Contamination (textual criticism)

In textual criticism, contamination refers to the phenomenon in which a manuscript or witness incorporates readings from multiple source texts, rather than following a single line of transmission. This undermines models that assume a single line of descent and complicates efforts to reconstruct an archetype.

Contamination

Descent of the manuscripts of Pseudo-Apuleius Herbarius
Field	Textual criticism
Origin	19th-century classical philology
Key people	Paul Maas, Giorgio Pasquali, Joseph Bédier
Purpose	Mixing of readings from different exemplars within one witness or transmission line

German classical philologist Paul Maas warned that there is "no remedy against contamination." ^[1] Since the late twentieth century, editors have treated contamination as common across Greco‑Roman, biblical, medieval, and non‑Western traditions, and they have adopted network or local geneaological models that can register intersecting lines of descent.^[2][3]

In New Testament studies the phenomenon is often called mixture. The development of phylogenetic and coherence‑based approaches has provided practical strategies for analyzing contamination without assuming a single global stemma.^[4][5]

Overview

Scribe's palette with styluses and residues of colors, from the Tomb of Kha and Merit. Between 1425 and 1353 BC (New Kingdom of Egypt). Museo Egizio, Turin.

Editors use contamination to describe any non‑vertical transmission in which a scribe, corrector, or editor draws on multiple sources, or transfers material from paratexts and parallels into the main text. Non‑vertical transmission occurs when textual material comes from multiple sources rather than a single line of descent. This includes cross‑contamination between different manuscript branches, borrowing from parallel texts, or incorporation of external material. Contamination can occur at the level of a single reading, a clause or sentence, or larger units. It includes both unintentional processes, such as harmonization from memory, and deliberate editorial practices, such as conflation. Because contamination violates the assumption of strict descent, it challenges classical stemmatics and favors network or local approaches that acknowledge intersecting ancestry.^[1][2][3]

Titivillus, a demon said to introduce errors into the work of scribes, besets a scribe at his desk (14th-century illustration).

Common contamination mechanisms include:

• Scribal mixing of exemplars, where a scribe may consult two exemplars alternately, or correct a primary copy against a secondary copy. The result is a witness whose ancestry varies by passage and whose placement in a single stemma is unstable.^[6]

• Memory based corrections and parallel texts, when copyists reproduced from memory or harmonized familiar wording across parallels. In the Greek New Testament, harmonization among the Synoptic Gospels is well attested and often reflects memory based substitution.^[7][8]

• Migration of marginalia into the main text, in which interlinear glosses, marginal notes, and lectionary marks can be copied as if they were authorial. Classical and biblical editors document many cases of paratext moving into the text.^[9][10] Research on the Comma Johanneum traces a Trinitarian gloss that circulated in margins before entering the main text in later transmission.^[11]

• Editorial collation and conflation, where edditors and revisers sometimes join divergent readings to preserve both. B. F. Westcott and F. J. A. Hort listed characteristic "conflate readings" in what they called the Syrian or Byzantine text. They used such readings as signs of later editorial activity.^[12]

While classical philologists traditionally speak of contamination, New Testament textual critics more commonly refer to the same phenomenon as mixture, reflecting distinct scholarly vocabularies develping in parallel, but separate fields of study.^[6]

Contamination differs from:

• Conflation, an intentional combination of two readings into one new reading.^[12]
• Harmonization, adjustment toward a parallel passage or familiar formula.
• Polygenesis, independent emergence of the same reading in different branches.
• Authorial revision, change by the author or authorized redactor. This is not contamination, but it can produce textual plurality that resembles contamination in transmission.^[13]

History

German classical philologist Paul Maas made contamination a central obstacle for Lachmannian practice in his 1958 work and wrote that there is "no remedy" when cross‑copying is pervasive.^[1] British classicist M. L. West set out practical criteria for identifying glosses, parallel intrusions, and secondary expansions in 1973, and he documented cases where marginal notes entered the text in Greek and Latin authors.^[9] Italian philologist Giorgio Pasquali argued in 1934 that many traditions are tradizioni aperte, or open traditions, marked by "horizontal contacts." He therefore recommended methods that do not presuppose a closed recensio.^[2]

French medievalist Joseph Bédier criticized routine stemmatic reconstruction and observed that editors often produced "arbres à deux branches", a warning about methodological circularity and issues with human collation increasing opportunities to blend sources.^[14] Later scholarship interpreted his critique as a call for caution rather than an endorsement of a perpetual best‑manuscript policy.^[15]

Italian classical scholar Sebastiano Timpanaro showed that "Lachmann's method" evolved historically and analyzed the strengths and limits of eliminatio in the face of contamination and lost intermediaries.^[15] Contemporary scholars have increasingly embraced methods that combine traditional stemmatic analysis with statistical and digital tools, recognizing that textual transmission often follows complex, interconnected pathways rather than simple tree-like descent.^[3]

Methods

Lachmannian

Lachmannian recensio groups witnesses by shared errors and seeks to eliminate direct copies of extant exemplars. Contamination breaks the assumption of single ancestry and yields contradictory signals across different passages. Maas's skepticism, Bédier's critique of "bipartite" stemmata, and Pasquali's insistence on open traditions define methodological limits to a single global stemma for many corpora.^{[1][14][2][15]}

Variant unit	Witness A	Witness B	Witness C	Analysis
v1	error ε1	error ε1	reading r	A and B share ε1, they look related
v2	reading s	error ε2	error ε2	B and C share ε2, they look related
v3	reading t	reading t	error ε3	C has ε3 alone, grouping is unstable

A and B appear related at v1, then B and C at v2. No single tree explains all three units without invoking contamination or coincidence.

Genealogical and cladistic

To address mixture, phylogenetic and network methods from biology have been adapted to texts. A 1998 study in Nature used shared variants to infer a "phylogeny of The Canterbury Tales."^[16] For the Letter of James, reduced‑median networks made reticulation explicit by drawing crossing connections among manuscripts.^[17] Phylogenetic network theory explains why reticulate graphs, not trees, are appropriate when horizontal transmission is present and warns about issues such as long‑branch attraction.^[18][4]

Coherence-based genealogical method

The Coherence-Based Genealogical Method models relationships locally at each variant unit, then infers global relationships among witnesses by aggregating "local stemmata of variants." The method quantifies genealogical coherence, allows for coincidental agreement, and identifies potential ancestors even when mixture is likely.^[19][5][20] Advocates report "considerable control" over complex genealogies when contamination is pervasive.^[6] Critics argue that pregenealogical choices can introduce "a bias at the heart" of the method.^[21]

Quantitative

Statistical analysis tools have transformed contamination studies by enabling large-scale collation or pattern detection across thousands of variants, quantifying degrees of mixture that were previously assessed only impressionistically, and providing objective measures of genealogical coherence that help distinguish genuine relationships from coincidental agreement. Tools include CollateX for automatic alignment and graph‑based comparison and Stemmaweb for hypothesis building and visualization.^[22][23] The stemmatology R package implements clustering, bipartite graphs of witnesses and variants, and disagreement‑based methods.^[24][25] Simulation studies benchmark accuracy under different contamination regimes and report sensitivity to alignment error and character dependence.^[26][18]

Open resources used for contamination studies include the New Testament Virtual Manuscript Room and CBGM workspaces at Münster, the Editio Critica Maior datasets for Catholic Epistles and Acts, Open Greek and Latin corpora, and the Chinese Text Project.^{[20][27][28][29]}

Examples

Greek New Testament manuscript tradition

Codex Regius (019) Mark

Editors have long described mixture among New Testament witnesses. Westcott and Hort cataloged "conflate readings" in the Byzantine tradition and treated them as signs of editorial activity.^[12] Later work reframed the Byzantine text as a process rather than a single recension and traced early phases by analysis of codices such as Alexandrinus and the Purple Codices.^[30] Network analysis of the Letter of James produced reticulate graphs rather than trees.^[17] The CBGM, developed by scholars at the University of Münster, provides a systematic approach to understanding local relationships between manuscripts, tracing potential textual ancestors, and measuring genealogical coherence as editors work through the complex mixture found in the Catholic Epistles, Acts, and John.^[20][19][5] Swedish New Testament scholar Tommy Wasserman demonstrated this approach through a detailed analysis of Mark 1:1, showing how editors can meaningfully connect "historical and philological" insights with the quantitative results that emerge from CBGM analysis.^[31]

Classical Latin literature

Scythians at the Tomb of Ovid c. 1640

Latin traditions often show cross‑copying and mixed descent. West describes common intrusions such as gloss migration and proposes cautious emendation when contamination is suspected.^[9] For Ovid's Ibis, the earliest stratum already shows widespread mixture, which complicates any global stemma.^[32] The tradition of Catullus is likewise notable for cross‑copying in medieval witnesses used by modern editors.^[33]

Medieval vernacular works

Joseph Bédier

Bédier's analysis of the Lai de l'Ombre catalyzed long debate about reconstructed stemmata in Old French narrative and about the frequency of contamination.^[14] For Chaucer's The Canterbury Tales, machine‑collated variation and phylogenetic analysis visualized clusters and crossing lines indicating mixture.^[16]

Sanskrit epics

Kurukshetra

Indian Sanskritist V. S. Sukthankar argued in the Prolegomena to the critical Mahabharata that contamination is structural in epic transmission. He recommended local evaluation in the presence of "polygenetic" readings.^[34] American Indologist Robert P. Goldman described analogous recensional variety and cross‑copying in the Ramayana, with editorial procedures that register diffusion across regional traditions.^[35]

Quranic manuscripts

Comparison of a 20th-century edition of the Quran (left) and the Birmingham Quran manuscript (right)

French Arabist François Déroche documented early corrections, marginal supplements, and comparison across copies in Umayyad Qurʾans, demonstrating composite profiles.^[36] Iranian‑American scholar Behnam Sadeghi and Mohsen Goudarzi analyzed the Ṣanʿāʾ palimpsest and argued that early transmission shows "textual plurality" with mixing across lines.^[37]

Rabbinic literature

11th-century Mishnah codex, Biblioteca Palatina, Parma

Israeli philologist J. N. Epstein analyzed the Mishnah across its principal codices and described cross‑fertilization in the medieval and early modern period, which produced mixed texts in printed editions.^[38] Hebrew textual critic Emanuel Tov has emphasized multiple streams of transmission for biblical and post‑biblical texts, which provides a comparative framework for mixture in rabbinic corpora.^[39]

Chinese classics

An 1827 version of the Kangxi Dictionary

Work on early Chinese texts describes collational practices and paratextual migration across editions in traditional scholarship, known as kaozheng. American sinologist Endymion Wilkinson shows how traditional Chinese scholars approached textual problems through careful collation and error analysis, observing that marginalia and parallel passages often migrated between editions, creating layered witnesses that reflect centuries of scholarly engagement with these foundational texts.^[40]

Debates and theories

While some readings arise once and spread through copying, others can emerge independently in different branches, leading CBGM literature to distinguish multiple emergence and coincidental agreement from genealogical dependence using coherence metrics, and prompting editors across fields to recommend passage‑by‑passage evaluation rather than assuming either monogenesis or polygenesis in advance.^{[19][31][5][3]}

Pasquali's tradizione aperta describes archetypes that are not closed, since copyists can import readings horizontally.^[2] Late twentieth‑century editorial theory reframed textual stability under the rubric of the "fluid text." This view explains why contamination is not always textual "pollution", since intersecting witnesses can preserve readings otherwise lost.^[13][41]

Modern practice builds local stemmata for each variant unit and tests points of contact iteratively, a procedure made explicit by German New Testament textual critic Klaus Wachtel and German computer scientist Gerd Mink in the CBGM, which allows editors to visualize repeated cross‑copying and its recursive effects on the tradition.^[6][19]

When contamination appears likely, editors step back from sweeping genealogical claims and embrace more nuanced approaches: careful local analysis, comprehensive collation, and transparent documentation of their reasoning. They learn to recognize contaminated witnesses, trace how marginal notes migrated into main texts, and distinguish between scribal conflations and genuine authorial revision or deliberate editorial synthesis. While digital tools and coherence‑based methods offer editors "considerable control" over even the most tangled transmission histories, these approaches demand intellectual honesty about the assumptions underlying editorial choices and an acceptance that multiple plausible reconstructions may coexist.^{[9][10][6][5]} In traditions where cross‑copying runs deep, some editors choose to honor textual complexity by publishing parallel versions or synoptic apparatuses that preserve competing forms rather than forcing them into a single, artificially unified text.^[13][3]

See also

• Stemmatics
• Textual criticism
• Lachmann's method
• Editio Critica Maior
• Critical apparatus