James R. Adair. An Inductive Method for Reconstructing the Biblical Text: Illustrated by an Analysis of 1 Samuel 3. JNSLMS 2. Stellenbosch: University of Stellenbosch, 2000. Pp. 380. Price: US $40.00. ISBN 0-7972-0785-6.

1. The books of Samuel and Kings are fertile ground for biblical studies from a variety of angles: the Masoretic Text, Qumranic studies, the witness of the versions such as the Septuagint (LXX)--and a host of related sub-topics such as the "Lucianic" text, the Targums (T), the Peshitta (P), and the Vulgate (V). Each of these is a recognized field in its own right. What is surprising is to find all of these topics as part of the central focus of this book. An ambitious project to say the least, and one that current conventional wisdom deems too extensive for a doctoral dissertation such as lies behind the work under review.

2. Provided I have understood it correctly, the overall thesis is not difficult: formulate a consistent inductive methodology centered on statistical analysis for evaluating variants for their suitability for reconstructing the primary text to which they bear witness, and then reconstruct the Vorlage along text-critical lines. The heart of the book is first the formulation, then the application, and encompasses all of the major versions and witnesses.

3. Adair has chosen to center the study in 1 Samuel 3. I find neither defense nor explanation for this choice. Was the chapter considered representative of something? If so, it is not evident in the results, particularly from the versions such as T, P, and V. Concomitantly, I am left with the impression that better representation might have been found that would have been more fruitful in demonstrating the relevance of the methodology, even within First Samuel, let alone beyond. In saying this I have in mind some chapter--given that the chosen focus is one chapter--with at least one thorny issue, and preferably more, such as chapter 14 or 17. I accept that one could argue that it is best to start with a chapter that is not particularly difficult to hone the process, and then move on from there. However, I do not expect that this book will spawn lots of other such examples, hence my desire to have seen the process--but not the author--put to such a rigorous test, because I believe it is headed in the right direction.

4. The first step in the analysis is preparation of a text base for each version. Adair observes: "Ideally, the steps of the model just described would be applied to each particular versional tradition to determine as nearly as possible the original text. However, that task is beyond the scope of the present study, so traditional methods alone will be employed to determine a standard text for each tradition" (p. 30). I am ambivalent at this point. On the one hand, Adair's position is defensible given the limitations of research time for the dissertation. On the other hand, a lot of time has been put into the research method and methodology. If at the foundation "traditional" methods are acceptable, where does the cutoff point lie between their adequacy here and inadequacy further on in the process? Personally, I am left with the equivalent of the dreaded stamp on the back of a CD music recording: ADD (Analog, Digital, Digital). Digital remastering does not produce a result superior to an original digital recording. Is it the same here? Ironically, this brings me back to the question of the choice of 1 Samuel 3. Much of the LXX, for instance, is already prepared in the Göttingen edition (a critical edition), and overlapping semi-critical editions of P exist. Since most of the interest lies in the relationship of the LXX to the Hebrew Ur-text, it seems to me, at least, that the Göttingen text would have been a preferable place to begin, rather than an ad hoc reconstruction, but that would ipso facto exclude the books of Samuel (and Kings) from consideration since the Göttingen text for these is yet to be prepared.

5. Adair notes that the Hebrew text of Samuel is often seen as corrupt, due to haplography. It is my understanding that this opinion arises in large measure from comparison with the texts of the various versions under study in this book. In the reconstruction of the text of the versions "along standard lines," I do not see sensitivity to this issue. Is the text of chapter three devoid of such issues? Note the reading at 3:9: <grc>se</grc> {Bya2 A dpqt ein}] + <grc>eti</grc> cx | + <grc>o kalwn</grc> boc2 e2 z f+ MNaghvb2; Thdt. Inter alia, Adair observes, "More significant is the longer reading that includes <grc>o kalwn</grc> and is supported by many witnesses, including the f+ and Lucianic groups" (p. 47). After considering the possibility that this longer reading may have arisen from a Hebrew original lost in our current text(s), he observes regarding the Lucianic text, "Since the Lucianic text is known for expansions of this type, it is probable that the variant is secondary and within the Greek tradition, so the first reading is preferable" (ibid.). At issue here is the problem acknowledged in Adair's abbreviation, add-om.1 In the larger issue beyond this verse, is it possible that the dictum drummed into all of us, "shorter is better," makes it difficult to admit longer readings in general? And what of the generally accepted conclusion that our current Hebrew text ("MT") has suffered from haplography? I may have missed it, but while there is acknowledgment of the assessment, I do not recall having seen discussion here of the issue in relation to specific variants.

6. Earlier, Adair noted that the Cambridge edition of the LXX is a diplomatic text (p. 30). However, on p. 36 he says, "The editions of the LXX edited by Holmes and Parsons and by Swete will also be checked." This later comment is presumably a mental lapse, since the diplomatic text that forms the basis of the Cambridge edition, as Adair noted earlier, is the text of Swete, which is itself a presentation for the most part of the text of ms B (Vaticanus). Thus it is unclear what one would expect to find different by referencing Swete.

7. While it probably has no influence on the final outcome, I am surprised to see that Adair has linked ms e with "the ungrouped manuscripts." My preference lies with Brock (Brock 1966: 18) to place it with mss fmsw. If this were done, one would have (with Brock) the e+ group (= mss efmsw), rather than the f+ group (only mss fmsw). In all fairness, Adair has chosen to follow Johnson (Johnson 1963) in his groupings without referencing Brock.

8. In 3:15 the question relates to an article: <grc>Hlei</grc> Bya2] pr <grc>tw</grc> A{cx boc2 e2 d+ f+} MN{aefhinvb2}. The larger context is the presence or absence of the article with datives of direct or indirect object. Since in Hebrew articles are not used with proper nouns, the addition is a semantic marker on the Greek side. Adair observes, "Of about twelve other cases of proper names used as datives of direct or indirect objects in the first twelve chapters of 1 Samuel, all of them have the article. Thus it seems probable that the article should be included in the present case as well (p. 50)." While I agree with the conclusion, I wonder at the narrow and rather haphazard analysis. What is it about the first twelve chapters (and this is not the only time the limit is set at this point), as opposed to all thirty-one chapters, that they should be deemed determinative, especially since they are all from the same hand in the best exemplar, ms B?

9. In 3:17 the question is again the presence or absence of an article: <grc>logwn</grc> Ba2 A] > cx e2 b2; ethvid | pr <grc>twn</grc> {boc2 d+ f+} MN{abeghinovyc2}. "In this unit of variation, the omission in the second variant is the result of parablepsis. Since more nouns have articles than do not, one's initial reaction, in spite of the extremely thin external support, might be to exclude the article as a later addition (p. 50)." Here the point of reference is not clear. Is it this chapter? Excluding substantives with adjectives and articles with proper nouns, my count comes out even for the chapter for nouns with and without the article. Conclusions such as this, based on dubious information, make me nervous about the outcome, since at issue is the reconstruction of the base text for the statistical analysis.

10. Next is the Peshitta base text. This time there is virtually no information about the manuscript families or their dates. For me, the creation of the text is a crucial phase, yet much of the background is assumed, since none is sourced in the footnotes. If a standard critical text were in use, I would be much happier, but when the text is being reconstructed, it is a different matter. I realize that only so much can be done in a dissertation, but I also recognize this cuts both ways. The converse is that such omissions are tacit indications that the advisors did a disservice by allowing the topic to range so far and wide. Nor is it adequate to say that ultimately, as the results show, the Peshitta is not a primary focus, since no deviations are considered original. If that was in any sense an assumption, the whole topic should have been omitted. As it is, the only footnote is to note 47 on p. 53: "OT in Syriac, VI-VII" in connection with manuscript groupings.

11. The putative relationship of P to LXX is heavily discounted on the authority of Cook, and denied by Mulder (p. 33), yet Adair muses when considering 3:6 that "9a1fam shows no particular tendency to correct toward LXX. Rather, it frequently demonstrates an independence in modifying the Syriac text that lay before it, apparently without the support of another exemplar" (p. 55). Is dependence on the LXX still a live option for Adair?

12. In relation to creating the text of T, most surprising is the decision not to create the earliest text, but the official one, even though this necessitates excluding even pre-Christian readings. The use of a different font for the Aramaic than that used for the Hebrew helps distinguish the two languages. Two deviations from the base text are accepted: 3:2 the addition of a conjunctive vav; 3:10 the omission of <arc>yqr) d</arc>.

13. Texts beyond the main four are also created, one of which is the Lucianic text for the chapter under study. It is interesting that when Adair came to deal with this text he created a majority text along the same lines as I did (Taylor1992-1993), except that where the text lacked a majority reading, he accepted "the reading that differs from the main LXX" (p. 66). This arose when the witness of the Lucianic manuscripts split 2 + 2.2 While such a mechanical solution is a defensible position when creating the text as a basis for later analysis, for the most part in the case of the Lucianic text--at least in 1 Reigns/Kgdms--while individual Lucianic manuscripts differ from the LXX tradition, difference from the LXX is not ipso facto helpful in selecting characteristic Lucianic readings as seen from both Adair's analysis later in the book and from my analysis. It is encouraging to see differing approaches being adopted and to see them dovetailing as here.3

14. Having decided in favor of the priority of textual criticism (TC) over literary/ theological analysis as the suitable starting point, Adair follows Tov's insistence that the place to begin is translation technique (TT) (p.3) (Tov 1986:130), and adopts four of Tov's five criteria for evaluating relative literalness: consistency (stereotyping), representation of the constituents of compound words in the source language by individual equivalents in the target language (segmentation), word order, and quantitative representation.4 Appendix 2, "Add-Oms and Preliminary Translation Techniques Tables," contains the raw data from the LXX, P, T, and V for the four analyses listed above, as well as for add-oms.

15. One of the dicta all textual critics learn is that manuscripts are weighed, not counted. It is a necessary and important lesson. However, I am certain that it has done harm despite its inherent truth. My sense is that more time is spent decrying "counting" than considering what it might mean to "weigh" a manuscript. The Sitz im Leben, as it were, behind each model is numeric. One could wish that the expression were a little closer to (current) reality: counting produces statistics, evaluation leads to conclusions; counting is optional, evaluation is mandatory. Adair's volume includes much "counting" and in some eyes is thereby rendered suspect. However, such a premature conclusion fails to take into account the careful "weighing" that also takes place.

16. In Chapter Three the statistical analysis begins. Adair provides helpful categories and careful attention to a plethora of details. Of necessity, there is a rigid one-to-one focus at this stage. Not unexpectedly, the LXX cannot stand up under such stress, and the preliminary conclusion is that the text is not literal. I personally am intrigued by this conclusion. I have spent the last several years translating the <grc>a</grc>, <grc>bb</grc>, and <grc>gg</grc> portions of Sam-Kgs (1-4 Reigns/Kgdms) (1 Rgns 1-31; 2 Rgns 1:1-11:1; 3 Rgns 2:12-21:43). My worksheet has been a printout of Tov's Hebrew/Greek parallel-aligned text (gopher://ccat.sas.upenn.edu:3333/11/Religious/Biblical/parallel), programmed to appear in the respective native fonts, etc. I have been fascinated to see how parallel the two texts are, at times resulting in less than felicitous Greek. Maybe it is time to ask--or, re-ask--what literal means. For most of us, it does not mean slavishly literal, overly literal, but rather, mutatis mutandis, bearing a close relationship to the extent that Hebrew syntax is imposed on Greek syntax. Ultimately, it depends on one's perspective. Adair's statistical analysis joins the ranks of those attempting to put the study on a firmer footing. For instance, for the LXX he presents 16 tables of analysis.5 Under "Consistency" he notes such things as that <heb>bn</heb> is translated by <grc>uios</grc> and <grc>teknon</grc>, and <heb>dbr</heb> by <grc>rhma</grc> and <grc>logos</grc> and tabulates frequency. Statistics for verbs, nouns, and adjectives in this category indicate that for the Hebrew words in the chapter occurring more than once (>1x) there are 40 words in the Greek, or 1.29 words/Hebrew word, yielding a deviation factor of 0.31. Further, the total number of words in the Hebrew occurring more than once is 31. Such information is an integral part of all the conclusions reached in the rest of the book.

17. All that said, the focus of the book is one chapter in (at least) thirty-one. How consistent was the translator overall? Are the differences noted here consistently non-conforming throughout? If not, how should that be factored in? How would one bridge from the analysis of one chapter to the analysis of the whole book? If the original impetus for creating databases that was seen in the early days of computing had continued unabated, and it were available in relational databases and SGML/XML6 tagged texts, we would have resources beyond the MT, NT, and LXX that were all (at the minimum) morphologically tagged. Armed with such resources it would be possible to do a lot of number crunching (semi-)automatically to produce raw data in a wide range of categories. While this would only be the beginning, and much evaluation would remain, it would be a great asset.

18. The same principles as were applied to the LXX are now applied to the Semitic languages in the study, Syriac and Aramaic. The introduction to the former essentially serves for both, since they are not only Semitic languages, but branches of the common Aramaic stock. It is helpful to keep this in mind, since the introduction to the latter appears meager. The paucity in the conclusions does not seem to warrant the effort involved to extract the information, but this is the nature of TC work. The danger--and it is very real--is to argue from the specific to the general and dismiss across-the-board efforts to gain evidence from these versions as either fruitless or not warranted given the effort involved. Needless to say, perhaps, but reader interest would have been easier to maintain if some tangible results were available as the end product, demonstrating at least one variant reflecting a different Vorlage not otherwise available.

19. In connection with the Vulgate, Adair notes: "Lacking further evidence of V's pattern of rendering the prepositions <heb>b</heb> and <heb>k</heb>, the freedom with which prepositions are rendered in general prohibits concluding that the Vorlage was different from MT" (p. 153). It is sensitivity to such issues that has kept Adair's research focused on the attainable.

20. Chapter Four ("Literary/Theological Characteristics of the Secondary Witnesses") is where the text meets interpretation--with a vengeance ... or, what to make of the variants? As Adair states and demonstrates, the primary focus in religious studies is still on the Hebrew text. Thus it is enlightening when one brings this text to the various translations/interpretations to see how different the versions reveal their tradition's understanding to be. Unfortunately, much of this testimony lacks systematic presentation in the literature. English readers will have a great asset when the NETS commentary series on the LXX comes to fruition,7 as French readers do in the emerging La Bible d'Alexandrie series.8

21. Chapter Five, "Consideration of Variants From Primary Witnesses," is by now an important discussion, though not only in the way intended by Adair. In my re-reading of the book for the review, it strikes me that the division of the texts into primary (those in Hebrew) and secondary (non-Hebrew, translated directly from the Hebrew) has some strict limitations. This especially came to the fore for me in relation to Adair's criticism of Kyle McCarter who "occasionally uses [ms] B as though it were equivalent to the OG, for example, in verse 9 where he accepts the reading 'Speak' (without 'Lord') on the basis of B, or when he omits 'and thus may he add' in verse 17, following only mss Ba2 (plus cop it) but not the vast majority of the Greek mss or Rahlfs" (p. 198). Irrespective of whether or not McCarter (McCarter 1980) is right at this point, in principle and in general, the criticism of overreliance on ms B is valid, and one that especially all of us who work in 1 Samuel need to be reminded of from time to time. I have been guilty of the same, as Herbert pointed out in his helpful review (Herbert 1994). It is easy to forget that it is not the pristine OG when it is such a "good" text and stands in such contrast to others, including the Hexaplaric group, mss Acx. However, I was taken aback by the criticism of accepting the B reading over "the vast majority of the Greek mss or Rahlfs." I suspect that careful editing distanced from the passion of the dissertation would have removed this comment. As it stands, Adair is being influenced by numbers alone. Textual critics are no strangers to adopting readings found in one manuscript alone when the evidence calls for such. Beyond that, though, there is real sense in analysis such as done here, in which the MT itself is a secondary witness, though obviously it now has nothing to do with the language in which it is written. Adair has criticized McCarter for adopting readings from ms B as if it were the OG. But has Adair not essentially done the same sort of thing when comparing the versional evidence to the MT? Again, scholars like Tsevat believe the MT to have suffered in transmission (Tsevat 1976). If we were to drop the designations "primary" and "secondary," what impact would it have? As Adair recognized from the outset, the area of add-oms is a thorny issue. My sense is that a priori Adair has found it more difficult to accept longer readings into his lists than to exclude them.9

22. In the next paragraph, Adair is back on track when he illustrates the danger of citing versional evidence as the basis for retroversion without considering translation technique. Wilhelm Caspari says that in v. 2 the text of P implies a Vorlage <heb>kbdwt</heb> since a form of <syr>yqr</syr> is used (Caspari 1926), but in 4:15 the same Syriac root is used to translate <heb>qmh</heb>, indicating that the word falls within the semantic ranges of both <heb>kbd</heb> and <heb>qmh</heb>.

23. A useful byproduct of the analysis is Adair's questioning whether it is appropriate to cite the evidence of the versions accumulatively, as is so often done in BHS notes, for instance. He cites a good example from v. 21, where the evidence of V is used to support a reading with wide manuscript support: <heb>kdbr yhwh</heb>, contra MT's <heb>bdbr yhwh</heb>. On the one hand, Adair has shown V not to be restrictive in the use of prepositions, but on the other hand the voice of V joins an already existing chorus. While not definitive, the reading of V is accumulative.

24. Chapter Six considers the retroversion of the variants. For too many, and for too long, dissertations have begun with the "mandatory" survey of the literature. Thankfully, Adair has eschewed that approach, and introduced surveys and evaluations at the relevant points of the ongoing discussion. In this chapter the discussion is based on an article by Margolis (Margolis 1910) and Tov's LXX textual criticism book (Tov 1981; a second edition is now available), along with studies by Miles (Miles 1985), Seeligmann (Seeligmann 1940), and Deist (Deist 1988). The conclusions are appropriate; in effect, proceed with all due caution, using all resources available, and Adair does just that. He has accumulated and weighed a lot of evidence and now comes to retroverting the non-Hebrew variants back into Hebrew. In this connection he considers--even, agonizes over--what to do when there is no clear choice available, especially because more than one is available with no clear contender. In this case Adair's final choice is the best: "retain the reading of the MT and simply note the versional variant." This is more difficult for Adair when his prior analysis has deemed such to be a "significant variant," but now it must be passed over for lack of sufficient evidence. It is clear from the careful work here that Adair understands that counting does not obviate evaluating, and he does very careful work in this chapter.

25. Throughout the analysis Adair evaluates each variant from the secondary witnesses to determine whether it should be considered significant or non-significant,10 and in the previous chapter the selected variants were retroverted into Hebrew. In Chapter Seven all of the work comes together, and all of the variants are evaluated together to determine the best reading at each point. Finally, in Chapter Eight the critical edition of 1 Reigns/Kgdms 3 is presented.

26. Having spent the time reading--or perhaps better, studying--the book, I find myself somewhat ambivalent. On the one hand, Adair has done significant work. Because of the scope of the book, there is much that is open to suggestions for refinement and even criticism,11 but I for one appreciate all the effort that has gone into it. On the other hand, the analysis of just one chapter, and from my perspective not one of the more thorny chapters, has produced a volume of some 360 or so pages. Is this much work necessary to achieve the same results? Or is much of this part of writing a dissertation? If it is the latter, then I can only wish it had been pared down to the essentials. If not, then I do not expect to see much application of the process beyond this volume. It is reasonable to assume that Adair hopes it will spawn others, because Appendix 4 gives the algorithm for calculating the deviation factor. Whether this is the methodology or not, I hope that scholars will continue to refine the process to the point where it can produce the raw data for evaluation without undue burden.

© TC: A Journal of Biblical Textual Criticism, 2002.


1Adair has adopted this term in acknowledgment of the perennial problem of defining what constitutes an "omission" or an "addition." To use either term is to imply prior presence or absence in a text of the reading under consideration, and some measure of volition in removing or adding that reading. Were we in possession of the Ur-Text the terms would be appropriate. In the absence of such, Adair has adopted "add-om" as a shorthand to indicate that at those places where the term is used, the texts under comparison have a difference in text length of the kind traditionally referred to as omission/addition. Further, Adair does not include consideration of add-oms in relation to consistency, segmentation, and word order, since "it is difficult to compare something to nothing, and calculations of consistency that tried to take add-oms into account might well produce less than meaningful results" (p. 25).

2As Adair notes, there are in fact five manuscripts, since ms b is the combined witness of mss b' and b. He does not indicate what he did in this case. For my text, I accepted the 3 + 2 witness where it occurred and counted the former as the majority reading.

3When Adair was doing his research, he did not have access to my volumes or to my dissertation, and of course vice versa.

4 The fifth--linguistic adequacy--cannot be subjected to statistical analysis.

5 The total number of tables in the Appendix for all the versions is 64.

6 Standard Generalized Markup Language, the parent text of HTML (HyperText Markup Language) so popular on the World Wide Web; and eXtensible Markup Language, the successor of SGML.

7 A New English Translation of the Septuagint and Other Greek Translations Traditionally Included under That Title. The first volume was translated by Albert Pietersma (Pietersma, trans. 2000).

8 The project is directed by Marguerite Harl, Gilles Dorival, and Olivier Munnich, with the collaboration of Cécile Dogniez. Each volume consists of a French translation and a philological and exegetical commentary. The first volume, Genesis, was published in 1986, and volumes continue to be published as they are finished.

9 This comment is based on my assessment of the commentary throughout the book rather than the statistical analyses, since the decision made here selected the variants for inclusion in the analysis that led to the statistics.

10 All variants in the primary witnesses are by definition significant.

11 As noted, perhaps the first is the scope of the book!


Brock, Sebastian P. 1966. "The Recensions of the Septuagint Version of 1 Samuel." D.Phil. dissertation, Oxford University.

Caspari, Wilhelm 1926. Die Samuelbücher. KAT 7. Leipzig: A. Deichertsche Verlagsbuchhandlung Dr. Werner Scholl.

Deist., Ferdinand E. 1988. Witnesses to the Old Testament. The Literature of the Old Testament 5. Pretoria: N. G. Kerkboekhandel.

Herbert, E. D. 1994. Review of The Lucianic Manuscripts of 1 Reigns. 1: Majority Text; 2: Analysis, by Bernard A. Taylor. In VT 44: 413- 416.

Johnson, Bo 1963. Die hexaplarische Rezension des 1. Samuelbuches der Septuaginta. Studia Theologica Lundensia 22. Lund: CWK Gleerup.

Margolis, Max 1910. "Complete Induction for the Identification of the Vocabulary in the Greek Versions of the Old Testament with Its Semitic Equivalents: Its Necessity and the Means of Obtaining It." JAOS 30: 301-312.

McCarter, P. Kyle 1980. 1 Samuel. AB 8. Garden City, NY: Doubleday.

Miles, John Russiano 1985. Retroversion and Text Criticism: The Predictability of Syntax in an Ancient Translation from Greek to Ethiopic. SCS 17. Chico, CA: Scholars Press.

Pietersma, Albert, trans. 2000. A New English Translation of the Septuagint and the Other Greek Translations Traditionally Included under that Title: The Psalms. New York/Oxford, Oxford University Press.

Seeligman, Isac Leo 1940. "Problemen en perspectieven in het moderne Septuaginta-onderzoek." Jaarbericht ex oriente lux 7: 359-390.

Taylor, Bernard A. 1992-1993. The Lucianic Manuscripts of I Reigns. Vol. 1: Majority Text; vol. 2: Analysis. HSM 50, 51. Atlanta: Scholars Press.

Tov, Emanuel 1981. The Text-Critical Use of the Septuagint in Biblical Research. Jerusalem Biblical Studies 3. Jerusalem: Simor.

Tov, Emanuel 1986. "The Story of David and Goliath in the MT and LXX." In David and Goliath, by Dominique Barthélemy et al., 129-137. OBO 73. Fribourg: Éditions Universitaires; Göttingen: Vandenhoeck & Ruprecht.

Tsevat, M. 1976. Interpreter's Dictionary of the Bible. Supplementary Volume. S.v. "Samuel I, II." Nashville: Abingdon: 777-781.

Bernard A. Taylor
Loma Linda, CA, USA