Monday, September 18, 2017

The Problem of P38 and the ‘Western’ Text in Acts

In his published dissertation, Eldon Epp was interested in theological tendencies in Acts. In particular, he was interested in theological tendencies in the “Western” text. (From here on, I’ll forgo the quotation marks because they get tedious.) His method, however, was to study one particular witness of the Western text, namely, Codex Bezae.

The problem for Epp is that Codex Bezae could not be treated as a simple proxy for the earlier Western text. The reason is that Bezae sometimes reflected various accretions to the Western text. In order to address the problem, he compared Bezae “with those witnesses which, along with D, are recognized as the best ‘Western’ evidence” (p. 28). In this way, Epp could confirm where Bezae was or wasn’t likely preserving the earlier Western text-type.

Setting aside for a moment the risk of circularity here, I want to point out a serious problem with one of Epp’s key control manuscripts, namely, P38.

In discussing the proper use of D as a witness to the Western text, Epp cites P38 as a key reference point for determining whether a reading in D is, in fact, Western. He writes:
If P38, because of its earlier date, is ipso facto assumed more accurately to preserve the early “Western” text, then a comparison of D with this papyrus shows, as H. A. Sanders concluded, that “D is a very imperfect source for the ‘Western’, or second-century, text’. Granting this, however, it must also be emphasized that D and P38 show such a degree of agreement over against the B-text that the papyrus can be used, at the same time, to show that ‘the D text existed in Egypt shortly after A.D. 300’; A. C. Clark could call P38 ‘a text almost identical with that of D’. Codex Bezae, then, at many points is an imperfect witness to the ‘Western’ text, and yet on this account it does not lose its leading place among those witnesses.
Later, Epp cites D, P38, and the Harklean Syriac margin as “the outstanding ‘Western’ sources for Acts” and they form, with the (then) recently discovered Coptic G67 as an “élite group” (p. 31). Epp cites Clark approvingly that P38 has a text “almost identical with D” and Epp says this agreement is especially prominent “over against the B-text.”

The problem is that this isn’t the case when one compares these manuscripts in more detail. Here are the results from the recently-released ECM for Acts:

P38 03/B 05/D
P38 69.4% (43/62) 59.0% (36/61)
03/B 69.4% (43/62) 68.4% (3,514/5,140)
05/D 59.0% (36/61) 68.4% (3,514/5,140)

P38 is, of course, fragmentary, containing only Acts 18.27–19.6, 19.12–16. This means that there is far less text to compare with B and D. But the problem for Epp’s Western text should be obvious. Far from P38 showing strong agreement with D “over against the B-text,” P38 actually agrees more with B than with D! And yet, Epp says that P38 is a member of the “élite group” of Western witnesses.

Now, perhaps Epp would argue that these texts shouldn’t be compared in all these places in our effort to identify the Western text. But until we can agree what variant points should be used and why, we cannot agree on whether or not P38 should assigned to the same text-type as D. If it should not be, then it obviously cannot be used to confirm that D’s readings are early Western readings and Epp’s thesis will need some revision.

Perhaps the issue of definition will move toward some resolution at this year’s SBL meeting in the ECM sessions. We shall see. But those in attendance will certainly want to read the ECM’s article (which I haven’t seen yet) on the Western text along with Epp’s recent, data-filled argument in NovT for its existence there.


  1. Could you provide the numbers for the agreements between B and D over just Acts 18.27–19.6 and 19.12–16?

    Also, how many agreements of P38 with B against D, and agreements of P38 with D against B are there? I assume that a lot of the numbers given are agreements of P38 with both of those manuscripts against other witnesses.

    1. You should be able to cull that data from the CBGM website.

    2. I assume you are referring to the "phase 4" section of this website right?

      For my first question, I can see that it will give me the data for whole chapters of Acts, but it doesn't look like I can refine the search to a string of verses that start and end in the middles of chapters. I can expand the data to show all the places where the manuscripts differ, and just look for the verses in the right range. But I can't do that for the variants in which they agree.

      On my second question, I think I can see how I could go about that the long way, but not where I could bring up the data on a single screen.

      I think this kind of comparison with more than two manuscripts at a time might be easier with the hard copy ECM.

      Incidentally, the website gives 37/62=59.67% for the comparison of P38 and D, just a very slight difference from the table you provided.

  2. As Eric Rowe noted, rather than raw statistical percentages, it would seem that the actual points of importance are those where p38 and D agree (some of these also with B), regardless of their individual aberrancies that otherwise might skew the overall data. In other words, something more like a clustering approach.

    1. "that otherwise might skew the overall data"

      Ah, but therein lies the debate. Which selection of data actually skews things?

  3. It would seem obvious that clustering per se would not depend on individual MS aberrancies, but only upon shared readings among those MSS that tend to cluster in a particular manner. Why should that principle be considered skewed as opposed to methods that consider readings in a manner regardless of how they otherwise align or cluster?

  4. I figured that 62 variation units (plus a few units where P38 is extant, but ambiguous) would be a manageable number to check manually, so I went over them in the CBGM website to see which ones seem significant on internal grounds. For brevity, I've included only the units where P38 agrees (or in a few cases, very nearly agrees) with exactly one of B or D.
    (TL;DR: the data seems to agree with Epp's position. P38 and D agree on more genealogically-significant readings, while P38 and B tend to agree on polygenetic readings or on readings supported by nearly all other witnesses.)
    1. 18:27/35 - P38 and D both have an addition related to Achaia, while B and most other witnesses omit, so probably significant. (P38, D | B, +1)
    2. 18:27/36-42 - P38 and D share the significant substitution "in the churches" for the reading "to those who had believed," found in one form or another in all other witnesses. (P38, D | B, +1)
    3. 18:27/44-48 - P38, D, and a handful of other witnesses omit δια της χαριτος, while all other witnesses, including B, have the longer reading. Haplography does not explain the omission, so the relationship is probably significant. (P38, D | B, +1)
    4. 18:28/13 - P38, D, and a handful of other witnesses share an addition that includes διαλεγομενος, so probably significant. (P38, D | B, +1)
    5. 19:1/2-30 - P38 and D agree on a substitution that is so extensive and precise that it could not be accidental; this is clearly a significant agreement. (P38, D | B, +1)
    6. 19:2/34-40 - P38 and D share the substitution "that some are receiving [the] Holy Spirit" for "that there is a Holy Spirit". This reading seems to have arisen by assimilation to Paul's question earlier in the verse. Given this cause, it might be coincidental (notice that P41 attests to a transposition of this reading), but given its isolation to these particular witnesses, it is probably due to a correlation. (P38, D | B, +1)
    7. 19:3/18 - P38 and D have ελεγον, while B and all other witnesses have ειπον. It seems unlikely for multiple scribes to have changed the imperfect to the aorist here (especially with the aorist ειπεν at the start of the verse), so this is probably a significant agreement (P38, D | B, +1).
    8. 19:4/2-6 - This is the first agreement of P38 and B against D, but it is not a very impressive one; the only difference concerns the presence or absence of the article before Paul's name, which is a common and genealogically insignificant change. This is therefore probably not an agreement worth counting. (P38, B | D, +0)
    9. 19:4/42 - Another agreement of P38 and B against D, but again, the variant concerns the presence or absence of the article τον. This is not significant enough to count. (P38, B | D, +0)
    10. 19:4/44 - A more significant agreement of P38 and B against D; the first two, with the support of a handful of other manuscripts, read ιησουν, while the third reads χριστον with one other witness. We will count this as a significant agreement between P38 and B, although it is worth noting that many of the later MSS supporting the P38-B reading here also support the P38-D reading in 18:27/44-48. (P38, B | D, +1)
    11. 19:5/5 - P38 and B omit, while 05 reads τουτο. This is not a terribly significant variant, as the addition of a clarifying object is an easy one for a scribe to have made inadvertently (indeed, L587 adds ταυτα), and the agreement of P38 and B is far from exclusive, with nearly all other MSS sharing the shorter reading. For these reasons, we will exclude this agreement from consideration. (P38, B | D, +0)

    1. Edit: on point 7, the second sentence should begin, "It seems unlikely for multiple scribes to have changed the aorist to the imperfect here."

  5. 12. 19:5/14 - Another agreement of P38 and B against D, but again, the variant concerns the presence or absence of the article, this time του. This is not significant enough to count. (P38, B | D, +0)
    13. 19:5/19 - P38, D, and a handful of other witnesses have the longer reading του κυριου ιησου χριστου εις αφεσιν αμαρτιων, while B and nearly all other witnesses have του κυριου ιησου. The addition is extensive and precise enough to be considered signficant. (P38, D | B, +1)
    14. 19:6/8-14 - P38, B, and a handful of other witnesses read του παυλου χειρας, while D reads χειρα του παυλου. While the differences are minor, there are enough of them to count this as a significant variant. (P38, B | D, +1)
    15. 19:6/15 - P38 and D add ευθεως, while B and all other witnesses omit. Given the isolation of the reading, it is probably significant. (P38, D | B, +1)
    16. 19:6/28 - P38, B, and virtually all other witnesses read αυτους, while D has the singular reading αυτοις. Given the minute nature of the change and the ubiquity of the P38-B reading, this should not be counted as a significant variant. (P38, B | D, +0)
    17. 19:12/24 - P38, B, and the majority of witnesses read η, while 05 has the singular reading η και. The reading και is also reasonably well-attested, so the D reading could be a conflation. In any event, it seems clear that the D reading is an isolated error / interpolation derived from the P38-B reading and that the widespread support for the P38-B reading makes it less genealogically informative, so this is probably not a significant agreement. (P38, B | D, +0)
    18. 19:12/40-46 - P38, B, and most MSS add τα after τα τε πνευματα, while D and a handful of other witnesses omit. As noted already, the presence or absence of the article a matter of little genealogical information, but beyond this, the shorter reading could have risen multiple times independently by homoioteleuton. Finally, the widespread support for the longer reading makes the P38-B agreement here uninformative. (P38, B | D, +0)
    19. 19:13/6-8 - P38 and B read τινες και, while D reads τινες. The majority is divided between these two readings. Between this fact and the ease of scribal transition between the two readings, it's likely that the variant is polygenetic and therefore genealogically uninformative. (P38, B | D, +0)
    20. 19:14/21 - Both P38 and D make similar extensive additions here; the agreement is definitely significant. (P38, D | B, +1)
    21. 19:15/2-12 - D has a singular reading that exchanges the first two words of the reading shared by P38, B, and all other witnesses. The change seems to be the result of a conscious innovation on the part of D's scribe, but the ubiquity of the P38-B reading makes it effectively uninformative. (P38, B | D, +0)
    22. 19:16/6-26 - P38, B, and a handful of other witnesses disagree against D. The P38-B reading is one transposition away from the common reading, and the common reading is one prepositional substitution away from D's singular reading. While not of the magnitude of many P38-D agreements, this should nevertheless be considered a significant P38-B agreement. (P38, B | D, +1)
    In total, we have 10 significant P38, D | B splits, 3 significant P38, B | D splits, and 9 insignificant P38, B | D splits.

    1. Is there a typo in your last sentence?

    2. I had to check a few times to make sure, but I don't think so. Of the agreements between P38 and B against D, I found more of them to be genealogically insignificant than significant.

    3. Oh sorry, I misread it. I see now that you have 3 significant and 9 insignificant P38-B agreements. How many insignificant P38-D agreements?

    4. I didn't find any, due in part to the fact that almost all P38-D agreements against B concern readings not likely to have arisen more than once, and these reading generally have very exclusive support). In my estimation, their shared reading at 19:2/34-40 is one of the weaker agreements, and it just might be insignificant.

    5. Great writeup, Joey. Thanks!

  6. Paolo Trovato9/19/2017 4:45 am

    I find almost all the above remarks of JMC quite convincing, both when he considers genealogically significant an agreement and when he rejects it