The One with the CBGM

A basic guide to what the CBGM is and why it matters for New Testament textual criticism—and beyond

The Coherence-Based Genealogical Method, or CBGM, has become one of the most important and most frequently misunderstood developments in New Testament textual criticism in the last several decades. It is central to the editorial work of the Editio Critica Maior and, through the ECM, has had an increasingly significant impact on the Greek New Testament used by scholars, students, pastors, and translators around the world. 

Since starting this blog, I’ve received several requests for a post introducing the CBGM to readers who already have a working knowledge of New Testament textual criticism but have not yet become familiar with the CBGM, or not as familiar as they would like to be. This post is my attempt to meet that need, though I expect it will continue to be revised and expanded over time.

There are already many excellent publications available on the method, although much of this material is detailed, highly technical, and not open access. At the end of this post, I will link to a comprehensive bibliography of printed resources on the CBGM for those who want to delve deeper. I will also highlight the three works I especially recommend. 

My aim here is modest. I do not intend to describe the more technical aspects of the CBGM, since others have already done that well. But since much of my own work involves applying the CBGM in the editorial work of the ECM, it seems only fitting to offer an orientation for readers who want to understand its basic purpose and significance for the field of New Testament textual criticism and beyond.

Understanding the CBGM is becoming more pressing than ever. While many discussions about it were drafted when only one volume of the ECM had been published, the situation today is quite different. The CBGM has now been applied across multiple published ECM volumes, including Mark, Acts, the Catholic Letters, and Revelation. As a result, nearly one-third of the Greek New Testament has been edited on the basis of the CBGM. Taken together, this has resulted in around 200 textual changes that have been adopted in the newly published UBS6 and the forthcoming NA29.¹ For this reason, it is more imperative than ever to have at least a basic grasp of the CBGM and the editorial process behind it.

1. Why the CBGM Was Developed

The CBGM was designed in response to one of the central problems in New Testament textual criticism: the complexity of the textual tradition.

The manuscripts that preserve the Greek New Testament do not form a simple, unmixed line of descent from the initial text to later copies. Gerd Mink describes this complexity as contamination. Although the term may sound negative, at its most basic level it refers to mixture and lost sources in the transmission of the text. A manuscript’s text may preserve early readings in one place, later readings in another, and corrections or influences from other sources.

In addition to mixture, the same variant can sometimes arise independently in different witnesses. This does not necessarily mean that those witnesses are closely related. Rather, copyists were often prone to the same kinds of changes.² This is referred to as “multiple emergence” or “coincidental emergence” of a variant.

If manuscripts descended neatly from one another in a single line, their relationships could be mapped with relative ease. But in the textual transmission of the New Testament there are many lost witnesses, and the ones that are extant do not always fall into tightly defined families. The presence of mixture makes it difficult to determine how one witness relates to another when considering its entire text.

Earlier text critics were, of course, aware of these problems and developed various approaches to manage them. One of the most dominant was the classification of witnesses into text-types, such as Alexandrian, Western, and Byzantine. These categories had enabled scholars to group witnesses together based on common readings, which gave them a way to organize an otherwise overwhelming body of evidence. In that sense, text-types served a useful purpose: it is easier to deal with three or four groups of witnesses rather than thousands of individual witnesses.

While text-types were a useful heuristic, they were limited as a tool for explaining the full complexity of the tradition. They are based on certain variants, but they do not fully account for the distinctiveness of individual witnesses. A manuscript may agree with one group of witnesses in one set of readings and with another group elsewhere. As Dirk Jongkind says, text-type labels mean “the voice of a whole group of manuscripts is reduced to a single voice,” which flattens “a complex reality” into “only a near-truth.”³ The CBGM offers an alternative way to analyze extant witnesses in light of all of their respective variants.

2. What is the CBGM?

The CBGM was first developed in the 1980s by Gerd Mink to analyze genealogical relationships between variant readings and the witnesses that preserve them.⁴ The CBGM calculates percentages of agreements among witnesses at individual points of variation and evaluates which readings are likely to have given rise to others. It uses this analysis to relate genealogical relationships among variants to the relationships between witnesses. In the CBGM, “witness” means the text recorded in a surviving manuscript. The CBGM distinguishes carefully between a manuscript as a physical object and the text that manuscript carries.⁵

One of the most important distinctions in the CBGM is the difference between the age of a manuscript and the age of the text that manuscript transmits. As Wasserman and Gurry explain, “A text of a manuscript may, of course, be much older than the parchment and ink that preserves it.”⁶ A younger manuscript may preserve an older form of the text, while an earlier manuscript may contain readings that were more prevalent later in the tradition. For this reason, when the CBGM maps the text of one witness as prior to the text of another, it is not referring to the age of the physical artefact itself, but to the readings preserved in that manuscript.

The method does not begin by assigning witnesses to predefined text-types.⁷ Rather it proceeds from the ground up by comparing witnesses at places where there are textual differences (i.e., variants). The CBGM then calculates how many agreements one witness has with another at these points of variation. The editors of the CBGM then use this quantitative evidence, together with traditional text-critical criteria, to ask at every variation unit which reading likely came first (i.e., prior) and which arose from the initial reading (i.e., posterior).

Put simply, the CBGM works directly with extant witnesses and their preserved readings. It does not construct a family tree of manuscripts which depends on reconstructed ancestors (i.e., hyparchetypes) to fill in missing links. This makes the method especially valuable because its conclusions remain grounded in the available evidence.

This also means that the CBGM does not set out to determine whether one manuscript was copied directly from another. Since many manuscripts have not survived, a witness that is genealogically prior to another should not automatically be understood as its direct physical ancestor.

With this basic description in place, we can now look more closely at two concepts that are especially important for understanding the CBGM: pre-genealogical coherence and genealogical coherence.

2.1. Pre-Genealogical Coherence

Pre-genealogical coherence refers to the degree of textual agreement between witnesses at places where variation occurs. Simply put, it tells you the percentage of textual agreement between any two witnesses.

Text critics have always noticed that some witnesses agree more often than others. What is new is that the CBGM applies this comparison systematically, using complete transcriptions and a full apparatus rather than collations only at a selection of passages.

The result is an overall percentage of agreement between witnesses based on their variants, which gives editors a clearer sense of which witnesses are textually close to one another. 

2.2. Genealogical Coherence

Genealogical coherence asks a different question. Once the degree of agreement between witnesses has been calculated, editors evaluate the direction of textual flow. They ask whether one reading can plausibly explain the origin of another. This is a matter of evaluating priority and posteriority, or which reading is likely earlier and if the other readings could reasonably have developed from it. It is also possible to say that the source of a reading is unknown. To judge the priority of a variant, editors must consider philology, immediate context, authorial style, transcriptional probability, scribal habits, and other traditional criteria. 

At each variation unit editors construct a local stemma. This diagram represents a hypothesis about the development of readings at individual points of variation. The local stemma portrays, for example, whether one reading is more likely to have given rise to another. 

A secondary reading could have arisen for a variety of reasons, for example if the copyist made a mistake during the transcription process or perhaps recalled a more familiar parallel wording and wrote that instead. This is where the method moves from determining objective agreements between witnesses to interpreting relationships: pre-genealogical coherence shows the agreement rate of witnesses, while genealogical coherence asks how the texts are related to each other.

After local stemmata have been created at variant passages, the relationships between witnesses can be evaluated more fully. From these local judgments about individual readings, the editors can then move outward to genealogical relationships between witnesses.

In this way, they can begin to identify which witnesses are the first potential ancestors of others, and which have the Ausgangstext as their first potential ancestor.⁸ Witnesses for which the Ausgangstext is the first potential ancestor are called “A-related witnesses,” even though all witnesses are related to the Ausgangstext to some extent.

The visualization of this data is given as a textual flow diagram.

Textual flow diagrams are often misunderstood as a family tree. While this analogy might be tempting because of its familiarity, it is misleading because a family tree implies a chronological sequence of physical manuscripts and direct lines of descent. The analogy breaks down in the context of the CBGM because the method does not necessarily show which manuscript was copied from which, especially since many intermediate manuscripts have not survived. Moreover, the age of a manuscript does not necessarily reflect the age of the text it preserves. For this reason, a witness that appears textually prior to another should not be understood as its direct ancestor, but as its first potential ancestor among the extant witnesses included in the CBGM.

2.3. Two Kinds of Coherence

To summarize, while the names of the two terms sound similar, the distinction between pre-genealogical and genealogical coherence is crucial. Pre-genealogical coherence measures agreement. Genealogical coherence evaluates direction. The first asks, “How often do these witnesses agree?” The second asks, “What do the readings suggest about the relationship between the witnesses?” The work of the former is largely objective and carried out by the software; the latter is the result of the text-critical judgments of the editorial team. 

3. Who or What is Behind the CBGM?

Descriptions of the CBGM often place too much emphasis on the method as digital or computer assisted. In reality, the work could be done on pen and paper, as was apparently done for ECM James (1997) and ECM Letters of Peter (2000), though digital tools have greatly accelerated the process.⁹ Computers help gather, organize, and compare the evidence, but they do not replace editorial decision making. As Gerd Mink emphasizes, “The Coherence-Based Genealogical Method makes no textual decisions. It merely reveals an image of the tradition which emerges from a text-critical philological study of all the variants.”¹⁰

In common parlance, we often speak as if “the CBGM” were a single agent that performs textual operations and reaches conclusions. When phrases are used such as “the CBGM shows,” “the CBGM determines,” or “the CBGM changes the text,” the agent behind the action is not always clear. Does “CBGM” refer to the software, the method as a whole, or the editors employing the method? These distinctions matter. The software can calculate and display coherence, but it cannot exercise philological judgment. Conversely, editors make textual judgments, but those judgments are informed by detailed computer-generated calculations and visualizations of data. “The CBGM,” then, is best understood as shorthand for a set of practices in which calculations, coupled with human judgment are both necessary for the editors to interpret textual evidence, evaluate variant readings, and assess genealogical relationships.

In my view, the CBGM is better understood not merely as a “method”—if that term implies a fixed procedure that automatically yields textual decisions—but rather as a broader methodology. This is because the CBGM makes explicit not only the procedures it uses, but also the rationale, assumptions, and principles that guide those procedures. Digital tools have become invaluable within this methodology, as they reliably tally percentages of agreement and coherence among witnesses; however, they do not themselves determine the text. Editorial evaluation, drawing on philology and established text-critical criteria, remains an integral part of the process.

To make this distinction clearer it is helpful to separate software-driven processes from editorial decisions. The following lists distinguishes these two aspects, showing more precisely what the computer does and what the editors do.

What the Software Does

What Human Editors Do

Before using the CBGM, editors:

Within the CBGM online interface, editors:

The editorial work within the CBGM is iterative, meaning that editors can revise their decisions at any point in the process. After an initial pass through the text, in which the local stemmata are constructed, editors can reassess how those decisions affect the coherence between witnesses. This may make it necessary to revise particular textual decisions, adjust local stemmata, or make another pass through the evidence.

4. The CBGM’s Impact on New Testament Textual Criticism

As noted above, the CBGM has already had a significant effect on the production of the ECM and, through the ECM, on the revision of the Nestle-Aland and UBS editions. But its influence is not limited to critical editions. It has also become a useful tool for research in New Testament textual criticism.

4.1. A Different Vantage Point of the Tradition

The CBGM is best known for its role in the ECM, where its impact on New Testament textual criticism has been most visible and where most readers first encounter it. But the method also offers a way to analyze the manuscript tradition beyond the confines of the ECM. By calculating textual agreements and the direction of textual flow, the textual-flow diagrams can display how readings spread, where they cluster, and how different streams of transmission developed.

Tommy Wasserman, for example, has argued that the CBGM is useful not only for reconstructing the New Testament Ausgangstext, but also for surveying the history of readings and explaining textual changes.¹⁶ Matt Whidden has likewise used the data from the CBGM to examine the relationship between individual witnesses, arguing that 205 is not the exemplar of 2886 since 2886 is the first potential ancestor of 205.¹⁷ These examples show that the CBGM can be used not only to make editorial decisions, but also to study the development of readings and the genealogical relationships among manuscripts, even when reconstructing the Ausgangstext is not the primary aim.

This wider potential is already suggested by open-source platforms such as Joey McCollum’s open-cbgm, which describes the CBGM as a “meta-method” that combines philological decisions with computer-based calculations to highlight genealogical relationships between different stages of a text.¹⁸

4.2 Greater Transparency

Another novelty of the CBGM is the increased transparency it brings about editorial decisions. This does not mean that every editorial decision is obvious or easy to evaluate from an outsider perspective. The CBGM tools still require training for the uninitiated to use, but they do allow users to insights into far more of the reasoning and evidence behind the ECM than was possible in earlier printed editions alone.

The CBGM is publicly available for Mark, Acts, Catholic Letters (temporarily offline), and Revelation. The CBGM program online displays the constructed local stemma at every variant passage; users can compare witnesses, see pre-genealogical and genealogical coherence, and interact with textual-flow diagrams. This data, along with the “Text-Critical Commentary” published in all ECM volumes (except for the Catholic Letters), which is freely available online, offers readers unprecedented access to the editorial decisions behind the text.

4.3. Customizable Digital Tools

The CBGM data is dynamic, allowing researchers to use the online tools the editors have created by changing the settings, which include displaying or removing certain witnesses from the textual-flow diagrams, adjusting the connectivity levels, and comparing witnesses at every possible variant passage. These tools do not simply present conclusions; they allow users to explore the source material, examine relationships between witnesses, and explore the effect on witness relationships when different decisions are made. In this way, the CBGM has affected not only how editors work but also how researchers can engage with the text.

The Docker installment of the CBGM can be downloaded on a personal computer, offering researchers a tool to experiment with the data by re-editing the local stemmata (see below). In the future the CBGM will be integrated into the NTVMR, so it will be feasible to make a critical edition beginning with one’s own transcriptions and ending with reconstructing the text in the CBGM. (Future developments in AI may also change how some of these preparatory tasks, especially transcriptions and apparatus construction, are carried out.)

For these reasons, the CBGM should not be viewed merely as a new method for establishing the initial text. Its significance lies also in the broader possibilities it has introduced: new technological advances, greater transparency in editorial decision-making, and access to digital tools that allow users to explore the textual tradition with a level of flexibility and precision previously unavailable. ¹⁹

4.4. Clarifying the Limits of the CBGM

Appreciating the method requires a clear sense of what it can and cannot do. It is often assumed that the CBGM should resolve textual contamination, reconstruct a complete historical genealogy of manuscripts, automatically determine the origin of readings, or produce objective results independent of editorial judgment. Some also expect it to account exhaustively for every dimension of the textual tradition, including versions, corrections, manuscript layers, editorial stages, and competing reconstructions of the text. These expectations, however, misunderstand the purpose of the CBGM. It is not a self-operating system, but an analytical tool used by editors to evaluate relationships among extant witnesses based on their readings.

The CBGM cannot resolve contamination in any absolute sense because the textual tradition is incomplete: many manuscripts and intermediary connections have been lost. Nor does it produce a full history of the text since it works only with extant evidence establishes relative genealogical direction rather than absolute chronology or historical descent. Likewise, it does not determine the priority of readings automatically because local stemmata depend on prior philological judgments rather than generating those judgments themselves.

For the same reason, the CBGM does not incorporate every possible type of data indiscriminately. Including evidence such as retro-translated versions, for example, could introduce ambiguity and distort genealogical coherence. Nor does it rely on excessive manipulation of the data, such as weighing variants or constructing hypothetical ancestors, since such procedures would increase subjectivity and weaken the method’s aim of providing a controlled and transparent analysis of textual relationships. In short, while there are still many ways the method could be improved, many criticisms of the CBGM arise from expecting it to perform tasks for which it was never designed.²⁰

5. Beyond New Testament Textual Criticism

Use of the CBGM in other textual traditions has also been carried out successfully. Leonardo Pessoa, for example, used the CBGM in work on the Septuagint.²¹ Work on the Zoroastrian text Yasna has proven that the CBGM can successfully be used for other languages.²² This shows that scholars working on other complex manuscript traditions are asking similar questions about transmission, variation, and the relationship between witnesses. The CBGM can also be helpful for addressing those questions.

For that reason, even when all ECM volumes have been published, the CBGM will likely remain a significant resource both within New Testament textual criticism and beyond. 

6. Additional Resources 

As promised, here are my top three must reads if you want to understand the basics.

The first resource is, in many ways, the “CBGM Bible”: Gerd Mink’s introduction to the method. While most readers will not grasp it on a first pass, and some may not make it very far before setting it aside for a break, it is an invaluable resource. Much like the CBGM itself, learning the method is an iterative process. You will likely need to return to it several times, each pass clarifying something that was obscure the first time around.

This is where Peter Gurry’s published dissertation, A Critical Examination of the Coherence-Based Genealogical Method, becomes especially helpful. Its first six chapters are an unparalleled explanation of the method and was formative for my own understanding of the CBGM. It is a great place to start. 

The introduction by Tommy Wasserman and Peter Gurry, A New Approach to Textual Criticism, is also extremely useful, and I highly recommend it as well. It has essentially become the standard for learning the method since it is very approachable and relatively short. Additional works written by Tommy and Peter will also be helpful to understand the method.

For readers who want to understand the practical application of the method, especially in the context of the ECM, Klaus Wachtel’s numerous publications on the CBGM are also indispensable.

And finally, for a hands-on approach, you should download the CBGM and play around with it. I give a video tutorial here for how to do that.

In closing, I wish to point readers to a wide range of resources on the CBGM for those who want to dig deeper. I have compiled a comprehensive bibliography of publications on the CBGM and would welcome suggestions if anyone notices an important item is missing. It should be noted that not all of the resources in the bibliography are useful for understanding the method—unfortunately some of them elicit gross misunderstandings.

7. CSNTM Conference Presentation

If you have made it to the end of what was meant to be a brief post, thank you very much for your time and attention. 

For those interested in this topic, and more broadly in the ongoing work of the ECM, this is also a fitting place to mention that I will be presenting on the CBGM at the CSNTM Text & Manuscript Conference, hosted in Texas, later this month.

My abstract is available here. This year’s theme is The Editio Critica Maior, Exegesis, and Translation. It is the first international gathering devoted entirely to the ECM, and the program includes a strong lineup of papers by great colleagues. I’m looking forward to learning from these presentations and to the conversations they will no doubt generate about the present and future of the ECM and New Testament textual criticism more broadly.

I welcome feedback on this post, and I hope to continue the conversation with many of you there.

---

1. These “textual changes” are in comparison to the text of the NA27. The text from the ECM was first taken over in the UBS5 and NA28, which resulted in 33 textual changes in the Catholic Letters. Now with the publication of the UBS6, 167 changes have been introduced in Mark, Acts, and Revelation, based on the ECM. ↩︎
2. Some of these common changes might be, for example, changing a singular verb that modifies a neuter plural subject to a plural verb, or replacing καί for δέ—equally understandable would be the reverse of each of these situations, especially considering that these types of common changes would likely have no impact on the meaning of the text and would go easily unnoticed by correctors or other copyists. ↩︎
3. Dirk Jongkind, “The Genealogy of Manuscripts and the CBGM,” Evangelical Textual Criticism Blog, Oct. 24, 2013, https://evangelicaltextualcriticism.blogspot.com/2013/10/the-genealogy-of-manuscripts-and-cbgm.html ↩︎
4. For a history of the method, see Gurry, A Critical Examination, chapter 1. ↩︎
5. See Wachtel, “Notes on the Second Edition,” 31*, who states, “the texts are the witnesses, not the manuscripts. The quality of a text is not defined by the date of the respective manuscript or by the writing material or style. The only reliable fact in this regard is that the text form of a manuscript existed at the latest from the time when the manuscript was produced. The text transmitted in a manuscript is, for the most part, older than the manuscript itself.” ↩︎
6. Wasserman and Gurry, A New Approach, 3. ↩︎
7. In light of the results of the CBGM, the broader theory of text-types has become increasingly difficult to sustain, except in the case of the Byzantine text, which represents a highly stable tradition in which many witnesses share 90 percent or more of their readings. ↩︎
8. Mink defines the Ausgangstext as “the text from which the entire tradition originates and which immediately precedes the first branching into various traditions.” (Original German: “Der Ausgangstext ist der Text, von dem die gesamte Uberlieferung ihren Ausgang nimmt und der der ersten Filiation in verschiedene Uberlieferungszweige unmittelbar vorausgeht.”) Mink, “Eine Umfassende Genealogie,” 482. ↩︎
9. Cf. ECM 1 John (2003), preface. ↩︎
10. https://www.uni-muenster.de/INTF/Genealogical_method.html ↩︎
11. This is pre-genealogical coherence, which is based on the variation units that are delineated by humans. ↩︎
12. The directional flow between witnesses is based on the human decisions made at all local stemmata. Genealogical coherence is visualized in various diagrams: Coherence at Variant Passages, Coherence in Attestations, and General Textual Flow. ↩︎
13. For example, this is displayed as a broken line between two witnesses in the various diagrams on the CBGM website. ↩︎
14. This is done by identifying the minimum number of witnesses required to explain all of the readings in that witness. ↩︎
15. This includes deciding what types of variant readings will go into the apparatus by classifying readings as meaningful or not (e.g., orthographic differences are not meaningful), regularizing non-standard forms of words, defining lacunae, and assigning readings and their attestations a letter to identify it. ↩︎
16. Wasserman, “The Coherence-Based Genealogical Method as a Tool.” ↩︎
17. Whidden, “Revisiting GA 205 and 2886.” ↩︎
18. Joey McCollum, open-cbgm, GitHub, https://github.com/jjmccollum/open-cbgm ↩︎
19. The CBGM was also the first published digital component of the ECM. Before there was a digital ECM, and before transcriptions were published in the NTVMR, the Genealogical Queries module for the Catholic Letters was offered online. Here, researchers could already access much of the essential data that the editors of ECM Catholic Letters used. From there, a new interface was developed for CBGM Acts, which is now the standard and has been used for CBGM Mark and Revelation. The digital ECM on the NTVMR links directly to the CBGM. Digital tools were also created for ECM Parallel Pericopes, found here: http://intf.uni-muenster.de/TT_PP/, though the CBGM was not applied to it. These tools calculate the pre-genealogical agreement of manuscript witnesses (1) at all Gospel test passages used in TuT (called Test Passages: Manuscript Clusters), or (2) using full transcriptions available for each of the 41 pericopes (called Parallel Pericopes: Find Relatives and Parallel Pericopes: Manuscript Clusters). ↩︎
20. For ways in which the CBGM can be further developed, see Mink, “ Manuscripts, Texts, History, and the Coherence-Based Genealogical Method (CBGM);” Mink, “The Necessity of Constructing Global Stemmata;” Wachtel, “Elements of a Global Stemma;” and Paulson, “Improving the CBGM.” ↩︎
21. Pessoa, “The Septuagint of Samuel.” ↩︎
22. Eventually their CBGM will be available here: https://muya.uni-muenster.de/. See their website: also https://muya.soas.ac.uk/ ↩︎