The Northern Origins of Old South Arabic Literacy

by George E. Mendenhall
(The University of Michigan and Yarmouk University)

Yemen Update 33 (1993):15-19

Among the specialists on the history of alphabets, there is general agreement that the complex of pre-Islamic Arabic alphabets separated from the Canaanite system about the time of the transition between the Late Bronze and Early Iron Ages, i.e., ca. 1200 B.C. (Cross 1979; Naveh 1982). The implications of these agreements seem to have been missed by some students of Old South Arabic who still derive the ONA alphabets from the Yemeni scripts.

Actually, the situation, as usual, is much more complex historically than present theories would have us believe. There is now very good reason to believe that there was already in the Late Bronze age a systematic contrast between what became the Phoenician and Arabic alphabetic complexes in the course of the Iron Age. The same is true of the languages themselves, but that is subject matter for another discussion. Here we deal simply with the writing systems.

All recent discussions of the history of Levantine alphabets attempt to develop a seriation of forms, based entirely (and necessarily of course) upon the present inventory of alphabetic inscriptions. So far the inventory of Bronze Age alphabetic inscriptions is far too small to justify the type of rigidseriation that the philologians have created. Instead, as is true everywhere else where evidence is available, in the formative period of writing systems there is a plethora of local alphabets within a given region, all related to each other, but rarely derivable one from another. Thus we have at least 32 different early Greek alphabets, about 14 early Etruscan, and it remains to be seen how many Bronze Age Semitic local alphabets existed.

Of one fact we can be sure: there was no such thing as the "standard Canaanite alphabet" until the 10thCentury B.C., (pace Cross) and even that conclusion rests upon very slender evidence indeed. Furthermore, that "standard Canaanite alphabet" existed (if at all) for little more than a century before it broke down again into a variety of local alphabets. The consternation produced by the alphabetic forms of the T. Fekheriyehstele is an amusing example of the fact that theories based upon evidence derived from a very limited geographical and cultural milieu(Palestine & Lebanon) are wrecked on the hard evidence of inscriptions from a different cultural region. One scholar went so far as to proclaim the inscription a "fake" because its letter forms did not conform to the orthodox theory.

Mentioned only in passing or ignored entirely is the extremely important small corpus of syllabic inscriptions from Byblos for which I published a decipherment in1985. Reviewers have been so shocked by the content that emerged that they have virtually ignored the purely formal ties between the syllabic system and the later alphabets. Those ties are valid entirely independent of the question concerning the accuracy of the translation of those very difficult texts.

There has been no convincing argument against Dunand's dating of the syllabic inscriptions no later than the Egyptian Middle Kingdom period, i.e., before about 1800 B.C. The internal evidence that the inscriptions yield is entirely compatible with that dating, indeed demands a dating before the Hyksos period. There is good reason to believe that they originated well before 1800B.C. and did not survive into the so-called Hyksos period of Egypt, the MB II period in Palestinian archaeology.

Others (Cross 1979; Naveh 1982) have collected the evidence for the MB II-LB I alphabetic inscriptions, and so far no inscriptions in the syllabic system later than about1800 B.C. have been discovered. Though some works on the history of the alphabet still cite the Balu'ah stele as a later example of syllabic writing (Naveh: 1982), Ward and Martin (1964) years ago exploded that myth by demonstrating that it is merely a very badly eroded Egyptian hieroglyphic inscription. I was very gratified with this conclusion, since years of work on the Byblos Syllabic inscriptions had already convinced me that there was no connection with the Balu'ah stele. An attempt to date the Byblos Syllabic corpus to the LB age by alleging syllabic signs on the reverse of analphabetic spatula is simply illusory (Mendenhall 1985: p.6).

Nineteen of the twenty-nine characters of the Sabaean alphabet have prototypes in the Byblos Syllabic system. Of those nineteen, six are signs that are not used at all in the various inscriptions regarded as Canaanite, but have the same consonantal values in these early syllabic inscriptions. In addition, several signs that do derive from the same source as the Canaanite characters have forms that are not likely to have been "borrowed” from the LB Age Canaanite alphabet, but do derive from a earlier common prototype.

Some of these signs thus are demonstrated to have their origin in the Levant before the existence of any known alphabet and nearly a millennium before any writing at all is known in the Old South Arabic milieu. Many of the letter forms of the later alphabets, whether Canaanite or proto-Arabic occur in the syllabic system, and can be presumed to have derived from it (Mendenhall 1985,Ch.4). These forms were thus inherited by the later alphabetic systems, but the syllabic system had three times as many forms as the alphabets needed, so two thirds of the syllabic characters dropped out of use as otiose. The importance of this simple fact lies in the contrast between the selection of signs that survived in the two alphabetic systems, for the two did not ultimately select the same character from the series of three that existed in the syllabic system for each consonant, namely Ca, Ci, and Cu.

What much later became the mainstream Canaanite (or Central Alphabet, to use the terminology that seems most appropriate) and the complex of pre-Islamic (or Eastern)alphabets had their major source in the Syllabic system of Byblos. Clearly this system was not the only source for both alphabetic traditions, but it is equally clear that contrary to prevailing theory, neither of the later alphabets can be derived simply from the other. They had inter-related and parallel histories until about the13th century B.C., when the Eastern alphabet tradition became almost entirely separated from the Central tradition by the process of geographical removal and social disruption.

The evidence for the historical reconstruction described above consists of three separate bodies of evidence. The first is portrayed in Table I, that shows the contrast between ONA, OSA and Central traditions in their derivations from separate signs of the Byblos signary. In addition to these nine characters that were not utilized in the Central Alphabet, there area dozen or more characters that the two alphabetic traditions share in common. With a very few exceptions, all of these letter forms are attested in the Levant (including the Sinai) in the Bronze Age. There should be little doubt that all the pre-Islamic alphabets exhibit a substantial continuity from the Syro-Palestinian complex of alphabets now attested from the 16th century B.C. to the end of the Bronze Age, and from Lebanon to the Sinai and Egypt (Cross: 1979).

The second body of evidence consists of inscriptions that are either ignored or dismissed as "by-forms" of the alphabet (Cross: 1979). The first example is the ostracon from Tell Jisr in the lower Litani River valley in the Biq`a of Lebanon(Mendenhall: 1971). (Figure 1) Though the sherd has not yet been securely dated, the double rope-molding feature is very typical of MB pottery in the middle Euphrates valley of Syria, and my colleague Dr.C. Lenzen informs me that the same features appear on MB pottery at Tell Irbid in northern Jordan. (Figure 2) The inscription is very difficult, but it clearly exhibits both the dâl and thethâ of later Eastern alphabets. If the ceramic evidence is reliable, the sherd is perhaps older than any other alphabetic inscription so far discovered.

The second example of Bronze Age inscriptions related to the Eastern Alphabetic tradition comes from Kamid el-Loz, just a few kilometers upstream from Tell Jisr (Figure3). These ostraca published by Mansfeld in 1969 are securely dated to ca. 1400 B.C., and their connections with the Eastern alphabets have already been noted, and summarily and arbitrarily dismissed by Cross (1979, p. 100): "...there is insufficient reason to assign the Kamid el-Loz sherds to the Proto-Canaanite corpus." Precisely! The prevailing theory simply dismisses the evidence that other corpora existed in the Bronze Age, and that each cultural area had its own alphabetic heritage, so to speak.

Ostracon 2 from Kamid el-Loz is virtually pure Arabic. It is true that the forms occur much later in other regions, and such occurrences simply illustrate a virtually universal principle: that relict areas typically exhibit extremely archaic features. The inscription reads from left to right (or top to bottom as the case may be): l mtry. The other ostracon is probably to be read: qr'. Both names occur in ONA inscriptions: mtr, and the fem.form mtrt, on the one hand, and qr'm, on the other. Neither name isin frequent use or has a reasonable explanation from within Arabic, which is another indication of their very archaic origins.

From the transition between LB and Early Iron Ages comes another possible connection between the Eastern Alphabet tradition and those of the North. In this case it is the still undeciphered and controversial clay tablet from Der'alla in the Jordan Valley. The character composed of a vertical stroke with a circular hole punched at its upper end is identical in form with theyod of the Eastern Alphabets, for which there has never been a convincing derivation. The vertical stroke, however, does represent/i/ in a number of Anatolian alphabets, and a very similar form occurs with similar phonetic connections in the hieroglyphic Luwiansignary. If this isomorph stood alone, it would not be convincing, but there are a number of other illustrations of connections between the Eastern alphabets and those of Anatolia (and Greece) including some evidence still unpublished.

Another indication of a continuity of alphabetic tradition in the inland area is furnished by the ON alphabets that exhibit an 'alif that cannot be derived from the OSA form. (Figure 4)

The Safaitic form that is used also in the relatively late Thamudic inscriptions occurs already in the late 7thcentury building inscription from the Umm Ruj‚m cistern, just North of Amman. Its form in multiple variations (as usual) derives from the 10th century 'alif from Byblos; there is no other reasonable explanation for it.

A third witness to the northern origin of the OSA literacy is the Beth-Shemesh tablet in Ugaritic cuneiform, which has now been proven to be an abecedary with the OSA order of letters of the alphabet, but with the typical Iron Age inventory of only 22 characters (Ryckmans 1988). It is clear that not only the forms of the Eastern alphabet characters were derived independently of the Central Alphabet from the Late Bronze Age of the Levant, but also that there existed in that region a tradition of the order of letters of the alphabet that was independent of the Canaanite order. There can now be little doubt that an Eastern Alphabet system, together with its order of letters existed in the Syro-Palestinian region in the 13th century B.C., and was subsequently transmitted and preserved in the Arabian peninsula until the rise of Islam.

For a half-century or more, there seems to have been a hermetically sealed academic compartmentalization between Ancient Near Eastern and Arabic studies. This is illustrated by two works of a past generation: first, my own teacher's presidential address at the SBL in 1938, in which he maintained that there was little or no continuity between the ancient Near East and the Islamic culture. Second, the influential and massive work by R. Dussaud on Lapénétration des arabes.... It is also amusingly illustrated by the "Instructions for Contributors to the Bulletin of the American Schools of Oriental Research" issued in 1976. This six page manual gave detailed instructions for transliteration of virtually every ancient language, but didn't even mention Arabic(BASOR 222:81-87). Both of these works, perhaps unconsciously, assumed the accuracy of the old 19th century concept of successive wave of cultureless barbarian nomads penetrating into the civilized regions of the Fertile Crescent, bringing with them nothing but another branch of the Semitic language family, the hypothesis described succinctly by J. Kupper as late as 1957 (p. xiv). This theory that underlays the concept of the common features that constitute the evidence for the entire Semitic language family is now absurd.

Those common features, especially vocabulary, were not created by the Dewey Decimal classification system of a university library, nor even by the great decipherers of the 19th century. They were created by populations in contact at some remote period that doubtless extends far back into pre-historic times. The place of contact can no longer be found in the Arabian Peninsula, but in the areas of high density of population in the Syro-Palestinian area, where sedentary village peoples are well attested from the end of the Old Stone Age into the historic period.

E. A. Knauf (1988) has presented a scenario that describes the historical processes underlying the relationships between those peoples of the Fertile Crescent, and the Semitic speaking populations of the Arabian Peninsula. He places the penetration of Semitic into the Yemen at the EB/MB transition period, but the evidence is extremely tenuous. It is a priori more likely that the population of that region at that time was related to the Hamito-Semitic peoples of Africa, and these peoples furnished the sub-stratum that introduced a number of innovations into the entire South Semitic/Ethiopic linguistic corpus. On the other hand, the evidence for the LB/EI transition is massive and detailed as pointed out above.

The history of the Semitic languages is the history of the populations of the Fertile Crescent, and that is characterized by an almost constant process of fission and fusion, the breaking down into local dialects (a language is simply a dialect with an army and a navy), and the production of a succession of lingua francas. Tracing this in more detail will be the rewarding task of future generations of scholars sensitized to the processes of historical and social linguistics.


Albright, W.F.
1939 The Ancient Near East and the Religion of Israel. JBL 59:85-112.

Cross, F. M. Jr.
1979 Early Alphabetic Scripts. Symposia Celebrating the 75th Anniversary of the Founding of the American Schools of Oriental Research. Ed. F.M. Cross, 105-111. ASOR, Cambridge, MA.

Dussaud, R.
1955 La pénétration des Arabes en Syrie avant l'Islam.

Knauf, E. A.
1988 The West Arabian place-name province: Its origin and significance. PSAS 18:39-49.

Kupper, J.
1957 Les nomades en Mésopotamie au temps des rois de Mari.

Mansfeld, Günter
1969 Deux ostraka incisés à écriture paléo-canaanéene du Tell de Kâmid el-Lôz. BMB 22 (1969), 67-75.

Mendenhall, G.E.
1971 A New Chapter in the History of the Alphabet. BMB 24, The Syllabic Inscriptions From Byblos. Beirut, AUB Press.
1988 The Syro-Palestinian Origins of Pre-Islamic Arabic. In Studies in the History and Archaeology of Palestine, III, pp. 215-224.
1989 Arabic in Semitic Historical Linguistics. PSAS 19:

Naveh, J.
1982 The Early History of the Alphabet.

Ryckmans, J.
1988 A.G. Lundin's interpretation of the Beth Shemesh abecedary: A presentation and commentary. PSAS 18:123-129.

Ward,W.A., and Martin,M.F.
1964 The Balu'a Stele. ADAJ 8-9 (1964), pp. 5-29.

Search Site

Search Library Collection