1

Chapter 504 and modern Arabic dialectology: What are kaškaša and kaskasa, really? With Sibawaih, small chapters often hide issues of larger importance, and their interpretation requires bringing to bear evidence from a variety of sources. These are, in the first instance, an understanding of Sibawaih’s goals, methodology and linguistic theory. What further sources may be relevant depends on the issue involved. A source that has traditionally been less appreciated among Arabicists in the historical interpretation of the language is the contemporary Arabic dialects, which play a crucial role here. In this small contribution I would like to relate a basic observation on object pronouns in Sibawaih to larger issues of Arabic historical linguistics. The discussion begins in sections 1 and 2 with Sibawaih’s observations on the four variants of the 2FSG oblique (object, possessive) pronoun suffix. Following a suggestion by Johnstone (1963), it is argued that of the two variants, nominally –kiš and –kis, the first definitely and the second possibly, is identical with the modern variants, -č ([tš]) and –ts respectively (section 3). Thereafter, the rest of the paper addresses the issue of why, if indeed Sibawaih would have heard –č and –ts, he nonetheless interpreted them as –kiš and –kis. To answer this question, three differentiated linguistic issues are examined in detail. The first is phonetic. It will be shown in section 4.2 that Sibawaih did in fact, among a set of eight proscribed sounds, observe the variant [č]. A key aspect of the argumentation in this section leads to an interpretation of Sibawaih’s system for identifying non-basic Arabic phonemes. Secondly, from a sociolinguistic perspective it will be shown that as it is among the proscribed variants, Sibawaih could not have recognized a variant –č of the 2FSG (section 5.1), since proscribed variants are not a functional part of Arabic grammar. Further issues militating against his recognizing a –č in the 2FSG are discussed in sections 5.2 and 5.3. Section 6 briefly reflects on the problems of interpreting Sibawaih. Thirdly, and much more briefly in section 7, it will be argued from evidence of the modern dialects *č reflexes of *k, including the 2FSG –ič, have such a wide distribution, including North Africa, the eastern Nile delta region (Sharqiyya) and Khorasan – besides its core area of the eastern Arabian peninsula, the Gulf area and parts of the Levant, that linguistic reconstruction leads to the same conclusion as does a close reading of Sibawaih’s linguistics, namely that already by his era the change *k > *č was a part of Arabic. 1. Chapter 504 Chapter 504 (Kitaab II: 322-3) contains three substantive observations about affixal pronominal behavior on the verb. First, Sibawaih notes that the 2FSG suffix can be realized as –ši, -kiš or –kis, as well as –ki. (1) 2FSG variants in Sibawaih1

1 Later sources add other variants, which I will not deal with here. Ibn Manḍur reports on the basis of ħadith (Lisan VI: 197), for instance, that a –si variant exists as well which, apparently, applies to both 2M

2

a. -ši b. -kiš c. -kis d. -ki Secondly he notes that in a double-object verb, the 2F and 2M can be realized as long vowels, as in ʔuʕṭiy-kiy-haa “I give you.F it.F” or ʔuʕṭiy-kaa-haa “I give you.M-it.F. Finally, he observes that some Arabs lengthen the vowel of the 2FSG perfect suffix before an object pronoun, as in ḍarab-tiy-hi “you.F hit him”. Each of these observations invites independent examination in its own right. Given that the linguist who this book is honoring made significant contributions to what will be argued to be the successor dialects of the –kiš and –kis forms, I will concentrate on these two forms.2 I will look at this chapter from two perspectives. First I will summarize Sibawaih’s treatment, and secondly I will raise the question of its plausibility in the light of the situation in the Arabian peninsula and adjacent areas today. Sibawaih notes that there are Arabic tribes whom he identifies as “many Tamimi” and “some Asad” who instead of using the suffix –ki for the 2FSG object suffix, use –ši. His stock examples are. 2. ʔinna-ši đaahib-at-un “Indeed you.F are going” maal-u-ši “your.F wealth” As ever with Sibawaih, each of his observations is embedded in what is for him a causal explanation for the phenomenon he observes. Pure descriptivism was not his methodology. In this case, he observes that the –ši form serves as disambiguation with the 2MSG object suffix in pausal position, the –š form serving to “affirm and emphasize” (taћqiyq wa tawkiyd) the difference. In pausal position, the M-F contrast is neutralized with the –k forms. The following paradigm illustrates this.

and 2F., ʔabuu-sa “your father”. Along the same lines, a reviewer draws attention to Suyuṭi’s treatment of the phenomenon in his Muzhir (chapter 11, p. 102), in which a -š variant is associated with 2FSG, a -s variant with 2MSG. While constraints of space prevent following up on these interesting strands here, it is my suspicion, subject of course to an examination of Suyuṭi, Ibn Manḍur and others, that it is Sibawaih’s eyewitness account which should be given primary credibility and attention. In the context of a footnote which addresses issues beyond the scope of this article, the same reviewer asks how Sibawaih’s phonetic variants were received by subsequent linguists, and whether the “jiym like a kaaf” discussed below couldn’t also have referenced an already-existing [g] reflex of “jiym”, as in contemporary Cairene (Woidich and Zach, 2009). To these I would add the question why Sibawaih, but not, apparently, his teacher al-Khalil, took an interest in phonetic variation (Sara, 2013). From my perspective, none of these interesting questions have self-evident answers, so let them be left for future interpretation. 2 Cf. Bruce Ingham’s many contributions to the study of Arabian peninsular and Gulf Arabic, e.g. his masterly concise study of Najdi Arabic (1994).

3

3. waṣl (context) waqf (pause) 2MSG bayt-u-ka bayt-u-k# 2MSG bayt-u-ki bayt-u-k# 2FSG bayt-u-ši bayt-u-š# He points out, analogically, that a contrast carried by a consonant is more perceptible than one shown by short vowels, adducing the pronominal contrasts FPL –na, vs. MPL –uw, or ʔantum vs. ʔantunna. Furthermore, Sibawaih explains the choice of –š as opposed to another consonant by phonological analogy: -ši is voiceless (mahmuwsa) like –k and an oral consonant is preferred to another theoretically possible choice, a pharyngeal (i.e. –ћ) because –k is also an oral consonant. Sibawaih observes two further variants from unnamed speaker (naasun min al-ʕarab). These are –kis and –kiš. The final –s or –š serve to “perceptualize” (yubayyinuwna) the vowel –i in pausal position. 4. (a) ʔaʕṭay-tu-kis# I gave you.F.SG (b) ʔaʕṭay-tu-kiš# These occur in pause only. In context (waṣl) there is no need for them, as the final –i of –ki is maintained. Sibawaih explains the choice of –s in (4a) as that of choosing one of the added consonants (ћuruwf zaaʔida), consonants which can be added for morphemic value.. 2. Internal logic and problems In post-Sibawaihian linguistics these two variants came to be subsumed under the rubric of kashkasha and kaskasa (e.g. Lisan VI: 196). These however, were not terms Sibawaih used. Sibawaih’s explanations are interesting for reasons beyond the way in which they shed light on his causal-based descriptivism. In particular, the explanations are phrased in perceptual and in morphological terms. The variants serve contrasts that make meanings clear. The choice is a very rational one. –š disambiguates two k’s in pausal position. They (Arabs) move from /k/ to /š/ and not from /k/ to /ћ/, because they remain in the oral tract. They choose –s in –kis because –s otherwise has morphemic value as an added consonant (e.g. in the form X verbal derivate –ist). The explanations are not without problems. –ši ~ -š occurs in all position, pausal and context, and no explanation is given as to how –š is arrived at. Given the morphologically-based explanation of –kis, -s would be the more expected form. Sibawaih is silent about why –š should occur at all in the –kiš variant, as –š is not among the added consonants. These are issues I will not address further here.

4

3. Modern dialects and language history This chapter is particularly interesting for students of Arabic, as it treats in very explicit form variation which appears to have a great deal in common with variants attested in the contemporary world. In particular, modern dialects have four basic variants of the 2FSG object suffix (5) Modern reflexes of the 2FSG suffix (a) -iš 3 Highland Yemen, southeastern Arabian peninsula (b) -its Najdi (Ingham 1994: 14) (c) -ič [itš] Gulf, “gilit” Iraqi, Jordanian and Syrian desert, rural Palestinian (d) -ik or –ki otherwise in the Arabic-speaking world A number of scholars have discussed these variants in relation to the values which Sibawaih gives for the 2FSG morpheme (see (1)). Early works by Barth (1911) and Schaade (1911: 62) accept Sibawaih’s values at face value, though Barth and Schaade have different explanations for these variants (Watson 1992: 67). Later scholars have more variegated interpretations. Cantineau (1960: 65) interprets the kashkasha as indicating a –š pronunciation, and kaskasa a –s pronunciation of the 2FSG suffix. As noted, “kashkasha/kaskasa” were not the terms Sibawaih himself used, and Sibawaih nowhere indicates a –s as a reflex of the 2FSG. Furthermore, Sibawaih clearly indicated three variants beyond –ki, namely -ši, -kiš and –kis, not two as Cantineau interprets the situation. Watson (1992) deals with the –kiš variant, suggesting that it be reinterpreted as –kç, citing (North) Yemeni dialects which have this realization of the 2FSG morpheme (1992: 77, see n. 24 below). Johnstone (1963: 225), without advancing specific arguments other than the striking coincidence of the correspondence, has a different interpretation, namely that (1b/c) in fact should be interpreted as (5b/c). Holes (1991: 670, n. 64) also appears to be favorable to this interpretation, even if his final summary reflects a different chronology of sound change. In this paper I would like to pursue the logic of this assumption, after interpreting Sibawaih’s treatment in detail, asking the question, if it is correct, why was it not so represented in Sibawaih? It should also be noted that although he does not discern a [-č] or [-ts] in the kaškaša and kaskasa variants, Cantineau (1960) does detect evidence elsewhere for [č] in Sibawaih’s phonetics/phonology, pace (6) below, as will be discussed in section 4.2 below. Before proceeding it is relevant to briefly summarize the detailed historical treatment of the development of (5b/c) in Arabic as developed by Holes (1991). Holes, correctly I believe, interprets the development of –ši as earlier than and independent from the development into –č and –ts. –š has reflexes throughout the Semitic languages, e.g. in Amharic, Harari and Gurage among Ethiopian Semitic (1997), among the modern South Arabian languages (1997) and, it can be noted, in contemporary Ma’lula Aramaic

3 -ič and -iš intermingle in Oman, with –ič dominant (Holes 1989). In the Gulf region, -iš is not found north of the Emirates, where it is in any case rare.

5

(Jastrow (1997: 337), as well as in Arabic dialects of eastern Arabian peninsula, from Yemen into the Emirates and Oman. It therefore can be inferred to belong to an old Semitic stratum, a retention from proto-West Semitic.4 –č and –ts on the other hand are Arabic-internal developments. Holes (following Johnstone 1963) points out that the change of *k > č/-ts extends beyond the 2FSG suffix to front vowel contexts generally where *k preceeds a front vowel, either /i/ or front /a/, e.g. čaan/tsaan “he was”. This general, conditioned change is paralleled by the change of *g5 > dž/dz, e.g. ğiddaam/dziddaam “in front”. Holes notes that of the two changes *k and *g, –č is the more widespread form: it occurs in Baghdadi/gilit and southern Iraqi, as well as in central Syrian, rural Palestinian and Jordanian dialects, whereas the [dž/dz] variants are largely limited to the eastern Arabian peninsula and Gulf region. He therefore proposes that the change of *k to > č preceded that of *g > dž (1991: 666). He furthermore suggests a linear development of *č > ts and *dž > dz in those dialects which have the latter forms, otherwise č/dž remaining as is. Overall one has: (6) Diachronic development of *k/g + front vowel: Holes 1991: 666 Front vowel environments: k > č > č > ts g > g> ğ ([dž]) > dz B1> B2 >B3 > B4 B1 is the original situation. On the basis of present-day distributions and what we know of historical events (e.g. the spread of Arabic groups from central Saudi Arabia to the Gulf in the eighteenth century), Holes suggests a chronology for these changes. B4 is the contemporary situation in central Arabia today, B3 was in place at least by the 18th century, perhaps earlier, and is the situation in the Gulf dialects, B2 was in place by the 13th century and is the situation in Baghdadi Arabic and elsewhere. I will return to the issue of the chronology of these changes in section 7. 4. Sibawaih At this point it is time to return to an interpretation of the variants as described by Sibawaih in (1), noting that disagreement with Sibawaih, a master linguist, is not to be undertaken without careful argumentation. Nonetheless, I think that Johnstone’s interpretation of (5b/c) as continuing (1b/c) is correct, i.e. that Sibawaih ‘meant’ –č and perhaps –ts.

Before beginning, it can be noted that contemporary (5a) and (5d) are unproblematic from a comparative perspective. Each continues forms which were observed to be present during Sibawaih’s time (see Owens 2009: 251). I will discuss the reflexes in (5b, c) under

4 See also Fischer (1956) who suggests borrowing as the source of –ši. The ultimate origin of this morphemic variant is not directly relevant to the current exposition. 5 Holes argues that *g in turn derives from *q, an issue outside the scope of this paper.

6

five headings: coincidence of forms (4.1), issues of Sibawaih’s phonetics and phonology (4.2), social status of the variants (5.1), and issues related to morphology and orthography (5.2) and to pausal position (5.3). 4.1 Coincidence First of all, it is striking that Sibawaih happens to mention two variants of the 2FSG whose final segments, -kiš and kis, correspond exactly to the final segment of the two contemporary variants, [-tš] and [–ts]. Moreover, it is odd that if –kiš and –kis were indeed the variants, they have disappeared with no trace in contemporary Arabic, quite in contrast to –ki and –ši. Still, coincidence alone is hardly a sufficient justification. 4.2.Sibawaih’s phonology and phonetics Sibawaih’s introductory chapter on al-idγaam “assimilation” (chapter 565, II: 452-5) is essential reading regarding his phonetics and phonology. For Sibawaih, phonetics was a functional phonology, hence his most important observations on phonetics and phonemic inventory come at the beginning of the chapter in which he begins to outline the rich phonemics and morphophonology of Arabic. Sibawaih defines 29 standard phonemes or ћuruwf for Arabic. I term these sounds “basic phonemes” and conventionally represent them in phonological slashes “/…/”. In addition, there are two sets of what can be called an extended variant list. As will become clear in the subsequent discussion, these two sets are crucial to an interpretation of the sounds/morphophonemes under discussion here, because they contain no less than four variants which belong to what can be called the jiym/shiyn complex. To date, there is no systematic integration of all of these sounds into an interpretation of Sibawaih’s phonetic and phonological thinking, though a number of linguists (see below) have cited different variants in one place or another. The first set consists of six further phonological variants, which are sanctioned for Quranic recitation and poetry. These can be listed without further comment. The variants for the forms immediately relevant to this paper are listed in brackets and will be justified in the following. I will conventionally represent these in phonetic brackets, “[…]”. Sanctioned variants 7. the medial hamza (bayna bayna) 8. imalized alif (see 5.1 below) 9. the light /n/ 10. the shiyn like a jiym, (= [ž]) 11. the ṣaad like a zaaʔ (= [ẓ] 12. the emphatic alif. Most members of the set of sanctioned variants are mentioned elsewhere as well, either in the Kitaab (e.g. II: 279 ff. for imala, II: 168 ff. for medial hamza), and/or in other linguistic traditions (e.g. Ibn Mujahid 105 for the ṣaad like a zaaʔ). Except for the shiyn

7

like a jiym (see (24) below), all of these are readily interpretable as allophonic variants of one of the 29 core phonemes. A classic case relates to the complex conditioning of the imala, a vowel-harmonic palatalization of [aa] (Owens 2009: chapter 7), but others are equally susceptible to a conditioned allophonic reading, for instance, the ṣaad like a zaaʔ is a voiced ṣaad in the environment of another voiced consonant. Beyond these, Sibawaih also notes a second set of eight further pronunciations which he does not sanction for the Qur’an and poetry, which are not considered good, and which he proscribes as not frequent among those who use good Arabic. Proscribed variants 13. The jiym like a kaaf (= [č] or [c] or [ts]) 14. The jiym like a shiyn (= [č]) 15. The weak ḍaad 16. The ṣaad like a siyn 17. The ṭaaʔ like a taaʔ 18. The ḍaa’ like a θaaʔ 19. The baaʔ like a faaʔ (= [p]) 20. The kaaf that is between the jiym and kaaf (= [?]) In addition in Chapter 525 Sibawaih identifies a: 21. sound (ħarf) that is between a kaaf and a jiym (= [g]) A number of different values have been suggested for various of these sounds. For (13) Cantineau, (1960: 58, following Bravmann) as well as Bakalla (1984), suggest the value [g]. For (14) Bakalla suggests [c] whereas Schaade (1911: 69, n. 33) speaks of a palatalized [z] (“palatisiertes z”), which he represents as [ź] (= [ž] ?). In addition, Cantineau, usually an expert interpreter of Arabic phonetics, introduces a “kaf comme jim” (1960: 65), citing his own example of ǧaafir for kaafir. Following Howell, he corrects this to čaafir.6 Leaving aside Cantineau’s interpretation, all of these representations of Sibawaih’s proscribed sounds have two attributes in common: (1) they are all, taken individually, plausible, and (2) all lack systematicity in that Sibawaih’s sounds are interpreted on an ad hoc case by case basis, not in terms of a common formula. It can be suggested here that the key to interpreting (10, 11, 13, 14, 19, 21) begins with a different chapter than that in which the sounds are described phonetically, namely chapter 525 (II: 375). This is the chapter in which he treats the adaptation of Persian sounds, including consonants, into Arabic. Two of the sounds are (19) and (21). Speaking of (21) Sibawaih writes, “They [Arabs, j.o] convert the [Persian j.o.] sound that is between the kaaf and the jiym into a jiym …” al-jurbuz “imposter” < gurbuz, al-jawrab

6 Cantineau’s “kaf comme jim” is also picked up by Holes (1991: 670, n. 64) and Watson (1992: 76, see n. 24 below).

8

“sock” < gawrab.7 Sibawaih more briefly notes that the sound between the faaʔ and baaʔ either is changed to either [f] or [b]. (22) Sibawaih’s designation firind/birind “decorative garment, decorated sword handle” < parand = [p] (Lane 1980, 6: 2389).8 baaʔ like a faaʔ (23) Sibawaih’s designation jurbuz < gurbuz = [g] sound between kaaf and jiym These two cases are instructive for postulating a systematicity behind Sibawaih’s designation. Sibawaih the phonetician would have thought in his phonetic classificatory terms. As in contemporary articulatory phonetics, Sibawaih classified sounds according to their place of articulation (muxraj or maxraj, pl. maxaarij), whether they are voiced (majhuwr) or voiceless (mahmuws), and their manner, which roughly in Sibawaih is expressed by how the sound (ṣawt) flows through the vocal tract, e.g. whether it is stopped = šadiyd, fricative = rixwa, and so on. In addition he describes the secondary articulation of emphasis (ʔiṭbaaq) which is irrelevant for current purposes. It can be assumed that in describing the extended variants, Sibawaih was thinking in these phonetic categories, even if he did not describe the extended sounds individually. The question is, what parameters he used to operationalize them. A first place to start looking for general phonetic parameters is in the two parts of the description itself. Sibawaih uses two formulations, either “sound X ka- sound Y”, “sound X like sound Y”, or “a sound between sound X and sound Y. There is either a likeness or a betweenness. Previous analyses, as noted already, have made suggestions for individual sounds, e.g. (13), a jiym like a kaaf is a [g]. What they haven’t attempted is to break X and Y into phonetic categories, generalizable to a common model. Clearly the extended sounds are a single sound. A “jiym like a kaaf”, whatever is meant, is certainly a single sound. Single sounds in Sibawaih’s phonetic theory can only consist of the discrete place, manner and voicing parameters which he defined. It can be suggested that Sibawaih used the “like” relation and the “between” relation to define these composite sounds, and if this is the case, he would have needed a system to make compound sounds regularly. The two relatively certain examples allow a phonetic model to be extracted, which describes such a system. Taking (19) for exemplification, repeated here,

7 Sibawaih (II: 375.21) observes that alternatively the in-between kaaf could be realized as [q]. I would like to thank Corey Miller of the University of Maryland, CASL (p.c’s. April 2011) for discussion and interpretation of these correspondences. 8 Ibn Manḍur gives both firind and birind, as loanwords (from *parand), with the meaning of “a decorative sword” (3: 89, 334), i.e either /f/ or /b/ as reflex of Persian *p.

9

24. The baaʔ like a faaʔ (= [p]) knowing that the outcome is a voiceless bilabial plosive, one need only look in the phonetic attributes of baaʔ and faaʔ to find which phonetic parameters Sibawaih took from each sound. These are marked in boldface in the following. baaʔ= voiced (majhuwr), stop (šadiyd), bilabial (al-šafataan) faaʔ = voiceless (mahmuws), fricative (rixwa), labio-dental The Persian [p] takes the following attributes: [p] = voiceless (mahmuws), stop (šadiyd), bilabial (al-šafataan) What Sibawaih has done is to take the voicing parameter of the second sound, Y, and the manner and place parameter from the first, X. The boldfaced attributes are taken from each sound and reassembled in the extended phoneme. A general formula can be postulated: (25) X = place/manner; ka/bayna Y = voicing This description serves as an hypothesis; Sibawaih used the X ka/bayna Y as a general model, whereby X = manner and place, Y = voicing. Sibawaih’s extended sounds are, in most cases, not as it were sounds at all, but rather instructions on how to combine the phonetic features of each constituent sound to interpret a composite extended sound. The model can be tested against (10, 11, 13, 14, 21). (26) Sibawaih’s designation composite features phonetic interpretation ħarf between kaaf and jiym place/manner of /k/+ voicing of /j/ [g] (27) Sibawaih’s designation composite features phonetic interpretation ṣaad like a zaaʔ alveolar/fricative + vd [ẓ] (28) Sibawaih’s designation composite features phonetic interpretation shiyn like a jiym alveopalatal/fricative + vd [ž] (29) Sibawaih’s designation composite features phonetic interpretation jiym like kaaf place/manner of /j/ + voicing of /k/ [č], [c] or [ts] (30) Sibawaih’s designation composite features phonetic interpretation

jiym like shiyn place/manner of /j/, voicing of /š/ [č] In all cases the formula gives a plausible result. Regarding (26) on the basis of the loanword chapter cited above, the phonetic value, based on the Middle Persian original, is

10

[g], i.e. the voiced counterpart of [k] (Saleman 1930: 13, Boyce 1975: 169). It is therefore unlikely that the “betweenness” resides in physical distance (place of articulation) since in Persian, there is no physical distance between [k] and [g]. Rather, “betweenness” should be interpreted in classificatory phonetic terms. [g] has the place and manner of articulation of a [k], but the voicing of a jiym, i.e. lies between two articulatorily-defined coordinates. All of the remaining sounds as interpreted in (26-30) are attested either in variants of Old Arabic, and/ or in the modern dialects. (27), [ẓ], is found inter alia in the Qiraaʔaat, (Ibn Mujahid 106-7), ṣiraaṭ “road, path”, the reading of Ħamza, as well as one interpretation of Abu ʕAmr ibn ʔAʕlaa’s pronunciation being a sound “between a zaaʔ and a ṣaad”, i.e. ẓiraaṭ.10 (28) is the widespread pronunciation of jiym found today in most of North Africa and in the Levant. (29) and (30) are “palatalalized k’s”, which as seen in (5) are widespread inter alia in the Arabian peninsula. There is an interpretive problem here, in that both (29) and (30), in the current formula, allow a [č] reading. Such a result runs against an expected precision in Sibawaih’s description. That the jiym is crucially implicated in the current issue is indicated on a prima facie basis by the basic observation that it is mentioned no less than five times in the description of the extended variants, i.e. in over a third of the phonetic descriptions. Here a closer look at basic jiym itself is necessary to elucidate the problem. Uncontroversially, the basic jiym is a stop (šadiyd) and voiced (majhuwr) sound. As is well known, in the case of jiym Sibawaih did not specify a contrast between an affricated and plain stop. It is simply “šadiyd”. As far as place of articulation goes (muxraj), it is placed after the kaaf (moving back to front), “in the middle of the tongue and … middle of the hard palate (al-ћanak al-ʔaʕlaa)” (Kitaab II: 453) in a class consisting of three phonemes, yaaʔ, shiyn and jiym. Further to the front of these three sounds comes [ḷ], an emphatic lateral fricative. A literal reading of this place of articulation would give the phonetic values to these three sounds of [y], [ç] (vl. palatal fricative) and [j] (vd. palatal stop), and in fact the latter two values have been suggested by various scholars.11 However, this literal interpretation runs afoul both of phonetic and Sibawaih-specific considerations.

9 Post-vocalic [g] could also be realized as [γ] in Middle Persian. Sibawaih does not discuss this variant, and his examples are all word-initial occurrences. I would like to thank Fabrizio Pennacchietti for the Middle Persian references. 10 One sees an interesting link between Sibawaih’s sanctioned variants, e.g. the ʔimaala and the “the ṣaad like a zaaʔ”, and the later ‘standardized’ Qiraaʔaat. 11 For jiym in Sibawaih, Cantineau would seem to recognize a [j], “occlusive palatale-dorsale” (1960: 57), though his formulation is not entirely clear (Pierre Larcher p.c May 2011). For shiyn Watson 1992: 74 suggests [ç] or [ɬ]. Daniels (2010) has a good summary of the debate over the last 50 years. Determining the phonetic value of shiyn is an issue in and of itself. Arguments have been made for its voiceless lateral status, [ɬ] (Rabin 1951: 33, Cantineau 1960: 62, Beeston 1962). Beeston (1984: 9) takes Sibawaih’s al-ћanak al-ʔaʕlaa to be limited to the hard palate, and hence would endorse

11

Beginning with general phonetics, phonetically, assuming that /j/ is, or is only [j], the palatal region has to accommodate not one, but three different [j] sounds (the basic one plus (13, 14)), as well as serve as a partial analogy for two other’s, (10, 20) above. However, the palatal region is typologically a somewhat underused region, probably for acoustic and perceptual reasons. It is implausible that such a logjam of sounds would cluster in the palatal region, particularly if this interpretation entails leaving the alveolar and alveopalatal region to the front empty of š-, č- and dž-like sounds. This general phonetic point can be underscored with a brief look at statistics readily available in the UPSID, the UCLA Phonological Segment Inventory Database, as available at the University of Frankfurt. This contains the segmental phonetic inventories of 450 languages. Searches can be defined by standard phonetic parameters, place, manner, voicing, various secondary articulations, etc. For the purposes of this paper two places of articulation were contrasted, the palatal vs. the alveo-palatal. The expectation is that the alveo-palatal region will show a greater degree of diversity than the palatal. This is measured by a simple statistic. Language tokens were classified as either having one segment at the palatal or alveopalatal position, or more than one (e.g. [š] and [dž], or only [š]). Segments with secondary articulations (e.g. labialized, laryngealized, nasalized, pre-nasalized), and special phonation types (breathiness, aspiration) were excluded from the count. The 2 x 2 Table 1 shows that there is a high degree of difference between the two places of articulation, with the palatal position having a far higher number of languages with only a single segment (usually [y]) than the alveopalatal. Table 1 Palatal vs. alveopalatal, single vs. more than one segment, UPSID sample palatal alveopalatal single segment 237 65 > 1 segment 167 223 p = .000, df = 1, chi sq = 89 In a very basic way this confirms the observation that differentiation in the palatal region is considerably restricted. Moreover, in the 450 language sample, only 105 have three or more palatal segments, and the suggested combination of Arabic palatal values, [ç, y, j] (see above) occurs in

the [y], [ç], [j] interpretation of the sounds. While Sibawaih’s description of ḍaad as a voiced emphatic lateral is fairly unambiguous, that of shiyn allows a lateral interpretation only inferentially (e.g. it is pronounced over an extended area of the mouth, which however, is compatible with the properties of alveopalatal [š] as well, see Sibawaih II: 467). The major problem with interpreting the sound as a lateral is that Sibawaih, an expert phonetician, does not describe it as a lateral, and does not pair it with the explicit lateral [ḷ]. Moreover, the lateral interpretation is difficult to relate to the extended sanctioned and non-sanctioned variants of shiyn (e.g. jiym like a shiyn).

12

only two languages in the entire sample.12 No language has as palatal values only [ç, y, j], the values of a “literal” reading of Sibawaih’s phonetic description. Looking further at the UPSID sample, no languages or ancestors of languages with which Arabic can be assumed to have been in close proximity in the late second/eighth century, Sibawaih’s era, have more than two palatal sounds and most have only [y]. On the other hand, alveopalatal sounds are common: Amharic: [š, ž, tš, dž, tš’] Tigre: [š, ž, tš, dž, tš’] Modern Persian: [š, ž, tš, dž] Kurdish: [š, ž, tš, dž] Neo-Aramaic (Iranian Azerbaijan): [š, ž, tš, dž] Soqotri: [š, ž, š’] Expanding the survey slightly, Neo-eastern Aramaic in general has [š, ž, tš, dž], while Western Neo-Aramaic has [š, tš] + either [dž] or [ž] on a dialectal basis (Jastrow 1997: 334, 348). Similarly Middle Persian has [š, tš, dž], with allophonic alternation (post-vocalic) of [dž ~ ž] (Boyce 1975:16). In the UPSID data base, of the languages ancestrally linked to the Arabic of the eighth century, only Modern Greek lacks alveopalatals (but also palatals). Both from a typological perspective and on the basis of an areal survey, it would therefore be unusual for Arabic to have three distinctive sounds at the palatal region in the combination which has been suggested. On the other hand, two alveopalatal sounds (e.g. [š], [dž] ~ [ž]) is not unusual, and alveopalatal sounds with values which can be interpreted from Sibawaih’s own description of Classical Arabic are common in languages (or their ancestors) Arabic is either closely related to genetically and/or in contact with. In Sibawaih-specific terms, it is noteworthy that after the hard palate (and as ever following Sibawaih’s progression of place, moving from the back towards the front of the mouth), Sibawaih hardly uses parts of the key passive articulator, the upper part of the mouth, to define further sounds. Instead, from the ḍaad onwards, the basic parameter defining the upper articulator is the upper teeth. For instance, the sound which comes “after” the jiym, the ḍaad, is described as being pronounced with the edge (ћaafa, “blade, front”?) of the tongue in the area adjacent (maa yaliyhu) to the molars (ʔaḍraas). There is no mention of constriction against the upper part of the mouth. On this point Sibawaih’s description lacks precision, as in articulatory terms the varying degrees of closure formed

12 Komi, a Finno-Ugric language with 9 palatal segments in all, and (Kwakiutl, North American Pacific coast) with 4. Tera (West Chadic) with 7 palatal segments has [y], [ç] and an implosive [>j]. Note that the palatal sound [y] is the most common palatal sound overall in the UPSID sample, with over 80% of all languages having it, so its presence in general is the rule rather than the exception.

13

by the active articulator, the tongue, are generally defined against the upper mouth, not the upper teeth.13 These two considerations suggest that the area of the al-ћanak al-ʔaʕlaa covered a much wider articulatory range than simply the hard palate, as is usually assumed by interpreters of Sibawaih (e.g. Cantineau 1960, Beeston 1962: 224, Watson 2002: 269). That is, al-ћanak al ʔaʕlaa was conceived by Sibawaih as extending from after the velar region (occupied by [k]) at least all the way to the alveopalatal region, if not beyond (see n. 13). This interpretation is compatible with the categorization of /y/, /š/ and /j/ as constituting a class. This interpretation is supported by Sibawaih’s designation of the laam as also being in the al-ћanak al-ʔaʕlaa area, very unlikely a lateral palatal [λ], a value which, to my knowledge, no one has interpreted for it.14 Thus, against a number of previous interpretations (e.g. Schaade 1911: 19, Cantineau 1960: 57), it is plausible to assign more than one place articulatory interpretation to the basic jiym. It could have been palatal (in the sense of universal articulatory phonetics), but could well have been alveopalatal as well, or post-palatal. All fit within the extended interpretation of al-ћanak al-ʔaʕlaa. Indeed, it could have had more than one value. Sibawaih classifies the jiym as a stop (shadiyd), but is not more specific than this, for instance giving no intimation as to whether it should be interpreted as a simple stop ([j]),15 or an affricate ([dž]). In fact, there is already strong evidence for jiym variation in Sibawaih’s description. Whatever the basic phonetic value of jiym was, (10) says that there was also a sanctioned variant that was [ž] (see (28)). If the basic jiym itself had the variants [dž], [j], two values of jiym widely attested among the dialects today16 would already have been in place in Sibawaih’s time and classified by Sibawaih as permissible pronunciations. Assuming this

13 Alternatively, if the al-ћanak al-ʔaʕlaa describes the top of the mouth from the hard palate all the way to the alveolar ridge, Sibawaih would have needed to differentiate the lingual sounds in another fashion, hence locating them relative to the teeth. 14 Much later in the thirteenth century, Sakkaki (p. 13) draws a representation of the mouth and where sounds are produced in it, in which the /l/ is placed much to the front of the /y, š, j/ class. His description (p. 12) is taken nearly verbatim from Sibawaih, so that one would conclude that the al-ћanak al-ʔaʕlaa accommodates an area all the way to where the stricture for the /l/ is made. The inference that the passive articulator of the jiym might have extended over a large area of the palate and alveopalatal region is also supported indirectly by Sibawaih’s description of the upper articulatory region of shiyn as extending all the way to where ṭ is pronounced (II: 462, 467, 473, see Schaade 1911: 74 n. 4). As noted, shiyn is of the same place category as the jiym. 15 That the palatal value would have been [j], and not [g] (as e.g. in Cairene today) is strongly suggested by Sibawaih’s chapter 525 in which he explicitly notes that the Persian “kaaf between kaaf and jiym” is realized as an Arabic jiym, because the Persian “kaaf between kaaf and jiym” is not among the sounds of the Bedouins, i.e. they had to shift the Persian [g] to something else, probably [j]. 16 [j] in various regions of Sudanic Arabic, while [dž] is generally the most widespread of the contemporary jiym variants.

14

perspective, in turn would potentially elucidate the variants (29) and (30) above. Both are correct, but reflect different dialectal variants.17 Assuming that Sibawaih would not have detected two [č] sounds, in (29) the most likely interpretation of the sound is probably [c], as this is the variant closest to /k/. In this case, both (29) and (30) use the “ka l-Y” to define a voiceless value. (29) additionally nuances a more backed variant, via the comparison to “kaaf” rather than “shiyn”. This would be the voiceless palatal stop corresponding to the voiced variant [j]. Still, it cannot be ruled out that (29) represents the second variant attested today, namely [ts]. To conclude this reading of Sibawaih, it is argued that among Sibawaih’s proscribed sounds (13-21) was a voiceless alveopalatal and/or palatal stop. Allowing that his description is vague in respect to the articulatory extent of the ‘palatal’ sounds, it is not implausible to assume that already in Sibawaih’s day the dialectal differentiation existed which is found today in the jiym. The basic jiym sound could have been /dž/ or /j/, it had a sanctioned variant [ž] (28), while the proscribed variant of the first would have been [č]. This model, moreover, resolves the ostensible incompatibility of (10) and (14), which appear to refer to a single value. In the formulation developed here, the relation is not symmetrical, “X like Y” does not imply “Y like X”, so there is no contradiction in Sibawaih’s formulation of the variants. Before returning to the question of kaškaša/kaskasa, a fundamental point needs to be dealt with, namely the extent to which Sibawaih’s phonetic description are tied to specific lexica, or are independent phonetic descriptions applicable to any word. This is a question which is pertinent in particular to the extended variants. In fact, both perspectives are applicable, depending on what variant is at issue. The imala ([imaala]), for instance, is at one and the same time an allophonic variant, and a variant which is always tied to particular words, namely those with an “alif” in the appropriate position for undergoing imala. A slightly different case pertains to the de-emphasized variants (17-19). Whereas the imala represents a conditioned split, the de-emphasized variants represent a merger. As Sibawaih represents them, however, these maintain their lexical integrity. Thus, whereas phonologically and phonetically, a “ṣaad like a siyn” is simply an [s], Sibawaih’s description continues to keep them lexically apart. Presumably [seef] in the sense of “summer” (as e.g. in Tlemcen, Dekkak 1979: 105), would be a “ṣiyn like a siyn”, whereas [sif] or [seef] in the sense of “sword” would simply be a “siyn”.18 In the case of (17-19), Sibawaih’s designations are etymo-phonetic.

17 For instance, if jiym had the values [dž], and [j], (28) and (29) would have had the respective voiceless values [č] and [c]. Inversely, if this is correct, the proscribed variants of jiym argue for variable sources of the basic jiym. 18 As for the taaʔ like a ṭaaʔ, one could, with Nassir (1993: 20), envisage a voiceless emphatic ṭaaʔ, since this in Sibawaih is voiced (majhuwr). This would follow the “like a = voicing” parameter. However, this interpretation makes little or no sense for the other two emphatic sounds. The only consistent reading is for lack of emphasis.

15

A different case, on the other hand, is represented by (21). Given the interpretation in this paper, by definition, the “sound between the kaaf and jiym” is a purely phonetic designation, at least as far as the only exemplification of the phenomenon in Sibawaih goes (chapter 525), since this is used for a non-Arabic sound. There is no Arabic etymological source to relate it to. A similar interpretation applies to (19/26). In this case, Sibawaih (II: 376) notes that Persian /p/ (baaʔ like a faaʔ) is realized in loanwords as /f/. A key question in this context is what perspective is intended in (13/29, 14/30). Clearly the analysis developed here requires the purely phonetic interpretation. It would be highly unusual, for instance, to interpret (13/22), as a [k] etymologically related to [dž], e.g. *kabal for jabal “mountain”. Indeed, this point is perhaps what led Cantineau (1960: 65) to introduce the spurious “kaaf like a jiym”. A “jiym like a kaaf” he would appear to suggest, is a rather absurd correspondence on etymological grounds, Cantineau adducing the nonsensical example *[džaafir] for [kaafir] (Cantineau’s, not Sibawaih’s example). Cantineau instead substituted the, for him, more reasonable “kaaf like a jiym” with a value [č]. In fact, (26-30) all require a purely phonetic interpretation. In any case, the crucial step in the current interpretation, is the recognition that Sibawaih thought both in etymological terms, and in purely phonological ones. The phonological perspective, in turn, allows Sibawaih’s extended variants to be interpreted in terms of phonological features, in particular the segment of his description “ka-l-Y” being shorthand for the voicing feature. From this perspective, the relevant (13, 14, 19, 21) are based on phonological parameters only. Before leaving this section commentary is necessary on (20), the kaaf that is between the jiym and kaaf. This is one sound which in the model in (25) would simply reproduce sounds already defined. Given the two potential original pronunciations of jiym, one can propose either of the following. (31) kaaf = alveopalatal/palatal, stop, voiceless = [č],[c] Alternatively, there is yet another palatal sound hidden here (e.g. [ç] or [ts]), but as the discussion above on the density of palatal sounds points out, this is unlikely. One could also think of the value [ts]. However, this formulation strictly speaking lies outside the model developed in (25), since the ‘kaaf’ is repeated twice, once as the name of the sound, and once as a point of orientation, in this case interpreted as a voicing parameter. One possibility is that (20) is in fact the same as (21). Sibawaih was inconsistent in his use of bayna “between”, and in chapter 525 the kaaf was placed first, in chapter 565 the jiym. However, given that there are only two sounds with the ‘bayna’ relation, (20) and (21), it may be that this question will remain forever open. 5. Why not kačkača?

16

It is now time to turn to the original question, the status of kaškaša and kaskasa. In this section I assume the correctness of the argument that Sibawaih included at least [č] among the eight proscribed variants, and possibly [ts] as well. Nothing of principle stands on the second assumption, however. If this is correct, however, a crucial question is why this wasn’t so formulated in chapter 504, why isn’t the 2FSG object variant designated as the “the jiym like a shiyn”. 5.1 Social status It can be suggested that the major reason has simply to do with the fact that [č] was among Sibawaih’s proscribed variants. Given this, Sibawaih would have had no interest in investigating its functionality in the language.19 Note that being a variant outside the basic 29 sounds does not automatically disqualify it from Sibawaih’s detailed attention. In fact, the imala, which is a highly legitimate variant of the non-imala alif and included among the six sanctioned variants, is the subject of the most detailed phonological observations in the entire Kitaab. 20 In contrast to [č], however, it is a sanctioned variant. The allomorphemic –č, on the one hand, the kashkasha and kaskasa variants, would have posed a problem. They must have been common enough in speech that Sibawaih took note of them (see 5.3 below). On the other, he could not have taken note of and expounded upon proscribed variants. Sibawaih’s solution, in this case, is to see (or hear) not [č], but rather [kiš], hence giving recognition to the variant, while keeping within the terms of both of his phonological precepts, and avoiding the domain of proscribed sounds. One indirect confirmation of this ‘avoidance’ interpretation can be found in what for Sibawaih is the sanctioned imala variant of long [aa]. Referring to today’s distribution of imala, all of them are purely [k] dialects. None of them underwent the split discussed in (32) below. What this suggests is that the language milieu Sibawaih worked in had as one dominant (leaving this term undefined) group of speakers who would not have had affrication and hence would not have been associated with the proscribed [č]. Following through on this reasoning, for Sibawaih it is not enough for a sound simply to be attested that it be taken up in his phonemic inventory of legitimate sounds. It must be attested among the right group of people. The imalized alif was used by the right group, while the “devoiced jiym” was not.

19 In describing Persian loanwords, the proscribed “in-between kaaf” describes a Persian, not an Arabic sound, and similarly the ‘baaʔ like a faaʔ’. 20 Indeed, it is an interesting issue in the history of Arabic sociolinguistics to ascertain why the imala pronunciation of alif didn’t become standardized, given its centrality in Sibawaih, the original arbitrer of Arabic grammar. In any case, the subsequent history of imala saw it spread out and/or survive in four geographical loci: in a northerly and easterly direction in northern Mesopotamia, including southern Turkey, and northern Syria, among the so-called qultu dialects as well as in Lebanon. Proceeding further west, it became established in eastern Libya and Maltese, it is well attested in Andalusia, and it has been recorded in a small pocket of Tiyaaha speakers (Shawarbah 2007) in the northern Negev. Today all of these regions are, broadly speaking, relic areas, and indeed, Andalusian Arabic died out some four centuries ago.

17

Related to the last point, two speculations regarding why the proscribed variants were proscribed can be made. It is noteworthy that of the eight proscribed variants, six are either obviously L2 variants, or explicitly other language sounds. (16-19) are all deemphasized variants, which can plausibly be attributed to L2 speakers (see. e.g Jaahiḍ, I: 40-2). (19) and (21) (and perhaps (20)) are explicitly Persian sounds, as Sibawaih describes in chapter 525. This leaves only the two sounds interpreted in this paper as [č] and probably, [c] or [ts], as in (13, 14). Generalizing across the entire class, it could be concluded that Sibawaih saw in all of the proscribed sounds something foreign or non-native. That is, as a first explanation, one basis of proscription could have been that Sibawaih judged the sounds according to the groups he heard uttering them. For whatever reason, speakers of [č] were, in his mind, something less than fully-fledged speakers of Arabic, whereas those who used imala were. Turning to a second alternative, referring only to (15) and (20), derived from Persian [g] and [p] respectively, it appears that Sibawaih proscribed clearly Persian variants from acceptable Arabic. [č], identified here as the ‘jiym like a shiyn’, was another Middle Persian sound.21 Sibawaih could thus have identified the [č] of the 2FSG –či with a Persian sound. However, to avoid having to allow ‘Persian’ sounds within his class of sanctioned variants, he interpreted –kiš/-kis instead, as argued above. 5.2 Phonology and orthography Besides the basic social factor, other potentially problematic factors relating to the recognition of a 2FSG –či suffix can be noted in this and the following section. As far as phonology goes, three potential problems are to be noted. First, given the morphological context in which Sibawaih discusses these morphemes, a purely morpho-phonemic variant which would be necessary to represent the variant could not be countenanced. Sibawaih gives no phonological conditions for the 2FSG variants, even if he does suggest certain communicatively-based constraints as to why speakers chose the forms they did on the basis of pausal position. It appears that Sibawaih conceptualizes speakers as picking their morphological variants from a pre-defined set of phonemes. This allows k + š or k + s, but not –ts or tš (= [č]), since these do not exist as members of a sanctioned set of phonemes, while choosing –t as a possible representation of the variants in this context is unlikely (t + š or t + s), since –t, in contrast to –k, is not appropriate as a morphemic value in this context. Alternatively, establishing a special morphophoneme for this context alone would be unprecedented in Sibawaih. Secondly, a broader approach to the variants would have involved dealing with allophonic conditioning. As noted above (see discussion around (5, 6)), the –č variant is phonologically conditioned, so one would need to ascertain the nature of the conditioned allophonic split. The basic rule is as follows.

21 [č], with the optional variant [z] post-vocalically (Boyce 1975: 16). Note that the inclusion of the “jiym like a shiyn” with the sound explicitly identified as Persian constitutes a further argument that it is indeed a [č], i.e. another of the Persian sounds.

18

(32) k > ts/č before a front vowel Otherwise, k kabiir or kibiir > čibiir/tsibiir “big” There is, to be sure, one instance where Sibawaih does describe a large-scale, regularly conditioned allophonic split, and that is the imala referred to above. In the word-internal imala which Sibawaih gives the greatest attention to, a long [aa] becomes [ia] in the context of an [i]. There are various conditions on this shift, but generally what Sibawaih describes is: (33) Imala in Sibawaih aa > ia / in context of [i] Otherwise [aa] (Kitaab II: 179) Formally this is identical to the conditioned “split” (allophony) of [k]. The question can be posed, if Sibawaih was aware of conditioned allophony, why wouldn’t he have heard –ts, -č in the present case? This leads to the third point. From a phonological and orthographic perspective the consonant split has a different status from the vocalic imala. Sibawaih was working in an era when the final orthographic inventory was still in the making. In the Kitaab are found a number of innovative orthographic signs, for instance those developed to describe pausal phenomena (II: 307-8). The imala itself was indicated by a special dagger alif written beneath the consonant (i.e. in the position of kasra) before the alif. What the orthographic reforms concentrated on were vowels and suprasegmental and contextual phenomena, not consonantal values. It is striking that Sibawaih himself, although he gave rather detailed information about consonantal variation both via assimilation and unconditioned variation, did not propose distinctive orthographic signs for them. Rather, he used the composite formulations described in 4.2 above. It would appear that in Sibawaih’s day there was a barrier to expanding the consonantal orthography. Thus, there was a consensus built around the 29 consonantal signs. Indeed, we see perhaps in Sibawaih’s treatment of the –kis and -kiš variants this predilection to think segmentally, with the already established –s and –š appended at the end of the morpheme. As Johnstone (1963) suggests, a comprehensive treatment of –ts, -č, dz/dž could have implied identifying new consonantal values, new phonemes. While this did eventually occur, it was not until after Sibawaih’s time that, for instance, Persian چ was introduced. Phonologically the allophonic k > k/č split is of a type whose conditioning Sibawaih did not deal with, namely a vowel causing a shift in consonantal value.22 The values which Sibawaih would have had to have dealt with are problematic from the perspective of

22 Consonants influencing vowels, on the other hand, are attested in the imala (e.g. emphatics prevent imala) and the influence of gutturals on the low vowel fatћa.

19

Arabic phonology. All of the component parts of the affricates exist independently in Arabic (breaking down the č into its component parts, t + š, for ease of reference): (34) ts = t + s tš = t + š dz = d + z dž = d + ž However, none of these could be recognized as sequences of two consonants, e.g. (35) bayt-u-tši = baytu-t-ši since this would violate basic syllable structure rules. (35), for instance, in pause gives, using a hyphen to separate phonemes, the following: (36) bayt-u-t-š# This, however, yields a disallowed CC# sequence. Word initially the change could be not be represented, since the words would begin with two consonants. (37) *d-ziddaam *t-šibiir etc. While the phonological context of (37) would require an alif al-waṣl, the phonetics does not support such. Clearly then, the affrication change does require the recognition of new consonantal phonemes, which as seen above, was not the direction which orthographic reform was taking. Over and above the issue of prestige, here would be a case where Sibawaih’s theoretical conceptualization of the issue would have led him away from perceiving forms which his model could not account for. 5.3 Why the pausal 2FSG? It can be asked why, if the supposed –č variant was general in the Arabic of Sibawaih’s day, he only singled out variants in pausal position, and only those in the 2FSG. These are two separate questions. Beginning with the second factor, it is clear that Sibawaih would have been drawn to the 2FSG in general by the striking variant –ši (1a).

20

Having considered this form in detail, he would have been sensitized to other allomorphs of the 2FSG. Beyond this structural, paradigmatic factor, it can be assumed that the 2FSG is a form of extremely high frequency in speech. Of all of the “k” and “k-like” sounds, the “k” of the 2SG is probably its most frequent realization. To lend at least prima facie support to this supposition, I counted in a spoken Hijazi-Arabic text of about 6,000 words, a conversation between two Saudi women (one of the textual bases of Owens et. al. 2009) all tokens of word-final /k/, including –ik, the 2FSG form in this dialect (-ki after a vowel). As can be seen, it is not any /k/ which constitutes the vast majority of word-final /k/’s in this sample, but rather the /k/ of the 2FSG –ik. (38) -ik: 91 = 63% Other –k: 5423 = 37% Total: 145 Conversation is the most typical ‘genre’ of language, so, even without claiming a scientific representativeness for this example, the statistics do give unequivocal support for the inference that Sibawaih’s interest in the 2FSG would have been reinforced by its high frequency in actual speech. Regarding Sibawaih’s concentration on the pausal forms, two alternatives can be considered. On the one hand, it could be that whatever variation was observed, even –č was indeed restricted to this morpheme. This would mean that the change from *k > č began in this morpheme.24 There is no way, of course, to confirm this, and this conclusion goes against the observations in 4.2 and 5.1 above, which argue that –č was a general variant, but a proscribed one. Alternatively, it can be observed that it is striking that Sibawaih in general was meticulously careful in documenting pausal phenomena (waqf) as opposed to context (waṣl) forms. A classic example of this, for instance, pertains to the realization of final vowels in pausal position (see II: 307-8). While this could simply be because there is more variation in pausal position, another aspect of the observation is that pausal position

23 23 of the non-2FSG k-final words were in one word token hinaak “there”. 24 As indeed Watson (1992: 77) argues in an analogous treatment. As already noted, she makes the interesting suggestion that kaškaša was –kç, as in certain North Yemeni dialects today. In this view, morphemic –kç subsequently spread to other phonological contexts, and to other geographical areas, where it eventually became the [č] variant as summarized in (5) above. As it stands this suggestion requires a great deal more justification. Leaving aside the observation that [kç] does not automatically entail or lead to [č], it is problematic that [-kç] is attested today only in this morphemic context, and that there are no dialects where a conditioned *k > kç is attested. This suggestion therefore entails, -kç > *kç > č, (morphophonemic [-kç], phonologically conditioned [kç], phonologically conditioned [č]) though the medial step is unattested both in the literature and among today’s dialects. It is simply a reconstructed medial step based on the assumption that conditioned [č] derives from –kç.

21

is also more susceptible to physical observation, particularly in the pre-electronic recorder era. Picking up allophonic variation in connected speech is more difficult.25 Furthermore, if indeed Sibawaih’s theoretical apparatus would have stopped him from picking up consonantal allomorphy, as suggested in 5.1, one distinguishing aspect of the 2FSG in all varieties where it is –č or –ts, is that it has this form, regardless of following vowel. That is, whereas elsewhere the allophonic split in (32) is operative,26 morphemic –tš/ts is invariable, even after back vowels: (39) abuu-č/abuu-ts (Johnstone 1963: 223) “your.F father” Thus, even if Sibawaih missed the general allophonic variation, he wouldn’t necessarily have missed the morphophonemic variants of the 2FSG. In short, the prominence of the 2FSG because of its frequency in speech, as well as the ability to observe pre-pausal forms more carefully than those in connected speech account for Sibawaih’s singling out the 2FSG in pausal position. 6. Sibawaih’s linguistic thinking Before moving to a reinterpretation of the history of the affricates, it can be noted that if the interpretation given here is correct, it further underlines the variegated theoretical complexity of Sibawaih’s thinking. Sibawaih has been described as a descriptivist, rather than a prescriptivist (Carter 1973). This is certainly correct if one should oppose Sibawaih’s linguistic practice to that of later grammarians, who, unlike Sibawaih, were working with a closed set of linguistic data. Sibawaih’s descriptivism, however, was not that of a living individual who blindly recorded all speech forms which he observed, a walking tape recorder as it were. Instead, all of his observations are filtered through and legitimized by an underlying theoretical justification. This involves in the first instance linguistic theory internal reasoning, but it can also involve, as suggested here, social evaluation based on groups of users. Note that from this perspective, Sibawaih’s theoretical constructs could, in certain instances, prevent him from analyzing phenomena which can be postulated to have existed in his lifetime, which he observed, but interpreted away, as it were, if they did not conform to his linguistic model.27 From an inverse perspective, the analysis of Sibawaih’s composite designation of sanctioned and proscribed sounds shows that western observers can be mislead by interpreting his observations through their own lenses. In this case, scholars have, quite understandably it may be said, interpreted a [g] as Sibawaih’s (13), “the jiym like a kaaf”.

25 Consider, for instance, the classic sociolinguistic research instruments of recording forms in wordlist format, i.e. each word bounded by pause, pausal forms. 26 Johnstone (1963: 211) points out that in lexically isolated cases, the affricated form can generalize, as in diič, dyuuč “cock, cocks” in Kuwaiti Arabic. 27 From a linguistic perspective, I consider this interpretation of Sibawaih yet another indication of the high degree of consistency in his linguistic thought. Theories elucidate, but they can also filter away objective reality.

22

This interpretation makes sense. However, the evidence from Persian loanwords, speaks decisively against postulating this value, as the [g] is explicity described as in (21). On the other hand, as soon as the composite forms are recognized as being built of a formula in which the likened-to phoneme is seen as a shorthand for defining the voicing parameter (see (25)), not only [č] and [č-like] sounds emerge from his description, but others as well which plausibly fit into Arabic variant forms. That is, as Carter (1972) first emphasized, understanding Sibawaih entails understanding the sometimes inimitable theoretical constructs and methodologies he built his grammar upon. 7. The history of the *k > č/c split revisited: Sibawaih and historical linguistic reconstruction The arguments given here obviously speak for a far earlier development of –č and perhaps –ts than in Holes’ (6 above) summary. Looking at a wider range of evidence from contemporary sources than has been done thus far reinforces this conclusion. Besides the palatalization summarized in (5) above, Seeger (2002) notes that conditioned [č], including the 2FSG –ič, occurs in Khorasan Arabic in eastern Iran, and Behnstedt and Woidich document the conditioned [č] “in non-/u/ contexts” (nicht -/u/- haltiger Umgebung”, 1984: 17 and p.c. May 2011) in the eastern part of the Sharqiyya (Egyptian Delta). Furthermore, Cantineau (1960: 66) writes that /č/ is found as an unconditioned reflex of /*k/ among Jewish speakers of Tlemcen, among Arabs in the Kabylie and in the area of Jijel (Djidjelli), fluuča < fluuka “boat”, and /ç/ in the mountains north of Tlemcen. Both Heath (2002: 139) and Vicente (2007: 131) note the [č] reflex of *k in northern Morocco (Jebli dialects).28 Cantineau considers the basis of these palatalized variants to be old, i.e. he relates them to the kashkasha and kaskasa of the Arabic grammars, without giving a detailed history of each variant. In various Levantine dialects (in particular, those referred to as the bəkuul group) as well as Sukhne in Syria (Cantineau 1960: 66), *k unconditionally has changed to č (Grotzfeld 1980: 174, Behnstedt 1997: 30). Between the conditioned and unconditioned reflexes, the historical development was undoubtedly, *k > conditioned *č, and thereafter, conditioned *č > č (unconditioned). The Arabs of Khorasan probably go back to the earliest Arabic incursions (late seventh/early eighth century) outside of the original Arabic homeland, while the [č] in the Sharqiyya can equally be interpreted as an old relic. As far as North Africa goes, it is plausible that the earliest wave of Arabic speakers in the seventh and eighth centuries brought the forms to Algeria and Morocco, the complete merger having already occurred in the Levant. The evidence from contemporary dialectology alone then allows a revision of Holes (1991) as well as of Owens (2009: 245). In the latter work two stages of reconstruction are postulated, one termed “pre-diasporic Arabic” which is broadly contemporary with Sibawaih’s grammar and marked with a single reconstruction star “*” the second a

28 Though Vicente ( 2005: 117) follows Heath (2002) in interpreting the palatalized kaf as a secondary development (see n. 29 for criticism of this interpretation).

23

reconstructed proto-form which antedates the pre-diasporic form by a to-be-determined period, and designated with a double star“**”. The evidence that both the conditioned k > č and the unconditioned change are found both within the Arabic pre-diasporic homeland, as well as outside of it, is a classic argument that the change occurred once, within the homeland, then spread outside of it. The k – č split thus already existed in pre-diasporic Arabic. When the origin of the proto split is to be situated is an open question. In any case, in the formulation of Owens (2009), one would have: (35) Arabic kaaf Proto Arabic: **k Pre-diasporic Arabic: *k ~ č (both conditioned and unconditioned; more precisely, *k > *č in conditioned context (27), thereafter *k ~ *č > č). The evidence from Sibawaih presented in this paper is compatible with the evidence of this reconstruction, with [č] already existing by Sibawaih’s times, and perhaps –ts as well.29 Reconstruction based on the comparative method using the rich source of Arabic dialectology, and Sibawaih’s description, here, as in many other cases, mutually reinforce and enhance our interpretation of Arabic language history. This paper addresses a further issue which has yet to be given independent prominence, namely the sociolinguistic status of the many variants which Sibawaih discusses in the course of his Kitaab, and how his evaluation of them might be crucial in his own interpretation of the ʕArabiyya. It is argued here that his own interpretation of the [č] as a proscribed variant led to his reinterpretation of the 2FSG object suffix as a –kiš (or –kis) rather than what it was, a –č. References Bakalla, M. 1984. Arabic culture, through its language and literature. London: Kegan Paul International. Barth, Johan. 1911. “Das arabische š-Suffix 2.P.Sg. fem.”. Wiener Zeitschrift für die Kunde des Morgenlandes. 25:281-9. 25. Beeston, Frederick. 1962. „Arabian sibilants“. Journal of Semitic Studies. 7: 222-33. --- 1984. Sabaic Grammar. Journal of Semitic Studies: University of Manchester. Behnstedt, Peter. 1997. Sprachatlas von Syrien. Wiesbaden: Harrassowitz.

29 The results of this study, as well as others of its kind, argue very strongly against the approach of many linguists and Arabicists who assume a homogeneous and idealized Classical Arabic as the source of all later forms of Arabic. Heath (2002: 139) for instance says that /č/ did not exist in Classical Arabic. The evidence discussed here shows that such a formulation begs the question of what Classical Arabic can be assumed to be or have been.

24

Behnstedt, Peter and Manfred Woidich. 1985. Die ägyptisch-arabischen Dialekte. Wiesbaden: Reichert. Boyce, Mary. 1975. A Reader in manichaean Middle Persian and Parthian. Leiden: Brill. Cantineau, Jean. 1960. Études de linguistique Arabes. Paris: Klincksieck.

Carter, Michael. 1972. “Twenty dirhams in the Kitaab of Sibawaih”. BSOAS 35: 485-96.

--- 1973. “An Arab grammarian of the eighth century A. D”. Journal of the American Oriental Society 93: 146-67. Daniels, Peter. 2010. “Arabian sibilants again“. Paper presented at Meeting of American Oriental Society. Dekkak, M. 1979. Sex dialect in Tlemcan. PhD diss., London Univ. Fischer, Wolfdietrich. 1956. “k → š in den südlichen semitischen Sprachen (kaškaša)“. Münchener Studien zur Sprachwissenschaft. 8: 25-38. Grotzfeld, Heinz. 1980 „Das syrisch-palästinensischen Raum“. Handbuch der arabischen Dialekte, ed. by W. Fischer and O. Jastrow, 174-91. Wiesbaden: Harrassowitz. Heath, Jeffrey. 2002. Jewish and Muslim dialects of Moroccan Arabic. London: Routledge. Hetzron, Robert. 1997. The Semitic languages. London: Kegan Paul. Holes, Clive. 1989. Towards a dialect geography of Oman. BSOAS 7:446–462.

—1991. “Kashkasha and the fronting and affricization of the velar stops revisited: a contribution to the historical phonology of the peninsular Arabic dialects.” Semitic Studies in Honor of Wolf Leslau, ed. by Alan Kaye, 652-78. Wiesbaden: Harrassowitz. Ibn Manḍur, Abu Fadl. Lisaan al-ʕArab. Beirut: Dar Al-Fikr. Ibn Mujahid. Kitaab al-sabʕa fiy al-Qiraaʔaat. Ed. by Shwaqi Dayf. Cairo: Dar al-Ma’arif.

Ingham, Bruce. 1994. Najdi Arabic, Central Arabian. Amsterdam: Benjamins.

Al- Jaahiḍ, Abu ‘Uthman. Al-Bayaan wa l-tabyiyn. Beirut: Dar al-Kutub al-‘Ilmiyya. Jastrow, Otto. 1997. “The Neo-Aramaic languages”. In Robert Hetzron (ed.), 334-77.

25

Johnstone Thomas. M. 1963. “The Affrication of ‘kaaf’ and ‘gaaf’ in the Arabic dialects of the Arabian peninsula”. Journal of Semitic Studies. 8: 210-26. Kaye, Alan. 1972. “Arabic /žiim/: a synchronic and diachronic study”. Linguistics 79: 31-72. Lane, Edward. 1980 (1877). Arabic-English lexicon. Beirut: Libraire de Liban. Nassir, Abdulmunim. 1993. Sibawayh the phonologist. London: Kegan Paul. Owens, Jonathan. 2009. A Linguistic history of Arabic. Oxford: OUP. Owens, Jonathan, Robin Dodsworth, Trent Rockwood. 2009. „Subject-verb order in spoken Arabic: Morpholexical and event-based factors“. Language Variation and Change, 21: 39-67. Rabin, Chaim. 1951. Ancient West Arabian. London: Taylor’s Foreign Press. Al-Sakkaaki, Muħammad. Miftaaћ al-ʕuluwm. Ed. By Na’i Zarzur. Beirut: Dar al-Kutub al-‘Alamiyya, 1984.

Salemann, C. 1930. A Middle Persian grammar. Bombay: British India Press.

Sara, Solomon. 2013. The Classical Arabic lexicographical tradition. In J. Owens ed., The Oxford handbook of Arabic Linguistics. Oxford: OUP.

Schaade, A. 1911. Sibawaihi’s Lautlehre. Leiden: Brill.

Seeger, Ulrich. 2002. “Zwei Texte im Dialekt der Araber von Chorasan”.“Sprich doch mit deinen Knechten Aramäisch, wir verstehen es! (Festschrift for Otto Jastrow), ed. by Hartmut Bobzin and Werner Arnold, 629-46. Wiesbaden: Harrassowitz.

Shawarbah, Musa. 2007. The Dialect of the Tiyaaha. PhD thesis, Hebrew University.

Sibawaih, Uthman. Al-Kitaab. Ed. By H. Derenbourg. Hildesheim: Olms, 1970.

Al-Suyuwṭiy, Jalaal al-Diyn. Al-Muzhir fiy ʕuluwm al-luɣa wa ʔanwaaʕuhaa. www.kotobarabia.com

Vicente, Angeles. 2005. Ceuta: une ville entre deux langues. Paris: L’Harmattan.

--- 2007. “Two cases of Moroccan Arabic in the diaspora”. Arabic in the City, ed. by C. Miller, E. Al-Wer, D. Caubet and J. Watson, 123-44. London: Routledge.

Watson, Janet. 1992. “Kashkasha with reference to modern Yemeni dialects”. ZAL 24. 60-81.

--- 2002. The phonology and morphology of Arabic. Oxford: OUP.

Woidich, Manfred, and Liesbeth Zack. 2009. The g/ǧ-question in Egyptian Arabic revisited. In Arabic dialectology. In honour of Clive Holes on the occasion of his sixtieth Birthday, ed. Enam Al-Wer and Rudolf de Jong, 41–60. Leiden: Brill.

Website

26

http://web.phonetik.uni-frankfurt.de/upsid.html (UPSID)