The Life and Death of the Sumerian Language in Comparative Perspective

Piotr Michalowski
University of Michigan
Ann Arbor

I. Introduction

It is a commonplace that history begins with writing. When the term prehistory was introduced in the nineteenth century, it was strictly associated with archaeology and geology, that is with sciences that could delve back into times of the "fog and the flood" (Daniel 1962). The Biblical account of the flood provided the stuff of many Victorian metaphors, as it allowed for a conceptual link between scholarly endeavors, morality, and theology. Assyriology was practically born as a result of the discovery of a text describing the flood, and many years later the search for the origins of the Sumerians was linked with purported archaeological evidence for such a flood that were found in the ancient cities of Ur and Kish. The decipherment of the cuneiform script and the recovery of the languages that were used in Mesopotamia provided a solid foundation for "history;" its early phases and the times that preceded it were inferred from a developmental view of how human societies evolved. Although much new evidence is available to us today, interpretations of this material are still captive to the fancies of an earlier age. Our histories of ancient Western Asia often remain constructed in terms of the rise and fall, of growth and decay of nation-states, and of the cyclical domination of "peoples," be they Sumerians, Babylonians, Hittites or Kassites. These same metaphors often dominate our thinking about the rise and fall of languages, especially as these are seen to have been the attributes of specific peoples or folk. Such thinking went out of fashion to a degree, but is now making a comeback of sorts. This is particularly true for the earliest epigraphically documented epochs: here the debate about the role of Sumerians and Akkadians, or more generally, of Semites, in the early history of Mesopotamia is inextricably linked with controversies about the Sumerian and Akkadian languages. The case of Sumerian and Sumerians-sometimes even flavored by modern nationalism and racism—is perhaps the most complex of them all (Cooper 1991).

The death of Sumerian has been the subject of speculation, or rather of asserted theses, but it has never been discussed at length. There is agreement on one thing only: the spoken language died in antiquity. The date of this death has been variously estimated; at one extreme there was Joseph HalJ vy (Cooper 1991), who claimed that there never was such a spoken tongue, on the other Stephen Lieberman (1977: 20) who seems to imply that there were Sumerian-speaking pockets in the south during the Old Babylonian period. HalJ vy aside, the death of the language has been variously placed as early as the beginnings of the third millennium or as late as the middle of the second (Cooper 1973). Banal as it may seem, there is only one undeniable fact in this discussion: no one alive today speaks Sumerian. Since the language was used in written communication for millennia, we infer that it was spoken at one time and that there had to have been a Sumerian vernacular. This assumption seems reasonable, but were we asked to support it with evidence, we would quickly find ourselves in difficulty. The problem is primarily methodological: what criteria would one use to argue for the life or for the death of an ancient language? Most recent statements on the subject have relied primarily on the language and distribution of personal names (e.g. Heimpel 1974-7), but such investigations have been based on a set of unquestioned assumptions about the relationship between spoken language and onomastic practice. If this is not good enough, what kinds of data, short of native testimony, would we want to marshal in a discussion of the death of a language like Sumerian? There are no easy answers. The fact that we are dealing with the written remains of a language and that most of our data on the language comes from some time after its putative demise, creates a situation that is unusual, but not unique. In view of the socio-linguistic context of the use of the Sumerian language I would suggest that we cannot simply ask the usual question: "when did the language cease to be spoken," or, as some would prefer to phrase it: "when was it no longer understood in vernacular conversation." This is obviously a complex matter that requires several different modes of investigation. In this short essay I shall limit myself to a few preliminary questions: what was it that actually died out, and what are possible socio-linguistic ramifications of language shift in early southern Mesopotamian society?

II. Languages in Early Western Asia.

The facts are limited and will probably stay that way for a long time to come. There is no way of avoiding the obvious: we cannot recover information on languages before the invention of writing, and all the languages that we know of from early Mesopotamia have been dead for a very long time. This has not always been an obstacle to speculation on fourth and fifth millennium languages, including reconstructions of putative "Proto-Tigridian" or "proto-Euphratean" tongues that served as the substrate for Sumerian (most recently Bauer 1999:436). The existence of these languages has been inferred from the analysis of certain "non-Sumerian" elements in the Sumerian lexicon, but the arguments do not hold up very well, as it turns out that most of these words are either native Sumerian, Semitic loans, or Kulturw` rter (Rubio 1999). One must admit, however, that most of the toponyms in Southern Mesopotamia are neither Sumerian nor Semitic.

In recent years there has been much speculation on the deep prehistory of languages, and brave linguists, primarily from the former Soviet Union and from the United States, have attempted to reconstruct linguistic families on a larger scale, such as Amerind, which would embrace all or most of the languages of the New World or Nostratic, which would include many of the languages families of the Europe and Asia (Diebold 1994, Campbell 1999: 311-326). There have even been attempts to link the distribution of language families with human genetic traits (e. g. Cavalli-Sforza 1997), although there has been strong opposition to such an approach (e. g. Noncentini 1993, Straus et al. 1998). Whatever one might think of these works, the new focus on language reconstruction has focused attention once again on historical linguistics and on such matters as the relationship between the study of language history, genetics, cultural evolution, and archaeology. Within the broad range of possibilities suggested by long-term linguistic history we can take a brief look at the context in which we must situate the isolate Sumerian.

What language or languages were spoken in Western Asia during the fourth, third and early second millennia B.C.? In Western Iran we have evidence for Elamite, possibly, but not assurebly the western tip of a Dravidian continuum (Diakonoff 1967:107-112, McAlpin 1981); in southern Mesopotamia there was Sumerian as well as a number of Semitic dialects, and in northern Mesopotamia and in Syria one has to posit a continuum of Semitic languages and dialects. Suddenly, in the middle of the third millennium one encounters Hurrian speakers (Michalowski 1986). Premature announcements to the contrary, there seem to be no traces of Hurrian presence in Syria in the early documents from Ebla and Mari. Towards the end of the third millennium, in Old Akkadian texts, we find the first occurrences of Hurrian words, personal and place names, the latter from northern Mesopotamia west of the Tigris (Steinkeller 1998). With the exception of Semitic, which I shall discuss briefly below, the broader family affiliations of these languages are a matter of speculation. Diakonoff (1967, 1984, 1986) and other Russian scholars have argued that Hurrian and Urartian were both part of the Nakh-Dagestan branch of the Northeastern Caucasian language family. Similar claims have been made for Hattic, the pre-Indo-European language of eastern Anatolia (Diakonoff 1984: 5). In addition, some have placed Indo-European speakers in Western Asia at an early time, albeit without much success. However one views this, one thing seems clear: there is no evidence of very early contact between IE and Semitic (Haarmann 1994) and so the IE dispersal either originated elsewhere or took place before the first Semitic spread.

In her highly influential book on Linguistic Diversity in Space and Time, Johanna Nichols (1992) has provided a new way of juggling genetic and areal linguistic history. She distinguishes between spread and residual zones; in her terminology the ancient Near East was a spread zone. Clearly, the area we are interested in has been affected by a number of Semitic spreads. In historical times we must reckon with three, if not, more such spreads: the early one that gave birth to the languages attested in the ED III documents, Amorite, and the controversial Aramaic spread.

The first spread is an oversimplification, as we are still at odds over the relationship between the earliest languages attested in Sumer and Akkad and those known from Syria. The matter of speakers of Amorite dialects is one of the more interesting and best-attested linguistic spreads in early Western Asia. The personal names that are attested in the third millennium Ebla, Mari, and Beydar tablets show no traces of Amorite. Although there is some variation in naming patterns from various regions of Syria, in general it appears that the geographical horizons attested in these documents bear witness to a prior spread of an earlier group of Semitic languages, including vernaculars whose written expressions are known as Eblaite and Old Akkadian. The relationships between the written and spoken tongues, as well as the range of local and social variation of these languages is a matter of some debate and, for obvious reasons, is also the subject of much speculation. There is very little written information between the time of these third millennium texts and the Mari tablets from the eighteenth century. The only post-ED texts from Syria that precede the main Mari archives are the so-called shakkannakku period tablets, some of which are earlier, and some of which overlap with the earliest texts from the reign of Yahdun-Lim. The written conventions of these texts are different, and although the language is Semitic, it is closer to Eblaite than to the Akkadian of the later archives. This does not mean that it was the vernacular, however. Just as the Akkadian of Mari was an import from elsewhere, specifically from Eshnunna, so it is equally likely that the earlier conventions are also expressions of an earlier written koine (Michalowski 1987).

For decades the question of the origin of Sumerian was predicated on the notion that the ancestors of the people who spoke the first attested language of southern Mesopotamia had to have come from somewhere else and were intruders in the area. The isolated position of Sumerian, tucked away in a corner at the end of the Semitic spreads, suggests a very different scenario. On comparative grounds, it is more probable that this language represents but a remnant of a much broader linguistic continuum, areal if not genetic, that had occupied much of Western Asia before the Semitic spreads. Even when a language belongs to an extensive family, the issues of how to identify a "language" and the moment of its very beginning an imposing if not an impossible task (Seebold 1996); to speculate about the origins and prehistories of ancient isolates may be somewhat foolhardy. Nevertheless, one could suggest that Sumerian and Hattic occupied a historical niche analogous to Basque and Etruscan in Europe—not that there is any agreement on the historical status of those languages. The isolated nature of Sumerian is further brought into relief when we take a look at certain typological features of the language. In contrast to all the surrounding languages, Sumerian has ergative argument marking and, as Nichols (1994) has shown, ergativity is relatively stable in areal terms and ergative languages tend to cluster together. On a very large scale her arguments are convincing, although once one begins to look more closely at the history of individual languages, problems do arise.

Consider, for example, the following broad-sweeping statement (Nichols 1994: 74):

"These spreads seem to have driven an accusative wedge into a continental interior that would otherwise be heavily ergative, for the languages all round the periphery are ergative: counterclockwise from the west they are Basque; three families of the Caucasus; Elamite (stative-active on an ergative base), Sumerian, and Hurro-Urartean of the ancient Near East; Burushaski and Tibeto-Burman languages to the south; Chukchi-Kamchatkan and Eskimoan on the northeast; Ket (stative-active on an ergative base) to the north. Only to the southeast are there accusative languages: Chinese, Japanese, Korean, Ainu, Nivkh (Gilak)."

Much depends on the level of generalization one is willing to live with; the ergative-based analysis of a Caucasian areal grouping has been questioned (Tuite 1999), the interpretation of Elamite offered above is open to serious doubt, and the chronological position of earliest Hurrian also creates problems. Nevertheless, the distribution of this pattern is suggestive.

All of this leads me to posit once again that Sumerian was not the language of some overland or maritime marauders who had settled in Mesopotamia sometime in the fifth or fourth millennia. The history of archaeology is filled with speculation on cultural rifts that may signal the arrival of some new population group that may perhaps be identified as "Sumerian" (Jones 1969). The evidence is tenuous no matter how one looks on the matter, but I would argue that several unrelated sets of data support the view that Sumerian was one, if not the major spoken language in Mesopotamia from very early on. The relatively late human occupation of the southern Mesopotamian alluvium does not leave very much room for too much linguistic history. We must also accept that there is at present no evidence at all for any other early language in the area, now that Gonzalo Rubio (1999) has convincingly smitten the Proto-Tigrido-Euphratean dragon. The Sumerian language is definitely present in some form or another in the Uruk III tablets and short of a miracle we shall never go back much farther than that as far as direct evidence for language history is concerned.

III. Sumerian and Akkadian

Building on the work of predecessors, in particular on the researches of I. J. Gelb and F. R. Kraus, Jerry Cooper (1973) two decades ago provided a succinct and balanced overview of the state of our knowledge concerning the relationships between Sumerian and Akkadian in Sumer and Akkad. His conclusion, that "in the absence of evidence to the contrary, it does not seem unreasonable to posit a situation in which the displacement of Sumerian as a spoken language in Sumer was in process in ED III (1973:242)," appears reasonable even today. Cooper clearly sided with Kraus, who saw little evidence for distinction between Sumerians and Akkadians in historical times; the position of Gelb, who was more inclined to see a separation between a northern Akkadian speaking core as opposed to a south dominated by Sumerian speakers, is maintained to this day by some of his students, primarily by Piotr Steinkeller (1993), but also by Aage Westenholz (1999: 25ff.) and others. Although some of the arguments used by these scholars are linguistic, they also stress cultural differences. As much as I admire their work, I have methodological qualms about notions such as "nation," "people," and "ethnic identity," notions that probably do not go back further than the seventeenth century in the West, but are often projected into the past (Smedley 1999). This is not the place, however, to debate these issues.

Lexical borrowings are the main evidence for interference between Akkadian and Sumerian. The most commonly cited evidence of syntactic borrowing has been the matter of word order in Akkadian. Since almost all Semitic languages have a basic VSO word order, the SOV structure of Akkadian has almost universally been attributed to the substrate influence of Sumerian, and is cited as evidence for a long period of co-existence of the two languages, if not for a prolonged period of bilingualism in Sumer. On the surface, this seems to be a reasonable assumption. There is an analogy to the situation: the SOV word order of Amharic is ascribed, together with other features, to the substratum influence of Cushitic. There have been challenges to the standard interpretation of the origin of Akkadian word order. Talmy Givon (1975, 1976) has argued that Proto-Semitic was verb final and that Akkadian, which branched off earlier than the West Semitic languages, actually preserved the original word order, and was therefore not influenced by Sumerian, although he does not explain how other languages in the group became verb initial. His analysis was based on the order of pronominal affixes in the verb. G. Haayer (1986), while acknowledging Bernard Comrie's critique of Givon, took the idea further and offered his own explanation of the SOV order of Akkadian. Following Diakonoff (and in a way similar to the work of H. P. Mh ller [1995]), he hypothezes that Proto-Semitic had ergative noun marking; from this he deduces that the basic word order of Proto-Semitic must have been SOV, since many ergative languages are verb final. His initial claim is that Sumerian should not have influenced Akkadian syntax. He does, however, contradict himself, and argues that while other Semitic languages changed, Akkadian, under the influence of Sumerian did not. Lieberman (1986), apparently independently, provided a more complex analysis of the order of bound morphemes on verbs; he observed that Proto-Afro-Asiatic probably had dominant SOV word order that changed to VSO when the verbal system shifted. He points to Cushitic and Omotic, both of which have SOV order. Like Givon, he suggested that Akkadian had branched off before the shift. His arguments may have to be revised in view of the interpretation of the relationships between constituent structure and word order recently presented by Dryer (1992), but for the time being one may accept Lieberman's formulations as a working hypothesis.

If we follow Lieberman’s argument then the main evidence for syntactic interference in Akkadian under the influence of substrate Sumerian evaporates. There may be other ways of looking at this problem that result in similar conclusions. Word order has been extensively studied and while there is some disagreement as to whether basic word order is really basic, if we accept such a category for the sake of the discussion, certain interesting consequences do follow.

Johanna Nichols (1992, 1995), in an extensive comparative diachronic as well as synchronic analysis of the matter, concluded that word order was a stable areal phenomenon, but was unstable genetically. Indeed, in her ranking of stability hierarchies word order was the most unstable feature in language stocks, and the most stable areal feature (the other features are alignment, head/dependent marking, and complexity). On the basis of this, one might conclude that Mesopotamia, Syria and surrounding regions constituted a linguistic area, after Akkadian broke off from its stock, but before the wide spread of Semitic, characterized by SOV word order, among other features. Most of the old languges of the area have verb-final syntax: Elamite, Hattic and Hurrian/Urartean, although in the case of the latter we must assume that its precursors were spoken on the fringes, further north than they are attested historically.

If we exclude the prize example of syntactic interference, we are left with loan words and loan morphemes to trace the influence of Sumerian on Akkadian. It is often said that Akkadian borrowed a large vocabulary from Sumerian, but no one has ever actually studied the matter in full. Lieberman (1977) catalogued Sumerian loans in Old Babylonian Akkadian. He found 529 nouns, of which 102 are known only from lexical texts. Not all of his etymologies would be universally accepted, but even if we accept a rough figure of approximately 400 loans, that is hardly a large number. One would need to study the semantic classes and frequency distributions of these words in order to arrive at any firm conclusions, and any full analysis of the matter would also have to take into account divine and personal naming patterns as well. However, from a quick look at Lieberman’s work and at the statistics provided by D. O. Edzard in his reviews of the Chicago Assyrian Dictionary volumes, one may conclude that while there was a small percentage of Sumerian loans in Akkadian, it was hardly an overwhelming phenomenon that would provide evidence of mass lexical interference.

The situation is somewhat different in Sumerian. Unfortunately, even less has been done on Akkadian and Semitic loans in that language. Many that have been concerned with the issue would probably agree that from the earliest readable texts on, we find extensive evidence of Semitic loans (noted already years ago by Falkenstein, Gelb and others; see, for example, Rubio 1999). Indeed, I would risk the statement that the Sumerian we know has a much larger percentage of Semitic and Akkadian loans than the other way around. Assuming that this is correct, what does it tell us about the linguistic situation in early Mesopotamia?

Briefly stated, it appears that one can entertain two mutually complementary hypotheses. First, the unequal evidence of lexical interference would suggest that most of it took place in a situation in which the social prestige of Akkadian was much higher than that of Sumerian. Second, and more radical, would suggest the working hypothesis that much of what we call Akkadian was a different languge, or dialect cluster, from the Semitic languages or dialects that were present in Babylonia when Sumerian was still a living tongue. This would find support in the work of Militariev (e. g. 1995, 1996), who argues that many early Semitic loans in Sumerian came from languages different from Akkadian, although his work is not without its critics. Without doubt, the history of what we call Akkadian is a complex matter, one that is not adequately handled by the charts and nomenclature in our grammars. In many of these we find an explicit or implied evolutionary progress from Old Akkadian, through the bifurcation into Old Babylonian and Old Assyrian, and then onwards and upwards through these two "dialects," ignoring cautionary warnings. More than three decades ago Erica Reiner (1966) observed that "no one has ever shown that Old Babylonian ever descended from Old Akkadian" and I think that this still holds true today. More recently, Westenholz (1999: 33) has summarized his own and Gelb’s thoughts on Old Akkadian in the following manner: "It has been known for some time that Sargonic Old Akkadian—certainly the dialect used by Sargon himself and presumably the dialect of Agade elevated to official language of record—is visibly different, both from the Pre-Sargonic Akkadian dialect spoken in Babylonia, and from the Akkadian of the Ur III sources, the latter being essentially an archaic version of Old Babylonian." Current debates over the status of Old Assyrian only serve to highlight the problem of the complex history of "Akkadian."

The complicated language situation in early Mesopotamia has been apparent to all who work on third millennium texts. Among the Early Dynastic literary tablets found at Abu Salabikh, there is one unusual text (Biggs 1974 no. 326). The tablet did not look like the other literary texts from the site, and the copyist noted (Biggs 1974: 32) that "noteworthy is no 326, in which j , "and," occurs; even at this early date Sumerian may have been under heavy Semitic influence." No progress was made until the recovery of the Pre-Sargonic archives from Ebla, in Syria, provided us with a wealth of new written documentation. Among the literary texts from the site was found a syllabically written duplicate that demonstrated immediately that the text from Sumer was a literary text in a Semitic language (Edzard 1984 no. 6). Biggs himself had earlier written about the difficulties of identifying the language of some early cuneiform texts, and his hunch proved to be more than correct.

In theory, one could say that there are only so many possible ways of using a system such as cuneiform for writing Sumerian and Akkadian. As has been noted (Civil 1984:76, Civil and Rubio 1999:264), a text could be:

1. written and read in Sumerian,

2. written and read in Akkadian or some other Semitic language,

3. written in Sumerian and read in Akkadian, or some other Semitic language

4. written in Akkadian and read in Sumerian.

5. written in Sumerian but read as Semiticized Sumerian

An early tablet inscribed exclusively with word signs would seem to be linguistically impenetrable; if we knew all the signs then we could understand it, but we could not be sure of the underlying language. In practice, there are a number of clues that one can use to establish the probable linguistic identity of a given text. The origin of the object provides some information: a text from northern Babylonia is more likely to be Semitic, while one from Sumer should be Sumerian. This is only a rule of thumb, however, since there are demonstrably Sumerian texts from the north and Akkadian ones from the south. Another indicator might be found in just one or more syllabic signs that provide readings from a specific language. Thus if a royal inscription seems to be written in Sumerian, but contains, before a place name, a syllable /in/ that can only be interpreted as the Akkadian particle "in," we may assume that the whole text is to be read in Semitic. Other indicators are less obvious. There are a few logograms and syllabic signs that were used exclusively in Akkadian and other Semitic languages such as Eblaite, but never in Sumerian. A good example of this is the word for "witness" which was written in Sumerian as ki inim-ma, corresponding to Akkadian š§bum, which means "elder," as well as "witness." The standard Sumerian signs were not used as logograms in third millennium Akkadian texts. Rather, this was expressed by means of the logogram that we transliterate as AB+ÁŠ, and which corresponds to the Sumerian word for "elder." These kinds of rules would seem to provide sufficient criteria for the linguistic identification of texts, but certain texts defy easy classification. Take, for example, the following fragmentary tablet (Westenholz 1975:36):

1. [1] lugal-a₂-zi-da                             Mr. Lugalazida
2. arad₂ Lugal-ki-gal-la                        the slave of Lugalkigal
3. ensi₂-da                                            from Mr. Ensi
4. in-da-zah₃                                         escaped.
5. ki zah₃-a-na                                      His hiding place
6. geme₂ ur-nigin3                                the slave girl of Urnigin
7. ba-dug₄                                             disclosed:
8. in maš-ka₃-ni^ki-[ŠA]BRA                "In (the town of) Mashkan-shapir..."
9. u-/ša\ -ab                                      he is dwelling;
10. /li-ru^?-u₃-nim\                            he should be brought here."
rest broken

Armed with the criteria enumerated above, how are we to establish the language of this legal text? On the basis of the sign forms we can date it to the time of the dynasty of Sargon. The provenance of the tablet is uncertain; it may have come from Nippur, but that is not a decisive feature since we have Sumerian as well as Akkadian texts from the city from this time period. One must add that the text is broken and that better information may have been contained in the missing lines, but the investigator must confront, more often than not, such incomplete tablets. Suddenly we find that the neat series of possibilities enumerated above may prove to be insufficient. Could one infer that the whole text was Akkadian, and that lines 1-7 are simply Sumerian logograms that were designed to be read in the Semitic tongue? More interestingly, however, one could claim that the two languages were intermingled; the main protocol of the juridical proceedings was redacted in Sumerian, and the testimony of the main witness, registered as direct speech, was written down in the language in which it was given, that is in Akkadian. Indeed, this is what the two editors of the text, Edzard (1968:137-8) and Westenholz (1975:36-7) claim, and they may be right. The language of writing for this particular scribe and his milieu was still Sumerian, but in reporting stylized direct speech he acknowledged that the vernacular in his world was already Akkadian.

Such obvious conclusions do not exhaust the inferences of this text. It appears to be written in two languages, corresponding to the two linguistic protagonists of our stories. It is highly improbable, however, that Mesopotamia was ever a strictly monolingual or even bilingual area. One must, of course, allow for localized pockets of such language uses, but from a larger perspective one must allow not only for areal and social dialect differentiation, but also for a variety of contact languages as well. The constant interactions of Mesopotamians—whatever their native languages—with neighbors who used a variety of unrelated vernaculars, certainly must have given rise to a variety of contact languages. The multilingual ancient Near East must have produced trade jargons, interlanguages, and possibly even pidgins and creoles. One usually associates the latter two with European expansion in the " age of exploration," but as Thomason and Elgibali (1986) have demonstrated, such linguistic phenomena can be documented much earlier and outside of the sphere of European expansion. Because of social restrictions on literacy, such contact languages would never make it into writing. The text cited earlier only serves to highlight one more the autonomous nature of writing and the formalization of written norms: the restricted and stylized versions of Akkadian and Sumerian that were preserved in permanent form do not in any way represent spoken language, alive or dead.

Our speculations can only hint at the socio-linguistic complexity of the cultural milieu of early Mesopotamia and little of what we suspect can be documented on the basis of the preserved textual record. One of the main reasons for this must be the immense chasm between spoken and written language. As we have seen, the vernacular, or rather the vernaculars, were constantly changing; they changed by language contact, and undoubtedly by local variation in cultural and historical circumstances. Most historical linguists today recognize that linguistic change and variation is primarily driven by socio-linguistic forces rather than by strictly structural matters (Thomason and Kaufman 1988). The ebb and flow of language variation, and the events that eventually led to language shift are hardly reflected in the texts. Think of how little we know of Sumerian dialectology. Most scholars, if asked about the matter, would answer with reference to the "main dialect" as opposed to the "emesal dialect." There is absolutely no indication, however, that the latter was ever spoken, and it seems to be very much a cultic reading convention rather than a true dialect. Although some have suggested that it had roots in the third millennium or earlier (Alster 1982, Maul 1998), there is little to support such a reconstruction. The word "dialect" seems to be used differently in Assyriology than in linguistics. In a recent cultural survey we read of a "dialect," but the only characteristic of this dialect is a phonetic difference in verbal prefixes (Westenholz 1999: 38). Natural language does not behave in this manner. I have sketched some of the cultural context of Sumerian, reflecting only on selected issues, for one purpose only: to posit major differences between spoken and written versions of the language.

IV. What is Sumerian?

What then is, or rather, what was Sumerian? Hoyrup (1992 and many other places) has argued that it was a Creole, but his arguments do not wash from a linguistic or Sumerological point of view. I do not want to debate the issue here, but will only note that one would be hard pressed to find an example of a language with the typological features of Sumerian whose origins could be found in creolization. He was undoubtedly influenced by the attempts to brand Middle English as a Creole, attempts that have been well refuted by now (Dalton-Puffer 1995).

In Assyriology we are used to collapsing broad diachronic and synchronic spans with a single linguistic label such as "Sumerian" or "Akkadian." Since Sumerian is known to us solely through the medium of writing it is extremely difficult to disentangle linguistic features of written language from anything else. Much of what we conceive of as historical language development can be conceived of as change in writing conventions. The vernaculars must have had more differentiation than we can detect in the written tradition, as there is simply not enough change in the language of the texts over a long span of time. Perhaps the best example of documented change is to be found in the early lexical texts. Since we read the language backwards, from the better documented and better understood early second millennium texts, it comes as no surprise to find that many words that have been discovered in the earliest lexical lists cannot be translated. Some of this is due to difficulties with the writing system, but in many cases we can be certain that we do not know the meaning of words because they had gone out of use and were replaced by others, sometimes in relatively early times. Nevertheless, from ED III times on, much of the change in the language that we can follow must be related to written conventions.

Such changes are usually described as changes in "spelling." Leaving aside the appropriateness of the term for such a writing system, I would like to suggest that much more is at stake here. We have no direct witnesses to any of the reforms that conditioned changes in the way cuneiform was written. Leaving aside for the moment earlier possibilities, there can be no doubt that one such reform took place during the reign of Naram-Sin of Akkad and that the reform was centrally imposed, with results that can be detected throughout the kingdom. How this was done we shall never know, but at least some of the effects must be attributed to central schooling. How else can we explain a relatively trivial matter, namely the abandonment of the use of an "upward" (from our point of view) stroke, a stroke that was by then limited to two signs, DA and CU? For us this is an important diagnostic feature, but at the time it could hardly have been a cosmic issue, and one can hardly imagine that it was enforced by royal edict. Most probably such a change, equivalent to crossing a t or dotting an i, was a random aesthetic change instilled at an early age by some influential schoolmaster in Agade. His students, and the fashion that went with them, spread throughout the realm, taking with them the new writing habit. But fashion in antiquity meant something completely different that it does today; communications were much slower and the role of tradition stronger. The speed with which this particular change spread is just one more indication of the way in which the administration of Naram-Sin imposed itself on local organizations together with a new set of bureaucrats and scribes.

There is little doubt that one or more writing reforms have to be attributed to the kings of Ur. The Sumerian literary texts written under their patronage have survived primarily in copies from the Old Babylonian period and, with a few exceptions, they have come down to us written in a manner that is quite different from the writing norms known to us from Ur III times. A few Shulgi hymns have writings that either preserve some earlier habits or were consciously remade at some point to approximate, perhaps infelicitously, earlier norms (see Klein, this volume).

We shall have to await the publication of Gonzalo Rubio’s dissertation on the Ur III literary texts, the majority of which remain unpublished, before we make any rash statements about the differences between the literary texts of that time and the Old Babylonian texts that we know. Can we imagine, however, that these changes were concerned only with writing conventions and a few lexical items that had gone out of use? Comparison with similar changes in written norms suggests that we have to allow for the contingency that much was at stake.

Roger Wright (1982), writing about the rise of the Romance vernaculars, and on the invention of a new way of pronouncing Medieval Latin insisted that (Wright 1991: 111) "Latin—Romance distinction of Later Middle Ages was created through such language planning, and that it would not have existed if it had not been invented." Charlemagne and Alcuin’s reforms of Latin pronunciation created a new oral form of delivery that had little to do with the language of the old empire. Citing the example of these Latin pronunciation reforms, the late Chaim Rabin (1985) claimed much the same for Byzantine Greek, Arabic, and Hebrew. In all cases the standardization of written versions of the language went hand-in-hand with reforms of norms for reading aloud. It seems reasonable to assume that in Mesopotamia the periodic changes in writing habits were accompanied by new grammatical as well as phonological norms, norms which may have taken the language quite far from its vernacular roots, including special cultic traditions that go under the name of eme-sal. I would suggest that the Early Dynastic "Semiticisation" conventions of Sumerian, as described by Civil and Rubio (1999:263-266), are but one instance of such formalized, invented processes of writing and reading. The proper metaphor for the history of the written Sumerian known to us is not gradual evolution but punctuated equilibrium (taken from paleontology and complex adaptive systems), a metaphor I have previously invoked in reference to the very origin of writing in Mesopotamia and elsewhere. If we accept a historical chasm between the written language, with its own complex history, and whatever vernaculars were once used in the land, the issue of the death of Sumerian has to be seen in a new light, since we must ask ourselves what exactly died and when.

V. The End of Vernacular Sumerian

Leaving aside the long second life that written Sumerian had in the schools and temples of the ancient world, we are confronted with complete silence on the demise, even on the very existence of everyday spoken version of the tongue. We assume that a large proportion of the population of Southern Mesopotamia once spoke the language, and that these people quietly switched to using Akkadian some time towards the end of the third millennium. How could something like this happen? Were the Sumerian speakers butchered to such an extent that the language died out with them? This hardly seems possible, although the Sargonic kings, did their utmost to decimate the inhabitants of the South, if we are to believe some of their claims. If we rule out such catastrophes, then we are probably dealing with but one more instance of language shift, that is the replacement of one spoken language with another. The best succinct statement on these matters was given to us almost thirty years ago by Cooper (1973: 241), who wrote: "Barring violent incidents, such as wholesale annihilation, deportation, or deliberate suppression, language displacement is a slow process, and occurs when the bilingual community expands to include all members of one mother-tongue group, who then neglect to teach the mother tongue (here Sumerian) to their children." How does this statement conform to the current state of our knowledge about early Mesopotamia and comparative research on language death? Without concrete data linguistics cannot predict specific historical events, in the future or the past, but some patterns are observable through the comparative study of language shifts.

In her pioneering work on East Southerland Gaelic in Scotland Nancy Dorian (1981: 51) observed that "in terms of possible routes towards language death it would seem that a language which has been demographically highly stable for several centuries may experience a sudden "tip," after which the demographic tide flows strongly in favor of another language." There has been a revival of interest in language attrition and death during the last decade or so, much of it strongly influenced by Thomason and Kaufman (1988) and the continuing work of Nancy Dorian herself. While details differ, there is much agreement that "heavy cultural contact is a necessary precondition for structural incursions of one language into another, but contact itself is not obviously the structural mechanism involved" (Meyers-Scotton 1998 :289). The study of dead and dying languages has revealed many instances of structural adjustment and attrition along the way, but none of the phenomena that accompany language shift can be said to be necessary and universal in such situations. Thus, while comparative linguistics can provide us with possible models for the shift from Sumerian to Akkadian, that is all it can do. Scholars working on the subject have claimed instances of case syncretism or merger (Huffines 1989), phonological, morphological, and syntactic reduction (Cambell and Muntzel 1989), or, more specifically, reduction in the use of relative clauses (e.g. Hill 1989, Rottet 1998).

Not all linguists working in this area agree with these findings. Can one truly blame external linguistic influence for language death or attrition? Perhaps no one has opposed this view more consistently than Eung-Do Cook, who writes (1995:218) that: "fluent speakers of a dying language maintain its conservative characteristics with no evidence of convergence, while the simplification in the speech of semispeakers in internally motivated." Cook (1995:227) further concludes that "the process of simplification and decay in language death is due to semispeakers’ impeded and prematurely terminated learning process." Hoenigswald (1989), Huffines (1989) and others have expressed similar views. This means that socio-linguistic patterns have an effect on internally motivated change in a dying language; such change has often been misinterpreted as borrowing from or convergence with a dominant tongue. Moreover, such change is often not gradual but relatively quick. This dovetails with Dixon’s (1997) recent arguments for a new model of historical linguistics that recognizes long periods of stability, disrupted by drastic, relatively rapid change. He even uses the metaphor of punctuated equilibrium to describe this model of language change.

Although the events we are trying to recover most probably took place centuries earlier, a cold, hard look at the historical evidence would suggest that the best historical backdrop for the "tip" of Sumerian was the reign of the kings of Agade. Since the shift often involves differences in social standing of languages, it need not have required any great new population to trigger the final move towards Semitic. We know that after Sargon defeated the south, he initially left many of the local elites in power, but after rebellions he and his successors, especially Naram-Sin, installed their own people in high places throughout the realm in political and religious offices. This created a new aristocracy in the realm and it is most probable that the new elites were primarily Akkadian speakers. Centralized schooling at the capital undoubtedly created a situation in which access to bureaucratic careers involved Akkadophones, or bilinguals. If we pursue this line of reasoning and accept its premises for the moment, we cannot avoid arriving at a fascinating paradox. The shift to a Semitic language of one sort or another, which had roots in earlier times, but was precipitated by changing social and political realities, probably reached maturity in Ur III times. This means that the reign of the family of Shulgi, long considered a "neo-Sumerian renaissance," and a reassertion of a putative Sumerian national pride (Becker 1985), was in actuality the epoch in which Sumerian was emitting its last breath. This would mean that the return to Sumerian as the language of accounting and bureaucracy throughout much, but not all of the new kingdom, took place just as the language was losing ground on the streets.

Such a scenario would fall in line with Walter Sallaberger’s recent assertion (1999: 129)—arguing against both Cooper (1973) and the present writer (1991)—that Sumerian had to be a living language, albeit under pressure from Akkadian, in Ur III times. While I am not unwilling to entertain this possibility, I remain unconvinced that personal names and choice of written language are in any way indicative of the language of the streets. Sallaberger (1999: 129 n. 28) is understandably suspicious of my comparative arguments, pointing out that "allerdings kann er dafür keine Argumente bringen, denn der Verweis auf Schreiberschulen Šulgis und ethnographische Parallelen allein reicht wohl nicht aus." The problem is that no convincing arguments can be mustered for the alternative view: that Sumerian was still a living language in Ur III times. Whatever the linguistic situation may have been on the streets of Nippur or the small villages of Sumer, there is a fascinating clue as to the language of the royal court. When the king of Mari decided to give his daughter in marriage to one of Ur-Namma’s princes, she took on a new name for the occasion, Taram-Uram, "she loves Ur," and the name was Akkadian not Sumerian (Civil 1962). One might be tempted to dismiss this as anecdotal—because Ur III royal children were named with both Sumerian and Akkadian names--but the serious diplomatic circumstances suggest a purposeful choice of homage in the prestige language of the court. This does not tell us if Sumerian was alive or dead, but it does provide a much better index of social prestige than does the officially imposed choice of written language.

It is an undeniable fact that vast majority of the 36,000 or so published Ur III texts were written in Sumerian. It is also often noted that these come from a limited group of southern cites; a small number of unprovenanced tablets as well as those from the northern town of IÓ~ n Mizyad were redacted in Akkadian (Mahmoud 1989). Other text collections such as the SI.A-a (Steinkeller 1989:305-307) and Tã ram-ili (van de Mieroop 1986) archives contain a mixture of Sumerian and Akkadian documents. Can this be used as evidence for vernaculars in these areas? It is interesting to note that the geographical distribution of Sumerian and Akkadian language texts is different in Sargonic and Ur III times. Many Old Akkadian letters and documents are official documents used in communication with the capital, and that explains the choice of language; nevertheless, the patchwork of Akkadian and Sumerian in archives from the time is instructive (Foster 1982). The choice of official language is a political and ideological issue and not an ethnic one. Nevertheless, as Bram Jagersma pointed out to me, it does appear that until the end of Ur III areas such as the Lagash-Umma region show a pattern of consistent Sumerian writing combined with a massive predominance of Sumerian names (see now Bauer 1999:437). One could therefore argue that already in this period there were regional differences in language use and that this area was the most conservatively Sumerophone part of Southern Babylonia. One should also note the decline in occupation of this region in post-Ur III times, which may have been a factor in the demise of Sumerian in this part of the land.

The second matter that is often invoked in discussions of the language of the streets is the choice of personal names. By the second millennium this issue is surely most problematic. There are Sumerian language personal names from the Old Babylonian period, but they cannot be lumped together and used as evidence for a living Sumerian language. Scribal names have to be taken out of the count, as they were undoubtedly either given to the sons of literate officials and priests in anticipation of future schooling or taken on during the educational process. In many cases it is not actually clear that a Sumerian name was truly Sumerian. For decades scholars have wondered if common names such as Sin/Shamash-iddinam and Nanna/Utu-maÓ um, "The Moon/Sun God has given (a son)" in Akkadian and Sumerian respectively, were distinct names or simply different ways of writing the same Semitic name. Tanret (1996) has recently shown that in the northern city of Sippar, if not in other places as well, Sumerian names were for the most part simply alternative ways of writing Akkadian ones.

All of this is complicated by the history of Akkadian. If Westenholz is correct, and there is a major historical divide between Ur III and earlier Akkadian, then we are at a loss to explain the Sumerian loan words in Old Babylonian. Significantly, there are few such borrowings in Old Akkadian. This can only mean that the ancestor dialects of OB, and I am not sure that this does indeed encompass the attested Ur III Akkadian, distinct from the Old Akkadian that we know, were the ones used at a time when the two languages were in use in a bilingual situation. When and where this might have been, I do not know.

Thus, while Sumerian might still have been spoken in certain areas in Ur III times, there are other possibilities than the scenario outlined above. For example, we might seek the social conditions that could have given rise to the "tip" much earlier, prior even to the invention of writing. The Middle Uruk period could be a prime candidate. This was a time of immense social change, and with it must have come rapid disruption of the linguistic fabric. First, we have evidence of massive demographic shifts, including movements from countryside to the city. This must have brought speakers of new languages as well as of rural dialects of Sumerian into the linguistically volatile urban environment. People from Uruk and its vicinity moved to locations in Syria, Elam, and Anatolia (Stein 1999). They lived there in enclaves and within one generation that would have produced internal language change: the homogenization of the native language as well as influence from the local vernaculars. The dynamics of change were different in each settlement, as were the local languages that interacted with the speakers from Sumer. The collapse of the "Uruk expansion" could have created two separate effects. Some settlers stayed where they had lived, and blended into the local population, and sometimes this may have had local consequences that we can only guess at. Others undoubtedly returned to the Uruk heartland; their speech would have been archaic, preserving elements that had been discarded or changed in the center, as well as innovative, changed through contact with other languages unknown to the residents of Sumer. This kind of linguistic change, which took place over hundreds of years, may have been the catalyst that led to the socio-linguistic tip needed for the process of Akkadian replacement to have begun. Even that statement is an oversimplification. If the arguments about early Semitic in Mesopotamia sketched above should prove to hold, we must take conclude that the period leading up to the tip was characterized by a broad range of interference phenomena from a succession of Semitic languages, and not by Akkado-Sumerian bilingualism, the latter being but the final element in the process.

VI. Conclusions.

So how did Sumerian die, when did the last person who could understand the vernacular leave this earth? Did the heavens roar or was it more like Breughel’s Icarus, a tiny splash far off on the horizon? What was one more or one less language in history? After all, it estimated that "during the coming century,… 3,000 of the existing 6,000 languages will perish and another 2,400 will come near to extinction (Hale: 1998: 192)." If that is how we phrase the question, I am afraid that we will never find the answer. Although some Assyriologists question the use of comparative data to illuminate problems such as the one we are wrestling with here, but I am afraid that without recourse to linguistic research we will never come close to any answers. Not only do we not have any direct evidence at present, but we also do not know what evidence we are looking for. If we agree that the linguistic identity of personal names and the "evidence" of written language are not indicative, we have little to argue with. We look for philological clues, but, to cite Henry Hoenigswald (1989: 353) "obsolescence itself is a sociolinguistic matter and not a specifically linguistic one. Demise can be predicted, it seeems, only at the terminal stage, where it is obvious, what with a last speaker surviving in California or on some Dalmatian island."

Perhaps the issue is not the death of the last native speaker, but the life of the language we know. The search for a living, albeit inaccessible, Sumerian seems driven by an unwillingness to accept a high degree of autonomy for written language (Michalowski 1987). I would suggest that in addition to a variety of vernacular multilingual situations, we much also reconstruct a parallel for of what one might call "complex monolingualism," to borrow and alter Wright’s (1993) felicitous term. As already noted, the term "Sumerian," like any linguistic identity label, is a metaphor that stands for a broad range of variation in time and space. The written language and the pronunciation norms that went with it brought its own set of similar variations. If we accept the model outlined above, according to which the reading and writing conventions were periodically reinvented, then we must accept the consequence that we will never discover an ideal Sumerian phonological system, to cite but one element of the grammar. This has serious implications for the study of specific corpora. The Ebla texts have provided new evidence of glides in Sumerian (Civil 1984 :80), but does this have a bearing on any phase of Sumerian in the heartland, or does it just point to reading conventions used in a Syrian town?

Moreover, I would argue that the apparent evidence of convergence between Sumerian and Akkadian does not support the idea of interference in the process of a slow gradual replacement of one language by the other, nor is it suggestive of a relatively late date for the death of the older tongue. Quite the opposite, if one takes seriously the work of Cook, Dixon and others cited earlier, one might interpret such data as evidence for interference not among vernaculars, but in the restricted domain of written tradition. Pronunciation traditions as well as writing conventions of "standard" Sumerian would carry imprints of Akkadian, or even Amorite, linguistic traits in a manner quite different from a living language, alive or dead.

Strange as it may seem, my conclusions are not pessimistic. I reiterate that we still do not even know what criteria we might be looking for to answer the classic question on the demise of Sumerian. It is therefore better not to ask that question, but to continue to investigate the various lives of Sumerian, eschewing any notions of any "classical" form of the language. Seen in such a perspective there are no periods of decadence or incorrect norms, only normal linguistic change within writing rules as well as reading conventions. There are interesting consequences of such a point of view. Linguists have lamented the loss of intellectual, affective and poetic aspects of human life that die together with languages (Woodbury 1998, Mithun 1998). In ancient Mesopotamia, however, the long life of written Sumerian and its coexistence with written Akkadian, guaranteed the preservation and expansion of these cultural elements, albeit within limited social circles. Sumerian was a movable feast.

REFERENCES CITED

Alster, Bendt (1982), ‘Emesal in Early Dynastic Sumerian? What is the UD.GAL.NUN Orthography?’, Acta Sumerologica 4, 1-6.

Bauer, Josef (1998), ‘Der vorsargonische Abschnitt der mesopotamischen Geschichte’, in J. Bauer, R. K. Englund, and M. Krebernik, eds, Mesopotamien: Späturuk-Zeit und Frühdynastische Zeit (Orbis Biblicus et Orientalis, 160/1). Freiburg/Göttingen: Universitätsverlag/Vandenhoeck & Ruprecht, 429-585.

Becker, A. (1985), ‘Neusumerische Renaissance? Wissenschafts-geschichtlische Untersuchungen zur Philologie und Archäologie’, Baghdader Mitteilungen 16, 229-316.

Biggs, Robert D. (1974) Inscriptions from Tell Abu Salabpkh (Oriental Institute Publications, 99). Chicago: The University of Chicago Press.

Bonechi, M., (1991) ‘Onomastica dei testi di Ebla: nomi propri come fossili-guida?’, Studi epigrafici e linguistici sul Vicino Oriente antico 8, 60-79.

Campbell, Lyle (1999), Historical Linguistics: An Introduction. Cambridge: MIT Press.

Campbell, Lyle and Martha C. Munztel (1989), ‘The Structural Consequences of Language Death’, in Dorian 1989: 181-196.

Cavalli-Sforza, L. Luca (1997), ‘Genetic and Cultural Diversity in Europe’, Journal of Anthropological Research 53: 383-404.

Civil, Miguel (1962), ‘Un nouveau synchronisme Mari-IIIe dynastie d’Ur’, Revue d’Assyriologie 56, 213-214.

____ (1984), ‘Bilingualism in Logographically Written Languages: Sumerian in Ebla’, in L. Cagni, ed., Il bilinguismo a Ebla. Atti del convegno internazionale (Napoli, 19-22 aprile 1982). Naples: Istituto universitario orientale, Dipartimento di studi asiatici, 75-97.

Civil, Miguel and Gonzalo Rubio (1999), ‘An Ebla Incantation Against Insomnia and the Semiticization of Sumerian: Notes on ARET 5 8b and 9’, Orientalia NS, 254-266.

Cook, Eung-Do (1995), ‘Is There Convergence in Language Death? Evidence from Chipewan and Stoney’, Journal of Linguistic Anthropology 5, 217-231.

Cooper, Jerrold S. (1973), ‘Sumerian and Akkadian in Sumer and Akkad’, Orientalia NS 42, 239-246.

____(1991), ‘Posing the Sumerian Question: Race and Scholarship in the Early History of Assyriology,’ Aula Orientalis 9: 47-66.

Dalton-Puffer, Christiane (1995), ‘Middle English Is a Creole and its Opposite: On the Value of Plausible Speculation’, in Jacek Fisiak, ed., Linguistic Change under Contact Conditions. Berlin: Mouton de Gruyter, 35-50.

Daniel, Glyn Edmund (1962), The Idea of Prehistory. London: C.A. Watts.

Diakonoff, Igor M. (1967), Iazyki drevnei Perednei Azii. Moscow: Nauka.

____(1984), The Pre-history of the Armenian People. Delmar: Caravan Books.

____(1986), Hurro-Uratean as an Eastern Caucasian Language. Munich: Kitzinger.

Diebold, A. Richard, Jr. (1994), ‘Paleontology, Linguistic’, in R. E. Asher, et al., eds., The Encyclopedia of Language and Linguistics. Oxford: Pergamon, 2906-2913.

Dixon, Robert M. W. (1997), The Rise and Fall of Languages. Cambridge: Cambridge University Press.

Dorian, Nancy (1981), Language Death. The Life Cycle of a Scottish Gaelic Dialect. Philadelphia: The University of Pennsylvania Press.

____(1989), Investigating Obsolescence: Studies in Language Contraction and Death. Cambridge: Cambridge University Press.

Dryer, Matthew S. (1992), ‘The Greenbergian Word Order Correlations’, Language 68, 81-138.

Edzard, Dietz Otto (1960), ‘Sumerer und Semiten in der frh hen Geschichte Mesopotamiens’, Genava 8, 271-258.

____(1968), Sumerische Rechtsurkunden des III. Jahrtausends aus der Zeit vor der III. Dynastie von Ur. Munich: Verlag der Beyrischen Akademie der Wissenschaften.

____(1984), Hymnen, Beschwörungen und verwadtes (aus dem Archiv L. 2769) (Archivi reali di Ebla, 5). Rome: Missione archaeologica Italiana in Siria.

Englund, Robert K. (1998), ‘Texts from the Late Uruk Period’, in J. Bauer et al.,eds., Mesopotamien. Späturuk-Zeit und Frühdynastische Zeit. Göttingen: Vandenhoeck & Ruprecht, 15-233.

Foster, Benjamin (1982), ‘Ethnicity and Onomastics in Sargonic Mesopotamia’, Orientalia NS 51, 297-354.

Gelb, I. J. (1960) ‘Sumerians and Akkadians in their Ethno-Linguistic Relationship’, Genava 8, 258-271.

Givon, Talmy (1975), ‘Topic, Pronoun and Grammatical Agreement’, in Charles N. Lee, ed., Subject and Topic. New York: Academic Press, 149-188.

____(1976), ‘On the SOV Origin of the Suffixal Agreement Conjugation in Indo-European and Semitic’, in Alphonse Juilland et al, eds., Linguistic Studies Offered to Joseph Greenberg: On the Occasion of his Sixtieth Birthday. Saratoga: Anma Libri, 481-503.

Grenoble, Lenore A. and Lindsay J. Whaley, eds., (1998), Endangered Languages. Current Issues and Future Prospects. Cambridge: Cambridge University Press.

Haarmann, Harald (1994), ‘Contact Linguistic, Archaeology and Ethnogenesis: An Interdisciplinary Approach to the Indo-European Homeland Problem’, Journal of Indo-European Studies 22, 269-288.

Haayer, G. (1986), ‘Languages in Contact: The Case of Akkadian and Sumerian’, in H. L. J. Vanstiphout, et al, eds, Scripta Signa Vocis. Studies about Scripts, Scriptures, Scribes and Languages presented to J. H. Hospers. Groningen: Egbert Forsten, 77-84.

Hale, Ken (1998), ‘On Endangered Languages and the Importance of Linguistic Diversity’, in Grenoble and Whaley 1998: 192-216.

Heimpel, W. (1974/7), ‘Sumerische und akkadische Personennamen in Sumer und Akkad’, Archiv für Orientforschung 25, 171-174.

Hill, Jane H. (1989), ‘The Social Function of Relativization in Obsolescent and Non-obsolescent Languages’, in Dorian 1989: 149-166.

Hoenigswald, Henry M. (1989), ‘Language Obsolescence and Language History: Matters of Linearity, Leveling, Loss and the Like’, in Dorian 1989: 347-354.

Hoyrup, Jens (1992), ‘Sumerian: The Descendant of a Proto-Historical Creole?’, AION, Annali del Seminario di studi del mondo classico, Sezione linguistica 14, 21-72.

Huffines, Marion Lois (1989), ‘Case Usage among the Pennsylvania German Sectarians and Nonsectarians’, in Dorian 1989: 211-226.

Jones, Tom, ed. (1969), The Sumerian Problem. New York: Wiley.

Kaufman, Steven A. (1974), The Akkadian Influences on Aramaic. (Assyriological Studies, 19) Chicago: University of Chicago Press.

Lieberman, Stephen J. (1977), The Sumerian Loanwords in Old-Babylonian Akkadian. Volume One: Prolegomena and Evidence. Missoula: Scholars Press.

____(1986), ‘Word Order in Afro-Asiatic Languages’, in Proceedings of the Ninth World Congress of Jewish Studies D/1. Jerusalem: World Union of Jewish Studies, 1-8.

Mahmoud, N. Ahmed (1989), "The Ur III tablets from Ishan Mizyad’, Acta Sumerologica 11, 330-352.

McAlpin, David W. (1981), Proto-Elamo-Dravidian: The Evidence and Its Implications (Transactions of the American Philosophical Society, 71). Philadelphia: The American Philosophical Society.

Maul, Stefan M. (1997), ‘Kh chensumerisch oder hohe Kunst der Exegese? g berlegungen zur Bewertung akkadischer Interlinearh bersetzungen von Emesal-Texten’, in Beate Pongratz-Leisten et al, eds., Ana šadi labnani lu allik. Beiträge zu altorientalischen und mittelmeerischen Kulturen. Festschrift für Wolfgang Röllig (Alter Orient und Altes Testament, 247). Neukirchen: Neukirschener Verlag, 253-267.

Michalowski, Piotr (1987), ‘Language, Literature and Writing at Ebla’, in L. Cagni, ed., Ebla: 1975-1985. Atti del convegno internazionale. Naples: Istituto universitario orientale, Dipartimento di studi asiatici, 165-175.

____(1991), ‘Charisma and Control: On Continuity and Change in Early Mesopotamian Bureaucratic Systems’, in R. D. Biggs and M. Gibson, eds., The Organization of Power: Aspects of Bureaucracy in the Ancient Near East. Chicago: University of Chicago Press, 45-58.

____(1993), ‘Tokenism’, American Anthropologist 95, 996-999.

Van den Mieroop, Mark (1986), ‘Tã ram-ili, an Ur III Merchant." Journal of Cuneiform Studies 38, 1-80.

Militarev, Alexander Yu. (1995), (Sumerian and Afroasians). Vestnik Drevniej Istori 113-126.

___(1996), ‘Home for Afrasian – African or Asian? Areal Linguistic Arguments’, in Catherine Griefenow-Mewis et al eds., Cushitic and Omotic Languages: Proceedings of the Third International Symposium, Berlin, March 17-19, 1994. Köln: Rüdiger Köppe Verlag, 13-32.

Mithun, Marianne (1998), ‘The Significance of Diversity in Language Endangerment and Preservation’, in Grenoble and Whaley 1998: 163-191.

Mh ller, H.-P. (1995), ‘Ergative Constructions in Early Semitic Languages’, Journal of Near Eastern Studies 54, 261-271.

Nichols, Johanna (1992), Linguistic Diversity in Space and Time. Chicago: Chicago University Press.

____(1994), ‘Ergativity and Linguistic Geography’, Australian Journal of Linguistics 13, 39-89.

____(1995), ‘Diachronically Stable Structural Features’, in Henning Andersen, ed., Historical Linguistics 1993: Selected Papers from the 11^th International Conference on Historical Linguistics, Los Angeles, 16-20 August. Amsterdam: John Benjamins, 337-355.

Nocentini, Alberto (1993), ‘Power and Limits of the Genetic Classification of Languages’, The Mankind Quarterly 33: 265-281.

Rabin, Chaim (1985), ‘Massorah and "Ad Litteras"’, Hebrew Studies 26, 81-91.

Reiner, Erica (1966) A Linguistic Analysis of Akkadian (Janua Linguarum, Series Practica, 21). The Hague: Mouton.

Rottet, Kevin J. (1998), ‘Clause Subordination Structure in Language Decline’, Journal of French Language Studies 8, 63-95.

Rubio, Gonzalo (1999), ‘On the Alleged "Pre-Sumerian Substratum’, Journal of Cuneiform Studies 51 (in press).

Sallaberger, Walther (1999), ‘Ur III-Zeit’, in Walther Sallaberger and Aage Westenholz, Mesopotamien. Akkade-Zeit und Ur III-Zeit. Göttingen: Vandenhoeck & Ruprecht, 121-392.

Seebold, Elmar (1996), ‘Wann hat eine Sprache begonnen?’, in Robin Sackmann, ed., Theoretical Linguistics and Grammatical Description: Papers in Honor of Hans-Heinrich Lieb on the Occasion of his 60^th Birthday. Amsterdam: John Benjamins, 287-296.

Smedley, Audrey (1999), ‘"Race" and the Construction of Human Identity’, American Anthropologist 100: 690-702.

Stein, Gil (1999), Rethinking World-Systems : Diasporas, Colonies, and Interaction in Uruk Mesopotamia. Tucson: University of Arizona Press.

Steinkeller, Piotr (1989), Sale Documents of the Ur III Period (Freiburger altorientalische Studien, 17). Stuttgart: Franz Steiner.
______(1993), ‘Early Political Development in Mesopotamia and the Origins of the Sargonic Empire’, in M. Liverani, ed., Akkad: The First World Empire. Padua: Sargon, 107-129.
______(1998), ‘’, in G. Buccellati and M. Kelly-Buccellati, eds., Urkesh and the Hurrians: Studies in Honor of Lloyd Cotsen. Malibu: Undena Publications.

Straus,-Lawrence et al (1998), ‘The Peoplings of Europe: A J.A.R. Debate’, Journal of Anthropological Research 54: 399-420.

Tanret, M. (1996), ‘Noms sumériens ou noms accadiens?’, in Ö. Tunca and D. Deheselle, eds., Tablettes et images aux pays de Sumer et d’Akkad. Mélanges offers à Monsieur H. Limet. Liège: Association pour la Promotion de l’Histoire et de l‘Archeologie Orientales, 191-206.

Thomason, Sarah G. and Alaa Elgibali (1986), ‘Before the Lingua Franca: Pidginized Arabic in the Eleventh Century A.D.’, Lingua 68, 317-349.

Thomason, Sarah G. and Terrence Kaufman (1988), Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press.

Tuite, Kevin (1999) ‘The Myth of the Caucasian Sprachbund: The Case of Ergativity’, Lingua 108, 1-26.

Westenholz, Aage (1975), Early Cuneiform Texts in Jena. Pre-Sargonic and Sargonic Documents from Nippur and Fara in the Hiprecht-Sammlung … Copenhagen: Munksgaard.
____(1999), ‘The Old Akkadian Period: History and Culture’, in Walther Sallaberger and Aage Westenholz, Mesopotamien. Akkade-Zeit und Ur III-Zeit. G` ttingen: Vandenhoeck & Ruprecht, 17-117.

Woodbury, Anthony C. (1998), ‘Documenting Rhetorical, Aesthetic, and Expressive Loss in Language Shift.’ in Grenoble and Whaley 1998: 234-258.

Winford, Donald (1997), ‘Creole Formation in the Context of Contact Linguistics’, Journal of Pidgin and Creole Languages 12, 131-151.

Wright, Roger (1982), Late Latin and Early Romance in Spain and Carolingian France. Liverpool: Cairns.

____(1991), ‘The Conceptual Distinction between Latin and Romance: Invention of Evolution?’, in Roger Wright, ed., Latin and the Romance Languages in the Early Middle Ages. London: Routledge, 103-113.

____(1993), ‘Complex Monolingualism in Early Romance’, in William Ashby, et all, eds., Linguistic Perspectives on the Romance Languages: Selected Papers from the 21^st Linguistic Symposium on Romance Languages, Santa Barbara, California, 21-24 February 1991. Amsterdam: John Benjamins, 377-388.

1.  Even though Englund (1998: 73ff.) argues that there is no trace of Sumerian in the Uruk texts, I still stand by my earlier opinion (Michalowski 1993), holding with those who identify syllabic elements in the script that could only point towards Sumerian.

2.  Another indication of the complexity of language contacts in Western Asia is encountered in the case of an isolated early borrowing into Ugaritic and Aramaic of a word that was originally Sumerian.  The word for "palace" in Sumerian was written ?-gal, and has a transparent internal etymology ("big house") and therefore one can be certain that the word is in fact Sumerian. How it entered Ugaritic and Hebrew as hkl and Aramaic as h(y)kl is a matter of some speculation (Kaufman 1974: 27).

3.  It is interesting to note, in this context, the comments on the Semiticization of Sumerian at Ebla by Civil and Rubio (1999:266): “Moreover Ebla would represent the written expression (or the peripheral and exclusively scribal expression) of the process which may have been taking place during the Early Dynastic period: the switch from bilingualism in the Late Uruk and  Jemdet Nasr period to diglossia in Ur III, when Sumerian was not a living language any more.”