Importance of Sumerian
For students of Mesopotamia, the need to study Sumerian is obvious. Alongside Akkadian, Sumerian is of prime importance for reconstructing all aspects of Mesopotamian civilization. However, a knowledge of Sumerian is also useful for those primarily interested in Semitic liguistics. Sumerian has a profound influence upon Akkadian – influence upon the phonology, morphology, syntax, and lexicon. Only through a knowledge of Sumerian can one differentiate between features of Ahhadian which are a product of its Semitic ancestry and those which have arisen secondarily under the influence of Sumerian.
The most Semitic language to be discovered is Eblaite. Even though Eblaite has only been known for a short while, its study has had a profound effect on Semitic linguistics. However, the majority of the texts found at Ebla are written completely in Sumerian, not in Eblaite. The remaining texts, although written in the Eblaite language, are couched in a Sumerian writing system which obscures many of the actual Eblaite forms. This means that a knowledge of Sumerian, especially a thorough understanding of the principles underlying the Sumerian writing system, is necessary for research in Eblaite.
Difficulties in the study of Sumerian
Sumerian is not as well understood as is Akkadian. Although there has been considerable linguistic progress in the last three decades, enough still remains unsure that scholars often have widely divergent views about Sumerian, Some of the reasons for these difficulties are summarized here:
-Sumerian is not genetically related to any other known language, living or dead. By contrast, it was discovered earl-on that Akkadian was a Semitic language. This genetic relationship aided early scholars in their reconstruction of Akkadian grammar and vocabulary. But in the case of Sumerian, there is no such help available.
-The writing system of Sumerian only imperfectly mirrors the spoken language; it does not indicate all the grammatical features which are known to have existed )or are assumed to have existed) in the spoken language. This schematic nature of the script makes it very difficult to reconstruct the morphology.
-With no comparative evidence and no native speakers to turn to, it is very difficult to determine what minor variations in morphology or syntax are meant to convey. Occasionally forms and or sentences are found which differ only a slightly from those occurring in other textsm but there is no simple way to find out what these differences signify.
In certain ways, it is actually easier to study Sumerian than it is to study , for example, Akkadian. This is because Sumerian does not have a great deal of morphology; there are not a large number of grammatical forms to learn. There is nothing like the weak-verb systems of Akkadian or Hebrew, which require a great deal of sheer memorization. . Rather, many students find the difficulties to be more conceptual in nature: the language works in way different than English, or other languages which students are likely to have been exposed to. It is occasionally difficult to understand some of these principles and even more difficult to observe these principles in action.
Historical background and texts used
The texts utilized here are all royal inscriptions of the Ur III Dynasty (approximately 2112-2004 BC), sometimes referred to as the Neo-Sumerian Dynasty. It grew out of the vacuum left by the collapse of the Dynasty of Akkad, which had been ruled byAkkadian- speaking kings of Semitic stock (approximately 2334-2193 BC).
The Ur III Dynasty was founded by Ur-Nammu, who ruled in the city of Ur from about 2112 to 2095. He had previously been governor of Ur under the suzerainty of the king of Uruk, Utu-Hengal; he may have been a relative of the latter. At some point he
declared himself independent. During his rule, and especially during the rule of his son Shulgi, the territory controlled by Ur expanded, until it reached most of the area previously controlled by the rulers of Akkad, that is, most of central and southern Mesopotamia. After
three more descendants of Ur-Nammu, the dynasty collapsed in 2004, partially due to pressures from the intrusion of nomadic, Semitic-speaking tribes. Thus, the Ur III period lasted a little more than a century; with the fall of Ur, Sumerian civilization, for all intents and purposes, also fell.
Ur III was a period of relative calm and stability in much of Mesopotamia. Because of the blooming of Sumerian art and literature, which had been somewhat submerged under the Semitic dynasty of Akkad, this period is often called the “Sumerian Renaissance”. Towns were fortified, many temples were rebuilt, and canals were dredged; trade with various foreign countries flourished.
The city of Ur itself, the capital of the Ur III Dynasty, was primarily excavated by Sir Leonard Woolley, perhaps the most famous of all Near Eastern archaeologists. The principal results were published by him and others in a series entitled Ur Excavations. Ten volumes have appeared: Volume I in 1929, and Volume VII in 1976 (Volume X appeared in 1951). Woolley popularized his results in a one-volume work entitled Ur of (1929). After Woolley’s death, P.R.S. Moorey revised and updated the work; it appeared as Ur ‘of the Chaldees’ (1982). This is a readable and interesting description of
the city at different historical periods.
Many Ur III texts have been preserved. The vast majority are economic and administrative; these number in the tens of thousands. Unfortunately, there are very few texts of what might be called a “historical” nature. There is much that is not known about
such matters as Ur-Nammu’s rise to power, the internal politics of the Ur III Dynasty, or even the physical extent of the Ur III “Empire”; C. J. Gadd refers to the “tantalizing want of information due to the singular unwillingness of the age to record even the triumphs, much less the failures, of its kings” (1971 :617).
Some original literary texts are also preserved from this period, as well as older works now committed to writing. Jacobsen says that the kings of Ur Ill, especially Shulgi, were much concerned to preserve extant older literary works and to encourage the creation of new ones. The court background of these works is unmistakable . … A major portion of Sumerian Literature as we have it traces back to the court of the kings of the Third Dynasty of Ur, where it was
composed and performed by the royal bards (1987 :xii, 277).
The royal inscriptions of the Ur III kings have been the object of study by W. W. Hallo. According to Hallo’s definition, royal inscriptions are texts which “were dedicated
either by, or to, or on behalf of the king” (1962:1). Hallo catalogued these texts, providing a standard system of reference. He also studied the different sub-types of royal inscriptions, categorizing them according to their function and according to their form.
These texts range in difficulty, from quite simple to very complex. They also contain a high degree of formulaity; many of the epithets of the king, for example, occur in a large number of the inscriptions. Even the phrasing of the verbal expressions is rather fixed. Since the genre of royal inscriptions existed both before and after the time of Ur III (in Sumerian and in Akkadian), a knowledge of the Ur III texts gives immediate access to other similar texts.
There has been much recent discussion about when Sumerian ceased to be a spoken language. This is not an easy question to answer; there are both historical issues and issues
of general linguistics to resolve. Most Sumerologists would say that Sumerian was a living spoken language in Sumer during the Ur III period, although some would say that it was already starting to die out during the latter part of this period. A minority would say that spoken Sumerian was either pretty far on its road to extinction, or might even have ceased to be a spoken language by the end of the Ur III period. Even the proponents of this view, however, would admit that the language of the Ur III royal inscriptions is “good” Sumerian, unlike some Sumerian of later periods.
THE SUMERIAN LANGUAGE
CLA S S IFICA TION
Sumerian appears to be what is called a language-isolate, that is, it has no genetic connection with any known language, living or dead. Attempts have been made to link Sumerian with many different languages – the most popular have been Hungarian, Turkish,
Caucasoid, Dravidian, and the Indus Valley language(s) – but none of these has found general acceptance. Such attempts have usually been based on surface-level resemblances with languages which are typologically similar.
A. Leo Oppenheim has pointed out:
The fact that S umerian is a complicated though very well understood language which cannot be linked to any other known language has created during the past hundred years a large literature attempting to relate Sumerian to practically all languages between Polynesia and Africa. The authors of such studies unfailingly “prove” that either their own language or a language
in which they happen to be interested is related to ancient Sumerian
( 1 97 1 :219).
Sir Gerard Clauson has summed this up: “Sumerian … has every appearance of being a ‘loner’, in spite of numerous attempts to foist relatives upon it, some grotesquely improbable”( 1973:38).
The possibility that a connection might be found with some other language is slim. Any related languages have probably died off without leaving any written records. The original homeland of the S umerians is unknown, so it is not even clear where its possible
linguistic relatives might be located. Wherever such a homeland might be, it was probably not in an area where writing developed very early.
The Sumerians referred to their own language by a term often transliterated as: emegir15. The value of the second sign is not sure, and so the term is variously transliterated as eme-gi7′ , etc., especially in older secondary literature. erne means “tongue” in Sumerian. The meaning of gir15 is unsure. Older scholars thought that it meant “Sumer”; in that case, the term would mean “language of Sumer”. More recently it has been argued that the term means something like “noble, prince”; erne-gir15 would then mean “the noble language”. Because of the uncertainties in reading this word, the term “Main Dialect” is often used instead.
There is also a “dialect” called erne-sal. The meaning of the second element of the name is uncertain; it may mean “fine, thin”. The “status” of this dialect is also uncertain. It has traditionally been called a “women’s language”, because it appears in literary texts of the Old Babylonian period, used by women when speaking to other women. For example, in the myth “Inanna’s Descent to the Netherworld”, when Inanna speaks to her aide Nin-Shubur, she does so in Emesal. There is no consistency in this usage; in other texts Inanna may speak in Main Dialect. Moreover, in texts of the later Old Babylonian period Emesal is also used for specific genres of text. Certain kinds of lamentations are always written in Emesal, even though recited by male priests. (Texts in some of these genres were preserved and even composed in schools for a thousand years after Sumerian had ceased to be a spoken language.) This use by men makes it difficult to determine exactly what Emesal is, and whether or not it should be classified as a “dialect”.
Emesal is well-attested from the beginning of the Old Babylonian period on. However, there appear to be at least one or two Emesal forms in the Gudea texts, and there has been a recent attempt to see Emesal forms in a group of texts written in an unusual orthography from Tell Abu Salabikh (approximately 2600 BC).
Emesal differs from Main Dialect in phonology and in the lexicon, but not apparently in morphology. In phonology, the Emesal forms often appear to be older. For example, the word for “lord” in Main Dialect is len/, in Emesal lumun/. It is difficult to say exactly what the more original form was; it may have been something like */ewenl or */uwun/. In any case, the Emesal form appears more conservative than the Main Dialect form. According to other scholars, however, Emesal forms are linguistically the more innovative; Emesal forms result from consonants being shifted to a more fronted or to a higher place of articulation. For example, Main Dialect Igl > Emesal Ib/; Main Dialect Idl > Emesal Iz/, etc. But there are several exceptions to these general principles, and there are a number of details of Emesal phonology which are not clear. As an example from the lexicon, the Main Dialect word for the interrogative “what?” is lana/; the Emesal form is /tal. These are apparently two etymologically distinct words.
It has been claimed that Emesal shares certain characteristics of “women’s languages” which occur elsewhere in the world. In particular, women’s languages are said to differ from “standard” dialects in phonology – the women’s dialect being more conservative than the standard dialect – and in the lexicon. More work needs to be done in defining the characteristics of Emesal, and in comparing Emesal with other women’s languages.
Not much is known about geographical variation within Sumerian. The extent of the Sumerian-speaking area is unsure; Sumerian texts are preserved from only a rather limited area. Moreover, the nature of the Sumerian writing system makes it difficult to see such variation. Only traces can be found, particularly in the later periods. There was undoubtedly more dialectal variation present than the writing system allows us to see.
Similarly, although Sumerian was spoken over a long period of time, there does not appear to be much variation before the Old Babylonian period. More differentiation is noticeable in post-Old Babylonian periods, when Sumerian was no longer a spoken
language. But here the differences may reflect the practices of different scribal schools and scribal centers, and not differences which were originally in spoken Sumerian.
There are occasional references in late Sumerian texts to what are apparently specialized languages, or jargons of particular occupations. For example, there are passing references to eme-utula, “the language of shepherds”, and to eme-ma-lah4-a, “the language of sailors”. It is hard to say what these dialects or jargons were like. Similarly, there are only passing references to what may be some kind of “literary dialects”: eme-gal, “great language”, eme-sukud, “high language”, etc. It is not known what these designations
There are several ways in which Sumerian works differently than the Semitic or IndoEuropean languages. Consider the Akkadian sentence, “The king went”:
(1) sarrum illik
Now, consider the Akkadian sentence, “The king built the house”:
(2) Sarrum bItam Ipus
king-NOM house-ACC VERB
In Akkadian, “king” is the subject in both sentences : It is the subject of an intransitive verb in sentence (1), and the subject of a transitive verb in sentence (2). Therefore, in both sentences it is put into the nominative case, Sarrum. In sentence (2), “house” is the direct
object of a transitive verb, and so it is put into the accusative case, bItam.
Languages in which the subject of a transitive verb and the subject bf an intransitive verb are marked one way (called the “nominative” case), and the direct object is marked a different way (called the “accusative” case), are often called “accusative” languages (or
Sumerian, on the other hand, is what is called an “ergative” language. In an ergative language, what we consider to be the subject of a transitive verb is marked by the “ergative” case. But, what we consider to be the subject of an intransitive verb, and what we consider to be the direct object of a transitive verb, are both marked by the “absolute” case.
In some ergative languages the ending for the ergative case, and the ending for the absolute case, may look completely different. In other ergative languages, the ergative case will have one marking, but the absolute case will be unmarked. (“Unmarked” can also be understood as “marked by zero”. This can be symbolized by “zero”: 0.) In other languages, there is no case-marking on any of the nouns; rather, ergativity is reflected in the way that certain elements within the verb cross-reference the case relationships.
In Sumerian, sentences ( 1 ) and (2) would be expressed as follows (Here and elsewhere, a period is used to separate morphemes; the verb forms have been slightly simplified) :
(3) luga1.0 i.gin
(4) luga1.e e.0 mu.n.du
king-ERG house-ABS VERB
In (3), the subject of the intransitive verb is marked by .0, the absolute case-marker. In (4), the subject of the transitive verb is marked by .e, the ergative case-marker, while the direct object is marked by .0, the absolute case-marker. This fits the definition of an
ergative language: The subject of a transitive verb is marked one way (in Sumerian, by .e), while the subject of an intransitive verb, and the direct object of a transitive verb, are marked a different way (in Sumerian, by .0).
Ergativity is a different way of marking the primary participants in a sentence. In an accusative language, the subject of a transitive verb and the subject of an intransitive verb fall into one grammatical category; in an ergative language, the subject of an intransitive
verb and the object of a transitive verb fall into one grammatical category. Consider the two English sentences, “The ball rolled down the hill”, and “The boy rolled the ball down the hill”. In English, “ball” in the first sentence is the subject, but in the second sentence it’s the direct object. Yet in each case, it’s the ball that is rolling down the hill. In an ergative language, “ball” would be in the absolute case in both the first and second sentences, and “boy” would be in the ergative case in the second sentence. In this example, an ergative language seems to capture our intuitions about the role of the ball in these two sentences better than does our accusative language.
In the above discussion, the terms “subject” and “object” were used. However, it is imprecise (and unjustified on theoretical grounds) to use these two terms when talking about an ergative language. Most linguists prefer to use the term “agent” to refer to the subject of the transitive verb (marked by the ergative case), and the term “patient” to refer both to the subject of the intransitive verb, and to the direct object of a transitive verb (both marked by the absolute case). Thus, in the examples above, “boy” is the agent, and “ball”
is the patient. In practice, it is very difficult to escape using such common terms as “subject” and “object”, especially in unambiguous contexts, even if these terms do not really fit Sumerian.
There are many ergative languages in the world, belonging to a number of different language families: many languages in Australia, many American Indian languages, the Caucasoid languages (for example, Georgian), Basque, to name a few. However, none of
what are sometimes referred to as the “major cultural languages” of Europe are ergative, and so the concept is unfamiliar.
There are two other important points about ergativity. First, the definition given above describes what may be called “minimally” ergative languages. However, ergativity can also be reflected in other parts of a language’s grammatical system – it may affect verbal
agreement, cross-referencing of case-markers, coordination and subordination, etc. This will be discussed in more detail later.
Second, there appear to be very few (if any) “pure” ergative languages. Most (perhaps all) ergative languages are “split”. In certain constructions, the language behaves in an ergative manner; in other constructions, the language behaves in an accusative manner. In Sumerian, for instance, the perfect aspect functions in an ergative manner, while the imperfect aspect functions in an accusative manner. That is, Sumerian is split along an aspectual axis. There are other languages in the world which are split along exactly such an axis, that is, the perfect aspect functions in an ergative manner, and the imperfect aspect functions in an accusative manner. Also, the independent pronouns in Sumerian function basically on an accusative, not an ergative, basis. Languages of the world show a rather bewildering variety and complexity in the ways that they are split.
In addition, there are languages which use an ergative – absolute differentiation to mark semantic distinctions which are not easily made in the Semitic or Indo-European languages. An oft-cited example is the sentence “We fell” in Bats, a member of the
Caucasoid language family, spoken in Georgia. If the act of falling is purely an accident, outside of our control, the subject of the sentence is in the absolute case. If we fell as a result of our own action, the subject is in the ergative case. Other languages use an ergative
– absolute differentiation to mark other kinds of information, such as degrees of animacy.
Because there are very few (if any) pure ergative languages, it is perhaps best not to think of “ergative – accusative” as a simple binary opposition. C.T. van Aalderen has said that “One suspects that the whole phenomenon is more a continuum than a set of
oppositions” ( 1 982:27). That is, some languages are closer to one “pole” than to the other. Several recent linguists, for example, speak of “degrees of ergativity” in different languages.
In the last twenty years or so, general linguists have shown a great deal of interest in ergative languages; the bibliography of recent works is vast. In one of the more recent articles, John Du Bois says:
Seemingly, ergativity stands as a challenge to the view that all languages are built on one universal archetype . … Why are there ergative languages in the world? … Ergativity … would seem somewhat perverse in splitting up an apparently basic category like subject, assigning half its contents to a contrasting category like object. This perception of unnaturalness is of course only an index of our failure to apprehend the actual basis of ergativity, a difficulty which is simply reinforced by traditional grammatical terminology ( 1 987 :805-7).
It is only somewhat recently that the term ergative has been systematically used for Sumerian. Although some early researchers had intimations that this was how Sumerian worked (even if all the details were unclear, as they still are), it is only in the last few years
that ergativity has been explicitly discussed in Sumerian. This means that in reading even fairly recent Sumerological literature, such concepts and terms as “ergative”, “agent”, “patient”, etc., may not be used at all. The material might be discussed in what would now be called an ergative model, without use of the term ergative, or in older works the material might be presented in an accusative model. Moreover, not all scholars believe that Sumerian functions on an ergative basis. Some S umerologists believe that not enough
evidence has been presented to prove the case, and also believe that there are too many “exceptions” to the model. Others disagree on the degree to which Sumerian can be said to be split. Given the complexities of split ergativity in the languages of the world, it may be that current presentations of ergativity in Sumerian are too simplistic. “Full” proof can only be forthcoming when there is more secure knowledge of Sumerian verbal morphology.
The first person to apply the term ergative to Sumerian was apparently Viktor Christian in 1 957, although he used the term a little differently than it is usually understood. Diakonoff ( 1965) sketched the system of ergativity in Sumerian and other Ancient Near Eastern languages, without explaining the details of morphology. The articles by Daniel Foxvog ( 1 975) and Piotr Michalowski ( 1980a) viewed Sumerian in an explicitly ergative
framework, while elucidating the verbal morphology. Van Aalderen (1982) has explored some of the theoretical issues in more detail. The grammar by Marie-Louise Thomsen ( 1984) also follows a split-ergative model. A recent survey of ergativity in Sumerian is by
Gong Yushu ( 1987).
Sumerian is often described as an “agglutinative” language. This term goes back to the nineteenth century, when linguists attempted to classify the languages of the world into a few basic types, based solely on typological (not genetic) criteria. For these linguists, the
three most common types of language could be classified as:
In isolating languages, virtually every morpheme forms a separate “word”. In Chinese, for example, there are no tense-markers on verbs; such information is conveyed by separate adverbs. There are also no plural-markers on nouns or verbs; this information
is conveyed by separate number-words.
In fusional languages, such as Akkadian or Latin, grammatical morphemes are expressed through endings on nouns or verbs, and several different morphemes tend to “fuse” together. Latin amo, for example, means “I love”. The /0/ ending on the verb signals several things: the verb is first person, singular, present tense, indicative mood, active voice. However, none of the morphemes for person, number, tense, mood, or voice can be segmented out – they are all fused into the ending /0/.
In agglutinative languages, as in fusional languages, several grammatical morphemes are combined into one word. However, the morphemes are distinct from each other; they do not fuse together. In an agglutinative language, strings of prefixes or suffixes tend to
occur; each affix is formally distinct, and expresses one morpheme. The parade example of a language of this type is Turkish. In Turkish, the phrase “from his houses” is expressed as: evlerinden. Ev means “house”, ler is the plural marker, in is the possessive pronoun “his”, and den is the postposition expressing the ablative “from”. In general, each affix expresses one morpheme; each morpheme is invariant: ler is the automatic plural marker for all nouns; den means “from” after any nominal phrase, etc. The morphemes are distinct, not fused into each other.
Sumerian is similar to Turkish. The verbal phrase, for example, consists of a string of prefixes, followed by the verbal root, and then a smaller string of suffixes. Each affix expresses one morpheme, and each affix is (basically) invariant. Nominal phrases can be very long, with a noun, modifying adjectives and appositives, genitive phrases, etc., with a case-marker at the end of the entire nominal phrase.
The typological scheme presented here has been somewhat simplified. Moreover, languages only tend to one category or the other; they are not “purely” isolating, fusional, or agglutinative. English, for example, is largely isolating, but it is also to some degree
fusional. It is occasionally agglutinative in its processes of word formation. In English words such as “predictability” or “antidisestablishmentarianism”, it is fairly easy to separate
several different morphemes, both as prefixes and as suffixes.
Most modern linguists who specialize in linguistic typology are not very interested in this particular “morphological typology”. They believe that such a scheme is not especially useful, because it does not offer any interesting or helpful intuitions or generalizations
about language. The methodological underpinning of this classification scheme has also been attacked on several grounds. For example, it was mentioned above that languages do
not usually fall neatly into one of these types. However, since the term agglutinative is still used in Sumerological literature, especially in popular descriptions of the language, it is useful to have some idea of what the term means.
The two terms ergative and agglutinative refer to different categories. The ergative – accusative distinction depends on how the primary participants in a sentence are marked in relation to each other. The isolating – fusional – agglutinative distinction refers to the
different ways that morphemes are combined into words. In theory, a language can be either ergative or accusative, and also either isolating or fusional or agglutinative, although not all of these possible categories seem to occur.
In discussing any writing system, there are two factors to consider: the external characteristics of the script, and the principles behind the script.
Because of the external shape of the signs in the Sumerian script, its writing system is called “cuneiform”. “Cuneus” is the Latin word for “wedge”; the term was coined because of the most striking characteristic of the script – the fact that the signs are built up of strokes looking like little wedges. (The term cuneiform was apparently first used by one Thomas Hyde in 1700. In his Historia religionis veterum Persarum, he refers to “dactuli pyramidales
The cuneiform signs were inscribed by means of a stylus probably formed from an actual reed (such as still grows in modern-day Iraq), by impressing the stylus upon a tablet of moist clay (or, occasionally, upon other surfaces). The stylus could also be made of bone, metal, hardwood, or even other material.
The first cuneiform texts discovered were all relatively late, from a period when the wedge-shaped characteristics of the script were most striking. In the earliest phases of the script, however, this wedge-shaped character is less pronounced; the script of most of the
Ur III inscriptions in this book does not look nearly as wedge shaped as do later texts.
The tenn cuneifonn refers solely to the external shape of the individual signs. Cuneifonn script was adopted and modified by many peoples of the Ancient Near East; it was used to write Akkadian, U garitic, Hurrian, Persian, etc. However, the fact that these languages use signs with the same general external characteristics says nothing about their possible genetic relationship. Sumerian, Akkadian, Hurrian, and Persian, for example, belong to four entirely unrelated language families. Expressions such as “cuneifonn language” are occasionally encountered, but this is a rather imprecise way of referring to one or several languages, which may or may not be related, which use a script with the
same external characteristics.
The writing system used for English is an attempt to render speech as closely as possible. Although English does suffer from numerous archaic spellings, and there are certain features (such as upper and lower-case letters) which are found only in writing, writing is basically an attempt to reproduce speech sounds. By contrast, the Sumerian writing system was never an exact, phonetic representation of speech; it was not “designed” to reproduce spoken language as such. Rather, to some degree the writing system is only a mnemonic device, to jog the memory of the writer and reader. The earliest uses of writing were for administrative texts, which were of a fonnulaic nature, and whose contents were familiar to the scribes. There was no need to write down what would be obvious to a scribe who was a native speaker of Sumerian, and who was familiar with the material being written. When such scribes “read” the texts, they knew how to supply the infonnation not indicated explicitly in the writing.
Thus, a certain amount of infonnation in the spoken language was not expressed in the writing. The further back in time one goes, the less the Sumerian writing system expresses grammatical elements which are assumed to have been present in the spoken language. For
example, the basic graphic shape representing the root for “to build” was originally a picture of a wooden peg. In the earliest Sumerian, this one sign could be used for any inflected fonn of the verb: any tense, mood, or person. S imilarly, the expression for “on that day” in Sumerian was: ud-bi- (“day-that-on”). But in the earliest Sumerian, only the ud-sign was written; the reader inferred the rest.
As might be imagined, this lack of explicitness in the script can cause much trouble in interpreting Sumerian texts. Nor is this problem limited to the earliest Sumerian texts; in late economic texts, for instance, it is often difficult to tell if something is being distributed
“to” or “from” somebody.
As time passed, the scribes wrote more and more down, that is, the writing became more and more explicit. For example, there is a Sumerian text known as the “Kesh Temple Hymn”, attested in several copies mostly from the Old Babylonian period (dating to around 1800 BC). In the 1960s, a version of the same text was found at Tell Abu Salabikh, dating to about perhaps,2500 BC. Unfortunately, only a few lines of the Tell Abu Salabikh
version survive. But if one compares the Old Babylonian version with the Tell Abu Salabikh version, it can be seen that although the text itself is relatively stable, the Old Babylonian version indicates more verbal affixes than does the Tell Abu Salabikh version.
This increase in explicitness may be connected with the fact that Sumerian was gradually dying out, and so scribes needed more help in their own understanding of texts.
Thus, a fundamental feature of the Sumerian writing system is its lack of explicitness. It does not fully represent the spoken language. This has been summarized by Jacobsen: “The history of Sumerian writing is one of progressively ever greater but never quite attained adjustment to Sumerian speech” ( 1 957:366 n.1). Similarly, Marvin Powell has pointed out that “We find traces of its mnemonic character enduring to the very end of the Sumerian orthographic tradition” (1981 :421).
A further complicating problem is that the writing system is to some degree morpheme-bound. There is indirect evidence to show that there were certain phonological changes which took place in Sumerian, such as contraction, vowel deletion, etc., but these
changes are masked by the script; the script often reproduces the basic morpheme, without showing the changes which are assumed to have taken place in the spoken language.
The view here presented, that the Sumerian writing system in origin and in practice is
basically mnemonic, has been especially expounded by Diakonoff ( 1976) and Stephen Lieberman ( 1977).
The script used for writing Sumerian is a combination of “logographic” and “syllabic” elements. Logographic means that a sign stands for a particular word. For example, the sign 4 stands for the word utu, “sun”; the sign stands for the word digir:, “god”. The
external shape of many of these signs is clearly pictographic in origin. Thus the sign for “sun” was originally a picture of the sun rising over a mountain. The sign for “god” was originally a picture of a star. The original significance of many signs cannot yet be determined.
The same sign can often have more than one logographic value. Thus, the same sign can represent digir, “god”, or it can represent an, “sky”. In general, it is only the context which determines the meaning of the sign, and its correct reading.
Syllabic signs are used to reproduce a sequence of phonetic elements. For example, the sign is used to represent the syllable /ga/. This particular syllable can form a component of several different morphemes: it may be part of the cohortative prefix on
verbs, or part of the ending of a genitive phrase on nouns, etc. The sign in these contexts does not stand for any particular word; rather, its purpose is to represent the phonetic sequence /g/- /a/, which may form part of a number of different morphemes.
Syllabic signs can represent several different kinds of segments of consonants and vowels. Some syllabic signs stand for single vowels, e.g., and i. More common are signs standing for the sequence consonant-vowel (ba, mu) or vowel-consonant (fill, in). There are some signs that stand for consonant-vowel-consonant, but these are not common; instead, the script uses a convention that represents /CVC/ by CV-VC. For example, the segment /nir/ is written by: ni-ir. A writing such as ni-ir does not imply a long vowel; this is purely an orthographic convention, to reduce the number of CVC-signs which would otherwise be necessary.
Many signs have more than one syllabic value. Many signs have both logographic and syllabic values – sometimes more than one of each. The correct value of the sign can usually only be derived from the context. Signs with more than one value are called “polyvalent”, or are said to have several “readings”.
Thus, the Sumerian writing system is both logographic and syllabic. The syllabic value of most signs derives from a logographic value. For example, the sign in its meaning as “sky” is pronounced /an/. This phonetic value was then generalized, so that this sign can stand for the syllable /an/ in other contexts.
In general, lexical morphemes are written logographically, and grammatical morphemes are written syllabically, but this is not always the case. The system is complicated by the fact that certain syllabic signs tend to be used for certain morphemes.
For example, there is a “conjugation-prefix” on the verb, pronounced /bi/. There are several different possible ways that this phonetic sequence could be represented in the script. In practice, however, the scribes almost always used only one of these possibilities, the sign. That is, certain morphemes tend to be indicated in only one way, and, conversely, certain signs tend to be used only for certain morphemes.
In addition to logographic and syllabic signs, there are a few other elements present in the script. One of these is “determinatives”. Determinatives are signs which are used to indicate the general semantic class to which a following (occasionally a preceding) noun
belongs. For example, almost all divine names are preceded by the sign ; this sign tells the scribe that “what follows is a divine name”. Most names of countries are followed by the sign ; this sign tells the scribe that “what precedes is the name of a country”. Determinatives were probably not spoken, even when Sumerian was read out loud. They were only a feature of the written language.
In other contexts, the cuneiform signs which function as determinatives can also function as logo graphic or syllabic elements. For example, the sign ¥ can represent digir, “god”; the sign can represent ki, “country”.
To sum up, Sumerian is mostly logographic, and only partially syllabic. Akkadian, on the other hand, is mostly syllabic, and only partially logographic. Persian cuneiform is almost entirely syllabic, and Ugaritic cuneiform is basically alphabetic. In practice, people
sometimes confuse the issue, and the term cuneiform is occasionally used to refer in general to any logographic-syllabic system of writing, but this is wrong; there are many logographic-syllabic scripts which have existed in the world, which are not cuneiform.
This has been a somewhat simplified discussion of the Sumerian writing system. There has been much recent discussion about the script, mostly hinging on theoretical questions, such as the difference between pictographic and logographic, or the degree to which the script is morpheme-bound.
When citing Sumerian texts, or when discussing Sumerian grammar or vocabulary, Sumerologists do not generally reproduce the original cuneiform signs. Rather, they cite the word or passage in transliteration into Latin characters. Transliteration is a sign-by-sign image of the original written text. It is designed specifically to reflect the actual cuneiform signs present. By looking at a transliteration, one should be able to determine exactly which cuneiform signs occur in the original text (excluding palaeographic niceties). Transliteration serves several purposes. It is more convenient, quicker, and cheaper to
produce Latin characters than it is to produce cuneiform characters. Also, it provides an approximate phonetic rendering of the signs occurring in the Sumerian. Since many
Sumerian signs have more than one reading, a scholar, by giving the text in transliteration, explicitly states his opinion about the reading of a particular cuneiform sign. For example,
the signcan be read Hkur (the name of a god), or im (“wind”), or ni (“self’). Based on his understanding of the text, a scholar decides the correct reading.
There are some complexities of transliteration. It is possible for several different cuneiform signs to have the same pronunciation. These signs must be differentiated in transliteration, so that the original cuneiform can be reconstructed from the transliteration.
For example, there are at least four different signs pronounced as /u/. If y were used as the transliteration for all four signs, it would not be possible to go backward from the
transliteration: Given a transliteration y, one could not tell which of the four possible signs actually was written in the cuneiform. To obviate this problem, scholars have devised the
following system: The most common (or most important) sign with a particular value is unmarked. The second most common (or most important) sign with this same value is marked with an acute accent: g. The third most common (or most important) sign with this
same value is marked with a grave accent: !l. The fourth, and higher, most common signs with this same value are marked with subscripts: .!4 , liS ‘ etc. This system is purely
arbitrary; it provides a convenient means to differentiate between signs pronounced alike, thus enabling us to reconstruct the cuneiform from the transliteration.
This use of the acute and grave accent-marks as “indices” has nothing to do with pronunciation. They do not indicate anything about accent, nor do they indicate anything about vocalic length, nor do they indicate anything about tone. They are used instead of a
possible Y2 and Y3 simply because it is easier to type accent marks (at least in Europe) than it is to turn the typewriter carriage up to make a subscript.
These indices are based largely on frequency. However, these frequencies were determined on the basis of Akkadian texts, not on the basis of Sumerian texts (for the simple reason that Akkadian was “discovered” before Sumerian). This produces a certain
inconsistency. In Sumerian, for example, the bi-sign is much more common than the bi-sign. This inconsistency is not really a problem; the only other alternative would have been to devise a separate system for Sumerian, based on values and frequencies in Sumerian.
But this would have engendered so much confusion and complication that it is far easier to work with the traditional system.
Confusion arises when indices are used on bisyllabic signs, that is, signs which represent a segment of two syllables, such as /kala/ or /Urim/. If there is more than one sign with the same bisyllabic reading, some scholars put the accent-marks on the first
vowel, then continue onto the second syllable if there are several signs with the same reading. Other scholars, however, begin with the last vowel, moving back to the first. Either system is prone to mechanical mistakes in printing, and the mere presence of the two different systems can cause problems in determining what the cuneiform sign actually was. To mitigate against this difficulty, some Sumerologists do not use acute or grave accent-marks on bisyllabic signs. Instead, they use a subscript 2 or subscript 3 when necessary.
For example, there are several signs with the value of /kala/. These are differentiated as: kala, kala2> kala3′ kala4’ etc. This is the system followed here. Some recent publications, including the Pennsylvania Sumerian Dictionary, use subscripts in place of accent-marks, even on monosyllabic signs. Thus, instead of ù, they use u2; instead of ù, they use u3. Determinatives are a feature of the written language, and were probably not spoken.To indicate that they were not pronounced, they are transliterated with superscript letters:
Xki, tugX, etc. For convenience sake, the determinative for god (the iligir-sign) is transliterated as a superscript d: dBtar. Because of the typographical difficulties of printing superscripts, some publications instead print the determinatives on the same print-line,
connected by a period: X.ki; tug.x. In transliteration, signs comprising one “word” are linked by hyphens: kalam-ma, illgir-ra-ni, etc. (Determinatives are an exception; no hyphens are used.) As will be seen below, it is not always easy to determine what constitutes a “word” in Sumerian.
Some Sumerologists use initial capital letters in their transliteration of Sumerian proper names; other Sumerologists do not. Those who do not use them, consider capital letters to be a feature particular to the English writing system; since capital letters have no correlate in
the Sumerian writing system, they should not be used in transliteration. Other scholars feel that since transliteration is an artificial device anyway, there is no harm in using capital
letters, if they help make the text clearer to the reader. This second practice is followed here.
Finally, it is necessary to say a few words about the typographic conventions used in transliterating Sumerian. Throughout this book, Sumerian is transliterated by Roman characters, underlined. The few Akkadian citations used here follow the same system.
However, it is occasionally inconvenient to use the same typographic conventions for two different languages. To solve this problem, many publications cite Sumerian in Roman characters, but widely-spaced. Thus, the word for “god” will be transliterated as: digir.
This may seem like a convenient procedure to differentiate citations from the two languages, but it is prone to produce mechanical errors in printing.
It is frequently the case that it is not known how a particular Sumerian sign (or word) is to be read. Some scholars elaborate the system just discussed, by presenting such doubtful or unsure readings in caps. For example, the word for “interest-bearing loan” in Sumerian is written: 1lUI . It is not sure how the first sign is to be read. For this reason, the word is often cited as: IjAR-ra. Some scholars do, however, believe that they now know how to read this word, and so nowadays one is likely to see the reading: urS ra.
That is, wide-spaced Roman is used for the “standard” transliteration of Sumerian, and caps Roman is used for unsure readings. Not all Sumerologists follow this system,
however, and what is sure for one scholar may be unsure for another scholar.
Transliteration is, by definition, a reflection of the written language, and so does not necessarily reproduce the spoken language well (as we think we understand it). For this reason, most Sumerologists use some form of transcription in their study of Sumerian.Transcription is not used as frequently as is transliteration; it occurs in discussions of grammar, and appears in scholars’ own notes.
Transcription attempts to reproduce Sumerian forms in their approximately correct phonological and morphological shape, disregarding the omissions, conventions, and idiosyncrasies of the written language. For example, signs appearing as kalam-ma in
transliteration, will appear as kalama in transcription, since that is probably how the word was actually pronounced.
There is no “official” or “standard” system of transcription of Sumerian. It tends to be somewhat personal and idiosyncratic, used by each Sumerologist to enable himself to understand the language behind the written form. This situation contrasts with that of Akkadian, for example. In Akkadian there is a standard way of transliterating texts, and also a reasonably standard way of transcribing them. This can be done for Akkadian, because scholars are generally confident of their understanding of the rules of Akkadian phonology and morphology; in general, transcriptions of Akkadian done by different scholars will be quite similar. In the case of Sumerian, there is much less confidence about the language. Because the script does not always express all grammatical elements, the morphology is not always sure. Moreover, there are several different analyses of the phonetic structure of Sumerian.
The system of transcription used by most Sumerologists is not always transcription in the precise sense of the term. For example, morpheme boundaries are often indicated. Also, full forms of morphemes are often indicated, even when it is assumed that some
vocalic or consonantal segment probably dropped. Thus, it is actually a kind of morphological transcription.
The system of transcription used in this book is based on the system of Jacobsen, and is similar to what many Sumerologists use. It is a morphological transcription, in that it separates morphemes from each other. In this system, morphemes are separated by periods. Features which are assumed to have been present in the spoken language, but which do not show up in the written language, are enclosed in parentheses. The different indices which appear in transliteration are ignored. Thus, é will be transcribed as e, and
Urim5 as Urim. Exceptions to this latter rule are sometimes made, particularly for grammatical morphemes which tend to be written in only one way. Thus, the “terminative” case-ending is normally transcribed by .sè, because it is always written with the sè-sign,
and never with the sè-sign or the M-sign. Similarly, the “enclitic copula” is normally transcribed as .am, since it is regularly written by the am-sign, and not by the am-sign or the am-sign. (Details of these conventions will be discussed below.)
The difference between transliteration and transcription should be kept in mind. Transliteration is essentially sign-by-sign, with the goal of representing the cuneiform signs which were used in the original. Transcription is essentially word-by-word, with the goal
of approximating the correct phonological and morphological shape of a word. (In practice, however, the terms transliteration and transcription are occasionally used promiscuously.)
Transcription is important, because transliteration alone masks too many morphological and phonological issues. Only a consistent transcription can reveal a thorough understanding of the language of the texts. Some of the simplest inscriptions, for example, could be translated without knowing much Sumerian, simply from a knowledge of Akkadian and of simple vocabulary; a transcription reflects the structure of the language hidden beneath the written form.
At certain times in this book, the purely phonemic structure of Sumerian will be stressed, ignoring any morphological considerations. In that case, normal linguistic practice will be followed, and the item will be put between slashes, e.g., /kalama/.
Thus, our understanding of Sumerian may be reflected in three different ways: a transliteration, reflecting the written shape; a phonemic transcription, reflecting the pronunciation; and a morphological transcription, reflecting our understanding of the
pronunciation and morphology.
It is not easy to reconstruct the phonological system of Sumerian, or the precise pronunciation of any of its sounds. There are two main reasons for this problem. Since Sumerian is a language-isolate, there is no comparative evidence to provide help. Moreover, most of the evidence for Sumerian phonology has been filtered through the
Akkadian phonological system; Sumerian phonology is seen through Akkadian eyes. For instance, it is quite likely that the word for “son” in Sumerian was pronounced /domu/, with an initial /o/-quality vowel. But Akkadian does not have an /o/-quality vowel, and
hence no /0/ -sign, and so this word is spelled out in syllabic Akkadian as: du-mu. If there were only Akkadian evidence, it might never even be known that Sumerian had an /0/- quality vowel. Thus, the picture of Sumerian of the Ur III period (21 12-2004 BC) is actually based on Akkadian of the Old Babylonian period ( 1894- 1595 BC), and later. (Similarly, much knowledge of Sumerian grammar derives from the interpretations given to it by Akkadian-speaking scribes and scholars.
Likewise, very little is known about the historical development of Sumerian phonology. Sumerian was spoken over a period of several centuries (and was used as a written language for even more centuries). The phonological system of Sumerian at the time of, say, Tell Abu Salabikh and that of the time of Ur III may have been significantly different.
To some degree, more is known about the value and pronunciation of Sumerian grammatical morphemes, than about Sumerian lexical morphemes. This is because grammatical morphemes are mostly written syllabically, while lexical morphemes are usually written logographically. Without the evidence of lexical lists, it is quite difficult to fix the value of a logogram. For the same reason, it is occasionally possible to see phonetic change through the course of Sumerian in grammatical morphemes, but it is more difficult to see such changes in lexical morphemes.
The upshot of this is that Sumerian probably possessed sounds which Akkadian did not, and which can only be determined using a variety of indirect evidence. Because of the difficulty of dealing with this indirect evidence, there have been several different reconstructions of the Sumerian phonological system. These reconstructions differ both in the number of phonemes present in Sumerian, and in the value attributed to certain phonemes.
In practice, however, most Sumerologists do not try to exactly reproduce the sounds of Sumerian. Rather, they use the standard values known from Akkadian. Thus, virtually all transliterations of Sumerian will use the value dumu for “son”, even though this is one
of the clearest cases where an /0/ -quality value can be postulated for Sumerian. Similarly, it is sure that Sumerian had a velar /lJ / , which did not exist in Akkadian. The sign �� , for example, represents /lJ u/ , the velar nasal followed by an /u/ -quality vowel; this is the morpheme for the first person singular possessive-suffix on nouns. But the normal value of this sign in Akkadian is /mu/. Therefore, many Sumerologists transliterate this sign as mu, e.g., lugal-mu, “my king”. Other scholars, however, transliterate this sign as gu10 e.g., lugal-gu. Still others, who wish to be more precise, in fact transliterate this sign as lJu10, or as some typographical equivalent, such as gu10, gu10 etc.; for example, lugal-gu10. This
means that transliterations of Sumerian will differ somewhat from scholar to scholar. The transliteration used here will reflect the conventional method of transliteration used by most Sumerologists, even if this reconstruction is somewhat shaky and incomplete.
Sumerian had at least the following vowels:
The precise phonetic value of these vowels, particularly the /e/, is unsure.
Many scholars also believe that Sumerian had an lol-quality vowel, but since no /0/ existed in Akkadian (at least on the phonemic level), there is only indirect evidence to it. It is very difficult to determine whether any particular Sumerian word had an
/0/ -quality vowel or an /u/ -quality vowel; its existence has been established for only a few cases. Under the assumption of the existence of this 101 -quality vowel, the vocalic system of Sumerian is more symmetrical:
Other Sumerologists have posited other vowels, such as both an open /e/ and a closed /e/. Others have posited the existence of nasalized vowels, but the exact number and quality of these varies from one scholar to another: /i/; /e/; /i/ and /a/; /i/, /a/ and /e/,
etc. Claude Boisson ( 1988) has investigated various reconstructions of the phonemic system of Sumerian, in comparison with what is known about language in general. He feels that if Sumerian possessed only four vowels, then the vowel normally represented as
/e/ was more likely IEI than le/. He also feels that none of the systems of nasals which have been posited for Sumerian is likely.
It is not sure if there was a phonemic distinction between short and long vowels; this cannot be told from the script. It has been postulated that there were no originally long vowels in Sumerian, but that they did arise through vocalic contraction, in particular the
contraction of final root-vowels with initial vowels of suffixes.
As discussed above, in practical terms most transliterations of Sumerian usually only reflect the vowels known from Akkadian; that is, the four vowels listed above.
Most analyses of Sumerian would include the following consonants:
b p m
d t n
g k IJ
z s s
(For ease in printing, the consonant indicated above as b is often simply transliterated as h, without the “dish”. Since Sumerian does not have a “simple” /h/, there is no ambiguity in this usage.)
Virtually all Sumerologists accept the existence of the velar nasal /IJ/ (although some scholars prefer to speak of a palatal nasal, and others have seen more complex phonemes, such as /IJm/). When Sumerian words containing this phoneme are loaned into Akkadian,
it is usually (although not always) reflected as !!g. For example, , “kind of priest” appears in Akkadian as sangu.
Transliterations of this phoneme vary. In older works, and in many contemporary works, it may simply appear as g. Some recent works use g, or some typographical equivalent (g, etc.). It will be transliterated here as g, in cases where it is assumed by most
Sumerologists to be present. With many words, however, it is not known whether a phoneme is /IJ /, /g/, or even /n/ or /m/, and so some variation in the transliteration of certain words appears. For example, the verb “to go” is understood by some Sumerologists to be /gin/, but by others to be /gin/ (or /gen/).
Many Sumerologists believe that Sumerian had a phoneme usually symbolized by /dr/; its exact phonetic significance is unsure. Its existence has been proven in only a few cases. Because of the difficulties of proving its existence in specific words, it is usually not
indicated in transcription; instead, in the standard sign-lists and in most transcriptions it is reflected as d.
Several other consonants have been posited for Sumerian: /h/, /w/, /y/; two (or more) types of /l/; two (or more) types of /r/; a labiovelar /kw/; a pre-nasalized labial stop /mb/; etc. Since none of these sounds exists in Akkadian, the evidence for their existence
in Sumerian is indirect at best, and individual Sumerologists have their own preferences. Transliterations of Sumerian do not normally try to reproduce these disputed phonemes.
As a typical example of a reconstruction of Sumerian phonology, it may be instructive to present that postulated by Lieberman:
e i b p m s
a 0 d t n z
u g k g
l r i
In the tables above, certain consonants are indicated as differing only in voice: /b/ /p/; /d/ – /t/; etc. It is not in fact sure what differentiated such pairs; Lieberman explicitly says that the distinction he marks as /b/ – /p/ was not one of voice. Some Sumerologists have speculated that the difference was one of aspiration; this is not an uncommon view today. Boisson, for example, says: “A correlation of aspiration seems to be the only hypothesis with a high probability of success” ( 1988:25). Other Sumerologists have speculated that the difference was one of glottalization.
There does not appear to have been a phonemic distinction between short and long consonants; it is not in fact sure if long consonants occurred at all.
One of the thorniest questions in Sumerian involves the status of word-final and syllable-final consonants. According to most Sumerologists, certain consonants, when in
word-final position, were not pronounced. For example, the root for “dais” is /barag/, with a word-final /g/. However, unless this /g/ was followed by a vowel, it was not
pronounced: this word would have been pronounced as Ibara/.
The word-final consonant in a root is usually referred to by the German term “Auslaut”. Thus, it is said that the word for “dais” (pronounced /barag/) had a “g-Auslaut”, or the word for “to live” (pronounced /til/) had a “l-Auslaut”.
The consonants which were regularly not pronounced in word-final position are called “amissable” consonants. Those which were pronounced in word-final position are called “non-amissable”. (These terms are apparently peculiar to Sumerologists; they are not used by general linguists.)
Sumerologists differ among themselves about which consonants were not pronounced. Some believe this affected all consonants, although perhaps not “to the same degree”. Others believe that it affected a smaller number of consonants (although no two
lists of such consonants seem to agree exactly). Also, it is not known if the amissable consonants were not pronounced in word-final position only; most Sumerologists believe that they were not pronounced in any syllable-final position. Arno Poebel, for example (the real father of Sumerian grammar), states that “As a rule, an amissable consonant is dropped whenever it stands at the end of a word or syllable” ( 1935: 147). Similarly, Samuel Noah Kramer says: “All final consonants in Sumerian are amissable. … The term ‘final
consonant’ as here used includes the consonant at the end of a syllable as well as the one at the end of a word” ( 1936: 19).
The existence of amissable consonants is certainly not impossible. There is a close parallel in French: In spoken French, word-final consonants are not pronounced (under certain conditions), although they still appear in the written form. A few Sumerologists,
however, are not convinced of the existence of amissable consonants. They interpret the problem as being orthographic in nature.
The reason this question is still unresolved is because of ambiguities in the writing system. At various points in this book, different pieces of evidence will be cited, some of which seem to indicate that word-final consonants were pronounced, and some of which seem to indicate that word-final consonants were not pronounced.
The existence of amissable consonants means that the cuneiform signs which represent words with these amissable Auslauts have two values: a “long” value, which includes the amissable Auslaut (e.g., kalag, Urim5′ til), and a “short” value, which does not (kala, Uri5 ‘
til). With some signs, the long value and the short value have different indices, e.g., til [with diacritic] and !i [without diacritic]. This annoying situation is partially due to the fact that indices were originally assigned on the basis of frequency in Akkadian, not Sumerian.
Some scholars transliterate Sumerian using basically only the long values; others transliterate Sumerian using basically only the short values. Other scholars use both, the choice being determined by syllabic conditions: the short form if word-final (or syllablefinal),
the long form if not. Others are less consistent, using a mixture of long and short values. This latter practice is particularly true of less recent Sumerological literature, where one finds a mixture of transliteration principles, based primarily on customary readings of
the cuneiform signs. Such customary readings have arisen from the piece-meal growth in understanding of Akkadian and Sumerian. For example, in 1940 Kramer published an edition of the “Lamentation over the Destruction of Ur”. This is a Sumerian poem, some 436 lines long, bemoaning the destruction of Ur at the end of the Ur III period; it was written probably about a century after its destruction. In his Introduction, Kramer says that “The time is not yet ripe for a thorough and scientific overhauling of the Sumerian system of transliteration”. Therefore, he “deems it best to follow the more or less established usage”. In this system,
In the case of signs representing roots that end in a consonant and may have either the long or the short value (e.g., the signs for pa(d), “to call”, du(g), “good”, etc., which may be read either pad, dug, etc. or pa, du10 ‘ etc.) the transliteration uses the longer value in spite of the fact that the shorter is scientifically the more correct. Only in cases such as u(d), “day”, and sa(g), “heart”, where the shorter value has become more or less standard, is that value used in our transliteration, although the inconsistency in transliterating
the signs for pa(d) and du(g) as pad and dug while giving those for u(g) and a(g) as u4 and sa is only too patent (1940:6).
Kramer is obviously irked by this inconsistency, but feels that there is nothing he can do about it. Although he wrote this passage almost fourty-five years ago, some editors of Sumerian texts still follow such customary usage. A compromise made by some Sumerologists is to put the Auslaut within parentheses, e.g., kala(g). However, if the short and long forms have different indices, this can create confusion; some scholars transliterate as ti(1), others as ti(1). All word-final consonants have been consistently transliterated (and transcribed) .
There were undoubtedly other features in the spoken language, which the writing system only hints at. There is only marginal evidence, for example, to determine wordstress, and it will not be dealt with here. Similarly, there is only the most indirect evidence
Because of what is claimed to be a large number of homonyms in Sumerian, it has several times been argued that Sumerian possessed phonemic tones. Diakonoff, for example, says: “Sumerian was certainly a tonal language, or else the many homonyms would have made spoken Sumerian quite unintelligible” ( 1983:86). However, the evidence is indirect and slight. In fact. many words which earlier Sumerologists believed to be homonyms have been shown to contain different Auslauts, and so are not actually homonyms.