Editing the Sumerians, How and Why?

Abstract

The science of pedagogy in Babylonia in the early second millennium BCE was based on a two-part curriculum: students began their study of Sumerian by copying and memorizing lists of signs and words, and then progressed to copying connected texts. All surviving manuscripts (in the form of cuneiform tablets) of Sumerian literary compositions from that time are school copies produced by students following the second part of the curriculum. This contribution examines how the editing of these compositions and the representation of individual text witnesses and their variants have evolved in the eight decades since the first scientific edition of a Sumerian literary composition in 1937, and defines what a reconstructed and edited Sumerian composition actually represents.

The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007–2013)/ERC Grant Agreement No. 269804.

Writing, the graphic representation of human language, was devised first in the ancient Near East, in the centuries just prior to 3000 BCE, by the Sumerians in the far south of Mesopotamia, in what is now Iraq, and by the Egyptians in the northeastern corner of the African continent.¹ The Sumerian language² was at home in southern Babylonia, Babylonia itself being the region between the neck of the Tigris and Euphrates rivers at about the latitude of Baghdad, and the Persian Gulf. The manuscripts—clay tablets inscribed with cuneiform signs—that I will be discussing are predominantly from Nippur, the early religious capital of Babylonia, with a smaller yet significant number from Ur (of Biblical fame), and much smaller groups of manuscripts from other sites or of unknown provenance.³ The reason why a discussion of literary manuscripts belongs in a volume on editing ancient scientific texts will be put forward after a brief introduction.

Sumerian literature⁴ appears, with some minor exceptions, at three moments in Mesopotamian history: the earliest literature is found on several hundred manuscripts from around 2600–2500 BC, very difficult to understand, and frequently written in a cryptographic writing system that has been only partially deciphered. The next large group of Sumerian literary manuscripts, the group that will concern us here, can be dated to a one-hundred-year period in the early second millennium, from approximately 1825–1725 BCE. And then, from the first millennium, there are a large number of Sumerian incantations and liturgical texts, many descended from compositions known from the early second millennium, and most with interlinear Akkadian translations.⁵ The earliest written Sumerian—though not literary—appears before 3000 BCE, and the latest Sumerian manuscripts, a few with transliterations into Greek, are dated to near the beginning of our own era.⁶

Sumerian, like Egyptian or Chinese, has an extraordinary record of continuous use spanning over three millennia. But unlike Chinese and Egyptian, after a bit more than the first 1000 years of documentation, Sumerian was a dead language, a learned language, certainly what Ong (1982: 115) has called a male language, taught by men to boys. The teaching was done in schools, which, in the period that interests us here (the early second millennium BCE), were located in private houses and probably never had more than a dozen students, most often less.⁷ All of what is considered Sumerian literature in the narrower sense (i.e. excluding magical and liturgical texts) comes from these schools, including mythological texts such as Inanaʼs Descent to the Netherworld, the Sumerian tales that were a major source for the Akkadian Gilgamesh Epic, hymns to gods and praise songs to kings. There were also minor genres, including scribal dialogues; debates between, say, Ewe and Grain, or Silver and Copper; and proverbs.⁸

Students began by learning how to shape clay tablets and inscribe wedges, and then learned lists of signs with their values, and a compendium of personal names. There followed thematic word lists, advanced sign lists, metrology and mathematics. Only then did they begin to produce continuous Sumerian texts, starting with model contracts and proverbs, and then a group of four relatively simple literary compositions. All of these were learned by copying a teacherʼs model, written on a tablet’s left-hand column, and copied by the student on the right-hand column. The studentʼs work was erased (smoothed over) and repeated until he got it right. Excerpts of just a few lines of these texts were sometimes written down on small round tablets. Only after mastering all of this material did the student begin to study Sumerian literature in earnest.⁹

Sumerian literary manuscripts, found—when find spots are known—in the same contexts as the practice tablets just described, are primarily of two kinds. They are either large, multi-column tablets containing an entire composition or half of a larger composition, or they are single-column excerpt tablets with from about thirty to one hundred or more lines excerpted from a composition.¹⁰ We imagine that students first learned a composition by writing a series of excerpts, and then, to demonstrate their mastery, wrote out the complete composition on multi-column tablets. For a given composition, the ratio of multi-column tablet manuscripts to single column excerpts is between 1:4 and 1:6.¹¹ Nothing from this period resembles a library copy or a teacherʼs master copy; all our manuscripts are the detritus of scribal education. These four thousand or so manuscripts were found on the floors of abandoned school rooms or used as building material in rebuilt school houses. Otherwise, they would have been recycled, soaked in a tub of water until soft and formed into new tablets for student exercises. These manuscripts, then, are the primary evidence for Babylonian pedagogy in the early second millennium BCE, and if pedagogy be a science, then Sumerian literary manuscripts are primary textual evidence for that science.

There are altogether about 6000 manuscripts of Sumerian literature in the larger sense from the first half of the second millennium, and about 4000 represent literary compositions in the narrower sense that I am speaking of today. Over 80% of these manuscripts were excavated at Nippur, and another 10% were found at Ur. About 90% of the manuscripts found at Nippur and Ur fall into the group that has been called curricular, representing about 106 separate compositions,¹² each of which can have anywhere from a few to over 100 manuscripts. Nearly half of these have over twenty manuscripts each; a given line of a composition can have fifteen or more witnesses. Because the manuscriptsʼ origins are so restricted geographically and chronologically, the texts of the Sumerian literary compositions are remarkably stable. Most variants are either orthographic, or errors typical of manuscript copying. Yet significant variants do occur, and when there are only a few witnesses for a line, even a choice between insignificant variants is not always simple. Clearly, the prime responsibility of an editor is to fully represent the contents of each manuscript, so that the evidence for the editorʼs choices of readings is available to the reader.

Before presenting a short history of the editing of Sumerian texts, with emphasis on how the representation of multiple manuscripts with their variants has developed, let me clarify what an editorʼs goal must be, that is, what the resulting text edition represents. The goal is certainly not to reconstruct the composition as it emerged through an authorʼs stylus. We know nothing of the authorship of Sumerian literary compositions, nor, with some exceptions, do we know when or where they were composed. We think that the majority was composed between 100 and 350 years before our manuscripts were copied. Fragments dating to around 2000 BCE of a few of the compositions have turned up at Nippur, and a couple of compositions are found on other fragments dating to c.a. 2500. But most of the tiny corpus of literary manuscripts from 2000, and the vast majority of the larger corpus dating to 2500, consist of compositions that do not appear in the Sumerian literary corpus studied and transmitted in the Babylonian schools of the eighteenth century BCE. We can imagine the context for certain compositions and categories of compositions: royal panegyrics may have been part of court ceremonial under the dynasties of Ur and Isin (twenty-first to twentieth centuries BCE)¹³; the so-called ‘epics’ of the legendary kings of Uruk—Enmerkar, Lugalbanda and Gilgamesh—may have been written under the patronage of the kings of the Third Dynasty of Ur, who felt a special kinship to those legendary rulers¹⁴; debates between Silver and Copper and the like may have been court entertainment; hymns to deities could have been used in the cult. But for many compositions, such as the great mythological tales of the gods Inana, Enki, Enlil and Ninurta, scholarly guesses of original context have been few and far between.

What Assyriologists of today aim to produce in a scholarly edition of a Sumerian literary composition is a text that a teacher at Nippur or Ur would recognize as correct, that is, to approximate the as yet undiscovered ‘master copy’ that the teacher would have taught from, or, if he taught from memory, a text he would recognize as an accurate representation of the composition.¹⁵ The editing process begins with careful study of the individual manuscripts of the composition upon which the reconstruction of the entire composition will be based. Unlike manuscripts in Biblical or Classical corpora, our manuscripts are not the work of scribes intending to produce copies for reference or posterity (Delnero 2012b: 10). Nor are any of the copied Sumerian compositions sacred scripture where absolute fidelity to the text would be essential. Because our Sumerian manuscripts are school exercises coming mainly from one place over a short span of time, we almost never can reconstruct manuscript families; manuscripts donʼt sort into groups that regularly incorporate the same set of variants, even when manuscripts from Ur are thrown into the mix. That is, a given variant can occur in some manuscripts from both Nippur and Ur, but not in others, and usually, there are no Ur manuscripts that regularly pair with individual Nippur manuscripts.

Nippur was first excavated by the University of Pennsylvania at the end of the nineteenth century CE; it was the earliest American archeological expedition to Mesopotamia.¹⁶ Excavations there were resumed in 1948 under the auspices of the American Schools of Oriental Research (now the American Society of Overseas Research), the University of Pennsylvania and the Oriental Institute of the University of Chicago (now the Institute for the Study of Ancient Cultures). These continued until 1990, and the later campaigns were undertaken solely by the Oriental Institute. Of approximately 5000 Sumerian literary manuscripts found at Nippur, 4000 came from the late nineteenth-century excavations, and were divided between the University of Pennsylvania’s University Museum (now the Penn Museum) and the Archeological Museum in Istanbul, with several hundred ending up in Jena. But although a considerable corpus of Sumerian manuscripts was available at the beginning of the twentieth century, the first scientific Sumerian text edition didnʼt appear until 1937. The reasons for this, having to do with museum accession policy and procedures, the fragmentary state of most of the manuscripts, as well as the difficulties of the Sumerian language itself, will not detain us here.

The first scientific editions of Sumerian literary texts were produced by S.N. Kramer, but the story of Sumerian literature has its own Rosalind Franklin or Alice Kober figure (albeit male).¹⁷ In 1923, the Assyriologist Edward Chiera went to Istanbul to copy Sumerian tablets from Nippur, and then spent 3 years copying and studying Nippur tablets in Philadelphia. In 1927, he was called to the University of Chicago’s Oriental Institute to work on the Assyrian Dictionary project,¹⁸ and Chicago agreed to publish the several hundred copies he made in Philadelphia. When Chiera died prematurely in 1933, he had not yet prepared the introductions to those volumes, and the task was given to Kramer, then a post-doc at the Oriental Institute. Kramer, in his own words, ‘got hooked’ on Sumerian when preparing those introductions.¹⁹ Working through Chieraʼs copies, and using the list of unpublished duplicates in Philadelphia and Istanbul compiled by Chiera, Kramer was able to reconstruct several entire Sumerian compositions. The first to be published was Inanaʼs Descent to the Netherworld, which appeared in 1937.

Despite Chieraʼs work, Kramer had only eight manuscripts for his initial edition, just 20% of the forty or so manuscripts known today (many of these would be identified by Kramer himself in the following four decades). He was able to list his manuscripts over a short footnote (see Fig. 10.1). The alphabetical manuscript sigla are assigned according to the first preserved line of the manuscript, and where several manuscripts begin with Line 1, the longest manuscripts are given priority. Variants are given in footnotes, but, as can be seen in Note 1 on his Page 99, although you are told that ‘B omits a’, you are not told what B has (that is, how much of the line is actually preserved in Manuscript B). This is a shortcoming in the editing of Sumerian manuscripts that would persist until the mid-1970s. But Note 3 reveals a strength in Kramerʼs method (though ‘method’ is not what heʼd have called it): Even though Manuscript A is a multi-column tablet that originally contained the entire composition, he is not following that manuscript and giving variants from other manuscripts in his notes. Rather, he is reconstructing an eclectic best text, preferring other manuscripts to A when he believes they are more correct. When scholars have edited Sumerian texts following the readings of a single best preserved manuscript, they have sometimes produced texts that enshrine erroneous, defective or atypical readings.²⁰ An unfortunate feature of Kramer’s edition is the interlinear English translation, a practice Kramer would soon abandon. Future editions of Sumerian compositions, by Kramer and other others, would put entire translations following the Sumerian text, or, better, on facing pages. Without a connected translation, a Sumerian composition appears to be a series of puzzles requiring line-by-line decipherment; a connected translation was essential for the composition to be received as a work of literature.

Fig. 10.1

Kramer 1937: 98–99

Kramer was a pioneer who continued throughout his long career to identify and publish Sumerian manuscripts, as well as to make the Sumerians known to the larger public beyond the academy. But Kramer was not the most careful philologist. His editions could be sloppy, and his lexicography and interpretation were superficial, as he was the first to admit. The standards for fastidious Sumerian text editing were set by Adam Falkenstein at Heidelberg in the 1950s and 1960s. But, as Fig. 10.2 shows, it remained impossible in Falkenstein’s editions to ascertain the state of individual manuscripts, since only the variant signs were indicated in the apparatus. The Falkenstein tradition reached its acme at the end of the 1960s in the work of his student, C. Wilcke, whose edition of the Sumerian epic now known as Lugalbanda II or Lugalbanda and the Anzud Bird (Black et al. 2004) or The Return of Lugalbanda (Vanstiphout 2003a) had the added convenience of providing the translation on facing pages (Fig. 10.3). Note, too, that in his Note 1 to line 53, he tells the reader that three manuscripts—AA, H and J—have the sign in question, whereas EE omits it. From this information, one can deduce that the other witnesses to the line—A, FFF and I—must be broken at the lineʼs beginning. The composition Wilcke edits here is one of the few Sumerian literary (in the narrower sense) compositions transmitted beyond the middle of the second millennium BCE, surviving in a couple of bilingual manuscripts from the seventh century BCE, that is, 1000 years after the bulk of the manuscripts was copied. Looking at Lines 58 and following, you can see that he enters the much later manuscripts B₁ and B₂ with their Akkadian translation separately, which is the right editorial choice to make. Whereas one can justify using contemporary manuscripts from different sites—usually Nippur and Ur—when editing a composition, manuscripts written after the passage of a millennium must be kept apart and utilized not to reconstruct the literary composition as it was known in 1800 BCE, but rather to understand how Sumerian language and literature were transmitted and received in 800 BCE.²¹

Fig. 10.2

Falkenstein 1959: 14

Fig. 10.3

Wilcke 1969: 96–97

Despite Wilckeʼs very thorough apparatus, which enables the user to imagine the state of the manuscripts for a given variant, Sumerian compositions were still presented as a succession of lines of reconstructed text, with a forest of footnotes constituting the critical apparatus. The characteristics of individual manuscripts could only be ascertained by assembling the various copies and photographs of tablets used by the editor. This changed in the mid-1970s when another Falkenstein student, D. Edzard, challenged the accepted method (Edzard 1974). He proposed that the entire text of each witness for each line be presented, one after the other (Fig. 10.4), in an arrangement he likened, mistakenly I think, to a musical score, and the name ‘score’, or ‘Partitur’ in German, has stuck. His solution threatened to bring on what Pollock (2009: 952) has called ‘Text Panic’, because, as is evident from his example, there is no editorial decision, no reconstructed ‘best text’, data but no interpretation. This is quite user unfriendly when there are as many individual manuscripts as in his example here, but not quite as bad in the case of the first complete composition to be so edited (Fig. 10.5), a Sumerian hymn with only a few extant manuscripts that Edzard published together with Wilcke in 1976.

Fig. 10.4

From Edzard 1974: 106

Fig. 10.5

Edzard and Wilcke 1976: 158

My own effort in 1983 to devise a method of presenting and editing Sumerian manuscripts strove to preserve the advantages of Edzardʼs ‘score’ method but to streamline the representation of individual manuscripts in order to enable the differences and similarities of the manuscripts to be grasped more quickly (Fig. 10.6). The manuscripts were arrayed beneath a reconstructed eclectic ‘best text’ for each line: when a manuscript was identical to the reconstructed text, a dash was placed beneath each preserved sign; an omitted sign was indicated by zero and only variants were written out.²² Note here the two manuscripts at bottom (R₃, S₃), spaced apart from the others. This composition is one of the few that has manuscript witnesses from ca. 2000 BCE—three centuries earlier than the others—and it is these that I have kept apart. For the userʼs convenience, as well to emphasize that what we have here is a literary text and not just a collection of Sumerian words and sentences, I separately laid out the eclectic ‘best text’ facing an English translation (Fig. 10.7), with only significant variants indicated in footnotes.

Fig. 10.6

Cooper 1983: 150

Fig. 10.7

Cooper 1983: 50–51

Thus, by the mid-1980s, Sumerian text editions looked quite a bit different from those of Kramer or Falkenstein, but one thing had not changed: manuscript sigla were alphabetic, with double letters or letters with numerical subscripts used when the number of manuscripts exceeded twenty-six. A 1996 publication of S. Tinney changed that (Fig. 10.8): Manuscripts are identified by place of origin: N for Nippur manuscripts, U for Ur manuscripts and X for unknown. Within geographical groups, manuscripts are distinguished only by subscript. But, as can be seen, most manuscripts are from Nippur (N), and it is rather difficult, especially for older colleagues, to distinguish the subscript numbers. Even if you can read the numbers easily, the manuscripts are harder to get to know than when they each have their own upper-case siglum. Nevertheless, Tinney made an important point, and other text editions, beginning in the mid-1990s (see below), have managed to present manuscript provenance while preserving the distinctiveness of the upper-case sigla.

Fig. 10.8

Tinney 1996: 225

Just before Tinneyʼs publication, M. Civil (1994) published an edition that altered somewhat the representation of manuscripts, adhering to a method he told me he had employed since the late 1950s in the hundreds of unpublished editions he has made of Sumerian literary compositions (Fig. 10.9). Civil considered each line of text with its witnesses a ‘textual matrix’, and the representation of the composition as a whole ‘textual matrices’, a far better term than the misleading ‘score’ (though I fear that the latterʼs conciseness ensures that it will continue to be the preferred term). A plus-mark is used when a sign is the same as that in the eclectic text, a minus when it is omitted, a dot indicates a partially preserved sign and a zero is used where the tablet is broken. Civil pointed out to me an important but as yet unnoticed difference in the way he treated variants: when variants to a sign in the eclectic text are present, the sign in every manuscript is spelled out, even when it agrees with the eclectic text. Should an editor at some time before publication decide that another of the variants is to be preferred to the one he already chose for his eclectic text, the editor need only change the eclectic text and not the individual manuscript lines.

Fig. 10.9

Civil 1994: 208

Meanwhile, our Germanophone colleagues were uneasy about the dots, dashes, plusses and zeros. For reasons that were at first difficult to comprehend, they wanted to see every sign of every manuscript spelled out.²³ Seemingly, this is a conflict between the maximal presentation of data and an economical one,²⁴ yet I remain doubtful that anything is gained by the former, whereas the advantages of the latter are clear. K. Volk (1995) (Fig. 10.10) uses upper-case sigla, adding provenience in lower case (n = Nippur, u = Ur), but his maximalist presentation makes it difficult to tell at a glance that in Line 129, the only variant is in manuscript Fu, which reads /ni/ where other manuscripts read /na/. Volk conveniently adds the column and line number for each line in each manuscript, also done by Zgoll (1997) (Fig. 10.11), who, however, introduces a very complicated system for her sigla: Each siglum is composed of a two-letter geographic abbreviation—Ni for Nippur, Ur for Ur, La for Larsa and Un for Unbekannt—and a capital letter, whereby each geographical category has its own sequence of letters. So, for example, there is a NiB as well as an UrB. Neither Volk nor Zgoll precede the individual readings for each line with their eclectic ‘best’ text, given elsewhere facing a connected translation, so a reader going through Zgoll’s ‘score’, for example (Fig. 10.11), will not immediately see that in Line 151, the reading of NiA is not the reading the editor accepts as correct. Mittermayer (2009) (Fig. 10.12) follows Volk’s construction of sigla (upper-case plus lower-case indication of provenience) and the addition of manuscript column and line numbers of both Volk and Zgoll, but precedes the manuscript readings with each line’s eclectic text. To my mind, this is the most successful of the maximalist formats.

Fig. 10.10

Volk 1995: 105

Fig. 10.11

Zgoll 1997: 292

Fig. 10.12

Mittermayer 2009: 182

More recently, a pioneering monograph on editing Sumerian texts has been published by P. Delnero (2012b). His primary concern is the motivation and evaluation of textual variants, based on a painstaking analysis of 740 manuscripts of ten frequently copied Sumerian compositions, thought to be among the first advanced literary compositions copied by students at early second millennium BC Nippur. Although Delnero is American, he studied for many years in Germany and hence presents the entire text of each manuscript, highlighting variants from his eclectic text in boldface (Fig. 10.13). His sigla are also maximalist in the information they provide and revert to Tinney’s construction of sigla using upper-case indicators of provenience: Nippur manuscripts are labeled N (Ur stands for a manuscript from Ur and X for unknown provenience), followed by a subscript identifying (only for the Nippur manuscripts) whether the manuscript is a multi-column tablet (Roman I), single-column (Roman III) or prism (P).²⁵ Each manuscript within those three groups is assigned a subscript Arabic numeral which appears after the format identification. While the information provided is useful for some purposes, the subscripts are not only difficult to read, but note how similar the subscript of the third manuscript in this example—N-Roman I-numeral 4—is to the that of the last, X-numeral 14. Also, as stated earlier, it is much harder to become acquainted with individual manuscripts when they are distinguished only by subscripts. When working through an edition, one begins to recognize the idiosyncrasies of Manuscript M and the frequent errors of Manuscript R; when both are instead identified by N plus a series of subscripts, their distinctive characteristics are more difficult to spot and keep in mind. But, as Delnero explains (2012b: 183), this system was devised as a research tool to enable him to quickly identify the provenience of each manuscript and its format for the purposes his study of variants across his corpus, and he is not advocating it as a standard for text editions.

Fig. 10.13

From Delnero 2012b: 182–183

The real importance of Delneroʼs work is that he is able to establish that when there are four or more manuscript witnesses, the variants of the majority are nearly always correct, and thus to be favored when reconstructing a ‘best’ text (2012: 185). There appears to be some circularity to his argument, but if we remember that what we are reconstructing is simply a text that would be recognized by a Nippur schoolmaster as correct, and that the manuscripts of individual compositions are 90–95% identical (Delnero 2012b: 10), then there is a compelling logic to his contention, all the more so because variation from the majority can almost always be explained as errors or as simple orthographic variation. Delnero, I believe, has defined what Pollock might call the proper ‘editorial strokes’ for studying the ‘historical folks’ learning Sumerian in eighteenth-century BCE Babylonia. Nearly all the textual variants can be considered, in Pollockʼs words, ‘textually transmitted disease’ (Pollock 2009: 952–953), keeping in mind that the symptoms of disease can teach us much about the patient.

What we have in the end is a reconstructed corpus of literature whose only certain context is the schoolroom. Not only is authorship and original context left to our educated guesses, but we have no idea if, in eighteenth-century BCE Babylonia, this literature had any other context at all outside the school, or any function other than the construction of an educated elite identity, an identity whose existence is detectable only in these same school texts (Cooper 2015). We can point to nothing resembling scriptoria with scribes laboring a lifetime to preserve ancient sacred traditions, no odd sectarians copying manuscripts in the Judean desert, no mandarins collating texts in a peony pavillion—just schoolboys learning a mostly irrelevant literature in a dead language. If ever Saidʼs statement (2004: 57) that ‘philology is just about the … least sexy … of any of the branches of learning’ were true, it would be true for the all-consuming, ceaselessly fascinating task of editing the Sumerian literary manuscripts produced by Babylonian students nearly 4000 years ago.