You are on page 1of 16

Linguistic Society of America

The Arabic Koine Author(s): Charles A. Ferguson Source: Language, Vol. 35, No. 4 (Oct. - Dec., 1959), pp. 616-630 Published by: Linguistic Society of America Stable URL: . Accessed: 18/09/2011 17:16
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.



Centerfor Applied Linguistics 0. It has usually been assumed' that the modern Arabic dialects are on the whole lineal descendants of Classical Arabic or of a variety very similar to this.2 Stated differently, this assumption holds that apart from borrowings and innovations the linguistic substance of the modern dialects is a direct continuation of an earlier stage of Arabic substantially identical with the Classical Arabic of the grammarians, with only a few isolated instances in which one or more of the modern dialects seem to preserve archaisms antedating the codification of the Classical language. Until clear contradictory evidence is produced, this assumption will have to stand as the most reasonable working hypothesis. The purpose of the present study is to offer one important refinement to this hypothesis, namely that most modern Arabic dialects descend from the earlier language through a form of Arabic, called here the koine,3 which was not identical with any of the earlier dialects and which differed in many significant respects from Classical Arabic but was used side by side with the Classical language during early centuries of the Muslim era. It is well known that there were great dialect differences in Arabia in preIslamic times, and it is widely accepted4 that the Classical language, the 'Arabi1 Three linguists read a draft of this study: Haim Blanc, Jacqueline Wei, and Joseph Van Campen. All made helpful suggestions about the substance and the presentation, many of which I followed. The responsibility for all facts and opinions, however, remains mine. 2 Cf. C. Brockelmann, Semitische Sprachwissenschaft 41-4 (Berlin-Leipzig, 1916); C. Bergstrasser, Einfuhrung in die semitischen Sprachen 156 (Munich, 1928); D. L. O'Leary, Comparativegrammarof the Semitic languages 16-20 (London, 1923); and J. H. Kramers, De semietische Talen 47-8 (Leiden, 1949). 3 This thesis is not new: the term 'koine' has been used before in approximately this sense for the history of the Arabic language; cf. AIEO 14.7 (1956), Enc. of Islam2 1.574 Col. 1, line 19 ff. (Leiden, 1957). This essay is, however, the first attempt known to me to establish the thesis by a full linguistic argument. Two diachronic studies of Arabic have appeared recently which attempt to sketch the phonological developments from the koine to a modern dialect, in one case that of Cairo, in the other Jerusalem: H. Birkeland, Growthand structure of the Egyptian Arabic dialect (Oslo, 1952), and I. Garbell, Remarks on the historical phonology of an Eastern Mediterranean Arabic dialect, Word14.303-45 (1958).My views on Birkeland's study were expressed in a review in Lg. 30.558-64 (1954); the writing of that review was the stimulus for putting the present article on paper. The Garbell study appeared after the article was completed, but footnote references to it have been added. Both studies are structuralist in approach and both present valuable material, the Garbell study being especially rich in historical and dialectal detail. In my view both studies err on the side of placing specific phonological changes at too recent a period and in too localized an area. Many of the changes they attribute separately and at a late date to Egyptian and Eastern Mediterranean Arabic are likely to have occurred much earlier, often in the formation of the koine itself. Further investigation is clearly required to determine the relative merits of these views. 4 C. Rabin, The beginnings of Classical Arabic, Studia Islamica 4.19-38 (1955); id., Enc. of Islam2 1.564-7; R. Blach&re,Histoire de la litterature arabe 66-82 (Paris, 1952). Certain 616



yyah of the grammarians, was based on a standard poetic language not necessarily identical with any one dialect, but in oral use by poets and orators of many dialects and known to us fairly directly from the remnants of pre-Islamic poetry and from the Qur'an. After the 'Arabiyyah became accepted throughout the world of Islam and was explicitly codified in the works of the grammarians, it remained essentially unchanged in phonology and morphology until the present time, when it is still accepted as the norm both for written and for formal spoken Arabic.6 During the centuries, however, spoken Arabic, even at the time of Muhammad quite different from the 'Arabiyyah in many parts of Arabia, diverged increasingly from this standard. It is a priori quite likely that some dialect differences in Arabic today continue the early dialect differences mentioned above, but on the whole there is little evidence of such continuation on any large scale. It is the thesis of this article (1) that a relatively homogeneous koine, not based on the dialect of a single center, developed as a conversational form of Arabic and was spread over most of the Islamic world in the first centuries of the Muslim era, (2) that this koine existed side by side with the 'Arabiyyah although it was rarely used for written purposes, and (3) that most modern dialects, especially those outside Arabia, are continuations of this koine, so that their differences are chiefly borrowings or innovations which took place subsequent to the spread of the koine. The situation is thus partly analogous to the frequently cited case of Greek,6 in which the modern dialects are not direct descendants of the early dialects but derive from the koine, and the present dialect differences are generally innovations which took place subsequent to the spread of the koine. The major differences between the two cases are the persistence of Classical Arabic virtually unchanged through the entire time span of this series of developments, and the fact that the Greek koine was based to a large extent on the spoken Greek of a single center. It seems highly probable that the beginnings of the koine already existed before the great expansion of Arabic with the spread of Islam, but it also seems probable that the full development of the koine coincided with this expansion, which brought about mingling of the original dialects, caused large numbers of speakers of other languages to adopt Arabic, and required intercommunication throughout
specific explanations of the origin of the 'Arabiyyah (e.g. the Meccan or Qurayshi dialect) can be rejected; others (e.g. from the dominant dialect of the Kinda confederacy, the dialect of HIra) can be neither accepted nor rejected until more evidence is available. Unfortunately, the term 'koine' (or 'poetic koine') has also been used to refer to the preIslamic standard which was the basis of the 'Arabiyyah. Rabin, Beginnings, has pointed out the inappropriateness of this term for a language apparently used little if at all for ordinary conversation. If the term 'koine' becomes generally accepted in this meaning, the Arabic koine which is the subject of the present article will have to be called Koine II or something of the sort to differentiate them. Cf. Enc. of Islam2 1.574, where both uses of koine occur in the same paragraph. 5For statements of the present relationship between Classical Arabic and the dialects, see A. F. Sultanov, National language and script reform in the countries of the Arab East,
Akademiku V. A. Gordlevskomu ... 252-74 (Moscow, 1953) [in Russian]; A. Chejne, The role

of Arabic in present-day Arab Society, The Islamic literature10.4:15-54 (April, 1958); C. A. Ferguson, Diglossia, Word 15.325-40 (1959).
6 Cf. A. Meillet,

Aperfu d'une histoire de la langue grecque 259-64 (Paris, 1913).



the whole world of Islam. Also, it seems highly probable that the koine developed chiefly in the cities and in the armies and that its spread coincided roughly with the spread of urban Arabo-Islamic culture. In some cases small pockets of spoken Arabic doubtless remained relatively unaffected by the koine, and in certain instances even fairly large-sized migrations (e.g. Bani Hilal in North Africa) established in certain areas varieties of Arabic quite distinct from the main mass of koine-based Arabic dialects. Generally, modern beduin dialects are not descended directly from the koine, and some sedentary dialects have been 'beduinized' by the incorporation of certain elements. But all these constitute only a small fraction of the total Arabic speech community: it is the dialects of the overwhelming majority-chiefly the sedentary populations outside the Peninsula-which are under discussion in this study. It must be noted that no attempt is made here to date the formation of the koine with any precision or to locate its boundaries at any period; the validity of the study does not depend on any historical verification of the TIME or PLACE at which the koine existed, much though historical documentation of the FACT of the koine's existence is welcome as a confirmation of the thesis. The basic argument is very simple. The modern dialects agree with one another as against Classical Arabic in a striking number of features. If these features can plausibly be interpreted as a natural development or 'drift' which continues early trends (e.g. loss of glottal stop, reduction of inflectional categories, increase of symmetry in the grammar) the agreement among the dialects as against Classical proves nothing, because it is perfectly possible that parallel changes of this sort could have taken place independently in the various dialects. But if some of these features are complicated, systemically isolated items difficult to account for by drift, and if there is a sizable number of such features, then the agreement among the dialects as against Classical shows that these dialects come from a common, non-Classical source. Once again it must be noted that no assumption is made here that all the features developed or became widespread at the same TIME (several may have appeared very early, before the full development of the koine), but the FACT of their existence is sufficient for the argument. It may even be true that a few of the features of the koine continued an original state while the corresponding forms of Classical were the innovations. Fourteen features in which modern dialects agree as against the 'Arabiyyah will be described here. Each 'feature' is in fact a constellation of minimum linguistic elements which, taken together, seem likely to have functioned as a unit in the historical development of Arabic. Most of the features are morphological, but three lexical features and one phonological feature are included. The features selected for description are those which seem most convincing to me. Many other features could be adduced as possible supporting evidence which are not as fully satisfying for the basic argument as the ones chosen. On the other hand, once the thesis is accepted, we may proceed with somewhat more confidence to a reconstruction of the koine, making judicious use of features of agreement which were not the basis of the original argument. Subsequent studies will do this, offering a fairly full outline of the sounds and forms of the koine so far as they can be inferred from the modern dialects or other evidence.



The assumption is made here that the koine came into existence through a complex process of mutual borrowing and leveling among various dialects and not as a result of diffusion from a single source. The reason for making this assumption is that the history of the Arabic-speaking world shows no evidence of long-continued linguistic predominance of a single center of prestige and communication. Great respect has always been accorded to beduin Arabic as opposed to the language of settled populations; since the 2nd century of the Muslim era some lip-service has been paid to the superiority of the Meccan or the Qurayshi dialect; and a great deal of discussion has always taken place about which spoken variety is the 'best' kind of Arabic, i.e. nearest to the Classical. But there is no evidence of conscious or unconscious normative influence on the whole spoken language from a single center over a long period of time. In this respect the modern Arab world remains unchanged. No variety of spoken Arabic is accepted as the norm or standard for the whole speech community, although of course important centers of prestige and communication may exert a considerable linguistic influence over a certain region (e.g. Cairo Arabic in Egypt). 1. Before listing the features themselves it may be useful to give some indications of the nature of the drift of Arabic. It is assumed here that a language or group of related languages (i.e. continuations of a single language) often shows a 'drift' or general direction of development consisting of a number of specific trends more or less integrated into a total pattern. Arabic is a good example of this: certain trends continue or recur throughout the history of the Arabic language. Several of these trends are found also in other Semitic languages and may be regarded as a part of the drift of the Semitic family as a whole; others are more particularly Arabic. The phonological drift of Arabic includes the following trends: loss of glottal stop, loss of final -h, increase in number and symmetry of 'emphatic' consonants, ay > e and aw > o, loss of unstressed short i and u (or phonemes derived from them) in open syllables, shortening of unstressed long vowels, and vowel assimilation (e.g. CaCiC > CiCiC). Some of these phonological trends have had morphological consequences; thus, loss of final glottal stop leads to merger of final-hamzah and final-weak verbs. Some of the more specialized phonological developments of particular dialects have had even more far-reaching morphological consequences; thus, merger of /a/ and /i/ in Maghribi leads to disappearance of the active/passive distinction in participles of derivative verbs. An important trend on the border of phonology and morphology has been the development of a difference between pause forms and context forms of words and then the generalization of pause forms to all positions.7 Pause forms are generally but not always shorter than the corresponding context forms, which in earlier Arabic often have final inflectional material lacking in the pause forms. Morphological trends include development of suffix alternants conditioned by the consonantal or vocalic nature of the preceding phoneme, reduction in the number of inflectional categories, and re-forming of nontriconsonantal roots into the triconsonantal norm.
7 This trend is treated at some length in Birkeland, Growth and structure, as well as in his earlier Altarabische Pausalformen (Oslo, 1940). Cf. also Lg. 30.560, 563 (1954).



All these trends appeared very early in the history of Arabic and are still in force today; they have worked at varying speeds and with great variation in detail at different times and places. But they tend to continue or to recur, and they are generally irreversible. Accordingly, features of dialect agreement as against Classical which seem to fit in with or exemplify these trends will not be used here as direct evidence for the existence of the koine. 2. The first two features to be described here are rather general in nature, i.e. they cut across major word classes. One is the special pattern of loss of the dual, the other is the unexpected presence of short i vowels in certain affixes in which Classical Arabic has short a.
I. Loss oF THE DUAL. Gradual loss of dual forms is a familiar story in the

history of Indo-European and Semitic languages, while good examples of the formation of a new dual are hard to find in the history of these languages. Also, the reduction of inflectional categories is part of the drift of Arabic. Accordingly, the absence of dual forms in the dialects in contrast with their presence in the 'Arabiyyah is not in itself an argument for our thesis. One might expect that all dialects would show fewer dual forms than Classical, with regional variation in the degree of retention and in the exact details. This is the case, however, only in the dual of nouns, which is regular and may be formed from almost any singular noun in Classical Arabic but shows considerable variation in the dialects. Thus, Moroccan has special dual forms only for nouns of measure and a few othersand even these are probably Classicisms-while Syrian has a regular and highly productive dual of nouns. But there are two striking elements of agreement in the details of the loss of the dual in the dialects. One is that the dual forms of adjectives, pronouns, and verbs have disappeared everywhere without a trace.8 If this were a natural development or a part of the drift of the language one would expect the same kind of differences as those found in the dual of nouns, with dialects varying in the amount of retention and with some dialects preserving some instances and other dialects preserving others. Such an argument from silence, however, is not completely convincing. The other element is the nature of the concord with the dual. In Classical Arabic a verb, pronoun, or adjective which refers to a preceding dual noun is also dual. On the other hand, in Classical Arabic as well as in the dialects a verb, pronoun, or adjective which refers to a preceding plural noun is either plural or feminine singular, the plural generally being used if the noun refers to human beings, the feminine singular if it refers to animals or objects. Accordingly, with the gradual disappearance of the dual forms one would expect that the same kind of concord would be found with dual nouns as with plural nouns. But this is not the case: the dual noun wherever it occurs in the dialects requires plural, not feminine singular agreement, whether it refers to persons or to things.
8 It seems quite likely that the dual forms of verbs and pronouns as well as the dual agreement of adjectives are analogical extensions in 'Ur-arabisch' from the dual form of the noun, which was presumably present in Proto-Semitic. But these additional duals were apparently well established, although with regional variations in detail, in the dialects of Arabia at the time of Muhammad, and were lost again in the development of the modern dialects.



Thus Classical9 baytani kabirani 'two large houses' and buyuitunkabtratun (f sg) 'large houses' contrasts with Syrian betrn kb&r(pl) 'two large houses' and byiit kbire (f sg) 'large houses. These two details in the development of the dual category in the dialects seem a good piece of evidence for a common non-Classical origin: complete loss of the concord with dual in the adjective, the pronoun, and the verb; obligatory PLURAL dual nouns. II. TALTALAH. number of inflectional affixes which contain /a/ in Classical A Arabic have in modern dialects the reflexes of /i/ (including zero) instead of those of /a/ whenever the dialect in question has retained the a-i contrast. The use of /i/ for /a/ in some of these affixes (the prefixes of the imperfect) was noted as dialectal even by the grammarians, who gave to this phenomenon, regarded as If a defect, the name taltalah.l10 the modern dialects to a considerable extent continued the earlier dialects one would expect either that some dialects should have reflexes of /a/ and others those of /i/, or that there should be variation in this respect in single dialects. Instead, all dialects outside Arabia seem to have the reflexes of /i/ instead of those of /a/ in the following affixes:
subject prefixes of the imperfect:loa

ta- in taftaHu 'you open', tiftaH ya- in yaftaHu 'he opens', yiftaH ta- in taftaHu 'she opens', tiftaH na- in naftaHu 'we open', niftaH intransitivizing prefix ta-: ta'allama 'he learned', t'allam yata'allamu 'he learns', yit'allam context form of the feminine suffix -at-:
gurfatuka 'your room', gurftak

gurfatuhum 'their room', gurfithum connective prefix wa- 'and': walbintu 'and the girl', wilbint alternant -al- of the definite article prefix after min 'from':11 mina Ibayti 'from the house', mnilbet
9 The examples in this article are kept to the minimum necessary to illustrate the points, and the same words are used repeatedly to illustrate different points in order that the nonArabist may be able to follow the argument without the burden of too many unfamiliar items to deal with. The usual order of citation will be: Classical form, gloss, colloquial form, with the two Arabic forms in commensurate phonemic notations. Unless otherwise specified, the colloquial items are in a slightly normalized Syrian (= Garbell's Eastern Mediterranean Arabic); they are usually Jerusalem Arabic, but where this is aberrant in the Syrian area a more typical form is supplied. Although a procedure of this kind has obvious pitfalls, it is hoped that no change has been made which affects the argument. Other procedures were rejected because the points to be made would have been obscured in a mass of irrelevant details. 10 Cf. Rabin, Ancient West Arabian 61-3, map 60 (London, 1951). lOa Cf. Garbell, Remarks 312. " Cf. Rabin, Ancient West Arabian 71-3, map 72; Garbell, Remarks 334.



prefixed 'a- of the 'plurals of paucity':

'alsunun 'tongues', 'ilsun

'argifatun 'loaves', 'irgfe preformative 'a- of the second-person pronouns: 'anta 'you (m sg)', 'inte It could be suggested that this /i/ for /a/ is either a general phonetic change or a morphologically conditioned change of some kind affecting all affixes. In either case the dialect agreement would not then be an argument for the assumption of a koine. But initial Ca- remains in the vast majority of instances apart from presumed vowel assimilations of the type CaCi > CiCi (examples: kataba 'he wrote', katab; katabtu 'I wrote', katabt;fahima 'he understood', fihim; fahimutu 'I understood', fhimt). Also, there are some affixes with /a/ in Classical which have the expected /a/ reflexes in the dialects. For example, the first-person subject prefix of the imperfect 'a- has generally remained, except where it has been replaced by the analogical n- of Maghribi.'2 Also, the ma- and mi- prefixes of Classical place, time, and instrument nouns have survived in the dialects, and in fact even with an increase in the proportion of ma- to mi-. Also, the prefix la- 'to, for' of Classical has la- or even ld- allomorphs in various dialects. li'

3. The verb system of the modern dialects generally continues the earlier system as represented by Classical Arabic, with most of the changes exemplifying the trends of generalizing pause forms and reducing inflectional categories. Even some of the differences between Classical Arabic and the dialects which seem at first sight to be striking agreements for my thesis turn out on closer inspection to be natural consequences of these two trends. As an example I may cite the agreement which the dialects show in having a long stem vowel in the imperative of middle-weak verbs (cf. Classical qum, sir, xaf, colloquial qum, s?r, xcf). However, the loss of the indicative/subjunctive/jussive contrasts in the imperfect and the generalization of the most frequent pause form of the stem would call for exactly this development. Even such widespread phenomena as the disappearance of the passive and of Form-IV verbs could be attributed in large part to phonological trends (vowel assimilation, loss of glottal stop). But there are at least three differences which cannot be accounted for in this way: the loss of finalwaw verbs, the re-formation of geminate verbs, and the development of a verb suffix -1- 'to, for'. III. Loss OF FINAL-WJW VERBS. Classical Arabic has five kinds of primary verbs with a 'weak' final root-consonant (e.g. rama, gaza, sa'a, baqiya, saruwa). Of these the first two types are by far the most common, followed in frequency by the fourth; the third is relatively rare and the fifth so rare as to be negligible. That the fifth type (saruwa) should vanish is not surprising, but that the second type (gaza) should disappear in the dialects is a significant feature of dialect
12 In some parts of the Syrian dialect area (e.g. Damascus, most of Lebanon) the 'aprefix has followed the analogy of the other prefixes of the imperfect; but this is clearly a subsequent development, since the 'a- remains in Iraq, Egypt, and much of the Syrian area, having even spread to Form II and III verbs where the prefix was 'u- in Classical. Cf. MdlangesMassignon 1.312-3 (Damascus, 1956).



agreement as against Classical. The only apparent vestiges are either obvious Classicisms such as the politeness formula 'arj0k 'please' [= 'I beg you'] or marginal phenomena such as the Tunisian baby word iHbu 'he goes on all fours'. Otherwise, the verbs of the second type have merged completely with those of the first type (rama). Examples: rami 'he threw', rama ramaytu 'I threw', ramet yarmz 'he throws', yirmi gaza 'he raided', gaza gazawtu 'I raided', gazet (expected *gazot) yagzu 'he raids', yigzi (expected *yigzu)
IV. RE-FORMATION GEMINATE OF VERBS. In all varieties of Arabic the verbs

of which the second and third root-consonants are identical ('geminate roots') have certain forms which differ from those of verbs with 'sound' roots. Classical Arabic showed some fluctuation in the imperfect (and imperative) of these geminate verbs, and it is of some general interest to note which of the alternative forms survived in the dialects, but this does not really advance the argument here. Also, in the dialects the active participle in verbs of this type is CaCiC, contrasting with the CaCC of the Classical language, but this can be regarded as a natural development since (a) all other primary verb types have CaCiC in Classical and (b) the syllable type CVC is extremely rare in Classical Arabic, occurring chiefly in these very participles. One difference, however, between Classical Arabic and the dialects in the inflection of these verbs is highly significant: in Classical the normal first and second person forms of the perfect are exactly analogous to those of sound verbs (e.g. Halaltu 'I untied': katabtu 'I wrote'; masistu 'I touched': 'amiltu 'I worked'); in the dialects the corresponding forms are similar to final-weak forms (e.g. Hallet 'I untied', masset 'I touched'; cf. katabt, ramet) and are in fact identical with the first- and second-person perfect forms of final-weak verbs of Form II, i.e. with doubled second root-consonant: Hallet 'I untied': xallet 'I let' (cf. the third-person forms Hall o/H 11, xalla V/x I y/w). This formation is mentioned by the grammarians as one of various dialect formations but is regarded as nonstandard in Classical Arabic. It is the only formation in the modern non-Arabian dialects, which seem to have no vestiges of the standard Classical forms.
V. THE VERBSUFFIX-1- 'TO,FOR'. In Classical Arabic there is a relational prefix

li- 'to, for' (with pronoun endings lz 'to me', laka 'to you (m sg)', lahu 'to him', etc.) as well as an independent preposition 'ila 'to, toward' (with pronoun endings 'ilayya 'to me', 'ilayka 'to you', etc.). In the modern dialects these two items have been combined and re-formed in a variety of ways. One feature on which the modern dialects agree, however, is that the reflex of li- with pronoun ending is added directly to verbs as a suffix -1-.There are differences of detail; for example, Syrian dialects show shortening of any long vowel in the final syllable of the verb to which the -1- is attached and have special 'heavier' alternants of the suffix such as -ill-, -all- after certain verb forms; Iraqi has shortening in the case of only one verb, gal 'say'; Egyptian may add to the same verb at the same time both



-1- with pronoun ending and another pronoun ending as direct object. But all agree in having the -I- suffix as an integral part of the verb phonologically and morphologically.l3 There seems to be no trace of this in Classical Arabic, where the lila- plus pronoun ending is an independent word, in no way attached to the verb and often separated from it by several intervening words.l4 4. The morphology and syntax of the cardinal numbers are fairly complicated in Classical Arabic and, on the whole, considerably simpler in the dialects. So far as this simplicity in the dialects is part of the general loss of inflections it can be disregarded for our present purpose, since such change is part of the drift of Arabic. But two points are relevant to my thesis: (a) certain features of the syntax of the numbers 3-10, and (b) the form of the numbers 11-19, in particular the presence of an unexpected emphatic /t/.
VI. CARDINALNUMBERS3-10. In Classical Arabic the numbers 3-10 occur

in two forms-a form with the feminine ending -ah -at- which is used with MASCULINE nouns, and an apparently masculine form without the feminine end'

ing which is used with FEMININEnouns. Examples:

baytun (m) 'house' kam baytan 'how many houses?' xamsatu buyuitin'five houses' xamsatun 'five' kam gurfatan 'how many rooms?' gurfatun (f) 'room' xamsu gurafin 'five rooms' xamsun 'five' It is hardly surprising that this strange feature of the Semitic number system should tend to disappear, since it has no support elsewhere in the grammatical structure-it is much more surprising that it exists in the first place; but the exact nature of the disappearance is of interest. In the modern dialects the long form (the form with the feminine ending) is used when there is no following noun at all,'5 and the shorter, apparently masculine form is used before any noun, regardless of the gender. Examples: bet (m) 'house' xams byut 'five houses' gurfe (f) 'room' xams guraf 'five rooms' kam bet 'how many houses?' xamse 'five' kamgurfe 'how many rooms?' xamse 'five'

Since this is only one of the various possible ways of eliminating the gender polarity of 3-10, the fact that all dialects agree in this point is significant in itself.
13 The suffixal nature of the -I- is shown incontrovertibly by the position of word stress, the lengthening of short vowel before -I-, and the existence of allomorphs of -I- conditioned by the preceding morpheme. Cf. 4drabu 'they hit', /arabiukum 'they hit you (p1)',darabdlak 'they hit for you'. 14 An early example of this construction is cited in G. Graf, Der Sprachgebrauch der

altesten christlich-arabischen Literatur 11 (Leipzig, 1905).

or 15 The long form is used before a noun under two special conditions: (a) in ordering listing items, where the following noun may be collective, singular, or plural depending on circumstance (e.g. xamse 'ahwe 'five coffees' in ordering at a restaurant, cf. xams 'ahawi 'five cafes'); and (b) with ethnic collectives having no proper plural (e.g. xamse badu 'five beduins').



Even more significant, however, is the nature of the one vestige of the use of the long form with following nouns which has survived in the modern dialects, with regional variation in the extent of the retention. In many modern dialects there is a handful of high-frequency masculine nouns with plurals beginning with a glottal stop (Arabic hamzah) which replace the hamzah with a t- when one of the numbers 3-10 precedes. This t-, while unmistakably pronounced as a part of the noun plural, is clearly a vestige of the feminine ending of the preceding number. There is some variation from dialect to dialect in the number of instances of this construction. Two nouns, yom 'day' (xams tiyydm 'five days') and 'alf 'thousand' (xams talaf '5000') are apparently found in this construction in all dialects which preserve the feature at all (many Maghribi dialects have lost even this vestige), while the number of other nouns rarely exceeds fifteen in any given dialect. Example: yawmun 'day', yom 'ayydmun 'days', 'iyyam'6 xamsatu 'ayyamin 'five days', xams tiyyam It is interesting to note that all the instances of this t- involve reflexes of plural patterns referred to in Classical Arabic as 'plurals of paucity' (jumu' al-qillah). The Classical patterns are 'aCCuC, 'aCCiCah, 'aCCaC.17 The grammarians assertl8 that when a given noun has several plurals in use of which one has one of these patterns, this plural is preferred when a small number of items (3-10) is referred to. Such an assertion has a ring of artificiality about it, and in fact it does not seem to be supported by extant texts; yet if the thesis of this study is correct, the statement was not just a meaningless creation of the grammarians but probably reflected a special construction of the spoken language ancestral to the construction described above. In any case, the notion of 'paucity' is misleading, since the association was probably with the actual cardinal numbers 3-10, and the so-called 'plurals of paucity' may just as well occur with, say, 103110 or 503-510. VII. /t/ IN THE NUMBERS 13-19. In Classical Arabic the cardinal numbers 11-19 consist of two parts, a form of the number '10' and a digit part corresponding to 1-9. The noun which follows is in the accusative singular, and the 'ten' part of the number always agrees in gender with the following noun (i.e. the long form of '10', with the feminine ending, goes with feminine nouns), as do the digit parts '1' and '2'; the digit parts '3'-'9', like the independent numbers '3'-'10', disagree in gender (i.e. the long form with feminine ending goes with masculine
16 The Classicism 'ayyamis also in colloquial use without preceding number. It is sometimes difficult to elicit isolated plural forms without preceding numbers for nouns used in this construction. Informants sometimes give a Classicism with 'a-, or a totally different plural (e.g. shura), or even a form with t- (e.g. tishur). 17 The pattern CiCCah, commonly included among the plurals of paucity, does not fit this discussion. 18 Cf. M. S. Howell, A grammar of the Classical Arabic language 1.885-8 (Allahabad, 1894); M. Gaudefroy-Demombynes and R. Blachere, Grammairede l'arabe classique 178-82 (Paris, 1937); H. Fleisch, L'arabeclassique 33 (Beirut, 1956).



nouns). The numbers '11'-'19' themselves remain invariable in case inflection (always accusative without indefinite -n). xamsata 'asara baytan 'fifteen houses' xamsa 'aWrata gurfatan 'fifteen rooms' In the dialects, forms originally associated with a following masculine noun have been generalized, becoming (if they did not already have that status in Classical Arabic) compound words of a kind rare in Arabic, usually with double stress, completely invariable internally, and showing no trace of gender (dis)agreement externally. All this could be attributed to drift, even though other ways of simplifying the system could be imagined. The unexpected feature here is the presence of 'emphasis' (velarization, tafxim) in the numbers '13'-'19'; in some dialects it has spread to '12' or even to '12' and '11'. The focal point of the emphasis is the -t-, which is apparently the continuation of the -t- of the feminine ending in the digit half of the Classical number. These numbers differ in certain formal details from dialect to dialect: some dialects have lost the /'/ of the '10', others not; some dialects have the final -r of the '10' only when followed by a noun, others have it always; and so on. But all agree in having an emphatic /t/ in these numbers, a phenomenon for which no convincing explanation has ever been found. It is sometimes asserted that the emphasis in some way reflects the loss of the /'/ of the '10', and there may very well be some connection between the /'/ and the emphasis; but there are two stumbling blocks: /'/ does not cause emphasis elsewhere in the language, and the emphatic /t/ appears even in dialects which have not lost the /'/. xamsta'? 'fifteen' xamsta'sar bet 'fifteen houses' xamsta'sargurfe 'fifteen rooms' 5. The adjectives of the modern dialects reflect quite closely those of Classical Arabic in form and function, with only the kind of simplification and reduction of categories found elsewhere in the grammar: case endings are gone, the dual is gone, the varieties of feminine endings are fewer. But in three features the dialects agree on a non-Classical form which is sufficiently unmotivated to serve as evidence for my thesis: the loss of the feminine form of the comparative, traces of a fu'dl plural, and the change of the nisbah suffix -iyy to -1. VIII. Loss OF THE FEMININE COMPARATIVE. The Classical Arabic comparative 'aCCaCu (e.g. 'akbaru 'larger, largest') has a special feminine form associated with it, CuCCa (kubra). The modern dialects have a comparative form derivable from a presumed koine form *'aCCaC just as might be expected (e.g. Syrian 'akbar, Moroccan kbarwith regular loss of initial hamzah). No modern dialect, however, seems to show any trace of the feminine except for set phrases clearly borrowed from the Classical. Since the feminine of the comparative was already of limited use in Classical Arabic and was a special formation, its loss might seem to be a natural instance of drift. But the feminine of ordinal numbers was similarly limited in use in Classical Arabic and is preserved in the dialects, and the feminine of 'color' words of pat-



tern 'aCCaC, which was also a special formation, is also preserved. The following ten examples illustrate these points. 't 'akbar bet , s 'akbaru baytin 'th l IbetI'akbar albaytu'l akbaru 'akbar gurfe , 'akbarugurfatin the largest room' ^ . , ?7 T n'h l IgurfeI'akbar 'algurfatu1lkubra fifth house' xamisu baytin x&misbet . k'the fifth house' ?_ ,, 7, 7 'albaytujlxamisu Ixamis hIbw

J f


algurfatujlxamisatu 'the red house' 'albaytuVl'aHmaru 'algurfatuUlHamra'u 'the red room'

'the fifth room'lm7

fifth room'

IgurfeIxamse Ibetl'aHmar IgurfelHamra
of the pattern

IX. ADJECTIVEPLURALFU'AL. In Classical Arabic, adjectives

In CaCZC(fa'Zl)normally have a plural CiCaC (fia'dl). adjectives of this kind the modern dialects generally have a singular CCtC, sometimes CaCiC, and a plural CCaC.19But there is one unexpected complication. Generally it is impossible to tell, apart from the evidence of Classical, whether the lost vowel of a modern dialect form in which a short vowel has been dropped was originally /i/, /u/, or /a/, but sometimes there are clues in the modern form itself. If the dialect in question has a contrast r-r, then /r/ often appears near a lost /i/, and /r/ near a lost /u/ or /a/. Again, in some dialects, notably Egyptian, phrase-initial CVCVC retains the short vowel. And in some dialects, such as Moroccan, the loss of /u/ often leaves labialized consonants.20It is noteworthy, then, that all dialects which can show one or two of these clues, give evidence for a lost /u/ in these adjective plurals, with modern forms such as kbar, kubar, k/br.21 Accordingly, we are probably justified in positing a plural *CuCaC in the koine, a striking feature of difference from Classical Arabic, which shows no trace of a fut&lplural.
X. NISBAH SUFFIX -iyy > *-i. In Classical Arabic pause forms there is con-

trast between final -iyy and final -1, e.g. 'arabiyy 'Arabic', sabiyy 'boy': 'alqadz
19For the singular some dialects keep the -a- throughout; some lose it completely. In areas where there is partial retention the -a- appears in adjectives of which C1or C2is a guttural (x g H ' h ') or in Classicisms of various periods. The loss of -a- is probably to be accounted for by vowel assimilation and loss of unstressed /i/ in open syllable (CaCIC > *CiCIC > CCXC).Cf. H. Blanc, Studies in North Palestinian Arabic 32 (Jerusalem, 1953); I. Garbell, Remarks 321. The plural seems perfectly regular CiCaC > CCaC. 20 The analysis of this labialization is uncertain. Some linguists recognize the labialized consonants as separate phonemes, others posit a rounded shwa vowel phonemically present but apparent only in the 'allophones' of these consonants. In either case the labialization is distinctive. 21 In this particular example, used here to keep the illustrative material as limited as possible, these apparent reflexes of /u/ could have developed simply because of the presence of the labial /b/, but other adjectives of the same pattern without labial consonants also show these reflexes (e.g. qusar, 'urJ4, gudfid). A striking piece of evidence for this *fu al plural is supplied by Haim Blanc: dialects with second and third degree 'imalah (e.g. Aleppo, Mosul, Jewish Baghdad) regularly have e or z in words derived from CiCaC but have a in 'tongue'; jmel, jmIl 'camels' these adjective plurals. Examples: kleb, klib 'dogs'; Isen, Ismn but smiin, koar, mlaH.



'the judge', baytz 'my house' 'uktubl 'write (f sg)!' A very common instance of final -iyy is the suffix added to a noun to form a relative adjective (Arabic nisbah). The dialects vary in their treatment of final vowel and semivowel contrasts, but all agree on having the nisbah suffix identical with the reflex of final _-.22 This is especially surprising for two reasons. First, the functional load of this contrast is fairly heavy. There are several suffixes and several regular stem forms ending in -s as opposed to the nisbah -iyy, and minimal or near-minimal pairs are fairly numerous (e.g. mdliyy 'financial': mall 'my property'). Second, there is strong support for the contrast from the feminine forms of the adjectives: the feminine of a nisbah in Classical Arabic ends in -iyyah, while that of an adjective with stemfinal -1 ends in -iyah, a contrast continued in the modern dialects and reinforced by shift of stress. Several examples will clarify this:
'the second (m)', ttani 'aOOdan

'the second (f)', ttanye 'aOOaniyah

'al'arabiyy 'the Arabic (m)', I'drabi

'al'arabiyyah 'the Arabic (f)', I'arabiyye bal 'my mind', bali taktubZ '[that] you (f sg) write', tikitbi 6. Recognizing definite differences between the lexical stock of Classical Arabic and that of the dialects is much more difficult than recognizing differences in morphology, since the grammatical system of the 'Arabiyyah is fairly well codified while the limits of its lexicon have never been established. As in any literary language in use over such a large span of space and time, Classical Arabic has coined new words, has absorbed and 'Classicized' words from the spoken dialects, and has shown great variation in the relative frequency of use of words at different periods and places. In spite of these and other difficulties it is possible to note several lexical differences which are sufficiently clear to be significant for the argument here. A striking feature of lexical difference between Classical Arabic and the dialects is the disappearance in the dialects of a group of high-frequency words such as ma 'what', 'aydan 'also', laysa 'it is not'; a number of particles such as 'inna, 'an, 'anna 'that'; qad, sawfa tense markers; and several prefixes such as ka- 'like'. The disappearance of the particles is connected with the loss of modal distinction in the verb, and their functions are carried out by other syntactic means. But words like ma and 'aydan have various equivalents in the dialects, and no satisfactory explanation has been offered for this replacement. If there
The pause forms of Classical Arabic show a three-way final contrast in each of the high ranges viz. -Cuww : -Cu : -Cw and -Ciyy : -Cs : -Cy. In many modern dialects still another possibility is added, the reflexes of the Classical -Cuh and -Crh. In the u/w range the dialects vary greatly, some even having a full four-way contrast, such as those variants of Syrian Arabic which differentiate the final sequences of 'adtww 'enemy', Hiliw 'sweet', kdtabu 'they wrote', katabu 'they wrote it'. But this is rare; usually only a two- or at most three-way contrast obtains, with one reflex for both -Cu and -Cw and, in dialects which have lost final -h, one for both -Cuwwand -Cuh. Dialects show similar variation in the degree of retention of final contrasts in the i/y range, but even where a final -iyy: -z contrast has been preserved, the nisbah ending has always merged with the reflex of -1, sometimes pulling along with it a few other nouns in -iyy.




were also instances of retention of the Classical forms in certain dialects this would constitute evidence against the thesis, but in fact the disappearance is universal, and the varied replacements must be accepted as instances in which the koine was not fully homogeneous.23Three items of lexical agreement will be treated here: 'bring', 'see', and the relative.
XI. THE VERB 'TO BRING'. Classical Arabic had two verbs 'to come' 'ata and

ja'a; both of these could be used with bi- 'with' in a sense equivalent to English 'bring'. The verb 'ata has disappeared from non-Arabian dialects, a reflex of ja'a being in the dialects the usual word for 'come'. The exact formation of this word 'to come' varies from one dialect to another, since with the loss of final hamzah this verb has too little substance to fit any normal pattern of Arabic verbs. The modern reflex of ja'a is not used with bi- to mean 'bring'. The regular word for 'bring' in the dialects is a new verb jab (imperfect yj3b), which clearly has arisen from a fusion, at some early date, of ja'a and bi-. This verb behaves like a middle-weak verb (vj y b) with full regularity of form and no evidence of any morphemic boundary remaining between the original ja'a part and the original bi- part. In the Classical language there is no trace of the fused verb. That such a fusion could take place at some point in the development of Arabic is perfectly conceivable, but this is the only clear-cut case of such fusion in the language, and the exact pattern common to the dialects is striking: loss of 'at&, retention (and varied re-formation) of ja'a, no use of reflexes of ja'a and bi- to mean 'bring'; fused verb jab 'bring'. To explain the persistence of this pattern throughout the Arab world one would have to assume that this unparalleled fusion was made at many times and places and always outlived the other forms. The common origin of the dialects is a much simpler explanation.
XII. THE VERB'TOSEE'. By far the commonest verb 'to see' in Classical Arabic is ra'a (imperfect yara); this is the ordinary word in all written and oral use of the Classical language today. On the other hand, as the ordinary word 'to see' the dialects have saf (imperfect ysuf). The verb ra'a appears in the dialects only in derivative forms (e.g. Moroccan warra 'show') or in marginal words such as the Maghribi rani 'I am', rak 'you are', etc. [= 'see me!', 'see you!', etc.].24The verb saf occurs in Classical Arabic, but not with the meaning 'see'. It might be argued that with the loss of final hamzah the verb ra'a would lack substance to fit the Arabic verb system, but this seems not to have prevented ja'a from continuing in the dialects, and parallel formations to those of ja'a could have been expected.
23 Certain preliminary reconstructions can of course be made for the various items but they do not lead very far. For example, it seems likely (a) that ma 'what' was very early replaced by *'dyji ^ *'ays (< 'ayyu say'in), which led to the modern dialect forms such as Syr. 'es and Moroccan as, and (b) that side by side with this *'aysi ^ 'ays in certain areas an extended form *'aysinhu (< 'ayyu say'in huwa) was used, which resulted in forms like Iraqi Xinuand Syrian su. But this still leaves unanswered such questions as the reason for the loss of ma, the origin of Egyptian 'eh, and many points of detail. Reconstructions of other items present similar problems. 24 Cf. W. Fischer, Die demonstrativen Bildungen der neuarabischen Dialekte 186-93 (The Hague, 1958).



XIII. THE RELATIVE *'illi. The relative 'pronoun' of Classical Arabic, 'allaYii, with its feminine, dual, and plural forms, has disappeared in the modern dialects. The forms of 'allaYii in Classical Arabic are isolated, having no support elsewhere in the grammatical structure, and there was already great dialectal variation in Arabia in the forms of the relative.25Accordingly, it is not surprising that the Classical form should have vanished, but it is significant that throughout the non-Arabian dialects the only forms found are those which may be derived from a presumed *'illi, invariable for gender and number, occasionally reduced to 1- or expanded to halli or yalli.26 7. The phonologies of the dialects continue to a remarkable extent the phonology of earlier Arabic as represented by the 'Arabiyyah.27Several consonant phonemes show varying phonetic shapes in the dialects, notably those represented by the letters jtm and qaf, and a few additional consonants have appeared in various dialects, either filling 'gaps' in the structure or, as a result of mass influx of loanwords, extending the basic structure. The long vowels have been relatively stable, and the so-called diphthongs ay and aw have generally been monophthongized. Only the short-vowel phonemes have been highly unstable; they have been lost and combined, and new phonemes have arisen, all in a bewildering variety of ways. Only one purely phonological feature seems to give clear evidence for the thesis of the koine. XIV. THE MERGER dad and <a'. The sound system of the 'Arabiyyah as OF described by the early grammarians included two 'emphatic' interdental phonemes, those represented by the letters Wa' and dad. The former was presumably velarized, voiced, interdental (spirant), of the kind heard in dialects such as Iraqi today. The other apparently had all the distinctive features of the SY' and in addition was lateral or lateralized and probably a stop or affricate. Whatever the phonetic details, the two were separate phonemes. Minimal pairs have been listed by Cantineau and others, and there are consistent correspondences with other Semitic languages. In no non-Arabian dialect today are there phonemically independent reflexes of these two phonemes. In dialects which preserve the interdental spirants /0 t/, the reflex /fi/ of the interdental emphatics is phonetically the sound described above for 5a'. In dialects which have lost the interdentals (0W > t d), the reflex /d./ of the interdental emphatics is a velarized voiced stop, now the voiced counterpart of the reflex /t/ of Classical td\ This clearly suggests that Sa' and dad had merged in the koine and that the interdentals were lost subsequently in various dialects.28
Rabin, Ancient West Arabian 39, 89-90, 154-5, 203-5, map 204. The agreement of the dialects on this point has been noted before. Cf. I. Anis, Fl al-lahajat al-'arabiyyah2218 (Cairo, 1952). 27 For a description of the phonology of the 'Arabiyyah cf. J. Cantineau, Esquisse d'une phonologie delarabe classique, BSL 43.000-0 (1946). The phonology of the koine as assumed in the present study is roughly equivalent to Garbell's Stage 2, Remarks ?2, 301-12, but with a number of differences of detail. 28 Cf. Fuick, 'Arabiyah 89; Garbell, Remarks 308. Dialects which have lost the interdentals may have instances of /?/ in Classicisms or in re-borrowing of Arabic items from Turkish, but not as the regular reflex of the earlier /?S/. Cf. Garbell, Remarks 317-8.

26 Cf.