The CBGM and the Claim of the ‘Best Manuscripts’

Article 7 of 27 in the series “When the Canons Collide.”

The CBGM is often praised by defenders and dismissed by critics in forms that its own practitioners would reject. It is not a machine that discovers a single physical family tree. Nor is it a mere disguise for Westcott and Hort. It is a powerful set of tools for comparing agreements, testing coherence, and displaying potential relationships across a vast corpus. Yet the direction assigned to local variants is still a philological act, and the global picture depends on those local orientations. The method can discipline judgment without abolishing it.

Controlling question: What does the CBGM calculate, what must editors decide, and can its genealogical gains settle the jurisdictional question?

The Coherence-Based Genealogical Method

The Coherence-Based Genealogical Method (CBGM) is sometimes imagined as a computer that solves the genealogy of New Testament manuscripts. That description is false. The CBGM is computer-assisted. It analyzes patterns of agreement, displays potential ancestors and descendants, and tests coherence across a large field. But editors make the local philological decisions that orient variant relationships.

The Institute for New Testament Textual Research describes the method in terms of local stemmata, pre-genealogical coherence, genealogical coherence, and an initial text. The initial text is not an extant autograph; it is the reconstructed starting point for the tradition represented in the edition. Local stemmata—editorial hypotheses about which variants derive from which—feed the global picture. The global coherence can then expose difficult local decisions and prompt revision.¹

This iterative structure is powerful. It can reveal that two manuscripts agreeing frequently are not independent. It can undermine crude text-type categories. It can make thousands of relationships visible in ways no unaided scholar could hold in mind.

It does not remove the problem of adjudication. Official explanations are candid that philological assessment remains central, that delimiting variation units contains a subjective element, and that coherence is one argument among others. Internal evidence can override coherent attestation. A local stemma may be provisional and revised when the global picture creates pressure.²

The result is neither “pure subjectivity” nor an algorithmic oracle. It is a sophisticated feedback system joining human local judgments to computational measures of coherence.

That description yields four cautions.

The starting orientation is human

To calculate directed genealogical relationships, variants must be oriented: one is treated as prior to another, or their relationship is left unresolved. Those decisions employ the familiar reservoir of internal and external considerations.

Coherence can discipline but not consecrate

If a local decision makes a witness's relationships globally incoherent, the editor gains a reason to reconsider. But coherence does not equal originality. A widespread error can be coherent. Contamination can transmit readings across branches. The model measures relations within its data and decisions.

Iteration can correct or amplify

Feedback is not inherently circular in a vicious sense. Scientific models often iterate between local data and global fit. Yet iteration can amplify initial assumptions if alternative orientations are not equally tested. Transparency about unresolved directions and sensitivity to changed inputs is essential.

A better genealogy does not supply a theology of authority

Even a perfect account of extant textual relationships would answer a historical question: how surviving states relate. It would not, without additional premises, answer why an editorially reconstructed initial text has authority to displace the Church's received Scripture. Computer assistance changes scale, not jurisdiction.

“Best Manuscripts” as a Conclusion

The phrase “the best manuscripts” should therefore appear at the end of an argument, not the beginning. A careful statement would specify:

• best in which book or corpus;
• best by which tested measure;
• best at preserving which kinds of readings;
• best relative to which independently established cases;
• best with what known exceptions;
• and best for what conclusion.

Without those qualifications, the phrase functions rhetorically. It invites the reader to imagine that quality is printed in the parchment.

This matters pastorally. A marginal note that says “the earliest and best manuscripts omit” compresses multiple layers of inference into a sentence. It may report a defensible scholarly consensus. It should not be mistaken for the voice of the artifacts themselves.

External Evidence Has Internal Premises

Dabney's claim that the foundation remained inferential can now be stated more precisely. External evidence is not unreal, but its weight contains internal premises. Dating uses internal marks of script and codex. Genealogy uses patterns of readings. textual character uses prior local decisions. Versional support uses translation judgments. Patristic support uses judgments about quotation, source, and transmission.

This does not collapse all evidence into arbitrary preference. It means the courtroom metaphor must be completed. Witnesses enter the court only after an advocate has established identity, competence, independence, and relevance. The judge then assigns weight. The testimony is real; the ranking is adjudicated.

When the judge is also constructing the statute by which the court itself will be bound, the question of jurisdiction returns.

Four Questions to Ask of Any CBGM Appeal

An appeal to the CBGM should be made concrete before it is allowed to settle a local debate.

1. Which local relationship is being claimed? State the competing variants and the editorial direction assigned between them. A global diagram cannot replace the local philological argument from which it was built.
2. What kind of coherence is at issue? Pre-genealogical coherence measures agreement. Genealogical coherence uses directed local relationships. Neither is a simple photograph of physical descent.
3. How sensitive is the result? Ask whether another defensible local orientation would materially change the potential ancestors, textual flow, or global fit. A robust conclusion should survive reasonable alternatives.
4. What conclusion is being drawn? The method may show that one decision fits the wider relationships better. That does not by itself establish the autograph, divine preservation, or the right to bind the Church.

These questions keep the critique current. They also keep defenders of the received text from dismissing a real advance merely because it does not answer the theological question. The right answer to overstatement is precise limitation, not denial of the data.

What This Article Establishes

• Computer assistance increases the reach and consistency of analysis but does not make local textual decisions impersonal.
• Coherence can expose strain in a decision, yet coherence is not identical with originality.
• Best manuscripts is a conclusion that must name its corpus, measure, exceptions, and intended use.

Conclusion

A stronger genealogy is a genuine scholarly gain. It may overturn a careless appeal to manuscript count or expose a local decision that does not fit the wider evidence. But scale is not jurisdiction. The CBGM cannot answer what its designers do not ask it to answer: by what revealed right an editorially reconstructed initial text displaces the Church's received Scripture as the final norm. That question remains when every diagram has been drawn.

Previous: Manuscripts Give Readings, Not Rankings · Next: Is the Shorter Reading Better?