Tuesday, October 25, 2011

Ehrman's Appeal to Intrinsic Evidence

Over at the Center for the Study of New Testament Manuscripts (CSNTM), Nika Spaulding and Robert D. Marcello summarize the recent debate at the Southern Methodist University campus (Dallas) between Bart Ehrman and Dan Wallace here. It should be noted in this connection that Wallace is the director of CSNTM, and that Spaulding and Marcello of course largely share Wallace’s views on the NT text.

I am happy to have contributed a little to the preparation of Wallace’s argument (in private correspondence) about the inconcistency on the part of Ehrman (and others), when he appeals to intrinsic evidence, and at the same time takes a completely agnostic position as to what a certain author wrote. In fact, I discussed this issue with Ehrman on a textual criticism discussion group in 2008 (here). Ehrman fully realized the problem as he responded:
Now, if someone can explain to me the logic of appealing to an author’s style when you don’t think you can get back to his words (hence his style), I’ll eat my Westcott and Hort!
Consequently, in the new edition of his The Orthodox Corruption of Scripture (recently released), he has a brief discussion of the “resulting theoretical problem,” i.e., the apparent contradiction of his own reconstruction of an early form of the text, and his claim that there is no way of getting to an original (pp. 350-52). There he states:
At the same time, I have not observed other critics wrestling with the issue; instead they continue to use intrinsic probabilities even while admitting that we have no access to an authorical text. I belive that is a problem, but I also believe it has a theoretical solution.
Then Ehrman proposes that although we are reconstructing an “author” with verbal, stylistic, literary and theological predilections, and although “recognizing them allows us to decide which readings go back to his imaginary pen and which were later creations of scribes,” we must at the same time acknowledge that this author is not a tangible human being of the past.

I actually agree with Ehrman – this is the bottomline of my own reply to him (and actually in accordance with the theoretical basis for the Coherence Based Genealogical Method, which seeks to reconstruct something more than the archetype of the tradition, but less than the authorial text – the term used is “the initial text”). So the question then is how far removed is our reconstructed author’s text from the historical author’s text? In my discussion with Ehrman I further suggested:
As I said, the simplest theory is that the initial text is the autograph (we do not know); the more complicated theory, the less is intrinsic evidence worth .... In practice, we assume that the text we reconstruct approximates towards the author’s text.
In sum, other text-critics have indeed wrestled with the problem, even directly in discussion with Ehrman. I also discuss the issue in my essay “The Implications of Textual Criticism for Understanding the ‘Original Text’” in Mark and Matthew I. Comparative Readings: Understanding the Earliest Gospels in their First Century Settings. Edited by Eve-Marie Becker and Anders Runesson (WUNT I 271; Tübingen: Mohr Siebeck, 2011). There is much to say about this problem. I have found Peter Shillingsburg’s works very helpful in this area. Perhaps I will post something on that in the future.

Finally, “the theoretical problem” is of course mainly a problem for those who are utterly pessimistic about reaching the initial text (“the more complicated theory, the less is intrinsic evidence worth”).


  1. Thanks, Tommy, for a thoughtful posting. Ehrman's claim to "have not observed other critics wrestling with the issue" prompts me to mention an 1990 article by Kobus Petzer, who was a visiting student at Princeton Seminary when Bart and I were students there: J. H. Petzer ("Author's Style and the Textual Criticism of the New Testament," Neotestamentica 24.2 [1990]: 185-197), a penetrating essay whose implications (though I referenced it in my 1995 article on reasoned eclecticism in the Metzger FS)I did not fully grasp at the time. Here are the last sentences of his article: "a consistent [authorial] use of language and style ... cannot be presupposed and used as a criterion for determining textual integrity. This kind of qualitative judgment as to the use of language can only be established after an analysis of the language and style of the whole document, which means that one has to decide beforehand what the document in question, looks like, before analysing the language and style employed therein. Such a qualitative judgment can only be the result of analysis and not the basis and presupposition for analysis, and can therefore not be used to establish what the original text of a document looked like." (pp. 195-6).

  2. Is establishing the text of a whole document really necessary before one can say anything definitive about an author's use of language and style? If so, NTTC is in deeper mire than previously thought.

    I would suggest that a reasonably accurate assessment of a given author's language and style can be obtained from those portions of scripture that basically are not subject to serious doubt (i.e., those places where textual variation either does not occur, or where the variants are so questionable or weakly supported as not to be received by anyone).

    For all practical purposes, then, I would suggest that the ca.94% agreement shared among all major editions of the Greek NT (Alexandrian-based eclectic or Byzantine) would be a reasonable starting point from which to determine authorial usage and style that then can be utilized when evaluating readings on internal grounds (e.g., despite numerous variant readings, one certainly can posit a heavy use of OUN as specifically characteristic of Johannine usage or initial KAI as typical of Markan usage).

  3. Thanks Mike, yes I have read that article by Petzer.

    Thanks Maurice too for your thoughts on this matter.

  4. Thanks Tommy for this post. Just thinking out loud for a moment from the standpoint of LXX text criticism, which is different from NT text criticism I realize, but there maybe an intersection on this topic of intrinsic probability.

    LXX text critics undergo translation technique studies for a variety of reasons, but one of them is to determine the tendencies and patterns of the translator in rendering his Hebrew text. There are two major assumptions in this project, 1) the Hebrew text resembles the Vorlage of the LXX translator and 2) Rahlfs' text is at least a good approximation of the Old Greek.

    We work under these assumptions realizing that both of these texts may need to be corrected at some point, but even still these texts seem to provide a sufficient starting point for gauging probabilities in translation technique.

    At the very least, just because one does not have the OG from the start does not mean that one cannot engage in translation technique studies for the purposes of establishing the Greek text itself. The starting point is sufficient to establish the tendencies of the translator.

    In a similar manner, I suspect the text of the NT is sufficient to establish intrinsic probability.

    Does this make sense?

    Also, James Voelz has an interesting article on this topic here: Voelz, James W. “The Greek of Codex Vaticanus in the Second Gospel and Marcan Greek.” Novum Testamentum 47 (2005): 209-249. I found it interesting, but you all are the experts.

  5. I admit that I'm more comfortable with the notion of the archetype than the "initial text," but I'd like to note that, even in a purely stemmatic approach, internal evidence plays a strong role in establishing the text of the archetype.

    Paul Maas (1958), for example, holds that the text critic, faced with a choice between two readings for the archetype, ought to base the decision on the author's "style and content" in the first instance and then, secondarily, look at transcriptional probabilities if the stylistic possibilities seemed balanced (pp. 13-14).

    In other words, Maas privileges stylistic (intrinsic) considerations over the transcriptional.

  6. Thanks John, yes I am aware of Voelz's article (which I note in the essay I referred to). Voelz rightly admits that his procedure contains an unavoidable element of circularity, “[A] given author’s Greek usage is determined on the basis of the readings of given mss., yet the selection of those mss. is determined to a significant extent on the basis of the author’s perceived usage” (p. 232).

  7. Thanks Stephen for your comment.

  8. Thank you, Tommy. I tend to put a high value on the style and usage of a N.T. author in making Text critical judgments (in the tradition of my mentor G.D. Kilpatrick). In this connection my thinking has been shaped by the comments of C. H. Turner,who at the beginning of his "Notes on Marcan usage" cited F.J. A. Hort's famous dictum, "Knowledge of documents should precede final judgment upon Readings." ("The New Testament in the Original Greek"section 38, set in capital letters). Turner wrote, "I want to enter a similar plea for what I conceive to be an even more important principle, namely that 'Knowledge of an author's usage should precede final judgment' alike as to readings, as to exegesis and - in this case - as to mutual relations in the Synoptic Gospels." (JTS 1924, 337) In the case of Mark, Turner's articles on Marcan usage, reprinted by J.K.Elliott in 1992, offer a valuable model for establishing an author's style, and its value for text-critical decisions.

  9. I don't think circularity can be avoided entirely in NTTC. I think Hort and Turner are both right. The task is to balance them (IMHO).

  10. Timo, Thanks. I agree, and this is essentially what I tried to argue in my article "The New Eclecticism", Nov. Test. 34 (1992) 388-97. Peter

  11. John, you state that the first assumption in LXX translation style studies is that "the Hebrew text resembles the Vorlage of the LXX translator." This is a huge assumption (depending on how thoroughly its logical conclusions are followed through). Apart from cases of wholesale revision, it does seem reasonable to assume that the LXX Vorlage is related to the MT text in a way that can be explained with reference to the normal habits of the scribes. But that is very much different from assuming the LXX is translating the MT. All too often it seems like false "styles" can be isolated when the LXX Vorlage may actually differ from the MT. It seems like methodologically in dealing with translation style you would need to deal with three levels.

    1) Corruption within the inner-Greek tradition
    2) Style exhibited in the act of translation
    3) Variation within the Hebrew tradition

    Rather than simply conflating these issues by comparing Rahlfs with the MT in toto, would it not be better to ignore problems clearly identifiable as operating on levels 1) and 3) and restrict your analysis of style to level 2)? If you limit your corpus for studying translation style to those texts where both the Old Greek and its Hebrew Vorlage can be identified with a high degree of certainty, you would avoid potentially false reading arising from comparison with the MT. Even if you end up having considerably less information about LXX translation style, it would be significantly more reliable.

    Similarly, when it comes to determining the author's style for the Hebrew OT text, it seems most reasonable to approach the problem by removing questionable texts from consideration and identifying reliable texts (those without attested variants or other reason to suspect corruption) on which to base one's analysis of the author's style. I think you would still find a large enough corpus of material from which to make conclusions about style, and yet your results would be a lot more reliable.

    In the NT, where there are significantly more variant readings, this might be more difficult, but I suspect that Maurice is still correct that there would be a significant enough corpus of reasonably certain text on which to base intial stylistic analysis.

    So in the end, while the use of intrinsic evidence will always of necessity be somewhat circular, reliable results can be attained by studying reliable texts, and in turn help in resolving the textual problems.

  12. Thanks Drew. I think we are in essential agreement, but you have provided much needed nuance.

    Of course one cannot rely in toto on MT or Rahlfs, which I thought I affirmed. One must be open to correcting both texts as a clearer picture of the relationship between the two emerges.

    My primary question is do these texts provide a sufficient (not comprehensive) starting point for the comparison of the Hebrew text with the Greek text? They may provide more than a starting point in some books and maybe less than one in others, but it seems all studies depart from here.

  13. To the OP,

    My first reaction is similar to Maurice's.

    While we acknowledge points of corruption in the text - even points that continue to elude resolution - the fact remains that the bulk of the text does not fit that description, being either transmitted without corruption or transmitted with corruption that was easily resolved.

    I myself have argued for the more extreme position that there remain points of total corruption (i.e. points where the initial text does not survive in any extant ms) but even I stress that the great bulk of the text is good and acceptable.

    This bulk provides enough data to establish authorial trends and justify the use of intrinsic criteria. As I have written before, it is the consistency of the surviving text that can reveal points of inconsistency: points of corruption or other trouble.

    Phrased as it is in the OP, it is an interesting theoretical puzzle, but I think that is all it is.

    It reminds me of what I have come to think of (only in my own head) as the philosophers problem. No offense to philosophers out there, just my own anecdotal experience of a tendency to apply ideas - usually ideas about what we can't or are unable to do - in a vast, all or nothing, winner takes all fashion. Probably this stems from a phil. major roommate in college, who I found sitting destitute one morning after reading Kant, explaining to me how we could no longer know *ANYTHING* about *ANYTHING*! "There's some things we can know, right?" "No, *NOTHING*!" he replied. "But I think I know I have to go to the bathroom..."

    I mean, yes, people like Kant or Ehrman in the OP can have a good point, but why do we have to accept (or reject) it so wholesale? What's wrong with acknowledging a limited degree of legitimacy but not allowing it to bring down the house? There's nothing wrong with injecting just a little uncertainty into the mix.

  14. Ryan, I agree with you. In my own doctoral work I did use Ehrman's arguments in some textual cases without agreeing with his total skepticism of the text. I think both Kant and Ehrman have a point, but only up to a point :)

  15. Thanks for the clarification, John. It just seems like LXX scholars tend to err on the side of assuming that the LXX is translating the MT text, when often other inner-Hebrew possibilities may be at play. I still worry that starting from the top down (MT vs. Rahlfs generally > understanding LXX translation style > corrections on individual texts and styles), as opposed to the bottom up (evaluating texts clearly translational from the MT or known Vorlage > understanding LXX translation style > corrections on individual texts and styles), can still be misleading, even if later refinements are made to the findings. This is particularly true in cases when a stylistic concern is only occassionally exhibited.

    An example might help:

    The LXX of Genesis shows consistent evidence of harmonization. But it is irrelevant to show 50 cases of harmonization (against the MT) as evidence of the LXX translation style to harmonize, because these could just as easily have been made on the level of the Hebrew Vorlage. The only way to prove that it relates to translation style is to find examples where the harmonization makes more sense on the translation level than the Hebrew level. I would cite, for example, Gen 8:13, which has two identical phrases in the LXX "the waters subsided from the face of the earth," but in the MT, they are quite different. The nouns and prepositions are different in the Hebrew, but the same in the Greek, which makes this harmonization much more likely to have been done in translation. The decision may also have been influenced by the translation of the verb as "quit, subside," which would be inappropriate with the subject "the surface of the land." For these reasons, harmonization makes much more sense in the process of translation into Greek than within the Hebrew tradition. With examples like these, we can then prove that at least some harmonization occurred in translation, which informs our understanding of the translation style more accurately and gives more reliable results when examining individual variants. In this case, the two approaches come to the same conclusion, but the latter seems much more objective. The standard approach would be the equivalent of determining an author's style from one of the great codices without reference to variant readings. Even if one goes back and adjusts one's findings in light of other readings, they may still be greatly skewed because of prior observations. But if a more limited pool of uncontested data is used to determine style, you may get more accurate results. Do you think this breakdown of the two approaches is accurate, or am I still missing something?

  16. Paul Goodfellow11/23/2011 3:55 pm

    The summary by Spaulding & Marcello states that "Ehrman demanded absolute proof that the New Testament had not been corrupted". Has anyone ever asked Ehrman what would constitute absolute proof? I am not sure that all copies of the NT text being in 100% agreement would constitute absolute proof. In fact, I am confident it wouldn't. Couldn't 100% agreement be met by someone gaining control of the text and destroying all manuscripts that differed from their preferred version? So then is it possible that discrepancies actually indicate a more reliable text?