Thursday, November 15, 2007

Mark-Q Overlaps V: the degree of verbatim agreement

This is the fifth in my current series of posts on the Mark-Q Overlaps (so-called), in which I would like to make a new point about the degree of verbatim agreement between Matthew and Luke in the pericope under discussion in the Q Section where I will be presenting, Matt. 3.7-12 // Mark 1.7-8 // Luke 3.7-9, 15-17 (John's Preaching).

5. The Degree of Verbatim Agreement

The point here is a simple one, but it is one that I have not seen anyone else make in discussion of the Synoptic Problem. The degree of agreement between Matthew and Luke in double tradition material is often too high for them to have been copying from another source. John Kloppenborg helpfully draws attention to this "extremely high verbal correspondence . . . in a significant number of instances" (John S. Kloppenborg, "Variation in the Double Tradition and an Oral Q?", ETL 83 (2007), 53-80, 53) including the one currently under discussion:
Matt 6.24 // Luke 16.13 (98%), Matt 12.43-45 // Luke 11.24-26 (93%), Matt 11.20-24 // Luke 10.13-15 (90%), Matt. 3.12 // Luke 3.17 (88%), Matt. 12.27-32 // Luke 11.19-23 (88%), Matt 23.37-39 // Luke 13.34-35 (85%), Matt. 3.7-10 // Luke 3.7-9 (85%).
These figures "are based on the number of common words divided by the total number of Lukan words" (53, n. 1). Kloppenborg helpfully draws special attention to our pericope:
In the last named pericope, Matthew has 76 words in Greek, 61 or 80% of which are identical with Luke in lexical form and inflection. This would rise to 63 or 83% if καρπόν and ἄξιον are included as agreements. Luke's version has 72 words in Greek, 61 or 85% are identical with Matthew, 63 or 87.5% if καρπούς and ἀξίους are counted as agreements. (53)
As Kloppenborg rightly makes clear, "the extraordinarily high degree of verbatim agreement" here makes theories of an oral mediation of such material impossible. And one might add that there are some striking verbatim strings of agreement here, of 12, 12.5, 20 and 24 words. The latter string, of 24 words, is Matt. 3.9-10 // Luke 3.7b-9.

In the handout that I have produced for the SBL session on Mark-Q overlaps where I will present my paper on this, I have produced a synopsis of Matthew, Q (IQP) and Luke on this so that one can see quickly and easily just how much verbatim agreement there is in this passage. Indeed, what is remarkable is how little disagreement there is. I have coloured all the agreement grey, and there is very little white left. One of the reasons for doing this in three columns is to remind ourselves of the fact that on the Q theory, what we have here are triple agreements. We get so used to thinking about this material as "double tradition", in the sense that it is present only in Matthew and Luke among the extant texts, that we can easily forget that for Q theorists, the agreements here are triple agreements, between Matthew, Q and Luke.

This leads us to an interesting question about the degree of verbatim agreement here between Matthew, Q and Luke. Does one ever see this kind of agreement between Matthew, Mark and Luke? Are the triple agreements in Matthew, Mark and Luke similar in extent to the triple agreements between Matthew, Q and Luke? The question that I am asking here is, I think, a new one. If anyone else has asked this question, then I have missed it. The question itself is an important one, so let me take a little time asking it in another way so that we can be clear about what is at stake.

According to the Two-Source Theory, Matthew and Luke are both independently using Mark and Q. We have access to Mark, so we have an idea what Matthew and Luke look like when they are working from a shared source. We know the degree of verbatim agreement to expect. Our question, then, is whether the degree of verbatim agreement is similar when they are using Q. In his recent ETL article, Kloppenborg reproduces a chart from Charles E. Carlston and Dennis Norlin, "Once More -- Statistics and Q", HTR 64 (1971): 59-78 (71):

Triple Tradition Matt Luke Avg. Double Tradition Matt Luke Avg.
Narrative 50.2% 46.9% 48.5%
55.7% 51.8% 53.7%
Words of Jesus 63.5% 68.3% 65.8%
69.5% 73.6% 71.5%
Misc. words 56.7% 60.6% 58.5%
87.5% 80.9% 84.1%
Average 56.0% 56.0%56.0%
69.8% 72.2%71.0%

For those who find figures instantly off-putting, let me express this in words. Matthew and Luke show consistently higher degrees of verbatim agreement in double tradition than they do in triple tradition. One cannot say in response to this, "But this is because double tradition is primarily sayings material" because the pattern is the same with respect to triple tradition sayings vs. double tradition sayings as it is with respect to triple tradition narrative vs. double tradition narrative and so on. Carlston and Norlin sum this up by noting that "the use of 'Q' is even more conservative than the use of Mark, possibly something like 27 per cent. more conservative" (Carlston and Norlin, 1971, 77). This is an anomaly on the Two-Source Theory. Why should Matthew and Luke apparently be so much more conservative in their use of Q, not least given their known respect for Mark's order?

The point of interest here is that the statistics make sense on the assumption that Luke is borrowing directly from Matthew in the double tradition (and Mark-Q overlap) material. They cohere with a scenario in which the double tradition is due to direct borrowing, Matthew to Luke, rather than mutual use of a shared source.

James Robinson once hinted that the all important clues to Q's existence might show up early in the document. I think Robinson was right. The remarkably high degree of verbatim agreement that shows up right at the beginning of Q is an important clue to the identity of the material as a whole. Here, as often elsewhere in Matthew and Luke, the agreement points to direct borrowing by Luke from Matthew, and not mediation via an unknown, hypothetical source.

Note: Carlston and Norlin's figures were criticized by Sharon Lea Mattila and subsequently revised downwards by them, but with the same relative degrees of agreement. Moreover, Carlston and Norlin noted that the same observations hold true with respect to the figures produced by Honore in 1968. (The issue relates to how one counts. Does one count only identical lexical forms, in the same number, case etc.? Can one count synonyms, etc.?). Mattila's 2004 article further criticizes the Carlston and Norlin case, but there are some difficulties with Mattila's re-count which I hope to outline on another occasion.

9 comments:

Bob MacDonald said...

It being my first SBL, I may change all my plans for Monday and make it to your talk - but with my current emphasis on the Psalms, I don't know since I think you conflict with John Hobbins session - anyway, I hope for the possibility of finally saying hello. And maybe I can get to two places at once.

steph said...

Terence Mournet is helpful for his critique of methodology on statistics. I think charts are useful but limited unless you still examine each unit independently and in context. I agree though - the high level of agreement between Matthew and Luke doesn't count in favour of an inbetween document generally speaking.

It is interesting that while Luke might have preferred Mark as his framework, the chunks he might have copied from Matthew he treated with greater respect verbally. Mark was embarrassing and needed to be rewritten. He revealed details like family quarrels for a start...

Scott Ferguson said...

First a question. When Kloppenborg compares Mark as a source to Q, does he use a reconstructed Q?

It seems to me that the next important step is to compare Luke's pattern of usage of Mark's material with the pattern that emerges if we assume that Luke is using Matthew as a source. If the statistics are similar between the Mark-as-source hypothesis and the Matthew-as-source hypothesis, then you have something pretty strong. If the patterns of Luke's theoretical usage diverge then you have to come up with further theories of why this is the case and Ockham's Razor may need to be sharpened a bit. Of course, Q has it's share of multiplying entities.

Jim Deardorff said...

In my comment to your previous blog, I noted that the modified Augustinian hypothesis (MAH)has a plausible explanation for this high Mt-Lk Q-verse verbal agreement. It comes about because of the role played by the translator of Hebraic Matthew into Greek, within the MAH framework.

The same kind of analaysis for the frequency distribution of strings of identical words leads to the same conclusion for the verbal qgreement between Mark and Matthew. This is also explained as the handiwork of the translator of Hebraic Matthew. As would be expected, this kind of verbal agreement does not exist between Mark and Luke, however.

I believe that Jack Poirer has also come to the same conclusions from his verbal-agreement studies of the word-string frequency distributions.

Jim Deardorff said...

In reply to Steph's comment, I don't see any logic ot purpose in the writer of Luke, if he extracted portions of Matthew that were omitted from Mark and carefully placed them in different contexts, then taking care to copy Mztthew's Greek so closely in his extractions.

Hence it's easier to see someone else doing this careful copying, namely the translator of Hebraic Matthew into Greek, by the modified Augustinian hypothesis. By the MAH, the translator had a plausible reason for his editorial behavior.

steph said...

In reply to Jim: I was merely noting that on the Mark without Q hypothesis, Luke uses Mark as his framework but according to statistics has higher verbal agreement with Matthew than Mark when he is not following Mark.

As far as taking care to copy Matthew's Greek carefully, that obviously didn't happen in every case and the units deserve to be examined separately and variations and agreements studied in context.

I don't think it follows that if Luke used both Matthew and Mark as sources, he would have used them in the same way, if that is what Scott is suggesting. He may use Mark for his structure but use extra sayings and more elaborate narrative from Matthew where he preferred the writing. His different approach to each source would not compromise Ockam's razor and necessitate extra entities. Nevertheless, while Ockam's razor strives for simplicity, it sacrifices detail and ends up delivering a "workable hypothesis" which is not a reflection of historical reality.

T LEWIS said...

Mark, I have often tried thinking this way in the past but unfortunately such a high Mt-Lk verbatim agreement still does not even make sense of Luke's author (someone who at other times clearly knows how to put things into his own words in varying degrees and not simply copy out huge chunks verbatim). I'm afraid it's still anomalous on the Farrer theory! I mentioned this recently as one of my eight myths/missasumptions on my almost retired blog: www.sourcetheory.blogspot.com/
Tim Lewis

Mark Goodacre said...

Thanks for all these comments. I am currently expanding the argument of this section of the paper into an article of its own.

Tim: I don't know that there is a problem for Luke as an author who varies the degree of his copying. There is a continuum from high verbatim passages to low verbatim passages, with many different levels in between. We know that Luke is a versatile author with a large vocabulary, and we know (on the assumption of, say, the Priority of Mark) that he sometimes copies directly from his source material. His versatility is expressed in the different extents of his use of his sources at different points in his narrative.

steph said...

I'm still troubled by what I think is the logic of this argument. Appealing to the scientific laws of probability doesn't account for the human persuasions of the authors and it doesn't allow for the intricacies of each little pericope. Also the bits of "Q" with such a high level of verbal agreement do happen to be sayings and the authors may tell the same story in a different way (when following Mark) but when quoting words of Jesus (or John), may try to be precise. It gets confusing to me to apply statistics to the Synoptic Problem.