1 Aramaic & Hebrew Metathesis Emmanuel Aïm University Paris 7 & University of Orléans [email protected] Abstact The aim of this paper is to offer an analysis of the well-known metathesis process of the sibilant+dental stops clusters in the reflexive/passive verbal forms of Ancient Aramaic and Ancient Hebrew, e.g. : (1) Aramaic / (cid:1)it + s(cid:2)(cid:3)ar / > (cid:1)ist(cid:2)(cid:3)ar *(cid:1)its(cid:2)(cid:3)ar “he has been visited” This process will not be considered as a proper feature of Aramaic and Hebrew: it will be integrated into the more general discussion on extrasyllabicity and on special status of the coronal obstruents at word edges. 1. Introduction 1.1. Extrasyllabicity It is notorious that the coronal obstruents at the margin of words have a special status. Languages like English and German exhibit constraints on word-initial and word-final consonantal clusters: they must respectively display increasing sonority and decreasing sonority. However exceptions occur when the initial consonant or the final consonant is a coronal obstruent. Let’s consider for example the case of English. The relevant word-initial constraints are given in (2). Word-initial onsets must have maximaly two consonants (2a), must display rising sonority (2b), the second consonant cannot be nasal (2c) and finaly coronal consonants are never followed by l (2d) (2) English word-initial onsets a. CC maximaly plaque b. rising sonority drum c. *Cn, *Cm *fn…, *fm... d. *coronal+l *dl… However, some clusters systematicaly violate these constraints. They have three consonants (3a), they don’t display rising sonority (3b), the second consonant is a nasal (3c), they exhibit a coronal followed by l (3d). (3) a. CCC scream1 b. decreasing sonority stem c. Cn, Cm snow, smell d. coronal+l slide Phonological studies traditionaly account for the special feature of initial s by a particular apparatus: appendix (Fudge 1969), extrasyllabicity (Kenstowicz 1994), coda preceded by an empty onset (Kaye 1992), and so on. The more current, extrasyllabicity, is given in (4): the s doesn’t belong to any syllable. However, since it is a constitutive element of the word, it is directly linked to it. 1 In the context #_r , s surfaces as š, e.g. [šred] shred. 2 (4) W (cid:1) x x x x | | | | s k (cid:4) p Let’s consider now the English word-final clusters restrictions. The relevant word-final constraints are given in (5). Word-final clusters must have maximaly two consonants (5a), and must display decreasing sonority (5b). (5) English word-final clusters a. CC maximaly hemp b. decreasing sonority elf However, some clusters systematicaly violate these constraints. They have three consonants (6a), they don’t display decreasing sonority (6b) or they violate both length and sonority constraints (6c). (6) a. CCC [wayld] wild [peynt] paint2 b. *decreasing sonority [(cid:5)ps] apse [(cid:5)dz] adze act c. CCC+*decreasing sonority [s(cid:4)ks(cid:1)] sixth [s(cid:4)ks(cid:1)s] sixths Each of these clusters end with one or more coronal obstruent, that is s, z, t, d, (cid:1). The treatment of these final consonants is broadly the same as the initial s: appendix, extrasyllabicity, onset followed by an empty nucleus, and so on. As shown in (7), the differences between word-initial extrasyllabicity and word-final extrasyllabicity are small. The main differences are firstly that the class of final extrasyllabic segments is more important including both coronal fricatives and stops, and secondly that final coronals can be flexional morphemes. (7) a. # s C(C) (C)C s # # š C(C) (C)C z # (C)C t # (C)C d # (C)C (cid:6) # b. monomorphemic word segment (snow) monomorphemic word segment (paint) morpheme (dog-s) 1.2. Another look at extrasyllabicity Another look at extrasyllabity has been recently proposed by Lowenstamm 2002. This reanalysis belongs to a wider theory in which the syllable structure is viewed as a strict alternation of C and V 2 Whether the glide must be treated as a part of the nucleus or as a part of the coda is another question. 3 slots. Within this frame, the representations of an open light syllable, a long vowel, a closed syllable and a geminate are given in (8). (8) a. C V b. C V C V c. C V C V d. C V C V C V | | | \ / | | | | | \ / | d a [da] d a [daa] d a k [dak] d a k a [dakka] Let’s return to extrasyllabicity. The word-initial consonant s is represented as the propagation of a lexical segment on an initial CV-site located on the left word-edge of every major lexical category (for independent arguments supporting the existence of an initial CV-site, see Lowenstamm 1999, 2003). For instance, the representation of the french word [sp(cid:7)r] “sport” is given in (9a), where the initial CV-site is in bold, while the representation of the french word [p(cid:7)r] “port” is given in (9b). (9) a. CV - CVCV b. CV - CVCV | | | | | | s p (cid:7) r p (cid:7) r The fundamental distinction between the traditional approach and Lowenstamm’s one is the following. In the traditional view, the extrasyllabic slot or appendix is generaly only required in languages like English, German and French. In Lowenstamm’s approach, the presence of the initial CV-site is assumed in all languages. Thus a question arises: if the initial site is the seat of a segmental restriction in languages like English, German, etc, why wouldn’t it be also the case in other languages? Here, a precision is needed : the segmental restriction on the initial CV-site in English, German and French is a restriction that arises mainly during the formation of monomorphemic words in the lexicon. This restriction doesn’t appear during other processes, like prefixation, reduplication or clitisization. For exemple, the French initial CV-site hosts non coronal obstruents like l, m, v, during clitisization, as illustrated in (10): (10) CV-CVCV CV-CVCV CV-CVCV | | | | | | | | | | | | | | | l a (cid:8) (cid:9) z ma (cid:8) (cid:9) z v o (cid:8) (cid:9) z [la(cid:8)(cid:9)z] [ma(cid:8)(cid:9)z] [vo(cid:8)(cid:9)z] “the chair” “my chair” “your chairs” Nevertheless, if a restriction on the type of consonants allowed in the initial CV is lexicaly determined in some languages, it is not excluded that a similar or a different restriction can at once exist and be morphologicaly determined in other languages. In other words, we would expect a language to perform a selection of a subset of segments during their association to the initial CV-site. Regarding this subset of segments, we of course expect it to constitute a natural class of sounds. So, the logical possibilities are manifold : glide, nasal, lateral, labial, velar and so forth. If we want to get closer to extrasyllabicty facts, two hypothesis are possible. First, the subset is the same as the word-initial subset: it only includes coronal fricatives. Secondly, the subset is identical to the word-final subset: it only includes the coronal obstuents, fricatives and stops at once. The aim of this paper is precisely to show i) that Aramaic and Hebrew exhibit a selection between the 4 coronal obstruents (fricatives and stops at once) and the other consonants during the derivation of the reflexive/passive verbal forms ii) that this selection can explain the metathesis that occurs in this verbal forms. 2. The facts 2.1. Aramaic & Hebrew verbal systems The Ancient Aramaic verbal system is given in (11). Each active stem has a reflexive/passive form.The reflexive/passive stems are discerned from their active counterparts by the presence of the prefix hit-, (cid:2)it- or (cid:2)et- (according to the dialects) and ablaut3. (11) Aramaic active passive/reflexive Basic p(cid:2)(cid:3)al (cid:2)etp(cid:3)(cid:4)el Factitive/Intensive pa(cid:3)(cid:3)el (cid:2)etpa(cid:4)(cid:4)al / (cid:2)etpa(cid:4)(cid:4)eel Causative (cid:1)ap(cid:10)(cid:3)el (cid:1)ettap(cid:10)(cid:3)al 4 In Hebrew, as shown in (12, 13), the stem that is caracterized by the prefix hit- only exists as the reflexive/passive form of the pi(cid:3)(cid:3)el stem. Note that Mishnaic Hebrew exhibits an alternative form with nit- instead of hit- in the perfective conjugation. The presence of n is traditionaly explained as an analogical formation from the nifal stem (cf. Wright 1890, Segal 1927, Bar-Asher 1999). For my part, I have nothing to say about it. (12) Biblical Hebrew active passive reflexive Basic paa(cid:3)al nip(cid:10)(cid:3)al------------------- nip(cid:10)(cid:3)al Intensive pi(cid:3)(cid:3)eel pu(cid:3)(cid:3)al hitpa(cid:4)(cid:4)eel Causative hip(cid:10)(cid:3)iil hop(cid:10)(cid:3)al - (13) Mishnaic Hebrew active passive/reflexive Basic paa(cid:3)al nip(cid:10)(cid:3)al Intensive pi(cid:3)(cid:3)eel nitpa(cid:4)(cid:4)eel ~ hitpa(cid:4)(cid:4)eel 5 Causative hip(cid:10)(cid:3)iil hup(cid:10)(cid:3)al The verbal stems in which sibilant metathesis occurs are in bold in (11, 12, 13). 2.2. Metathesis & assimilation. Let’s start with the forms that present a fricative coronal as first radical consonant. Consider the data in (14) 6. With roots beginning with s, š, (cid:1), z and (cid:1)s there regularly is metahesis of this consonant and the t of the prefix. As shown in (14b), a further voicing assimilation of t takes place with roots 3 In (11), (12) and (13) the meaning of each pattern given is mere indication. 4 Internal causative passive hu/op(cid:4)(cid:3)al is attested in Biblical Aramaic instead of the more recent (cid:2)ettap(cid:4)(cid:3)al. 5 The vocalisation nitpa(cid:3)(cid:3)al / hitpa(cid:3)(cid:3)al also occurs. 6 The next data of Ancient Aramaic will be abstracted from Onqelos and Jonathan Targums’ Judeo-Palestinian Aramaic but they are also valid for Biblical Aramaic, Syro-Palestinian, Samaritan, Syriac, Mandaic and Judeo-Babylonian Aramaic (data from other ancient dialects are scanty). The data of Ancient Hebrew will be abstracted from Biblical and Mishnaic Hebrew without distinction. For reason of place, philological issues as orthographic variants, script problems notably in some Syriac forms, uncommon verbs where metathesis or/and assimilation doesn’t occur, etc., will be ignored. I will also disregard the total assimilation of the prefix t to any first radical consonant that sometimes occurs in Babylonian, Mandaic, Galilean, Samaritan, in the late period of Syro-Palestinian and sometimes also in Biblical and Misnaic Hebrew. About all these points, see Aïm 2003. 5 beginning with z, and, as shown in (14c), a further emphasis assimilation of t takes place with roots beginning with (cid:1)s. {Please notice that for Aramaic data, the first form is always the basic reflexive stem (cid:2)etp(cid:5)(cid:3)el and the second is the intensive reflexive stem (cid:2)etpa(cid:3)(cid:3)a/eel.} (14) a. Aramaic (cid:1)it - s(cid:2)(cid:3)ar > (cid:1)ist(cid:2)(cid:3)ar “to be visited; to be inflicted upon” (cid:1)it - sakkeel > (cid:1)istakkeel “to look at, reflect ; to be/become wise” (cid:1)it - š(cid:2)lee > (cid:1)išt(cid:2)lee “to be abandonned, forgotten” (cid:1)it - šaggeeš > (cid:1)ištaggeeš “to be confused, perplexed, excited” Hebrew hit - sabbeel > histabbeel “to stuff oneself, grow fat” hit - (cid:1)akkeer > hi(cid:1)takkeer “to profit; to deal in” hit - šappeel > hištappeel “to be humble, gentle ; to be lazy, indolent” b. Aramaic (cid:1)it - z(cid:2)ra(cid:3) > (cid:1)izd(cid:2)ra(cid:3) “to be sown, to be stocked with seed” (cid:1)it - zayyeen > (cid:1)izdayyeen “to be equipped, armed” Hebrew hit - zayyeep(cid:10) > hizdayyeep(cid:10) “to be falsified, forged” c. Aramaic (cid:1)it - (cid:1)(cid:1)s(cid:1)(cid:1) (cid:2)lee > (cid:1)i(cid:1)(cid:1)s(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) (cid:2)lee “to be moved; to tremble, shake” (cid:1)it - (cid:1)(cid:1)s(cid:1)(cid:1) abba(cid:3) > (cid:1)i(cid:1)(cid:1)s(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) abba(cid:3) “to be dipped, immersed; to be soaked” Hebrew hit - (cid:1)(cid:1)s(cid:1)(cid:1) addeeq > hi(cid:1)(cid:1)s(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) addeeq “to justify oneself, excuse oneself” Let’s now consider in (15) the forms with a coronal stop as first radical consonant. As shown in (15a), voicing assimilation of t occurs with roots beginning with d. As shown in (15b), emphasis assimilation of t occurs with roots beginning with (cid:1)t. (15) a. Aramaic (cid:1)it - d(cid:2)(cid:11)eeq > (cid:1)idd(cid:2)(cid:11)eeq “to be pressed, squeeze oneself” (cid:1)it - daheen > (cid:1)iddaheen “to drip, be fat” Hebrew hit - dabbeer > hiddabbeer “to hold communion, converse” b. Aramaic (cid:1)it - (cid:1)(cid:1)t(cid:1)(cid:1) (cid:2)(cid:3)an > (cid:1)i(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) (cid:2)(cid:3)an “to be laden ; to be carried” (cid:1)it - (cid:1)(cid:1)t(cid:1)(cid:1) abba(cid:3) > (cid:1)i(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) abba(cid:3) “to be sunk” Hebrew hit - (cid:1)(cid:1)t(cid:1)(cid:1) ammaa > hi(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) ammaa “to be unclean ; to make oneself unclean” Finally, as shown in (16), whith roots beginning with t, the t of the prefix does not undergo the spirantization since it constitutes a geminate cluster with the first consonant of the root. (16) Aramaic (cid:1)it - t(cid:2)nee > (cid:1)itt(cid:2)nee “to be repeated” (cid:1)it - taggar > (cid:1)ittaggar “to make profit, to be benefited” Hebrew hit - tabbeel > hittabbeel “to be mixed, seasoned ; to defile oneself” 6 It is very important to note : i) that the metathesis also applies in Aramaic when an epenthetic i/e is (optionaly) inserted between the affix t and the first radical consonant, e.g. Syriac7 : /(cid:1)et-šqel/ > (cid:1)ešetqel ~ (cid:1)ešt(cid:2)qel “he has extoled himself” The metathesis is not a mere phonological process triggered to avoid an illicit sibilant+dental stop cluster: even disjoined by a vowel, the sibilant and the dental change places. ii) the metathesis doesn’t apply outside the reflexive stems, as illustrated in (17): (17) Aramaic netšat “she has teared off” *neštat (cid:2)ntš Hebrew notšii “my uprooting” *noštii (cid:2)ntš Aramaic qudšaa “the sanctity” *qušdaa (cid:2)qdš Hebrew qodšii “my sanctity” *qošdii (cid:2)qdš Obviously, the metathesis is a morphological process: the succession of any sibilant and any dental stop is not a sufficient condition. To trigger the metathesis, the sibilant must be the first consonant of a root and the dental stop must be the passive/reflexive affix. 3. The analysis 3.1. Preliminary analysis. Previous works on this facts (Greenberg 1950, Malone 1971) only deal with the sibilants metathesis. Certainly, the assimilation of t with the dental stops is considered as another phenomenon. On the contrary, my goal is to link these two processes. In order to obtain the surface forms with roots beginning with d and (cid:1)t, only a regressive assimilation is required, as shown in (18) : (18) / hit+dabbeer / / hit+(cid:1)(cid:1)t(cid:1)(cid:1) ammaa / regressive assimil. hiddabbeer hi(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) ammaa [ hiddabbeer ] [ hi(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) ammaa ] In order to obtain the surface forms with roots beginning with z and (cid:1)s, two rules are required: a metathesis rule and an assimilation rule. According to rules order, the assimilation rule can be regressive, as in (19a) or progessive as in (19b). (19) a. b. / hit+zayyeep / / hit+zayyeep / regressive assimil. hid-zayyeep metathesis hiz-tayyeep metathesis hiz-dayyeep progressive assimil. hiz-dayyeep [ hizdayyeep(cid:10) ] [ hizdayyeep(cid:10) ] 7 This epenthesis appears in Judeo-Palestinian Aramaic, Syriac and Mandaic. It is not restricted to the passive/reflexive stems but belongs to a wider phonological process (see Stevenson 1924, Bohas 1999, Duval 1881, Nöldeke 1875, Macuch 1965). 7 At first sight, (19a) seems to be the best solution. Firstly, other assimilations attested in Ancient Aramaic and Ancient Hebrew are generaly regressive. Some examples from Syriac are given in (20). (20) /tuub-taanaa/ > tuup(cid:5)taanaa “happy” /p(cid:2)šii(cid:1)(cid:1)t(cid:1)(cid:1) -taa/ > p(cid:2)šiittaa “simple” /ragguuz-taanaa/ > ragguustaanaa “irascible” /(cid:11)esdaa/ > (cid:11)ezdaa “clemency” /metbar/ > medbar “to break” Secondly, regressive assimilation is also required with roots beginning with d and (cid:1)t. Recall previous data in (18). On the other hand, as pointed out by Bolozky (1997) regarding the same data in Modern Hebrew, the solution (19a) orders a mere phonetic process, the assimilation, before a restricted morpho-phonological one, the metathesis. So, this order is not desirable and the solution (19b) must be prefered. But a difficulty appears: we need two assimilation processes, one regressive as in (18), an other progressive as in (19b). However, we can suppose that the forms with roots beginning with d and (cid:1)t are also obtained with firstly a metathesis rule and secondly a progressive assimilation rule, as shown in (21), since the surface forms obtained are correct. (21) / hit+dabbeer / / hit+(cid:1)(cid:1)t(cid:1)(cid:1) ammaa / metathesis hid-tabbeer hi(cid:1)(cid:1)t(cid:1)(cid:1) -tammaa progressive assimil. hid-dabbeer hi(cid:1)(cid:1)t(cid:1)(cid:1) -(cid:1)(cid:1)t(cid:1)(cid:1) ammaa [ hiddabbeer ] [ hi(cid:1)(cid:1)t(cid:1)(cid:1) (cid:1)(cid:1)t(cid:1)(cid:1) ammaa ] In the same way, we can suppose that the forms with roots beginning with t are obtained with a metathesis rule, as in (22). Note that the idea that coronal stops also undergo metathesis has been previously suggested by Kaufman 1997 and Lipinski 19978. (22) / hit+tabbeel / metathesis hit-tabbeel [ hittabbeel ] The fact that progessive assimilation occurs only there is not a problem, since it is the only assimilation systematicaly noted in scripts. On the contrary, regressive assimilations are just sporadically written. So, we can conclude that the general case is the progressive assimilation, not the regressive one. Therefore, the metathesis implies both coronal fricatives s, š, (cid:1), (cid:1)s, z and coronal stops t, d, (cid:1)t. The problem now is to understand first of all why the metathesis occurs only in the reflexive/passive stems and secondly why the metathesis implies coronal obstruents only. A brief draft of the answer can be given. As shown in (23), the left margin of the word in the surface forms is always constituted by a subset of segments: the coronal obstruents. Does it mean that the left margin is restricted to a subset of segments ? 8 Note also that the akkadian reciprocal infix -t- / -ta- undergoes metathesis with all coronals, stop and fricative; on this facts, see Aïm 2003: 271-274. 8 (23) regular case (hi) t qa(cid:1)t(cid:1)teel (hi) s tabbeel (hi) š tappeek (hi) (cid:3) takkeer metathesis (hi) z dayyeep(cid:10) (hi) (cid:1)s (cid:1)taddeeq (hi) t tabbeel (hi) d dabbeer (hi) (cid:1)t (cid:1)tammaa I am now going to expose the theoretical framework adopted here. Next, a representation of the reflexive stems and an analysis of the metathesis will be offered. 3.2. Theoretical framework. I follow in this paper the templatic analysis of Classical Arabic proposed by Guerssel and Lowenstamm 1990. In this frame, all Classical Arabic verbal forms are derived from a single template. This template is given in (24). (24) CV CV CV CVCV The derivation of each verb is performed by the association of segments to specific positions called “head positions”. There are two head positions: an initial CV-site and a medial CV-site. These two sites are in bold and underscored in (24). For instance, verbal forms such as kaatab (stem III) and kattab (stem II) are respectively derived by identification of the medial CV-site by means of Vocalic-spread (25a) and Consonantal-spread (25b). (25) a. k a t a b b. k a t a b | | | | | | | | | | CV CV CV CVCV CV CV CV CVCV kaatab- kattab- As depicted in (26), verbal forms such as (cid:2)iswadad (stem IX) and (cid:2)iftaxar (stem VIII) are derived by anchorage of the first consonant of the root to the initial CV-site9. (26) a. s w d b. f x r | / \ | | CV CV CV CVCV CV CV CV CVCV | t ((cid:1)i) swadad- ((cid:1)i) ftaxar- Regarding the two other reflexive stems, that is stem V takattab and stem VI takaatab, Guerssel and Lowenstamm argue that these forms are merely obtained by concatenation of the prefix ta- to their 9 Note that Guerssel & Lowenstamm assume that the infix t- of ftaxar (26b) is a non-derivational (reflexive) feature of the base. Note also that the (cid:2) and the i on the left are not represented since they are prostethic segments. For convenience, vocalic melodies are not represented. 9 non-reflexive counterparts, respectively kattab and kaatab. So the derivation of these forms doesn’t involve templatic morphology as represented in (27). (27) ta + kattab > takattab- ta + kaatab > takaatab- The problem is how to represent the reflexive stems of Aramaic and Hebrew. According to the traditional Semitic studies, we may suppose that the segments hi , (cid:2)i or (cid:2)e of the prefix are prosthetic. Now, regarding the representation of the reflexive stems, several hypothesis are possible: the morpheme t could be either a prefix or an infix. So, let’s make the following hypothesis. 3.3. Reflexive/passive stems and metathesis: a proposal. I suppose that the reflexive forms are derived from their non-reflexive counterparts (see Aïm 2003: 275-279 for a discussion on this point). Let’s begin with the Aramaic basic reflexive stem (cid:2)etp(cid:5)(cid:3)el. I assume that it is built from the basic stem p(cid:5)(cid:3)al represented in (28): (28) p(cid:2)(cid:3)al R R R 1 2 3 | | | CV CV CV CV CV The basic reflexive stem (cid:2)etp(cid:5)(cid:3)el is obtained by anchorage of the first consonant of the root to the initial CV-site as it is represented in (29) : (29) R R R 1 2 3 | | CV CV CV CV CV t and next by the anchorage of the morpheme t. I suppose that this morpheme is a floating segment. It is linked to the only available consonantal slot, as shown in (30) : (30) R R R 1 2 3 | | CV CV CV CV CV t The reflexive stem (cid:2)etpa(cid:3)(cid:3)al / hitpa(cid:3)(cid:3)eel is obtained in the same way. It is built from the stem pa(cid:3)(cid:3)el / pi(cid:3)(cid:3)eel depicted in (31) : (31) pa(cid:3)(cid:3)el / pi(cid:3)(cid:3)eel R R R 1 2 3 | / \ | CV CV CV CV CV 10 Its reflexive counterpart is obtained firstly by anchorage of the first consonant of the root to the initial CV-site : (32) R R R 1 2 3 / \ | CV CV CV CV CV t and next by association of the floating t to the available consonant slot : (33) R R R 1 2 3 / \ | CV CV CV CV CV t Now, let’s assume that the initial CV-site of the template has the following property : (34) the initial CV-site can host coronal obstruents only So, when the first consonant of the root is a coronal obstruent, it links to the initial CV-site ; the floating t links to the only available consonantal slot : (35) R = coronal obstruent 1 a. (cid:1)est(cid:2)(cid:3)ar b. (cid:1)estakkeel s (cid:3) r s k l | | / \ | CV CV CV CV CV CV CV CV CV CV t t But, when the first consonant of the root is not a coronal obstruent, it cannot link to the initial CV-site; it then stays in its initial position. Therefore, the floating t must link to the only available consonantal slot, that is the initial CV-site. Since it is a coronal obstruent, it can do so :