See George Howard's response in TC 4 (1999).

[Click here, here, and here to see discussion of this review in the tc-list archives.]

Some Observations on a Recent Edition of and Introduction to Shem-Tob's "Hebrew Matthew"

William L. Petersen
The Pennsylvania State University
Netherlands Institute for Advanced Studies

1. The point of departure for this study--which is, properly speaking, a review article--is the publication of a book titled Hebrew Gospel of Matthew, edited, translated, and introduced by George Howard (Howard 1995). Our exploration begins by describing this Hebrew Matthew, and commenting upon some of its unusual readings. We also remark upon some of its characteristics and Howard's translation of it. Our study then moves to its main subject, namely, a summary of Howard's evidence for stipulating the genesis and dating of this Hebrew Matthew; this is interwoven with a critical examination of his evidence and theories. New evidence is introduced to suggest that Howard's analysis is defective; this new evidence also suggests a different genesis and dating for the text. Finally, our study concludes with a reflection on some of the lessons which can be learned from this exploration.

2. The focus of our study, Howard's Hebrew Gospel of Matthew, is a revised, second edition of a work originally published in 1987 under the title The Gospel of Matthew according to a Primitive Hebrew Text (Howard 1987). The Hebrew text and Howard's English translation appear essentially unchanged; Howard's analysis of the text and its ancestry is, however, considerably revised.

1.0 Shem-Tob's Even Bohan and Its Hebrew Matthew

3. Hebrew versions of the Gospel of Matthew have been known to and used by New Testament scholars for centuries. Often referred to in scholarship as the "Old Hebrew Matthew," Sebastian Münster produced the first edition of such a manuscript in 1537 (Münster 1537); in 1555 Jean du Tillet edited a different but similar Hebrew Matthew manuscript (du Tillet 1555). However, in addition to these two, at least five other Hebrew Matthews (mostly fragmentary) are known: (1) the Book of Nestor, (2) the Milhamot HaShem, (3) the Sepher Joseph Hamekane, (4) the Nizzahon Vetus, and, now, (5) the work edited by Howard, a Hebrew Matthew contained in the twelfth (in some manuscripts, the thirteenth) book of the Even Bohan (The Touchstone), a Jewish polemical treatise directed against Christians (Howard 1995: 160-161).

4. The Even Bohan was composed in Spain (presumably in Aragon) by a Castilian-born Jew named Shem-Tob ben-Isaac ben-Shaprut in 1380 (ibid.: xi). He revised his work at least three times: in 1385, around 1400, and once again, still later (ibid.).

5. Shem-Tob's Even Bohan is, then, clearly Iberian and clearly a late medieval work. The manuscripts from which Howard produced his edition (nine, dating from "between the fifteenth and seventeenth centuries" [ibid.: xii; curiously, for an edition, Howard--while giving the manuscript registrations--never gives the dates of the individual manuscripts]) are also all late. The Hebrew Matthew in the Even Bohan is, however, older than Shem-Tob: this is demonstrated by, among other features, his occasional critical remarks on the translation (ibid.: 160-175). As we will see below, there is evidence--both internal and external (in the form of another Semitic-language translation made in Spain during this period)--to suggest that the original translator/compiler of the Hebrew Matthew contained in the Even Bohan was a Christian who knew Hebrew (perhaps a Jew who had converted to Christianity?).

1.1 The Text of This Hebrew Matthew

6. The text of the Hebrew Matthew embedded in Shem-Tob's Even Bohan is intrinsically interesting, for it contains many arresting readings. Here are four examples.

1. Matt 12:37

   7. The standard Greek text (cited here as elsewhere in this article from Nestle-Aland²⁷) reads: "For by your words you will be justified, and by your words you will be condemned"; Shem-Tob's Hebrew Matthew reads, however: "According to your words you will be judged, and according to your deeds you will be convicted." Note that the Hebrew Matthew juxtaposes "words" with "deeds" (the standard Greek text has "words" twice); its text is completely negative, speaking only of "judged" and "convicted," leaving aside the standard text's optimistic "be justified." The theological difference (a stress on "deeds" and without the certainty of "justification") is striking.

2. Matt 28:19-20

   8. In Shem-Tob's Hebrew Matthew, the "Great Commission" becomes: "Go and teach them to carry out all the things which I have commanded you forever." No mention is made of "making disciples of all nations," nor does Jesus promise to be "with you always, even unto the end of the world."

3. At Matt 28:9

   9. The standard Greek Mehrheitstext reads: "And as they went to tell his disciples, behold, Jesus met them, saying, 'Hail'"; Shem-Tob, however, reads: "As they were going Jesus passed before them saying: 'May the Name deliver you.'"

4. Matt 24:40-41

   10. The standard Greek text reads:

   40 Then two will be in the field; the one is taken, and the other is left. 41 Two women are grinding in the mill; the one is taken, and the other is left.

   Shem-Tob, however, reads:

   40 Then if there shall be two ploughing in a field, one righteous and the other evil, the one will be taken and the other left. 41 Two women will be grinding at a mill; one will be taken and the other left. This is because the angels at the end of the world will remove the stumbling blocks from the world and will separate the good from the evil.

   As can be seen, Shem-Tob interpolates an explanation which stipulates who the two are (one is righteous, the other evil [the Greek text makes no such identification]), and gives a new--and inconsistent--definition of "stumbling blocks": they are the evil people who will be removed (the parallelism of v. 40 in Shem-Tob suggests--and standard understanding is--that it is the good who will be "taken" [to heaven] and the evil who will be "left"; but Shem-Tob's v. 41 states that it is the evil who will be "taken," presumably leaving the good here!).

11. One of the distinctive characteristics of Shem-Tob's Hebrew Matthew is its pronounced tendency to harmonize. Howard twice comments upon this: "A characteristic feature of Shem-Tob's Matthew is its harmonistic readings" (Howard 1995: 196); "There are many more readings in Shem-Tob's Matthew that agree with Mark and Luke than with John" (ibid.: 199). Interpolations from the other synoptics are very frequent. Usually these consist of just a word or two, but at one point a whole sequence of verses is added en bloc: Mark 9:20-28 (nine verses!) is dropped into the text of Matthew between Matt 17:17 and 17:19 (Matt 17:18 is omitted).

1.2.2 Tendency to Abbreviate

12. Another distinctive characteristic of Shem-Tob's Hebrew Matthew is its tendency to abbreviate. Despite its occasional interpolations (such as in example 4, above, from Matt 24:40-41), the overwhelming tendency of the compiler of this Hebrew Matthew is to give a shorter text, omitting much. Examples abound: see above, example 2, from Matt 28:19-20. Also consider the following examples: at Matt 8:4, Shem-Tob omits "in witness to them"; at Matt 8:24, Shem-Tob omits "but he was asleep"; Shem-Tob entirely omits Matt 16:6-8a and radically abbreviates Matt 16:9-12. Curiously, Howard never remarks upon this tendency to abbreviate, which is one of the Hebrew Matthew's most consistent, striking characteristics.

1.3 The Edition's English Translation

13. Howard's English translation is generally accurate, but at times frustrates. An example is the treatment accorded the distinctive Semitic grammatical construction known as the waw-consecutivum ("and" + the perfect [or imperfect] tense of a verb). As we will see below (see infra, sec. 2.1.1), Howard suggests that presence of the waw-consecutivum in Shem-Tob's Hebrew Matthew indicates that it was composed in the early Christian era, for, according to Howard, the waw-consecutivum ceased to be used after the decline of Biblical Hebrew (which occurred in the early Christian era). It is very odd, then, that Howard's translation often fails to render this feature, which he considers to be a key piece of evidence for the antiquity of this Hebrew Matthew. An example is Matt. 2:11, where, if one reads Howard's translation, one presumes there is one waw-consecutivum (italicized in the text below):

They came into the house, found him and his mother Mary, knelt before him, worshipped him, opened their sacks, and-brought to him gifts of gold, frankincense, and myrrh, that is, mira.

And-they-entered into the house and-they-found him and his mother Mary, and-they-knelt before him and-they-worshipped him and-they-opened their sacks and-they-brought to him gifts of gold and frankincense and myrrh, that is, mira.

14. Another problem with the translation is that it sometimes inverts the sequence of nouns in nominal sentences, against their sequence in the Hebrew text. While this is not grammatically incorrect, the result is that the translation does not accurately reflect the word order of the Hebrew. An example is Matt 3:9, where the Hebrew reads "Father-our Abraham"; Howard translates this as "Abraham is our father," instead of "Our father is Abraham." There is, however, no reason to abandon Hebrew word order here, especially when it is congenial with English word order. Furthermore, reversing the sequence of the nouns changes the emphasis found in the Hebrew: "Our father is Abraham" (= "We are the chosen people").

15. The result of these subtle moves away from a literal translation is that a knowledge of Hebrew remains requisite for accurate use of this document, for the English translation neither gives all of the words found in the Hebrew, nor presents them in the order in which they occur in Hebrew--even when the Hebrew word order could have been carried over, without awkwardness, into English.

16. In conclusion, Howard is to be heartily congratulated for producing a serviceable edition of the Hebrew Matthew known to Shem-Tob, and for giving us a useful and readable--although not entirely accurate--English translation.

2.0 Howard's Analysis of the Text and a Critique

17. The matter of the origin of Shem-Tob's Hebrew Matthew is, in Howard's analysis, dependent upon its dating. In his first edition, Howard presented numerous arguments to demonstrate that Shem-Tob's Hebrew Matthew was "primitive," as the title of that first edition proclaimed: The Gospel of Matthew according to a Primitive Hebrew Text (Howard 1987). A summary of the arguments and evidence in that first edition is given in a review done by your author for the Journal of Biblical Literature in 1989 (Petersen 1989). Virtually all of the arguments and evidence employed in that first edition are absent from this second edition. Despite the wholesale abandonment of this earlier argumentation, Howard still maintains that the Hebrew Matthew found in Shem-Tob is very ancient. What evidence does Howard muster to support his claim? He presents three sorts of evidence: (1) linguistic features, (2) lists of agreements with early Christian sources, and (3) parallels with the Gospel of John. Each of these three sorts of evidence will now be presented in turn, followed immediately by our critique.

2.1.0 Linguistic Features, Especially the waw-consecutivum

18. Howard describes the Hebrew of this Hebrew Matthew as "basically biblical Hebrew with a healthy mixture of Mishnaic Hebrew and later rabbinic vocabulary and idiom" (Howard 1995: 178); on the same page he also describes it as "standard biblical Hebrew (BH) with a mixture of Mishnaic Hebrew (MH) and even late medieval vocabulary." Among the features Howard identifies as being specifically "Biblical Hebrew" is the waw-consecutivum. Howard states that "a pronounced difference in BH and MH is the virtual disappearance of the consecutive tenses in MH" (ibid.: 179). Howard elaborates no further, so it is not entirely clear what his point is. One surmises, however, that what he wishes to argue is this:

1. Shem-Tob's Hebrew Matthew contains a feature which, according to Howard, is distinctive of "Biblical Hebrew"; and
2. this feature fell from use, according to Howard, after the decline of Biblical Hebrew; therefore,
3. Shem-Tob's Hebrew Matthew must have been composed in the period when Biblical Hebrew was still in use--in other words, in early Christian antiquity.

19. There are at least three rather obvious problems with this argument, none of which are considered by Howard.

20. (A) Howard never considers the possibility that the waw-consecutivum might be an artificial, "antique" feature used by the translator to give his medieval Hebrew translation of Matthew an "antique" flavor. Recourse to such "archaizing" is hardly unknown: many liturgies composed in the second half of the twentieth century contain anachronistic grammar and vocabulary ("O Lord, we beseech you . . . ."; etc.).

21. (B) Howard seems unaware that there is empirical, textual evidence that the waw-consecutivum was used long after the decline of Biblical Hebrew. Codex Vaticanus 32 is a late sixteenth- or early seventeenth-century translation of the gospels into Hebrew, made by a certain Dominicus of Jerusalem. Although raised a Jew, Dominicus eventually converted to Christianity. Familiar, as any person fluent in Hebrew would be, with biblical style, Dominicus' translation is redolent with the waw-consecutivum where a similar construction or sequence (conjunction + simple verb) is lacking in the Greek. Dominicus' translation, then, is empirical proof that the use of the waw-consecutivum in Hebrew did not end in Christian antiquity when Biblical Hebrew was replaced by Mishnaic Hebrew (as Howard seems to think), but continued to be used--especially when translating biblical texts, such as the gospels--even into the modern period.

22. (C) Finally, although distinct from the question of the use of the waw-consecutivum itself, there is the matter of the arrangement of the text--its sequence--where the construction is used. Greek hypotaxis, with its many participles, was unpopular in vernacular gospel translations made in the West during the medieval period. Rather than imitating the complex Greek grammatical structures, the vernacular translations used paratactic constructions, linked by interpolated ands. An example drawn from a particular manuscript, in Middle Dutch, copied about 1280, will illustrate the point. Howard offers a list of eight examples of the waw-consecutivum in Shem-Tob's Hebrew Matthew (Howard 1995: 179); we will use the first of his examples, which is from Matt 1:24, as our first illustration. The standard Greek of this verse reads:

And (de) arising from sleep, Joseph did . . . .

23. Now let us turn to our Western, medieval, vernacular gospel translation, in Middle Dutch. The holdings of the Liège University Library include MS 437, a gospel harmony known as the Liège Harmony. It was copied about 1280 C.E. (the latest date would be 1350), and, via a Middle Dutch archetype, goes back to a Latin Vorlage. The Middle Dutch Liège Harmony is written in the Zuid Limburgs (South Limburg) dialect; its text of Matt 1:24 reads:

And when Joseph awoke he arose and did that which . . . (Plooij 1929-1970: 22-23).

24. Now let us compare the text of Shem-Tob's Hebrew Matthew (where we restore the waws which Howard's translation inexplicably omits [see supra, sec. 1.3]):

And-he-awoke Joseph from his sleep, and-he-did according to all . . . .

25. There is, of course, no equivalent for the waw-consecutivum--with its "conversion" of the verb from a perfect to an imperfect, or vice-versa--in any language other than Hebrew. But the structural similarity of the Middle Dutch text and the Hebrew text at the point where the waw has been interpolated is beyond dispute.

26. The example we have just seen is not an isolated one. Consider the following, from Matt 3:14-15 (Matt 3:15 is Howard's third example of the use of the waw-consecutivum). The standard Greek text reads as follows:

14 And (de) John prevented him, saying: . . . .
15 And (de) Jesus answering said: . . . .

27. The Middle Dutch Liège Harmony avoids the hypotaxis of the Greek by using simple verbs joined by "and." It renders these verses in the following manner:

14 And John the Baptist refused it him and said: . . . .
15 Then Jesus answered him and said thus: . . . (Plooij 1929-1970: 45-46).

28. Let us see if the structure of Shem-Tob's Hebrew Matthew has any similarities with the Middle Dutch. The Hebrew, following Howard's translation, reads:

14 But [rather: And] John was doubtful about baptizing him and-said: . . . .
15 And-answered Jesus and-said: . . . .

29. This similar pattern of structure in the transformation of Greek participles into simple verbs, accompanied by the interpolation and identical positioning of conjunctions--indeed, even the moving of subjects ("Joseph," in the first example, at Matt 1:24)--need not indicate that the Middle Dutch Liège Harmony and Shem-Tob's Hebrew Matthew are mutually dependent upon the same source, for each may have independently hit upon the same solution for avoiding Greek hypotaxis. At the same time, however, one cannot preclude the possibility that these similarities in structure stem from dependence upon a common archetype. And since we know the Middle Dutch manuscript's Vorlage, it suggests that if there is dependence upon a common Vorlage, then that Vorlage is Latin. But whether such a common archetype might lie behind the two traditions can only be determined by a much more extensive analysis of distinctive variants (such an analysis will be found later in this article).

30. In conclusion, then, there are not only theoretical reasons (an attempt by the Hebrew translator to give his translation an artificial "biblical" flavor; the tendency of many vernacular translations to avoid Greek subordination and participial constructions by using simple verbs joined by "and"; possible mutual dependence of the Middle Dutch Liège Harmony and the Hebrew Matthew on a common Latin Vorlage which also circumvented Greek subordination by using simple verbs joined by "and"--which could have been rendered in Hebrew by either a simple waw-conjunctivum, or by the waw-consecutivum), but also empirical, textual reasons (Dominicus of Jerusalem's Codex Vaticanus 32) for disputing Howard's easy assumption that the use of the waw-consecutivum in Shem-Tob's Hebrew Matthew is evidence of composition in the first centuries of the Christian era.

31. Since we will be returning time and again to the Middle Dutch manuscript introduced above, namely, the Liège Harmony, it deserves a fuller introduction. This is provided in the following excursus.

2.1.3 Excursus: The Middle Dutch Liège Harmony

32. As noted above, this famous manuscript (Liège, Bibliothèque de l'Universitè, MS 437), written in Belgium in one of the local dialects, Zuid Limburgs, dates from about 1280 (at the earliest, but possibly as late as 1350). It was apparently copied from a now-lost Middle Dutch archetype, for it shares a common error with another Middle Dutch text composed in 1271: Jakob van Maerlant's famous Rijmbijbel (Maerlant 1858-1859; at Matt 27:7, both the Liège Harmony and the Rijmbibel read "poorters" [citizen's] for "potters" [potter's]). However, we know that behind this lost Middle Dutch archetype stands a Latin gospel harmony. This is demonstrated by (1) the preface of the Liège Harmony, which states that it was translated from Latin; (2) the manuscript's numerous Latinisms; (3) its numerous Vetus Latina and sometimes Vulgate readings; and (4) the practice of the time and place (Middle Dutch ecclesiastical literature in this period was all translated from Latin). Since Middle Dutch literature only begins to appear about 1200 (at the earliest), we can be certain that the Middle Dutch archetype of the Liège Harmony was translated from Latin no earlier than 1200. Later--probably about 1280 (but possibly as late as 1350)--a copy was made of this Middle Dutch archetype: that copy is the Liège Harmony.

33. Our knowledge of the Liège Harmony is so detailed because it has been known to scholarship for nearly two centuries (it was first edited in 1835 by G. J. Meijer). Further, because of its great textual significance (it has been used as manuscript evidence for the text of the Gospel of Matthew in the editions of the New Testament of Augustinus Merk [Merk, ed. 1964 and previous editions)] and in the first three editions of The Greek New Testament [UBS^{1, 2, 3}]), it has been the subject of much careful investigation, resulting in books by Daniël Plooij (including a magnificent edition with one of most impressive apparatuses ever constructed, published in Amsterdam between 1929 and 1970 in eight fascicles by the Royal Netherlands Academy of Arts and Sciences [Plooij 1929-1970]) and others, as well as articles by Adolf Jülicher (Jülicher 1924a, 1924b), F. C. Burkitt (Burkitt 1924), and Gilles Quispel (Quispel 1958, 1969, 1975b), among others. (For more on the Liège Harmony, see either Metzger 1977: 22-24 or Petersen 1994: 144-147, 170-195, et passim.)

34. The Liège Harmony is only one of a variety of vernacular gospel translations which stem from this earlier Latin harmonized gospel tradition: this Latin Vorlage spawned the Old High German, Middle Dutch, Old French, and Middle Italian gospel harmonies. Because of this common Vorlage, all of these Western medieval gospel harmonies share common sequences of harmonization and variant readings. But each also has its own unique textual complexion. This uniqueness stems from at least three factors: (1) the word choices made in the vernacular languages by the translator, (2) the individual transmission-history of each of these vernacular families after its translation, and (3) the fact that each of them rests upon a slightly different Latin Vorlage, for (a) they were translated at different places and at different times, and (b) the Latin Vorlage itself was neither monolithic nor static.

35. With this background in place, let us now return to our main text, and examine the second of Howard's classes of evidence for the antiquity of Shem-Tob's Hebrew Matthew.

2.2.0 Agreements with Early Christian Texts which Vanished in Antiquity

36. Howard's most important evidence for the antiquity of Shem-Tob's Hebrew Matthew consists of lists of agreements between it and texts known in ancient Christianity, but which then disappeared until recently. He adduces parallels with, among others, the Gospel of Thomas (found in 1945), Codex Sinaiticus (), 01) (first seen by Western eyes--those of Tischendorf--in 1859), the Vetus Syra (the two manuscripts were first edited in the nineteenth century), and the Vetus Latina. Howard's argumentation from each list is identical and involves three steps. (1) Parallels exist between the Hebrew Matthew in Shem-Tob's Even Bohan--a late fourteenth-century document, copied in the West--and these early Christian texts whose circulation was in Eastern Christendom. (2) None of these early Christian texts was known in or circulated in the West during the medieval period when Shem-Tob composed the Even Bohan. (In the case of the Vetus Latina, Howard feels it was not the source of this Hebrew Matthew, since the Vulgate had become the standard Latin gospel text after Jerome created it in the late fourth century; the possibility that Latin gospel texts other than the Vulgate [and Vetus Latina] might have circulated in the medieval period, and been the model from which the translator of Shem-Tob's Hebrew Matthew worked, does not seem to have occurred to Howard.) (3) Since none of these sources is known to have circulated in the West during the long sweep of medieval history, one is--according to Howard--driven to conclude that these common readings must be the result of this Hebrew Matthew having been composed in a time and in a place where works such as the Gospel of Thomas, the Vetus Syra, and Codex Sinaiticus were in circulation--in other words, in ancient Christianity. Let us examine four of Howard's lists and then critique them. We begin with Thomas, then address the Vetus Latina, followed by the Vetus Syra; finally we consider the parallels with Codex Sinaiticus (), 01).

2.2.1.1 Howard's List of Parallels with the Gospel of Thomas

37. Howard presents a list of 22 parallels with the Coptic Gospel of Thomas, recovered at Nag Hammadi, Egypt, only in 1945. Of these parallels, Howard writes, "It is highly unlikely that Shem-Tob had direct contact with the Gospel of Thomas. The agreements of his Matthew with Thomas, therefore, must be traced to the early centuries of the Christian era" (Howard 1995: 205, italics added), where, presumably, such "direct contact" occurred, resulting in these parallels.

2.2.1.2 Critique of the Parallels with the Gospel of Thomas

38. Here we will present and examine only the first seven of Howard's 22 readings, comparing them with the reading at the same point in the Middle Dutch Liège Harmony, the manuscript introduced above when discussing the waw-consecutivum (see secs. 2.1.2 and 2.1.3).

1. Matt 19:30

   39. Shem-Tob and Thomas logion 4 omit "and the last first." But the Liège Harmony (Plooij 1929-1970: 353) omits the whole verse. Thus, the Liège Harmony and Shem-Tob deviate in the same direction: both have omissions at this point, and the words "and the last first" are missing from both.

2. Matt 13:47 (sic! Howard must mean 13:48)

   40. Shem-Tob and Thomas logion 8 interpolate the word "fisherman (so Thomas)/fishermen (so Shem-Tob)." But the Liège Harmony (Plooij 1929-1970: 180) also interpolates the word "fishermen." (Note that, against the singular "fisherman" in Thomas, both Shem-Tob and the Liège Harmony interpolate the plural "fishermen.")

3. Matt 13:48

   41. Shem-Tob and Thomas logion 8 substitute the word "choose" for the Greek "harvest." But the Liège Harmony (Plooij 1929-1970: 180) also makes the identical substitution.

4. Matt 7:3

   42. Shem-Tob and Thomas logion 26 substitute the word "see" for the Greek "notice" (thus Shem-Tob and Thomas read "see" twice in this verse: Shem-Tob: "Why do you see the straw in the eye of the other person but not see the . . ."). But the Liège Harmony (Plooij 1929-1970: 82-83) also makes the identical substitution, and thus also reads "see" twice.

5. Matt 5:14

   43. Shem-Tob and Thomas logion 32 substitute "being built on" for "set on." But the Liège Harmony (Plooij 1929-1970: 67) also makes a substitution here: "that stands on" for the Greek "set on." Thus, both Shem-Tob and the Liège Harmony make similar (albeit not identical) substitutions at the same point in the text.

6. Matt 5:15

   44. Shem-Tob and Thomas logion 33 interpolate "in a hidden place." But the Liège Harmony (Plooij 1929-1970: 67) also interpolates "in a hidden place" in this verse.

7. Matt 5:15

   45. Shem-Tob and Thomas 33 interpolate "he puts (so Thomas)/they place (so Shem-Tob)." But the Liège Harmony (Plooij 1929-1970: 67) also gives an interpolation here: "one sets." Thus, both Shem-Tob and the Liège Harmony make very similar--albeit, not identical--interpolations at the same point in the text.

46. We could go on like this, through all 22 of Howard's parallels. In point of fact, of the 22 parallels with the Gospel of Thomas which Howard presents, eleven--fully half of Howard's total--are found in the Liège Harmony in either the identical or a very similar form (Howard's nos. 1, 2, 3, 4, 6, 7, 13, 17, 20, 21, 22 = Plooij 1929-1970: 353, 180-181, 181, 83, 67, 67, 125-126, 169, 110, 110, 111). Additionally, Howard's no. 10 is found in the Middle Dutch Stuttgart Harmony (Bergsma, ed. 1895-1898: 52), a harmony which is intimately related to the Liège Harmony, and Howard's nos. 5 and 16 are perhaps related to the similar deviating (from the standard Greek) readings in the Liège Harmony (Plooij 1929-1970: 67, 177). But this is not all.

47. The first scholar to notice the common readings between the Gospel of Thomas and documents such as the Liège Harmony was Gilles Quispel, in his article "L'Évangile selon Thomas et le Diatessaron" (Quispel 1959). The existence of such agreements is now well known to scholars (it is remarked upon by all standard works in the area: e.g., Metzger 1977: 29-30). Numerous other scholars (e.g., A. F. J. Klijn [Klijn 1969], R. McL. Wilson [Wilson 1982], Tj. Baarda [Baarda 1983, 1997], J.-É. Ménard [Ménard 1975], C. Tuckett [Tuckett 1988], et al.) have investigated these agreements as well and generated an enormous body of scholarship on the matter. It seems that either because Thomas and Tatian's Diatessaron used the same early gospel text (the preference of Quispel and Klijn), or because Thomas used the Diatessaron (the preference of Baarda), or because the Diatessaron used Thomas (the suggestion of Ménard), common readings are shared between witnesses to the Diatessaron (such as the Liège Harmony) and Thomas (for a summary of the debate, bibliography, and examples of the parallels, see Petersen 1994: 272-279; 292-300). Thus, while the precise mechanism by which this sharing came about remains obscure, it is simply a fact--agreed to by all parties, and textually beyond dispute--that such common readings exist.

48. In his study Tatian and the Gospel of Thomas (Quispel 1975b), Quispel included in Appendix III ("Variants Common to the Diatessarons and the Gospel of Thomas," pp. 174-190) an exhaustive list of Thomas readings in (1) the Eastern harmonized gospel tradition, (2) the Western harmonized gospel tradition (of which the Liège Harmony is part), and (3) other sources. Although Howard ignores this source (as, indeed, he ignores the whole discussion of the relationship between Thomas and the Western harmonized gospel tradition [pace his lone reference to Baarda [Howard 1995: 203-204]), each and every one of the 22 parallels Howard lists is on Quispel's list, which was published twenty years ago! Some examples: of Howard's first parallel (#1 above, omit "and the last first" from Thomas logion 4/Matt 19:30), Quispel notes the same omission in two manuscripts (MSS Q H) of the Middle Italian Tuscan Harmony. Of Howard's sixth parallel (#6 above, interpolate "in a hidden place" at Thomas logion 33/Matt 5:15), Quispel notes the same interpolation in the West in three related Middle Dutch Harmonies (the Liège, Stuttgart, and the Haaren Harmonies), the Middle Italian Tuscan Harmony, the Old High German Harmony (the Old High German column in bilingual [Latin and OHG] Codex Sangallensis), and three Latin harmonies (Codices Fuldensis, Sangallensis, and Cassellanus); it also appears in several Eastern witnesses to the harmonized gospel tradition, as well as some Greek manuscripts.

49. Howard has said of his list of 22 agreements with Thomas that, since "it is highly unlikely that Shem-Tob had direct contact with the Gospel of Thomas," therefore, these "agreements . . . must be traced back to the early centuries of the Christian era" (Howard 1995: 205, italics added). It is easy to make such grand assertions if one is unaware of the relevant modern scholarship and one is unfamiliar with the pertinent medieval sources. But anyone familiar with the requisite materials will find three problems with Howard's claim.

50. (1) Howard appears confused when he states that "it is highly unlikely that Shem-Tob had direct contact with the Gospel of Thomas," for it is clear to everyone--and Howard himself has argued this point (see supra, sec. 1.0)--that Shem-Tob did not translate this Hebrew Matthew, but incorporated an already existing (and, therefore, older) Hebrew Matthew into his Even Bohan. Therefore, Howard should not be focusing on whether Shem-Tob had access to Thomas, but whether the unknown translator of this Hebrew Matthew--working in, say, 1100 C.E.--might have had access to Thomas readings.

51. (2) Howard casts the problem as one focusing on the question of whether "Shem-Tob had direct contact with the Gospel of Thomas"; but again, even if we substitute "unknown translator of this Hebrew Matthew" for "Shem-Tob," Howard's description of the problem still shows a misunderstanding of the issues involved. To the best of our knowledge no medieval author "had direct contact with the Gospel of Thomas," which apparently did indeed "vanish" from circulation in the first three or four Christian centuries. Yet we know (cf. the history of Thomas research, and the articles of Quispel [Quispel 1959], Klijn [Klijn 1969], Ménard [Ménard 1975], Baarda [Baarda 1983, 1997], et al.) that numerous Western medieval sources (the Middle Dutch Liège Harmony, the Middle Italian Tuscan and Venetian Harmonies, the Old High German Harmony, the Latin Harmonies [Codex Fuldensis, Codex Cassellanus, Codex Sangallensis, etc.], etc.) contain parallels with the Gospel of Thomas (cf., e.g., the discussion just above concerning Thomas logion 33 [Matt 5:15] "in a hidden place," which survives in no fewer than three Latin harmonies). If one were to treat this evidence (Thomas parallels in Western medieval gospel texts) as Howard has treated the identical evidence in his Hebrew Matthew, then we would have to argue that all of these Western medieval gospel texts "must" date back "to the early centuries of the Christian era," whence they acquired these readings from "direct" contact with Thomas. Yet no scholar makes such a claim. Why?

52. It is because these scholars know the transmission-history of the gospels in the Western medieval world--which is, after all, the period from which Shem-Tob's Hebrew Matthew stems. Scholars know that Thomas readings are found in the harmonized gospel texts, both East and West, which indirectly descend from the Diatessaron of Tatian. In the West, sometime between the second and beginning of the sixth century, a Latin translation, revised, of the Diatessaron was made. This Latin gospel harmony underwent many further revisions and modifications (including resequencing and removing non-standard readings) during its transmission-history. It fractured into numerous families, of which at least four large ones can be identified today ([1] pre-Codex Fuldensis harmonies [i.e., pre-546, the date of Codex Fuldensis]; [2] Fuldensis and its closest allies [sixth to ninth centuries]); [3] later harmonies [twelfth to fifteenth centuries]; and [4] the so-called "Munich" family (the name was coined by H. J. Vogels) [Munich, MSS Clm. 10025 and Clm. 23977]). Each of these has its own distinct character.

53. Sprouting from one or another of these preserved or now-lost Latin traditions are the Western vernacular gospel harmonies, each translated from its own distinct Latin Vorlage: the Old High German in the mid-ninth century; the Middle Dutch at about 1200; the Middle Italian in the twelfth or thirteenth century; an Old French harmony in the twelfth century. Some of these families went on to spawn other vernacular harmony families: the Middle English was translated from the Old French in about the thirteenth century; the Middle High German was translated from the Middle Dutch in the late thirteenth or early fourteenth century. Nevertheless, all of these various vernacular gospel texts trace back to a Latin Grundlage, and that Latin Grundlage ultimately traces back--through numerous intermediary texts--to the Diatessaron of Tatian. Therefore, when we find a Thomas parallel in the thirteenth-century Middle Italian Harmonies, or in the Middle Dutch Liège Harmony, or in the late thirteenth-century Middle English Harmony, no informed scholar presumes that this reading came from "direct contact" with the Gospel of Thomas, or that this Middle Italian or Middle Dutch or Middle English Harmony is a translation of some ancient gospel text. Rather, because we are well informed about the genesis of the Western medieval harmonized gospel tradition, an informed scholar knows the route by which the Thomas reading arrived in Western medieval vernacular texts: it was taken over from the Latin archetype from which the vernacular translation was made, perhaps as late as the thirteenth or fourteenth century.

54. (3) Howard's posing of the problem completely ignores not just (a) the possibility that Thomas readings might have entered intermediary gospel texts which circulated in the medieval world, from which the translator of this Hebrew Matthew might have obtained the Thomas reading, but also (b) the fact that we know with absolute certainty that such intermediary texts, in Latin, existed (we possess over twenty such Latin harmonies today), and that they are the conduit by which Thomas readings reached the Western medieval gospel harmonies.

55. In order to investigate texts in a responsible manner, one needs to know not just the pertinent documents from the past (in this case, Western medieval harmonized gospel texts), but also the pertinent modern scholarship (in this case, the vast body of literature pertaining to Thomas readings in the Western medieval gospel texts). Howard, however, appears unaware of both; the result is that his list of parallels with Thomas is meaningless, for every single one of the 22 parallels with Shem-Tob's Hebrew Matthew is found in other, directly contemporary gospel texts circulating in the West. Why, then, "must" (Howard's word) the point of "direct contact" between this Hebrew Matthew and Thomas "be traced to the early centuries of the Christian era"? There is only one thing which Howard's list of parallels with Thomas demonstrates: the closeness of the text of this Hebrew Matthew to the Western medieval harmonized gospel tradition, especially as represented in the Middle Dutch Liège Harmony.

2.2.2.1 Howard's List of Parallels with the Vetus Latina

56. Howard presents a list of twelve parallels with the Vetus Latina ("occasionally accompanied by the Vulgate," N.B.!). He claims that these parallels are "against all other Matthean witnesses" (Howard 1995: 195).

2.2.2.2 Critique of the Parallels with the Vetus Latina

57. Ignoring the problem posed by the fact that the parallels "occasionally" occur in the Vulgate (which means that Shem-Tob's Hebrew Matthew might have acquired these readings from no more antique or exotic a source than a late-medieval Vulgate manuscript), Howard's claim that these readings are "against all other Matthean witnesses" is simply not true. Of his twelve readings, six--once again, half of Howard's list--are found in just a single source, namely the Middle Dutch Liège Harmony (if the comparison were expanded to include all the Western gospel harmonies, the number would increase). Specifically, Howard's nos. 1, 2, 6, 7, 10, 11 are paralleled in the Liège Harmony (Plooij 1929-1970: 19, 115-116, 252, 127, 282, 292). Howard's claim of exclusivity for these parallels with the Vetus Latina is simply untrue; the readings are widespread in the Western medieval harmonized gospel tradition. What, then, can be the relevance of presenting a list such as this, upon which no argument can be built?

2.2.3.1 Howard's List of Parallels with the Vetus Syra

58. Next Howard presents a list of fourteen readings in which Shem-Tob's Hebrew Matthew parallels the Vetus Syra. Of these, Howard writes:

Since the Old Syriac perished in early antiquity, it does not seem likely that Shem-Tob, writing in fourteenth century Spain, had access to it. The many readings shared by Shem-Tob and the Old Syriac, therefore, strongly suggest a relationship, whose roots go back to the early centuries of the Christian era (Howard 1995: 196, italics added).

59. Once again, however, Howard's claim that his parallels between Shem-Tob's Hebrew Matthew and the Vetus Syra are unique is simply false. Numerous readings are paralleled in Western harmonized gospel texts; indeed, singling out just one such text, four of his twelve readings are found in the Liège Harmony, and a fifth approximates its reading. Specifically, Howard's nos. 10, 11, 13, 14 (partial) are found in the Liège Harmony (Plooij 1929-1970: 217, 217, 476, 560), and Howard's no. 5 is similar to the Liège Harmony's reading (Plooij 1929-1970: 105). Placed in context and stripped of the demonstrably false claim that the parallels are unique, what possible scholarly use can such a list have? The answer, of course, is: none.

60. Apropos of parallels between the Vetus Syra and Shem-Tob's Hebrew Matthew, Howard has stated on the TC-List (Howard 1997) that Shem-Tob's Hebrew Matthew "perhaps has more parallels with the Old Syriac than any other textual tradition." There are two problems with this assertion.

61. First, before making this assertion, one must ask whether Howard has examined the Western harmonized gospel tradition, including the Liège Harmony, for these same readings. If he has not (and there is no indication that he has), then this claim proves nothing more than that Howard still does not understand either the issues involved or how one constructs an argument of dependence in textual criticism. Any informed scholar would immediately ask: How many of these parallels with the Vetus Syra also show up in the Vetus Latina and/or the Western harmonized gospel tradition? Next, an informed scholar would ask: How many of these parallels with the Vetus Syra are unique? The reason for asking these questions is that (1) the Vetus Latina and the Western harmonized gospel tradition are awash with parallels with the Vetus Syra, and (2) they go to the heart of a basic principle of textual criticism: if one desires to demonstrate dependence between two documents, then the textual links must be unique and cannot be found in any intervening sources from which the later document might have obtained the reading. For this reason, any list of parallels between Shem-Tob and the Vetus Syra which Howard presents, but which has not first been vetted by comparison against the readings of the Western harmonized gospel tradition, the Vetus Latina, and the Greek manuscript tradition, would be as useful as his list of parallels with the Vetus Syra, Thomas, or the Vetus Latina, which we have already dissected (see supra, secs. 2.2.2.1-2.2.3.2).

62. Second, Howard's claim ignores the fact that numerous parallels between the Vetus Syra and Shem-Tob's Hebrew Matthew would be due to the syntax, idioms, and grammar of sister Semitic languages: e.g., pleonastic pronouns, paratactic constructions joined by "and," Semitic word order, etc. In short, when the matter involves two Semitic languages (as it does in Howard's claim of "more parallels with the Vetus Syra"), no inference as to dependence can be drawn from linguistic evidence. Given these circumstances and facts, Howard's assertion on the TC-List is meaningless.

2.2.4.1 Parallels with Codex Sinaiticus (), 01)

63. Another of Howard's lists consists of five parallels with Greek Codex Sinaiticus (), 01). To Howard, these five parallels are "significant" (Howard 1995: 192), for they once again link Shem-Tob's Hebrew Matthew with a text of Christian antiquity. Howard writes:

The type of text [Codex Sinaiticus] represents, Alexandrian with a strain of "Western" type readings, fell out of general use during the Middle Ages and was replaced by the Byzantine text. . . . The disparity in time and geography between Shem-Tob and Codex Sinaiticus strongly suggests that [Shem-Tob] had no direct knowledge of or contact with this biblical manuscript. The roots for their agreement, therefore, must go back to the early centuries of the Christian era (ibid.: 192, italics added).

64. None of these five readings shows up in the Liège Harmony, and your author has not invested the time necessary to check them against the twenty or so harmonies and other texts and editions necessary to test Howard's assertion. The possibility exists, however, that--as in the two other cases where Howard claimed uniqueness for his parallels--upon examination these readings will also appear in other documents. But leaving that speculation aside, a new problem bedevils this list. No less than four of these five readings are omissions (Howard's other lists also contain omissions: his list of parallels with the Vetus Syra contains three, his list of Vetus Latina parallels has four, his Thomas list has five omissions, etc.). Including omissions in such lists ignores two elementary canons of textual criticism.

65. First, it ignores the dictum that arguments from omissions are intrinsically weak. This is because omissions can arise from any number of reasons other than an omission in the exemplar: from fatigue, parablepsis, lacunae, homoioarcton, homoioteleuton, shortage of materials, etc.; none of these requires an omission in the archetype. In short, an omission marks a lack of evidence. (On when and with what restrictions omissions might be used in textual arguments, see Petersen 1985: 108-109.)

66. Second, and even more to the point, Howard appears oblivious to the fact that Shem-Tob's Hebrew Matthew is, by and large, an abbreviating text. He never comments upon this distinctive and patently obvious feature of its text (see supra, sec. 1.2.2). Even under the best of circumstances, arguments from an omission are highly suspect; in a case such as this, where the document itself is (generally speaking) an abbreviating text, venturing an argument from an omission is absurd.

2.3.0 Johannine Elements in Shem-Tob's Hebrew Matthew

67. While the four preceding lists, together with the evidence of the waw-consecutivum, show (according to Howard) that Shem-Tob's Hebrew Matthew was composed in early Christian antiquity, the evidence which stipulates the date of this composition is, according to Howard, a series of Johannine fragments found in the text of Shem-Tob's Hebrew Matthew (Howard 1995: 197-199). Howard registers eighteen places where such Johannine elements are found, tucked into the Hebrew text. What does Howard make of these?

68. According to Howard, these are not interpolations of Johannine material into the Matthean text. Rather, says Howard, they are not Johannine at all, but part of a very ancient version of Matthew which was used by the Fourth Evangelist when he composed the Gospel of John. In other words, when these words crop up in our present text of John, they are not originally Johannine; rather, they are words from a "Shem-Tob type of text of Matthew" known to the author of the Gospel of John when he was composing his gospel about the year 100. According to Howard, the author of John used this "Shem-Tob type of text of Matthew," appropriated these fragments of text from it, and inserted them into his autograph of the Fourth Gospel. Here is Howard's statement: the presence of these snippets of (what is today) Johannine text in Shem-Tob's Hebrew Matthew "suggest[s] that the author of the Fourth Gospel knew a Shem-Tob type of text of Matthew and used this text when he wrote his gospel" (Howard 1995: 199).

69. This is an astonishing statement. Howard does not pursue its implications, but your author will. Leaving aside the matter of synoptic dependency, the most significant inference concerns the date of this "Shem-Tob type of . . . [Hebrew?] Matthew." ⁵², which is dated to about 125 C.E., is generally accepted as setting the terminus ante quem for composition of the Gospel of John. Howard's statement that "the author of the Fourth Gospel knew a Shem-Tob type of text of Matthew and used [it] when he wrote his gospel" can only mean that a Shem-Tob type of (Hebrew?) Matthew, sharing these same "Johannine" features, must have existed before 125 C.E., the date of ⁵². In short, although Howard never states it, one can only conclude that he feels these Johannine elements in Shem-Tob's Hebrew Matthew set a terminus ante quem for a text very similar to Shem-Tob's Hebrew Matthew as well: it existed before 125 C.E.

2.3.2 Critique of the Parallels with the Gospel of John

70. There are, once again, many problems with both the logic and the facts of Howard's assertion and his list of Johannine parallels. We address the logical errors first and then the textual ones.

71. (1) Was it not Howard himself who noted that "harmonistic readings" were a "characteristic feature" of this Hebrew Matthew? And was it not Howard who stated that "There are many more readings in Shem-Tob's Matthew that agree with Mark and Luke than with John" (Howard 1995: 199; already cited above, in sec. 1.3)? In view of this acknowledged tendency of Shem-Tob's Hebrew Matthew to harmonize, why, then, must these eighteen bits of John in Shem-Tob's Hebrew Matthew be pre-Johannine? As obvious as this question is, Howard neither answers it, nor does he present any additional argumentation or evidence; one is simply left with his assertion that these Johannine snippets suggest that "the author of the Fourth Gospel knew a Shem-Tob type of text of Matthew and used [it] when he wrote his gospel"--an assertion which, nota bene, runs counter to his own description of the work as one which has a pronounced tendency to harmonize.

72. (2) Howard's assertion ignores the rather obvious problem of explaining how these "Johannine" elements--if, as Howard proposes, they were originally part of a "Shem-Tob type of text of [a Hebrew?] Matthew" in pre-Johannine times--presumably came to be expunged from all copies of Matthew other than Shem-Tob's Hebrew Matthew. Normally, one would think that, if such an ancient text of Matthew were important enough to have been used by the Fourth Evangelist, then some trace of it would have survived in other ancient manuscripts of Matthew, or in patristic citations of Matthew from the first Christian centuries. But, no: elements of this pre-Johannine Matthew apparently survive only in the text Howard is editing. (If the situation is otherwise--if parallels for these [for Howard] very ancient and significant fragments of "John" exist in other manuscripts of Matthew--then Howard has not deemed it important enough to inform his readers of the fact. [But see infra, sec. 3.0, reading "D".]) It is incumbent on the proponent of such a remarkable thesis to anticipate and answer such obvious objections. Howard, however, does not.

73. (3) When Howard encounters the identical phenomenon in relation to the other gospels--i.e., fragments of Mark and Luke embedded in Shem-Tob's Hebrew Matthew--he does not come to the same conclusion. This is very odd for, in the absence of any qualifications (and Howard provides none), one would assume that identical phenomena ([1] the bits of Mark and Luke in Shem-Tob's Hebrew Matthew, and [2] the bits of John in Shem-Tob's Hebrew Matthew) should be interpreted identically. If Howard were consistent, then he should also argue that the fragments of Mark and Luke in Shem-Tob are evidence that Mark and Luke (like John) used a Shem-Tob-type of Matthew when they wrote their gospels. For reasons not explained, Howard does not.

74. The three logical flaws just described are not, however, the only problems facing Howard's list of eighteen parallels with the Gospel of John; there are significant textual problems as well.

75. First, apparently Howard has not bothered to screen his examples for the normal word choice of his Hebrew Matthew. For example, the first "Johannine" element cited in Howard's list is what he takes to be the substitution of acioj from John 1:27 for the Matthean ikanoj at Matt 3:11; Howard points to the use of yw)r here in the Hebrew Matthew's text of Matt 3:11, indicating that this is the Hebrew Matthew's translation of acioj. But at Matt 8:8, in another place where the Greek text reads ikanoj, Shem-Tob's Hebrew Matthew also translates ikanoj by the Hebrew yw)r. In other words, it appears that Shem-Tob's Hebrew Matthew uses yw)r as its standard translation of ikanoj, as shown by his use of it in Matt 8:8--and Matt 3:11; it may also use it to translate acioj. But if it is used to translate both, then how can one argue that a Johannine element has been interpolated here? The answer is simple: one cannot.

76. Second, by now it should come as no surprise that nine of these same "Johannine" elements--once again, half of Howard's list--are also found in the Middle Dutch Liège Harmony, harmonized in the same fashion, at the same point in its text. Specifically, Howard's nos. 1, 2, 3, 7, 13, 14, 15, 16, and 17, are found in the Middle Dutch manuscript (Plooij 1929-1970: 45, 45, 46, 196, 691, 691, 699, 723, 723). If one were, mutatis mutandis, to apply Howard's logic to this identical evidence in the Liège Harmony, then one would have to argue that the text of this Middle Dutch gospel harmony, like the text of Shem-Tob's Hebrew Matthew, antedates the Gospel of John. This is, of course, absurd, for we know that we are (1) dealing with a gospel harmony, (2) from the medieval period, whose direct ancestry is, after first passing through a Middle Dutch archetype, (3) Latin.

77. Here is an example from Howard's list of Johannine elements in the Hebrew Matthew, placed side-by-side with the text of the Liège Harmony. In Howard's list of eighteen parallels, his no. 17 is the interpolation from John 18:39 of the word "Passover" into Matt 27:15. The standard Greek text of Matthew reads:

And during [the] feast (eorthn) the governor was accustomed . . . .

On the day of the honored feast of Passover, it was their custom . . . .

At the feast of the Passover it was the custom . . . .

78. It is sufficient to say that Howard's list of Johannine elements in Shem-Tob's Hebrew Matthew proves nothing beyond the fact that we are dealing with a harmonized gospel text, which has extensive links with the textual tradition found in the Middle Dutch Liège Harmony.

3.0 Parallels between the Liège Harmony and Shem-Tob's Hebrew Matthew

79. By now it is apparent that there are numerous parallels between Shem-Tob's Hebrew Matthew and the Middle Dutch Liège Harmony. For example, four of Howard's lists of parallels (with [1] Thomas, [2] the Vetus Syra, [3] the Vetus Latina, and [4] the Gospel of John) contain a total of 66 readings. Although these lists consist of readings from Coptic (Thomas), Syriac, Latin, and Greek sources, and despite Howard's claim that the parallels in two of the lists were "against all other Matthean witnesses," nevertheless, a minimum of 30 (and a possible maximum of 34) of the parallels are also found in a single manuscript--the Middle Dutch Liège Harmony--which is separated from the time of Shem-Tob by less than a century, and in distance by less than 1,000 km (650 miles). And note that not only do these lists come from a wide range of documents, they also were not compiled by your author: Howard compiled them. Nevertheless, half of the Thomas parallels (a minimum of 11 out of 22 readings), half of the Vetus Latina parallels (6 out of 12 readings), 4 of the 14 Vetus Syra readings, and half of the Gospel of John parallels (9 out of 18) were also found in the Liège Harmony.

80. Many of these parallels, however, also occur in other members of the Western medieval harmonized gospel tradition as well (see supra, sec. 2.2.1.2, apropos of Quispel's list of Thomas parallels), or in the normal Greek or Latin manuscript tradition. Here are some additional examples of agreements between Shem-Tob's Hebrew Matthew and the Liège Harmony, accompanied by an indication of where else these same readings occur.

1. Matt 3:10 + Luke 3:10

   81. Shem-Tob and the Liège Harmony both interpolate Luke 3:10 immediately after Matt 3:10; the Hebrew Matthew reads: "(Matt 3:10) Already the axe has reached the root of the tree; the one which does not produce good fruit will be cut down and burned in the fire. (Luke 3:10) The crowds asked him: if so what shall we do? John answered them: He who has two shirts let him give one to him who has none." But the identical interpolation of Luke 3:10 after Matt 3:10 is also found in the Arabic and Persian Harmonies in the East, and Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), and the Middle Italian Tuscan Harmony in the West.

2. Matt 7:16

   82. Shem-Tob's Hebrew Matthew and the Liège Harmony both substitute "deeds" for "fruits." But the same substitution is also found in the Georgian Version (Geo²: "operibus", as per Legg, ed. 1940: ad loc.).

3. Matt 8:1

   83. Shem-Tob's Hebrew Matthew and the Liège Harmony both substitute "Jesus" for "he." But the same substitution is also found in the Middle Italian Venetian and Tuscan Harmonies, and the Vulgate.

4. Matt 8:1

   84. Shem-Tob's Hebrew Matthew and the Liège Harmony both read the singular "crowd" for the Greek plural "crowds." But the same substitution is also found in Greek MS 33 and Vetus Latina MS h.

5. Matt 9:23

   85. Shem-Tob's Hebrew Matthew and the Liège Harmony both interpolate "weeping" (a word drawn from Mark 5:38 and/or Luke 8:52) in this verse. But the word is also interpolated in this verse in the Arabic and Persian Harmonies, in Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), the Middle Italian Venetian and Tuscan Harmonies, the Middle High German Zürich Harmony, and the Middle English Pepysian Harmony.

6. Matt 11:2

   86. Shem-Tob's Hebrew Matthew and the Liège Harmony both interpolate "two" into this verse ("sent two of his disciples"). But the same interpolation is also found in numerous Greek MSS, in Vetus Latina MSS g¹ aur ff¹ l, and the Vulgate.

7. Matt 12:1

   87. Shem-Tob's Hebrew Matthew and the Liège Harmony both interpolate (from Luke 6:1) "to crush them between their hands" (with minor variations). But the same interpolation is found in Vetus Latina MS c and the Curetonian Syriac (Syr^c), as well as the Arabic and Persian Harmonies, Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), the Middle Italian Venetian and Tuscan Harmonies, and the Middle English Pepysian Harmony.

8. Matt 22:4

   88. Shem-Tob's Hebrew Matthew and the Liège Harmony both substitute "fowl" for the standard "fattened calves"; they read "I have killed oxen and fowl." But this same substitution is also found in the Vetus Latina (MSS ff¹ aur l) and the Vulgate, as well as in Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), and the Middle Italian Venetian and Tuscan Harmonies.

   89. (N.B.: The singular reading of Shem-Tob's Hebrew Matthew and the Liège Harmony presupposes a Latin Vorlage which read altilia (as do the Latin MSS listed above) and not saginata [the reading of the rest of the Vetus Latina], for although both altilia and saginata mean "fat, fattened" (and thus correctly translate the standard Greek reading sitista), only "altilia" also has the meaning "a fattened bird" [cf. Lewis and Short 1879, s.v. altilis]. In other words, the reading "fowl" in Shem-Tob's Hebrew Matthew and the Liège Harmony cannot be explained by any reading in the Greek or the Latin manuscript tradition except the Latin "altilis." This is only one of several instances where a Latin Vorlage is clearly visible behind a shared "mistranslation"--or, better, a shared tendentious translation--in Shem-Tob's Hebrew Matthew and the Liège Harmony.)

9. Matt 24:31

   90. Shem-Tob's Hebrew Matthew and the Liège Harmony both interpolate "and with a great shout" after "with a trumpet." But this same interpolation is also found many Greek MSS (D 1241, etc.), the Vetus Latina, the Vulgate, Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), as well as the Middle Italian Tuscan Harmony.

10. Matt 25:1

    91. Shem-Tob's Hebrew Matthew and the Liège Harmony both interpolate "and a bride" after "bridegroom" in this verse. The same interpolation is also found in Codex Fuldensis, Codex Sangallensis (Latin and Old High German columns), and the Middle Italian Tuscan Harmony.

92. Despite the fact that all ten of these readings are found in both the Liège Harmony and Shem-Tob's Hebrew Matthew, one cannot use them as evidence of a common textual Vorlage; the reason is that there are numerous other sources which contain the same variants. Shem-Tob's Hebrew Matthew might have acquired some of these readings from the Western harmonized gospel tradition, some from the Vetus Latina, some from the Vulgate, and still others from Greek manuscripts; there is no way to prove from which of these sources Shem-Tob's Hebrew Matthew acquired these readings. (This is precisely the problem which Howard failed to consider when he presented his lists of agreements between Shem-Tob and Thomas, the Vetus Syra, and the Vetus Latina, and which makes these lists useless as evidence.)

93. While the sheer number of agreements with the Liège Harmony suggests that some sort of relationship exists between it and the Hebrew Matthew, one cannot stipulate the proximity or character of that relationship unless one can find unique variant readings or sequences of harmonization between the two documents. Furthermore, these unique readings must be striking enough (Burkitt [1899: xi] called them "well marked variations," that is, readings which are not just a "this" interpolated here, or a plural noun in place of a singular noun) so as to preclude the possibility that they arose by chance.

94. Here, now, are four such unique agreements. They have been checked against and found wanting in the editions of Tischendorf, von Soden, Legg, Jülicher, Wordsworth and White, and Burkitt's edition of the Vetus Syra; because of the widespread and diverse nature of the harmonized gospel tradition, they have also been checked against and are, except as noted, absent from all the major gospel harmony traditions: Codex Fuldensis, Codex Cassellanus, Codex Sangallensis (Latin & Old High German), the Middle Italian Venetian and Tuscan Harmonies, the Middle High German Zürich Harmony, the Middle English Pepysian Harmony, the Arabic Harmony, the Persian Harmony, and Ephrem's Commentary (Syriac and Armenian).

1. Matt 25:6

   95.

   • Greek = . . . at midnight there was a cry: "Behold . . . ."
   • Shem-Tob = . . . at midnight that behold a voice was heard: "Behold . . . ."
   • Liège = . . . to midnight there came a cry and a voice which spoke thus: "The . . ." (Plooij 1929-1970: 617).

   The point of comparison is the identical interpolation of the word "voice"; other than Shem-Tob's Hebrew Matthew and the Liège Harmony, this variant is, to the best of your author's knowledge, unattested.

2. Matt 25:9

   96.

   • Greek = "Perhaps there will not be enough for us and for you; go rather to the ones selling and buy for yourselves."
   • Shem-Tob = "Go now to those who sell and buy for yourselves because there is not enough oil for us and you."
   • Liège = "Go to those who sell oil and buy there, lest there be any wanting to both us and you" (Plooij 1929-1970: 617-618).

   The point of comparison is the inversion of the position of the first and second clauses. This inversion is found only in Shem-Tob and the Liège Harmony; to the best of your author's knowledge, it is otherwise unknown. It is possible that this inversion is due to a scribal error: the scribe erred by copying the second clause first, and then, rather than erasing it, decided (since the order is irrelevant in the sentence) to continue by copying the omitted first clause after the already written second clause. If this is the correct explanation for this inversion, then this reading is a common error, one of the surest signs of textual dependence.

3. Matt 26:2

   97.

   • Greek = . . . and the Son of Man will be delivered up to be crucified.
   • Shem-Tob = . . . and the Son of Man will be delivered into the hand [sic!] of the Jews for the gallows.
   • Liège = . . . and the son of man shall be delivered to the Jews, who shall crucify him (Plooij 1929-1970: 627).

   The interpolation of "the Jews" is unique to Shem-Tob, the Liège Harmony, and Liège's two Middle Dutch daughters, the Stuttgart and The Hague Harmonies (dated 1332 and 1473, respectively), and a Middle High German cousin of the Liège Harmony, the Zürich Harmony (dated 13/14th century, and dependent upon the Middle Dutch tradition). Note: this is one of several readings which suggest this Hebrew Matthew was the product of Christian hands, possibly even the hands of a Jewish convert to Christianity.

4. Matt 27:15

   98.

   • Greek = And at the feast the governor was accustomed . . . .
   • Shem-Tob = On the day of the honored feast of Passover, it was their custom for the commander of the city . . . .
   • Liège = At the feast of the passover it was the custom that the judge . . . (Plooij 1929-1970: 723).

   The interpolation of "of [the] Passover" is unique to Shem-Tob, the Liège Harmony, and Liège's two Middle Dutch daughters, the Stuttgart and The Hague Harmonies, and the Middle High German Zürich Harmony (which is dependent upon the Middle Dutch tradition). No other sources known to your author contain this interpolation. Ironically, your author's attention was drawn to it because it stood (as #17) on Howard's list of Johannine elements found in the Hebrew Matthew (Howard contends it comes from John 18:39; see supra, sec. 2.3.2).

99. But this is not all. In addition to these four unique readings, your author presents nine more such unique readings in an article which has been accepted for publication in the journal New Testament Studies (Petersen 1998); additionally, that forthcoming article contains one non-unique parallel. This means that, between this TC article and the forthcoming NTS article, thirteen unique readings between Shem-Tob's Hebrew Matthew and the Liège Harmony will have been published.

100. But there is still more. In the 1989 JBL review of the first edition of Howard's book, your author adduced eleven additional agreements between the Liège Harmony and Shem-Tob's Hebrew Matthew, at least one of which was unique. This means that your author has published (or had accepted for publication) 62 readings in which Shem-Tob's Hebrew Matthew agrees with the Middle Dutch Liège Harmony. And of the 62, at least fourteen are unique readings.

101. Your author's researches indicate that probably more than a thousand non-unique agreements (i.e., readings found also in Greek or versional manuscripts, fathers, or non-Middle Dutch-dependent gospel harmony witnesses [Ephrem's Commentary, the Venetian Harmony, the Old High German Harmony, the Arabic Harmony, etc.]) exist between Shem-Tob and the Liège Harmony (this is a conservative figure: the actual number might well be above 1,500, or even 2,000).

102. As for the all-important unique agreements (readings restricted to Shem-Tob and the Liège Harmony [or, in a few cases, also with harmonies which are dependent upon the Middle Dutch tradition]), these probably number (conservatively) somewhere between 150 and 300. It is these unique agreements which allow one to stipulate the precise position of the Vorlage of Shem-Tob's Hebrew Matthew within the Western medieval harmonized gospel tradition and also to give an approximate date for this Vorlage. First, the high number of these unique agreements and the character of many of them (in the forthcoming NTS article, for example, one of the unique agreements is, in the Liège Harmony, an interlinear gloss, while in Shem-Tob's Hebrew Matthew it has been substituted for the standard reading; another unique agreement presented in the NTS article is a reading well-known in medieval commentaries--first appearing in the Englishman Bede [fl. 720]--which, in the Liège Harmony, is conflated with the standard Matthean reading, while in Shem-Tob this reading from Bede is once again substituted for the standard text [only these two documents are known to place the remark from the commentary into the text of Matthew]; etc.) indicate that there is an intimate textual link between the Vorlage of Shem-Tob's Hebrew Matthew and the (Latin) Vorlage of the Liège Harmony.

103. If anyone wishes to venture an alternative Vorlage for Shem-Tob's Hebrew Matthew, then that explanation must account for the known empirical evidence, namely, the more than 60 non-unique and the fourteen unique parallels which your author has published over the last decade.

4.0 The Genesis of Shem-Tob's Hebrew Matthew

104. There is no mystery about the genesis of Shem-Tob's Hebrew Matthew. It is obvious that it incorporates material from a variety of sources (e.g., from the Toledoth Jeshu, the Vetus Latina, etc.). But because of the high number of agreements with the Liège Harmony, many of them unique, the tradition behind the Liège Harmony--which we know to be a Latin gospel harmony--must also be the principal element responsible for the textual complexion of Shem-Tob's Hebrew Matthew: it explains the Vetus Syra readings, it explains the Vetus Latina readings, it explains the Thomas readings, it explains the harmonizations, it explains the Johannine fragments, it explains the many parallels with the rest of the Western medieval harmonized gospel tradition (with, e.g., the Venetian Harmony, Codex Cassellanus, etc.).

105. Turning to the matter of the date of this common Vorlage, our textual researches, especially the unique parallels and the character of some of them (viz. obviously medieval glosses), set the boundaries.

106. All the Western vernacular harmonies derive from a Latin base. As one might expect in such a case, the further back one goes (the earlier the date), the more wide-spread a reading is within the tradition. However, the later one goes in the tradition (the more recent the date), the more singular and localized the variants become.

107. Since Middle Dutch literature begins only at about 1200 (at the earliest), we can be certain that the archetype of the Middle Dutch tradition was not translated from its Latin Vorlage before 1200. And it cannot be later than 1250 or so, for the Liège Harmony (copied about 1280) is at least a first-generation copy of that Middle Dutch archetype (recall the common error which the Liège Harmony shares with van Maerlant's Rijmbijbel). We can be certain, then, that the Latin Vorlage from which the Middle Dutch tradition derives was in circulation between about 1200 and 1250 in Belgium (the provenance is dictated by the Zuid Limburgs dialect in which the Liège Harmony is written).

108. How old might this particular Latin Vorlage, which circulated in Belgium between 1200 and 1250, have been? Could it, for example, date from the fifth century? No, it could not, for if it were that early, then many of the other vernacular harmonies (the Middle Italian, the Old High German, etc.) would also share these "unique" readings with Shem-Tob's Hebrew Matthew and the Liège Harmony. But they do not. Rather, the quality (unique to Shem-Tob and the Liège Harmony) and character (some being obviously medieval glosses, one even found in an interlinear position in the Liège Harmony) of these unique readings mean that this Latin Vorlage which circulated between 1200 and 1250 must have been fairly far out on (near the tip of) a specific branch of the Latin harmony tradition.

109. If one arbitrarily stipulates the date of the translation of the Middle Dutch archetype from this Latin Vorlage as 1200, and if one arbitrarily attributes to this same Latin Vorlage a century of life before it spawned the Middle Dutch tradition, then one arrives at a date of about 1100 as the date of copying of this Latin manuscript, which carried the distinctive readings it later bequeathed to the Middle Dutch tradition. If one takes that date (1100) as the mid-point, then one might presume that this specific Latin sub-family (i.e., distinguished by these unique glosses and variants) of the larger Latin harmonized gospel tradition evolved in and saw circulation in the period between approximately 900 and 1300 (i.e., our mid-point, 1100, +/- 200 years).

110. Earlier than 900 is unlikely, because of the uniqueness of many of the readings (which are restricted to Shem-Tob's Hebrew Matthew and the Liège Harmony), and their high-medieval character (the glosses). (Put differently: if this Latin Vorlage were much earlier, then these distinctive readings would be much more widespread within the Western harmony tradition.) Later than 1300 is unlikely, for we know that this Latin tradition must have been in existence by the date of the copying of the Liège Harmony (ca. 1280).

4.1 Additional Evidence from History and from the Text

111. While our textual evidence has indicated a source and given us an approximate date for the genesis of Shem-Tob's Hebrew Matthew, one might also wonder: Are there any historical parallels which are relevant to this discussion? For example, a critic might ask, with justification: Are there any precedents for the text of a separate gospel (which Shem-Tob's Hebrew Matthew clearly is [although recall that harmonizations are one of Shem-Tob's distinguishing characteristics]) being influenced by a gospel harmony? A critic might further point out that while this Latin Vorlage appears to have circulated in Belgium (the provenance of the Liège Harmony), Shem-Tob was writing in Spain; is there, then, any evidence to show that this Latin tradition also circulated in Spain during the period under discussion, say from about 900 to about 1300? Although Howard never mentions it, the answers are, once again, already in the literature, both medieval and modern.

112. In 1888, Ignazio Guidi (Guidi 1888: 28-29) remarked on the existence of an Arabic translation of the separate gospels made in Spain, in 946 C.E., by a Christian named Isaac Velasquez (Ishaq ibn Balisak), of Cordoba. Velasquez's work is well-enough known that Metzger mentions it twice in his Early Versions (Metzger 1977: 21, 260). Although the text of Velasquez's translation has never been published, Anton Baumstark analyzed its text of Mark 2 in a study titled "Markus Kap. 2 in der arabischen Übersetzung des Isaak Velasquez Veröffentlicht und unter dem Gesichtspunkt des Zusammenhangs mit dem Diatessaron gewürdigt," published in the journal he founded, Oriens Christianus (Baumstark 1934).

113. Baumstark discovered numerous agreements between this Arabic translation of the separate gospels by Velasquez and the Middle Dutch harmonies, especially the Liège Harmony (see the remarks and readings in Baumstark's apparatus, ibid.: 232-239). Since these agreements cannot stem from Arabic-to-Dutch or Dutch-to-Arabic influence, we are compelled to conclude that the common denominator was, from Velasquez's side, the Latin exemplar of the separate gospels from which he worked, a Latin exemplar which had been profoundly influenced by a Latin gospel harmony akin to the Latin Vorlage of the Liège Harmony. Because we find not just identical readings, but also evidence of identical harmonization in the separate gospels both in Velasquez/Liège and in Shem-Tob/Liège, it is clear that this Latin harmonized gospel tradition influenced not just vernacular harmonies (such as the Liège Harmony), but also a Latin edition of the separate gospels. It is apparently from this harmony-influenced separate gospel text that both Velasquez's Arabic translation and Shem-Tob's Hebrew translation of Matthew derive.

114. The striking evidence from Isaac Velasquez--so crucial for Howard's self-proclaimed goal of placing Shem-Tob's Hebrew Matthew "in a proper historical . . . context" (Howard 1995: vii)--closes the circle, for it provides a historical parallel for Shem-Tob's Hebrew Matthew. Velasquez offers incontrovertible historical evidence that (1) Latin-to-Semitic language translations of the gospels were being executed (2) in Spain (3) during the medieval period; furthermore, although these translations were of the separate gospels, their text was nevertheless (4) influenced by the same harmonized gospel tradition responsible for the Middle Dutch Liège Harmony. It is within this matrix--not (as Howard proposes) in Christian antiquity--that the genesis of Shem-Tob's Hebrew Matthew lies.

115. Textual evidence confirming the link between Velasquez, Shem-Tob's Hebrew Matthew, and the Liège Harmony became available for the first time in November 1997. While no edition of Velasquez's translation has ever been published, Baumstark collated much (precisely how much, and how thoroughly, is unknown) of one of the extant Velasquez manuscripts, Munich Aumer 238 (dated 1394 C.E., thus, a direct contemporary of Shem-Tob) on small 10 x 7 cm. slips of paper, one variant per slip. Through unusual circumstances, which will be published in another forum in due course, Baumstark's collations came into the possession of two of your author's fellow team-members (Dr. August den Hollander and Dr. Ulrich Schmid) here at the Netherlands Institute for Advanced Studies (NIAS), in late November 1997. The archive (which also contains collations of many other Western gospel harmonies and related witnesses; the archive was, in part, the Fundgrube for Baumstark's posthumously published Die Vorlage des althochdeutschen Tatian [Baumstark 1964]) will eventually be placed with a major Dutch university library. With the assistance our team's Arabist, Dr. Peter Joosse, of the University of Frankfurt, your author checked fourteen verses (the same fourteen verses in which your author found the singular agreements between the Liège Harmony and Shem-Tob) against Baumstark's collations of Velasquez. Several new agreements among Shem-Tob, Velasquez, and the Liège Harmony were found (e.g.: read "it came to pass" for "and" at Matt 25:6; interpolate "Again Jesus told his disciples another parable" between Matt 25:13/14), but these mean little, for the same variants are also found in numerous other sources (in both cases, in Greek manuscripts). Two readings of significance did, however, emerge. Both are unique (they have been screened against our same list of sources listed supra, sec. 3.0). The first reading is found, to the best of your author's knowledge, only in Shem-Tob's Hebrew Matthew, Isaac Velasquez, and the Liège Harmony. The second is found only in Velasquez and Shem-Tob's Hebrew Matthew. Here are the two readings.

116. (A) A reading in the NTS article (which will now acquire a postscript), which was singular to the Liège Harmony and Shem-Tob's Hebrew Matthew (at Matt 11:25: for "babes" read "the humble") is also found in Velasquez, according to Baumstark's collation. This reading, as part of the text of Matthew, is now know from only three sources: the Middle Dutch Liège Harmony, Shem-Tob's Hebrew Matthew, and Isaac Velasquez's Arabic translation of the gospels. We know that two of these documents have a Latin Vorlage (Liège and Velasquez); we know that two of these documents individually share common readings with the Liège Harmony (Shem-Tob = Liège; Velasquez = Liège); we know that two of these documents have a Spanish provenance (Shem-Tob and Velasquez); we know that all of them were copied between 946 (the date of Velasquez) and 1380 (the date of Shem-Tob's Even Bohan). The common denominator would seem to be a Latin manuscript of the separate gospels which circulated in Spain in the high-medieval period and which had been influenced by the same Latin harmony which debouched in the Liège Harmony.

117. (B) At Matt 9:24, the standard Greek reads: "He said, 'Depart, for the girl is not dead but sleeping.'" Shem-Tob, however, reads: "He said to them, 'Go outside all of you and stop weeping because the girl is asleep and not dead,'" reversing the order of "asleep" and "dead." Baumstark's collation states that Velasquez's text reads, like Shem-Tob, "asleep and not dead." Other than these two medieval Iberian sources, your author knows of no other texts which contain this variant.

118. This, then, is additional textual evidence which confirms our earlier findings. And it comes from a manuscript (dated 1394) of Isaac Velasquez which is directly contemporary with Shem-Tob. The Hebrew Matthew in Shem-Tob clearly incorporates older traditions, but these are merely derived from its Latin Vorlage, a Vorlage which was closely related to Latin Vorlage of the Middle Dutch Liège Harmony. The same for the Liège Harmony: while it contains ancient readings (e.g., the agreements with Thomas, with other witnesses to the second-century Diatessaron of Tatian), these are not the result of it having direct contact with Thomas or dating from the second century; rather, through a long chain of dependence, going back first through Dutch ancestors, and then Latin ancestors, and then, perhaps, Syriac ancestors, we sometimes reach an antique text related to Thomas, or to the Diatessaron. In no way is Shem-Tob's Hebrew Matthew a relic from early Christianity, or even directly related to texts from early Christianity. Rather, many--perhaps even most--of the singular readings in Shem-Tob are distinguished by their presence in other medieval texts related to the harmonized gospel tradition, especially in those texts related to the Vorlage of the Middle Dutch harmonized gospel tradition.

119. (Readers may also wish to refer to the excursus by William Horbury, "The Hebrew Text of Matthew in Shem Tob ibn Shaprut's Eben Bohan" [Horbury 1997]. Although using different evidence and different methods, Horbury's conclusions are very similar to the findings presented here [see esp. p. 731 for Latin dependence and p. 738 for a medieval genesis of the text of the Hebrew Matthew].)

5.0 What Can Be Learned

120. There are very important lessons to be drawn from this investigation. Each of the missteps in Howard's study comes from a failure to respect some of the most elementary rules of scholarship. These rules have not been concocted on a whim; they are, rather, the product of millennia of scholarly experience. Careful scholars hew to these rules, for they prevent them from indulging in precisely the errors we encounter here. What are these rules?

121. First, when studying a text--in this case, a medieval gospel text, suffused with harmonizations, copied in Spain--research must commence with the documents which are in closest chronological, geographic, and generic proximity to the text under examination. In this case, had this rule been followed, then one would have begun searching for parallels in medieval gospel texts in the West, which evinced harmonizations. Here, this has not been done. Howard completely ignores the documents closest in time, provenance, and genre to Shem-Tob's Hebrew Matthew; instead, he commences his search for parallels by trolling in waters at the opposite end of the Mediterranean, and a millennium away.

122. Second, before making sweeping claims ("against all other Matthean witnesses"; "must go back to the early centuries of the Christian era"; etc.), a scholar is obligated to conduct the requisite investigations which such statements presuppose. If one does not conduct the necessary investigations, then one must not make such statements. Here, no heed has been taken of this basic tenet of scholarship. While Howard's study abounds with broad claims and bold assertions of what "must" be, examination shows these claims to be demonstrably false.

123. Third, before theorizing, one must scrupulously collect all of the pertinent evidence. Only then, with all of the evidence one can muster in hand, only then should one begin to theorize. Here, this has not been done. Note that, as soon as the full range of parallels with Shem-Tob's Hebrew Matthew is made known, as soon as the manner by which Thomas readings and other ancient readings entered the Western harmonized vernacular gospel texts is disclosed, then the problem of the Vorlage of Shem-Tob's Hebrew Matthew is placed in an entirely new light. Howard's failure to gather all the pertinent evidence before constructing his theories means that they are incapable of performing even their most basic function, namely, explaining the observed phenomena: Howard's theories cannot account for or explain the presence of these identical, sometimes unique variants in Middle Dutch and other earlier vernacular gospel texts in the West.

124. Fourth, after developing his or her theories, a scholar must subject them to rigorous self-criticism. Where lapses of logic (e.g., treating the identical phenomenon differently; asking the reader to make an unwarranted jump in reasoning) are detected or where obvious gaps in the argument are detected, then either additional evidence and/or explanations must be offered, or the argument must be abandoned. Here, this has not been done. There is a conspicuous lack of logic in Howard's book; there is also a lack of familiarity with how one constructs a scholarly argument. The consequence is that we find textual arguments based on omissions (in other words, on the absence of evidence) and significance claimed for parallels between Shem-Tob's Hebrew Matthew and some early Christian text--even though numerous intermediate texts between these two documents contain the identical variants. We also find identical phenomena treated differently (snippets of John come from a "pre-Johannine" Matthew, but snippets of Mark and Luke apparently do not), and a constant "privileging" of a particular text (apparently the only place where these "pre-Johannine" fragments survive is--mirabile dictu!--in the text Howard is editing), and more--all this chicanery simply to lend a patina of credibility to stillborn theories.

125. Fifth, before commencing work in a particular historical and/or textual area, one must thoroughly familiarize oneself not only with relevant modern scholarship, but also with the relevant antique texts. Here, this has not been done. As we have seen, both modern scholarship (on Thomas; on the Western medieval gospel tradition; on the manner by which readings from early Eastern Christian texts were transmitted to the Western medieval world) and relevant medieval sources (Cod. Vat. 32; the medieval Latin and vernacular gospel harmonies; Isaac Velasquez's translation of the gospels) have been ignored. Familiarity with the transmission-history of the gospels in the period when Shem-Tob was writing would have led one to this literature and these sources.

126. Sixth and finally, one is obliged to respond to founded criticism. Here, this has not done. A full decade ago, in 1988, the Journal of Biblical Literature invited your author to review the first edition of Howard's book. In the review, which was published a year later, in 1989, your author wrote:

Howard's analysis of the genesis of the text is unsatisfactory, however, for significant evidence--much of it textual--contradicts his conclusions. As proof, consider the readings from a single MS, Liège University 437 (known as the Liège Harmony), copied ca. 1280 (prior to Shem-Tob) in a dialect of Dutch, Zuid Limburgs (Petersen 1989: 723).

127. If the eleven textual parallels adduced in the 1989 review were incorrect or irrelevant, then there is an obligation to demonstrate why they were incorrect or irrelevant. If, however, the parallels adduced in the review were correct, then there is an obligation to modify the hypothesis so that it can account for these parallels. Here, however, neither has been done; this empirical, textual evidence has simply been ignored.

128. Failure to adhere to these elementary scholarly principles has had consequences. In the January 1997 number of New Testament Studies, following Howard's model of comparing Shem-Tob's Hebrew Matthew with ancient Christian texts, R. F. Shedinger published nine parallels between Shem-Tob and ⁴⁵(!) (Shedinger 1997). But all of Shedinger's parallels are completely useless, for, like Howard, he does not appreciate that it is impossible to demonstrate dependence when the identical variant is found in intervening texts. And in Shedinger's case, all of the readings he adduces to demonstrate the dependence of Shem-Tob's Hebrew Matthew on ⁴⁵ are paralleled in much later standard Greek and/or Latin manuscripts of the gospels: two of his readings are that of the textus receptus; one reading is found in the Vulgate (!), and--most damning of all--three are found in a single Greek manuscript of medieval date! Furthermore, like Howard, Shedinger appears unaware of the dangers of arguing from omissions: three of the five readings he himself characterizes as "strong" are omissions. And, in a final insult to logic, only two of his nine readings compare ⁴⁵'s Matthew with Shem-Tob's Matthew; the remaining seven readings compare Shem-Tob's Matthew with ⁴⁵'s Mark or Luke--for no better reason than that the papyrus is defective at the Matthean parallel!

129. The most devastating blow, however, comes from a source of which Shedinger is completely ignorant. Apparently Shedinger's research was not thorough enough to discover the 1989 review nor, apparently, was his command of the transmission-history of the gospels competent enough to lead him to the pertinent medieval documents. The result? As the reader might have guessed, many of Shedinger's nine parallels are also found in the Liège Harmony. The reader is invited to determine the number for him- or herself; to assuage your curiosity, however, your author will stipulate that the number is more than three, but less than nine.

130. (Shedinger's evidence is also examined in detail in your author's forthcoming NTS article [Petersen 1998]. Note that, once again, on a list not of your author's creation, about half of the readings have a parallel in a single source: the Middle Dutch Liège Harmony. The test of any hypothesis is how it performs on data not selected by the theory's proponent. Here, the hypothesis of Shem-Tob's Hebrew Matthew's dependence upon the Vorlage of the Liège Harmony is once again vindicated by its performance on a list drawn up by Shedinger, just as it was on the lists drawn up by Howard.)

131. It is here, in the perpetuation of misinformation in the form of factual errors and logical nonsense, that the real evil of such scholarship lies. Placed on the wrong track by Howard's false assertions, misleading lists of parallels, and unsubstantiated theories, hobbled by his own superficial research, his own naïve understanding of the transmission-history of the gospels, and his own lack of familiarity with the sources closest to Shem-Tob (i.e., Western, medieval vernacular gospel texts, evincing harmonizations), Shedinger has repeated all of Howard's blunders, wasted both his time and energy, and possibly misled others in the future.

132. In 1989, your author wrote: "By clearly isolating the earliest complete Hebrew text of Matthew and then producing a critical edition of it, Howard has provided scholarship with a significant new research tool" (Petersen 1989: 723). That statement remains true nearly a decade later. However, the following assertions, also nearly a decade old, also remain true:

The extensive--often unique--textual agreement with the Liège MS (far surpassing the number and degree of agreements adduced by Howard with any other source) suggest that Shem-Tob's Matthew derives from certain medieval traditions. . . . [Howard's] analysis is hobbled by a lack of familiarity with studies into the origin of the Vetus Latina, Vetus Syra and the Diatessaron by such scholars as T. Zahn, A. von Harnack, F. Baethgen, A. Hjelt, H. J. Vogels, and D. Plooij. . . .

[Howard's] attempt to locate Shem-Tob's Matthew in the textual tradition of the gospels is fundamentally flawed, not only because it ignores past research and fails to engage in rigorous self-criticism, but also because it overlooks extensive textual evidence which indicates Shem-Tob's Matthew is dependent upon sources such as the [Vorlage of the] Liège MS, the Vetus Syra and the Vetus Latina, not anterior to them (Petersen 1989: 724-725).

133. It must be repeated: Howard's edition is to be warmly welcomed; his serviceable translation is very helpful, but it cannot be used for text-critical purposes without a knowledge of Hebrew; his evidence and theories concerning the origin and transmission of this Hebrew Matthew must be dismissed completely. You author trusts that now, after having--for a second time--signaled the pertinent literature (both medieval and modern), provided numerous additional unique and non-unique parallels between the Shem-Tob's Hebrew Matthew and the Middle Dutch Liège Harmony, and pointed out the factual, logical, and philological errors which have caused previous attempts to stipulate the genesis of this Hebrew Matthew to collapse, he will not find it necessary to do so again in the future, for a third time.

134. In the first edition of this book, Howard devoted two pages to setting out the parallels in other Hebrew Matthews for a very interesting variant at Matt 12:32 (Howard 1987: 173-174). In Shem-Tob's Hebrew Matthew, the text gives the standard reading "Everyone who says a word against the Son of Man, it shall be forgiven him." However, a comment appended directly after v. 37 in Shem-Tob's Hebrew Matthew reads: "Behold the Father and the Son will forgive him but the Spirit will not forgive him." (The text of this comment is, oddly enough, given only in Howard's first edition [ibid.: 173]; the second edition omits it, even though the reading is #9 on Howard's list of parallels between Shem-Tob's Hebrew Matthew and Thomas [Howard 1995: 204].) Howard compares the reading of this comment in Shem-Tob's Hebrew Matthew with the Gospel of Thomas, logion 44: "He who blasphemes against the Father will be forgiven, and he who blasphemes against the Son will be forgiven, but . . . ."

135. Howard's exploration of this variant (". . . against the Father") in earlier Hebrew Matthews is interesting, but it ignores one important parallel: this same reading is found in both branches (or families) of the Middle Italian harmonized gospel tradition, whose extant manuscripts date from the thirteenth to fifteenth centuries. The text of the Middle Italian Tuscan Harmony reads: "Chiunque dirà parola contra 'l Padre, gli sarà perdonato . . ." (Vattasso and Vaccari 1938: 244, lines 14-15).

136. This Middle Italian harmony tradition is translated from a Latin Vorlage and belongs to the same broad Western medieval gospel harmony tradition as does the Middle Dutch Liège Harmony. Hence, the Middle Italian harmonies and the Middle Dutch harmonies have many readings in common. Nevertheless, their two Vorlagen are not identical: in this case, the Middle Dutch harmonies lack this reading. But notice: Shem-Tob's Hebrew Matthew is once again sharing a unique reading with a Western medieval harmonized gospel text (this time, the Middle Italian Tuscan Harmony) and Thomas. This example once again demonstrates the dependence of Shem-Tob's Hebrew Matthew upon the Western medieval Latin harmonized gospel tradition. (For a description of the Middle Italian harmonies, see Metzger 1977: 24-25 or Petersen 1994: 247-251; cf. the stemma of the entire Diatessaronic tradition, ibid.: 490.)

137. But the real significance of this variant is that it also occurs in one other source, another Middle Italian harmony, the Venetian Harmony. Like the Tuscan Harmony, the Venetian Harmony also has a Latin Vorlage and is related to the Liège Harmony because of their mutual dependence, ultimately, upon the same Latin Western medieval harmonized gospel tradition. What arrests one's attention, however, is that while this Venetian Harmony (just like Shem-Tob's Hebrew Matthew) reads the normal text in Matt 12:37 (". . . against the Son of Man"), it also contains a comment inserted post Matt 12:37 (just like Shem-Tob's Hebrew Matthew), and in this comment we read: ". . . peccato contra lo Pare" (Todesco 1938: 59, line 29). Although the text of the two comments is not identical (at least the two lines which Howard presents in his first edition), this Venetian Harmony provides a remarkable parallel for the structure of Shem-Tob's Hebrew Matthew, which places the standard (= Greek/Vulgate) reading in the body of the text, and the deviating (= Thomas) reading in a comment inserted after v. 37: not only is the variant the same, they are presented in the same context, namely, in a comment following v. 37. (For a detailed study of this logion, see Baarda 1997.)

138. We know that this variant is a Thomas reading, and we know that Thomas readings entered the West via the Latin gospel harmony tradition, which dates back to at least 546, the date of Codex Fuldensis. Therefore, we presume that this reading, well-represented in the Hebrew Matthew tradition and in Western harmonies dependent upon the Latin harmonized gospel tradition, was made available in the West via this same Latin harmonized gospel tradition, as were the other Thomas readings discussed above in secs. 2.2.1.1 and 2.2.1.2.

139. Credit must also be given where it is due: Quispel, of course, logged this parallel between Thomas and the Middle Italian harmonies more than twenty years ago, in his Tatian and the Gospel of Thomas (Quispel 1975b: 181); however, he does not report that the parallel in the Venetian Harmony is found only in a comment.

140. Finally, two concluding observations may be offered. First, it must be stated that, despite this interesting parallel, the number of parallels between Shem-Tob's Hebrew Matthew and the Liège Harmony far surpasses the number of parallels between this Hebrew Matthew and the Middle Italian harmonies. While this reading once again points to the Western medieval harmonized gospel tradition as occupying a prominent role in the genesis of the text of Shem-Tob's Hebrew Matthew, in no way does the degree of textual intimacy which exists between the Middle Italian harmonies and Shem-Tob's Hebrew Matthew even remotely approach the much greater degree of textual intimacy which exists between the Middle Dutch Liège Harmony and Shem-Tob's Hebrew Matthew.

141. Second, as noted above, although this reading from Matt 12:32 is present in the Middle Italian harmonies, it is absent from the Middle Dutch harmonies. No single vernacular family preserves all the readings of its Latin Vorlage; indeed, none of the extant Latin manuscripts preserve all of them (for example, no extant Latin manuscript offers this variant at Matt 12:32). However, it is recognized that all of these various vernacular harmony families stem from the Latin harmonized gospel tradition, and it is this Latin tradition which is, therefore, presumed to be the carrier of, in this instance, this Thomas reading, for we know it carried other Thomas readings--some of which are found in the extant Latin harmony manuscripts (e.g., "in a hidden place" in Thomas logion 33 is found in three extant Latin harmony manuscripts; see supra, sec. 2.2.1.2). Therefore, it does not follow that the absence of a reading from one or another of the individual harmony manuscripts (in this instance, the absence of the reading "the Father" at Matt 12:32 from the Liège Harmony) means that the reading did not stand in the Latin Vorlage, for--as this example demonstrates--it is often found in (an)other harmony manuscript(s): here, in the Middle Italian harmonies. The importance of this is the following. If, in the case of Howard's list of twenty-two parallels with Thomas, a critic were to object that since only half of the twenty-two readings were paralleled in the Liège Harmony, the remainder of the readings--the other eleven--might well have arisen from contact between Shem-Tob's Hebrew Matthew and Thomas in Christian antiquity (as Howard has argued), then it shows that the critic does not yet understand the issues. The response to such a critic is simple: all of Howard's twenty-two parallels are found in one or another branch of the harmonized gospel tradition (e.g., here in Matt 12:32, we have just found a variant which occurs only in the Middle Italian tradition). The reason that our study has focused on the Liège Harmony is that, of Howard's list of twenty-two agreements, the Liège Harmony has more agreements with Shem-Tob's Hebrew Matthew than any other single harmonized gospel manuscript. And this same pattern of the Liège Harmony having more agreements with Shem-Tob's Hebrew Matthew than any other single harmonized gospel manuscript extends through Howard's other lists examined above, as well.

This article was written while Fellow-in-Residence 1997/98 at the Netherlands Institute for Advanced Studies (NIAS), Wassenaar, the Netherlands, the research arm of the Royal Netherlands Academy of Arts and Sciences. The Institute's generous hospitality and excellent support are gratefully acknowledged, as is the delightful colleagueship of the other members of "Team Diatessaron."

© TC: A Journal of Biblical Textual Criticism, 1998.

Bibliography

Aland, Barbara; Aland, Kurt; Karavidopoulos, Johannes; Martini, Carlo M.; and Metzger, Bruce, eds. 1993a. The Greek New Testament. 4th ed. Stuttgart: Deutsche Bibelgesellschaft.

Aland, Barbara; Aland, Kurt; Karavidopoulos, Johannes; Martini, Carlo M.; and Metzger, Bruce, eds. 1993b. Novum Testamentum Graece. 27th ed. Stuttgart: Deutsche Bibelgesellschaft.

Baarda, Tjitze 1983. Early Transmission of the Words of Jesus: Thomas, Tatian and the Text of the New Testament. Edited by J. Helderman and S. J. Noorda. Amsterdam: Boekhandel Uitgeverij.

Baarda, Tjitze 1997. "'Vader � Zoon � Heilige Geest.' Logion 44 van 'Thomas.'" Nederlands Theologisch Tijdschrift 51: 13-30.

Baumstark, Anton 1934. "Markus Kap. 2 in der arabischen Übersetzung des Isaak Velasquez Veröffentlicht und unter dem Gesichtspunkt des Zusammenhangs mit dem Diatessaron gewürdigt." Oriens Christianus 31 [= III.9]: 226-239.

Baumstark, Anton 1964. Die Vorlage des althochdeutschen Tatian. Edited by Johannes Rathofer. Niederdeutsche Studien, no. 12. Cologne: Bohlau.

Bergsma, J., ed., 1895-1898. De Levens van Jezus in het Middelnederlandsch. Leiden: Sijthoff.

Burkitt, Francis C. 1899. "Introduction" to The Biblical Text of Clement of Alexandria in the Four Gospels and the Acts of the Apostles, by P. M. Barnard. Texts and Studies, vol. 5, no. 5. Cambridge: Cambridge University Press.

Burkitt, Francis C., ed., 1904. Evangelion da-Mepharreshe. 2 vols. Cambridge: Cambridge University Press.

Burkitt, Francis C. 1924. "Tatian's Diatessaron and the Dutch Harmonies." Journal of Theological Studies 25: 113-130.

Davies, W. C., and Allison, D. C. 1997. A Critical and Exegetical Commentary on the Gospel according to Saint Matthew., Vol. 3 [Matthew XIX-XXVIII]. International Critical Commentary. Edinburgh: T. & T. Clark.

Guidi, Ignazio 1888. Le traduzioni degli evangelii in Arabo e in Etiopico. Memorie, Accademia nazionale dei Lincei, Classe di scienze morali, storiche e filologiche, Ser. 4, v. 4, pte. 1a. Rome: R. Accademia dei Lincei.

Horbury, William 1997. "The Hebrew Text of Matthew in Shem Tob ibn Shaprut's Eben Bohan." In Davies and Allison 1997, 729-738.

Howard, George 1987. The Gospel of Matthew according to a Primitive Hebrew Text. Macon, GA: Mercer University Press; Louvain: Peeters.

Howard, George 1995. Hebrew Gospel of Matthew. Macon, GA: Mercer University Press. [Revised edition of Howard 1987]

Howard, George 1997. "Re: P45 and SHEM-TOB's Hebrew Matthew." Post to TC-List electronic mailing list, 10 March; 8:38AM. URL: http://www.findmail.com/list/tc-list/1925.html.

Jülicher, Adolf 1924a. "Der echte Tatiantext." Journal of Biblical Literature 48: 132-171.

Jülicher, Adolf 1924b. "Ein neu entdecktes Leben Jesu." Die Christliche Welt 38: 162-169.

Jülicher, Adolf 1938-1970. Itala. 4 vols. Berlin: de Gruyter.

Klijn, A. F. J. 1969. A Survey of Researches into the Western Text of the Gospels and Acts. Vol. 2. Supplements to Novum Testamentum, no. 21. Leiden: Brill.

Legg, S. C. E., ed. 1940. Nouum Testamentum Graece. Oxford: Clarendon.

Lewis, C. T. and Short, C. 1879. A Latin Dictionary. Oxford: Clarendon; reprint ed., 1969.

Maerlant, Jakob van 1858-1859. Rijmbijbel. Edited by J. David. 3 vols. Brussels: M. Hayez.

Meijer, G. J., ed. 1835. Het Leven van Jezus, een Nederlandsch handschrift uit de dertiende eeuw. Groningen: Oomkens.

Ménard, J.-É. 1975. L'Évanglie selon Thomas. Nag Hammadi Studies, no. 5. Leiden: Brill.

Merk, Augustinus, ed. 1964. Novum Testamentum Graece et Latine. 9th ed. Rome: Biblical Institute Press.

Metzger, Bruce M. 1977. The Early Versions of the New Testament. Oxford: Oxford University Press.

Münster, Sebastian 1537. Evangelivm secvndvm Matthaevm in lingva Hebraica, cum uersione latina .... Basel: Henrichus Petrus.

Petersen, William L. 1985. The Diatessaron and Ephrem Syrus as Sources of Romanos the Melodist. CSCO 475 (Subsidia 74). Louvain: Peeters.

Petersen, William L. 1989. Review of G. Howard, The Gospel of Matthew according to a Primitive Hebrew Text, in Journal of Biblical Literature 108: 722-726.

Petersen, William L. 1994. Tatian's Diatessaron: Its Creation, Dissemination, Significance, and History in Scholarship. Supplements to Vigiliae Christianae. no. 25. Leiden: Brill.

Petersen, William L. 1998. "The Vorlage of Shem-Tob's 'Hebrew Matthew.'" New Testament Studies forthcoming.

Plooij, Daniël 1923. A Primitive Text of the Diatessaron. Leiden: Sijthoff.

Plooij, Daniël 1925. A Further Study of the Liège Diatessaron. Leiden: Brill.

Plooij, Daniël 1929-1970. The Liège Diatessaron. Verhandelingen der koninklijke Nederlandse Akademie van Wetenschappen, Afdeling Letterkunde, vol. 31. Amsterdam: Koninklijke Akademie Wetenschappen te Amsterdam.

Quispel, Gilles 1958. "Het Luikse 'Leven van Jezus' en het jodenchristlijke 'Evangelie der Hebreeën.'" De Nieuwe Taalgids 15: 241-249.

Quispel, Gilles 1959. "L'Évangile selon Thomas et le Diatessaron." Vigiliae Christianae 13: 87-117.

Quispel, Gilles 1969. "The Latin Tatian or the Gospel of Thomas in Limburg." Journal of Biblical Literature 88: 321-330. [Also in Quispel 1975a: 159-168]

Quispel, Gilles 1975a. Gnostic Studies. Vol. 2. Uitgaven van het Nederlands Historisch-Archeologisch Instituut te Istanbul, no. 24.2. Istanbul: Nederlands Historisch-Archaeologisch Instituut in het Nabije Oosten.

Quispel, Gilles 1975b. Tatian and the Gospel of Thomas. Leiden: Brill.

Shedinger, Robert F. 1997. "The Textual Relationship between P45 and Shem-Tob's Hebrew Matthew." New Testament Studies 43: 58-71.

von Soden, Hermann Freiherr 1913. Die Schriften des Neuen Testaments. 2 vols. Göttingen: Vandenhoeck & Ruprecht.

du Tillet, Jean 1555. Evangelium Hebraicum Matthaei .... Paris: Martinus Juvenis.

Tischendorf, Constantin von 1869-1890. Novum Testamentum Graece. 8th ed. 3 vols. Leipzig: J. C. Hinrichs.

Todesco, V. 1938. Il Diatessaron in Volgare Italiano. Pt. 1: Diatessaron Veneto. Studi e Testi, no. 81. Vatican City: Biblioteca apostolica vaticana.

Tuckett, C. 1988. "Thomas and the Synoptics." Novum Testamentum 30: 132-157.

Vattasso, M. and Vaccari, A. 1938. Il Diatessaron in Volgare Italiano. Pt. 2: Diatessaron Toscano. Studi e Testi 81. Vatican City: Biblioteca apostolica vaticana.

Vogels, Heinrich J. 1919. Beiträge zur Geschichte des Diatessaron im Abendland. Neutestamentliche Abhandlung 8.1 Münster: Aschendorff.

Wilson, R. McL. 1982. "Nag Hammadi and the New Testament." New Testament Studies 28: 289-302.

Wordsworth, I., and White, H. I. eds. 1889-1898. Quattuor Evangelia: Secundem Editionem Sancti Hieronymi. Nouum Testmaentum Domini Nostri Iesu Christi Latine, Pt. I. Oxford: Clarendon.

Zahn, Theodor 1881. Tatian's Diatessaron. Forschungen zur Geschichte des neutestamentlichen Kanons und der altkirchlichen Literatur, no. 1. Erlangen: Deichert.

This version of the review article has minor corrections from the original (ed.).