Sumerian is a language isolate and it is highly unlikely that we will ever find a language related to it. Nearly all of the six thousand or so languages known to us today are separated from Sumerian by a time gap of thousands of years, which is all but impossible to bridge with the methods of historical and comparative linguistics. More crucially, the linguistic landscape of which Sumerian was a part is largely unknown and has long since disappeared, mostly without
1 Postgate (1992) is still the best overview of Mesopotamian society from ca. 3000-1500 BCE. Pollock (1999) is a highly influential analysis that pays more attention to prehistoric developments, treating the period 5000-2100 BCE. Historical overviews with a wider scope are Edzard (2004) and Van de Mieroop (2004). The fundamental survey of ancient settlement and land use is Adams (1981).
a trace. Linguistically, modern Iraq is three language shifts removed from Sumerian: from Sumerian to Akkadian to Aramaic to Arabic. Similar language shifts have happened in neigh-bouring Iran, Turkey, and the rest of the Middle East. Any traces of related languages have thus been obliterated thousands of years ago.
In the third millennium BCE, Sumerian was surrounded by unrelated languages. In Meso-potamia itself, Sumerian was spoken in the area closest to the Persian Gulf, while the Semitic language Akkadian was at home in the neighbouring area more upstream on the Euphrates and Tigris. Farther away, in northern Syria, we find other Semitic languages, Eblaite and Amorite.
Since the Semitic languages belong to the Afro-Asiatic language family, their ultimate origin lies in Africa, but they spread into the Middle East very early. By 2600 BCE, they had ex-panded so far into Syria and Mesopotamia that Akkadian was already used in parts of southern Mesopotamia, steadily reducing the area where Sumerian was spoken. This process of lan-guage shift in Mesopotamia did not only lead to the death of the Sumerian lanlan-guage itself, but most probably also obliterated its closest relatives (Michalowski 2000: 180).
The most important neighbour of Sumerian outside Mesopotamia was Elamite (Stolper 2004), at home in present-day Iran. The earliest documents date from about 3100 BCE but are written in an as yet undeciphered script, so that their linguistic assignment to Elamite is uncer-tain. The oldest unambiguously Elamite texts date to about 2300 BCE. Although the language is still poorly understood, enough is known to make language comparison possible. Elamite seems to be related to the Dravidian languages. The language became extinct in the late first millennium BCE, being replaced by Iranian languages. Although there were trade relations and wars between Elamite and Sumerian speaking areas, Elamite seems to have had hardly any linguistic impact on Sumerian: no Elamite loanwords have as yet been identified in Sumerian.
Elamite was not the only language spoken in the neighbouring areas of Iran in Sumerian times, but very little is known about the others (Rubio 2005: 316-7). A people called the Gutians are attested from about 2300 BCE onwards, but of their language we know little more than a few proper names. Coming from the modern area of Iran, the Kassites first turn up in Babylonia during the second millennium BCE. Of their language we know a few dozen words and proper names. Both Gutian and Kassite, as well as any other ancient pre-Indo-European language from Iran, became extinct thousands of years ago.
Directing our attention further west, we find traces of two more ancient Near Eastern lan-guages from Sumerian times: Hurrian (Wilhelm 2004) and Hattic (Klinger 1996). The former was spoken in present-day northern Iraq, northern Syria and southeast Turkey from at least the later third millennium BCE until the end of the second millennium BCE. Hurrian is closely related to Urartian, which is another ancient Near Eastern language and which was in use during the first millennium BCE in the area from the Caucasus to northeastern Iraq. Some scholars have argued for a genetic relationship between Urarto-Hurrian and Northeast Cauca-sian, but others remain unconvinced.
Hattic is a language indigenous to Anatolia. It became extinct in the early second millen-nium, being replaced by the Indo-European language Hittite. Hattic is mainly known from loanwords into Hittite and from some Hittite sources which document the use of Hattic in a cultic context. For Hattic, too, genetic links with Caucasian languages have been proposed but without decisive proof.
There is little or no evidence of linguistic contact between any of these ancient Near Eastern languages and Sumerian, with one important exception. Sumerian and Akkadian were not only neighbouring languages but there was also extensive linguistic contact between the two, so extensive in fact that we can speak of a Sumero-Akkadian linguistic area in Mesopotamia (Edzard 2000; 2003: 173-8). In a general situation of widespread bilingualism and language
shift, the two languages greatly influenced each other in their phonology, grammar, and lexi-con. Under Sumerian influence, Akkadian, for instance, acquired a ventive, lost several typic-ally Semitic consonants, shifted to a verb-final word order, and borrowed many words from Sumerian. Sumerian likewise converged with Akkadian in a number of ways. Its phoneme inventory, for example, was reduced at the expense of phonemes unknown to Akkadian. It also shifted to the general word order noun – adjective under Akkadian influence (cf. §10.4.1).
Sumerian, too, borrowed many lexical items from Akkadian.
Such lexical borrowing occurred already quite early. Sumerian actually shows several layers of loanwords borrowed from Akkadian during the third millennium. The most recent layer consists of nouns like za-ba-lum (a tree) (Cyl A 12:5; L; 22) from Akkadian supālum.
They show the Akkadian nominative suffix -um. An earlier layer consists of nouns like dam-h~~~~a-ra ‘battle’ (Ent. 28 1:26; L; 25) from Akkadian tamh~ārum. They have a suffix -a. Loan-words that are not nouns lack these suffixes -um and -a: e.g., silim ‘whole’ (Nik 1:287 2:2; L;
24) from Old Akkadian *śalim. An even older layer of endingless forms has been proposed, but remains unproven (Civil 2007; Sommerfeld 2006).
Nothing firm is known about the linguistic landscape of Mesopotamia before the third millennium. Attempts to identify a pre-Sumerian substratum have failed so far (Rubio 1999).
As it is, many of the Mesopotamian place names lack a plausible etymology. This is not really surprising, though, because the main settlements may already have been established as early as the sixth millennium BCE. We have, of course, no way of knowing which language their occupants actually spoke. But even if they spoke a language that was a direct ancestor of Sumerian and gave their settlements perfectly transparent names in this language, both this language and its place names would have changed beyond recognition during the following millennia. What we do know about these early place names, however, is that similar ones are found across the entire area of Mesopotamia. Place names with final -ar or -ur are found in both northern Mesopotamia (Nagar, Assur, Gasur) and in the south (Sippar, Nippur). This suggests some homogeneity in the earlier linguistic landscape.
1.2.2. Sources
We have written sources for Sumerian from a three thousand year long period beginning ca.
3200 BCE. This grammar is based on only a large subset of them, those dating to the second half of the third millennium BCE. Earlier texts are written in a script which is so defective that grammatical research is all but impossible. Most (if not all) later texts have been produced by scribes for whom Sumerian was not their mother tongue but only a language which they had learned during their scribal education.
The earliest texts, about 6000 in number, date to the Uruk III and Uruk IV periods (ca.
3200-3000 BCE) and are written in a somewhat rudimentary script called proto-cuneiform.
They are so difficult to decipher that Englund (2004: 101; 2009: note 18) still sees no defini-tive evidence identifying the language of these texts. But most scholars are not so sceptic and see clear proof that they are written in Sumerian (Wilcke 2005). Four centuries later, the ar-chaic texts from Ur (ca. 2800 BCE) provide the first unambiguous spellings of grammatical elements, so that it is beyond any doubt that they are in Sumerian. All these early texts are of an administrative nature, except for a few sign lists (Englund 1998).
The Fara period (ca. 2600 BCE) has yielded a much wider range of texts, with the sites of Fara (ancient Shuruppak) and Abu Salabikh providing the main text finds. In addition to the usual administrative texts, there are not only sign lists but also legal documents, incantations and numerous literary texts (Krebernik 1998). The spelling of grammatical elements has
become more common, making these texts much more comprehensible than the earlier ones.
But such spellings also remain to a large degree optional and for this reason I have excluded them from the corpus on which this grammar is based.
The earliest texts included in our corpus date from the Old Sumerian period and more specifically from the reign of Eannatum of Lagash and his contemporaries (ca. 2470 BCE) until the unification of the entire country under the kings of Akkad (ca. 2340 BCE). There are about 2200 published Sumerian texts from this period, which include administrative texts, royal and dedicatory inscriptions, a few letters and legal documents, as well as a small number of mostly fragmentary literary texts (Bauer 1998: 432-3). The majority of these sources come from a very small number of sites:
• Girsu and Lagash: well over one hundred inscriptions (ca. 2470-2350 BCE) on a variety of objects excavated from various temples and a group of more than 1700 administrative texts (ca. 2380-2350 BCE) concerning the estate of the queen of Lagash, probably excavated from the palace in Girsu.
• Nippur: about 220 inscriptions and administrative texts with a variety of backgrounds.
• Zabalam: an earlier group of about 50 administrative texts (ca. 2430 BCE) and a later group of about 100 administrative texts from the Inanna temple (ca. 2350 BCE).
• Adab: about 100 inscriptions and administrative texts.
Several other sites have yielded small numbers of texts from this period, including Ur, Umma, and Isin. It is clear, however, that most sources for the Old Sumerian period come from Girsu and Lagash, in fact almost 87 per cent of the total.
During the Old Akkadian period (ca. 2340-2200 BCE), Mesopotamia was politically unified under the kings of Akkad. This period has so far supplied about 3000 published texts in Sumerian, primarily administrative, but also a few legal documents, letters and inscriptions, with the largest groups coming from the following sites (Westenholz 1984: 17-24):
• Lagash: about 1000 mostly administrative texts from the palace archives. They record among other things the governor’s dealings with the king and the management of land, livestock, and personnel in the province of Lagash.
• Nippur: a few hundred administrative texts belonging to several different groups, including about 100 of the so-called ‘onion archive’ and two dozen from the Enlilemaba archive.
• Umma: 500-600 administrative texts from the palace archives.
• Adab: 500-600 administrative texts from the palace archives.
For the early Neo-Sumerian period (ca. 2200-2113 BCE) we need to distinguish between Lagash and the rest of the country. Outside Lagash we speak of the Guti-period, which has yielded only a few royal inscriptions. For Lagash, however, the picture is quite different. After the Old Akkadian period, the so-called second dynasty of Lagash was in power there (Lagash II, for short). This dynasty and especially its principal ruler, Gudea, has provided us with a set of royal inscriptions that has so far been the single most important text group for Sumerian grammatical studies. Apart from over two hundred regular royal and dedicatory inscriptions, this group contains 26 inscriptions on statues, including a few very substantial ones, and, above all, two clay cylinders with over 1300 lines of narrative text.
During the later Neo-Sumerian or Ur III period (ca. 2112-2004 BCE), Mesopotamia was again politically unified, this time under the Third dynasty of Ur. It was a time when the king and his representatives relied more than ever before on written documents in their management of their resources. This period has yielded over 60,000 published Sumerian texts, mostly ad-ministrative but also including about two hundred royal inscriptions, three hundred court
decisions, hundreds of letters, a few dozen incantations, and some literary texts. Most of these sources derive from a few sites and archives (Sallaberger 1999):
• Umma: over 25,000 texts from the palace archives.
• Lagash: over 13,000 texts from the palace archives.
• Drehem: over 12,000 texts from different royal archives, mostly dealing with livestock but also with treasure.
• Ur: about 3500 texts originally mostly from the royal archives but excavated from second-ary contexts where they had been discarded in antiquity.
• Nippur: about 2500 texts from smaller archives, including one of the Inanna temple and those of various individuals.
• Garshana: about 1500 texts from the archives of a member of the royal family.
The latest group of texts to be included in the corpus consists of the over 900 administrative texts from the Isin craft archive (ca. 2014-1982 BCE), which show the same spelling as the texts from the Ur III period and which overlap with them in time.
The early Old Babylonian period (ca. 2017-1722 BCE) is the last to produce unilingual Sumerian texts, most of them dating to the 18th century. It has yielded a great many administra-tive and legal documents in Sumerian, but most of all it has been the source of thousands of literary texts and fragments which have made it possible to piece together hundreds and hundreds of literary compositions, among them myths, epics, hymns, and wisdom literature. In addition it has provided us with numerous lexical texts, which are far more comprehensive and informative than ever before. But it is also the period when Akkadian had become the primary language of the scribes. That we have all these wonderful literary and lexical texts is only thanks to their efforts to learn proper Sumerian. They are the products of a scribal education (§1.2.4). Because these texts were produced by scribes who were not native speakers of Sume-rian, I have generally excluded them from my corpus.
After the early Old Babylonian period, scribes continue to produce Sumerian texts until the late first millennium BCE but with interlinear Akkadian translations added. Sumerian had become a language of scholarship and cult. These later texts are primarily lexical lists, literary texts, incantations, and cult songs.
1.2.3. Dialects
Sumerian was spoken during a long period and over a large area, in many different states, towns, and villages. It comes therefore as no surprise that it had dialects, distinct varieties of the language spoken in specific geographic areas. Unfortunately we know preciously little about them. Dialects were primarily a feature of the spoken language, whereas the written language tended to be quite uniform. Yet, in the earlier periods, when political unification was still the exception, local scribal traditions were stronger and those usually reflected traits of the local dialects.
Legal and illegal excavations have yielded texts from a limited number of larger towns.
Many regions are poorly documented. On the available evidence we can identify two main dialects during the second half of the third millennium BCE. In this grammar we will call them Northern and Southern Sumerian, following an old tradition to call the downstream area closer to the Arabian Gulf ‘South’ and the upstream area ‘North’. Strictly looking at the compass, we could just as easily call them Western and Eastern Sumerian. The relationship between these two main dialects changes across time and so do the linguistic properties in which they differ.
In the Old Sumerian period, the two dialects have at least two distinct linguistic properties.
In contrast with Northern Sumerian, Southern Sumerian follows a rule of vowel harmony, whereby certain verbal prefixes have two different forms, one with the vowel /e/ and the other with /i/ (see §3.9.3 for details). The second difference between the two dialects lies in their passive verbal forms. Northern Sumerian uses the prefix {÷a} as its passive marker and South-ern Sumerian {ba} (see §11.5.3 for details). On the basis of these two criteria, we can assign the general area of Lagash, Umma, Ur, and Uruk to Southern Sumerian, while Nippur, Adab, and Isin turn out to belong to Northern Sumerian.
After the early Old Akkadian period, the Southern Sumerian rule of vowel harmony ceases to operate and from then onwards the Southern dialect behaves in this respect in exactly the same way as the Northern. At the same time, however, the other distinctive property becomes even more marked. Both dialects continue to use their own distinctive passive markers, but in addition the verbal prefix {÷a} all but disappears from Southern Sumerian in its other, non-passive uses. Only the allomorph /al/ is retained and even that form only survives in subordin-ate clauses (see §24.4 for details). Thus, from the Old Akkadian period onwards, the main difference between the two dialects lies in their use or non-use of the verbal prefix {÷a}. (See chapter 24 for a full discussion of this development.) At the same time, a new but less promin-ent dialectal difference arises: the phoneme /ř/ becomes /r/ in Southern but /d/ in Northern Sumerian (see §3.3.2).
In the earlier periods, the two dialects were also the written norm in each of the two areas where they were spoken. This reflects the political fragmentation of the times. By the end of the third millennium this changes. During the Ur III period, when all of Sumer (and much more) was part of a single empire, the written language became also much more uniform than before. Most texts were written in what could be called standard Neo-Sumerian, which was linguistically a variety of Southern Sumerian. It used {ba} as a passive marker and had all but lost the verbal prefix {÷a}. Texts written in standard Neo-Sumerian are also found outside the area where Southern Sumerian was indigenous. All texts from Drehem are in that dialect and so are many texts from Nippur. The latter city, however, has also yielded many Ur III texts with Northern Sumerian or mixed features.
The dialectal differences mentioned thus far are certainly not exhaustive. They just happen to be rather frequent features that are prominently present in our texts. That the rather uniform scribal traditions hide a great deal more from us is obvious from occasional slips into non-standard spellings. Thus, the verb normally written se12 ‘live (said of more than one person or animal)’ occurs in an Umma text written as zé (SANTAG 6:154 obv 9; U; 21) but in Nippur texts as še (e.g. ECTJ 81 3; N; 24). Such differences in pronunciation must have been common but are only rarely reflected in the written language.
While the dominant written language during the Ur III period was a variety of Southern Sumerian, this changes during the subsequent, Old Babylonian period. The first centuries of the second millennium have yielded a great many unilingual Sumerian texts. Not only adminis-trative and legal documents, but also a wealth of literary texts. They, too, reflect different dialects, including standard Neo-Sumerian. However, the dominant written language of these later texts differs from standard Neo-Sumerian in a number of important ways.
One crucial difference lies in the behaviour of the two vocalic prefixes {÷i} and {÷a}, treated in chapter 24. In standard Neo-Sumerian, they are lost before the form /ni/ of the local
One crucial difference lies in the behaviour of the two vocalic prefixes {÷i} and {÷a}, treated in chapter 24. In standard Neo-Sumerian, they are lost before the form /ni/ of the local