of 52 /52
Arab. arch. epig. 2000: 11: 28–79 Copyright C Munksgaard 2000 Printed in Denmark. All rights reserved ISSN 0905-7196 Reflections on the linguistic map of pre-Islamic Arabia M. C. A. MACDONALD Oriental Institute and Wolfson College, Oxford, UK At a workshop on ‘Civilisations de l’Arabie pre ´islamique’ in Aix-en-Provence in February 1996, I was asked by the organizers to give a survey of the state of our knowledge of the languages and scripts of pre-Islamic Arabia and to pro- pose a coherent set of definitions and terms for them, in an attempt to clarify the numerous misapprehensions and the somewhat chaotic nomenclature in the field. I purposely concentrated on the languages and scripts of the Arabian Peninsula north of Yemen, and only mentioned in passing those of Ancient South Arabia, since these were to be the subject of another paper. Unfortunately, four years after it took place, the proceedings of this workshop remain unpub- lished. In the meantime, the contents of my paper have circulated widely and I, and others, are finding it increasingly frustrating having to refer to it as ‘forthcoming’. I am therefore most grateful to the editor of AAE for allowing a considerably revised version of my paper to be published here. It should be seen as an essential preliminary ground-clearing for my detailed discussion of the Ancient North Arabian languages and scripts which will appear early in 2001 in The Cambridge Encyclopedia of the World’s Ancient Languages (ed. R.D. Woodard, Cambridge University Press) and my book Old Arabic and its legacy in the later language. Texts, linguistic features, scripts and letter-orders, which is in preparation. Terminology The epigraphy of pre-Islamic Arabia is lit- tered with labels which were misnomers even when they were first applied (eg. ‘Safaitic’, ‘Thamudic’) or have become so as research has progressed (eg. ‘Minaic’). I shall preface this paper with an attempt to present a more coherent taxonomy, taking account of what we now know, and do not know, of the linguistic situation at various periods in pre-Islamic Arabia. When suggesting new terms (1) I have tried, as far as possible, to follow system- atically the use of the ending -ic (Sabaic, etc.) for languages and scripts, and -aean/ ian or -ite for peoples and cultures (Sa- 28 baean, Qatabanian, Lihyanite, etc.). The two cases in which this is not possible are the terms ‘North Arabian’ and ‘South Ar- abian’, where the -ian ending is necessary to distinguish these groups of languages and scripts from Arabic. The use by some scholars of the terms ‘North and South Ar- abic’ for ‘North and South Arabian’ is therefore to be regretted, particularly since others use the term ‘North Arabic’ to refer not to Safaitic, Thamudic, etc., but to what is normally called ‘Arabic’ (2). In the case of ‘South Arabic’ the term is particularly misleading since neither the Ancient nor the Modern South Arabian languages are in any sense ‘Arabic’ (3).

Reflections on the Linguistic Map of Pre-Islamic Arabia

Embed Size (px)


Linguistic history of Arabia

Text of Reflections on the Linguistic Map of Pre-Islamic Arabia

  • Arab. arch. epig. 2000: 11: 2879 Copyright C Munksgaard 2000Printed in Denmark. All rights reserved

    ISSN 0905-7196

    Reflections on the linguistic map ofpre-Islamic Arabia

    M. C. A. MACDONALDOriental Institute and Wolfson College, Oxford, UK

    At a workshop on Civilisations de lArabie preislamique in Aix-en-Provencein February 1996, I was asked by the organizers to give a survey of the state ofour knowledge of the languages and scripts of pre-Islamic Arabia and to pro-pose a coherent set of definitions and terms for them, in an attempt to clarifythe numerous misapprehensions and the somewhat chaotic nomenclature inthe field. I purposely concentrated on the languages and scripts of the ArabianPeninsula north of Yemen, and only mentioned in passing those of AncientSouth Arabia, since these were to be the subject of another paper. Unfortunately,four years after it took place, the proceedings of this workshop remain unpub-lished. In the meantime, the contents of my paper have circulated widely andI, and others, are finding it increasingly frustrating having to refer to it asforthcoming. I am therefore most grateful to the editor of AAE for allowing aconsiderably revised version of my paper to be published here. It should beseen as an essential preliminary ground-clearing for my detailed discussion ofthe Ancient North Arabian languages and scripts which will appear early in2001 in The Cambridge Encyclopedia of the Worlds Ancient Languages (ed. R.D.Woodard, Cambridge University Press) and my book Old Arabic and its legacyin the later language. Texts, linguistic features, scripts and letter-orders, which is inpreparation.

    TerminologyThe epigraphy of pre-Islamic Arabia is lit-tered with labels which were misnomerseven when they were first applied (eg.Safaitic, Thamudic) or have become soas research has progressed (eg. Minaic). Ishall preface this paper with an attempt topresent a more coherent taxonomy, takingaccount of what we now know, and do notknow, of the linguistic situation at variousperiods in pre-Islamic Arabia.

    When suggesting new terms (1) I havetried, as far as possible, to follow system-atically the use of the ending -ic (Sabaic,etc.) for languages and scripts, and -aean/ian or -ite for peoples and cultures (Sa-


    baean, Qatabanian, Lihyanite, etc.). Thetwo cases in which this is not possible arethe terms North Arabian and South Ar-abian, where the -ian ending is necessaryto distinguish these groups of languagesand scripts from Arabic. The use by somescholars of the terms North and South Ar-abic for North and South Arabian istherefore to be regretted, particularly sinceothers use the term North Arabic to refernot to Safaitic, Thamudic, etc., but to whatis normally called Arabic (2). In the caseof South Arabic the term is particularlymisleading since neither the Ancient northe Modern South Arabian languages arein any sense Arabic (3).


    i) Ancient North Arabian [ANA] (nordarabique ancien, Altnordarabisch) [pre-Islamic h- (hn-) and apparentlyzero dialects]Oasis North Arabian [ONA]

    Taymanitic [formerly Thamudic A/Taymanite]Dadanitic [formerly Dedanite and Lihyanite]Dumaitic [formerly Jawfian]Dispersed ONA [texts in the ONA scripts from Mesopotamia (formerly called Chaldaean) and other

    places, which cannot be classified as Taymanitic, Dadanitic or Dumaitic]SafaiticHismaic [formerly Thamudic E; also erroneously called Tabuki Thamudic or South Safaitic]Thamudic B, C, D, Southern ThamudicHasaitic [?]

    ii) ArabicOld Arabic (vieil arabe, Altarabisch) [pre-Islamic $l- dialects]

    Pure Old Arabic texts [wholly in Old Arabic but written in the Sabaic, Nabataean, early Arabic or Greekscripts]Mixed texts: Safaeo-Arabic, [Sabaeo-Arabic], Dadano-Arabic, Nabataeo-Arabic, Aramaeo-Arabic [textswritten in the Safaitic, [Sabaic], Dadanitic, Nabataean, or other Aramaic languages and scripts but with OldArabic features]

    Middle ArabicClassical ArabicModern Standard ArabicSpoken Arabic Dialects

    [Appendix: Undifferentiated North Arabian](1) Pure Undifferentiated North Arabian texts [can be in any script, but are clearly North Arabian in language

    although they cannot be assigned either to Old Arabic or to a particular ANA dialect](2) Undifferentiated North Arabian Mixed texts eg. Sabaeo-North-Arabian [formerly pseudo-sabeen: texts

    written in the Sabaic (etc.) languages and scripts but including North Arabian features which cannot be assi-gned either to Old Arabic or to a particular ANA dialect]

    Fig. 1.Suggested terminology for languages and scripts: I. North Arabian (nordarabique, Nordarabisch).

    Languages (4)The ancient and modern languages of theArabian Peninsula fall into two quite dis-tinct linguistic groups: North Arabian andSouth Arabian. While they share some fea-tures which mark them off from most otherSemitic languages, there are many morewhich distinguish them from one anotherand it is now realized that their relation-ship is not particularly close.

    I. North Arabian (nordarabique,Nordarabisch) (5)The term at present in general use is a sens-ible neutral label covering:


    i) Ancient North Arabian [ANA](nordarabique ancien, Altnordarabisch) (6)which comprises the pre-Islamic h- (hn-)and apparently zero (7) dialects:

    Oasis North Arabian [ONA] viz.TaymaniticDadaniticDumaiticDispersed ONA (8)

    Sw afaiticHw ismaic (9)Thamudic B, C, D, and SouthernThamudic

    and possiblyHw asaitic.


    At present, we know too little about thelinguistic features of this group to makeany further subdivision.

    ii) Arabicin all its forms and at all stages of its devel-opment. The earliest of these stages is:

    Old Arabic (vieil arabe (10), Altarab-isch (11)). Just as Old English, OldFrench, etc. refer to the earliest sur-viving stages of these languages, Old Ar-abic refers to the pre-Islamic $l-dialects ofwhich traces remain in a handful of textsand in names (see below).

    In the Islamic period we have evidence ofother varieties of Arabic of which the prin-cipal are:

    Middle ArabicClassical ArabicModern Standard ArabicSpoken Arabic Dialects

    Appendix to the North Arabian groupThere are a number of texts from the pre-Is-lamic period which are wholly or partiallyin a North Arabian language but whichcannot be classed either as Old Arabic or asANA because they contain only featureswhich are common to both. I have labelledthese Undifferentiated North Arabian,until such time as they can be classifiedmore precisely. See the discussion below.

    II. South Arabian (sudarabique,Sudarabisch) (12)A clear, neutral, geographical term to indi-cate both the ancient and modern non-Ar-abic, Semitic languages of the region cov-ered by modern Yemen and Oman. Withinthis overall grouping, two sub-groups canbe distinguished:

    i) Ancient South Arabian [ASA](sudarabique ancienAltsudarabisch)is perhaps a safer term than either Epi-


    graphic South Arabian, which describes alanguage group by the materials on whichit is written, or Old South Arabian whichimplies (incorrectly) that it is the direct an-cestor of Modern South Arabian (13). Thiscollective label covers two subdivisions.a) The Sw ayhadic (14) languages, ie. those

    traditionally called Epigraphic SouthArabian, Old South Arabian, Altsud-arabisch, etc., viz.

    SabaicMadabic (15)QatabanicHw adw ramitic (16)

    b) The non-Sw ayhadic languages, ie. theother ancient languages of southern Ar-abia, of which so far we have only rareglimpses. Among these is the spokenlanguage of the Hw imyarites, who usedSabaic in their inscriptions. At present,this is known only from reports bywriters of the Islamic period, but ispossibly the language of the hymn tothe sun-goddess at Qaniya (17). The lan-guages of two other texts (18) which areat present incomprehensible seem alsoto fall under this heading, as wouldprobably Native Minaic, the languagespoken by the Minaeans (as opposed tothe Madhabic which they wrote), if anexample were to be found. It is possiblethat the dipinti recently discovered inDhofar (19) should be included as well,but they have yet to be deciphered.

    ii) Modern South Arabian [MSA](sudarabique moderne, Neusudarabisch)is the common collective term for

    Batw hw ar,Hw arsus,HobyotJibbal,MehrSuqutw r (Socotri)

    the unwritten, non-Arabic Semitic lan-guages spoken today, or in the recent past,


    i) Ancient South Arabian [ASA] (sudarabique ancien, Altsudarabisch)Sayhadic

    SabaicMadhabic [formerly called Minaean/Minaic. The written language used by the Minaeans and apparently

    inherited from their predecessors in the region of Wad Madab, in the Yemeni Jawf].QatabanicHadramitic

    Non-SayhadicHimyaritic [the native language of the Himyarites, of which a handful of possible examples remain]Other non-Sayhadic texts (ZI 11?, Ja 2353?)? Native Minaic [this should be restricted to any evidence that may appear for the language the Minaeansspoke]?? The language of the Dhofar dipinti [at present undeciphered]

    ii) Modern South Arabian [MSA] (sudarabique moderne, Neusudarabisch)Batw hw arHw arsusHobyotJibbalMehrSocotr

    Fig. 2.Suggested terminology for languages and scripts: II. South Arabian (sudarabique, Sudarabisch).

    in Oman, southeastern Yemen and Socotra.Although the terms North Arabian and

    South Arabian are both drawn from ge-ography, the groups they describe are de-fined by very different criteria. North Ar-abian describes a group of dialects (poss-ibly languages?) which appear remarkablyhomogeneous linguistically. For each ofthese, its phonemic repertoire, morphologyand (as far as we can tell) syntax, findcloser parallels within the group than withany language outside it.

    By contrast, the term South Arabian isbased more on geographical than linguisticcriteria. While there appear to be fairlyclose internal relationships within theModern South Arabian group, it is ques-tionable to what extent all the languageswithin the Ancient South Arabian Sayhadicgroup really belong together on linguisticgrounds. Indeed, future discoveries mayprompt a drastic regrouping and relabel-ling of all the Ancient South Arabian lan-guages. Equally, there seems to be no ques-


    tion of any lineal descent, at least from theSayhadic languages to the Modern SouthArabian tongues (20). Thus, South Ar-abian is a geographical term which atpresent covers three quite distinct types oflanguage group, each defined by differentcriteria. Sayhadic represents the official,written, languages of the ancient South Ar-abian kingdoms, a grouping based as muchon the fact that they are all relatively welldocumented, as on linguistic features.Non-Sayhadic is simply a Restklassenbil-dung for any language indigenous to an-cient South Arabia, which cannot be de-fined as Sayhadic; while MSA is the onlyone of the three to be defined by linguisticcriteria.

    To some extent, the contrast betweenNorth Arabian and South Arabian reflectsthe extent of our knowledge of the two. Thelinguistic data available for Ancient NorthArabian is relatively sparse and it is possiblethat if we had as much material for it as wedo for Ancient South Arabian we might dis-


    cover that the group was a good deal lesshomogeneous than it appears.

    ScriptsThe Arabian alphabetic tradition (21)The Arabian and the North-West Semiticalphabetic traditions are the two great al-phabetic writing systems of the AncientNear East. It is generally assumed that theystemmed from a common source in thenorth and separated some time in thesecond millennium BC. Each has its owntraditional alphabetic order (the North-West Semitic $bgd (22) and the Arabianhlhw m (23)) and both have been found atUgarit (24). This in itself, of course, is onlyevidence that both letter orders were in usebefore the beginning of the twelfth centuryBC. It seems likely that the hlhw m was notnative to Ugarit, where the local alphabeticorder ($bghd) is amply attested, the infer-ence being that it was in use elsewhere. Theonly region where it seems to have been innormal use is Ancient South Arabia, al-though the evidence is from a much laterperiod. It is true that at sites in the YemeniJawf, fragmentary inscriptions in the SouthArabian alphabet have been found in levelsdated by Carbon 14 to between the ninthand thirteenth centuries BC (25) but, whilethe juxtaposition of these unrelated frag-ments of evidence raises intriguing pos-sibilities, there is as yet no firm basis fordating the origins of the Arabian alpha-betic tradition.

    The individual scripts within the Ar-abian group are as follows:

    I. The North Arabian scriptsTaymaniticDadaniticDumaiticDispersed Oasis North ArabianSafaiticHismaic


    Thamudic, ie. a number of differentscripts which have not yet been fullyidentified and distinguished, repre-sented in some 11,000 graffiti scatteredthroughout the Arabian Peninsula.

    II. The South Arabian scriptsMonumental South Arabian of whichthere are relatively minor variations in the

    SabaicMadhabicQatabanic andHadramitic inscriptions; theHasaitic script (developed from theSabaic) (26); thezabur or minuscule scripts (27); andpossibly the script(s) of theDhofar dipinti and inscriptions.

    III. The Ethiopic syllabary (or vocalizedalphabet)the only form of the Arabian alphabetic tra-dition still in use today.

    New nomenclature for languages and scriptsThe different North Arabian alphabetswere related to each other and to the SouthArabian alphabets in ways that are not yetfully understood. Among the North Ar-abian scripts there are a number of sub-categories that urgently need redefinition.Because the majority of ANA inscriptionsare short, very often consisting solely ofnames, genealogies of varying lengths andintroductory particles, it is often impossibleto identify the language of the text (28). Itis therefore customary, faute de mieux, toapply the name of the script to the languagein which (it is assumed) the text waswritten, unless there is evidence to the con-trary (29). This is most unsatisfactory but atpresent unavoidable. The discussion whichfollows is therefore primarily concernedwith the taxonomy of the North Arabianscripts (for which we have ample evi-dence). The nomenclature of the languages


    (on which we have much less data) will in-evitably follow that of the scripts in whichthey are habitually expressed, exceptwhere there is evidence to the contrary.

    There is an increasing recognition thatthe distinction between the Dedanite andLihw yanite scripts is artificial and that theyrepresent the same script at different stagesof development. Moreover, the chrono-logical limits of these different stages donot necessarily coincide with the kingdomsof Dadan (see note 1) and Lihw yan, afterwhich they are named. However, if weabandon the term Dedanite it would stillbe anachronistic to talk of Lihyanite be-fore the kingdom of Lihw yan. It would seemto be more sensible to name the script (atall stages of its development) after the oasisin which it developed, rather than after aspecific kingdom. I would therefore sug-gest that, from now on, the labels Lihy-anite and Dedanite be abandoned andthe script and language throughout theirhistory be referred to as Dadanitic. Itwould then be possible to distinguish dif-ferent phases within the development ofthe script (early, middle, late, etc.)without tying them to political events, thedating of which is anyway uncertain.

    Similarly, I would refer to the scriptswhich developed in the oases of Tayma$

    and Duma (modern al-Jawf) as Tayman-itic (30) and Dumaitic (31) respectively.These terms would also refer to the dialectsnormally expressed by these alphabets.

    The scripts of the Early Dadanitic, Taym-anitic and Dumaitic texts, together withthose of the so-called Chaldaean inscrip-tions (of which more below) are very closein form, and it is often difficult to assign ashort text to one or other type (see Fig. 3).It therefore seems to me useful to think interms of an Oasis North Arabian [ONA]script, which was employed, with smalllocal variations in letter-form and ortho-graphic practice, in the three major oasis-


    towns of North Arabia: Dadan, Tayma$ andDuma, and probably also among the Arabcommunities settled in Babylonia and else-where (32). I would suggest that OasisNorth Arabian be subdivided into Dadan-itic, Taymanitic, etc. only when there isclear evidence to justify this (33).

    A new name is also urgently needed forthe so-called Chaldaean inscriptions, thebrief texts in the ONA script, on seals, pot-tery, bricks, etc. which have been found invarious parts of Mesopotamia and else-where (34). They are clearly in no senseChaldaean (35) and Burrows descriptionof the script as Old Arabic is equally mis-leading (36). While it is likely, although inmost cases unprovable, that at least someof these texts are to be associated with theArab communities settled in Babylonia,others appear to be connected with Syriaand Transjordan (37), and yet others seemto be from Arabia itself (38). They thereforedo not form a homogeneous group. Iwould suggest that this be reflected in a de-scriptive term such as dispersed OasisNorth Arabian inscriptions, failing amore concise and elegant title.

    The name Thamudic was the inventionof western scholars even though there isvirtually no evidence to connect any of thetexts gathered under this rubric with theancient tribe of Thamud (39). However, thelabel is far too well established to bechanged and its very inappropriateness,when recognized, serves to emphasize theartificiality of the category. For Thamudicis no more than a Restklassenbildung (a termI owe to E. A. Knauf, ZDPV 97, 1981; 189,n. 7), a sort of undetermined pigeon-holeinto which one can put everything whichdoes not fit into one of the better-definedcategories. In a field such as Ancient NorthArabian, in which so much of the evidenceis uncertain and so much work still needsto be done, this serves a useful purpose.

    However, while this is, or should be,












































































    well known to those working in the field,it is necessary to warn those less familiarwith the subject who may be misled intoassuming that the items brought togetherunder this rubric can in some way betreated as a whole. If scholars in otherfields are misled or bewildered it is largelythe fault of epigraphists such as Jammeand van den Branden who, against all theevidence, insist on lunite de lalphabetthamoudeen and, in the case of van denBranden, on a connection with the tribe ofThamud (40).

    In an exhaustive study (41), GeraldineKing has recently identified the distin-guishing characteristics of the script whichWinnett labelled Thamudic E, thus en-abling us to remove it from the ThamudicRestklassenbildung. This script, and the dia-lect normally expressed in it, therefore re-quire a new name (42). The surveys con-ducted by Geraldine King and others in theHw isma` (43) desert of southern Jordan,where to date at least 5000 texts have beenrecorded, suggest that this region of Jordanand northwest Saudi Arabia holds themajor concentration of these inscriptions,although naturally smaller numbers arescattered over a much wider area. In a re-cent article, Geraldine King and I havetherefore suggested that the script and dia-lect of these texts should be called Hw is-maic (44).

    Like Thamudic, the term Sw afaitic is amisnomer, although for different reasons. Itrefers to the mainly barren area of extinctvolcanoes and unbroken lava flows, south-east of Damascus and northeast of Jabal al-cArab (45), which is known as the Sw afa. Itwas near the eastern edge of this regionthat the first Safaitic inscriptions were dis-covered. The name is thus a purely modernlabel which bears no relation to what theauthors of these tens of thousands of textscalled themselves or the script they used (ifindeed it had a name). Unfortunately, al-


    though these inscriptions are spread over ahuge area of desert in southern Syria,northeastern Jordan and northern SaudiArabia, the one place where they havenever been found is in the Sw afa itself. How-ever, the misnomer is far too well estab-lished to be changed, although the fact thatit is a purely conventional term needs con-stantly to be borne in mind in the face ofsuch artificial conceptions as Safaitictribes or a Safaitic name (46).

    The term Hw asaitic (47) refers to inscrip-tions almost entirely funerary foundprincipally in northeast Arabia at the sitesof Thaj, Qatw f, etc. but occasionally furtherafield (48). Because of the small number ofinscriptions so far published and the re-petitive character of their content, little isknown of the linguistic features of Hasaitic.

    It is generally held that the Hasaitic in-scriptions are written either in the monu-mental South Arabian alphabet (albeit withoccasional unusual letter-forms) or in oneclosely derived from it (49). However, avery different theory has recently been ad-vanced which claims that

    Comme on trouve des textes en ecriture hw aseennedans le sud de lIraq a` partir du VIIIe sie`cle avantle`re chretienne, il semble vraisemblable que la re-gion du Golfe a emprunte son ecriture directementa` la source, probablement dans la region Syrie-Pa-lestine ... plutot quen Arabie du sud. (50)

    Unfortunately, this theory is based on sev-eral misconceptions. Firstly, the onlyHasaitic text so far discovered in Mesopot-amia is CIH 699, which Loftus found in situin an underground tomb at the foot of amound where Seleucid tablets occurred(51). The chamber had already been robbedand no dating evidence for either the tombor the inscription is known.

    Secondly, the only North Arabian in-scriptions found in Mesopotamia whichmay date to the eighth century BC aresome of the Dispersed Oasis North Ar-abian texts. However, the Dispersed ONA


    inscriptions do not form a coherent corpusrepresenting one script which could beconsidered native to this region. They are arandom assortment of small finds from avery wide area, which display a consider-able range of letter-forms and clearly repre-sent a number of different varieties of theONA scripts. This suggests that they repre-sent imports rather than the products of anative form of literacy (52). It is for thisreason that I have suggested calling themDispersed Oasis North Arabian, a labelwhich I hope emphasizes their heterogen-eous character and the fact that they areunlikely to be indigenous to the placeswhere they have been found.

    Thirdly, the script of the Hasaitic inscrip-tions is clearly very different from that ofany of the Dispersed ONA texts, as can beseen by comparing the relevant rows inFigure 3, and there is nothing to indicatethat the former developed from the latter(53). Moreover, if it had, one would needto explain how this supposed independentdevelopment produced a script (ie.Hasaitic) which was identical, in all but thetiniest details, to the monumental SouthArabian alphabet.

    The dating of the Hasaitic inscriptions isat present uncertain. Pirenne made a tenta-tive comparison of the script of oneHasaitic text (Ry 687) (54) with elements ofher Stade D2 and Stade E, for which sheoffered rough dates of c.280 BC and 150100 BC respectively, and more confidentlyassigned another, fragmentary and muchdamaged text (Ry 688) (55), to her StadeC, dating it to 350300 BC (56). However,Pirennes dating was based on the shortSouth Arabian chronology, which archae-ological finds in recent years have shownto be untenable (57). Furthermore, Pirenneherself admitted that cette graphie duGolfe Persique ne sinte`gre pas exactementa` levolution de lecriture sud-arabe; elleest specifique..., (58) the implication of


    which is surely that it may be misleadingto try to date it by fitting it into a SouthArabian palaeographical schema (59).

    Thus, at present we unfortunately haveno direct evidence for either the origin orthe dating (60) of the Hasaitic script. In theabsence of such evidence, I can only sug-gest that the Hasaitic monumental script isso similar to the monumental South Ar-abian that it is highly unlikely that theycould represent parallel developmentsfrom an early period. It seems far moreprobable that, at a date unknown, theSouth Arabian script began to be used formonumental inscriptions in the northeastof the Peninsula, possibly for reasons ofprestige. Once this had begun, this al-phabet ceased to be South Arabian be-cause it was being used at too great a dis-tance from Southern Arabia for it to main-tain the same evolutionary course as thetrue South Arabian script. Instead it be-came the Hasaitic script in the sense thatit began its own palaeographical develop-ment. Alas, the tiny number of inscriptionsso far known and the lack of direct externaldating evidence makes any true palaeo-graphical analysis impossible at present,and we must await the discovery of muchmore evidence before this, or any other,hypothesis can be tested.

    Old Arabic seems to have remained apurely spoken language until the late fifth/early sixth centuries AD (61) which means,of course, that no specific script was associ-ated with it before that period. Thus, on therare occasions when it was written, thescript associated with the local language ofprestige was used: South Arabian in thesouthern half of the Peninsula; Dadanitic inDadan; Nabataean at Hw egra, cEn cAvdat inthe Negev, and at al-Namara; a form ofeastern Aramaic at Mleiha (Malayhw a) onthe Oman Peninsula; Greek in an ecclesias-tical context in Syria (62); and early Arabic(63), again mainly in Syria. Some aspects


    of the significance of this will be discussedbelow. Here I would only explain some dis-tinctions which recent studies have re-vealed and which require some refine-ments of terminology.

    The documentary evidence for Old Ar-abic is of two kinds, see Figure 1: (64)(a) Texts in Pure Old Arabic are those ex-

    pressed wholly in the pre-Islamic $l-dialect(s), although written in a varietyof scripts, as outlined above.

    (b) Mixed Old Arabic texts are those ex-pressed in the language normally as-sociated with the script (eg. Safaitic,Dadanitic, Aramaic, etc.) but which in-clude some linguistic features identifi-able as Old Arabic. In order to distin-guish these from the pure Old Arabictexts, I would suggest that these becalledSafaeo-Arabic (65)[Sabaeo-Arabic] (66)Dadano-ArabicNabataeo-ArabicAramaeo-Arabic (67).

    I would only apply such a label to a text ifthe intrusive elements can clearly be iden-tified as Old Arabic as distinct from ANAor ASA. For the diagnostic criteria avail-able at present, see below.

    Given the paucity of material, the inad-equacies of the scripts used and the appar-ently close similarities between the twosub-groups of North Arabian, there will in-evitably be a number of cases where it isimpossible to decide whether the intrusiveelements in a particular document are OldArabic or ANA. To avoid classifying theseincorrectly and thus creating confusion inthe future when the characteristics of thetwo groups are described, it is useful tohave a Restklassenbildung, or pending cat-egory into which such texts can be placeduntil such time as their language can beidentified more exactly. I suggest, there-fore, that these texts be called Undifferen-


    tiated North Arabian. Within this class,subdivisions can be made to reflect thescript and basic language of each text. Thusthose Sabaic inscriptions from Haramwhich include intrusive elements from anundifferentiated North Arabian language(see below), I would call Sabaeo-North-Ar-abian, within the category of Undifferenti-ated North Arabian (see Fig. 1).

    Reflections on the linguistic map of pre-Islamic ArabiaAt the outset it is important to distinguishbetween languages and scripts, especiallysince, as mentioned above, the same labels(Sabaic, Safaitic, etc.) are usually appliedto both. Any script can, of course, be usedto express any language (more or less ef-ficiently) and there is no inevitable or in-dissoluble link between a particular tongueand the written form in which it is habitu-ally expressed. The fact that certain writingsystems have become associated with par-ticular languages is no more than conven-tion and, more often than not, the scriptadopted is ill-suited to the phonological re-quirements of the language it is used to ex-press. Indeed, phonology seems seldom tohave been considered, when writing sys-tems which had been developed to expressone language (with varying degrees ofcompetence) were employed to express an-other. One need only think of Akkadianwritten in a syllabary which had evolvedto express Sumerian, or the Old Aramaicphonemic repertoire squeezed into an al-phabet designed for that of Phoenician.

    The fact that the same script may be usedto write several different languages has far-reaching effects on our perception andclassification of the ancient documents westudy. To take just one example: scripts as-sociated with languages of political or cul-tural prestige, or which have become themedium of international correspondence,


    are liable to be used to express other, morelocal, languages. We are not particularlysurprised, therefore, when we find aHasaitic text in the South Arabian script, oran Old Arabic one in the Nabataean. Butwhat of the names in a graffito in the SouthArabian script, where there is no visiblegrammatical context to indicate the lan-guage; or a seal with a single name in theAramaic script? Failing any internal indi-cators, such names have traditionally beenassigned to the language usually associ-ated with the script: they become SouthArabian or Aramaic names (68). Then, alltoo often, we use linguistic or onomasticfeatures to define the boundaries betweencommunities. At best, these are extremelyinexact indicators after all, most of ushave at some time written in languagesother than our mother tongue and if, bychance, such a document was the only oneto survive, future historians might, for in-stance, class me as French on the basis of aletter I had written in that language.

    Similarly, we need to be extremely cir-cumspect in the way we use onomastic ma-terial. Synchronically, a name represents theperson who bears it, it does not mean any-thing else (69). Of course, like all words,names have etymologies. But in the vast ma-jority of cases those who give and those whobear them are unaware of, and unconcernedwith their original signification. For, in mostcases, the nugget of linguistic information aname contains is a fossil, and there is no wayof relating it to a particular stage of the lan-guage in question or to a particular societyusing that stage. We therefore have no wayof knowing whether the linguistic featuresof a name can give us accurate informationabout the language used by its bearer. Rela-tively few names at any one time are coinedwithin the community which uses them,most are considerably older, and in manycases were borrowed from another culture.Given that a man or womans name is the


    most intimate of personal identifiers, thepersonal choice of the name-giver must al-most always have been paramount, andeven in societies where practices such as pa-ponymy are the norm, the unexpected is al-ways possible. We therefore need to be veryclear about what exactly we mean when wespeak of a North Arabian or a South Ar-abian or a Nabataean name, and weshould avoid the temptation to use names ofany sort as evidence for the language, letalone the ethnicity of those who bear them(70).

    In what follows, I shall divide Arabiavery roughly by a northwest-southeast lineand refer to the western two-thirds of thePeninsula as western Arabia and to theeastern one-third as eastern Arabia (seeFig. 4).

    On present evidence, eastern Arabia(both at its northern and southern ends)appears to have had a development whichis rather different from that of the culturesfurther west. This apparent difference mayin part be due to an imbalance in theamount and type of research which has sofar been devoted to the two sides of thePeninsula. The study of the ancient historyof both the northwest and southwest is stilldominated by epigraphy, and the results ofthe archaeological work of the last twenty-five years are only now beginning tochange and fill out the traditional picture.By contrast, our understanding of easternArabia is due entirely to archaeologicalwork and, with the exception of Dhofar,this side of the Peninsula has producedonly a handful of inscriptions, containingvery little information (71).

    Thus for pre-Islamic western Arabia(especially the southwest) we have a greatdeal of information about those aspects oflife which are the stuff of monumental in-scriptions: social and political systems,land and water rights, religious practice,etc., whereas in the east, we are almost en-


    Fig. 4.Sketch map of pre-Islamic Arabia showing the east-west division used in this paper (map drawn by A. Searight).

    tirely ignorant of these. Similarly, the ar-chaeological exploration of the Gulf coastand Oman has provided information aboutchronology, city plans, diet, burial practicesand much else, which is only slowly be-coming available in the west of the Penin-


    sula. Nevertheless, even accepting this,there are clear differences between the tworegions, and these are worth exploring.

    To the epigraphist, the most striking con-trast between them is in the use of the artof writing. In northwest Arabia there was


    a multiplicity of different native scripts, inaddition to the imported Aramaic, Greekand South Arabian alphabets. From at leastthe mid-first millennium BC, forms of theOasis North Arabian script appear to havebeen in use in the major oases of Duma,Tayma$ and Dadan and, from a period notmuch later, the nomads seem also to haveused a number of different alphabets (72).The question of the development and theinterrelationships of the different Arabianscripts is still very uncertain, but it seemspossible that most, if not all, of these NorthArabian alphabets developed in parallel tothe South Arabian scripts and not fromthem.

    By contrast, the only native script to de-velop on the eastern side of the Peninsulawas that of Dhofar. Apart from this, onpresent knowledge at least, writing ineastern Arabia appears to have been funda-mentally derivative. However, I shouldemphasize that the amount of material sofar discovered is derisory. A few dozentexts in cuneiform have been found at thenorthern end of the Gulf (73), and a scat-tering of Greek texts and fragments in thesame region (74). The script of the Hasaiticinscriptions clearly derives directly fromSouth Arabian monumental, although inindividual texts there are occasional ratherodd variations in the forms of certain let-ters (75). The few Aramaic texts from theGulf are, with one major exception, shorttexts, fragments or coin legends. They dis-play a variety of forms of the script in boththe lapidary and cursive traditions andsome texts notably the coin legends butalso others contain letter-shapes appar-ently derived from both (76). Indeed, insome texts the scripts seem to have de-veloped in unusual directions and, for in-stance, in the one long inscription the signsfor d, r, k and c are identical, as are thosefor hw and m (77). Although certain of itsindividual features can be compared with


    those of Hatran and Parthian (Arsacid),this particular form of the Aramaic scriptis not known from elsewhere. It is thereforepossible that it represents a local develop-ment, of which as yet we have only twoexamples (78). However, the total numberof Aramaic texts from the Gulf area is sotiny and their geographical and chrono-logical range apparently so wide that itwould be dangerous to draw any con-clusions at present.

    In the south, the contrast between westand east is slightly different. In Yemen anumber of individual settled, literate cul-tures grew up and yet there appears tohave been the most extraordinary degree ofuniformity in the appearance of the monu-mental script (79). Since the palaeograph-ical work of Jacqueline Pirenne, it has beenassumed that this developed at a more orless uniform rate throughout the regionand even in outlying areas (80). The differ-ences between one inscription and anotherin the form of the monumental script aretherefore generally attributed to chrono-logical developments, rather than regionalvariations. At the same time, there are twoother types of the South Arabian script: theadapted majuscule of the graffiti andsome of the inscribed sticks mentionedbelow, and the zabur, or minuscule script,used on the majority of the inscribed stickswhich have been appearing in northernYemen, in their hundreds, since 1970 (81).In both cases these are groups of scripts for,in contrast to the monumental writing,there is no uniformity here and, particu-larly in the case of the zabur, there is aplethora of different script forms. Nor wasliteracy confined to the settled population,for in the deserts of southern Saudi Arabiathere are thousands of graffiti in the SouthArabian (82) and Southern Thamudic (83)scripts, some of which may well have beenwritten by nomads, like the Safaitic andThamudic graffiti of the north.


    Thus, there is at present a very clear con-trast between eastern Arabia, wherewriting seems hardly to have taken root,and the western two-thirds of the Penin-sula where it seems to have been endemic.This said, however, there are also clear dis-tinctions between the northwest and thesouthwest in languages, scripts, the typesof document available and relations withother societies. In what follows, I shall beconcentrating on what we know of thelinguistic situation in north and central Ar-abia in the first millennia BC and AD. Ishall not deal with the languages of ancientSouth Arabia, except when they impingeon those of the north (84).

    Ancient North Arabian

    The hn-dialectsIn northwestern Arabia a multiplicity ofdialects developed within the group I havecalled Ancient North Arabian, which usesh- as the definite article. The earliest occur-rence of this article is in the name of thegoddess hn-$lt in the Aramaic dedicationson silver bowls found at her shrine at Tellal-Mashutw a, in the Nile delta. These havebeen dated to the late fifth century BC (85).One of the dedicators was Qaynu barGesem, king of Qedar, and it has recentlybeen suggested that this is evidence thatthis tribe could have spoken a h-/hn-dialect(86). However, this is to stretch the evi-dence too far. The other dedications foundat the shrine are also to hn-$lt. Of the twoothers which name the dedicator, one ismade by a man with an Egyptian nameand a North Arabian patronym (87), whilein the other both names are SemiticizedEgyptian (88). It seems clear, therefore, thatthe goddess of this shrine was worshippedas hn-$lt, regardless of how this epithetwould be realized in the languages of indi-vidual pilgrims, or the Aramaic of thededications. This epithet cannot therefore


    be used as evidence for the languagespoken by the tribe of Qedar or any otherof the dedicants.

    Similarly, this example should warn usthat the epithet $Alilat mentioned by Her-odotus (89) is not necessarily evidence forthe dialect of his Arab informants (whowere probably living in eastern Egypt orSinai). In ancient Near-Eastern religionsepithets rapidly became names and, oncethis had happened, the form of the namewas fixed. All we can say is that in easternEgypt in the fifth century BC both the h-and the $l-dialects were represented in theepithet the goddess (90), but with no cer-tainty as to who was speaking these dia-lects at the time, or where they originated.

    DadaniticThe one dialect of Ancient North Arabian inwhich it is certain that hn- was used was Da-danitic. The normal Dadanitic article is h-but before /$/ (91) and /c/ (92) it is hn-. Ithas been suggested that hn- is a survival ofthe original article in all members of thisgroup, but that in the other dialects it was re-duced to h- (93). It is impossible to prove ordisprove this theory, but if this were so, onewould expect the odd irregularity of usagein Dadanitic and sporadic survivals in otherdialects. In fact, however, there is only onepossible instance in Dadanitic of hn- beforea phoneme which is not a pharyngal or aglottal (94), and no certain examples (exceptin names) in the other dialects. So this fea-ture could just as well, or perhaps morelikely, be a euphonic development withinDadanitic, as a survival from an earlierperiod.

    A Safaitic inscription, by an author whogives his nisba as hn-hw wly (95), provides aninteresting footnote to this discussion. It isclear from many other Safaitic texts that theHw wlt came from outside the normal mi-gration areas of the tribes east of the Hw aw-ran whose members wrote these graffiti,


    and it is likely that the Hw wlt should beidentified with the Avalitae, whom Pliny as-sociates with the oases of Duma and Hw egra(96), the latter at least being very muchwithin the Dadanitic hn- dialect area. Un-fortunately, the statement in the text w nfrmn rm and he escaped from the Romans is too brief to identify its language.

    Outside Dadanitic ie. principally at Tellal-Mashutw a and in the Hasaitic inscrip-tions the evidence for the article hn- comesentirely from names, all but two of whichare compounded with hn-$lt (97) and, asnoted above, these are not evidence for thelinguistic features of the texts in which theyoccur. This is particularly clear in the case ofthe name hn-$mlt, found in a Safaitic textfrom northern Saudi Arabia (98), since thereis ample evidence that the article in Safaiticis h- before all sounds, not hn-.

    The h- and possible zero dialects

    Hasaitic (?)It is important to reiterate that the etymolo-gies of personal names are not evidence forthe language of their bearers, since Hasaitichas been classed with the hn- dialectssimply because of the presence of the ar-ticle hn- in theophoric names in the texts.Other than in names, there is as yet onlyone possible instance of the article inHasaitic, and here it is the South Arabiansuffixed -n (99). So limited is the materialyet discovered that we have no other con-texts in which the definite article might beexpected (100). It is therefore far from cer-tain what form it took in Hasaitic. It is evenpossible that it did not use a definite articleat all, or else employed one, such as a vo-calic affix as in Aramaic, which did notshow up in the purely consonantal SouthArabian script.

    Despite its script, Hasaitic is classed asAncient North Arabian rather than Ancient


    South Arabian because it uses the typicallyANA expression d$l and, with the one ex-ception just quoted, lacks any characteristi-cally ASA features. This classification isprobably correct, but it is as well to beaware that it is, and can only be, based onthe tiny amount of non-onomastic materialavailable.

    Oasis North ArabianThe report, in Akkadian, by the eighth-cen-tury BC governor of Suh8 u on the middleEuphrates, concerning a raid he carried outon a caravan of the people of Tayma$ andSaba$, is by now well known. This is ourearliest evidence for the involvement of theSabaeans and Taymanites in the caravantrade (101). However, there may be anotherreference to Tayma$, this time in a hiero-glyphic Luwian document also of the eighthcentury, which is by a certain Yariris, apalace official who seems to have becomeregent of Carchemish during the minority ofits king. In this inscription, Yariris boaststhat he knew twelve languages and at leastfour scripts (102). The latter were, the scriptof the City (ie. his own hieroglyphicLuwian), and those of Sura and Assyria andfinally the Taymani script. Sura has beenidentified as Tyre and its script, therefore, asPhoenician (103). Rather more tentatively, ithas been suggested that ta-i-ma-ni-ti refersto the script of Tayma$ (104). Like the otheritems in the list, this is an adjectival form(105), and this could explain the -n-. Such aninterpretation is very attractive, not least be-cause it provides a schematic map of thetrading relationships of Carchemish at thisperiod: with Phoenicia to the west, Assyriato the east and Tayma$ and Saba$ to thesouth (106).

    If this identification is correct, it providesthe earliest historical reference to a NorthArabian script. It does not seem to menecessary to assume that the South Ar-


    abian script is meant here (107). The gov-ernor of Suh8 u was quite capable of distin-guishing between Taymanites and Sa-baeans and there is no reason to supposethat Yariris would have said Taymanitic ifhe had meant Sabaic. If Sass is correct indating to the eighth century BC some of theseals with Oasis North Arabian inscrip-tions found in Mesopotamia, this would fitvery well with this reference.

    Although I am unconvinced by many ofSasss arguments, unless one takes the ex-treme position that all the inscriptions onthe seals are later additions, his dating isprobably more or less correct in somecases. I would therefore suggest that wecould take as a working hypothesis theexistence of the Oasis North Arabian script,in one or more forms, by the eighth centuryBC. The way in which this was adapted toexpress the different dialects of the regionis only just beginning to be worked out.

    It would appear that the Taymaniticscript had signs for all three unvoiced non-emphatic sibilants, represented by s1, s2

    and s3 (see Figs. 3 and 5). It seems probablethat this s3 sign originally had a formsimilar to Monumental South Arabian s3,for a shape very close to the latter is foundon a Taymanitic seal inscription (RES 2688)(108). What appears to have happened isthat the distinction between the phonemesrepresented by s3 and s1 was retainedlonger in Tayma$ than in other places (109).Developed forms of the s3 sign appear inat least two other Taymanitic inscriptions(110), but in none of the other North Ar-abian scripts.

    Similarly, it would appear that in itsearliest form, the ONA script had no signfor /d/ (see Fig. 3). In Taymanitic, whichin this case again appears to be very con-servative, the z sign was also used for /d/,while the d signs in early and late Dadan-itic look very much as if they are adap-tations of the z sign. At Duma, on the other


    hand, a different solution was found. Theinherited z sign, with the horizontal bar setdiagonally, was used for /z/, and a signresembling the South Arabian dw was em-ployed for /d/ (in the verb cwd) (111). Atthe same time a sign similar to the SouthArabian d was used for /dw / in the N.Div.rdw w (112). This interchange of the d and dwsigns is paralleled in Safaitic and ThamudicB which also employ a sign identical to themonumental South Arabian d for their /dw /.On the other hand, in Hismaic, an iden-tical sign was used for /t/, the common tsign being used for /g/. No explanationhas yet been found for this bizarre feature.There are many other examples and a thor-ough study of the relationships betweensigns and the sounds they represent in thedifferent forms of the Arabian alphabetictradition is long overdue.

    ThamudicIn the last sixty years, the study ofThamudic has progressed considerably,and thanks to the work of F.V. Winnett andGeraldine King, it is now possible to re-move from the Restklassenbildung twogroups of inscriptions, and to make a veryrough preliminary subdivision of some ofthe rest. Of the two groups which can beremoved one is Winnetts Thamudic Awhich is now recognized as Taymaniticand has been discussed above. The other ishis Thamudic E which I have suggestedshould be renamed Hismaic (113). Geral-dine King has made a detailed analysis ofthis dialect and script (114) which, it ishoped, will be published before long, andso I shall add only the few brief remarksbelow. The distinctive features of language,orthography, style and content of both Tay-manitic and Hismaic are now fairly wellknown and each constitutes a clearly de-fined type. Of course, the categories are nothermetically sealed and there are textswhich could be either Taymanitic or early


    Dadanitic, and others which could be His-maic or Safaitic, or which show a mixtureof the features of both groups. However,this is to be expected.

    What is left as Thamudic is a mass ofover 11,000, mainly short, texts scatteredover the whole of the western part of thePeninsula from Syria to Yemen (inclusive).Approximately 2,000 of these were copiedand published by early scholars and travel-lers in northwest Arabia and were ratherroughly divided by Winnett in 1937 intothree sub-groups, Thamudic B, C and D(115). In 1970 he revised the division andthe labels, calling Thamudic B Najdi andcombining C and D under the heading Hi-jazi (116). However, this only led to con-fusion and his original classification hasgenerally been retained (117). Of thesegroups, by far the largest is Thamudic Band it is almost certain that future workwill show that this should be subdivided.Similarly, on the basis of different signs forr and n, Geraldine King has suggested thatThamudic C should be subdivided (118).

    Most of the texts are known only inhand-copies of uncertain accuracy and thiscombined with their brevity makes themextremely difficult to interpret. In addition,the values of certain signs in Thamudic B,C and D are still in doubt. An interestingfeature of Thamudic D, which is the scriptof the Thamudic counterpart of the Raqosinscription (119) and which can thereforebe dated to AD 267, is the apparently ar-chaic forms of many of its letters (see Fig.3). This is particularly clear in the case of $,b, z, s1, sw and n. This should serve as awarning to those who propose palaeo-graphical developments between onescript and another. For had we not had theRaqos text with its date, Thamudic Dwould have been classed as one of theearliest of the North Arabian scripts.

    The vast majority of the known Thamudicinscriptions (some 9,000) were discovered


    by the Philby-Ryckmans-Lippens ex-pedition mainly in the southwestern part ofSaudi Arabia. As might be expected, thesedisplay many features which are very dif-ferent from the Thamudic inscriptions ofthe north and they do not fit easily into theexisting sub-groups. They are calledThamudic simply because they are con-sidered to be North Arabian but cannot beclassified as Oasis North Arabian, Hismaic,Safaitic or Hasaitic. They are termedSouthern Thamudic to distinguish themfrom the existing rough subdivisions B, Cand D. Jacques Ryckmans described some oftheir features in his article Aspects nou-veaux du proble`me thamoudeen (120)which, in my opinion, even after half a cen-tury, remains the best available descriptionof Thamudic as a category. We must awaitthe full analysis of these texts before we canknow how many more sub-groups are rep-resented in the collection and whether anyof these can be removed from the Thamudicpending file and given labels of their own.

    HismaicSome years ago, E. A. Knauf tried to showthat the script and dialect which Winnettcalled Thamudic E should be classed as asub-group of Safaitic which he suggestedrenaming South Safaitic (121). However,as Geraldine King has shown, the script,orthography, content and some linguisticfeatures of these texts differ markedly fromthose of Safaitic and the attempt to sub-sume them under the Safaitic rubric blursimportant distinctions, bringing confusionrather than clarification (122). For instance,no fewer than six signs have identicalshapes, but completely different values, inthe two alphabets (Fig. 3). Thus the signrepresenting g in Hismaic is used for t inSafaitic, the sign for tw in Hismaic is hw inSafaitic, and so on (123). It will thus beclear that, although geographically, andperhaps chronologically, these scripts are


    Fig. 5.Sibilant shifts in Ancient North Arabian, Arabic, Nabataean and other Aramaic dialects.

    close neighbours, the palaeographical re-lationship between the two is fairly distant.

    Similarly, although the two dialects werealmost certainly mutually comprehensible,there are significant differences in usage.To take just one example, whereas Safaiticuses h- for both the definite article and the,much rarer, nearer demonstrative (this),Hismaic uses h- before nouns so in-frequently that Geraldine King has sug-gested that when it is used it may have hada specifically demonstrative force. In mostcases where one would expect the article,and where in Safaitic it would be present,there is no visible mark of definition in His-maic. This suggests that Hismaic either hadno definite article (as in Syriac) or else em-ployed a vocalic affix (like the Aramaic em-phatic ending) which would not have beenrepresented in the purely consonantalscript. In this regard it may be significantthat affiliation to a social group is never ex-pressed by the nisba in Hismaic (124).

    It is for this reason that I have tentativelydescribed Hismaic as neither a h- / hn- noran $l- dialect, but as an apparently zerodialect, as far as the definite article is con-cerned. As noted above, Hasaitic may also


    be of this type, but as yet there is too littleevidence to tell.

    SafaiticMuch progress has been made in the studyof Safaitic in the last twenty years and it ishoped that this will accelerate in the nearfuture. It is the dialect of Ancient North Ar-abian about which we know most and yetthe very nature of the texts puts a severelimit on the amount and type of infor-mation we can glean from them.

    The texts were written largely by no-mads who migrated across southern Syria,western Iraq, northeastern Jordan andnorthern Saudi Arabia probably betweenthe first century BC and the fourth centuryAD. Some 20,000 inscriptions are known atpresent and there are tens of thousandsmore awaiting discovery. They thus repre-sent a very dense concentration of evidencefor a particular Ancient North Arabianscript and dialect over a considerable areaat the northern end of the Peninsula. Theextent of the material available, and in par-ticular the huge onomasticon, can some-times distort comparisons with other dia-lects which are less well represented (125).


    I have published elsewhere (126) alengthy survey of the content of these in-scriptions and I will not repeat that dis-cussion here. Instead, I would like to drawattention to a few interesting, but uncon-nected, details which may help to place thedialect in its linguistic context.

    Like the other North Arabian languages,Safaitic had two unvoiced non-emphaticsibilants, represented by the letters s1 ands2 (see Fig. 5). It is clear from the loan-words, or rather loan-names, found inSafaitic that these letters did not have thesame phonetic values as their etymologicalequivalents in standard Arabic, repre-sented by the letters (sn) and (sn) re-spectively. It is for this reason that I nowuse the phonetically neutral notation forthese letters, s1 and s2 (borrowed from An-cient South Arabian), rather than the tra-ditional s and s which imply the pronunci-ation [s] and [e] respectively.

    In his brilliant studies of Arabian Sibi-lants, Professor Beeston showed that inAncient North Arabian and in Arabic upto at least the eighth century AD, the twounvoiced non-emphatic sibilants, repre-sented by s1 and s2, and in Arabic by theletters (sn) and (sn), were pronouncedrespectively [e] and [c] and that there wasno pure [s] (127). This explains why theAramaic divine name Bacal-samn was speltin Safaitic with s1 (thus bcls1mn) not s2. It isinteresting to compare this loan form withthe calque (ie. loan translation) bcls1my, alsospelt with s1, which is found in a fewSafaitic texts (128). If s1 and s2 in Safaitichad had the values of their equivalents(sn) and (sn) in standard Arabic, onewould have expected the loan form fromAramaic Bacal-samn to be spelt with s2 andthe calque bcls1my to be spelt with s1, sincewe know that the word for heaven wass1my (cf. Ar. sama$) in Safaitic. The fact thatboth are spelt with s1 suggests that this waspronounced [e].


    It follows from this that, if the name ofthe deity which we pronounce (anachron-istically) Dushara, had been pronouncedwith a [e] in Nabataean, it would havebeen spelt in Safaitic with s1, and this wenever find. Instead, it is always spelt withs2 and probably had a pronunciation ap-proximating to the ich-laut [c], which Sba-wayhi describes for (sn), the reflex of s2,in eighth-century Arabic (129). As I explainbelow, we know that this name came toSafaitic via Aramaic because it is generallyspelt with a d rather than the etymologicald. We must therefore conclude that in theNabataean Aramaic of the Hw awran, atleast, the phonemes /s/ and /s/, whichwere represented by the same letter ( ) inthe Aramaic alphabet, were still pro-nounced differently. This maintenance ofthe distinct pronunciation of the pho-neme /s/ marks out Nabataean Aramaicfrom most of the other Middle Aramaicdialects which underwent the sound-shift[s][s]. Simplistic dismissals of this as Ar-abic colouring (130) mask the really inter-esting questions which it raises (131).

    In Safaitic and Hismaic, the lack of [s]obviously caused problems in the translit-eration of Latin and Greek names and titlescontaining S or S. There are a few caseswhere s1 has been used ($qlds1Claudius,tts1Titus) but the more usual expedientseems to have been to use the emphatic sw(qsw rkaisa|, grfsw $Ag|ippaV, hrdsw cH|}dhV, flfsw FilippoV, grmnqsw Ger-manicus, etc.). It is almost certainly for thisreason that sw was put in the place ofsemkath in the Hismaic ABC found atKhirbat al-Samra$ (132).

    Nabataean (133)For many years it has been widely heldthat the Nabataeans wrote in Aramaic butspoke Arabic. This view is based on thepresence of Arabic loan-words in the Naba-taean inscriptions and on the claim that the


    vast majority of the names in Nabataeaninscriptions are Arabic (134). However, ifone excepts the Namara inscription, whichis in Old Arabic, and JSNab 17, which is inNabataeo-Arabic (see below), only a smallnumber of Arabic loan-words can be iden-tified in the Nabataean inscriptions. I havecounted a maximum of twenty-eight, six ofwhich could also be Aramaic (135). More-over, with two exceptions (136), thesewords occur only in the texts from Hw egraor Rawwafa: exactly where some Old Ar-abic and Ancient North Arabian influenceis to be expected. If these loans provide evi-dence of an Arabic substrate in NabataeanAramaic, it is therefore only in the lan-guage used in North Arabia from the firstcentury AD onwards, not in the Nabataeanof Petra or the other parts of the kingdom.It cannot therefore be used as evidence thatArabic was the spoken language of theNabataeans in general. This demonstratesthe danger of treating as a homogeneouscorpus all the inscriptions which we labelNabataean.

    Turning to the onomastic evidence, it isfirst necessary to distinguish what we meanby Nabataean names. Generally, this istaken to mean all names found in Naba-taean inscriptions unless there is a par-ticular reason for attributing them to an-other group (eg. Greek or Jewish names).However, we use the term Nabataean fortexts in a number of local varieties of the Ar-amaic script from the Hw awran to Arabia,Sinai and Egypt, and while it may some-times be helpful to use the label in this flex-ible way, we should not allow it to misleadus into assuming that the authors of all thesetexts belonged to a single homogeneousgroup, the Nabataeans.

    Names which are clearly of Arabic form,eg. those containing the definite article $l, orthe word $bn son (as opposed to Aramaicbr), or the $f cal nominal form, occur verylargely in Sinai. But the script of these texts


    from Sinai has its own characteristics andinternal development, which is distinctfrom that of the Nabataean heartlandaround Petra. It is therefore misleading torefer to them as Nabataean and to use themas evidence for the Nabataeans. It wouldbe wiser to return to the practice of earlierscholars and refer to them as Sinaitic.

    Moreover, while the vast majority of thesetexts consist simply of names accompaniedby slm, dkyr, bryk, b-tw b, etc. (137) the languageof the few statements which do occur is al-ways Aramaic, not Arabic. It seems to meprobable that someone writing a graffitowould express himself in his language ofnormal use, rather than a literary tongue(138), and therefore the case for spoken Ar-abic even in Sinai seems to me unproven.Certainly, the widespread use of nameswhich are linguistically Arabic is no evi-dence for the spoken language at the timethe texts were carved. Names have an extra-ordinary record of survival and it is usuallyimpossible to guess at the reasons for using,retaining, or discarding them in any par-ticular society at any particular period.

    Just as my name Michael has been pre-served through the survival of two re-ligions, so the popularity of certain names inSinai, or among the Nabataeans, could insome cases have been connected with theworship of deities who were also veneratedby speakers of Old Arabic. Unfortunately,however, we do not know which of thesecults, if any, were ancestral among the Nab-ataeans and which were later introductions.It is usually assumed that the Nabataeanscame from Arabia, but I know of no argu-ment for this origin that will stand scrutiny(139). In the earliest reference to the Naba-taeans, when they were nomadic, they werealready in southern Transjordan, which re-mained their heartland throughout theirhistory and I know of no clear evidence thatthey came from anywhere else. Given theirgeographical position and their involve-


    ment in the frankincense trades they nat-urally had links with peoples in northwestArabia and eventually came to rule thatarea. They worshipped many of the samedeities as their neighbours in that region.But if their chief deity, Dushara, was the godof the Petra mountains, as is generally be-lieved, this surely suggests a very old andclose attachment to this area. It is interestingthat in the Hismaic graffiti from the area ofWad Ramm, not far from Petra, invocationsto Dushara outnumber by approximately62% to 38% those to Lt/$lt (140), despite thefact that there was a major temple to $lt atRamm (141) and that in the Safaitic graffiti,written by the nomads to the north and eastof the area, she is by far the most populardeity. Moreover, the N.Div. Dushara inthese inscriptions is most often spelt in itsNorth Arabian form D-s2ry (142), whichcontrasts with the entirely Aramaized form(ds2r) normally found in the Safaitic inscrip-tions.

    This may contribute a small piece of evi-dence to the debate on which language theNabataeans spoke. If, as is commonly be-lieved, D-s2ry/Dwsr$ means He of theShara` (the mountains, north of Petra)(143), he was presumably a local deity forthe population of the Hw isma` of southernJordan, which produced the Hismaic in-scriptions. The meaning of the epithetwould have been obvious to them and itwould not be surprising, therefore, thatthey spelt it in its etymological form. Onthe other hand, it is also unsurprising thatit should appear as a loan-form in theSafaitic texts, written several hundred kilo-metres away, in an area where most peoplewould be unaware of the geographical ref-erence (144). What is significant is that theloan in Safaitic is from Aramaic. It seemsto me out of the question that ds2r inSafaitic could be a spelling pronunciation,since this would require that all thosewriters of Safaitic inscriptions who in-


    voked Dushara as ds2r had read his namein an Aramaic text, without ever hearing itfrom a North Arabian speaker. This is obvi-ously preposterous. However, the onlyalternative is that they received the cultfrom Aramaic speakers.

    It is, of course, likely that they came intocontact with the cult of Dushara in theHw awran, rather than at Petra, and it is gen-erally assumed, though there is virtuallyno evidence one way or the other, that thespoken language of the Hw awran was Ara-maic (145). But the cult of Dushara was avital and distinctive element of Nabataeancommunal identity. Other deities in Syriahad important cult centres (eg. Bacal-shamn at Sc) but were also worshippedelsewhere (eg. Bacalshamn at Palmyra).Dushara was one of the very few to beuniquely associated with a particularpeople or state. It therefore seems likelythat Nabataeans, from the heartland,would have been involved in his cult evenin the Hw awran, indeed one might expectthem to have dominated it. If these werespeakers of Old Arabic there would havebeen no point in their communicating withthe nomads in Aramaic, since Old Arabicand Safaitic would have been mutually in-telligible. In this case, the name Dusharawould have come into Safaitic with an ini-tial /d/ and a final, consonantal /y/. In-stead, it was borrowed in its Aramaic formwith an initial /d/ and (presumably) afinal vowel, and this suggests that thoseNabataeans from whom the cult wasadopted spoke a dialect of Aramaic. Need-less to say, this is very far from being con-clusive. It is intended simply as a smallcontribution to a continuing debate.

    Old Arabic

    Diagnostic featuresAs I have said, Herodotus statement thatthe Arabs worshipped a goddess named


    $Alilat, tells us that this form of the Ar-abic definite article was in use by the fifthcentury BC but, of course, gives us no otherinformation about the dialect. So we cannotassume that this is a direct ancestor of theOld Arabic attested in later texts (146).However, there is a handful of inscriptionsdating from the late first century BC on-wards which are either expressed in, or dis-play features of, Old Arabic and whichhave been found right across the Peninsula,in central, north and eastern Arabia, Syriaand the Negev.

    However, since Old Arabic appears tohave been normally an unwritten languageits features are not at all well defined. Themost obvious characteristic is, of course,the use of the definite article $l-. But, evenhere it is necessary to be cautious. As I re-marked earlier, the presence of $l- in a namedoes not mean that the text in which thatname occurs is in Old Arabic. The inscrip-tion from Qaryat al-Faw, a photograph ofwhich was published by Wafik Ghoneimand which was discussed by ManfredKropp at the 1991 Seminar for ArabianStudies (147), was included in a recent listof the earliest inscriptions in the Arabiclanguage, as the first attestation of the defi-nite article $l- (148). However, the onlyclear example of $l- in this text is in thename of an ethnic group the $(a)l $l-$hw nktwhich one would expect to retain the ar-ticle in its native form, whatever the lan-guage of the text (but see below under Da-dano-Arabic). A second instance of the ar-ticle has been assumed in this text in thephrase l-lt which Robin interpreted as *li-l-lat, for Al-lat. However, we know fromthe Safaitic and Hismaic inscriptions thatone of the most popular deities in NorthArabia was Lt, who is distinguished inboth theophoric names and in invocationsfrom $lt (*ilat?/Allat?) (149) there aresometimes prayers to both in the same text(150) and from the epithet (h)-$lt (the


    goddess) (151). It is therefore unsafe to as-sume that the article $l- has been assimi-lated in this phrase, since llt could as easilyrepresent *li-Lat as *li-l-lat.

    One characteristic of Old Arabic, whichhas not so far been recognized, is that the3rd. person masc. sg. of the suffix conju-gation (perfect) of verbs tertiae y/w endedin/-a/ in Old Arabic, as in the later stagesof the language, thus *bana he built (152).This is in contrast to the ANA treatment ofthis class of verbs, in which the final radicalretained its consonantal value, thus bny(presumably representing *banaya). Thisprovides us with another important dis-tinction between Old Arabic and ANA andone which is visible not only in scripts suchas Dadanitic and Aramaic which use matreslectionis for final vowels, but also in thosewhich do not. In the latter, the pronunci-ation *bana is expressed simply as bn, as inthe inscription of cIgl bn Hfcm (see below),as opposed to the regular ANA bny.

    Another distinctive mark of Old Arabicis the feminine singular relative pronoun,$lt (cf. Classical Arabic $allat), which occursin JSLih 384, see below. This is not foundin ANA, which uses dt (eg. in Safaitic andHismaic) or d$t (eg. in Hasaitic).

    Other features of Old Arabic distinguishit from the Sayhadic languages, but areshared with Ancient North Arabian. Thusit has two, rather than three, unvoiced non-emphatic sibilants, although the notation ofthese in scripts which have signs for threeis sometimes unpredictable. This is a fea-ture which Old Arabic shares with all butthe oldest stages of Ancient North Arabian.The causative stem of the verb is $f cl ratherthan hf cl or s1f cl and it uses the prepositionmn rather than bn from.

    Until recently, it was thought that the useof the negative particle lam plus the prefixconjugation (the jussive in Classical Ar-abic) was a distinctive feature which couldbe used to identify Old Arabic (153). How-


    ever, this construction has recently beenfound in a Safaitic inscription from north-eastern Jordan (154), the language ofwhich, in other respects, could be eitherSafaitic or Old Arabic (155).

    As explained in the section on nomencla-ture, before the sixth century AD, no onescript was associated exclusively with OldArabic and, on the rare occasions on whichit was written, a number of different alpha-bets was used. The reasons for this will bediscussed below, but the result is that,while there are relatively few inscriptionswhich are linguistically more or less pureOld Arabic, there are several others whichdisplay a mixture of Old Arabic featuresand those of the language normally associ-ated with the script in which the text iswritten: in particular Safaitic, Dadanitic,Nabataean or other Aramaic.

    Texts in pure Old ArabicThe earliest document which is indisput-ably in Old Arabic is the inscription of cglbn Hfcm, written in the Sabaic script,which was found at Qaryat al-Faw (156)and may date from the end of the first cen-tury BC (157).

    This text uses the article $l- (158), the*bana, rather than *banaya, form of the 3 m.sg. of the suffix stem of verbs tertiae w/7(159), the $f cl form of the causative stem,and the preposition mn rather than bn(160). On the other hand, it has certain Sa-baic features such as the form of the 3 m.sg. enclitic pronoun -hw which alternateswith -h (161) and the sporadic attachmentof -m to the end of words, possiblyattempting to imitate Sabaic mimationwithout understanding its purpose (162).

    W. W. Muller (163) has recognized thatJSLih 384 should also be classed as Old Ar-abic rather than Dadanitic, since it containsthe Arabic relative pronoun $allat in theform $lt. The text also contains an exampleof the *bana rather than *banaya form of the


    3 m. sg. of the suffix stem of verbs tertiaew/7. Dadanitic is the only ANA script inwhich matres lectionis are employed (164)and the letter -h is used to represent final/a/ (165). Thus *bana appears in this textas bnh (166), as opposed to Dadanitic (andother ANA) bny, (ie. *banaya) (167).

    The other documents which can be saidto be in more or less pure Old Arabic arethe Namara inscription (AD 328) (168),lines 45 of the cEn cAvdat inscription (169)(of uncertain date) (170), both in the Naba-taean script; and the inscriptions of Ummal-Jimal (of uncertain date), Zebed (AD512), Jabal Usays (AD 528), and Harran(AD 568) all of which are in what is recog-nizably the Arabic script (171).

    In addition, there are two leaves ofparchment from a Psalter found in the ge-nizah of the Umayyad mosque in Dam-ascus. They bear part of the Septuagint textof Psalm 78 (LXX, 77) with an Arabic glossin Greek transliteration, set out in parallelcolumns. Following a detailed study of thistext I am convinced that it is pre-Islamic(172). This is the most valuable text in OldArabic so far discovered since the Greektransliteration seems to have been madewith great care and consistency from anoral source, and thus is uncomplicated bythe orthographic conventions of anotherscript. It also, of course, provides thevowels and has the additional advantagethat there can be no doubt as to themeaning.

    Old Arabic Mixed textsAs described in the section on nomencla-ture, the Old Arabic mixed texts can besubdivided into Safaeo-Arabic, Dadano-Arabic, etc. These are texts written in theSafaitic (Dadanitic, etc.) scripts, and whichare predominantly in the language nor-mally associated with that script (ie. inSafaitic, Dadanitic, etc.), but which also


    contain elements which can be attributedto Old Arabic.

    [Sabaeo-Arabic]At present, there are no clear examples oftexts in the Sabaic script and languagewhich contain elements clearly attributableto Old Arabic rather than ANA (173).However, when all the inscriptions fromQaryat al-Faw are published it is veryprobable that some will be found to fallinto this category.

    Safaeo-ArabicThere are a handful of Safaitic inscriptionswhich appear to contain the definite article$l-. The clearest of these is C 2446 which isknown only from a hand-copy (Dn 68c).This reads

    l S1cd bn Mr$ bn Nr w-wgm c[l] $h-h Nrqtl.-h $1 (N)btw y mbcy ncm cwd w-dw f f-h-lt mcmn w-$lt dtn w-gd-[c](w)d w-gd-dw f t$rm-d $s1lf w-wlh kbr s1hw r cl $h-h hw bb-h l-$bd(174).

    By S1cd son of Mr$ son of Nr and hemourned for his brother Nr (whom) the Na-bataean killed during the attack (175) on theflocks of cwd and Dw f (176). And so, O Ltmcmn (177) and goddess of Dtn (178) andGd-cwd and Gd-Dw f (179) [grant the oppor-tunity of] revenge on him who did this thing(180). And he mourned greatly s1hw r (181) forhis brother, his beloved, for ever.

    On Dunands copy, the first letter of (n)btw yis the same length as the previous l, as op-posed to the very short ns in the rest of thetext, so the reading is not entirely secure.However, regardless of whether the nisba isNbtw y or Lbtw y, it seems probable that the let-ters $l represent the Arabic article, ratherthan the word $l (cf. Ar. $al) meaning socialgroup, since in Safaitic this is always fol-lowed by the name itself ($l rm the people


    of Rome, the Romans, $l dw f the tribe ofDw aif, etc.) and never by the nisba (rmy, dw fy).Moreover, in Safaitic, the Nabataeans arealways called simply nbtw , rather than *$l nbtw .If therefore, as seems likely, $l in this con-text is the Arabic article, this text containsa record of what I have called the northernOld Arabic isogloss (182) in which the /l/of the article is not assimilated, even beforethe sounds called by the grammariansof Classical Arabic al-hw uruf al-samsiyya orsun letters.

    However, other features of the text arecharacteristic of Safaitic, such as the voca-tive particle h- and the assimilation of then in the preposition mn. I would thereforeplace C 2446 among the mixed texts as anexample of Safaeo-Arabic.

    Another possible text of this type is WH589 which contains the statement $s2r.ql-l-mdbr and he migrated to the inner de-sert. The two ls are clear on the photo-graph, but it is quite possible that they arethe result of dittography on the part of theauthor, who had just written, and thencrossed out, a second s2 between the r andthe q of $s2rq. In Safaitic, the word mdbr isnever used with the article (183), but it ispossible that it was referred to as *$l-mdbrin Old Arabic, cf. the place-name Rhw btwhich also does not take the article inSafaitic, but appears as al-Ruhw bah in laterArabic (184).

    There are also several Safaitic inscrip-tions in which the definite article appearsas $- rather than h-. Given that in theSafaitic script the two letters are dis-tinguished only by a short side-stroke (seeFig. 3), I restrict myself to those for whichphotographs are available and where it isdifficult to suggest another explanation(other than an error by the author). ThusSIJ 37 has s1lm w-fsw yt m-b$s1 $-s1snt securityand deliverance from misfortune this year(185), and KRS 125, an unpublished in-scription from northeastern Jordan, has cwr


    l-m ycwr $-s1fr blindness to whoeverscratches out the writing (186). In bothcases, if my analysis is correct, these wouldbe records of what I have called the centralOld Arabic isogloss in which the /l/ of thearticle is assimilated before a sibilant (187).

    However, all these inscriptions displayfeatures typical of Safaitic, as well as poss-ible examples of the Old Arabic definite ar-ticle and it is for this reason that I haveplaced them among the mixed texts, asSafaeo-Arabic.

    Dadano-ArabicThe Dadano-Arabic texts (JSLih 71 and 276)and the Nabataeo-Arabic inscription(JSNab 17) were found in the very centreof an Ancient North Arabian (ie. h- dialect)area. The two former were found at Dadanand the latter at Hw egra, c.25 km to thenorth. In JSLih 71 and in JSNab 17 we findthe name of the town in its Arabic form($)l-hw gr. This contrasts with the forms Hw gr$

    in Nabataean, possibly hw gr in Safaitic (188)and the form behind Plinys transcriptionHaegra (189).

    JSLih 71 is an honorific inscriptionwritten in the late Dadanitic script. Thename and genealogy take up the first twolines and are linguistically indeterminate,in that they could be either ANA or OldArabic, although the mater lectionis -h, re-presenting the final -a in the first name, is adistinctive mark of Dadanitic orthography(190). Line 3 gives the tribal affiliation ofthe man honoured in the form d $l hn-$hw nkt.As noted above, we cannot tell as yetwhether the expression d $l was used inOld Arabic or whether it is a distinguishingmark of ANA, therefore, at present thismust also be treated as linguistically inde-terminate. However, the situation is furthercomplicated by the fact that, although theexpression d $l is typical of other ANA dia-lects (Safaitic, Hismaic, Thamudic B and


    Hasaitic), it has not yet been found in Da-danitic where, on the relatively rare oc-casions when affiliation to a social group isexpressed, it is usually marked simply byd pi the group name (191). As a final twist,however, the name of the honorands socialgroup is given as hn-$hw nkt, ie. with the dis-tinctive Dadanitic definite article, eventhough the same name is found elsewherein the form $l-$hw nkt, with the Old Arabic de-finite article $l- (192).

    It is customary to link these exampleswith various South Arabian texts in whichthe nisbas hw nkyn, hw nkytn, and $hw nkn arefound (193). However, this does not seemto me possible. The form of the tribal name,$hw nkt, as attested in North Arabian, wouldproduce the nisba * $hw nky in the masc. sg.,* $hw nkyt in the fem. sg. and presumably* $hw nkyn in the pl. and this would surelybe the same in Ancient South Arabian.I cannot therefore agree with Robin(and some earlier scholars) that the formshw nkyn (194), hw nkytn (195), and $hw nkn (196)are respectively the masc. and fem. sg.and the pl. nisbas from the clan/tribal name$hw nkt (197).

    It is now generally accepted that lines410 of JSLih 71 are in Old Arabic and itis certainly true that the definite article $l-occurs at least twice in this section of thetext. However, although Professor Bees-tons interpretation of this part of the in-scription as near-classical Arabic (198)was of great importance in drawingattention to its true language, I have diffi-culties both with individual details of hisreading and, more importantly, with thefact that (like all the interpretations beforeit) it did not take account of the fact thatthe ends of lines 49 are missing. I ampreparing a new reading of the textwhich I hope to publish in the near fu-ture. Here I will merely describe some ofits Old Arabic features.

    The article $l- occurs at least twice


    (199), once in b-l-hw gr (ie. in al-Hw igr), andonce, following the demonstrative particleh- (also found in Ancient North Arabianand, as a preposed ha-, in early Arabicpoetry) in the phrase h-l-mfl in line 8.Failure to recognize the demonstrativehere has led some writers to assume thathl- represents another, otherwise unat-tested, form of the definite article: thuspresupposing the use of three differentforms in the same text.

    The implications of JSLih 71 are in-triguing. The inscription was set up in atown which was a centre of the AncientNorth Arabian (h-) dialect, Dadanitic, andis written in the Dadanitic script. Yet theonly part of the text which is linguisticallyDadanitic is the definite article in the nameof the honorands social group (hn-$hw nkt),even though this is introduced by the non-Dadanitic ANA expression d $l. The rest ofthe text is in Old Arabic. Why should thisbe? If the honorand, or the dedicator, wasa speaker of Old Arabic, as might be sug-gested by the use of this language in thebody of the text, why was the Dadaniticarticle used with his social group, particu-larly since the groups name occurs else-where as $l-$hw nkt? Whatever the expla-nation, this text is a dramatic example ofthe close co-existence of ANA and Old Ar-abic in this area.

    A second Dadano-Arabic inscription isJSLih 276. This reads

    zdlh / bn / {k}{l}b / d-cmrtc / f-crr / dgbt /crr / $-s1fr dh

    Zdlh son of {Kl}b of the lineage of cmrtc.And so, may Dgbt defile the destroyerof this writing.

    The reading $-s1fr dh is clear (200) and therecan be no doubt that the $ was intentionalsince (in contrast to their Safaitic equiva-lents) the letters $ and h have entirely dif-ferent forms in the Dadanitic script. The ex-


    pression finds an exact parallel, but thistime with the Dadanitic article, in HE 1:

    f-crr / dgbt / crr / h-s1fr/dh.

    The protective formula is typically Da-danitic and for this reason I have classedJSLih 276 as Dadano-Arabic, rather thanpure Old Arabic. The treatment of the OldArabic article in this text is identical to thatin the cgl bn Hfcm inscription, ie. the as-similation of the /l/ before a followingsibilant, and this makes it a record of thecentral Old Arabic isogloss (201).

    Nabataeo-ArabicIf we except the cEn cAvdat inscription, inwhich Aramaic and Arabic are not mixedbut are used in different sections, the onlyNabataeo-Arabic text so far identified isJSNab 17, at Hw egra/Mada$in Sw alihw , whichis composed in a mixture of Nabataean Ar-amaic and Old Arabic and written in theNabataean script (202). The interpretationof parts of the text are still disputed, but itis clear that some elements are in Aramaic(eg. br, brt, the date) and others in Old Ar-abic, and yet others could be in either (eg.w-lcn mry clm$ mn ysn$, where, apart fromthe divine name, the rest could be eitherNabataean Aramaic or Old Arabic). As inJSLih 71, the article $l- occurs twice, oncein the place name $l-hw grw (line 4) and oncebefore a common noun $l-qbrw (line 7).

    The presence of these Old Arabiclinguistic features in the Dadanitic and Na-bataean Aramaic milieu suggests a degreeof multilingualism in these North Arabianoases which parallels that at Qaryat al-Faw,and possibly at Nagran (203). In each ofthese places there was a dominant writtenlanguage, respectively Dadanitic, Naba-taean Aramaic or Sabaic, but these mixedtexts suggest that speakers of Old Arabicwere also present, although, in the north at


    least, we have no way of knowing whetheror not they were residents.

    Aramaeo-ArabicFinally, there is an Aramaeo-Arabic inscrip-tion, from Mleiha (Mulayhw a) in the U.A.E.,which suggests that Old Arabic was also inuse in eastern Arabia, at least in the secondcentury AD (204). The text is composed ina rather barbarous Aramaic but contains anumber of Old Arabic features includingthe use of the article $l- and the prepositionf in, where Aramaic and Ancient NorthArabian use b- (205). It may also be signifi-cant that the deities invoked are Manat,here spelt mnt (ie. presumably with along /a/ which would not be representedin a medial position), rather than mnwtw asin Nabataean (206), and possibly Khl, theeponymous deity of Qryt dt Khl (ie. Qaryatal-Faw) (207). There are so few writtendocuments from eastern Arabia that it isimpossible to be sure what languages werespoken there at any time in the pre-Islamicperiod. However, this text from Mleihaprovides a tiny fragment of evidence tosuggest that Old Arabic may have been oneof the vernaculars.

    Undifferentiated North ArabianIn addition to the texts discussed above,there are a number of others which areclearly in a North Arabian language butwhich cannot be exactly classified withinthat grouping. In some cases it is not evenpossible to decide whether they are in OldArabic or in ANA. But for the majority itis possible to exclude Old Arabic withoutbeing able to assign them to a particularANA dialect.

    Since most of the latter group come fromQaryat al-Faw it is quite possible that theyrepresent the dialect of ANA spoken in thecity at one or more periods. However, thenumber of published inscriptions fromFaw is, at present, so small that it seems


    safer to refrain from classifying this as aseparate dialect until the full epigraphy ofthe site is published, after which thelinguistic situation there should become agreat deal clearer. I have therefore left thesetexts in the Undifferentiated Restklassen-bildung for the present.

    On Figure 1, I have divided the Undif-ferentiated North Arabian documents, intotwo sub-groups: (1) Pure UndifferentiatedNorth Arabian texts and (2) Undifferenti-ated North Arabian Mixed texts (see theexplanations on Fig. 1 and the discussionbelow).

    Pure Undifferentiated North Arabian textsThere are five such inscriptions from Qa-ryat al-Faw, and possibly two from Na-gran. These are:(i) Ja 2122 (208) which is to all intents

    and purposes an ANA text. Whilemuch of the vocabulary could, ofcourse, also be Ancient South Ar-abian, the $f cl Form, $hw dt, must beNorth Arabian. In this context, thespelling bny (ie. *banaya) as opposedto *bn (209) (ie. *bana) for the verb tobuild, suggests that the language isin a dialect of ANA rather than OldArabic (see the discussion of JSLih 348under Pure Old Arabic texts, above).

    (ii) The tomb inscription of Mcwyt bnRbct (210). With the exception of theAncient South Arabian form of thenisba, qhw tw nyn, the language of this textis pure North Arabian. Given this, Iwould classify it as ANA rather thanOld Arabic on the basis of the formbny. Kropps conclusion that, if theclan name at the end of line 3 is re-stored as $[hw nkt], then at least four let-ters must be lacking at the end of eachline, is not necessarily correct (211).The name is clearly incomplete, but(regardless of whether or not thename was $hw nkt) it is far more likely


    that the inscription continued onto afourth line, now lost (212). The textruns perfectly from the existing endof line 1 to the beginning of line 2, asit does from lines 2 to 3. The chancesof this happening by coincidencemust be extremely remote.

    (iii) Ansary, Qaryat al-Fau: 147, no. 6 isalso ANA for similar reasons, eventhough here the 3rd person masc. sg.enclitic pronoun alternates between-h and -hw. As already explained, thedivine name $l-$hw wr, while itselflinguistically Old Arabic, cannot, ofcourse, be used to identify the lan-guage of the text.

    (iv) Ansary, Qaryat al-Fau: 143/2, whichreads w$l bn sw qn d $l ntn; and possibly

    (v) Ja 2142 (213) which is very fragmen-tary.

    Similarly, two texts from Nagran whichcontain the phrases d$t $hl and d $hl respec-tively, should probably also be classed asPure Undifferentiated North Arabian.These are(vi) Mu 2 (214)(vii) Ja 859 (215).