Monday, April 30, 2012

Quantification of Variants in OTTC

Recent conversations in the comments of a past blog post prompted me to think a bit about the overall number of variants in the OT text. I also just finished an article by David J.A. Clines "What Remains of the Hebrew Bible? The Accuracy of the Text of the Hebrew Bible in the Light of the Qumran Samuel (4QSam(a))" in Studies on the Text and Versions of the Hebrew Bible in Honour of Robert Gordon, edited by Geoffrey Khan and Diana Lipton, 211-20 (Brill: Leiden, 2012). This article will serve as a good starting point for considering the quantification of variants in OTTC.

Right from the start, Clines takes on a very polemical and adversarial tone. With regard to Robert Gordon (for whom the volume was edited), Clines says, "If nothing else, this paper adds to the documentary evidence he will need if he is to refute the ideas (211)." Indeed, he retains the same tone throughout the article, eventually concluding that the Hebrew Bible is in a "state of radical uncertainty (219)," where every word is in doubt. In the end, however, I doubt Robert Gordon (or anyone else for that matter) will have much difficulty refuting Clines' ideas.

His basic argument is as follows. If we compare 2 Samuel 22 in the MT and 4QSam(a) and its parallel Psalm 18 we note that there are variants for one out of every two or three words. If we then extrapolate this ratio to the entire Hebrew Bible, we would have about 111,090 variants out of 305,500 words. Furthermore, McCarter accepts 6 variants to the MT as superior out of 66 words in 2 Sam 22, which would imply 27,700 places where the MT text is inferior in the whole Hebrew Bible. Since one out of two or three words on average would have had a variant at one time and we do not know which words had the variants, every single word is thrown into doubt, "as if we possessed an entirely uncertain text (218)." Yet he himself admits that we generally know the contents of the Hebrew Bible, even if not every detail (219).

There are many interesting things to note from his statistics, but also many problematic elements. He claims to have dealt with complicating factors in a previous paper (217), but as these pertain to fundamental logical and evidential flaws in his approach, he should have nuanced his current argument as well. Since he fails to do so, his chapter reads as little more than sweeping generalizations leading to unwarranted conclusions. A few points are important to keep in mind when encountering such claims.

1) Choice of Test Passage: Clines' choice of 2 Sam 22 and Psalm 18 does not appear to be random, but to have been purposely selected to bolster his argument. First, Samuel is well-known as being one of the most corrupt texts in the MT. Second, 2 Sam 22 and Psalm 18 are well-known for significant differences between these parallel passages. Indeed, looking at BHS, almost half the page for each page for 2 Sam 22 is devoted to variants, in contrast to the few lines characteristic of most other books. If you want to prove the text is disputed, base your statistics on one of the most disputed texts available! Studying other passages would have dramatically changed his results. If he had compared Isaiah 36-39 with the parallels in 2 Kings as I have recently been doing, he would have found a much more stable base text.

2) Redactional Differences in Parallels: Clines treats all differences between 2 Sam 22 and Psalm 18 as equivalent, when there is in fact good reason to believe that many of the differences arose from intentional redactional activity as the literary unit was transferred from one context to another. If this is the case, then these differences have no bearing on the accuracy of the transmission of the text of Samuel or Psalms. Parallels cannot be properly used without a nuanced appreciation of their individual contexts. By treating all of the variants in the same way, he has further increased his number of "variants."

3) Extrapolation of Statistics: Clines takes the ratios derived from his study of 2 Sam 22 and then extrapolates the data to see what a similar ratio would yield for the entire Hebrew Bible. This is problematic for two reasons. First, the mere fact that his numbers are actually extrapolations and not actual evidence makes them little more than hot air, especially since they come from such a small pool of texts. His statistics on 2 Sam 22 tell us nothing about the state of the text of the rest of the Hebrew Bible. Second, this extrapolation is misleading, since his pool of texts is non-representative. The books of the Hebrew Bible are very different in nature and preservation, and we cannot safely generalize about books other than those included in our study. His statistics on 2 Sam 22 tell us how many variants there are for 2 Sam 22, and nothing more. His results are significantly smaller when he looks at Isaiah, for instance.

4) Evaluation of Variants: While Clines thankfully does not include merely orthographic variants, he does include all variants without respect to their value or the characteristics of the witnesses which attest to them. This method is quite problematic. For instance, I have heard that the Samaritan Pentateuch has around 6000 variants from the MT. Many of these, however, are clearly secondary, and do not throw the text of the MT into doubt. In a sense, these variants are mostly irrelevant for the question of the preservation of the MT. Likewise, centuries of collations of medieval manuscripts and late translations have shown that variants can be multiplied ad absurdam without significantly altering our understanding of the text. Simple numbers, apart from detailed textual analysis, are meaningless at best and misleading at worst. To show corruption in the MT text, you have to demonstrate the secondary nature of its text, not simply state that there are variants.

5) Simple Mathematics: And of course, as with any statistical argument, we must be sensitive to the statistical methods employed. Clines counts pluses and minuses by word, rather than by variation units, for instance. You have to consider which base text he counts for his word numbers as well. I do not want to get into all the statistical details, but suffice it to say that these are something to be aware of.

6) Unwarranted Conclusions: And the final major point I would make is simply that the primary conclusion he draws is simply not warranted by his evidence. Depending on how you define a variant and how exhaustively you examine the manuscript evidence, I imagine you could probably claim to have found even more than 100,000 variants and speculate about far more. But this number would prove absolutely nothing. It is like Bart Ehrman's classic 400,000 variants in the NT statistic, which is designed more to wow uninformed audiences than to make a nuanced point about the preservation of the text. The simple reality is that the preservation of the OT is far more complex than any such simplistic arguments could possibly cover. Some texts are quite well preserved, while others are more problematic. Some variants to the MT are earlier, and some variants are later. The preservation of any given text must be studied in its own right by collating, sorting, and evaluating the extant evidence.

In contrast to Clines' extreme skepticism, I would point out the statements of Gene Ulrich (the general editor for the cave 4 Qumran manuscripts) to the effect that, "The base text of most books remained relatively stable... Clearly the books were copied with a care and fidelity that fills us with awe and admiration (DSS and the Origins of the Bible 109, 114)." According to Ulrich, though occasionally certain scribes made significant editorial revisions to biblical books, most of the text was accurately copied down from generation to generation. Unfortunately, because of the immensity and complexity of the evidence, we will never be able to quantify the variants in the OT text. But the preserved evidence by no means requires a radical pessimism about the text. It is not very often that you will hear scholars call their own positions "radical," but in Clines' case, he is clearly correct.


  1. Drew,

    As I have continued to study this topic I have a question concerning textual criticism from the LXX and syriac.

    When you are trying to determine the original reading for a text and you are going to utilize the LXX or some translation. Is it possible to determine the Hebrew word that it would be translated from?

    So if you are looking for the original Hebrew reading if the MT might need a different Hebrew word can the original be determined from translations?

  2. Hi Peter,

    You have hit on one of the fundamental problems in OTTC! It continues to be (and will probably always be) debated. Scholarly opinion has historically wavered back and forth between overconfidence and radical skepticism about reconstructing Hebrew texts from ancient translations. Recent Septuagint scholarship has tended to stress the difficulty of using the LXX to reconstruct Hebrew variants, but others have argued that the DSS show that the LXX was generally literally translating Hebrew variants, and so we can use it. I personally would be somewhere in between. I think the LXX does often literally translate real Hebrew variants that can be reconstructed with a fair degree of confidence, but I also think that at many points the LXX translation style better accounts for the semantic differences. I guess a couple of points would give you an idea of how I approach it.

    1) If the version agrees with the MT and the other Hebrew witnesses, we can be relatively certain that it read the same Hebrew text.

    2) If the version has a different meaning or form from the MT or other Hebrew witnesses, we must consider it on three levels.

    -Inner-version: Some differences are best explained by corruptions within the language of the version and do not reflect a Hebrew variant.

    -Translational: Some differences are best explained by the translation style of the version. You have to study the version in detail to learn how it characteristically treats its Hebrew text. Some are more literal (and more likely to reflect real Hebrew variants) and others are less literal (and more likely to change the meaning in translation). This stage requires good knowledge of the requisite languages, lexical equivalents, exegetical tendencies, etc.

    -Hebrew: Some differences are best explained by different forms of the Hebrew text. This might be supported by other Hebrew manuscripts, which confirm that there really was a variant at that point. We must look to see whether the change is more likely to have occurred in the Hebrew than in translation or the transmission of the version. An example would be the date in the LXX in Genesis 7:11, where it says "27th" instead of "17th". The two look almost identical in Hebrew, but are very different in Greek, which proves that the difference goes back to a Hebrew variant.

    I hope that helps clarify an incredibly complex process. :)

    1. Drew,

      That is helpful thanks. My reason for asking is because there are NT critics who would say that God has preserved the original readings in the manuscripts.
      For the OT it is more tricky and relates to the issue discussed above. Where the MT needs correction being able to determine the readings which was the original.
      This of course brings the presupposition that God has preserved the original readings in the OT. Since we are able with some relative confidence then in theory it could be the same in the OT.
      But given what you have said in the past why would you disagree?

    2. Certainly there are a number of older readings preserved only in the LXX, rather than in any Hebrew evidence. But there also apear to be some texts that aren't really preserved in any witnesses. 1 Sam 13:1 is probably a good example, as none of the oldest traditions have the original numbers. Some later manuscripts apparently make educated guesses, and modern translators do as well, based on Acts 13:21, but there is no reliably preserved tradition for this verse. An enlightening example may be the long addition at the end of 1 Sam 10 in 4QSam(a), which was previously known only indirectly through Josephus. This paragraph may actually have been original, but unknown to us for two thousand years. Many earlier conjectural emendations were confirmed by Qumran evidence, and it is likely that at least some others are also original.

      Maybe two examples from my work on Genesis will help as well. In Genesis 9:7, the second "and multiply on it" should probably read "and subdue it." In the old Hebrew script, two similar letters were confused, infecting basically the entire tradition. Only a few late Greek manuscripts have the reading, and it is difficult to tell whether they actually have any direct link with a good Hebrew reading, or whether they created the reading later. 6:13, "I am going to destroy them with the earth," is also very difficult in Hebrew, and probably corrupt. This is the one place I proposed an emendation, but I am not at all certain what the original reading would have been.

      Many examples like these with little or no textual evidence surviving could be cited. Examples like these make me hesitant to say that all of the original readings have been preserved.