Saturday, March 17, 2018

Alan Taylor Farnes on Scribal Habits in Copies with Extant Exemplars

We are delighted to feature the newly baked Dr. Alan Taylor Farnes in this guest blogpost where he summarizes his work on scribal habits in copies where the exemplar is preserved. Well done Dr. Farnes!

Scribal Habits in New Testament Copies with Extant Exemplars

As many of you may know, I have recently completed my dissertation at the University of Birmingham. The following is a summary with some conclusions, ramifications, and next steps to take.

In 2007, James R. Royse published his exceptional study on the scribal habits of six early New Testament papyri. In his work, Royse revolutionized text critics’ understanding of the text critical canon lectio brevior potior or, “the shorter reading is preferred”[1] by demonstrating that the scribes he studied tended to omit more than they added. In its place he coined a new canon which he called lectio longior potior or, “the longer reading is preferred.”

One disadvantage of Royse’s method is, because the papyri he studies had no known exemplar. he was forced to reconstrcut what the hypothetical exemplar probably said and then determine how the scribe copied the hypothetical exemplar. This is obviously a completely normal procedure in textual criticism. Royse admitted that his method had flaws and called for an examination of another set of manuscripts—those with surviving exemplars. Royse wrote: “there has been (it seems) a failure to explore the problem of scribal habits for the text of the New Testament in the best possible case, namely where the Vorlage of an extant manuscript is also known to be extant. In such a situation we can virtually look over the scribe’s shoulder and compare the text he is copying with his result.”[2]

My research has attempted to support or disprove Royse’s new text critical canon that, in fact, the longer reading is preferred. Rather than analyzing early papyri for which no exemplar remains, I chose to identify and analyze manuscripts which have a known exemplar.

I have therefore identified twenty-two New Testament manuscripts which have known extant exemplars (see Table here). Of these twenty-two I chose four manuscripts, which are italicized in the Table, and their copies to transcribe, collate, and analyze to determine how well the scribes copied the text of their exemplar.

In the past, unfortunately, manuscripts with a known exemplar have actually been tossed aside and ignored because they seemingly provide no new information. Previous scholarship has attempted to locate manuscripts which are direct copies of a known exemplar in order that they may discount and ignore them from text-critical consideration. My motivation is exactly the opposite. By locating and analyzing these directly copied manuscripts we gain special insight into the scribal habits of the copy. Rarely have previous scholars employed direct copies as a way to understand scribal habits better. Eberhard Nestle typified the attitude that has largely been held toward direct copies when he wrote of 0319: “in the Greek merely an incorrect transcription of [06], and may therefore be dismissed.”[3] Hort agreed saying “These instructive phenomena naturally receive little consideration now, because the exact knowledge that we possess of the original [Claromontanus] renders attention to the copy [Sangermanensis] superfluous.”[4] While it is correct to exclude a manuscript from critical editions when we can know that it is a copy, it is not sufficient to ignore them altogether since they provide a unique glimpse into scribal activity. The tradition of ignoring direct copies has persisted for some time. Kirsopp Lake ignored 205 because he thought it is was copy of 209, saying: “It is for this reason that no further notice has been taken of 205.”[5] But, unfortunately for Lake, more recent scholarship thinks that 205 was not actually a copy of 209 but that they were simply very closely related. Most believe that 2886 is a copy of 205 as is found in the Liste. For this reason, Amy Anderson ignored 2886 from her study of Family 1 in Matthew.[6] Frederick Wisse did the same.[7] But now more recent scholarship believes that 205 is actually a copy of 2886.[8] The only time a manuscript which is a known direct copy of an extant manuscript should be excluded from text critical consideration is in the formation of critical editions. Above all, such manuscripts are invaluable in revealing scribal habits.

But we must carefully determine which manuscripts are indeed direct copies of extant exemplars. I have therefore formulated six steps to ask of a manuscript to determine if it is indeed a direct copy of an extant manuscript: (1): Does the proposed copy share a high percentage of textual agreement with another manuscript? (2): Do these manuscripts share a good number of peculiar dual agreements, or readings which are found only in these two manuscripts? (3): Historical considerations: can one of the manuscripts be demonstrated to be older than the other or were the two manuscripts created contemporaneously to each other? This is important in order to discern the direction of borrowing between the manuscripts. (4): Paleographical concerns: is there any evidence from the appearance of the text itself that one is a copy of the other? If perhaps the proposed exemplar is damaged or faded in a certain location, perhaps the proposed copy will show difficulty or commit an error here. (5): Corrections: does the proposed copy stumble over corrections in the exemplar or show their hand in any way? (6): Codicological concerns: do the two manuscripts share similar formatting, i.e. line breaks, page breaks, columns, pages, etc.? The last three steps provide almost irrefutable evidence that one manuscript is a copy of another. If, however, no paleographical evidence corroborates the textual evidence then concluding that a manuscript is a direct copy of another should only be done tentatively while accepting that the two manuscripts may actually be sibling manuscripts or simply very closely related. Such is the case in my analysis of two manuscripts previous mentioned: 205 and 2886. Many have believed that 2886 is a direct copy of 205 but recent scholarship has argued for a reversal of the dependence that 205 is a copy of 2886. Because I have not been able to find paleographical evidence that one of these manuscripts directly depends on the other, I have concluded, following Josef Schmid, that they are likely sibling manuscripts descended from a now lost exemplar.[9]

Codex Claromontanus (06), copied in the fifth century, is the earliest extant manuscript with extant copies being copied not only once but at least twice by Codex Sangermanensis (0319, ninth century) and Codex Waldeccensis (0320, tenth century). I have transcribed, collated, and analyzed all three of these codices by test passages in order to determine the scribal habits of the scribes of 0319 and 0320. All three of these manuscripts are Greek-Latin diglots. There were two surprising conclusions from the analysis of these manuscripts: first, the scribes of both direct copies neither added nor omitted any words. They broke even completely on word count. Not only did they break even but they had no variants of adding or omitting any words. They copied their exemplar almost exactly. They did make many substitutions, spelling errors, and nonsense errors, but they did not add or omit a single word. Which leads to the second surprising conclusion: these scribes were this accurate because of their ignorance of Greek. These manuscripts are diglots with Greek and Latin and an analysis of the scribal habits shows that these scribes were more proficient in Latin but had very little knowledge of the Greek language. They therefore copied the text extremely accurately but when they made a mistake it was an egregious mistake which usually resulted in such a nonsensical error that the result was not even a real Greek word. I have therefore tentatively concluded that scribes who do not know the language they are copying may copy extremely well for the most part but when they reach a difficulty they may produce an extremely obvious error. More research on the habits of scribes copying a language with which they are unfamiliar is needed.

The Latin scribal habits of the scribes who copied Claromontanus were more in line with what we would expect to see in light of Royse’s new text critical canon. On the Latin side of the page the scribes—who were the same scribes who copied the Greek text in the case of 0319—lost words on the whole and made the types of errors that scribes who know the language are prone to make. Most of the variants between Claromontanus’ Latin text and its copies were a result of the copies updating their text to match the Vulgate. This resulted in numerous substitutions but very few nonsense errors.

Another manuscript pair are two catena manuscripts of John: the 10th century 0141—the exemplar—and its 16th century copy, 821 copied by a man named Camillus Venetus and commissioned by Cardinal Francisco de Mendoza. 0141 seems to actually have two extant copies being copied also by the 16th century 1370, but I was unable to procure digital images of this manuscript and I was unable to travel to Berlin to transcribe it in person. I hope to complete the study of this family in the near future. Venetus acted as we would expect most Greek scribes to behave. Over the course of more than 4,000 words he added one word and omitted seven for a net loss of six words. Therefore, while being incredibly accurate, on the whole Venetus lost more words than he added and had an error rate of 2.22 total variants per thousand words. This is the lowest error rate of any scribe in this study and in Royse’s study. But this is not unexpected since it is likely that a 16th century scribe will copy better than an earlier scribe. In fact, the data bears this out: the error rate of the scribes in this study decreased as the centuries rolled by—meaning, as time went on the scribes became better copyists and the text became more stable.

I also compared the error rates found using this method to those found if I were to use the method employed by James Royse who tabulated only singular readings—or, readings which are only found in the manuscript he was analyzing. I found that his method was less accurate than this direct copies method. Sometimes the inaccuracy was minor but in the case of 821’s relationship to 0141, Royse’s method yielded an error rate which was less than half of the scribe’s actual error rate.

I have found that the scribes in this study did their best at a difficult job of copying manuscripts. At no point did I find any example of intentional corruption for theological reasons. The most blatant intentional changes found in this study were the Latin scribes copying 0319 and 0320 who consistently altered the wording from Claromontanus’ Old Latin to update their text to the Vulgate. Additionally, some of the scribes in this study lost words on the whole, as did Royse’s, but other scribes broke even. None of the scribes in this study gained words on the whole. Therefore, with respect to the scribes in this study, we can reject the older canon lectio brevior potior. We are unable, however, to confirm Royse’s new canon of lectio longior potior but rather I caution that length should not be used in any way to determine which reading is more original. Here I add my voice to Stephen Carlson’s[10] and Peter Malik’s[11] that length is not a valuable metric for determining which reading is more original.

Another main conclusion of the dissertation was the role of the patron. We see hints of the hidden hand of the patron in Codex Sangermanensis (0319) which was actually copied by two different scribes. The two scribes who copied 0319 were very different individuals with different scribal habits. 0319A seems to have had at least enough Greek knowledge to pronounce (or mispronounce) words which led to many orthographic variants. 0319B, however, made only one orthographic variant. Neither 0319A or 0319B added or omitted any text. That 0319B was a better copyist with respect to significant variants and total variants further suggests that 0319B knew less Greek than 0319A. So 0319A and 0319B were very different people with different scribal habits. What is striking, however, are their shared scribal attributes. Both 0319A and 0319B usually followed the same corrector (06***). Additionally, they both ignored marginal corrections preferring instead only corrections in the main body of the text. These shared scribal attributes, in spite of these scribes’ distinct individuality, suggest that a patron was behind the production of this manuscript. The patron instructed both of these scribes on how to copy the text and what changes, if any, to make. We therefore see that patrons play a substantive role in textual transmission—a larger role than previously thought. We should therefore at each point of variation endeavor to determine who in the editorial process created each variant: the patron, a reader, or, only lastly, the scribe. I wholeheartedly agree with Barbara Aland that scribes wanted to copy as best as they could and that scribes would not have been authorized to make any changes in the text.[12]

There are many avenues for further study from this study. I would like to know more about the scribal habits of those copying a language with which they are unfamiliar and if my hypothesis holds true that non-native scribes will actually copy the text more accurately most of the time but will make rare but egregious nonsensical errors. 010 is a possible manuscript with which to begin.

We should be honest, however, about the limitations of this method. While this method gives great insight into copying procedures it is not without its shortcomings. For example, the earliest direct copy of the Greek New Testament is 0319 from the 9th century. Additionally, there are currently only twenty-two known directly copied manuscripts. Nonetheless, in order to best understand New Testament scribal habits, more attention should be paid to manuscripts which are direct copies. All twenty-two directly copied manuscripts should be wholly transcribed, collated, and analyzed for their scribal habits. Doing so will provide a much more complete picture of how scribes copied the New Testament text.

In the end, our analysis of these codices was greatly enhanced by access to their exemplar. Had we not had access to the exemplar but rather analyzed these codices based on singular readings alone, then our understanding of their scribal habits would be less accurate. While the singular readings method captures many scribally created readings we can be confident that the direct copy method captures all scribally created readings.

Alan Taylor Farnes is currently an adjunct instructor of Ancient Scripture at Brigham Young University. You can find him on Twitter @alanfarnes. His dissertation can be found as: Alan Taylor Farnes, “Scribal Habits in Selected New Testament Manuscripts, Including those with Surviving Exemplars,” (PhD dissertation; University of Birmingham, 2017).


  1. James R. Royse, Scribal Habits in Early Greek New Testament Papyri (NTTSD 36; Atlanta: Society of Biblical Literature, 2008), 732.
  2. Royse, Scribal Habits, 34.
  3. Eberhard Nestle, Textual Criticism of the Greek New Testament (tr. William Edie; New York: Putnam’s Sons, 1901), 77.
  4. Brooke Foss Westcott and Fenton James Anthony Hort, The New Testament in the Original Greek: Introduction and Appendix (New York: Harper and Brothers, 1882), II:255.
  5. Kirsopp Lake, Codex 1 of the Gospels and Its Allies (TS 7; Cambridge: Cambridge, 1902), xxii.
  6. See Amy Sue Anderson, “Codex 1582 and Family 1 of the Gospels: The Gospel of Matthew,” (PhD. Diss., University of Birmingham, 1999), 118-19; Amy S. Anderson, The Textual Tradition of the Gospels: Family 1 in Matthew (NTTS 32; Leiden: Brill, 2004), 116.
  7. See Frederick Wisse, The Profile Method for Classifying and Evaluating Manuscript Evidence (SD 44; Grand Rapids: Eerdmans, 1982), 106. 
  8. See Alison Welsby, A Textual Study of Family 1 in the Gospel of John (ANTF 45; Berlin: De Gruyter, 2014), 82-83 = Alison Sarah Welsby, “A Textual Study of Family 1 in the Gospel of John,” PhD diss., University of Birmingham, 2011, 122-24.
  9. See Josef Schmid, Studien zur Geschichte des griechischen Apokalypse-Textes, vol. I: Der Apokalypse-Kommentar des Andreas von Kaisareia. Einleitung (Munich, 1956), 288.
  10. See Stephen C. Carlson, The Text of Galatians and Its History (WUNT 2:385; Tübingen: Mohr Siebeck, 2015), 90.
  11. See Peter Malik, P. Beatty III (P47): The Codex, Its Scribe, and Its Text (NTTSD 52; Leiden: Brill, 2017), 114-15.
  12. See Barbara Aland, “Sind Schreiber früher neutestamentlicher Handschriften Interpreten des Textes?” in Transmission and Reception: New Testament Text-critical and Exegetical Studies (Jeff W. Childers and D. C. Parker, eds; TS 3.4; Piscataway, NJ: Gorgias Press, 2006), 114 and Barbara Aland, “Der textkritische und textgeschichtliche Nutzen früher Papyri, demonstriert am Johannesevangelium,” in Recent Developments in Textual Criticism: New Testament, Other Early Christian and Jewish Literature: Papers Read at a Noster Conference in Münster, January 4-6, 2001 (W. Weren and D-A. Koch, eds; Studies in Theology and Religion 8; Leiden: Brill, 2003), 36.


  1. Thank you for your study! I look forward to reading it in full. Out of curiosity, regarding your first criterion for accuracy, what percentage do you classify as high?

    1. Great question. I didn't set a limit but rather used a high percentage as a springboard to look at it with human eyes. The manuscripts I studied had about a 99% agreement.

  2. Anticipating Farnes by several years and similarly differing from Royse (based on my own dissertation research, 1982):

    "Neither the shorter nor longer reading is to be preferred" (MAR 2001 "Case" essay)

  3. Congratulations on the completion of your dissertation and PhD Alan! A fantastic piece of scholarship.

  4. Alan, this is extremely helpful; I didn't know you were working on this. I had been proposing this as a topic for a dissertation at Dallas Seminary for many years. One student started on it about a decade ago but didn't finish. When CSNTM went to Munich to digitize some of the Bavarian State Library's MSS we specifically wanted to digitize the Abschriften both for the reasons you articulated in this blog but also because of the theoretical possibility of seeing some Abschrift with leaves of text that the original no longer has. If the Abschrift is an accurate copy, then this could be used to push back the date of the reading to the earlier MS. Further, an Abschrift might make corrections to the exemplar (as Gurry implies in his NTS article on number of variants) where the exemplar simply made some nonsense alterations).

    Two other considerations going forward: in several MSS there are supplemental leaves. In some cases I would think that the supplements were made precisely because the original leaves were getting worn out, so the scribe makes an exact copy of the leaves that then must be thrown out. Using your criteria, perhaps we could determine with a large enough sampling whether such a supplement was indeed an Abschrift. Second, there must be a whole lot more Abschriften than have not been identified. But distinguishing between sister MSS and mother-daughter MSS is very difficult. I suspect that among the Kr group and others we have quite a few. Do you know of anyone doing research in this area? That so few NT MS Abschriften have been identified is almost baffling.

    When will your dissertation be published? I look forward to reading it!

    1. I believe identifying Abschriften would be far easier if one were to examine nonbiblical texts, where very often only one older manuscript may exist for a given work, yet several other later manuscripts exist that were very likely copied directly from the earlier exemplar.

      This could provide extremely useful comparanda for how scribes copied out the New Testament, with of course the caveat that scribes may act differently when copying a biblical text as opposed to when copying other Christian or even secular texts. Another issue is the overall skill of the scribe, perhaps scribes who copied secular texts may have, on average, been more skilled than the average scribes of biblical texts. Who knows?


    2. Thank you Dan. I agree, and as stated in the dissertation, I do think there must be many more Abschriften out there which is why I have given the criteria. I don't know of anyone doing more work in this area besides myself.
      Concerning supplemental leaves, I only have one example off the top of my head. There are supplemental leaves in the exemplar D06 and it is clear that these leaves were copied by the Abschriften. But that isn't exactly the situation you are thinking about.
      Dan, yes the full dissertation discusses non-Biblical Abschriften but doesn't go into as much detail as I had originally hoped due to scope and time considerations. There is an article by Rachel Yuen-Collingridge and Malcomb Choat that is helpful in this regard. I do point out many Abschriften identified in the LXX.
      The wheels of publication move slowly. The dissertation PDF should be available someday soon on the British Library ETHOS but for some reason it has taken a very long time to post online. I will alert the blog when it is publicly available.

  5. I found different dates for GA 018. The CSNTM dates it to the 9th century. Yet Wikipedia has a quote from Aland and INTF from stating a colophon dating it to 1364 A.D.

    I'm inclined to believe the CSNTM over Wikipedia, however I'm wondering what caused the change in dating the Manuscript to the 9th century, if it was indeed changed?

    1. Hi Marius,
      For 018 I followed the dating by Theodora Panella as found in her “Resurrection Appearances in the Pauline Catenae,” in Commentaries, Catenae and Biblical Tradition: Papers from the Ninth Birmingham Colloquium on the Textual Criticism of the New Testament, in association with the COMPAUL project (ed. H. A. G. Houghton; Piscataway, NJ: Gorgias Press, 2016), 122.
      She is the one who has identified that 018 and 0151 are manuscript pairs and dates them both to the 9th c. but she does not commit to a direction of copying. I have done no further research on this manuscript pair but am relying on Panella here.
      If 018 were dated to 1364 then it would help determine the direction of borrowing.

    2. Hello Alan,
      The mistake is mine. I confused 2 different manuscripts for 1. Minuscule 18 is dated to 1364 and uncial 018 aka Codex Mosquensis I is from the 9th century.


  6. Thank you for this important work.

    Do I understand correctly that the six steps listed (or other similar lines of reasoning) are the only means available for saying when a manuscript is a direct copy of another?

    Are there no cases in existence where the specific provenance of a manuscript is known historically with enough detail to say exactly what it was copied from (i.e. by its scribe explicitly saying so)?

    Regarding these 6 steps, I have two questions:
    1) How do these steps really narrow down the relationship between two manuscripts so precisely as to say one was directly copied from the other? I can see how they can prove that manuscripts are very closely related. But two manuscripts can also be very closely related by both being copied from the same exemplar, or by having other slightly more distant relationships that still put them quite close to one another. It seems to me that all 6 criteria allow for multiple possibilities.

    2) If two manuscripts are this closely related, how do you know for sure which is the exemplar and which is the copy? It seems like deciding this based on prior commitments to ideas about scribal habits would beg the question (i.e. if the two manuscripts, A and B, differ at a point where one has a mistake, is it because A did not have that mistake and B, in copying A made it, or is it that A had the mistake and B corrected it?).

    1. Hi Eric,
      I don't think I would claim that these are the only 6 six but that these are steps that I have taken to determine relationships and steps that could be followed by others. As Tommy points out below, perhaps the possibility of manuscript pairs increases when looking in the same repository.
      If there are instances of a scribe explicitly stating their exemplar I have never heard of such an instance in Greek New Testament manuscripts or in any form of copying. That doesn't seem to be something that any scribe is in the practice of noting. But I would be delighted to hear of examples.
      It is in the last 3 steps where dependence becomes more proven than suggested. At my 2016 SBL presentation and in the dissertation I show images where there is no doubt that the scribe is looking directly at the exemplar while copying. So if a scribe stumbles over corrections in the exemplar or if faded writing in the exemplar leads to an error in the copy then these are pretty solid evidence that they are indeed copies. Then historical evidence that the exemplar is hundreds of years older than the copy is useful to determine borrowing.
      When they are contemporaneous it is much more difficult. Such is the case with 2886 and 205. Many have argued for their dependence in both directions but at this point with the evidence gathered I think it is safest to say that they are siblings rather than parent-child and that they have descended from a lost exemplar. So it is precisely because I was unable to find sure signs from the last 3 criteria that I have taken a step back to concede that they are likely not a copy of the other.

  7. Yes Eric, I was going to suggest a seventh step (or criterion) relating to the third step (historical), and that has to do with provenance. For example, I pointed Alan to the MSS 1065 and 1068, which are deposited in the same Greek monastery. They were virtually identical in my collation for my Patmos project.

    1. Thanks. Yes, I did neglect to account for the third step in my questions. In cases where one manuscript is definitely older than the other, that narrows down the possible relationships I mention in my second question, although I don't know that it automatically narrows them down all the way to the single relationship of exemplar and direct copy.

  8. Here is a BIG question regarding Dr Farnes' chart: MS 2193 is cited as original and 872 as its copy. However, in the PA border regions (Jn 7.51-52; 8.12-13) both MSS differ significantly from each other (examples would take up too much space, but can be provided) -- it simply doesn't appear there that one was copied from the other. Any thoughts?

    Also (and only for the record): the presumed original 2193 lacks the PA in its main text, but it is inserted by a later hand at the end of Jn; in contrast, 872 includes the PA in its main text at the normal location, but from an entirely different textual line than that found in 2193.

    1. Hello professor,
      Concerning 2193's relationship to 872 here I relied upon Amy Anderson, “Codex 2193 and Family 1 in Mark,” in Studies on the Text of the New Testament and Early Christianity: Essays in Honor of Michael W. Holmes (NTTSD 50; Daniel M. Gurtner, Juan Hernández Jr., Paul Foster eds.; Leiden: Brill, 2015), 122–27. I did no further research into this manuscript pair aside from relying on Anderson.
      I state in the dissertation that for many of the 22 proposed Abschrift manuscripts in the chart I am simply compiling evidence from others who have claimed that they are copies. So Tommy argued for 1089 and 1218 and also for 1065 and 1068. So this list of 22 is an attempt to compile anyone's statement who has claimed that they have found an Abschrift. As stated in the dissertation, for most of these I have done no further research aside from noting that they have been argued for. The meat of my dissertation looked closely at 06 and its copies 0319 and 0320 in Greek and Latin, 0141 and its two copies 1370 and 821, and the tangled relationship of 2886 and 205 which I now believe not to be copies.

    2. The PA crept gradually in the Byzantine text. GA 2193 is adapted according later liturgical requirements. The anagnosmata numbers, for example, are added in the margin. Lesson 28, the PA, was not in the proper place and is added the end. (By the same hand? Later then GA 872 is composed?)
      The valuable discussions on the PA in GA 2193 and GA 872 in the dissertations of Welsby (2011, published 2104) and Koch (2013) need some additions and reconsideration.

  9. Speaking of scribes, now available online (thanks to Charlotte Hempel) is an illustrated lecture by Eibert Tigchelaar. "The lecture, titled 'Beautiful Bookhands and Careless Characters: An Alternative Approach to the Dead Sea Scrolls', focused on his recent work on the scribal hands of the scrolls and what they might reveal about who wrote the scrolls, particularly in light of several 20th and 21st century suspected forgeries."

  10. Does anyone know if there's been a study of the Peshitta MSS for which exemplars are extant?

  11. This dissertation is now freely available online.