- 1 Sáenz-Badillos and Yahalom published a special article on the Maḥberet textual problems (Yahalom & (...)
1This article aims at examining the textuality of Maḥberet Menaḥem ben-Saruk, the first biblical Hebrew-Hebrew dictionary ever written, composed in Cordoba, Andalusia, c. 950 CE. A test case of a particular text will be examined through a philological analysis based on the oldest manuscripts preserved at the British Library (and compared with relevant Italian Genizah fragments), in order to follow in the footsteps of the copyists’ interventions. The Maḥberet has reached us in several Genizah fragments and about twenty manuscripts, the earliest being London, The British Library, Add. 27214 (aka London 950) dated 1091 (hence L1) and London, The British Library, Arundel Or 51 (aka London 951) dated 1189 (hence L2). All the others, MSS Oxford, Hamburg, Paris, Krakow, Parma, Bern, Bologna, Florence, Berlin, and Leiden, to name just a few, were copied during the 13th century or later. To date, two editions have been published, one popular by Filipowski in 1854 (Menaḥēm Ben Saruq 1854), and the other critical by Ángel Sáenz-Badillos in 1986 (Menaḥēm Ben Saruq 1986). Both are eclectic, but the editors’ choice between variants was not always made according to objective terms, so that in many places their decisions may be arguable.1
- 2 A commentary to the Maḥberet may be regarded as a complete innovation. Filipowski’s comments are qu (...)
2After cycles of teaching and studying Menaḥem’s grammatical thought, and especially after Hananel Mirsky’s comprehensive study in his PhD dissertation (Mirsky 2018), it appears that the text of the Maḥberet is far from satisfactory, not only in the printed editions but in the manuscripts as well. The time has come for a new edition, with a thorough textual apparatus and a detailed commentary.2 As we realized that the chances of reproducing the author’s original version are very limited, we embarked on the preparation of a diplomatic edition (in press) based on the oldest manuscripts.
3In objective terms, the Maḥberet lost its validity one generation after its publication in 950 CE. This is because the linguistic theory on which it is based became obsolete when Yehuda ben David Ḥayyūj (Fes, Morocco; Cordova, Spain) developed his revolutionary Hebrew root theory towards the turn of the first millennium. Ḥayyūj proved that even the so-called mono- or biliteral verbal roots were in fact triliteral (known since then as “irregular”, because their root included either aleph, waw, yod or final he, e.g., roots q-r-’, p-n-y/h, q-w-m, or as “geminate”, e.g., s-b-b, š-m-m). Nevertheless, the Maḥberet has survived against all odds and has been transmitted continuously until the present day. In an article a few years ago, I tried to account for this phenomenon (Maman 2010). Here I argue that its survival is due to its status as a fundamental text at the origin of Hebrew linguistics and one which gave its development great impetus.
4In this study, I would like to present a test case of one particular textual problem in the Maḥberet. If one follows the evidence for the text of the Maḥberet, one may realize that even the earliest comprehensive manuscript (L1) appeared about 140 years after the Maḥberet was first published in Cordova, enough time for copyists to introduce changes or make unintentional errors. Indeed, even L1 is full of errors, as may be deduced from the type of “absolute errors,” i.e., in biblical quotes. The basis for any entry in a biblical dictionary or lexicon is the linguistic evidence from the Bible. Knowing that medieval Hebrew philologists used very reliable copies of the Bible, it should not be assumed these were the sources of the mistakes discovered in the quotations. Even without paying attention to deviations from the Masoretic text of the type of plene vs. defective spelling, or the addition or omission of conjunctive waw at the beginning of quotations, about 725 errors in quotations can be found in L1. Proportional to the 11,863 biblical verses quoted in the Maḥberet, 725 errors may seem few, as they amount to only 6% of the total, but in absolute terms this is a considerable number. Moreover, one can derive a fortiori that, if a little over 6% of quotations of verses from the Bible are incorrect, many more mistakes may have occurred in the transmission of Menaḥem’s wording (his lexical definitions or discussions) for which there is neither a reading tradition nor a sacred status.
5Interestingly enough, some of the errors in quotations do not appear in later manuscripts, a fact which could conceivably be due to their being based on a preserved text preferable to that used by their earlier counterparts. It appears however that in absolute errors such as biblical quotations late copyists simply corrected their predecessors’ mistakes.
6Another fundamental problem is that whereas biblical quotes may be verified against the text of the Bible itself to check where deviations occurred, there is no similar Urtext to check Menaḥem’s wording. The only way to check whether and where textual errors occurred is based on philological speculations or on Menaḥem’s grammatical theory. In fact, errors were discovered both in the lemmas as well as in the lack of concord between the quoted verses and the entries in which they were embedded. For instance, in the entry דל״ק “burn” the pseudo-verse ודלקו ב(הם) אש שבע שנים “fire will be burning in them seven years,” does not exist in the Bible. The actual biblical verse is וּבִעֲרוּ בָהֶם אֵשׁ שֶׁבַע שָׁנִים (Ez. 39:9) (with the same meaning). Similarly, the entry ח"ף “cover” includes the quotation חפה ראשיהם וחשק אותם, again a non-existing verse, instead of וְצִפָּה רָאשֵׁיהֶם וְחִשַּׁק אֹתָם “he wrapped their capitals and hooped them” (Exod. 38:28). These are absolute errors, either of the author or of later copyists.
7A comparison of the texts of the two earliest manuscripts L1 and L2 is also instructive, as it exhibits many differences. A particularly striking difference concerns the root-entries arrangement. Yahalom and Sáenz-Badillos assumed (Yahalom & Sáenz-Badillos 1985: 257) that the text of the Maḥberet split into two main branches: the Western tradition represented mainly in Ashkenazi manuscripts, which spread rapidly from Spain to Provence, then to France and Ashkenaz; the Eastern-Italian tradition represented in the Genizah fragments and L1. However, Yahalom and Sáenz-Badillos only paid attention to features that crystallized in the final form of the text, whereas our interest here is to get as close as possible to the earliest form of the Maḥberet. In this regard, the Genizah fragments and MS L1 provide better evidence in every aspect, except where disruptions are certain.
- 3 For instance, the copyist of L2 added 23 times the name of Dunash where he criticized Menahem, such (...)
8Against Yahalom and Sáenz-Badillos’ assumption, one may make an argument based on circumstantial evidence. As soon as the Maḥberet came out, it was sharply criticized by Dunash Ben Labrat in the form of a book containing 180 both general and specific criticisms. Menaḥem’s students responded to Dunash in their own book (Benavente 1986), and a student of Dunash responded in turn to Menaḥem’s students (Dunaš Ben Labrat. 1980; Stern 1870). In all that tumultuous public debate, there is no hint that Menahem produced a revised version. There is no doubt however that in later periods, some copies were updated reflecting the controversy literature with additions, corrections, or references to Dunash’s comments.3
9The most noticeable variation between L1 and L2 is in the location of the biliteral root entries, according to Menaḥem’s or any pre-Ḥayyūj grammatical theory. In L1 they appear in their proper place within a general alphabetical order and are intermingled with the triliteral roots, as may be seen in the case of the daleth entries. In L1 the order is as follows (emphasis added to the biliterals):
10דא. דאב. דאג. דב. דבר. דבא. דבק. דבלה. דבש. דג. דגל. דגן. דגר. דד. ד'. דהב. דהר. דהם. דו. דח. דחף. דחון. דחן. דחל. דחק. די. דַּיֵק. דִכֵן. דל. דלף. דלק. דלת. דלג. דלח. דם. דֹמֵן. דמע. דן. דֹנַג. דע. דעך. דף. דפק. דק. דקר. דר. דרור. דרש. דרך. דרדר. דרג. דרכמון. דש. דשא. דשן. דת.
- 4 See Perani 2014:147, 149. Although the specific part of the fragment relevant to our case was not p (...)
11This same order is attested in the oldest fragment of the Maḥberet preserved in the Italian Genizah, Bologna, Archivio di Stato, Hebr. frag. 330.1, originating from Otranto, southern Italy, 11th century.4
Figure 1. Bologna, Archivio di Stato, Hebr. frag. 330.1
Photograph by Mauro Perani, courtesy of him.
- 5 In L2 one may also note that the triliteral roots are rearranged so that the third radical is consi (...)
12On the other hand, L2 (or its source) rearranged the Maḥberet so that the biliteral roots were singled out and are assembled as a group on their own at the head of every letter in the Maḥberet, preceding the group of the triliteral entries, as may be seen below in the example of the same letter:5
13דא. דב, דג, דד, דו, דח, די, דך, דל, דם, דן, דע, דף, דץ, דק, דר, דש, דת, דאב. דאג. דבל, דבא, דבק, דבר, דבש. דגל. דגן. דגר. דהב. דהם. דהר, דחל, דחן. דחף, דחק, דיק, דכן, דלג. דלח, דלק. דלף, דלת. דמן. דמע. דנג. דפק. דקר. דרג, דרדר, דרכמון, דרור. דרש. דשא. דשן.
14In L1, however, the intermingled order is only maintained up to letter צ, whereas the letters ת-ק are ordered according to the L2 method. This section was somehow completed by a copyist who used a source of the L2 type. L1 is also missing at its head until mid-letter ג, but this part was never completed. The striking feature of some late manuscripts, as is the case for MSS Hamburg, Bologna, Paris, and others, is that though influenced by L1 they rather follow the ordering of L2.
15The selected test case for the textual problem to be analyzed below is related to the location of bi- and triliteral roots, and regards the location of the entries חור and חר:
16In L2 the root ח"ר follows ח"ץ and ח"ק and precedes ח"ש and ח"ת, all preceding the triliteral חו"ר, whereas in L1 entry חו"ר follows ח"ו and חו"ח and precedes ח"ז and חז"ק , according to the general alphabetical order. Subsequently, the entry ח"ר is located in its expected place between entries חק"ר and חרצ"ב. The following table presents the location of the entries ח"ר and חור among the entries of the letter ח in both manuscripts:
L2 (biliteral first)
|
L1 (general alphabetical order)
|
חב. חג. חד. חו. חז. חֹ. חח. חט. חי. חך. חל. חם. חן. חס. חף. חץ. חר. חת.
חביון. חבר. חבל. חבש. חבתים. חבצלת. חבט. חבק. חגר. חגוי. חגב. חדר. חדק. חדש. חדל. חור. חוח. חזיון. חזיז. חזק. חטא. חטף. חטב. חטם. חטר. חיב. חיל. חיק. חכליל. חלף. חלק. חלון. חלט. חלץ. חלם. חלמיש. חלאתה. חלך. חלד. חלב. חלש. חמק. חמת. חמד. חמש. חמץ. חמל. חמם. חמן. חמאה. חנם. חנף. חנך. חנט. חנק. חנמל. חסד. חסם. חסן. חספס. חסל. חסף. חסר. חפר. חפש. חפץ. חפן. חפש. חפז. חצב. חצר. חצף. חצן. חרב. חרם. חרמש. חרץ. חרצן. חרט. חרג. חרד. חרך. חרז. חרצב. חקא. חקר. חרם. חרש. חרק. חרול. חרף. חשג. חשל. חשב. חשך. חשר. חשן. חשד. חשף. חשק. חשחן. חשמנים. חשמל. חתף. חתך. חתל. חתם. חתר. חתן
|
חב. חביון. חבר. חבל. חבש. חבתים. חבצלת. חבט. חבק. חג. חגר. חגוי. חגב. חד. חדר. חדק. חדש. חדל. חו. חור. חוח. חז. חזיון. חזיז. חזק. חֹ. חח. חט. חטא. חטף. חטב. חטם. חטר. חי. חיב. חיל. חיק. חך. חכליל. חל. חלף. חלק. חלון. חלט. חלץ. חלם. חלמיש. חלאתה. חלך. חלד. חלב. חלש. חם. חמק. חמת. חמד. חמש. חמץ. חמל. חמם. חמן. חמאה. חן. חנם. חנף. חנך. חנט. חנק. חנמל. חס. חסד. חסם. חסן. חספס. חסל. חסף. חסר. חף. חפר. חפש. חפץ. חפן. חפש. חפז. חץ. חצב. חצר. חצף. חצן. חרב חרם. חרמש. חרץ. חרצן. חרט. חרג. חרד. חרך. חרז. חרצב. חקא. חקר. חר. חרם. חרש. חרק. חרול. חרף. חשג. חשל. חשב. חשך. חשר. חשן. חשד. חשף. חשק. חשחן. חשמנים. חשמל. חת. חתף. חתך. חתל. חתם. חתר. חתן.
|
17When looking into the entries חור and חר in both manuscripts, many questions arise, due to the re/arrangement. Below is the full text of both entries with their various explanations in both L1 and L2:
L2
|
L1
|
A1
|
A
חור. לֹא עַתָּה פָּנָיו יֶחֱוָרוּ: (Is 29:22). חוּר׀ כַּרְפַּס וּתְכֵלֶת (Est. 1:6)... ענין עין הלבן הוא. וכל לבן בלשון ארמית חוור.
|
B1
|
B
|
ח"ר. מתח' לה'.
הא'. חוריה ואין שם מלוכה יקראו. את החורים ואת הסגנים. בין חורי יהוד'. שפטים הם.
הב'. ויקב חור בדלתו. על חור פתן. שלח ידו מן החור. חורי עפר וכפיס. עני' נקב הם.
הג'. לאכול את חוריהם.
וישימה למחראות. עני' טנף צואת אדם. ויתכן להיות מגזרת זה העליון. מחראות מחור כמו צואים מצואת האדם.
הד'. וחרה נחשתה. ועצמי חרו מני חורב. נחר מפוח. והעצמות יחרו לחרחר ריב. ושכן חרירים. בחרחר ובחרב. כמוקד נחרו. ויחר עלי אפו. חרו יושבי תבל. ענין חרבוני קר[ ].
הה'. ולא עתה פניו יחורו. חור כרפס. ואורגי חורי. ענין הלובן הוא והלבן בלשון ארמי' חיור. לבושיה כתלג חיור
|
חר. מת' לחמש מח'.
המ' הא'. חוריה ואין שם מלוכה. ואקבצה את החורים ואת הסגנים. בן חרים. את חורי יהודה שטרים המה.
המ' הש'. ויקב חר בדלתו. ושעשע יונק על חור פתן. שלח ידו מן החור. חורי עפר וכפים. ענין נקב המה.
המ' השל'. לאכל את חוריהם.
וישימיה למחראות. ענין טנף צאת אדם. ויתכן להיות זה מגזרת העליון מחראות מחור כמו צואים מצאת האדם.
המ' הרב'. וחרה נחשתה. ועצמי חרה מני חרב. נחר מפוח. והעצמות יחרו לחרחר ריב. ושכן חרירים. בחרחור ובחרב. כמוקד נחרו. ויחר עלי אפו. חרו יושבי תבל. ענין חרבוני קֵד המה.
(המ' החמ' פניו יחורו. חור כרפס. ואורגים חורי. ענין לובן הוא והלובן בלשון ארמית חור. לבושיה כתלג חור. וכבר נקדמה מלה זאת למעלה ולא השניתיה כי אם עקב המחלקות)
|
C1
|
C
|
חו"ר. לֹא עַתָּה פָּנָיו יֶחֱוָרוּ: (Is 29:22). חוּר׀ כַּרְפַּס וּתְכֵלֶת (Est. 1:6)... ענין עין הלובן הוא. וכל לבן בלשון ארמי’ חור.
וכבר נקדמה מלה זאת במחבר’ ו"ו ולא השניתיה כי אם עקב נענוע הו"ו. א’ע’פ’ שאין הו"ו יסוד.
|
|
18A, B, and C mark separate entries. The comparative table makes it clear that for meanings 1 to 4 in the entry marked B and B1 (המ[חלקה] הא', המ[חלקה] הש[נית], המ[חלקה] השל[ישית], המ[חלקה] הרב[יעית] // הא', הב', הג', הד') both manuscripts are basically the same with only minor differences. When it comes to חו"ר, however, there are major differences as detailed below:
191. Whereas L2 lists 5 classes/meanings, L1 only lists 4, the fifth (in parenthesis, at the bottom of entry B) being added in the margin by a different hand, probably a third copyist, the content of which (המ[חלקה] החמ[ישית] “the fifth class”) is nothing but a repetition of what has already been said in its original location (entry A) or a transposition from L2, (entry C1). Moreover, a closer look at this addition raises questions and concerns. In a verbatim repetition of the entry חו"ר (entry A), an apology is included that states, “it has already been mentioned and I only repeat it here because of the meanings’ partition.” This apology has no direct counterpart in L2.
202. More importantly, whereas according to L1 the root חו"ר is clearly triliteral in the cases of 'יחורו' and 'כתלג חור', etc., the appendage to ח"ר is very strange, as it suddenly turns it into a biliteral. Such a dramatic change in the grammatical analysis would require an explicit statement by Menaḥem on the root issue. The more so the question holds for L2 where the repetition of the fifth meaning of B1 in C1 is no longer a gloss in the margin but part of the normal text of the dictionary.
213. The following example is how a necessity to apologize arises in L2: “this word already appeared in letter waw and I only repeat it because the waw is consonantal, though it is not radical.” Now, if the waw is not radical, why present it among the triliterals? Moreover, one can immediately notice a gross mistake: instead of “in letter waw” one would expect “in the letterח '” or “root ח"ר'”. The reasoning also seems strange. If eventually the root is ח"ר, the apology should have come in the first occurrence of חו"ר by the end of B1, since according to the apology, the waw is not radical.
224. The final sentence of the fifth meaning in L1 is artificially added, since in L2 there is rightly no such sentence.
23We will now eliminate from the previous table what is unnecessary for our purpose and focus on what needs to be compared.
L2
|
L1
|
C1
חו"ר. לא עתה פניו יחוורו. חור כרפס. ואורגי’ חורי. ענין עין הלובן הוא. וכל לבן בלשון ארמי’ חור.
וכבר נקדמה מלה זאת במחבר’ ו"ו ולא השניתיה כי אם עקב נענוע הו"ו. א’ע’פ’ שאין הו"ו יסוד.
|
A
חו"ר. לא עתה פניו יחורו. חור כרפס ותכלת. (ואורגים חורי.) ענין עין הלבן הוא. וכל לבן בלשון ארמית חוור.
|
B1
|
B
|
ח"ר. מתח' לה'.
הה'. ולא עתה פניו יחורו. חור כרפס. ואורגי חורי. ענין הלובן הוא. והלבן בלשון ארמי' חיור. לבושיה כתלג חיור.
|
ח"ר. מת' לחמש מח'.
(המ' החמ' פניו יחורו. חור כרפס. ואורגים חורי. ענין לובן הוא והלובן בלשון ארמית חור. לבושיה כתלג חור. וכבר נקדמה מלה זאת למעלה ולא השניתיה כי אם עקב המחלקות)
|
245. At the end of the marginal note in L1, an apology was added but based on a completely different reasoning than in L2. It reads: וכבר נקדמה מלה זאת למעלה ולא השניתיה כי אם עקב המחלקות “this word (= entry) has already been mentioned above and I only repeat it here because of the meanings’ partition.” But as the number of meanings is not essential, the reasoning does not make sense. At any rate, if the root is triliteral, its addition in the margin is superfluous. If, on the other hand, it is biliteral it should not have appeared earlier under חו"ר. Thus, one should deduce that the “apology” was imitating the pattern of חו"ר in the L2 “corrected” version of the Maḥberet or even transposed from there to L1, entry ח"ר as a fifth meaning.
256. The term השניתיה (lit., I doubled it, repeated it) is unique in the Maḥberet and is not one of Menaḥem’s language markers. It also seems that a repetition of a matter that has already been discussed or presented before is not commonplace in the Maḥberet. An exception is for some entries beginning in א which have counterparts without א with the same meaning, for which Menaḥem clearly expressed that that aleph may be considered either radical or appended. This is the case in the couples אגף/גף, אגר/גר and the like. According to Mirsky (2018: 133-134), a “subtractable aleph” (aleph nigra‘at) in Menaḥem’s view may be viewed both as radical or appended. In the opinion of Ofer (1999: 208) and Dotan (2005: 70), however, Menaḥem was undecided whether the initial aleph was part of the root or not. This is why some of these entries appear both in letter aleph of the Maḥberet and elsewhere.
267. It therefore appears that root ח"ר was initially divided into four meanings only, and was marked ד' מחלקות [=4 classes]. In L2 or its source, a certain copyist assigned חו"ר to ח"ר as a fifth class, and consequently had to “update” the opening expression ד' מח' [=4 classes] to ה' מח' [=5 classes], by only adding a stroke to the daleth, later rendered as חמש “five”.
278. Finally, there is no other case in the Maḥberet where an entire entry is repeated as is in another location, just because the author was hesitant in determining the root. In all cases where there are similar roots distinguished only by a medial waw or yod, they also differ in meaning, and there is no repetition of the same examples or the same definition. This is true for the couples בין/בן, גוע/גע, דד/דוד, הין/הון/הן, חוח/חח, חיל/חל, ליל/ל/לול, עור/ער, עות/עת, צוח/צח, and תה/תוה.
28To sum up, these data allow us to assume that the initial arrangement was of the L1 type. It was rearranged to the L2 type to make it look more “compatible” with Menaḥem’s method, probably because he himself assembled the monoliteral roots in one place where he discussed the root question at the very beginning of letter ב. The case of חו"ר/ח"ר proves that once the biliterals were rearranged and regrouped, duplicates were created, which are incompatible with Menaḥem’s grammatical thought and lexicographical method.