53e Rencontre Assyriologique Internationale, Moscow, July 23, 2007

Sumerian: A Uralic Language

Simo Parpola (Helsinki)

In the early days of Assyriology, Sumerian was commonly believed to belong to the Ural-Altaic language phylum. This view originated with three leading Assyriologists, Edward Hincks, Henry Rawlinson and Jules Oppert, and other big names in early Assyriology such as Friedrich Delitzsch supported it (Fig. 1). The Frenchman François Lenormant, who wrote on the subject in 1873-78, found Sumerian most closely related to Finno-Ugric, while also containing features otherwise attested only in Turkish and other Altaic languages.

The wind turned in the early 1880s, however, as two prominent Finno-Ugrists, August Ahlqvist and Otto Donner, reviewed Lenormant's work and concluded that Sumerian was definitely not a Ural-Altaic language (Fig. 2). This was widely considered a death-blow to the Sumerian-Ural-Altaic hypothesis, and since then Assyriologists have generally rejected it. Typically, when a Hungarian scholar in 1971 tried to reopen the discussion in the journal Current Anthropology, a few linguists welcomed the idea but the reaction of the two Assyriologists consulted was scornfully negative.

Attempts to connect Sumerian with other languages have not been successful, however, and after 157 years, Sumerian still remains linguistically isolated. This being so, there is every reason to take another look at the old Ural-Altaic -hypothesis, for it has never been properly investigated. In the 19th century, Sumerian grammar and lexicon were as yet too imperfectly known to be successfully compared with any languages, while all more recent comparisons suffer from the lack of Assyriological or linguistic expertise and are hence for the most part worthless. This does not mean, however, that they are all garbage: at least 194 of them seem perfectly acceptable both phonologically and semantically (Fig.3). That is a number large enough to deserve serious attention. Of course, it does not prove that Sumerian was related to Ural-Altaic languages, but it does indicate that the possibility exists and should be carefully re-examined in order to be either substantiated or definitively rejected.

To this end, I started in November 2004 a project called "The Linguistic Relationship between Sumerian and Ural-Altaic," on which I have been working full time since May 2006, with funding from the Academy of Finland. The aim of the project is to systematically scrutinize the entire vocabulary of Sumerian with the help of modern etymological dictionaries and studies, identify all the words and morphemes that can be reasonably associated with Uralic or Altaic etyma, ascertain the validity of the comparisons, convert the material into a database, and make it generally available on the Internet.

The database under construction will contain all the attested phonetic spellings and meanings of the compared Sumerian and Ural-Altaic lexical items, as well as, for control purposes, all Indo-European etymologies proposed for these items. The relevance of each comparison is assessed separately for form and meaning on a scale from 4 to 1 (Fig. 4). The highest score, 4+4, indicates perfect agreement in form and meaning; a low score correspondingly poor agreement and doubtful relevance. In deciding whether a comparison is relevant or not, the governing principle has been that all compared items must match reasonably well in both form and meaning, and any differences in form or meaning must conform with the phonological and semantic variation attested in the languages compared.

To date, I have systematically gone through about 75 per cent of the Sumerian vocabulary and identified over 1700 words and morphemes that can be reasonably associated with Uralic and/or Altaic etyma, allowing for regular sound changes and semantic shifts. Somewhat surprisingly, words with possible Altaic etymologies constitute only a small minority (about seven per cent) of the total, and it is unlikely that the picture will essentially change by the time the project has been finished. Although a close relationship of Sumerian with the Altaic family as a whole thus seems excluded, a genetic relationship with Turkish seems possible, as most of the matches are with Turkic languages, and they are basic words and grammatical morphemes also found in Uralic languages.

Practically all the compared items are thus Uralic, mostly Finno-Ugric. The majority of them are attested in at least one major branch of Uralic beside Finnic and thus certainly are very old, dating to at least 3000 BC. A large number of the words are known only from Finnic, but this does not prevent them from being ancient as well, since they have no etymology and are for the most part common words attested in all eight Finnic languages.

This collection of words runs the gamut of the Sumerian vocabulary (Fig. 5) and includes 478 common verbs of all possible types, such as verbs of being, bodily processes, sensory perception, emotion, making, communication, movement etc., adjectives, numerals, pronouns, adverbs, interjections, conjunctions, and 589 nouns including words for body parts, kinship terms, natural phenomena, animals, plants, weapons, tools and implements, and various technical terms reflecting the cultural level of the neo- and chalcolithic periods (in the fields of agriculture, food production, animal husbandry, weaving, metallurgy, building technology, etc.). I would like to emphasize that the majority of the words in question are basic words, and 75 per cent of them show a very good match in form and meaning. This does not mean that they are necessarily all correct, but they stand a very good chance of being so. About 20 per cent of the comparisons are more problematic and about 5 per cent of them are conjectural only. All clearly impossible comparisons will of course be excluded once the material has been thoroughly analyzed.

Over 1700 lexical matches with Uralic surely sounds like an awful lot, "too good be true," if compared with all the previous fruitless attempts to find a cognate for Sumerian. But it is not at all much for genetically related languages; on the contrary, it is what must be legitimately expected of languages that are related. Who marvels at the fact that members of the Indo-European language family, even ones widely separated in time and place, have a large number of words in common? The large number of common words is precisely the reason why these languages can so easily and securely be identified as members of the same family.

It may be asked why all these numerous lexical matches with Uralic have not been found earlier. The explanation is simple. It takes a good knowledge of the Uralic languages plus familiarity with the intricacies of Sumerian phonology and cuneiform writing system to recognize the connections between Sumerian and Uralic, and such a combination of special expertise is rare. Very few Assyriologists know any Uralic languages, and experts in Uralic studies do not know any Sumerian. Of course, beside the required special expertise one would also need the will to study the matter seriously, and such will has been entirely lacking in Assyriology for the past 120 years.

In order to get a better idea of the relationship between Sumerian and Uralic, let us now have a look at some of the comparisons to see what they are like and how they work in practice.

34 years ago, Miguel Civil in his article "From Enki's headache to phonology" showed that late Sumerian ugu, "top of the head," is the same word as earlier a-gù; and from the alternation of a-gù with the divine name dab-ú, he concluded that it probably originally contained a labiovelar stop in the middle (Fig. 6). Recently, Joan Westenholz and Marcel Sigrist have shown that beside "top of the head," ugu also means "brain." { Hungarian agy=brain} Both formally and semantically, the Sumerian word thus matches the Uralic word *ajkwo "brain, top of the head," which can be reconstructed as containing a labiovelar stop in the middle based on its reflexes in individual Uralic languages. Remarkably, Sumerian ugu4 "to give birth," a homophone of ugu, likewise has a close counterpart in Finnic aiko-, aivo-, "to intend; to give birth." The semantics of the Finnic word show that it derives from the word for "brain," and the alternation of /k/ and /v/ in the stem confirms the reconstruction of the labiovelar in the middle of the word.

Several other words discussed by Civil also display an alternation of /g/ and /b/, including gurux or buru4 "crow," and gur(u)21 "shield," also attested as kuru14, e-bu-ùr and íb-ba-ru (Fig. 7). These two words certainly were almost homophonous, since they could be written with the same logogram. The common Uralic word for "crow," *kwarüks, indeed contains the posited labiovelar stop and provides a perfect etymology for the Sumerian word. The original labiovelar is preserved in Selkup, but has been replaced by /v/ in other Uralic languages except Sayan Samoyed, where it is appears as /b/. Sumerian gur(u)21 "shield" can be compared with Finnic varus "protection," whose original form can be reconstructed as *kwaruks and thus provides a perfect etymology for the Sumerian word. {?Hungarian óv=to protect from harm, vár=a fort}

The regular replacement of the labiovelar by /g/, /k/ or /b/ in Sumerian and by /v/ in Uralic amounts to a phonological rule and helps establish further connections between Sumerian and Uralic words displaying a similar correlation, for example Sumerian gíd "to pull" and Uralic *vetä- "to pull," {Hungarian huz t>z} and Sumerian kur "mountain" and Uralic *vor "mountain." {also common as kur in many FU languages} The reconstruction of an original labiovelar in the latter case is strongly supported by Volgaic kurok, "mountain." The phonological correspondences between Sumerian and Uralic remain to be fully charted, but a great many of them certainly are perfectly regular. For example, in word initial position Sumerian /š/ regularly corresponds to Finnic /h/, while Sumerian /s/ regularly corresponds to Finnic /s/ (Fig. . {In Hungarian its often s, ch, sh }

The word a-gù just discussed was written syllabically with two cuneiform signs, A and KA, both of which have several phonetic values and meanings based on homophony and idea association (Fig. 9). All these phonetic values and meanings have close counterparts in Uralic, and the homophonic and semantic associations between the individual meanings work in Uralic, too; compare the homophony between a, aj "water" and aj, aja "father" in Sumerian, and jää, jäj and äj, äijä in Uralic. And this applies not only to the signs A and KA but, unbelievable as it may sound, practically the whole Sumerian syllabary. Consider, for example, the sign AN (Fig. 10), whose basic meaning, "heaven, highest god," was in Old Sumerian homophonous with the third person singular of the verb "to be," am6. The Uralic word for "heaven" and "highest god" was *joma, which likewise was virtually homophonous with the third person singular of the verb "to be," *oma. These two words would have become totally homophonous in Sumerian after the loss of the initial /j/. The loss of the initial /j/ also provided the homophony between Sumerian a "water" and aj "father" just mentioned.

Such a close and systematic parallelism in form and meaning is possible only in languages related to each other. Accordingly, the logical conclusion is that Sumerian is a Uralic language. This conclusion is backed up by the great number of common words and the regularity of the phonological correspondences between Sumerian and Uralic already discussed, as well as by many other considerations. Sumerian displays the basic typological features of Uralic; it has vowel harmony, no grammatical gender but an opposition between animate and inanimate, and its grammatical system is clearly Uralic, with similar pronouns, case markers, and personal endings of the verb. In addition, many Uralic derivational morphemes can be identified in Sumerian nouns and verbs. The non-Uralic features of Sumerian, such as the ergative construction and the prefix chains of the verb, can be explained as special developments of Sumerian in an entirely new linguistic environment after its separation from the other Uralic languages.

The Sumerians thus came to Mesopotamia from the north, where the Uralic language family is located (Fig. 11), and by studying the lexical evidence and the grammatical features which Sumerian shares with individual Uralic languages, it is possible to make additional inferences about their origins. The closest affinities of Sumerian within the Uralic family are with the Volgaic and Finnic languages, particularly the latter, with which it shares a number of significant phonological, morphological and lexical isoglosses. The latter include, among other things, a common word for "sea, ocean" (Sumerian ab or a-ab-ba, Finnic aava, aappa), and common words for cereals, sowing and harvesting, domestic animals, wheeled vehicles, and the harness of draught animals (Fig. 12). A number of these words also have counterparts in Indo-European, particularly Germanic languages. These data taken together suggest that the Sumerians originated in the Pontic-Caspian region between the mouth of the Volga and the Black Sea, north of the Caucasus Mountains, where they had been living a sedentary life in contact with Indo-European tribes. I would not exclude the possibility that their homeland is to be identified with the Majkop culture of the North Caucasus, which flourished between 3700 and 2900 BC and had trade contacts with the late Uruk culture (Fig. 13). Placing the Sumerian homeland in this area would help explain the non-Uralic features of Sumerian, for the Kartvelian languages spoken just south of it are ergative and have a system of verbal prefixes resembling the Sumerian one. The Sumerian words for wheel and the harness of draft animals that it shares with Uralic show that its separation from Uralic took place after the invention of wheeled vehicles, which were known in the Majkop culture since about 3500 BC.

About 3500 BC, the Indo-European Yamnaya culture that had emerged between the Danube and the Don began to expand dynamically to the east, reaching the Caucasian foreland by about 3300 BC. This expansion is likely to have triggered the Sumerian migration to Mesopotamia. It would have proceeded through the Caucasus and the Diyala Valley, and since wheeled transport was available, could easily have been completed before the end of the Late Uruk period (c. 3100 BC). The arrival of the Sumerians would thus coincide with the destruction of the Eanna temple precinct at the end of the Uruk IVa period.

The lexical parallels between Sumerian and Uralic thus open up not only completely new possibilities for the study of Sumerian, but also a chance to identify the original homeland of the Sumerians and date their arrival in Mesopotamia. In addition, they provide a medium through which it becomes possible to penetrate into the prehistory of the Finno-Ugric peoples with the help of very ancient linguistic data. Of course, it is clear that the relevant evidence must first pass the test of verification or falsification before any part of it can be generally accepted and exploited.

I am currently preparing an Internet version of the database in collaboration with the Department of General Linguistics of the University of Helsinki. This web version is planned to be interactive and will contain a search engine and a program to check the regularity of the sound changes involved in the comparisons. I heartily invite all skeptics to visit the site once it is ready and falsify as many of the comparisons as they can, and everybody else to look at the evidence, check it out, and contribute to it by constructive criticism and new data.