My musings on Biblical Studies, Biblical Hebrew, Dead Sea Scrolls, Septuagint, Popular Culture, Religion, Software, and pretty much anything else that interests me!

Old Testament Commentary Survey

  • Searches

The Practice of Textual Criticism (TCHB 8)

30th July 2006

With some of the theory surrounding textual criticism of the Hebrew Bible behind us, with this post I am going to discuss how to actually go about text criticism.

This is the eighth post in a series on the textual criticism of the Hebrew Bible/Old Testament. Other posts include:

All posts in this series may be viewed here.

More specifically, in this post I will discuss how to identify and evaluate variants for the reconstruction of this Ur-edition. In the practice of textual criticism, critics traditionally distinguish between external criticism (i.e., the evaluation of the textual witnesses), and internal criticism (i.e., the transcriptional and intrinsic probability of the readings). For the former, critics need to know the textual witnesses and their history; for the latter, they need to be aware of the kinds of errors scribes made and have sensitivity to the context and inner clarity of the text itself.

The Preliminary Task: Collect the Variants

Before the variants can be evaluated, they need to be collected. They should be first collected from the textual witnesses and then compared with the MT, more specifically with the Leningrad Codex (L) as found in BHS. Even if you do not know Hebrew, you can identify the most significant variants in the text notes to most modern English translations (e.g., NIV, NRSV) and may even be able to detect others in differences between the English versions. For instance, Psalm 19:4 [19:5 in BHS] in the NIV reads “Their voiceb goes out into all the earth.” The superscript “b” leads the reader to the footnote, which reads: “b4″ Septuagint, Jerome and Syriac; Hebrew line.” By this note the translators are informing the reader that the variant reading of the text, “voice,” (which they used in the translation) is found in the Septuagint, Jerome’s Juxta Hebraica, and the Syriac Peshitta; while the MT variant is “line.”

As far as what English translations to use, the best translations, from the standpoint of OT textual criticism are: NRSV, NIV, TEV, NASB, NEB, and NJB. All of these translations carefully considered the available evidence when making their textual decisions (in addition, the NJPS will provide a good translation of the MT). Exegetes should avoid using paraphrases like the Living Bible, as it is primarily based on other English translations, as well as old translations such as the KJV, which is about four centuries out of date when it comes to text critical matters.

Exegetes using BHS (or BHQ) will find significant variants in its apparatus. Unfortunately, the apparatus is not the easiest to decipher. In this regard you may want to get yourself a copy of one of these two guides to BHS:

  • William R. Scott, A Simplified Guide to BHS: Critical Apparatus, Masora, Accents, Unusual Letters & Other Markings (3rd ed.; Bibal Press, 1995). Buy from | Buy from
  • Reinhard Wonneberger, Understanding BHS: A Manual for the Users of Biblia Hebraica Stuttgartensia (2nd. revised ed.; trans. D. R. Daniels; Pontifical Institute, 1990). Buy from | Buy from

Since the apparatus of BHS, however, contains errors of commissions and omissions with respect to the Versions and the DSS, the advanced exegete will appeal to the critical editions of the ancient Versions and to the DSS themselves (see my posts on the Versions and DSS for more information on critical editions).

Another great resource that I will be posting on in the near future is the massive multi-volume Critique textuelle de l’Ancien Testament edited by Dominique Barthélemy. This is a truly massive undertaking that collates, discusses, and evaluates all of the variants in the Hebrew Bible.

External Criticism: The Evaluation of Textual Witnesses

As noted above, external criticism involves an examination of the textual witness themselves. This primarily entails evaluating the variant in relation to the “original edition” of the MT.

1. Evaluate Relationship to “Original Edition”
“True” variants are restricted to those that arose in the transmission of the “original edition” behind the MT. Before selecting any variant for further evaluation, the critic needs first to determine whether or not it is the product of a tendency within one of the primary text types (e.g., MT, LXX, SP, and unaligned). For example, on the one hand, the shorter variants of Jeremiah should be passed over if they belong to the text’s earlier literary development. On the other hand, the longer variants in the Torah of the pre-Samaritan text, such as an interpolation of Deuteronomy into Exodus, should also be passed-by because they represent a later stage of the text than the “original edition.” When the critic has excluded variants that stand apart from that Ur-edition, he or she will then proceed to evaluate the variant by internal criticism. But before turning to internal criticism, we need to rule out the traditional approach to external criticism.

2. Reject Traditional External Criticism
Sometimes text critics evaluate variants on the basis of the textual witness in which it is found. Some critics prefer a variant in the MT over the SP, or a variant in the LXX over the Tgs., because normally the MT and the LXX are superior to the other two. For example, E. Würthwein notes: “The various witnesses to the text should be examined, beginning with MT, and continuing with the rest in roughly the order of their significance for textual criticism, e.g., SP, LXX, Aquila, Symmachus, and Theodotion, Syriac Peshitta, Aramaic Targums, ….” (The Text of the Old Testament, 112). Then too, some think that a variant in an early text has a prior claim over a variant in a later one, or that a variant in the majority of texts should be preferred.

Such external criteria should be ruled out, however, for four reasons. First, the fact that early corruptions obviously affected all our witnesses, requiring conjectural emendations, shows that one bad “gene” early on could have corrupted numbers of MSS. Second, the Qumran scrolls show an intricate web of relationships, so that one cannot predict a corruption in any given MS. Third, scribes tend to commit the same sort of errors, and therefore the same error could have arisen independently in several sources. Finally, the simple fact is that the Qumran scrolls, though a millennium earlier, do not normally contain better variants than the MT because the scribes in the later tradition tried harder to preserve the original than those at Qumran.

In sum, because we do not know the genetic relationship of any MS to the original edition, in principle a variant in the MT and/or in many witnesses has no prior claim as the better variant; and a variant in an otherwise poor witness, or in only a few, cannot be ruled out.

Internal Criticism: Transcriptional and Intrinsic Probability

Having decided that a variant may stem from the original text, the critic should not evaluate it further on the basis of the textual witness but on its own merits. There are two facets to internal criticism: transcription probability and intrinsic probability.

The task of evaluating a reading on its intrinsic and transcriptional probabilities is both an objective science and a subjective art. The basic rule of thumb is: “that reading is preferable which would have been more likely to give rise to the other”; or, turning that around, “the variant that cannot be explained away is more probably the original.” To explain away a variant, however, demands a firm grasp on the manuscripts, scribal practices, and a lot of exegetical knowledge and of common sense (this is why I believe while many aspire to be textual critics, few can gain enough mastery to actually do it well). Unfortunately, there are no simple rules. Some have likened textual criticism to a dog catching flees. As a dog catches flees not by following rules but by treating each flee individually, so also the text critic must address each variant individually, deftly, and reasonably.

1. Intrinsic Probability
Evaluating a textual variant according to its intrinsic probability involves taking the author’s style and the immediate context into consideration. Inasmuch as the inner clarity of the passage itself is the standard for evaluation, this is a subjective enterprise. It is sometimes difficult to determine what the author’s style or particular vocabulary is, as well as what fits the immediate context best. Nevertheless, while difficult, this is one of the major procedures of OT textual criticism.

2. Transcriptional Probability
Here the text critic needs to keep in mind the kinds of errors scribes committed either unintentionally or intentionally during the transmission of the text.

Unintentional Errors. Within the restraints of this post, I am only able to mention the most common types of unintentional scribal errors.

a. Confusion of similar consonants. Sometimes scribes confuse consonants that are similarly formed, depending on the script, or similarly sounded, such as the gutturals. For example, ד (d) and ר (r) are readily confused both in the Hebrew angular and square script. This is apparently what happened with the name of one of Javan’s sons. Sometimes he is called דדני×? (ddnym), “Dodanim” (Gen 10:4, MT), and other times רדני×? (rdnym), “Rodanim” (1 Chr 1:7, MT; Gen 10:4, SP, LXX). Other consonants that often are confused in the square script are: ב / ×›, ב/ מ, ב / × , ×’ / ו, ×’ / ×™, ×” / ×—, ו / ×–, ו / ר, ×› / × , מ / ס, and ×¢ / צ.

b. Haplography (“writing once”). Due to homoioteleuton, words with similar endings, or, homoiarcton, words with similar beginnings, sometimes a letter or group of letters accidentally drops out of the text. Compare the following readings of Gen 47:16:

  • MT: ו×?תנה לכ×? במקניכ×?
    I will give you for your cattle (cf. KJV)
  • SP, LXX: ו×?תנה לכ×? לח×? במקניכ×?
    I will sell you food… for your livestock (cf. NIV, NRSV)

“Food,” לח×? (lhm), comes after the similarly sounding and appearing “you,” לכ×? (lkm). The scribe likely skipped over “food” when copying the text. Another example comes from Judges 20:13 where the MT refers to the tribe of Benjamin as only בנימן  “Benjamin” instead of the expected בני בנימן  “sons of Benjamin.” The LXX reads “sons of Benjamin” and the Masoretes evidently thought that a scribe must have skipped over בני “sons of”, since they included the vowel pointing for בני even though the consonants are lacking.

c. Dittography (“writing twice”). Sometimes scribes accidentally repeated letters, a word or a phrase. For example, Isa 30:30 in the MT, LXX, Tgs., Syr., and Vulg. all read: הש×?מיע יהוה  “The Lord shall make heard,” while 1QIsa reads: הש×?מיע הש×?מיע יהוה  “The Lord shall make heard, shall make heard.” Apparently the scribe inadvertently repeated הש×?מיע “make heard.”

d. Doublets. This is the conflation of two or more readings, either consciously or unconsciously. For example, the LXX and 1QIsaa of Isa 37:9 conflate the accounts of Hezekiah’s consultation of Isaiah in the MT of Isa 37:9 and 2 Kgs 19:9. Compare the following:

  • MT : ויש×?ב ויש×?לח מל×?×›×™×?
    he again sent messengers (2 Kgs 19:9)
  • MT : ויש×?מע ויש×?לח מל×?×›×™×?
    and when he heard it, he sent messengers (Isa 37:9)
  • LXX, 1QIsa: ויש×?מע ויש×?ב ויש×?לח מל×?×›×™×?
    and when he heard it, he again sent messengers (Isa 37:9)

e. Metathesis. This is the accidental exchange or transposition of two adjacent letters within a word. For instance, Deut 31:1 reads:

  • MT : וילך מש×?×” וידבר ×?ת־הדברי×? ×”×?לה
    And Moses went [vylk] and spoke these words (cf. NIV)
  • 4QDeut, LXX : ויכל מש×?×” וידבר ×?ת־הדברי×? ×”×?לה
    And Moses finished [vykl] speaking these words (cf. NRSV).

The scribe evidently miscopied and reversed the order of ל (l) and כ (k) . The NRSV follows the reading in 4QDeut and the LXX, while the NIV opted for the MT.

f. Different concepts of word and verse divisions. Sometimes scribes, for unknown reasons, divided words and verses differently. For example, a scribe evidently divided the words in Hos 6:5 incorrectly:

  • MT : ומש×?פטיך ×?ור יצ×?
    And your judgments, light goes forth (cf. NASB, KJV).
  • LXX : καὶ τὸ κÏ?ίμα μου ὡς φῶς á¼?ξελεÏ?σεται
    = ומש×?פטי ×›×?ור יצ×?
    And my judgment goes forth as light (cf. NIV, NRSV).

The copyist of the MT evidently attached the ×› (k), of ×›×?ור (k’vr), “as light,” to the preceding word. Compare the following variants in Ps 102:[101 LXX]:24-25a involving different vocalization and misdivision of the verses:

  • MT : עִנָּה בדבך כחו קצר ימי ×?מר ×?לי (The Qere reads ×›×—×™ “my strength”)
    He broke my strength on the way, he cut short my days. 25 I said, “My God….”
  • LXX : ἀπεκÏ?ίθη αá½?Ï„á¿· á¼?ν á½?δῷ ἰσχÏ?ος αá½?τοῦ Τὴν ὀλιγότητα τῶν ἡμεÏ?ῶν μου ἀνάγγειλόν μοι
    = עָנָהוּ בדרך כחו ימי ×?מר ×?לי
    He answered him in the way of his strength: The fewness of my days report to me.

The LXX is different from the MT in reading ×¢× ×” (‘nh), as (Qal) “to answer,” rather than (Piel) “to humble”; taking בדרך כחו (bdrk khv), as a construct; and besides other vocalization changes, it also does not divide the verse in the same place.

Intentional Errors. Sometimes the scribes took liberty to change the text deliberately. Four different types of intentional changes can be noted.

a. Linguistic changes. Scribes often modernized archaic features of a verse, primarily in relation to spelling and grammar. For example, the SP replaces the old infinitive absolute construction of the MT with an imperative or finite verb form. In Num 15:35, the MT reads רָגוֹ×? (ragom), but the SP reads רִגמוּ (rigmu).

b. Contextual changes. Sometimes scribes change the text in order to harmonize certain passages. For instance, in Genesis 2:2, according to the MT, the Tgs., and the Vg, God completed his work on the seventh day, but according to the SP, LXX, and Syr (perhaps independently of each other), he completed it on the sixth day. The scribe(s) evidently changed the text to avoid the possible inference that God worked on the Sabbath.

c. Euphemistic changes. Sometimes scribes changed the text for euphemistic reasons. In Gen 50:23 the SP changes the phrase על־ברכי יוסף (‘l-brky yvsp), “upon the knees of Joseph” into על־נימי יוסף (‘l-bymy yvsp), “in the days of Joseph” because it seemed improper that Joseph’s grandchildren should be born upon his knees. In Deut 25:11 בִּמְבֻש×?ָֽיו (bmbshyv), “his private parts” is changed to בִּבְשָׂרוֹ (bbsrv), “his flesh,” because it seemed too obscene to mention that in a fight a woman would grab a man’s genitals. Similarly, in Deut 28:30 ש×?גל (shgl), “rape, have sex (?)” was deemed way too obscene for public use and so it was changed to ש×?כב (shkb), “sleep,” in both the SP and the MT-Qere. (Perhaps the equivalent of the ancient Hebrew f-word!).

d. Theological changes. We noted above how the Samaritans altered the pre-Samaritan text to defend Mount Gerizim as God’s place of worship. Theological changes also occur in the MT. Compare the following renditions of Prov 14:32:

  • MT: וחסד במותו צדיק
    But a righteous man in his death finds a refuge (cf. NIV).
  • LXX: á½? δὲ πεποιθὼς τῇ ἑαυτοῦ á½?σιότητι δίκαιος
    = וחסד בתומו צידק
    But the righteous man in his integrity finds a refuge (cf. NRSV).

The change from בתומו (btvmv), “integrity” in the LXX to במותו (bmvtv), “death” in the MT could be a case of simple transposition of מ (b) and ת (t). But some scholars think the change in the MT was intentional and reflects an anti-Sadducean point of view. Better known are the changes of early names with the theophoric element בעל (b’l), “Baal,” by the derogatory element בש×?ת (bsht), “shame.” For example, Esh-Baal (“man of Baal”), the name of Saul’s fourth son, in 1 Chr 8:33 is changed to Ish-Bosheth (“man of shame”) in 2 Sam 2:8.

On the whole, however, theological changes are rare in the MT. G. R. Driver notes: “Theological glosses [in our terminology, interpolations] are surprisingly few, and most are enshrined in the tiqqune sopherim [scribal changes], which are corrections of the text aimed chiefly at softening anthropomorphisms and eliminating the attribution of any sort of impropriety to God.”


Sometimes none of the transmitted variants satisfy exegetical expectations. In cases where all witnesses seem “hopelessly corrupt” the text critic may find emendation (a conjectured variant based on the known variants) necessary. Qumran scrolls have now validated this procedure in some cases. F. M. Cross comments: “No headier feeling can be experienced by a humanistic scholar, perhaps, than that which comes when an original reading, won by his brilliant emendation, is subsequently confirmed in a newly-found MS.”

Emendations must satisfy the same criteria by which known variants are evaluated. That is, they must be plausible. There are many emendations proposed where it is very difficult to see how the purported error took place. That being said, there are a number of places where emendation appears to be the best alternative. For example, there seems to have been a confusion of consonants in the angular script in Ezek 3:12.

  • All texts; ברוך כבוד־יהוה ממקומו
    May the glory of YHWH be praised in his dwelling place (cf. NIV).
  • Emendation: ברו×? כבוד־יהוה ממקומו
    As the glory of YHWH arose from its place (cf. NRSV).

The NIV’s “be praised” is based on ברוך (brvk), “be praised,” which is attested in all textual witnesses. The clause, however, is unique, awkward and contextless. Scholars salvage the line by emending ברוך (brvk) to ברו×? (brvm), “when [it] arose.” In the angular script ך (k) and ×? (m) are easily confounded. This emendation nicely satisfies exegetical expectations, Hebrew syntax, and the context of the verse (cf. Ezek 10:4, 15-18).

In sum, McCarter wisely counsels that a text critic should keep the image of a scribe clearly in mind, look first for conscious errors, know the personalities of your witnesses, treat each case as if it were unique, and beware of prejudices

2 Responses to “The Practice of Textual Criticism (TCHB 8)”

  1. Paul V.M. Flesher Says:

    Dear Dr. Williams,
    I have enjoyed your postings on Hebrew Bible text criticism. I will “borrow” some of your discussion to teach the biblical text in my course on Ancient Judaism this semester.
    I would like to make a couple of minor suggestions about your “History of the Old Testament Text…” chart.
    First, it’s a great chart and well organized. Useful for students.
    Second, as a targum scholar, I would suggest that you put the Aramaic targums a bit later. While I agree that the earliest targums may be first century CE (Onkelos and Jonathan have the best claim for this), most are second century or later. Qumran’s Job Targum should not be included in this grouping; it is something different (see David Shepherd’s recent book).
    Third, the targums do not feed into the Peshitta. See the articles in my _Targum Studies_, vol. 2, Scholars Press.
    Fourth, this is a bit picky. Your discussion identifies 5 stages of texts, but the chart only has 4. Any chance of getting all 5 into the chart?

    Thanks again for your excellent discussions of these key matters.

    Paul Flesher
    Religious Studies Program
    University of Wyoming

  2. S. Says:

    Its worth noting that the ברו×? כבוד emendation was Samuel David Luzzatto’s (Shadal).