Triggers for Harmonisation

When we had to work through the whole of the New Testament in a more systematic way, we started with the Pauline corpus. The assumption was that the letters of Paul did not pose as many problems as some other parts of the NT, and this assumption bore out. Apparently there is something in tightly argued prose that puts it in less danger of textual change than simple narrative, especially narrative with synoptic parallels. Yet even within the Pauline corpus the same phenomena are present that you can find in the Gospels. Ephesians and Colossians contain sufficient parallel material to allow for cross-contamination, and the same happens with Galatians and Romans.
However, influence from parallel passages is not limited to similar sentences or similar narratives. There are all sorts of phenomena that can spark off cross-contamination. And, true to the reputation that the Greek-Latin manuscripts have, a number of these are found in the D(06) F(010) G(012) cluster. Two obvious examples to illustrate the point.

Gal 4:17 ζηλοῦσιν ὑμᾶς οὐ καλῶς, ἀλλὰ ἐκκλεῖσαι ὑμᾶς θέλουσιν, ἵνα αὐτοὺς ζηλοῦτε. (‘They are not zealous for you in a good way, but they want to shut us out so that you may be zealous for them’)

Paul finishes the sentence with ζηλοῦτε and after a negative sentence such as this one often introduces a positive contrast, which more or less follows in the next verse. Somehow, however, linked by the contrasting pattern and specifically the link work ζηλοῦτε we find in the D(06) cluster the extra words ζηλοῦτε δὲ τὰ κρείττω χαρίσματα (‘but be zealous for the better gifts’). These words are a clear echo of 1 Cor 12:31, though with some minute differences. Is this addition simply a marginal note that slipped into the main text? Is it the result of someone who is copying Galatians from memory more than from a document? Who knows, but the extra words are there now. The link is tenuous but we could reconstruct the triggers, and therefore learn something about the way in which copying can be affected.
The second example is just as gorgeous and concerns influence from within Galatians.

Gal 3:1: Ὦ ἀνόητοι Γαλάται, τίς ὑμᾶς ἐβάσκανεν (‘O foolish Galatians, who has bewitched you’).

The D(06) cluster, now with a whole lot of additional support, adds the words τῇ ἀληθείᾳ μὴ πείθεσθαι (‘to not obey the truth’). The source of these words is Gal 5:7 and the trigger here is the start of the actual question, τίς ὑμᾶς, followed by a verb (ἐβάσκανεν and ἐνέκοψεν), and both in an accusatory question. Not that much to go on, mainly the τίς ὑμᾶς part. Still it is enough of a trigger to import wording from elsewhere.

Both Galatian cases are quite clear to me (the second of course not for those who favour a Byzantine text, though I assume they would acknowledge the mechanism in the first case). Which leads me to the conviction that if this mechanism is at work in clear instances, it might well be at work in many less obvious cases. Therefore, if there is an explanation available that can explain the longer text as being the result of influence from elsewhere, the shorter reading has a strong transcriptional case.


Romans 1:1 and Manuscript Tendencies

“Knowledge of documents should precede final judgement upon readings.” Useful though this adage by Westcott and Hort is, it is also a little bit of an open door: lots of things should precede ‘final judgement’ (and when is anything ‘final’ in our discipline?) But what are the things that one needs to know about a document? It seems to me that ‘knowledge of documents’ includes also ‘knowledge of its readings’. Because manuscripts are not just characterised by their appearance, their dating, and their paratextual apparatus, first and foremost they are carriers of a text with a specific wording. The particulars of the wording of a manuscript, its ‘readings’, make a manuscript textually different from others and are an accumulation of inherited readings and scribe-created ones, but for our purposes today this latter distinction is irrelevant. What I want to demonstrate is how important it is to know about the tendencies a manuscript exhibits in the wording of the text, that is patterns within their readings. None of these individual readings needs to be unique to the witness, as long as one can make a case for a certain inclination to a type of reading. So we are talking ‘manuscript habits’ rather than ‘scribal habits’, or perhaps better, ‘manuscript tendencies’. Today I will share an example that explains a difference between the Tyndale House Edition and the modern critical editions.

Romans 1:1 starts as follows in the current Nestle-Aland, UBS, and SBL texts: Παῦλος δοῦλος Χριστοῦ Ἰησοῦ, ‘Paul a servant of Christ Jesus’. The Tyndale House Edition breaks the mold of modern critical texts and has, Παῦλος δοῦλος Ἰησοῦ χριστοῦ, ‘Paul a servant of Jesus Christ’ [edited: Initialy I claimed wrongly that we had retained the reading of Tregelles]. The difference is that of the word order in ‘Jesus Christ’. Both orders are found widely elsewhere in the Pauline corpus, though ‘Christ Jesus’ is typically Paul, with hardly any occurrence outside his writings.

The evidence for ‘Christ Jesus’ here in Rm 1:1 is limited though, only B(03) and P10, a roughly written papyrus that doesn’t seem to have been part of a whole text of Romans, but contains nothing more than the opening of the epistle. Both manuscripts are palaeographically dated to the fourth century. In addition there are the minuscules 81 and 1838. Neither the two minuscules or P10 in themselves would be sufficient to tip the scales in favour of considering ‘Christ Jesus’ in the main text, but it is the presence of B(03) that opens up the possibility.

I assume that one of the main reasons why ‘Christ Jesus’ is preferred is that it is more likely that typically Pauline usage is harmonised to the more general pattern than the other way around. It would almost require a scholarly and editorial mind in order to make a text more ‘Paul’ than Paul himself was, and since we don’t know of such source, the reading of B(03) and P10 is more likely original than not.

This is where manuscript tendencies come in.

There are a number of variants elsewhere in Romans where B(03) occupies also a heavy minority position in its readings. In the following five places B(03) has Christ Jesus where almost everyone else has Jesus Christ: 5:17; 5:21; 13:14 (’Christ Jesus’ for ‘Lord Jesus Christ’); 16:25 (together with minuscule 1739); 16:27. In addition there is 3:22 where B(03) has ‘Christ’ for ‘Jesus Christ’, whilst A(02) has ‘Christ Jesus’, and also 5:15 where minuscule 1739 has ‘Christ Jesus’ for ‘Jesus Christ’. The changes in B(03) are not systematic or exhaustive by any means, but they are remarkable, and concentrated in this particular witness.

There are two things one can do: accept the text of B(03) in all these places, or conclude that B(03) has in this particular feature the tendency to do what we above rejected as highly unlikely, namely to make Paul’s text more like Paul. I believe it is the latter, but this implies something quite interesting: the text-critical canon to prefer the reading that is most in line with the style of the author has been applied apparently already in this fourth century manuscript! In B(03) we find an intensification of Pauline style, at least when it comes to the order ‘Christ Jesus’ (I believe that I have found similar intensification of author’s style elsewhere in B(03), but that is not for now).

Summing up, B(03)’s tendency throws a large shadow over the value of B(03) in Rm 1:1 and makes it hard to accept its reading. More broadly, I hope this shows how important it is to know the tendencies of some of the key witnesses (and constellation of witnesses) in order to understand and explain the evidence at any given point. And that is why we need good and deep studies of our main manuscripts. To come back to Παῦλος δοῦλος Ἰησοῦ χριστοῦ in the Tyndale House Edition, now you know the story—it is more than just a numbers game.


Getting the Readings Right

A textual apparatus is useful as a quick summary of the evidence at a particular location and also to raise questions as to unexpected manuscript combinations or readings. However, by its very nature, an apparatus presents the evidence in an atomistic way and runs the risk of fostering a view of an artefact as little more than a collection of mutually independent readings. So it is advisable to have a large number of images open on your screen – and nowadays that is not much of a problem. But then we get the small problem of understanding what is actually there. The following two examples posed considerable problems, even though there is no problem with the physical clarity of the writing.

The first example is from Gal. 5:26.
Μὴ γινώμεθα κενόδοξοι, ἀλλήλους προκαλούμενοι, ἀλλήλοις φθονοῦντες.
‘Let us not become conceited, provoking one another, envying one another.’

There is variation on the case ending of the second occurrence of the reflexive pronoun αλληλοις / αλληλους. In many aspects this is an interesting variant: not only are the early big codices split but also the Textus Receptus has a different reading from the Majority Text (provided that the CNTTS apparatus in BibleWorks is correct).

From a scribal habit perspective the solution is straightforward as the second instance underwent influence from the first, resulting in twice αλληλους.

What I am interested now is the reading of a minuscule in the BnF, GA 6.

Often it is possible to ‘predict’ the reading of a manuscript given its affinity to other manuscripts, but in this case it is much trickier. Does the ligature indicate -οις or -ους? The problem is that this minuscule does not use many ligatures for these endings and I couldn’t find any parallels in the vicinity. So I fired off an email to Maurice Robinson who I trust to have seen a few more minuscules than I, with, what I thought, was a ‘simple’ question. Questions are never simple. The solution he came up with, which I think is likelier than any alternative I had, was that the ligature for the accusative plural -ους is quite distinct; and he sent me kindly some examples from Hebrews. When trying to locate these I stumbled over a clear parallel in Heb 2:18 for -οις in the expression δυναται τοις πειραζομενοις βοηθησαι, ‘he is able to help those who are being tempted’.







This solves the problem of GA 6’s reading in Gal 5:26; it reads αλληλοις.

The second example is from Codex Vaticanus, B(03). What does it read after τις ουν in 1 Cor. 9:18? μοι or μου?

It is not particularly clear to me and could go either way. Coupled with the correction that happens in Sinaiticus, ℵ(01), there is plenty of uncertainty around. In this case I judged that the direction of change would be from μου to μοι, in part because of the position of the pronoun. Had μου been at its neutral position after μισθος the problem would never have arisen. Still, what B(03) actually says is not immediately clear.

These are just two examples where the granularity of the data is starting to pose limits of what we can know. As a wise author once wrote, reality is not digital but analogue. With sufficient study we can push back the grey areas, yet only to find others instead.


The Text of the New Testament, of an Edition, and of a Manuscript

Lots of ink has been spilled and many pixels been lit up on the question of what we mean when we say ‘the text of the New Testament’. Is the only thing we are left with ‘texts of the NT’, in which every imperfectly copied manuscript constitutes a new and different text? How much can the text of a ‘work’ (that is here, the New Testament) be changed before the text of a particular manifestation of that work (a printed edition, a manuscript – the latter is also an ‘artefact’) is no longer that of the original ‘work’?1

In a sense this question is no different from what philosophers have been discussing since the days of Plato. Intuitively, or naively, most people who think for a moment about the text and the various forms in which it appears, solve the question the same way as Plato did. Different manuscripts with their slightly different wording, and even different translations of the text in a wild variety of languages, all constitute different instances of the same text, a perfect idea reflected in the wordings of the various manuscripts.
Relevance theorists would say that this is true because an utterance has no meaning in itself but only functions to point to the intention of the speaker, and listeners will interpret an utterance accordingly. When it comes to textual criticism, it may follow that the majority of differences we are talking about will hardly affect the listener’s (reader’s) construction of the speaker’s (writer’s) meaning. The same mental construction of the speaker’s meaning can be formed using an array of different utterances, and we can find evidence of this in that few, if any, of the various sub-strands of Christianity are based on particular manuscripts or depend on specific translations. If there were only ‘texts’ and not a single ‘text’, it is quite remarkable how few problems this has created in the course of history.
Perhaps a fresh way of doing Plato is found in linguistics, namely in prototype theory. Prototype theory says that often concepts are used in ways that are close to the mental prototype of that concept, but also allows for uses that are quite different. We all have the image of what a dog is, though many of us will have the experience to think about some actual dogs as more ‘dog’ than others; they conform more closely to the prototype we have formed. Perhaps it is possible to force an analogy with the ‘texts’ and ‘text’ discussion. There is the ‘prototypical text’ (perhaps better the archetypical text when we throw in chronology) and manuscripts, or again translations, conform more or less to the prototype yet still are all instances of that particular category / prototypical field. This is how we seem to organise concepts in our mind, and it works pretty well in the practice of doing textual criticism.

So what we have set out to do in the Tyndale House Edition is to present a text that approximates as closely as possible the oldest recoverable text since we hold that this is the best approximation and representation of the ‘ideal text’, the text of the ‘work’ as it was produced in the first place.

The theology of all this is of course quite a different matter.

1 Thanks to Michael Dormandy, who pointed me to this helpful distinction made by Driscoll: “Hamlet is a work. The New Swan Shakespeare Advanced Series edition of Hamlet by Bernard Lott, M.A. Ph.D., published by Longman in 1968, is, or presents, a text. My copy of Lott’s edition, bought from Blackwell’s in Oxford in 1979 and containing my copious annotations, is an artefact. (93)” I am quite conscious (and relaxed about this) that by talking about ‘original work’, I may be accused of misappropriation of the term as used by Driscoll.

Driscoll, M.J. “The Words on a Page: Thoughts on Philology Old and New.” In Creating the Medieval Saga: Versions, Variability, and Editorial Interpretations of Old Norse Saga Literature, edited by Judy Quinn and Emily Lethbridge, 85-102. Odense: Syddansk Universitetsforlag, 2010.


The First Step: Digitising Tregelles

Where do you start when preparing a Greek New Testament? Of course you can start absolutely from scratch, by typing in each and every letter and accent manually, with all the associated risks, but somehow this did not appeal very much. So we needed an existing text that we could adjust towards the desired wording of our edition.

What we did was to settle on the text as prepared by Samuel Prideaux Tregelles as our point of departure. The actual procedure was that we started with an electronic edition of the Nestle-Aland 26 / UBS3 text and adapted every paragraph, word, punctuation, and accent to match the print edition of Tregelles. We even asked permission to start out with the Nestle-Aland / UBS text as we wanted to avoid any shade of doubt as to existing copyright arrangements, yet in reply we were told that no permission was needed to create a new text in this way.

Interestingly, when Tregelles prepared his text, he started off with a Textus Receptus and adapted this text so that it matched his own. Especially in cases where Tregelles did not make an explicit decision or weighed the evidence, his text still reflects the TR. I assume that changing the TR to Tregelles included more changes than the NA26 / UBS3 to Tregelles.

It is surprisingly difficult, though, to get everything right. Even though two people worked independently on creating the digital Tregelles, and every difference between the two versions was resolved by reference to the printed Tregelles, a considerable number of errors remained. The whole exercise was a valuable lesson in the psychology of textual work. When starting from the same base text two transcribers of a second text may make at times the same error.

It is our digital Tregelles that functioned as a point of comparison for the SBL Greek New Testament project, and its editor, Mike Holmes, notified us of a good few errors in our transcription that came up in comparing Tregelles with a number of other editions. It is often impossible to see if you have typed a Greek ‘ο’ or a Latin script ‘o’.

There are a number of reasons why Tregelles was chosen as our starting text. One is that by starting from Tregelles we go back beyond Westcott-Hort and their influential and lucid textual theories, but not as far back as the Textus Receptus. We could have opted for the text of Lachmann too, but I think that Tregelles is more explicit, and certainly more accessible, in justifying his methodology and theoretical approach. Another reason is that Tregelles is the most recent critical text that was not included in the triad of texts used to create Nestle’s first edition (Westcott-Hort; Tischendorf 8th; Weymouth [in itself the result of a comparison of editions]) or fourth (Weymouth replaced with Weiss).

Whilst working on the digital Tregelles, we were hoping / expecting that this would lead to a full-scale, fresh edition of the GNT, and this is reflected this in the abbreviations used for the Tregelles New Testament: TNT1 for the text that reflects the print as accurately as possible, TNT2 for the text in which obvious print errors are corrected. The next edition could then be TNT3 (Tyndale New Testament), which gives an instant ‘editional’ pedigree, but it turned out this name was never going to work. Regardless, there is some history behind our text.

For Tregelles see:

S.P. Tregelles, An Account of the Printed Text of the Greek New Testament: with Remarks on its Revision upon Critical Principles. Together with a Collation of the Critical Texts of Griesbach, Scholz, Lachmann, and Tischendorf, with that in Common Use (London: S. Bagster and Sons, 1854).

———. The Greek New Testament, edited from ancient authorities, with their various readings in full, and the Latin version of Jerome (London: S. Bagster & Sons, 1857-79).

———. Introduction to the Textual Criticism of the New Testament. Vol. 4 of An Introduction to the Critical Study and Knowledge of the Holy Scriptures, edited by Thomas Hartwell Horne and John Ayre, 12th ed. (Longmans, Green, and Co., 1856 [1869]).