The Proto-Semitic language is the hypothetical proto-language ancestral to the historical Semitic languages of the Middle East. Its original Urheimat has been proposed to be in northern Mesopotamia, the Arabian Peninsula and the Levant, with a 2009 study proposing that it may have originated about 7400 BC. Semitic is considered a subfamily of the larger Afroasiatic macrofamily.
- 1 Dating
- 2 Homeland
- 3 Phonology
- 4 Comparative vocabulary and reconstructed roots
- 5 See also
- 6 Notes
- 7 References
- 8 Sources
- 9 External links
The earliest attestations of a Semitic language are in Akkadian, dating to around the 23rd century BC (see Sargon of Akkad) and the Eblaite language, but earlier evidence of Akkadian comes from personal names in Sumerian texts around 28th century BC. Researchers in Egypt also claim to have discovered Canaanite snake spells: "The passages date from between 2400 to 3000 BC and appear to be written in Proto-Canaanite, a direct ancestor of biblical Hebrew."
Semiticists have put importance in locating the Urheimat of the Proto-Semitic language since all modern Semitic languages can be traced back to a common ancestor. The Urheimat of the Proto-Semites cannot be determined without considering the larger Afro-Asiatic family to which it belongs. The previously popular Arabian Urheimat hypothesis has been largely abandoned since the region could not have supported massive waves of emigration before the domestication of camels in the second millennium BC.
Out of the Levant hypothesis
Some geneticists and archaeologists have argued for a back migration of proto-Afroasiatic speakers from Western Asia to Africa as early as the 10th millennium BC. They suggest the Natufian culture might have spoken a proto-Afroasian language just prior to its disintegration into sub-languages. The hypothesis is supported by the Afroasiatic terms for early livestock and crops in both Anatolia and Iran. Recent Bayesian analysis identified an origin for Proto-Semitic language in the Levant around 7400 BC and for Semitic around 3800 BC, with a later single introduction from South Arabia into the Horn of Africa around 800 BC.
Out of Africa hypothesis
According to the proponents of this theory, the non-Semitic toponyms preserved in Akkadian and Palaeosyrian languages suggest that Syria and Mesopotamia were originally inhabited by a non-Semitic population. Edward Lipiński believes that support for an African origin is provided by what he describes as a possible relationship between Afroasiatic and the Niger–Congo languages, whose Urheimat probably lies in Nigeria–Cameroon. He states that the most numerous isoglosses and lexicostatistical convergences link proto-Semitic to Libyco-Berber and that proto-Semitic speakers were still living in the Neolithic Subpluvial in the 5th millennium BCE when the Sahara was much wetter, retaining a link with Berber long after other Egyptic and Proto-Chadic separated.
According to Lipiński, rock art attests a Neolithic in the Sahara that collapsed due to desertification and climate change around 3500 BC, forcing proto-Semitic speakers to emigrate en masse through the Nile Delta to Western Asia. They were probably responsible for the collapse of the Ghassulian culture in Palestine around 3300 BC. Another indication of the arrival of the proto-Semitic culture is the appearance of tumuli in 4th and 3rd millennium Palestine, which were typical characteristic of Neolithic North Africa. It is possible that at this point, the ancestors of the speakers of Elamite moved towards Iran, although the inclusion of Elamite in Afroasiatic is only contemplated by a tiny minority. The earliest wave of Semitic speakers entered the Fertile Crescent via Palestine and Syria and eventually founded the Akkadian Empire. Their relatives, the Amorites, followed them and settled Syria before 2500 BC. The collapse of the Bronze Age in Palestine led the southern Semites southwards, where they reached the highlands of Yemen after 20th century BC. Those crossed back to the Horn of Africa between 1500–500 BC.
The reconstruction of Proto-Semitic (PS) was originally based primarily on Arabic, whose phonology and morphology (particularly in Classical Arabic) is extremely conservative, and which preserves as contrastive 28 out of the evident 29 consonantal phonemes. Thus, the phonemic inventory of reconstructed Proto-Semitic is very similar to that of Arabic, with only one phoneme fewer in Arabic than in reconstructed Proto-Semitic. As such, Proto-Semitic is generally reconstructed as having the following phonemes (as usually transcribed in Semitology):
|Nasal||*m [m]||*n [n]|
|Stop||voiceless||*p [p]||*t [t]||*k [k]||*ʼ [ʔ]|
|voiced||*b [b]||*d [d]||*g [ɡ]|
|emphatic||*ṭ [tʼ]||*q [kʼ]|
|voiceless||*ṯ [θ]||*š [ʃ]
|*ś [ɬ]||*ḫ [x]||*ḥ [ħ]||*h [h]|
|voiced||*ḏ [ð]||*z [z]||*ġ [ɣ]||*ʻ [ʕ]|
|emphatic||*ṱ [θʼ]||*ṣ [sʼ]||*ṣ́ [ɬʼ]|
|Approximant||*l [l]||*y [j]||*w [w]|
The fricatives *s, *z, *ṣ, *ś, *ṣ́, *ṱ may also be interpreted as affricates (/t͡s/, /d͡z/, /t͡sʼ/, /t͡ɬ/, /t͡ɬʼ/, /t͡θʼ/), as is discussed below.
- Dental stops d t ṭ
- Velar stops g k ḳ (normally written g k q)
- Dental sibilants z s ṣ
- Interdental ð θ tʼ (normally written ḏ ṯ ṱ)
- Lateral l ɬ ɬʼ (normally written l ś ṣ́)
The probable phonetic realization of most consonants is straightforward and is indicated in the table with the IPA. Two subsets of consonants, however, deserve further comment.
The sounds notated here as "emphatic" occur in nearly all Semitic languages as well as in most other Afroasiatic languages, and they are generally reconstructed as glottalized in Proto-Semitic.[nb 1] Thus, *ṭ, for example, represents [tʼ]. (See below for the fricatives/affricates).
In modern Semitic languages, emphatics are variously realized as pharyngealized (Arabic, Aramaic, Tiberian Hebrew (such as [tˤ]), glottalized (Ethiopian Semitic languages, Modern South Arabian languages such as [tʼ]), or as unaspirated (Turoyo of Tur-Abdin such as [t˭]); Ashkenazi Hebrew and Maltese are exceptions and all emphatics merge into plain consonants under the influence of Indo-European languages (Italian/Sicilian for Maltese, German/Yiddish for Hebrew).
An emphatic labial occurs in some Semitic languages, but it is unclear whether it was a phoneme in Proto-Semitic.
- Hebrew developed an emphatic /ṗ/ phoneme to represent unaspirated /p/ in Iranian and Greek.
- Ge'ez is unique among Semitic languages for contrasting all three of /p/, /f/, and /pʼ/. While /p/ and /pʼ/ occur mostly in loanwords (especially from Greek), there are many other occurrences whose origin is less clear (such as hepʼä 'strike', häppälä 'wash clothes').
The reconstruction of Proto-Semitic has nine fricative sounds that are reflected usually as sibilants in later languages, but whether all were already sibilants in Proto-Semitic is debated:
- Two voiced fricatives that *ð, *z eventually became, for example, *z for both in Hebrew, but /ð/ and /z/ in Arabic, respectively
- Four voiceless fricatives
- *θ (*ṯ) that became *š in Hebrew but /θ/ in Arabic
- *š (*s₁) that became *š in Hebrew but /s/ in Arabic
- *ś (*s₂) that became *ś in Hebrew but /š/ in Arabic
- *s (*s₃) that became (*)/s/ in both Hebrew and Arabic
- Three emphatic fricatives (*θ̣, *ṣ, *ṣ́)
The precise sound of the Proto-Semitic fricatives, notably of *š, *ś, *s and *ṣ, remains a perplexing problem, and there are various systems of notation to describe them. The notation given here is traditional and is based on their pronunciation in Hebrew, which has traditionally been extrapolated to Proto-Semitic. The notation *s₁, *s₂, *s₃ is found primarily in the literature on Old South Arabian, but more recently, it has been used by some authors to discus Proto-Semitic to express a noncommittal view of the pronunciation of the sounds. However, the older transcription remains predominant in most literature, often even among scholars who either disagree with the traditional interpretation or remain noncommittal.
The traditional view, as expressed in the conventional transcription and still maintained by some of the authors in the field is that *š was a Voiceless postalveolar fricative ([ʃ]), *s was a voiceless alveolar sibilant ([s]) and *ś was a voiceless alveolar lateral fricative ([ɬ]). Accordingly, *ṣ is seen as an emphatic version of *s ([sʼ]) *z as a voiced version of it ([z]) and *ṣ́ as an emphatic version of *ś ([ɬʼ]). The reconstruction of *ś ṣ́ as lateral fricatives (or affricates) is certain although few modern languages preserve the sounds. The pronunciation of *ś ṣ́ as [ɬ ɬʼ] is still maintained in the Modern South Arabian languages (such as Mehri), and evidence of a former lateral pronunciation is evident in a number of other languages. For example, Biblical Hebrew baśam was borrowed into Ancient Greek as balsamon (hence English "balsam"), and the 8th-century Arab grammarian Sībawayh explicitly described the Arabic descendant of *ṣ́, now pronounced [dˤ] in the standard pronunciation or [ðˤ] in Bedouin-influenced dialects, as a pharyngealized voiced lateral fricative [ɮˤ]. (Compare Spanish alcalde, from Andalusian Arabic اَلْقَاضِي al-qāḍī "judge".)
The primary disagreements concern (1) whether the sounds were actually fricatives in Proto-Semitic or whether some were affricates and (2) whether the sound designated *š was pronounced [ʃ] (or similar) in Proto-Semitic, as the traditional view posits, or had the value of [s]. The issue of the nature of the "emphatic" consonants, discussed above, is partly related (but partly orthogonal) to the issues here as well.
With respect to the traditional view, there are two dimensions of "minimal" and "maximal" modifications made:
- In how many sounds are taken to be affricates. The "minimal affricate" position takes only the emphatic *ṣ as an affricate [tsʼ]. The "maximal affricate" position additionally posits that *s z were actually affricates [ts] [dz] while *š was actually a simple fricative [s].
- In whether to extend the affricate interpretation to the interdentals and laterals. The "minimal extension" position assumes that only the sibilants were affricates, and the other "fricatives" were in fact all fricatives, but the maximal update extends the same interpretation to the other sounds. Typically, that means that the "minimal affricate, maximal extension" position takes all and only the emphatics are taken as affricates: emphatic *ṣ θ̣ ṣ́ were [tsʼ tθʼ tɬʼ]. The "maximal affricate, maximal extension" position assumes not only the "maximal affricate" position for sibilants but also that non-emphatic *θ ð ś were actually affricates.
Affricates in Proto-Semitic were proposed early on but met little acceptance until the work of Alice Faber (1981) who challenged the older approach. The Semitic languages that have survived often have fricatives for these consonants. However, Ethiopic languages and Modern Hebrew, in many reading traditions, have an affricate for *ṣ.
The evidence for the various affricate interpretations of the sibilants is direct evidence from transcriptions and structural evidence. However, the evidence for the "maximal extension" positions that extend affricate interpretations to non-sibilant "fricatives" is largely structural because of both the relative rarity of the interdentals and lateral obstruents among the attested Semitic language and the even-greater rarity of such sounds among the various languages in which Semitic words were transcribed. As a result, even when the sounds were transcribed, the resulting transcriptions may be difficult to interpret clearly.
The narrowest affricate view (only *ṣ was an affricate [tsʼ]) is the most accepted one. The affricate pronunciation is directly attested in the modern Ethiopic languages and Modern Hebrew, as mentioned above, but also in ancient transcriptions of numerous Semitic languages in various other languages:
- Transcriptions of Ge'ez from the period of the Axumite Kingdom (early centuries AD): ṣəyāmo rendered as Greek τζιαμω tsiamō.
- The Hebrew reading tradition of ṣ as [ts] clearly goes back at least to medieval times, as shown by the use of Hebrew צ (ṣ) to represent affricates in early New Persian, Old Osmanli Turkic, Middle High German etc. Similarly, Old French c /ts/ was used to transliterate צ: Hebrew ṣɛdɛḳ "righteousness" and ʼārɛṣ "land (of Israel)" were written cedek, arec.
- There is also evidence of an affricate in Ancient Hebrew and Phoenician/Punic ṣ. Punic ṣ was often transcribed as ts or t in Latin and Greek or occasionally Greek ks; correspondingly, Egyptian names and loanwords in Hebrew and Phoenician use ṣ to represent the Egyptian palatal affricate ḏ (conventionally described as voiced but possibly instead an unvoiced ejective).
- Aramaic and Syriac had an affricated realization of *ṣ until some point, as is seen in Classical Armenian loanwords: Aramaic צרר 'bundle, bunch' → Classical Armenian 'crar' /tsɹaɹ/.
The "maximal affricate" view, applied only to sibilants, also has transcriptional evidence. According to Kogan, the affricate interpretation of Akkadian s z ṣ is generally accepted.
- Akkadian cuneiform, as adapted for writing various other languages, used the z- signs to represent affricates. Examples include /ts/ in Hittite, Egyptian affricate ṯ in the Amarna letters and the Old Iranian affricates /t͡ʃ d͡ʒ/ in Elamite.
- Egyptian transcriptions of early Canaanite words with *z, *s, *ṣ use affricates (ṯ for *s, ḏ *z, *ṣ).
- West Semitic loanwords in the "older stratum" of Armenian reflect *s *z as affricates /tsʰ/, /dz/.
- Greek borrowing of Phoenician 𐤔 *š to represent /s/ (compare Greek Σ), and 𐤎 *s to represent /ks/ (compare Greek Ξ) is difficult to explain if *s then had the value [s] in Phoenician, but it is quite easy to explain if it actually had the value [ts] (even more so if *š had the value [s]).
- Similarly, Phoenician uses 𐤔 *š to represent sibilant fricatives in other languages rather than 𐤎 *s until the mid-3rd century BC, which has been taken by Friedrich/Röllig 1999 (pp. 27–28) as evidence of an affricate pronunciation in Phoenician until then. On the other hand, Egyptian starts using s in place of earlier ṯ to represent Canaanite s around 1000 BC. As a result, Kogan assumes a much earlier loss of affricates in Phoenician, and he assumes that the foreign sibilant fricatives in question had a sound closer to [ʃ] than [s]. (A similar interpretation for at least Latin s has been proposed by various linguists based on evidence of similar pronunciations of written s in a number of early medieval Romance languages; a technical term for this "intermediate" sibilant is voiceless alveolar retracted sibilant.)
There is also a good deal of internal evidence in early Akkadian for affricate realizations of s z ṣ. Examples are that underlying ||*t, *d, *ṭ + *š|| were realized as ss, which is more natural if the law was phonetically ||*t, *d, *ṭ + *s|| → [tts], and that *s z ṣ shift to š before *t, which is more naturally interpreted as deaffrication.
Evidence for *š as /s/ also exists but is somewhat less clear. It has been suggested that it is cross-linguistically rare for languages with a single sibilant fricative to have [ʃ] as the sound and that [s] is more likely. Similarly, the use of Phoenician 𐤔 *š, as the source of Greek Σ s, seems easiest to explain if the phoneme had the sound of [s] at the time. The occurrence of [ʃ] for *š in a number of separate modern Semitic languages (such as Neo-Aramaic, Modern South Arabian, most Biblical Hebrew reading traditions) and Old Babylonian Akkadian is then suggested to result from a push-type chain shift, and the change from [ts] to [s] "pushes" [s] out of the way to [ʃ] in the languages in question, and a merger of the two to [s] occurs in various other languages such as Arabic and Ethiopian Semitic.
On the other hand, it has been suggested that the initial merged s in Arabic was actually a "hissing-hushing sibilant", presumably something like [ɕ] (or a "retracted sibilant"), which did not become [s] until later. That would suggest a value closer to [ɕ] (or a "retracted sibilant") or [ʃ] for Proto-Semitic *š since [ts] and [s] would almost certainly merge directly to [s]. Furthermore, there is various evidence to suggest that the sound [ʃ] for *š existed while *s was still [ts]. Examples are the Southern Old Babylonian form of Akkadian, which evidently had [ʃ] along with [ts] as well as Egyptian transcriptions of early Canaanite words in which *š s are rendered as š ṯ. (ṯ is an affricate and the consensus interpretation of š is [ʃ], as in Modern Coptic.)
Diem (1974) suggested that the Canaanite sound change of *θ → *š would be more natural if *š was [s] than if it was [ʃ]. However, Kogan points out numerous objections: *s was then [ts] so the shift from *θ to *š is the most likely merger, regardless of the exact nature of *š at the time.
Evidence for the affricate nature of the non-sibilants is based mostly on internal considerations. Ejective fricatives are quite rare cross-linguistically, and when a language has such sounds, it nearly always has [sʼ] so if *ṣ was actually affricate [tsʼ], it would be extremely unusual if *θ̣ ṣ́ was fricative [θʼ ɬʼ] rather than affricate [tθʼ tɬʼ]. According to Rodinson (1981) and Weninger (1998), the Greek placename Mátlia, with tl used to render Ge'ez ḍ (Proto-Semitic *ṣ́), is "clear proof" that this sound was affricated in Ge'ez and quite possibly in Proto-Semitic as well.
The evidence for the most maximal interpretation, with all the interdentals and lateral obstruents being affricates, appears to be mostly structural: the system would be more symmetric if reconstructed that way.
The shift of *š to h occurred in most Semitic languages (other than Akkadian, Minaian, Qatabanian) in grammatical and pronominal morphemes, and it is unclear whether reduction of *š began in a daughter proto-language or in Proto-Semitic itself. Some thus suggest that weakened *š may have been a separate phoneme in Proto-Semitic.
Correspondence of sounds with daughter languages
See Semitic languages#Phonology for a fuller discussion of the outcomes of the Proto-Semitic sounds in the various daughter languages.
Correspondence of sounds with other Afroasiatic languages
See table at Proto-Afroasiatic language#Consonant correspondences.
Comparative vocabulary and reconstructed roots
- Afroasiatic languages
- History of the alphabet
- Proto-Afroasiatic language
- Proto-Indo-European language
- That explains the lack of voicing distinction in the emphatic series, which would be unnecessary if the emphatics were pharyngealized.
- Kitchen, A.; Ehret, C.; Assefa, S.; Mulligan, C. J. (29 April 2009). "Bayesian phylogenetic analysis of Semitic languages identifies an Early Bronze Age origin of Semitic in the Near East". Proceedings of the Royal Society B: Biological Sciences. 276 (1668): 2703. doi:10.1098/rspb.2009.0408.
- Milstein, Mati (5 February 2007). "Ancient Semitic Snake Spells Deciphered in Egyptian Pyramid". news.nationalgeographic.com. Retrieved 10 April 2017.
- Lipiński 2001, pp. 42
- Dziebel, German (2007). The Genius of Kinship: The Phenomenon of Human Kinship and the Global Diversity of Kinship Terminologies. Cambria Press. p. 366. ISBN 978-1-934043-65-3. Retrieved 13 August 2013.
- Nöth, Winfried (1 January 1994). Origins of Semiosis: Sign Evolution in Nature and Culture. Walter de Gruyter. p. 293. ISBN 978-3-11-087750-2. Retrieved 13 August 2013.
- Quantitative Approaches to Linguistic Diversity: Commemorating the Centenary of the Birth of Morris Swadesh. p. 73.
- Kitchen, A; Ehret, C; Assefa, S; Mulligan, CJ. (2009). "Bayesian phylogenetic analysis of Semitic languages identifies an Early Bronze Age origin of Semitic in the Near East". Proc Biol Sci. 276 (1668): 2703–10. PMC . PMID 19403539. doi:10.1098/rspb.2009.0408.
- Lipiński 2001, pp. 43
- Lipiński 2001, pp. 44
- Blench 2006, p. 96
- Versteegh, Cornelis Henricus Maria "Kees" (1997). The Arabic Language. Columbia University Press. p. 13. ISBN 978-0-231-11152-2.
- Sáenz-Badillos, Angel (1993) . "Hebrew in the context of the Semitic Languages". A History of the Hebrew Language (Historia de la Lengua Hebrea). trans. John Elwolde. Cambridge, UK: Cambridge University Press. pp. 18–19. ISBN 0-521-55634-1.
- Dolgopolsky 1999, p. 29.
- Hetzron 1997, p. 147.
- Woodard 2008, p. 219.
- For an example of an author using the traditional symbols but subscribing to the new sound values, see Hackett, Joe Ann. 2008. Phoenician and Punic. The Ancient Languages of Syria-Palestine and Arabia (ed. Roger D. Woodard). Likewise, Huehnengard, John and Christopher Woods. 2008. Akkadian and Eblaite. The Ancient Languages of Mesopotamia, Egypt, and Aksum (ed. Roger D. Woodard). p. 96: "Similarly, there was a triad of affricates, voiced /dz/ (< z >) voiceless /ts/ (< s >), and emphatic /tsʼ/ (< *ṣ >). These became fricatives in later dialects; the voiceless member of this later, fricative set was pronounced [s] in Babylonian, but [š] in Assyrian, while the reflex of Proto-Semitic *š, which was probably simple [s] originally, continued to be pronounced as such in Assyrian, but as [š] in Babylonian." Similarly, an author remaining undecided regarding the sound values of the sibilants will also use the conventional symbols, for example, Greenberg, Joseph, The Patterning of Root Morphemes in Semitic. 1990. p. 379. On language: selected writings of Joseph H. Greenberg. Ed. Keith M. Denning and Suzanne Kemme: "There is great uncertainty regarding the phonetic values of s, ś, and š in Proto-Semitic. I simply use them here as conventional transcriptions of the three sibilants corresponding to the sounds indicated by samekh, śin, and šin respectively in Hebrew orthography."
- Lipiński, Edward. 2000. Semitic languages: outline of a comparative grammar. e.g. the tables on p.113, p.131; also p.133: "Common Semitic or Proto-Semitic has a voiceless fricative prepalatal or palato-alevolar š, i.e. [ʃ] ...", p.129 ff.
- Macdonald, M.C.A. 2008. Ancient North Arabian. In: The Ancient Languages of Syria-Palestine and Arabia (ed. Roger D. Woodard). p. 190.
- Blau, Joshua (2010). Phonology and Morphology of Biblical Hebrew. Winona Lake, Indiana: Eisenbrauns. p. 25–40.
- Ferguson, Charles (1959), "The Arabic Koine", Language, 35 (4): 630, doi:10.2307/410601.
- Versteegh, Kees (1997), The Arabic Language, Edinburgh University Press, ISBN 90-04-17702-7
- For example, Huehnengard, John. 2008. Afro-Asiatic. In: The Ancient Languages of Syria-Palestine and Arabia (ed. Roger D. Woodard). P.229–231
- Dolgopolsky 1999, p. 33.
- Kogan, Leonid (2011). "Proto-Semitic Phonetics and Phonology". In Semitic languages: an international handbook, Stefan Weninger, ed. Berlin: Walter de Gruyter. P. 62.
- According to Kogan "Proto-Semitic Phonetics and Phonology" (2011), Steiner 1982a is the "classic exposition" (p. 62).
- Kogan, (2011), p. 63.
- Dolgopolsky 1999, p. 32.
- Kogan (2011), p. 66.
- Kogan (2011), p. 67.
- Kogan (2011), pp. 67-68.
- Kogan (2011), p. 69.
- Quoted in Kogan (2011), p. 68.
- Kogan (2011), p. 68.
- Kogan (2011), p. 70, quoting Martinet 1953 p. 73 and Murtonen 1966 p. 138.
- Kogan (2011), p. 70.
- Kogan (2011), pp. 92-93.
- Kogan (2011), p. 80.
- Dolgopolsky 1999, pp. 19, 69-70
- American Heritage Dictionary of the English Language (Fourth ed.). Boston & New York: Houghton Mifflin Company. 2000. ISBN 0-395-82517-2.
- John Huehnergard, Proto-Semitic Language and Culture, pages 2056-2059
- Semitic Roots, pages 2062-2068
- Lipiński, Edward (2001). Semitic Languages: Outline of a Comparative Grammar. Peeters Publishers. ISBN 978-90-429-0815-4. Retrieved 30 June 2013.
- Blench, Roger (2006). Archaeology, Language, and the African Past. Rowman Altamira. ISBN 978-0-7591-0466-2. Retrieved 30 June 2013.
- Huehnergard, John. (2003) "Akkadian ḫ and West Semitic ḥ." Studia Semitica 3, ed. Leonid E. Kogan & Alexander Militarev. Moscow: Russian State University for the Humanities. pp. 102-119. ISBN 978-5-728-10690-6
- Kienast, Burkhart. (2001). Historische semitische Sprachwissenschaft.
- Dolgopolsky, Aron (1999). From Proto-Semitic to Hebrew. Milan: Centro Studi Camito-Semitici di Milano.
- Hetzron; Robert (1997). The Semitic languages. Cambridge University Press. p. 572. ISBN 0-415-05767-1.
- Woodard, Roger (2008). The Ancient Languages of Mesopotamia, Egypt and Aksum. Cambridge University Press. p. 250. ISBN 0-521-68497-8.