1.0. General Considerations

1In his path-breaking article ‘Ha-Shiṭot ha-rashiyot shel he-ḥaruz ha-ʿivri min ha-piyyuṭ ʿad yamenu’, Benjamin Harshav established the two chief rhyme-norms that are in use in classical piyyuṭ.1 As is well known, the first of these is the norm that demands identity of sound from the consonant before the last vowel in the poetic line onwards. The second, which is termed ‘Qillirian’ in honour of the great poet who invented and introduced it into Hebrew piyyuṭ, is the discontinuous rhyme norm, which demands the participation of two root consonants in the rhymeme (i.e., the formally defined sound unit whose repetition at the ends of the poetic lines constitutes the presence of rhyme) in addition to identity of sound from the consonant preceding the last vowel onwards.

2These two norms (along with any conceivable rhyme norm in general) are rooted in the concrete language material of which the poet avails himself in the composition of his piyyuṭ. In other words, the poet orders the words in the poetic lines in a certain way in order to create the required acoustic impression in the ears of the audience of his listeners. This impression is created by means of the presence of sound parallelism in the expected places in the poetic lines, and if the audience is familiar with the rhyme system being employed in the poem that is being recited, this parallelism is anticipated and perceived as an integral part of the poetics of the poem.

3From the philological perspective, it is accepted that the linguistic material simultaneously consists of several layers—the phonetic, the phonological, the morphological, etc. Not all rhyme norms are equal in their relation to these layers. In the case under discussion at present, the norm of the consonant preceding the last vowel operates on the phonetic layer, since the parallelism that it entails is purely that of sounds. On the other hand, the Qillirian norm activates both the phonetic as well as the morphological layers, since the demand for the participation of two root consonants in the rhymeme is predicated on the existence of a root, which is a morphological unit. This important distinction notwithstanding, the common denominator of both norms is the equivalence of actual linguistic entities—phonetic or morphological. Ab initio, the notion of equivalence is concrete. However, as the consciousness of the rhyme norm spreads in the poetic culture, the notion of equivalence may be altered—it may be gradually liberated from the concreteness of the sounds on the basis of which it first came into being and become abstracted, i.e., formalised, to a certain degree. In other words, both the poet as well as the audience are prepared to process as equivalents linguistic (phonetic) units that are not in fact equivalent, but are nevertheless placed within the poetic line at points that are known to be points of equivalence.

2.0. Rhyme and Philology

4By itself, the fact that the notion of equivalence may become increasingly abstracted has no bearing on our understanding of piyyuṭ as a literary phenomenon. One may simply characterise it as belonging to the category of poetic license, and content oneself with listing poetic equivalents as an aid to the editing of piyyuṭim. For example, we find the following string of rhymes in the Qillirian seder ʿolam יי יננק תישאר וכרד for Shavuot: (lns. 246–54).2 From here, we may conclude that the sounds /t/ and /d/ are treated as being equivalent, if only in a small number of instances, for the purposes of the rhyme norm employed by Qillir. This is not surprising from the phonetic point of view, since tav is a voiceless dental plosive and dalet a voiced dental plosive, so that the degree of phonetic similarity is sufficiently great as to render possible the use of these two sounds as equivalents for the purpose of rhyme.3

5On the one hand, this example is clear-cut, since there is no doubt that despite the equivalence between tav and dalet within the abstract rhyme norm, the two sounds involved were kept distinct on the concrete phonological level of the linguistic material on which Qillir drew in composing his piyyuṭim. On the other hand, in creating sound-equivalents at the ends of poetic lines (which are loci that are relatively protected from spelling errors and scribal corrections, since every locus constitutes a link in the rhyme-chain, whose integrity guarantees the stability of every one of its links), the rhyme norm serves as an important philological tool, which enables researchers to uncover changes in pronunciation in the poet’s time and place. In such cases, however, great caution must be exercised, and it is incumbent upon the researcher to prove that a certain sound equivalence that seems to bear witness to a phonetic phenomenon is indeed a concrete one, rooted in the linguistic material itself rather than merely in the function served by it within the formal rhyme norm, as in the example of tav and dalet above.

6An instructive example in this connection may be found in the case of rhymes in which a syllable-opening ʾalef immediately follows an open syllable with a shewa, as in מ ְּ- א ֹ דו . We occasionally find in Qillirian piyyuṭ that such an ʾalef is elided, along with the shewa that precedes it—i.e., becomes (a double asterisk represents a hypothetical form). Thus for example in the following examples: ( טפשמב ךלמ , lns. 58–62);4 ( ךיבאא םויב ךבמ , lns. 52– 54;)5 שודק עיפוה (  , lns. 33–34).6 In these examples, in fact, the ʾalef is elided before both a vocal and a silent shewa. Here, moreover, as opposed to the case of tav and dalet that I cited above, it appears that we are faced not merely with poetic license, but with an actual phonetic phenomenon that may be traced throughout the development of Hebrew, from Biblical Hebrew to the language-form represented by Qillir. In the Bible, we already find a number of cases of the elision of ʾalef when following shewa—e.g., (Gen. 25.24; = ( ,) םימואת 2 Chron. 12.39 תיראש = ;), etc.7 This phenomenon is known also in the Dead Sea Scrolls: תירש ,) הדואמ =( הדומ (= .) היואר =( היור ,) תיראש8 And in manuscripts of Rabbinic Hebrew9.etc ,) באכ =( ביכ ,) םילהת =( םילית ,) וליאכ =( וליכ  :we find

7From this string of examples, it is clear that the phenomenon as it is found in Qillir’s poetry is the result of a historical phonetic development, and that there is consequently no reason to see it as stemming merely from rhyme pressure and, therefore, as a purely formal phenomenon. It is possible to highlight the phonetic status of the phenomenon under discussion as it is found in the Qillirian corpus by comparing it to an apparently similar phenomenon in the piyyuṭim of Pinḥas ha-Kohen, who postdates Qillir.10 In a new edition of his piyyuṭim Shulamit Elizur points out the rhyme (piyyuṭ יק , lns. 1–4).11 In this case, a root consonant ʾalef is elided between two full vowels—i.e., ראופמ becomes as it were, apparently for purposes of rhyme. On the one hand, in light of the examples that I have cited above from the Qillirian corpus and the dialects of Hebrew that preceded his, it appears that the roots of the phenomenon that is attested in the poetry of Pinḥas—i.e., the elision of ʾalef— are indeed phonetic. On the other hand, I have not found any support from the history of Hebrew for the phenomenon in the form in which it is attested there.12 It is therefore reasonable to conclude that in the present case we see in Pinḥas an emancipation from the concrete realm of phonetics and entry into the formal realm of poetics.

8This impression is strengthened by an additional example cited by Elizur: (poem 23, lns. 28–30).13 She notes that, for purposes of rhyme, there are two pairs here, the element ףאנת being shared by both of them: and In the first pair the ʾalef is reckoned as an actual consonant, the rhyme being ףא -. In the second pair, however, the ʾalef is elided, as the rhyme is ףנ -. It is entirely clear that from a phonetic point of view a bivalent consonant, which simultaneously exists and does not exist, is an impossibility. Therefore, the conclusion presents itself that we are not here dealing with the glottal stop /ʾ/ as it remains or is elided in various linguistic contexts, but rather with a rhyme-unit that may or may not be reckoned, in accordance with the needs of the formal rhyming system. At the same time, it is nevertheless important to underscore the fact that Pinḥas’ treatment of the ʾalef is rooted in phonetic developments.

3.0. The Rhyming of /a/ and /e/ in QIllirian Piyyuṭ

9In the Qillirian qedushta for Shemini Aṣeret וזלוא זל זלל there appears to be attested a rare and unusual phenomenon. In a number of instances in this composition, it seems that the vowels /a/ (pataḥ, qameṣ) and /e/ (segol, ṣere) are paired in the same rhymeme, which is contrary to expectation in the case of a vowel system of the Palestinian/Sephardi type, in which the distinction between these two vowel qualities is maintained. The examples are indeed few, but the phenomenon nevertheless appears to be real: (lns. 9–12); (lns. 112–15); in piyyuṭ 4, whose lines mostly rhyme in or the following rhyme-words are found: (ln. 63), (ln. 66), (ln. 73), (ln. 76).14 Aside from these examples, I am aware of three more cases of rhyming of this type in the Qillirian corpus. In the qerova for 14 benedictions רוכז הכיא for the Ninth of Av we find the following: (lns. 36–39).15 In three places, it might be possible to disagree with the vocalisation given in Goldschmidt’s edition—perhaps one ought to vocalise , ,16 —but I see no way of casting doubt on the vocalisation of the last word, , which is based on the language of Scripture in Lam. 5.6. Here are two examples from another source, the shivʿata ורוז יביא עפצ for Shabbat Zakhor: (lns. 7–10); (lns. 67–70).17 We are, therefore, in possession of additional examples from the Qillirian corpus of the rhyming of /a/ and /e/.

10Another example of such rhyming is cited by Elizur from the piyyuṭim of Pinḥas: (piyyuṭ ולק , lines 61– 64).18 Elizur (2004, 175 n. 69).expresses doubt in this case, indicating that “it is possible that here R. Pinḥas has divided the strophe into two rhyme pairs sharing a similar rhyme.” It seems, however, that it, too, is to be reckoned. Up to this point, I have attempted to demonstrate that even in cases of rhyme based on the formal poetic system, rather than phonetics, it is nevertheless possible to discover the influence of the phonetic level in the background, constituting the basis of the formal system. In the cases under discussion here, on the other hand, it is very difficult to discover the phonetic background of this odd alternation. As is well known, in the Genizah fragments alternations between segol and ṣere as well as between pataḥ and qameṣ are found in abundance.19 But as far as I am aware, there is no evidence of alternation between these two pairs.

11Perhaps it is possible to seek the background to this phenomenon in the phonetics of the Tiberian vocalisation system. In it, we do in fact encounter an alternation between pataḥ and ṣere, which may be defined in terms of morphology. In the suffixed verbal forms of the piʿel stem, as well as in the suffixed, prefixed and imperative forms of the hitpaʿel there are many instances attested in which the final syllable bears pataḥ rather than ṣere— e.g., alternating with with with etc.20 It is important to stress that in these cases the vocalic alternation is located in the ultima of the forms under consideration— i.e., in the syllable that functions as the locus of rhyme within the poetic line. In this context, it is instructive to compare the vocalisations offered by two different editors in a place of potential /a/ - /e/ rhyme. In his edition of the Qillirian shivʿata ינבו םתא ןויצ for Shabbat ha-Ḥodesh Spiegel vocalises / / / (lns. 11–14), thereby apparently acknowledging the existence of the alternation in the Qillirian rhyme system.21 Elizur, on the other hand, who doubts the existence of this alternation in the Qillirian corpus, vocalises in her edition / / (lns. 17–20).22 Examining the two first rhyme words (without taking into account the fourth, with regard to whose reading the editors differ), we must admit, on the one hand, that in Biblical Hebrew there is no piʿel infinitive of the form a fact that seems to vindicate Spiegel’s vocalisation. On the other hand, it is perhaps possible to claim that Qillir changed the quality of the expected vowel on the basis of an analogy to the forms of the prefixed verb and the imperative in the hitpaʿel, thereby justifying Elizur’s approach. In the final analysis, however, both Speigel’s and Elizur’s vocalisations take into account the possibility of the alternation /a/ - /e/, the difference being that, while in Speigel’s version the alternation is realised on the phonetic level—i.e, the vowels /a/ and /e/ participate in the rhymeme and are heard by the ear—in Elizur’s it is incorporated into the morphological level—i.e., she assumes the possibility of the existence of an infinitival form instead of the expected Thus also in the vocalisation given by Goldschmidt in the Qillirian qina (lns. 3– 4).23

12The alternation attested in Biblical Hebrew obviously occurs only in certain verbal forms, and this morphological specificity prevents us from pointing to the Biblical Hebrew phenomenon in order to explain those other occurrences of the /a/ - /e/ alternation in the Qillirian rhyme system that are not based on such verbal forms. Nevertheless, if we accept that the alternation does exist in Qillir’s piyyuṭim, it is perhaps possible to see in the Biblical Hebrew phenomenon at least a part of the phonetic background of the phenomenon that is attested in the formal Qillirian rhyme system. In other words, the instance that I have cited above makes it plausible that Qillir was indeed aware of the possibility of an /a/ - /e/ alternation within his rhyme system, if only on rare occasions, and if only under specific morphological conditions. From such a locus, in which the alternation is, so to speak, legitimate from the point of view of the morphophonology of Biblical Hebrew according to the Tiberian tradition, it spread to other points within the rhyme system, which are found outside of the original morphological context.

13In the final analysis, whether or not the specific considerations offered above provide a full explanation for the alternation /a/ - /e/ within the Qillirian rhyme system is not so important. The methodological point made here is more significant: when setting about to explain an unusual phenomenon in the rhyme system, we must attempt to seek its roots in the phonetics and morphophonology of the speech form (or forms) of which the poet makes use in composing his poetry.

Appendix—The Rhyming of /a/ and /e/ in Hebrew Maqāmas in Thirteenth-century Egypt

14The rhyming of /a/ and /e/ in Hebrew is also encountered in an entirely different linguistic and cultural context. In the course of editing a number of Hebrew maqāmas in the style of the Taḥkemoni of Yehuda al-Ḥarizi that were probably composed thirteenth-century Egypt,24 I have encountered the following cases:

(1) (Introduction, ln. 4)
‘As for this edifice that I am building (cf. 1 Kgs 6.12), / its foundations are suspended from (i.e., built upon) a true outline’.25
(2) (maqāma 7, lns. 30–31)
‘Behold a corpse in a coffin, / found upon the ground and in everyone’s hands. // When it is buried and goes down to hellfire, / it rises to life—breaks the coffin’s bonds and comes forth’.26
(3) (maqāma 10, ln. 2)
‘And he furthermore replied with unflagging strength / and composed in the letter ḥet’.27

15In all of these cases, an /a/-vowel (always qameṣ) rhymes with an /e/-vowel (ṣere or segol). Alongside them we probably ought to consider the following case, encountered in a homonym poem (ṣimmud):

(4) (maqāma 8, lns. 49– 50)
‘Would that my friend gave my due portion / of all [his love] and abandon Time and its vanities. // Were he to seek to take my spirit, my heart’s / response would be “He’s the friend, yield to him”’.28

16Here, the fact that the terminal elements ‘its vanities’ and ‘yield to him’ are supposed to be homonymic implies that segol and pataḥ are being treated as equivalents.

17The cases of /a/ and /e/ rhyming under examination here occur in a composition by an author whose native language we can safely assume to have been Arabic. With this background in mind, we ought to examine a related phenomenon, encountered in bilingual Hebrew-Arabic poetry: the treatment of Hebrew /a/ and /e/ vowels together as being the equivalents of Arabic etymological ā. Garbell has collected numerous examples from Spanish Hebrew poets.29 For the present purposes, it is sufficient to illustrate this point from the trilingual, Hebrew-Arabic-Aramaic poem רבד לא ןמאי by al-Ḥarizi, found in maqāma 20 of the Taḥkemoni.30 The poem is written in tristichs, the first stich of every line being in Hebrew, the second in Arabic and the third in Aramaic. In every line, the Hebrew and Arabic stichs rhyme with one another, whereas the Aramaic stich bears a rhyme that embraces the whole poem. The following Hebrew-Arabic pairs are relevant to our purpose:

(5) (ln. 31)
‘God’s word is sure / as is obedience to the Merciful’.31
(6) (ln. 33)
‘His name is exalted forever, / He is the sublime Lord’.
(7) (ln. 37)
‘He imposes his rule on the world, / which takes in all its quarters’.
(8) (ln. 42)
‘And my Salvation becomes manifest / and He raises the lowly’.

18More examples could be cited from this source, but these suffice to establish the point. In these cases, we may say that, for purposes of rhyme, the opposition between Hebrew /a/ and /e/ is neutralised in the Arabic phoneme /ā/.

19It would appear that both sets of cases are best explained in terms of the vocalic shift known in the Arabic grammatical tradition as ʾimāla, the fronting and raising of /ā/ (as well /a/).32 In the case of the bilingual rhymes, this could simply mean that as a result of ʾimāla Arabic /ā/ had become the closest possible rhyme-equivalent to a Hebrew /e/-type vowel.33 Furthermore, the /a/ - /e/ rhymes in the Hebrew maqāma would seem to indicate that the process of ʾimāla had come to affect the /a/ vowel in the Hebrew pronunciation of Arabic-speaking Jews, with the result that /a/ and /e/ became sufficiently close to serve as equivalents for purposes of rhyme. The rarity of the phenomenon is presumably to be explained by the fact that Hebrew poets and authors of rhymed prose tended to maintain the historical distinction between /a/ (pataḥ, qameṣ) and /e/ (segol, ṣere), with the result that the shift in vowel quality is masked. In fact, the maqāmas in which the rhyming of /a/ and /e/ is encountered employ a decidedly lower register than those of al-Ḥarizi. In plain words, they are more ‘folksy’. It is, therefore, not surprising that in them the ‘Hebrew ʾimāla’ that I am positing occasionally breaks through.