radiocarbon dating:** In June 2025, a University of Groningen–led team (Mladen Popović) published *Enoch*, a machine-learning model trained on radiocarbon-dated Dead Sea Scroll samples and handwriting features, in *PLOS One*. Results suggest many scrolls are **older than previously thought**, revising the development of the "Hasmonaean" and "Herodian" scripts; the team argues two biblical fragments (4Q114 Daniel, 4Q109 Qohelet) date to roughly the time of their books' presumed composition. Experts judged Enoch's estimates "realistic" in 79% of tested cases. **\[new; will be debated\]**
- **Computational stemmatology and phylogenetics** borrowed from evolutionary biology are being tested on textual traditions.

## Where to go next (pack navigation, not source text)

- **What committees do with these decisions** — §6.2 and §6.4(a) → `bible-how-translation-is-done`.
- **The manuscripts as physical, dated, provenanced objects** — dating in §17 → `bible-archaeology`; forged fragments and Codex Sinaiticus's removal in §30 → `bible-forgeries-looting-and-provenance`.
- **Integrity as one question of source criticism** — §25.3 → `bible-history-and-historical-method`.
- **A variant worked through end to end** — §33 (Deuteronomy 32:8) → `bible-three-verses-through-the-pipeline`; §32 shows a case where the text is secure.
- **Introductions and handbooks** — §35.2 → `bible-reference`; open questions 4–6 and 11 in §38.
- **A neighbouring pack reads the manuscript argument differently.** `evangelism-apologetics-and-church-planting` (§8 of that pack) treats an "over 99%" wording-reconstruction figure as defensible as stated — and as indefensible the moment it is stretched; this source (§11.6) calls popular "99% certainty" claims advocacy summaries rather than measured statistics. That pack also gives "over 5,800" Greek manuscripts, the kind of rounded popular figure §11.2 here says comes from older counts or different counting rules. Both are carried as their sources give them; neither pack arbitrates the other.