Definiteness Marking in Earlier Stages of Arabic Areal Evidence from Classical Arabic Idols’ Names

This study investigates definiteness marking in earlier stages of Arabic, based on a corpus of Classical Arabic idols’ names attested in technical prose. The latter mentions not only Arabic idols, the names of which bear definite markers, but also the tribal units that worshipped them. Relying upon attestations of this sort, the present study investigates the areal distribution of definiteness marking in earlier stages of Arabic. Finally, it compares the results of such an investigation with the current knowledge on the areal distribution of definiteness marking in earlier stages of Arabic, based on epigraphic and dialectological evidence.


Topic and Aim
This paper investigates definiteness marking as attested in some Classical Arabic idols' names that plausibly reflect earlier stages of the language. The investigation concentrates on the areal distribution of these idols' names, and in particular of the definite markers they bear. The idols' names that are the object of this investigation are attested in literary primary sources, a category broadly understood here as including any kind of non-epigraphic written source, technical prose included (e.g. lexicography). In particular, the paper considers literary primary sources that associate Classical Arabic idols' names with tribal units, which allows for a better understanding of their areal distribution. Although the main focus of this paper is on literary primary sources, the importance of the results achieved by Semitic linguistics and Arabic dialectology in the study of definiteness marking in earlier stages of Arabic, and therefore the advantages of an integrated approach, will also appear in due course.

Scope and Limits
Some philological caution is needed in dealing with the onomastic corpus selected for this study, on the levels of both methodology and data. On the level of methodology, an investigation of Classical Arabic idols' names in terms of areal distribution implies a resort to basic descriptive units of linguistic geography, such as 'dialect' and 'locality', that are anachronistic with respect to the literary primary sources. Rabin (1951, 15) suggests that this difficulty may be partially overcome by considering the terms luġa and qabīla (or whatever tribal unit), found in technical prose, as approximate equivalents of 'dialect' and 'locality' respectively. 1 But only rarely in technical prose is a luġa defined along the lines of a modern dialect, in terms of some core or distinctive linguistic features. A case in point is the phonological feature that literary primary sources (cited by Rabin 1951, 10;al-Sharkawi 2017, 48) ascribe to the dialect of Tamīm and label ʿanʿana 'pharyngealization of hamza' (that is, a shift from ʾ toʿ). Likewise, when literary primary sources ascribe a linguistic feature to a qabīla, or other tribal unit, they rarely locate it in a well-defined place using a toponym. Al-Hamdānī's (d. 334/946) mention of the toponym Tihāma in connection with the Ḥakam and their usage of the definite marker (a)m-is a quite isolated case (Rabin 1951, 44). Typically, the primary sources simply ascribe a linguistic feature to a tribal unit, as is shown in the following example (in which the linguistic feature is 'articled' and 'nunated' forms related to šams-, as a whole, as an authentic instance of heterogeneity. From a diachronic perspective, this instance of heterogeneity in all likelihood dates back to pre-Islamic times, as evidenced by the fact that the Arabic lexicographers, among them Ibn Sīda, trace the 'nunated' form šamsin back to a remote Yemenite past (represented by the ancestor Sabaʾ).
Accordingly, this study will mainly take into account, among the pre-Islamic idols' names transmitted in the sources, those characterised by heterogeneity. Another instance of heterogeneity discussed in the literature is the definite article (a)m-, a Yemeni variant of l-, which Rabin (1951, 34-7) also illustrates using lexicographical sources. However, because none of the Classical Arabic idols' names that have been thoroughly studied in the literature bears a definite article (a)m-, this potential pre-Islamic variant of l-will not be dealt with here.

Methodology
This study investigates definiteness marking in the earlier (pre-Islamic) stages of Arabic, with particular reference to its areal distribution, through an onomastic corpus that results from the usual stages of lexicographical work: the definition, selection, collection, and classification of descriptive units. These stages are illustrated in what follows.

Definition of Descriptive Units
This stage of lexicographical work establishes a pre-Islamic idol's name as the basic descriptive unit, based on two criteria.

Unambiguous Pre-Islamic Referent
The first criterion is a semantically oriented or, more precisely, a referent-oriented, criterion of dating: a Classical Arabic name is regarded here as genuinely pre-Islamic if it unambiguously denotes a pre-Islamic referent. Names of idols indeed satisfy this criterion. A word of caution, however, is in order. Since this criterion is semanterms of objecthood: see the previous footnote. Also, their semantic equivalence in terms of definiteness requires an additional consideration: the definite meaning shared by šamsin and l-šamsa has a dedicated locus of realisation on the level of form, as will be clarified in the course of this study.
tic, it can only be used to confirm the pre-Islamic dating of an idol's name on the level of meaning (i.e. names of idols are, by definition, pre-Islamic). It says nothing about the authenticity of an idol's name on the level of form, which should be assessed on the grounds of heterogeneity, as discussed above.

Speaking Name
The second criterion is the semantic category of 'speaking name'. The sources record idols' names that are not only proper names, but also clearly denote animals, plants, or stars. In this respect, a pre-Islamic idol's name is often more than a theonym -it is also a zoonym, a phytonym, or a star name. 14 A pre-Islamic idol's name that exhibits a double semantic nature of this sort in synchrony is a speaking name. For instance, the pre-Islamic instance of heterogeneity l-šamsa, šamsin, discussed at the outset of this study, consists of lexical variants of a speaking name, to the extent that their reference to a star ('the sun') clearly co-exists with their reference to a deity ('the Sun'); this may be inferred from the semantic context in which they occur (the noun ʿabdu and the verb ʿabada, both denoting the act of worshipping). The criterion of the speaking name is particularly relevant from a linguistic perspective for two reasons. First, this criterion highlights the concrete physical referent of a pre-Islamic idol's name, such as a zoomorphic idol (cf. al-nasr 'the vulture'), and so automatically locates it in a space, such as a shrine, thereby facilitating the description of a given idol's name in terms of areal distribution. Second, in an idol's name that has the form of a speaking name, definiteness marking appears to be overt and easily observable. In fact, on the one hand, in its function as a proper name (theonym), a pre-Islamic idol's name unambiguously conveys definiteness, as is exemplified, again, by the aforesaid pair of lexical variants l-šamsa, šamsin in the sense of 'the Sun'. 15 On the other hand, in its function as a common name (zoonym, phytonym, star-name), the same idol's name may exhibit a form that marks definiteness through an overt element that is usually observed in the pre-Islamic prestigious forms attested in the Koran: through the dedicated definite marker (or 'definite article') l-. Thus, since the lexical variant l-šamsa is still semantically transparent in the sense of 'the sun', it clearly shows that the well-known 14 Since the nineteenth century, this data has nourished a totemic approach to pre-Islamic religion; see Robertson Smith 1885and, contra, Zaydān 1906. A recent study on this question is Dirbas 2019.
15 For simplicity's sake, in this paper the English translation of idols' names does not include any attempt to represent the case-marker. strategy of definiteness marking through l-may also apply to a pre-Islamic idol's name.
Similarly, in its function as a common name, an idol's name corresponding to a speaking name may mark definiteness through another overt element that is perhaps less known: a definite marker -n. This is shown by the application of the commutation test to the aforesaid pair l-šamsa, šamsin, again in the sense of 'the sun'. This test reveals that, ceteris paribus, with respect to a constant surrounding context that consists of the lexical stem šams and an object case-marker, it is possible to replace the sound l-with the sound -n, with no change in the meaning of definiteness. It follows that in the pair l-šamsa, šamsin, definiteness is associated with, or marked by, l-in one case and -n in the other. The first result is in line with the traditional interpretation of l-as a definite marker for Arabic nouns in general, while the second result prompts an extension of this interpretation to -n as well. However, the interpretation of -n as a definite marker cannot be generalised, as it rests upon a definite reading of the nouns that exhibit this element. In the pre-structuralist era, Brockelmann (1908-13, 1: 466-74) must be credited as the first scholar to have provided a definite interpretation of -n along these restrictive lines.

Residual Issues
Overall, the criteria of the unambiguous pre-Islamic referent and of the speaking name define the descriptive unit of this study as a class of proper names of Arabic idols whose original concrete meaning (animal, plant, star) is still palpable, and whose intrinsic definiteness is conveyed by means of the markers l-or -n. For the sake of completeness, the corpus of this study will also include idols' names that comply with the first criterion only and whose pre-Islamic authenticity is thus less secure. The proper names of idols falling into this class do not necessarily denote animals, plants, or stars, nor do they exhibit a definite marker. An interesting fact concerning such idols' names is that their intrinsic definiteness may co-occur with diptotism. The idol's name manātu, attested as an hapax in the Koranic text, exemplifies the weak class of idols' names thus defined: wa-manāta l-ṯāliṯata l-uḫrā (and Manat the third, the other; Koran LIII,20). 16 In particular, in this class, the intrinsic definiteness of an idol's name correlates with the lack of a definite marker attached to it. For practical purposes, this phenomenon will be referred to here as a 'zero definite marker'. 17 16 Arberry's translation. This idol is traditionally said to have the form of a stone.
17 It is important to bear in mind that this kind of marker co-occurs with a diptotic case-marker (manātu, manāta). To put it briefly, the descriptive unit of this study is an idol's name that is genuinely pre-Islamic semantically and that tends to convey definiteness heterogeneously, through l-, -n, or a zero marker.

Francesco Grande Definiteness Marking in Earlier Stages of Arabic
This descriptive unit is graphically rendered here in vocalised transcription, in order to clearly signal the instances of definiteness marking consisting of -n and zero (with the latter covert marker being inferred from the overt diptotic case-marker).

Selection of Descriptive Units
The selection of the descriptive units is mainly based on two criteria illustrated in the previous sections: the philological criterion of heterogeneity and an approximate version of the dialectological criterion of locality, in which a tribal unit may signify a locality (or, more precisely, the area semi-nomadically inhabited by the tribal unit). Accordingly, this study considers only those idols' names that in the sources appear to be heterogeneous on the level of form and that are mentioned in connection with tribal units (and possibly toponyms). A third selection criterion is practical: the idols' names of this sort should be attested in the sources in the form of lists. This criterion eases the retrieval of data.

Collection of Descriptive Units
The descriptive units are mainly drawn from two sources, which each present a list of idols' names. One source is the thematic dictionary al-Muḫaṣṣaṣ (The Categorised [Lexicon]) by the Andalusi lexicographer Ibn Sīda (d. 458/1066), 18 the other is the genealogical work Ǧamharat Ansāb al-ʿArab (The Extensive Genealogies of the Arabs) by the Andalusi polymath Ibn Ḥazm (d. 456/1064). 19 Ibn Sīda offers a list of thirty-two idols' names, including their lexical variants, which are indicative of dialectal heterogeneity, in compliance with the first selection criterion (al-Muḫaṣṣaṣ, 13: 104-5); whereas Ibn Ḥazm offers a list of twenty-five idols' names, often mentioning the tribal units that worshipped the idols these names refer to, in compliance with the second selection criterion (Ǧamharat Ansāb al-ʿArab, 2: 491-4). 20 Sometimes, Ibn Ḥazm also mentions a toponym in connection with the idol worshipped by a given tribal unit, again in compliance with the second selection criterion. To a certain extent, the idols' names and related tribal units or toponyms included in Ibn Ḥazm's list are also attested in the antiquarian work al-Muḥabbar (The Adorned [Treatise]) by Ibn al-Kalbī's pupil Ibn Ḥabīb (d. 245/859), 21 but the exact nature of this intertextual link has not yet been assessed with certainty (Tritton 1964, 472). 22 In one case, notably the idol's name l-filsu/l-falsu, heterogeneity is not recorded by Ibn Sīda, but can be indirectly observed by comparing Ibn Sīda's and Ibn Ḥazm's lists, which record the forms l-filsu and l-falsu, respectively. 23 In another case, namely the idol's name ʿabʿabun/ġabġabun, the indication of the tribal unit is provided by Ibn Sīda rather than by Ibn Ḥazm. Similarly, while Ibn Sīda's list includes the idol's name l-šāriqu/šarīqun, probably meaning 'the dawn', 24 Ibn Ḥazm makes no mention of it, so this source cannot be used to locate the name geographically. However, the name of the pre-Islamic Qurayšite leader (and relative of Muḥammad) Al-Aḫnas b. Šarīq provides a cue for locating this idol's name, at least in the variant šarīq, in the neighbourhood of Mecca.
Two of the idols' names studied here must be singled out for comment. In the first case, data has been collected from another source besides Ibn Sīda or Ibn Ḥazm. For the idol's name bāǧaru/bāǧiru, data concerning both heterogeneity and tribal units have been taken from the Kitāb al-Aṣnām (The Book of Idols) by Ibn al-Kalbī (d. 204/819), since the critical edition of al-Muḫaṣṣaṣ consulted here proposes a double vocalisation for this idol's name with no principled explanation (whereas Ibn Ḥazm's list does not mention it at all). 25 In the second case, the treatment of the idol's name naṣrun/naṣṣaru (?) is partly conjectural, since it relies upon the etymological identification of the form naṣṣaru, drawn from Ibn Sīda's list, with the form naṣrun, which Ibn Ḥabīb mentions in connection with the toponym 20 However, the distinction between the two lists is not so clear-cut, as Ibn Sīda also indicates the tribal units and/or toponyms related to some idols' names, as will become apparent shortly.     Semantically, the two forms have a common referent, to the extent that they both denote the name of a king (who would have been later deified and worshipped). Concerning the form naṣrun, in fact, Ibn Ḥabīb describes the tribal unit that bears this name as kings: banī naṣrin mulūki l-ḥīra (the Banū Naṣr, the kings of al-Ḥīra; al-Muḥabbar, 369). Turning to the form naṣṣaru, Fahd (1968, 134) remarks that it is usually attested as a part of the king's name buḫt naṣṣar (or nuṣṣur), the Arabic version of Babylonian Nebuchadnezzar. Historical sources report buḫt naṣṣar (or nuṣṣur) to be the founder of an Iraqi settlement of Arabs, called ḥayr (Retsö 2003, 157): The presence of Arabs in lower Mesopotamia as early as 700 BC may be remembered even by much later historians. In al-Ṭabarī's account of the earliest history of the ʿarab he says that 'Buḫt Nuṣṣur' settled some tradesmen of the ʿarab in a ḥayr, a fenced-in camp near al-Naǧaf in Iraq. 26 Retsö (2003, 477) also remarks that the sources identify the ḥayr founded by the king buḫt naṣṣar (or nuṣṣur) with al-Ḥīra, which, it should be added here, Ibn Ḥabīb reports to have been later ruled by the aforesaid Banū Naṣr: [The purpose of the sources is] to tell about the origins of al-Ḥīra. The tendency is evident in the parallel between Nebuchadnezzar's ḥayr, the earliest Arab settlement in Iraq, and the name al-Ḥīra, and there is no doubt that the storyteller wants us to accept the identity between the two names.
If correct, the etymological identification of naṣrun with naṣṣaru satisfies the selection criteria adopted here, to the extent that it brings to light a certain heterogeneity for this king's and, later on, idol's name, as well as an accurate indication of the locality in which it was worshipped.

Classification of Descriptive Units
The classification of the idols' names reported in Ibn Sīda's and Ibn Ḥazm's lists rests on two parameters: the tribal unit or, whenever available, the toponym that is related to a given idol's name; and the kind of definite marker the idol's name bears, namely l-, -n, or zero.
26 Retsö's transliteration has been adapted to match the conventions of this paper. The latter marker co-occurs with a diptotic case-marker. What makes this kind of classification possible is an intertextual fact: Ibn Sīda's and Ibn Ḥazm's lists share seventeen idols' names, so in principle, for each of the names in this subset a definite marker can be paired with geographical information. 27 These seventeen idols' names form the onomastic corpus of this study, which will yield only provisional results because of its limited size. 28 The seventeen idols' names shared by the two lists can be compiled in a separate list, which uses the following abbreviations concerning heterogeneity and their sources: 27 Ibn Sīda and Ibn Ḥazm were coevals and each spent part of his scholarly life at the court of Denia. These biographical facts suggest that a personal exchange of knowledge between them may explain the intertextual commonalities between the two lists of idols' names they drew up. However, neither Andalusi scholar explicitly mentions the other when spelling out the sources of his list. This topic warrants further investigation, which lies beyond the scope of this paper.
28 A recent study by Bellino (2018) suggests that the corpus of idols' names under investigation may be expanded. When dealing with the idol Hubal, Bellino (2018, 118) observes that in the maġāzī literature, "The style of naming is generally al-Hubal (with article)". In pre-Islamic times, this variant probably co-existed with the diptotic variant hubalu, attested in al-Muḫaṣṣaṣ, 13: 104 and other sources. If so, the resulting pair l-hubal(u)/hubalu would be an instance of heterogeneity. This state of affairs raises the possibility of an areal description of the idol's name l-hubal(u)/hubalu, provided that the tribal units associated with its variants in the sources are better understood. For instance, Ibn Ḥazm mentions several tribal units in connection with the variant hubalu (cf. Ǧamharat Ansāb al-ʿArab, 2: 492) and, broadly speaking, the maġāzī literature includes references not only to idols' names, but also to the tribal units worshipping them: "Some of the ġazawāt narrate expeditions against tribes that own or worship specific idols" (Bellino 2018, 120).

Francesco Grande
Definiteness Marking in Earlier Stages of Arabic 31 tested in the Koran), the critical edition of al-Muḥabbar reads suwāʿu instead. 29 In both cases, differences in case-marking amount to an alternation between, on the one hand, a triptotic case-marker that co-occurs with the definite marker -n (bāǧa/irun, suwāʿun) and, on the other hand, a diptotic case-marker that co-occurs with a zero definite marker (bāǧa/iru, suwāʿu). Precisely the non-prestigious nature of this alternation is good evidence in favour of interpreting it as an instance of authentic heterogeneity rather than of interpolation. Had the author or copyist interpolated the original text, he would have introduced the prestigious definite marker l-, which is, however, lacking in this alternation. It should be added here that the critical editions of al-Muḫaṣṣaṣ, Kitāb al-Aṣnām, and al-Muḥabbar are not dependent on each other, at least with respect to the two idols' names under scrutiny.
As regards al-Muḥabbar, its editor Lichtenstädter explicitly states, in connection with the idol's name suwāʿu, that the established text reflects the original manuscript. 30 Turning to al-Muḫaṣṣaṣ, its editors ʿAbdah and al-Šanqiṭī were unable to consult the manuscript of Kitāb al-Aṣnām later discovered by Zaki Pasha. The latter, in turn, did not avail himself of the critical edition of al-Muḫaṣṣaṣ by ʿAbdah and al-Šanqiṭī when establishing the text of Kitāb al-Aṣnām, since he never mentions Ibn Sīda among his Arabic sources. 31 On these grounds, the following itemisation of the seventeen idols' names under scrutiny incorporates all the case-markers observa- 29 See the list below for detailed references. These differences in case-marking do not seem to be typographical errors. In 2006 Hindāwī issued a new edition of al-Muḫaṣṣaṣ that is based on the same Cairo manuscript as the critical edition consulted here, by ʿAbdah and Al-Šanqiṭī (the Cairo manuscript is the only integral manuscript of this work). Although Hindāwī's edition is not critical, it nonetheless avowedly aims at correcting typographical errors and, generally speaking, at improving the quality of ʿAbdah and Al-Šanqiṭī's critical edition (al-Muḫaṣṣaṣ, 1: 7, ed. Hindāwī). As in ʿAbdah and Al-Šanqiṭī's edition, Hindāwī's edition reads bāǧiru and suwāʿun (al-Muḫaṣṣaṣ, 6: 145, ed. Hindāwī).
30 See al-Muḥabbar, 369. In this passage, the manuscript and the critical edition read ʾn swāʿ in scriptio defectiva. It is evident that this passage attests the name of the idol in question without a case-marker. Still, it is equally evident that a diptotic case-marker is easily inferred from the particle ʾn (inna) that precedes the form swāʿ, as the consonantal ductus of the latter only allows for the scriptio plena suwāʿa. A classification of the idols' names attested in Ibn Sīda's and Ibn Ḥazm's lists, which is carried out according to the two parameters of tribal unit or toponym and form of definiteness marking, results not only in a reduced list of seventeen items, but also in three major categories that basically represent three different facets of one and the same phenomenon -namely, the areal distribution of a given kind of pre-Islamic definite marker. The first, second, and third categories illustrate the areal distribution of l-, -n and zero, respectively. Each category consists of entries that generally have the form of a pair: idol's name; tribal unit(s). In addition, the entry may include, whenever available, a toponym, enclosed in brackets.
A different manner of classifying these lexical materials concentrates on the tribal unit or, sporadically, the toponym, and presents all the attested kinds of pre-Islamic definite markers within the area represented by a given tribal unit or toponym. This category consists of entries having the form of a pair: tribal unit (or toponym); idol's name (or idols' names). Such entries are arranged in geographical order from north to south. The four categories are illustrated in detail in what follows. An asterisk indicates a tribal unit that is not included on Rabin's (1951, ii) map. Capital letters, as explained above, indicate the sources attesting to a given idol's name. For simplicity's sake, the idols' names included in the four categories are present- ed in full vocalization with no indication of the critical editions that exhibit their case markers. For this aspect, the reader is referred to the full list of seventeen idols' names presented above.
34 Only the lexical variant šarīqun is associated with a tribal unit. It is likely that the worship of l-šāriqu was widespread among all Arabs, as may be inferred from the following statement by Ibn Durayd: wa-qad sammat al-ʿarabu ʿabda l-šāriq (the Arab used the proper name ʿabdu l-šāriq; al-Ištiqāq, 1: 305).
35 Only the lexical variant nasrun is associated with a tribal unit.
36 Ibn al-Kalbī, the source mentioning these tribal units, reports that each of them used both lexical variants (bāǧarun, bāǧirun).

Results and Discussion
From the observation of the previously defined categories (which basically consist of sets of pairs), and especially of the fourth one, some patterns of areal distribution emerge. They remain tentative because of the limited size of the corpus.
37 Ibn al-Kalbī, the source mentioning these tribal units, reports that each of them used both lexical variants (bāǧaru, bāǧiru).
38 Only the form naṣrun, which is regarded here as a probable lexical variant of naṣṣaru, is associated with a tribal unit and a toponym.
39 In their lists of idols' names, neither Ibn Sīda nor Ibn Ḥazm mentions the form l-ʿuzzā in connection with Mecca. Nonetheless, the idol's name l-ʿuzzā can be associated with this toponym on the basis of the tribal unit mentioned by Ibn Ḥazm, the Qurayš, as well as on the basis of the toponym Naḫla (i.e. Naḫlat al-Šāmiyya), which Ibn Ḥazm reports to be the place of worship of this idol: Naḫlat al-Šāmiyya is a location near Mecca. Another factor that renders these patterns tentative is the availability of geographical information. In fact, not all of the tribal units that Ibn Sīda's and Ibn Ḥazm's lists associate with idols' names, and the related definite markers, are among the tribal units that Rabin (1951) placed on his tribal map of the Arabian peninsula (see Map 1). The location of some tribal units named in the lists thus remains uncertain. However, this drawback is partly compensated by the ancient toponyms that the same lists associate with some idols' names, since the position of these toponyms on the map of the Arabian peninsula is generally well known.
Likewise, the location of some tribal units mentioned in Ibn Ḥazm's (or Ibn Sīda's) list has been the object of careful investigation, as exemplified by Kister's (2002, 21-2) description of the Banū Usayyid's settlement in Yamamah (Najd).
The tentative patterns resulting from the examination of the categorised corpus of idols' names can be explicated according to two different viewpoints, which are referred to here as the 'static' and the 'dynamic' scenarios.

The 'Static' Scenario
The areal distribution of the different kinds of definite markers, as they are attested in the idols' names, may be analysed without considering the communication structure 41 of the Arabian peninsula at their time. The resulting picture is static, and the analysis can pay particular attention to the criterion of locality, in which case three main patterns of areal distribution seem to emerge. They are quantitatively sizeable enough to qualify as plausible, as they involve six, five, and four tribal units, or toponyms. In order of quantitative importance, these patterns are the clustering of the zero definite marker (co-occurring with diptotism) in Southern Hijaz; the coexistence (or competition) between the definite marker -n and the zero definite marker, with apparently no areal clustering; and the clustering of the definite marker l-on the border between Northern/Central Hijaz and Northern/Central Najd. Turning the focus of analysis to the criterion of heterogeneity, an interesting phenomenon appears to manifest itself only in Mecca. This is the clustering of all three kinds of definite markers that are attested in the corpus of idols' names. Such a unique and concomitant attestation of the three definite markers is likely to be a pattern in and of itself. A fourth pattern of areal distribution, therefore, emerges in Mecca. The four tentative patterns sketched out so far are discussed in detail in what follows, and illustrated in Map 2. Idols' names are represented on the map as numbers according to the above itemisation. Other symbols include L and N, for an idol's name bearing, respectively, the definite marker l-or -n, as well as a box, for an idol's name bearing the zero definite marker that co-occurs with the diptotic marker.

Northern/Central Hijaz -Northern/Central Najd Border
The clustering of the definite marker l-is presumably located on the border between Northern/Central Hijaz and Northern/Central Najd. This pattern involves four localities, all corresponding to tribal units: Ṭayyiʾ + Ġaṭafān + Ġanī + Bāhila l-filsu, l-falsu + l-ʿuzzā + l-ʿuzzā + l-ʿuzzā

The 'Dynamic' Scenario
The communication structure of a major geographical region consists of all the observable centres of radiation, as well as their diffusion areas. The areal distribution of the different kinds of definite markers, as they are attested in the idols' names, may be investigated by taking into consideration the communication structure of the pre- Islamic Arabian peninsula at their time. A survey of this sort would yield a dynamic picture of their areal distribution. In practice, however, in the sources consulted here, direct clues to a communication structure that might have facilitated the lexical transfer of pre-Islamic idols' names are, perhaps expectedly, very scanty. A case in point is Ibn Ḥazm's brief description of a battle and a migration, in consequence of which a tribal unit carried with it an idol and, as it were, the latter's name. In fact, after stating that yaġūṯu kāna li-madḥiǧ (the Madḥiǧ had an idol named Yaġūṯ; Ǧamharat Ansāb al-ʿArab, 2: 492), Ibn Ḥazm adds: fa-qātalat-hum ʿalay-hi banū ʿuẓayfin ḥattā harabū bihi ilā naǧrān (the Banū ʿUẓayf struggled with them for [the custody of] that idol, so that they [scil. the Madḥiǧ] fled to Naǧrān bearing it with them; Ǧamharat Ansāb al-ʿArab, 2: 492).
Given the scarcity of similar accounts, the only alternative left for attaining some knowledge of the communication structure of pre-Islamic Arabia is to gather indirect clues. In particular, it is proposed here to glean them from the pre-Islamic religious practice of sidāna, which literally means 'shrine-keeping'. In basic terms, a sādin was at once the minister of an idol's shrine and an older inhabitant of the area in which that idol was located. By contrast, the idol's worshippers may represent subsequent ethnic strata, especially when they belong to a tribal unit other than the sādin's.
Interestingly, Robertson Smith ([1890-91] 1995), 47) illustrates this practice by citing the Madḥiǧ and their dramatic historical vicissitudes involving the idol Yaġūṯ, as reported above: In Arabia we do not find priests at every sacred spot, but only where there is a temple with treasure and equipments and especially an idol (waṯan, ṣanam). The names used for priests show this (sādin, ḥāǧib). The priesthood was hereditary in certain families, whose property the sanctuary was, and this was often a noble family, for it was noble families, we are told, who had idols of their own. In some cases, it was a family foreign to the tribe that held the land, a relic of older inhabitants. Such families had difficulties in maintaining their privilege. For the idol Yaġūṯ there was a battle. 42 On these grounds, the toponym or tribal unit that the sources describe as the milieu or the ministers of a particular practice of sidāna roughly corresponds, from a linguistic perspective, to the place of origin of a given idol's name, and especially of the definite marker carried by it. That is, technically speaking, the milieu or the ministers of a particular practice of sidāna corresponds to a centre of radia-42 The transliteration system used by Robertson Smith ([1890-91] 1995) and by the modern editor of his Lectures, Day, has been adapted to match the conventions of this paper. tion of the definite marker carried by an idol's name. Conversely, all the remaining toponyms and tribal units that the sources associate with the same idol's name with no mention of the practice of sidāna can be tentatively regarded as points along a line of communication, such as a trade route or a seasonal migration route, or as localities in a diffusion area through which that idol's name and the definite marker carried by it spread because of mutually related social factors that were typical of the pre-Islamic Lebensform: semi-nomadism, trade, religion, and so on.
Therefore, some fragments of the communication structure of pre-Islamic Arabia, which revolve around its definite markers, may in principle emerge if the researcher introduces the following simple distinction within the tribal units, or toponyms, attested in Ibn Ḥazm's and, sporadically, Ibn Sīda's list: tribal units or toponyms that textually co-occur with the root s-d-n, versus those that co-occur with other roots, such as ʿ-b-d 'to worship' or k-w-n followed by li-, in the sense of 'to have', and the like.
In practice, the researcher fine-tunes the already existing classification of descriptive units in two steps. First, he focuses on any entry in the first three categories that includes more than one tribal unit (or toponym). Then, within an entry of this kind, he isolates the tribal unit (or toponym) that is associated with the root s-d-n from those that are not.
This simple distinction reshapes the pair of which such an entry originally consists into a triad: idol's name; tribal unit(s) associated with s-d-n; and tribal unit(s) associated with ʿ-b-d, k-w-n or similar roots. 43 Besides this triad, the entry may include, whenever available, a toponym enclosed in brackets. Table 1 illustrates the reformulation of the first three categories according to the criterion of sidāna. 43 In the sources consulted here, the root k-w-n may be followed by the prepositions bi, fī, which are in turn followed by a toponym or a tribal unit, in the usual sense of 'to be at'. While this instance of k-w-n is useful for locating idols' names in the Arabian peninsula in the study of the 'static' scenario, as discussed above, it is not relevant to the 'dynamic' scenario under scrutiny, as it may ambiguously refer to a centre of radiation as well as to a diffusion area. Accordingly, in what follows only the root k-w-n followed by li ('have') will be taken into account. For simplicity's sake, the construction 'k-w-n followed by li' will be referred to hereafter as k-w-n.  -ẓ-m: tuʿaẓẓimu-hā (they worship [scil. al-ʿUzzā]; Ǧamharat Ansāb al-ʿArab, 2: 491). ** In this case, the act of worship is expressed through the noun ʿabd rather than through a verb derived from the same root, and it is implied in the personal name ʿabd šams 13: 104). *** The root is q-r-r: aqarru-hu (they installed [scil. Yaġūṯ]; Ǧamharat Ansāb al-ʿArab, 2: 492).

Francesco Grande Definiteness Marking in Earlier Stages of Arabic
Under the 'dynamic' scenario, two major tentative patterns of areal distribution seemingly emerge, each consisting of a centre of radiation and a related diffusion area. They can be inferred from the geographical information gathered from the sources consulted here. In both of these patterns the centre of radiation is, as discussed above, a place related to a given idol's sidāna, whereas its diffusion area corresponds to the areal clustering of tribal units and/or toponyms. One tentative pattern consists of a centre of radiation of the zero definite marker that co-occurs with diptotism and is clustered in Southern Hijaz. Another tentative pattern consists of a centre of radiation of the definite marker l-in Northern Hijaz that is clustered at the border between Northern/Central Hijaz and Northern/Central Najd. Actually, from the geographical information available, another centre of radiation of the definite marker l-also seems to emerge, which is perhaps located in Central Hijaz, notably in Mecca, but its diffusion area, if any, is hard to discern.
If not only geographical but also genealogical information is considered, a third pattern seems to show up, which is more tentative than the previous ones. Such a pattern consists of a centre of radiation of the definite marker -n, perhaps located in Yemen, with its diffusion area in Najd.
The three tentative patterns sketched out so far are discussed in detail in what follows, and illustrated in Map 3. In Southern Hijaz, a centre of radiation of the zero definite marker, co-occurring with diptotism, corresponds to the tribal unit Madḥiǧ, whereas the clustering of the same co-occurring markers is observed in its neighbourhoods and specifically in four localities. All but one of these localities is a tribal unit (the only exception being Naǧrān), and all four correspond to the diffusion area of the centre of radiation under scrutiny. Another possible centre of radiation of these cooccurring markers corresponds to the tribal unit Huḏayl, although it is not entirely clear whether it is located in Central or Southern Hijaz. Nor is it clear whether the Huḏayl are also a centre of radiation of -n: In Northern Hijaz, a centre of radiation of the definite marker l-corresponds to the tribal unit Banū Bawlān, a tribal subdivision of the Ṭayyiʾ, whereas the clustering of the same marker is observed in its neighbouring regions, and specifically in three localities at the border between Northern/Central Hijaz and Northern/Central Najd. These localities actually are all tribal units, and they correspond to the diffusion area of the centre of radiation under scrutiny. For simplicity's sake, in what follows the Banū Bawlān are represented as Ṭayyiʾ: Ṭayyiʾ + Ġaṭafān + Ġanī + Bāhila l-filsu, l-falsu + l-ʿuzzā + l-ʿuzzā + l-ʿuzzā

Centre
The relation between the centre of radiation under scrutiny and the other conjectural centre of radiation of the same marker, which is associated with the idol's name l-ʿuzzā and is perhaps located in Naḫla, in the proximity of Mecca, is not clear. Nor is it clear whether the tribal units of Ḫaṯʿam and Ḥāriṯ (which are both associated with the idol's name ḏū l-ḫalaṣati) represent the diffusion area of the latter centre of radiation.

Centre of Radiation of the Definite Marker n-
According to Ibn Ḥazm, the tribal unit responsible for the sidāna of the idol Šams (šumsun) was the Banū Usayyid, whose genealogy he reports as follows: sadanatu-hā min banī usayyidin bni ʿamrin bni tamīm (the ministers of Šams were [chosen] among the Banū Usayyid, the son of ʿAmr, the son of Tamīm; Ǧamharat Ansāb al-ʿArab, 2: 493). 44 In the genealogical tradition, two brothers sharing the eponym Tamīm, namely Arāšah and Ġawṯ, are said to have migrated to Yemen, and Ġawṯ is also said to be of Yemenite descent through his moth-44 That the Banū Usayyid worshipped Šams in the form of an idol is apparent from Ibn Ḥabīb's description: kāna la-hu bayt (it was a betyle; al-Muḥabbar, 317). er, a woman of the Ǧurhum. 45 According to Kister (1965, 156), the genealogical details concerning the Yemenite migration or origin of Arāšah and Ġawṯ "point clearly to their connection with South Arabia". The question, therefore, arises whether similar information is available concerning the Tamīm and his descent: the eponym ʿAmr and especially the eponym Usayyid. Ibn al-Kalbī provides an affirmative answer to this question, reporting that Tamīm married Sulmā, a Yemenite woman; and that their son ʿAmr in turn married ʿAmra, a Yemenite woman as well, who gave birth to Usayyid:  Kister (1965, 154 Interestingly, Ibn al-Kalbī traces back ʿAmra's tribal unit, the Baǧīla, to the Yemenite, or Southern, ancestor Qaḥṭān (Nasab Maʿadd wa-l-Yaman al-Kabīr, 1: 131-3): The genealogical reference to Qaḥṭān implies a location in Yemen, since Ibn al-Kalbī explicitly states that some of Qaḥṭān's descendants, among which he counts the Baǧīla, tayāmanū (settled in Yemen; Nasab 1: 133).
In light of Kister's (1965) historicist interpretation of genealogical sources, the Yemenite grand-maternal and maternal lineage of the Banū Usayyid amounts to saying that they have a 'double', so to speak, South Arabian origin. This historicist interpretation is plausibly grounded on some epigraphical evidence: the Yemenite matrilineality of the Banū Usayyid reported by genealogists finds a historical parallel in one ancient South Arabian inscription (third century CE) that describes a maternal lineage of descent (Korotayev 1995, 91-2).
Therefore, while the sādin status of the Banū Usayyid is likely to indicate that their settlement in Yamamah (which has been briefly discussed above) is particularly old, their genealogically grounded South Arabian origin strongly suggests that such a settlement is the consequence of a migration from Yemen to Najd. In turn, a migration scenario of this sort may be indicative of a radiation of linguistic features, among them -n, from Yemen, and a diffusion area including Najd -and perhaps Southern Hijaz, to judge from Map 3.
The existence of other centres of radiation of -n besides Yemen is not clear. Ibn Ḥazm's description of the idol named suwāʿu, suwāʿun suggests that the tribal unit Huḏayl, in their quality of sadanatu-hu (its ministers; ǧamharat Ansāb al-ʿArab, 2: 493) may constitute a centre of radiation not only of -n but also of the zero definite marker (co-occurring with diptotism). Moreover, while the centre of radiation in question may be said to be located somewhere between Central and Southern Hijaz, its exact location is unknown at the present stage of research.

An Integrated Approach
So far, this study has investigated the areal distribution of definiteness marking in earlier stages of Arabic by examining a limited corpus of pre-Islamic idols' names drawn from primary literary sources, especially technical prose. This section places the main results of this investigation within a broader perspective by comparing them with results concerning the areal distribution of definiteness marking in earlier stages of Arabic that derive from Semitic linguistics and Arabic dialectology. 51 51 One reviewer has raised some philological issues on the levels of both methodology and data, and because of their broad epistemological scope it seems useful to consider them here. While these issues are valuable in principle, they are not central to the heuristic and interdisciplinary approach pursued in this paper. They can be summarised as follows: (I) The only opposition in terms of definiteness marking observed in the consonantal ductus of the idols' names is the presence or absence of the l-marker. The n-and diptotic markers are not found at all in the consonantal ductus. (II) The presence of the n-marker in the idols' names is also questionable, because it implies a case-marker (un/an/in) that was already lost or decaying in pre-Islamic times. (III) A study of manuscripts is required to assess the real form of the idols' names bearing no l-marker, i.e. to assess whether they end in an n-marker or a diptotic marker. (IV) The theoretical background of definiteness marking in Arabic and Semitic is not mentioned in the paper. The following is a critical discussion of these issues: (I) The reviewer assumes that the only opposition in terms of definiteness marking observed in the consonantal ductus of the idols' names is the presence or absence of the lmarker, i.e. A vs statement of non-A. But what is really observed in the consonantal ductus is the opposition of the l-marker vs some unknown marker, i.e. A vs non-statement of A. The consonantal ductus does not record the absence of the n-marker or of the diptotic marker, or any other marker. It is simply underspecified in this respect. What is opposed to the l-marker is not lack of l-, but some unknown marker. The question, therefore, arises how to determine the unknown marker(s). The most straightforward and honest answer, which is consistent with the heuristic approach of this paper, is cautious conjecturing. Convergence between the data recorded by Arabic linguistic tradition and the epigraphic and dialectal data is itself proof that data from the linguistic tradition has a certain degree of authenticity, as will become clear throughout this section. A second answer is heterogeneity, as illustrated at the outset of this paper. The conception of a member of an opposition as either negative (statement of non-A) or underspecified (non-statement of A), as illustrated immediately above, owes much to Jakobson's original formulation of markedness theory. (II) The loss of a case-marker, or of its functional yield, is sensitive to sociolinguistic contexts. Pre-islamic idols' names may have retained their case-markers, whether dip- The first main result of the present investigation is the identification of a clustering of the definite marker l-at the border between Northern/Central Hijaz and Northern/Central Najd, which indicates that this marker diffused from a centre of radiation located in Northern Hijaz. 52 Briefly, it confirms the origin of l-in Northern Hijaz. Such a result is to a good extent consistent with the results independently arrived at through the study of definiteness marking in North Arabian inscriptions, which no doubt attest to pre-Islamic stages of Arabic. Al-Jallad (2018, 3) affirms that "Pre-Islamic ʾl-dialects attested in various scripts", by which he refers to epigraphical dialects, 53 are "attested in various places, but concentrated in North Arabia and southern Levant". In the same vein, he adds (12): the ʾl-and ʾ-articles are concentrated in the north and northwest, in the Ḥigāz (Dadanitic), the southern Levant (Nabataean).
The second result of the present investigation is the identification of a centre of radiation of the definite marker -n in Yemen, whose diffusion area includes Najd. Briefly, it confirms the Yemenite origin of -n. Such a result is to a good extent consistent with the results independently arrived at in Semitic linguistics through comparison of the Epigraphic South Arabian definite marker -n with peripheral intotic or co-occurring with the n-marker, for religious reasons, such as the formulaic usage of ritual language, at a moment when case-markers, generally speaking, were already lost or decaying. (III) Copyists may have adjusted manuscripts to conform to the rules of Classical Arabic, just as grammarians and lexicographers had done before them when collecting data, and as modern scholars do when preparing a critical edition. Thus, a particular linguistic feature directly observed in a manuscript, such as an n-marker, does not necessarily reflect a genuine fact of language. Moreover, the collation and comparison of several manuscripts to determine the original reading itself involves external human intervention. It follows that resorting to manuscripts is not necessarily a conclusive method when assessing whether the idols' names that bear no l-marker end in an nmarker or in a diptotic marker. (IV) The avowed aim of the paper is descriptive. The rich debate and wide range of definitions concerning definite markers and case-markers in Classical Arabic and Semitic languages are the product of several theoretical orientations, in both historical and general linguistics, that one may be interested in validating or falsifying. However, this paper positions itself in an (ideally) pre-theoretical stage of research, which consists in bringing to light data so far unknown: the linguistic dimension of the pre-Islamic idols, which have been investigated so far mostly in their religious and cultural dimensions. The linguistic dimension crucially includes their areal distribution.
52 The hypothesis that Mecca (more accurately, Naḫlat al-Šāmiyya) is part of this area as (another) centre of radiation has been touched upon above. This hypothesis does not seem to have much effect on this result, Mecca (or Naḫlat al-Šāmiyya) being located in Central Hijaz.

53
The relationship between such epigraphical dialects and the ancient Arabic dialects described in Arabic linguistic tradition, and attested centuries later, is not clear at the present stage of research. stances of the definite marker -n, or definite nunation, in Classical Arabic (which is exemplified by the aforementioned form šamsan 'the sun'). In this field of study, the Epigraphic South Arabian definite -n is usually regarded as the ancestor of Classical Arabic definite (and indefinite) nunation, based on chronological and distributional reasons -primarily the productivity of Epigraphic South Arabian definite -n, as opposed to the peripherality of the same marker in Classical Arabic. For instance, Zaborski (2000, 32-3) remarks: in other Epigraphic South Arabian languages it is -n, for example hgr-n 'the city', and this -n is normally interpreted as nunation which was originally used for determination The third result of the present investigation is the identification of a clustering in Southern Hijaz of a particular instance of the diptotic marker, one that correlates with definite meaning or, according to a different formulation, that co-occurs with a zero definite marker. Southern Hijaz thus corresponds to the diffusion area of a centre of radiation of the diptotic marker plus the zero definite marker, which at the present stage of research can be approximately identified at least with the Madḥiǧ, a tribal unit (also) settled in Southern Hijaz. The fourth result of the present investigation clarifies the dialectal status of the diptotic marker plus the zero definite marker: this is the coexistence (or competition) between this kind of marker and the definite marker -n within a broad area which is not limited to, but also includes Southern Hijaz, as shown by the tribal unit Banū Kināna and by the toponym Naǧrān. Briefly, the third and fourth results confirm the Southern Hijazi origin of Classical Arabic diptotism, and its coexistence (or competition) with -n in the same area (as well as in others).
These results are rather consistent with what is currently known of the modern dialects spoken in that region. Greenman (1979) offers a description of the present-day Arabic dialect of Central Yamani Tihamah, which is spoken in the Al-Hudaydah governorate, in the coastal part of Southern Hijaz. In particular, Greenman (1979, 60-1) reports that in this dialect indefinite masculine nouns, or noun-like items, such as participles, exhibit the indeclinable ending -u, also in pause: e.g. ana nāširu msōq (I am leaving the market), štuktub kitābu (will you write a book?) (in pause). 54 He also reports that a sub-dialect, spoken in the Wadi Mawr area, which is situated about one hundred kilometres north-northeast of the city of Al-Hudaydah, exhibits the indeclinable ending -un instead of u in the same distributional contexts: e.g. ana nāširun ǝmsōq (I am leaving the market). Before Greenman's (1979) study, scholars had partially observed these phe-54 Greenman's transliteration has been adapted to match the conventions of this paper. nomena, as pointed out by Rabin (1951, 57), who mentions previous research on the field by Rossi and Landberg:

Francesco Grande Definiteness Marking in Earlier Stages of Arabic
The present-day colloquial of the old ʿAzd country, in Hodeida and part of the Tihāma, still pronounces nouns in the indeterminate state with a final -u: they say burru 'wheat', but al-burr 'the wheat'.
According to Blau (2006, 27, 29), the phonological realisation of this dialectal ending as -u, as well as its syntactic inability to co-occur with the article, 55 constitutes an argument strong enough to identify it, notwithstanding its indeclinability, with the Classical Arabic diptotic marker. Blau (29) even considers such a dialectal ending as the ancestor of Classical Arabic diptotism: 56 It seems not unreasonable to assume that Yemenite dialects reflect the original situation and that in Classical Arabic its use was limited to proper nouns, in which diptosy is especially frequent. Blau's (29) assumption that the diptotic-like ending -u of this dialect, which is ultimately attested in Southern Hijaz, may "reflect the original situation" of Classical Arabic diptotism, is to a good extent consistent with the third result of the present investigation -the Southern Hijazi origin of Classical Arabic diptotism. Moreover, "The u/un alternation which may be observed", according to Greenman (1979, 54), "in the masculine form of this [scil. indefinite nominal] paradigm" (cf. the aforementioned pair nāširu/nāširun) is to a good extent consistent with the fourth result of the present investigation -the coexistence of (or competition between) the definite diptotic marker and the definite marker -n.
However, as far as the Southern Hijazi origin of Classical Arabic diptotism and its coexistence (or competition) with -n are concerned, the correspondence between the results of the present investigation and those already existing in the literature is less straightforward than the previous correspondences concerning the origin of l-in Northern Hijaz and the Yemenite origin of -n. The reason for this is that the correspondence requires more theoretical elaboration than the previous ones, especially on a diachronic level. In Blau's (2006) study, the developmental pattern from (the ancestor of) a modern Yemenite dialect to Classical Arabic diptotism is based on external linguistic evidence, coming from Nabatean Arabic. Thus, it comes as no surprise that Greenman (1979, 60) assumes the reverse developmental pattern, considering the dia-55 In Blau's (27) own words: "when lacking the definite article" diptotic nouns "behave as if they were indefinite." 56 Blau (28-9) also adduces a comparative argument, based on a phonological parallel with Nabatean Arabic, that is not relevant here. lectal ending -u to be "a residue" of the Classical Arabic case system. Furthermore, while in the idols' names investigated here the diptotic marker correlates with definite meaning, in the Yemenite data recorded by Greenman (1979) and his predecessors, the diptotic-like ending -u (that is, a marker having the same syntactic distribution as the Classical Arabic diptotic marker) correlates with indefinite meaning.
Finally, a fifth result of the present investigation is the concomitant attestation, in Mecca, of l-, -n along with the zero-marker that co-occurs with diptotism. This result is to a good extent consistent with the well-known koineization theory relative to the beginnings of Classical Arabic, according to which Mecca is the main core of this process. 57

Conclusions
This study provides additional evidence, drawn from primary literary sources, that the definite markers l-, -n and the zero-marker that co-occurs with diptotism originated in well-defined areas of the pre-Islamic Arabian peninsula. It also provides a unified areal view of their origin, seemingly bringing to light a quadripartite division of the pre-Islamic Arabian peninsula with regard to the rise of definiteness marking: l-in Northern Hijaz, -n in Southern Hijaz, the zero-marker plus diptotism in Yemen, and a murky picture of the original definite marker, if any, in Najd. In fact, from an areal perspective, the 'birthplace', as it were, of a definite marker is its centre of radiation, but Najd does not qualify as such at the present stage of research and is treated exclusively as a diffusion area. Nor can the origin and areal distribution of the definite marker (a)m-be clarified here, as noted at the beginning of this paper. The results summarised so far pertain to the linguistic materials themselves. A final result of this study pertains to methodology, specifically the value of the reappraisal of Classical technical prose, even though it raises the philological problem of the authenticity of its data. In particular, the results of this study encourage a reappraisal not only of lexicographical sources, but also of sources showing a direct or indirect antiquarian interest in pre-Islamic times, among which are al-Muḥabbar and genealogical works. The linguistic materials investigated here highlight that a reappraisal of technical prose is warranted by its valuable attestation of heterogeneity (linguistic variation), a fundamental aspect of any realistic language description.