Evaluation of the Theory of Literary Dependence. 
(draft Jun 20, 1998)

My intention here is not to duplicate the work of those 'dissenting scholars' discussed by Wenham (see the preceeding article), but to try to survey the arguments and data and arrange the issues logically. In this way, the reader who wishes to see the trade-offs and areas for further investigation will have a background from which to proceed.



What specifically is LD?

At its simplest, it is the theory that a gospel writer composed his gospel by copying from, and modifying in the process, a written exemplar of different gospel(s) or source(s).

This source might be:

(A1) read aloud by someone else, while the author wrote;

(A2) read silently by the author, and then the 'new' gospel dictated orally to scribes; or

(A3) conceivably have been so familiar to the author (via VERY detailed memory that extended to linguistic forms and long stretches of order) that it was 'on the mind' of the author as he prepared his new work.

In the case of the Two-Source theory of Literary Dependence, this would either (B1) have the author looking at two documents in front of him, while he writes the third; or

(B2) have two readers, reading the documents orally to the author, who composes the new document after brief periods of reading.



What are the main arguments used by advocates of LD?

There are essentially three main arguments for LD: (1) overlap in content; (2) parallels in the order of passages; and (3) verbal agreements.

  1. Overlap in content: This argument points out the obvious overlap in the events described and the words recorded in the Synoptic gospels. This agreement of content is not considered a strong argument, of course, since this would be expected in ANY of the theories: Eyewitness, Oral Tradition, or Literary Dependence. As long as the same stock of history is being described we SHOULD expect a considerable sharing of stories. (Accordingly, we will not be discussing this aspect of the problem here.)


  2. Parallels in the order of passages: This argument is generally considered quite strong. Even Wenham, who has all but abandoned any belief in LD, considers this a strong indication of some kind of dependence (or at least interdependence). The thrust of this argument is that many of the various pericopes (i.e., narrative or literary sub-units) occur in the same order within the Synoptics, even though there does not seem to be any thematic or 'internal' clues/connections (e.g., chronological, locale, topical) that would suggest such an order. Wenham describes the logical force of this [RMML:43]:

  3.  
      "Where several pericopes, which have no apparent logical or chronological succession are found in the same order, a natural possible inference is a literary connection. Similarly, if a sequence of material is broken by an omission or by the intrusion of new matter and is then resumed again, a literary connection is a natural explanation. Such sequences are found at a number of places in the synoptics, especially in the triple tradition"

    This suggests that one of the gospels 'set the standard' by creating this order, and that the other gospels-having read that order-stuck to it.

    The case for 2GH is very strong here, in that Luke keeps to the order of Matthew in each 'strip' of his narrative: he uses a sequence, breaks it off for some of his own material, but RESUMES Matthean order later [BQI:Introduction].


  4. Verbal agreements: This argument makes the point that there seems to be a surprising (and non-trivial) number of agreements in specific language expressions between the synoptics. The example that James used, and that we started this article with, is considered to be an obvious case of 'borrowing' from one gospel to another. It has the appearance of being lifted right out of one gospel, and planted-with three changes-into another. Proponents argue that this level of "verbal agreement" is so high throughout the gospels as to make the allegedly 'looser' OTH much more improbable.
I have noted that only arguments 2 and 3 are currently used, so I will restrict my evaluation to these.




Evaluation of the argument from "Verbal Agreements"

The interesting thing here is that this has a VERY high level of ambiguity in it.

The standard method, of course, is to identify 'parallel passages' and then compare the words. But even this initial first step is fraught with difficulty-what exactly is a rigorous definition of 'parallel'?

For example, Neville points this out [NT:AOSSC:227-228]:

"In this connection one must also examine the criteria for determining true parallelism between pericopes. Synoptic critics currently work without clearly defined guidelines for discriminating true parallels from nonparallel pericopes..." And the old 'synopsis' diagrams have been shown to be biased, with the Griesbachian approach more neutral than the 2SH. So Neville [NT:AOSSC:236]: "The methodological procedure adopted by Lachmann, Kummel, Neirynck, and Tuckett (Tanknote: leading advocates of 2SH) determines to some extent what one's conclusion will be. Griesbach's procedure is more neutral because it does not preclude possible explanations of the synoptic data in the way that Lachmann's does." This problem is obviously highlighted by the simple fact of the nature of Jesus' teaching ministry. He undoubtedly taught the same truths to multiple audiences, with slight adaptations to local conditions, and with varying levels of detail and exposition. When two such similar sayings show up in the gospel, we would clearly be mistaken to presume that one was 'more original' than the other(!). Plus, the often generic (and therefore 'stretchable') semantic content of some of the public sayings (e.g., the "poor" vs "poor in spirit" quote given above) creates that same 'reverse engineering' problem.

But let's assume that we find a passage that is clearly parallel, determined by a unique episodal framework (e.g. baptism of Jesus), how should we proceed to analyze similarities and differences?

The first thing we need to consider is what literary 'borrowing' would look like. What kinds of text would indicate (clearly) literary identity. In the case of NT Greek, you have several components to work with: word substitutions, word order, word root, word endings, word omissions, word additions. In English, we do not have nearly as many word endings for nouns and verbs as does Greek, and few endings on modifiers (e.g., adverbial '-ly' or comparative senses, such as '-er' or '-est'). But let me take a stab at illustrating some of the factors. Consider the following lines:
 

  1. "The man took his mother to see her two cousins, but the hotel where they were staying was closed." (original)
  2. A man took his mother to see her two cousins, but the hotel where they were staying was closed. (modifier change, no real change unless rhetorical)
  3. The young man took his mother to see her two cousins, but the hotel where they were staying was closed (word addition, brings out detail).
  4. The man and his mom went to see her two cousins, but the hotel where they were staying was closed. (word order, plus verb change, different emphasis)
  5. The woman took her mother to see the mother's two cousins, but the hotel where they were staying was closed. (word root, major change in meaning, all else the same)
  6. The man took his mother to see her cousin, but the hotel where they were staying was closed. (ending change, plus word omission)
  7. The man took his mother to see her two cousins Polly and Molly, but the Excelsior where they were staying was closed. (word additions, plus word substitution)
  8. Joe took Martha to visit Polly and Molly, but the Excelsior wasn't open (massive changes-same event)
  9. The man took his cousin to see their mothers, but the hotel where they were staying was closed. (word order, but vastly different meaning)
  10. The man took his mother to see her two cousins, but the inn was closed (massive word changes, but clearly the same event)


Now, the question we are studying is simply this: given than author READ #1 in a manuscript above, which of the items 2-10 would be clear evidence of influence/borrowing?
  1. Could easily be seen that way, since the author could have decided to cast the story in a different light;
  2. Likewise, if the author had knowledge that the man was young, and he had a specific reason to change the original
  3. Probably NOT...this would require a much larger number of changes, and require much more thought as to what was trying to be done
  4. This is questionable-roots generally get changed into some very, very close. Opposites or radical substitutions do NOT occur, and apart from suspected textual errors, we don't have these in the NT at any level (to my knowledge)
  5. This could be, but the slight change from two to one, and one ending change (the plural on 'cousin') might be a bit extreme for even an copyist error
  6. This is likely not, but the 'form' could be suggested by the exemplar, with the author knowing more detail and adding it. But could we call this 'borrowing'?-I doubt it.
  7. This is radically different, and looks like an eyewitness account (due to the absence of explanatory material about 'who' the various characters were)
  8. This, strangely enough, MIGHT BE, even thought it represents a complete change of meaning. But it would most likely be a copyist errors-similar to dyslexic phenomena.
  9. Probably not-the event is the same, but there are so many word changes that a 'copy' paradigm is too much of a stretch.
So, we have some that obviously are NOT, some that probably ARE, and the rest tend to be doubtful.


Now let's look at the NT data and the methodology involved.

First, let's note some issues of method:

First, we have already noted above the less than rigorous definition of 'parallel pericope,' and this will be quite an issue. Indeed, the very nature of the teaching mission of Jesus will militate against facile identification of parallels. Wenham gives the example of the Beelzebub accusations [RMML:65]:

"It is by no means obvious that a common document lies behind the two passages in this first section. They could be independent accounts of a single occasion considerable modified in transmission; or they could be accounts of different occasions when the same slander was similarly dealt with by Jesus. It is likely enough that this line of attack was deliberately thought out and then pressed home against Jesus during the latter part of his ministry."
In documenting the failure of the argument from order relative to the "Q" source in Luke's Central Section (9.51-18.14), Wenham points out this issue in a different way [RMML:76-77]: "If the Q-material of the Central Section does not come from the one or more Qs or from Matthew, what is the alternative? The simplest answer is the most revolutionary. The answer could be that these Q-passages have no common literary, or even oral, origin, but derive from different sayings of Jesus. Their similarities derive from a common source in the mind of Jesus, rather than from a single utterance of his lips. It is inevitable that an itinerant preacher must repeat himself again and again, sometimes in identical words, sometimes with slight variations, sometimes with new applications; sometimes an old idea will appear in an entirely new dress. All Q-passages set in different contents in the two gospels (from cases of complete identity to cases of similar imagery) could quite well have come from a preacher who on occasion had used the Matthean form and on another the Lukan. It could be that we are distorting the material when we insist upon asking which of the two is the more original. And again in RMML:78-79: "It is quite improbable that the great sayings which find a place in both Matthew and Luke, many of which are hauntingly memorable, were uttered only once by Jesus. He would have enjoyed them and would have wished them remembered far and wide. He would have repeated them again and again. These sayings would have become known in a common Greek form, through the instructions given by the Jerusalem church to the diaspora Jews who came to the festivals from all over the world. The identical or almost identical sayings in Luke's Central Section, which at first suggest literary dependence, could in fact easily have come from a saying uttered on two different occasions. It is exceedingly unlikely that Jesus taught about prayer only once; it is natural that he should have given the Lord's Prayer and his 'ask ...seek...knock' sayings more than once, and not always in identical words. As for the passages where common ideas are presented in quite different dresses, this is all so natural. Indeed if we release the Central Section from the procrustean bed of literary dependence and accept it for what it purports to be, the strains and distortions of the other theories vanish. What this means is that the very material with which the evangelists were concerned-the words and deeds of Jesus-had its own substrate of 'identity and difference' within itself. In other words, this identity/difference would not have been created by the evangelists at all!



Second, "verbal identity" must be at the word and linguistic form level. It will simply not do to appeal to some vague "parallel content", since this would be available for ALL competing theories of gospel formation (i.e., LD, OT, Eyewitness).

Linnemann drives this home [NT:ITSP:70-71, 110, 111]:

"Literary dependence can only be proven or disproven from the actual wording; one must restrict study to the linguistic data. Here it will not do to point to individual formulaic agreements encompassing just a sentence, or even a half sentence or just a few words. Such a practice often obscures the fact that there is no real contextual conformity at all from a linguistic point of view. Agreement in individual formulations does not automatically show literary dependence. That kind of clarity can be attained only after a general and thoroughgoing investigation of all data. The extent of agreement, as well as the differences, must be understood quantitatively if one wishes to come to an objective, well-grounded result. It is not enough to quantify by verses, which has been done, since considerable differences are found within parallel verses when one scrutinizes the actual wording rather than simply the general content. The basis for quantification should be the word as the smallest component of meaning."

"The extent of the similarities, as well as the differences, must be noted carefully. And this must take place, not at the (general) level of content, but at the (precise) level of linguistic form, for this alone will show whether a literary relationship exists. In order to arrive at usable results, the investigation must expressly avoid equating general agreement in content with identity in linguistic form; the two are by no means the same."

"Showing similarities among verses only discloses general agreements in content. To arrive at precise results at the level of linguistic formation, it is necessary to proceed from the foundational insight that the word is the smallest significant unity. Dependence is established in comparable passages by the occurrence of words in identical form. Not only the root of the word but also its specific form in the case at hand must be scrutinized. Similarities related simply to a common root have no relevance for the question of literary dependence, since the similarity of common root is offset by the dissimilarity of divergent form.

And even identical strings of words will not provide evidence of borrowing, IF this string is misplaced within the pericope. In other words, if the passage is "scrambled", then it obviously wasn't copied in a meaningful sense [NT:ITSP:112].


To illustrate this, consider the story of the Rich Young Ruler in Mark 10.21-22 and Luke 18.22-23. Wenham points out that in these two verses there are more than a score of differences, although the meaning is obviously close. This is hardly evidence for literary borrowing [RMML:21-22]. And the story of the Great Banquet (Luke 14.15-24 and Matthew 22.1-14) is supposedly Q-material, but out of 180 words in Luke and 223 words in Matthew there are only eleven identical words. Wenham draws the point: "This is no basis for a theory of a common literary source" [RMML:73].



Third, even when we find verbal identity it may be difficult to determine (in many cases) when the identity is due to literary dependence, or rather due to general oral transmission and memory dynamics. So Wenham [RMML:51]:

"Before proceeding to the argument from wording a serious question of methodology must be faced, which necessitates a digression. Redaction critics tend to see the redactionary process as something quite complex, so it may seem naive to try to assess the probability of a literary connection by simply laying parallel passages side by side and asking ourselves whether they look as though one is adapting the text of the other. It has been believed for so long that the synoptic problem must be solved in terms of literary relationships, that likenesses between passages are explained almost exclusively in a literary way however great the differences may be. It must be remembered, however, that the gospels were produced in a society where much learning was acquired by rote, and they were produced for communities who were trying to propagate the common teaching of the apostolic church. In such circumstances verbal likenesses would survive in oral transmission, and there is no need to invoke literary sources unless there is consistent evidence of actual copying." This factor alone would put the burden of proof on the LD'er, to demonstrate long sections of pure copying-which simply cannot be found in the gospels (we shall see more about this below).

And Wenham points out that this is a difficult issue, but that there is no compelling reason up front for LD (versus OTH) [RMML:5]:

"The great question is: Have we been justified in placing so much emphasis on documentary relationships? Lying behind this is the even more basic question: Can one distinguish documentary dependence from indebtedness to a common oral tradition? A frequently used method of approach is simply to look at the parallel narratives and ask oneself about the closeness of parallelism. If the wording is largely different, one rejects a literary connection; if the wording has a good deal in common yet is not very close, one keeps an open mind; but as soon as there is identity of expression for more than about a dozen words in succession, one leaps to the conclusion that the connection must be literary But his is not a safe conclusion. Even in our print-ridden era many Christians know a large number of ringing passages from the gospels and can quote them in their favourite version verbatim: 'Foxes have holes...', 'The harvest truly is great...', 'Ask and it shall be given you...', 'O Jerusalem, Jerusalem...' Much more would Greek-speaking Christians in the first century have memorised many of the sayings of Jesus in whatever form they were commonly taught...There is no reason therefore why sayings of dozens of words in length should not occasionally be found in identical or nearly identical form and yet have no literary connection. As far as the working of the individual pericopes is concerned, nearly all (if not all) could be explained by oral tradition." [RMML:4-5] And finally, we must note that too often critical scholarship simply assumes that the similarities are more significant than the differences. And, as Linnemann points out, this is all too unscientific [NT:ITSP:60]: "Here an all too facile equation is set up: In spite of considerable differences, the extensive agreements still provide proof for literary dependence. Not thought is given to alternative explanations of how such agreement might have come into being. I have already mentioned one possibility-the writers' efforts to pass on Jesus' words as precisely as possible. Similarities might also arise because identical content limits the variety of words available to a faithful recorder."

Okay, so let's move to the raw data on "verbal identity". Of the words, forms, and linguistic patterns, HOW MUCH is really 'identical'? [I will be drawing from Wenham's analysis (plus his use of other dissenters), and the first-hand detailed analysis of Linnemann, the ex-Bultmannian Form Critic).]
  The data above leads to some rather sobering conclusions, relative to what might now be called "alleged" verbal identity(!), and LD theories seem almost in the "What's wrong with this picture?" category. Some of the implications for the gospel authors seem so counter-intuitive as to render the theory of LD in need of 'extraordinary evidence'! 



What are the implications to be drawn from the raw data above? Now, if you haven't noticed by now, this is (1) rather detailed linguistic data; and (2) it is disastrous for the argument from "Verbal Identity". We have seen that this argument is rather 'fuzzy'; that the details in the text basically contradict our often facile judgments based on superficial similarities.
 
 

Let's do a quick summary of the above comments:

  1. There is an initial methodological problem with identifying parallel passages. It is very 'fuzzy' and lacks the rigor needed to really pursue this in a scientific fashion.

  2. The very nature of the teaching ministry of Jesus would produce an original deposit of data that was already characterized by 'similarities and differences', and therefore any attempts to explain these by redactional theories would be manifestly wrong.

  3. Bases for comparison must go beyond the superficial 'it LOOKS similar' (i.e. content-based) and get into the very verbal roots, forms, tenses, number, order, etc.

  4. It is very difficult (if not virtually impossible) to distinguish verbal identities based on oral tradition from those based on documentary sources. There is no prima facie case for literary dependence that can be made for these texts that would render it a 'reasonable assumption' (much less 'assured results')

  5. The differences within a parallel pericope must be taken seriously, and probably accorded even more weight than similarities.

  6. The actual linguistic data indicates HUGE amounts of very detailed (and almost senseless) changes.

  7. The actual linguistic data indicates VERY LITTLE actual verbal identity.

  8. The interpretation of this 'very-little-borrowing' finding forces one to doubt LD at a very significant level.

Accordingly, it is difficult to make this argument from 'verbal identity' carry much of the load in support of LD. There are simply too many methodological holes in it, the data seems overwhelmingly against it, and the redactional scenario envisioned by it (given the data) seems highly unnatural and bizarre at best.

Hence, we look now at the argument from Order.




What is the Argument from Order?

We have already seen the basic statement of this argument: that similarity of order implies literary borrowing (versus other explanations of this similarity).

Wenham describes the logical force of this [RMML:43]:

"Where several pericopes, which have no apparent logical or chronological succession are found in the same order, a natural possible inference is a literary connection. Similarly, if a sequence of material is broken by an omission or by the intrusion of new matter and is then resumed again, a literary connection is a natural explanation. Such sequences are found at a number of places in the synoptics, especially in the triple tradition" But here again we are going to run into similar problems-that of dislocations in sequence. And we will have to ask if the dislocations are MORE SIGNIFICANT than the similarities (assuming that we can come up with a meaningful explanation for the similarities, of course).

Let's turn to the data again:
 

Again, these criticisms are leveled at the very foundation of the argument from order-the very data of the sequences. If the sequences are only followed at less than half of the cases, it is hard to build a secure case upon that evidence!

I should also note here the methodological problem with arguments from order. Neville points out [NT:APSSC]:

"perhaps the most fundamental methodological issue to have arisen in synoptic studies this century centers on two basic options for analyzing the literary data. The first option is to compare the pattern of agreement and disagreement in order among all three synoptic gospels simultaneously. The second is to compare the similarities and differences in order between Matthew's and Mark's Gospels first, and then to do the same with mark's and Luke's Gospels" (p232)

"The concern about an appropriate methodological procedure for analyzing the phenomenon of order is significant because it is obvious that one's approach may predetermine one's 'results.'" (p234)

What we seem to be stuck with is this: the LD hypothesis, although a reasonable explanation for the similarities in order, is rendered useless by its abject inability to account for the manifold differences in order and sequence in the Synoptics. And there is no way to account for these differences-within the theory-without presupposing LD to begin with. Hence, the argument from order is either forceless or circular.


So, we have seen that the two main arguments for LD-the argument from verbal identity and the argument from order-floundered on the actual 'disorderly' character of the details in the NT. The verbal identifies weren't significant and there were more out-of-order sequences than in-order sequences.

But there is yet one other argument against the theory of LD: the argument from scribal practice.



The Argument from scribal practice

The argument from scribal practice is simply that the theory of gospel formation alleged by LD'ers, and especially 2SH'ers, is totally out of sync with what is known about scribal and literary praxis in the first century A.D. In other words, the only 'control data' we have, argues AGAINST LD, in its various forms, but esp. 2SH.

What we must understand is that many of the notions we have in the modern world about working with texts simply do not apply to antiquity. Let's note a couple of the dynamics of that time:

"In the first century tables and chairs such as we know them did not exist. Diners reclined propped up on an elbow, at the low tables. To consult more than one scroll an author would presumably have had to spread them out on such a table or on the floor and either crawl around on hands and knees or else repeatedly crouch down and stand up again, looking at first one and then another. He could either make notes or commit what he read to memory before writing the matter up on a sheet of papyrus or vellum, or, possible, sitting down and transferring it directly to his new scroll. Finding the place, unless he was prepared seriously to deface his scrolls, would be difficult." [RMML:205]

Secondly, for 'commoners' to undertake the difficult task-given the lack of textual helps at the time-is extremely unlikely.

"Copying with some adaptation was also common in the ancient world, but it was the work of highly educated scholars. For one who was not a professional to take a lengthy manuscript with no chapter, verse or even word divisions and select, rearrange and revise it was a formidable task. It is highly unlikely that one gospel was produced as the results of an author working directly on the scroll of another; even less that he worked on two or three at once." [RMML:206]

Third, the historical data that we have from compositional/scribal praxis of the period is at significant variance with the praxis assumed by LD'ers:

"We are in a position to tell with a considerable degree of certainty what compositional procedures for making use of existing writings would have been readily available in the first century. We can tell on the basis of many examples of practice and some indications of theory: even the most highly literate and sophisticated writers employ relatively simply approaches to their 'sources'...Conflation was itself only rarely attempted, and then very simply effected...the long debate on the sources of the Synoptic Gospels seems to have been conducted without paying much or any attention to this issue of whether any indications of 'sensible' compositional procedures in the first century C.E. were available" [F. G. Downing, cited in RMML:206]


Fourth, the data we have gives us NO REASON to believe the theory of LD:

"If this general picture be accepted for a fairly sophisticated and wealthy writer's preparation for writing, then the implications for the study of the Synoptic Gospels is clear. Even had one of our evangelists wanted to emulate the well-staffed and well-equipped compositional procedures of a sophisticated literary figure, nothing would have suggested that he should begin by analysing his source material, nor on that or any other basis that he should plan some complex conflation of this sources." [Downing, cited in RMML:207]


Finally, the very nature of the physical texts argues against LD [HI:JTOT:343-344]:

"And a purely literary approach to the written text would require too complicated a technique of cross-references and cross-checking to account for the change of order. As F. G. Kenyon pointed out long ago, there were normally no visible signs of where various parts of a composition written on a scroll began or ended at this time. The matter was further complicated through the need to roll and re-roll the scroll. Once a passage in a scroll was located, there would be no way to refer to it by paragraph or page, The intimate knowledge of the writing that was required in order to locate a passage in a scroll actually obviated the need to identify it exactly. And we have yet no clear evidence that Mark was originally written on anything else than a common scroll. Therefore, the very physical nature of written tests makes it likely that people quoted from memory rather than located the passage in a writing. In addition, we should not exclude the possibility that the school of the Matthean community had secretaries and that Matthew himself composed his narrative by dictating orally."
The upshot of this is clear: the theories of LD, esp. those involving conflation of texts, do not hold up when compared to the 'control data' of first-century scribal and literary practice. As the case with Documentary Hypothesis of the Old Testament, these theories are scholarly creations 'in a vacuum' without any connection to the real world of the past and the real word of the data.




So, what do we have so far?
  1. There is a significant erosion of the confidence in theories about the Synoptic problem, with special erosion from the camps of LD and 2SH
  2. There is a significant growth in adherents to 2GH
  3. One's approach to the issue of redaction is generally brought to the practice of redaction criticism
  4. LD itself is NOT a matter of faith: many, many strong evangelicals are firm believers in LD
  5. Cases of 'redaction' (in spite of the difficulty of identifying them) cannot reveal 'motives' by themselves.
  6. The argument from verbal identify, used to support LD, is not very strong at all.
  7. The argument from order, used to support LD, is not very strong at all.
  8. The argument from scribal practice, used to deny LD, is fairly strong, since it uses the only control data available from the period.
So, I have to conclude that James' statements are simply out-of-sync with the internal gospel data, out of sync with the external 'control' data,  and out-of-date with the scholarly trends.

So, now let's move on to how I understand the gospel formation process to have looked...


The Christian ThinkTank...[http://www.Christian-thinktank.com] (Reference Abbreviations)