Thursday, May 23, 2024

What level of confidence we should have in reconstructions of fragmentary papyrus texts?

15

With a fragmentary papyrus manuscript—as the majority of our NT papyri are—and a known text being copied, editors very often reconstruct the missing portions of text beyond the boundaries of the extant material. This is all well and good, and totally necessary. But one might ask the question as to what level of confidence we should have in the reconstructed (=non-extant) text?

A preliminary answer could be: “no confidence at all, it is not extant, reconstruction is basically speculation, try to ignore it.”

Another option might be: “well the scholar who did this had all the time with the actual manuscript, and had studied the general way of the scribe closely, and (potentially) long experience in such things, so it makes sense to trust it”.

But someone might say: “I wish we had a test case where a published text of a NT papyrus manuscript was later supplemented by the publication of a fragment or two which gave total clarity on the beginnings and endings of lines and could help us with an assessment of what level of confidence we should have in reconstructed texts.”

Recently I had occasion to look much more closely than I had before at two fragments of P66 from Köln (photos below) (remember that the later portions of P. Bodmer II was published by Martin and then in a revised edition by Martin and Barns in 1962). So we can compare the reconstructions (in John 19) of Martin & Barns with those of Gronewald.

 V. Martin & J.W.B. Barns, Papyrus Bodmer II, Supplément. Evangile de Jea chap. 14-21 (Geneva: Bibliotheca Bodmeriana; rev. 19622), 35-38.

M. Gronewald, ‘Johannes-evangelium, Kap. 19,8-11.13-15.18-20.23-24, Kölner Papyri 5’ Papyrologica Colonensia [Abhandlungen der Rheinisch-Westfälischen Akademie der Wissenschaften] 7 (1985), 73-76, 296-298.

Across the passages in John 19.8-11, 13-15, 18-20, 23-24 Martin (& Barns) was dealing with a fragmentary text (P. Bodmer II) somewhere in the middle of the four columns, and on that basis reconstructing total lines. With the new discoveries we had either beginnings or ends of lines for nine or ten lines in each of these columns.

The result of a comparison is that although Martin (& Barns) made some brilliant anticipations, the 1985 lines reconstructed by Gronewald match exactly the 1962 lines in only 18 out of 38 lines (47%). Although a fair number of the twenty are errors of word division of no great consequence; there are also new word order variations, spelling variations, additional punctuation, different cases, corrections, and one major difference in the text reading (in John 19.10 where alterations across three lines result in a quite distinct reading).

This test case confirms that we should be extremely cautious about appeals to reconstructed text in the use of fragmentary material.


15 comments

  1. Alexander Thomson5/23/2024 1:10 pm

    The brouhahas over such fragments as 7Q5 has made me sceptical - possibly permanently. And the bitter behaviour of some scholars was an eye-opener!

    ReplyDelete
  2. Alan Bunning5/23/2024 2:11 pm

    Indeed, having compared hundreds of New Testament electronic transcriptions there are many such variations in the editors’ supplied words. I think some editors are better at it than others. In particular, there are many notable differences between the supplied words in Comfort’s transcriptions and the INTF’s. Quite often either set is plausible, but sometimes I think it shows the editor’s bias for certain readings. The CNTR transcriptions follow a particular standard of where supplied words are offered, and I have later changed my mind about what they should read on more than one occasion. The CNTR, however, gives supplied words absolutely no weight in terms of processing. That is, except in cases of vid, which was not mentioned in the post.

    ReplyDelete
  3. A third approach is to accept the reconstruction as worthy of consideration, while refraining from accepting it as probable until having had a chance to consider the details of that specific case.

    Some reconstructions will prove themselves to be quite likely, while others will prove to be highly speculative. And in some cases we may be left, not with an ability to provide a highly probable positive reconstruction, but still have good reason to accept a negative one (i.e. no particular variant reading is supported, but one or more can be ruled out as impossible).

    ReplyDelete
  4. "A preliminary answer could be: 'no confidence at all, it is not extant, reconstruction is basically speculation, try to ignore it.'"

    Response: Inference is necessary for making any guess. Our confidence very may well be lower, but that doesn't negate the fact that without inferences Textual Criticism is worthless. Are we willing to deny that any other evidence beyond explicit text is worthless?

    "Another option might be: 'well the scholar who did this had all the time with the actual manuscript, and had studied the general way of the scribe closely, and (potentially) long experience in such things, so it makes sense to trust it'"

    Responses: Yes, but with varying levels of confidence depending on the strength of the inferences.

    "But someone might say: 'I wish we had a test case where a published text of a NT papyrus manuscript was later supplemented by the publication of a fragment or two which gave total clarity on the beginnings and endings of lines and could help us with an assessment of what level of confidence we should have in reconstructed texts.'"

    Response: This simply is a hasty generalization waiting to happen, and maybe even a composition fallacy. Just because something is true this one time does not apply to the rest of fragments. They are all diverse in a plethora of ways.

    "The result of a comparison is that although Martin (& Barns) made some brilliant anticipations, the 1985 lines reconstructed by Gronewald match exactly the 1962 lines in only 18 out of 38 lines (47%)....This test case confirms that we should be extremely cautious about appeals to reconstructed text in the use of fragmentary material."

    Response: This is the right posture --caution, not complete dismissal.Giving them "no weight" would be committing a serious fallacy.

    ReplyDelete
  5. This post raises some great questions, and I agree that we have to approach reconstructions of lacunose portions of manuscripts with a critical eye. In a couple recent papers (https://academic.oup.com/dsh/article/38/2/647/6936676 and https://brill.com/view/journals/nt/66/2/article-p257_6.xml), I've argued that we can assign numerical likelihoods to such reconstructions on the basis of measurable statistics regarding a manuscript's line lengths in letters. If a manuscript has been studied extensively enough to have tabulations of its line lengths readily available, then this approach can distinguish the relative plausibility of competing reconstructions of its lacunose passages. In my papers, I apply it to P46 (a manuscript with copious codicological statistics available thanks to the work of Ebojo, Peterson, and their predecessors) with promising results regarding its likely text in Eph 6:19 and other passages. As I show in the second paper, you can easily incorporate conjectures and otherwise unattested combinations of variant readings into the reconstructions that you evaluate.

    Obviously, as I make clear in the papers, there are limitations. If two competing reconstructions of lacunose text have the same number of letters (and have potential line breaks at the same places), then they will have the same likelihoods. In other words, an approach based on line lengths is really only robust to relative additions and omissions of letters, not substitutions or transpositions. Such an approach is also limited by what we bring to it: if we want to know if a lacunae once contained some text not found in any other witness, then we have to supply such a text. Ultimately, no reconstruction will ever be guaranteed, but with a statistical approach, we can at least quantify how (un)certain we can be about a given reconstruction.

    ReplyDelete
    Replies
    1. Stephen Carlson5/24/2024 10:41 am

      Yes, Joey’s technology is pretty cool here. It can indicate (and even quantify) which alternatives are relatively more likely than their competitors.

      Delete
    2. Alan Bunning5/25/2024 1:12 am

      The CNTR tries to handle the issue through various categories of vid, regarding readings or parts of readings that are highly probable. Although each reading could be given a probability, I don’t see much use for it in situations on the low end where say an unknown word could be 5 or 6 characters, and they both fit quite nicely in the gap. I don’t think line length probability would be very convincing in such cases, especially with manuscripts with line lengths that vary considerably. While each reading could be assigned such a probability, I don’t think that probability necessarily relates to the likelihood of a given reading in such cases, perhaps giving an unrealistic bias. I have a program that calculates line length stats on each manuscript, and the variance of line lengths can be quite large, and even change in different parts of a manuscript. I met someone who plans on using AI to predict the readings in lacuna based on a number of criteria, and will be quite interested in seeing what he comes up with.

      Delete
    3. Maurice A. Robinson5/25/2024 5:57 pm

      Not just papyri, but even among minuscules damaged portions exist that require intelligent reconstruction. In just the PA, for example, reconstruction was needed for portions of 11 MSS and 3 lectionaries; for these, I and Joey McCollum worked on reasonable solutions (thank you, Joey!)

      Delete
    4. Happy to help! I look forward to working with manuscripts of the PA with you again soon!

      Delete
    5. (Oops, forgot to comment using my Google account...the previous comment was obviously by me, haha)

      Delete
    6. Alexander Thomson5/26/2024 6:33 pm

      Joey,
      Now that you’re here, are you able to say when the revised Solid Rock Scholar’s Edition will appear, and whether it will include Scrivener 1881 and Tyndale House GNT’s?

      Delete
    7. Alexander,

      Thanks for your interest! I can say for sure that it won't be within the next year (as I'm still doing my PhD). But Stephen Brown has been adding to the list of things to add or change little by little. If I recall correctly, we'd talked specifically about waiting for the next edition of the THGNT to be released, because it would be good to include it. We listed some other editions that would be nice to add, but I don't remember what the rest were.

      Delete
    8. Alexander Thomson5/26/2024 9:02 pm

      Joey,
      Thanks! I look forward to publication! Please, please do include Scrivener 1881!

      Delete
  6. What Joey McCallum said, but louder.

    ReplyDelete
  7. P16 is an interesting instance. The ed.pr. assumes conflation at Philippians 4:7 when dittography is also possible.

    ReplyDelete