Collation of NA28 and THGNT

The introduction at the back of the THGNT teases us by noting a collation done with NA28, but gives little more detail. That’s why it’s very helpful that Hefin Jones has shared a computer collation of NA28 with THGNT on Facebook. He says, “Caveat emptor: It’s automatically generated. There will be some curious items in there.” From there, Nelson Hsieh, a doctoral student at Southern Seminary, expands:

Thanks for producing this. I’m using Accordance for my own collation as well, but adding a lot more. Some problems with the Accordance collation vs. what I’m producing:

1. Accordance can mark accentuation and punctuation differences, but you turned those features off since it would produce thousands of more differences. I’m not sure what font size you used, but when I created a collation that includes punctuation and accentuation, the collation came out to 522 pages vs. the 94 pages you collated. I can’t get mine to look just like yours, so it might be less than 522 pages, but it will still be much longer if you add accents and punctuation. I’ve looked at the accentuation differences and it does become significant, for example, with liquid verbs, where the difference between the present and future tense is just accentuation, an acute vs. a circumflex accent (see Rom 2:16; 8:34; 1 Cor 3:14). And punctuation differences become significant, for example, with questions (see Matt 6:31; Mark 7:18-19; 8:18; Rom 11:24; 1 Thess 2:19; Heb 2:2-3; 9:13-14; 2 Pet 3:11-12; 1 John 4:20). Or check out John 1:3b-4 as well on the placement of a period.
2. The Accordance collation does not do a good job handling word order differences. Take a look at what the collation produces at, for example, Matt 14:4; 15:30; 22:43; Eph 6:8; 2 Tim 1:10; Heb 3:13 vs. the textual differences themselves. Writing out the variant in context allows you to see that it is a word order difference. It’s hard to identify and compare word order differences in the Accordance collation.
3. The Accordance collation cannot compare paragraphing/macro-structural differences. See, for example, John 1:1-18, the so-called “Prologue” of John. Peter Williams, co-editor of the THGNT, wrote an article, “Not the Prologue of John,” JSNT 33, no. 4 (2011): 375-86, where he argues that 1:1-18 is not really a prologue based on how ancient MSS structured the text, which informed paragraphing choices for the THGNT in John 1:1-18. THGNT has a paragraph beginning at v. 18 (not v. 19 like in NA28) and that paragraph continues until v. 20.
4. The Accordance collation cannot compare certainty levels, esp. important in the Catholic letters with the 43 diamond readings from the ECM. My collation will compare certainty levels and include UBS ratings.
5. The Accordance collation cannot compare the apparatuses of the THGNT vs. NA28 and their use of vid. in the citation of MSS. This is one area where the THGNT apparatus is better than the NA28 apparatus. THGNT is more transparent and will use vid. and transcribe the variant in MSS where the NA28 does not use vid., which gives the impression that the reading in the MSS is clear (compare apparatuses in Matt 5:22; 10:2; 13:40; James 4:9; 1 Pet 3:1).
6. Somehow the Accordance collation completely misses the significant variant in John 1:18. It just notes that THGNT has the article and NA28 lacks it. But the real variant is: ὁ μονογενὴς υἱός (THGNT) vs. μονογενὴς θεός (NA28). Although Dirk Jongkind told me at SBL that he regrets the textual decision here and would like to change it. But this example shows that Accordance can make mistakes and not display important variants in a meaningful way. I’ve linked to a PDF below to compare and contrast what kind of collation I am producing vs. what Accordance produces. Not every variant will be that detailed in my collation, but I will be listing witnesses (so you can evaluate quickly, for example, where 01, 03, and Majority Text stand) and I will describe the issues, so that you can search for every instance of differences in word order, verbal aspect, verbal voice, verbal mood, liquid verbs, adding the article, particle, conjunction, etc.
7. Overall, the Accordance collation can give you a big picture sense of differences, but you still need a human to categorize the differences, pick out the more significant differences, provide some context for each difference, and summarize the results. I’m presenting my paper comparing the THGNT vs. NA28 at SBL Midwest regional on Feb 7-9, so I’ll post a draft around that time. I’ve also attached a PDF of my collation for Hebrews to give you an example of what I’m working on. Bold Scripture references mean I think they are significant differences. 

On another post, Nelson says this:

I’m working on a full collation of textual differences, differences in certainty levels (esp. for the Catholic Letters), and differences in orthography for the SBL Midwest regional meeting. I’ve collated most of Paul, Hebrews, about 1/3 of the gospels, all the catholic letters. I’ve got 16 pages so far and expect the textual differences to reach maybe 35+ pages. I think the total number of textual differences (excluding orthography) could reach up to 500 differences or more. For example, I found 50 textual differences just in Matthew. But most differences will be minor. Here are 11 bullet points to summarize the differences between the THGNT and the NA28: (1) the most significant textual differences (from my perspective) are Matt 19:9; 27:16; John 1:18 (although Jongkind told me he regrets the textual decision here); Rom 5:1; Eph 5:22; 1 Pet 4:16; 2 Pet 3:10; Jude 22. Less significant (but still grammatically or theologically interesting) are Matt 17:9; 27:24; Rom 8:11; 1 Cor 2:1; Gal 5:21; Col 4:8; 1 Thess 2:7; 2 Thess 2:13; Heb 9:11; 11:11, 37; 1 John 2:20. (2) On different accentuation of liquid verbs (creating the present vs. the future tense), see Rom 2:16; 8:34; 1 Cor 3:14. (3) On different punctuation of questions, see Matt 6:31; Mark 7:18-19; 8:18; Rom 11:24; 1 Thess 2:19; Heb 2:2-3; 9:13-14; 2 Pet 3:11-12. 

And he continues with more detail from there.

Thanks to both for sharing these. They are both very helpful.

Collation Advice from the Past

Following from Tommy’s recent transcription suggestions, here is some advice on the subject from A. A. Vansittart, written to Hort on October 15, 1869.

…. How I wish I had seriously taken to collating and the like when I took my M.A. degree! Then I might have been able to follow your plans of collation which are in many respects admirable. Now alas I have 45 strong reasons against it! But I think I should recommend it to any young man beginning betimes: only with two modification. First I should impress on his mind always to collate to the best text within reach: never for instance to use a Lloyd’s Testament if he could beg, borrow, or steal a ‘Tregelles’. The best plan I think is what Wright was doing this year with his Chaucer, to take or make a text and have a lot of copies printed (with large margin, on writing paper: neglect nothing which may help one to write with speed what can be read with ease) and collate two or three MSS in each of them. Secondly I should decidedly recommend the use of coloured inks. They lose no minute of time: and they gain distinctness which is an equivalent of time: very likely they may save you from the dilemma of either having to do the work of weeks over again or not being able to rely on it. But perhaps I may have misunderstood your monochromania: perhaps it may bear the innocent nay laudable meaning that one should only write with one ink at a time? …

Given that Vansittart wrote this from the Hotel du Louvre in Paris, I think I would add one more tip: always try to do your collating from nice hotels in Paris.

Report from the Digital Collation Conference in Münster

The following is a report from Peter Gurry who attended the Research Summit on Collation of Ancient and Medieval Texts in Münster on 3-4 October.

* * *

A few weeks ago I attended the Research Summit on Collation of Ancient and Medieval Texts in Münster, Germany and I thought I would offer a brief summary of some of the papers. The conference was designed to introduce textual scholars to the ins and outs of electronic collation in general and CollateX in particular. The first day was primarily focused on papers from invited speakers and the second day was set up to be more hands-on with CollateX. Readers of this blog will be interested to know that versions of Collate have been used for the Editio Critica Maior (ECM) since 1 John (published in 2003).

The first presentation was from Caroline Macé of the Universität Frankfurt who spoke about her experience editing Gregory of Nazianzus. She spoke about the choice between collating and transcribing and suggested that the right choice depends on the purpose of the edition being made. For Gregory, she had around 140 manuscripts and decided that transcribing these would have been too much work with too little benefit. Her own preference, in fact, would be to have automated transcriptions from digital collations rather than automated collations from digital transcriptions.

Next up was Peter Robinson whose pioneering work as a student at Oxford in the 1980s led to the first version of Collate (history here). Robinson spoke about misconceptions of digital collation, the main one being the belief that the computer does all the work. In actual fact, Robinson wrote Collate only after becoming dissatisfied with other collation software because he felt it was too mechanical; he wanted something that required editorial input during the collation process itself. He went on to argue that the purpose of a digital collation should not simply be to record differences but to use those differences to understand the relationships of witnesses. Like the CBGM, Robinson wants to use all textual variants for genealogy rather than just a selection. The use of complete collations is what led to a revision in previous genealogies in the recent electronic edition of Dante’s Commedia. 

The third presentation was offered by Klaus Wachtel and David Parker about their use of Collate for the ECM. For John’s Gospel, the team in Birmingham has incorporated Collate into their own editing software (mentioned here) which allows them to move from transcriptions, to regularization of spelling, to construction of the apparatus, all in one place. It was impressive. In all, Parker said that the new software has made constructing the apparatus faster and more accurate. If my notes are right, he said it took them about 6 months to construct a full apparatus for the Greek witnesses of John.

Barbara Bordalejo presented next on the praxis of collation and gave some fun examples of how hard but also important it can be to electronically encode the complexities encountered in a manuscript. She showed examples of the change in the first draft of the Declaration of Independence including the change discovered in 2010 from “our fellow subjects” to “our fellow citizens”—a small change that makes a big difference! (But given my current home I shall say no more about that.) At the end of her talk there was a brief but lively back-and-forth over whether an expunction dot should be marked in a transcription as a “deletion” or as “marked for deletion” in order to distinguish it from the ways other scribes in the same manuscript deleted text.

The final talk was offered by Ronald Dekker, one of the programmers behind CollateX, who talked about some of the principles behind the software’s collation algorithms. The hardest part, as any human collator knows, is deciding how to segment the texts for comparison; the actual comparison is the easy part. Peter Robinson told us at one point that only about 1–2 percent of his original code was actually for comparing the texts; most of the rest was used to identify which parts of each text to compare with each other (a process known as “alignment”). Dekker illustrated the complexity of programming these decisions by showing that two witnesses with 100 segments (or “tokens”) could potentially produce as many as 10,201 possible points of disagreement (or “nodes”).

Unfortunately I had to catch a train the next morning so I wasn’t able to attend the second day of the conference. But the first day provided a good sense of where digital collation is and how it is being used. And as always, it was good to meet and talk with scholars editing a variety of other texts. The only real disappointment for me was learning that the location had originally been set for Iceland. I guess there’s always next time.

Finally, my thanks to Joris van Zundert and Klaus Wachtel for all their behind the scenes work in organizing the conference for us.