Sage Journals: Discover world-class research

Abstract

Sumerian, an agglutinative language of unknown affiliation, surfaced in mankind’s earliest written sources around 3300 BCE. It continued to play a salient role in the transmission of cuneiform cultures for more than three millennia, even after its disappearance as a vernacular (around 2000 BCE). From its beginnings, semantic classification played an important role in this initially predominantly semasiographic-ideographic script. Here, such classifications are contrasted with and compared to the normalized ‘consolidated’ standard set of pre- and post-nominal classifiers – generally known as ‘determinatives’ – which function chiefly as sortal classifiers. It is further suggested here that the relationship between classifying nouns and classified nouns is best described in terms of nominal apposition. This article further provides updated examples of how the relevant information was entered into the iClassifier digital tool, specially designed for such comparisons in order to improve our understanding of how cuneiform classifiers mark their host words and to make them comparable with similar features in other complex scripts and other languages. We will then briefly survey how the answers to queries in the iClassifier may help to illuminate ‘the diverse mindsets of ancient Mesopotamia’.

Keywords

Determinatives/classifiers knowledge mapping nominal apposition semasiography semantic classifiers

Introduction

Writing surfaced in ancient Mesopotamia shortly after the middle of the 4th millennium BCE. Its evolution from earlier symbolic systems – accounting devices and visual representations – is generally accepted. The various stages are well documented: sealed number tablets and spherical clay balls bearing numbers and seal-impressions are understood as devices developed in order to ease accounting and controlling the movement of goods.

An important role in this development was played by the ‘token system’, a term that refers to small clay pieces of different shapes that are mostly abstract and therefore based on arbitrary semantics, but also include a considerable number resembling or representing objects from the natural environment (see Schmandt-Besserat, 1992, 1996; and especially Michalowski, 1993; also Wagensonner, 2009 and the collection of articles in Woods, 2010; and see now Selz, 2020a: 192–200 and Selz, 2022). Counting could be done by the manipulation of such objects in an abacus-like style; there is no need to pronounce such operations and linguistic aspects might be neglected. Such clay tokens were occasionally impressed on clay tablets, small cushion-like lumps of clay functioning as a kind of proto-signs.

Almost at the same time, the ancient accountants began using reed styluses in order to impress such ‘signs’ on tablets. The obvious motivation for these processes was an attempt to make the embedded information lasting and accessible for all sorts of control. One might therefore say that the combination of (durable) visual information and the rather ephemeral accounting processes stood at the birth of the Mesopotamian writing system(s). Of course, it was possible to pronounce such information, but a link to any specific language could be ignored (cf. Selz, 2021b).

The set of signs used in the early script was quite limited: some show their pictographic origin while others seem – at least to us – to be rather arbitrary ‘abstract’ signs. However, we observe in the earliest script rather sophisticated means for enlarging the body of such signs: the turning and mirroring of a sign, alteration by omitting or adding glyphs (‘strokes’) to a specific sign, and, above all, a synecdochic and metonymic use of some signs. Depictions of animal heads, for instance, are used to designate a member of animal species, and the depiction of several vessels is used to refer to offices with which they are typically linked, thus making reference to various social statuses (see D’Anna et al., 2016; Johnson, 2019; also Selz, 2020a: 214–216 and Selz, 2022). Also well attested is the combination of different signs, either by simply compounding signs or by sign-combinations, combined signs, in which often one or more ‘miniature’ signs are inscribed into a larger ‘framing’ sign.

It is important to note that the material (clay) and instrument (reed stylus) certainly contributed to the rapid loss of iconicity in the Mesopotamian script. The individual impression of a sign, the glyphs, soon took the shape of ‘wedges’ – hence the labelling of the typical Mesopotamian writing as cuneiform (‘wedge-shaped’).

The history of sign formation and the traceable use of these signs show, beyond reasonable doubt, that this script is best described as a kind of conceptual writing, despite the widespread labelling of these signs as ‘logograms’. They do not correspond to one word in a given language but rather to a set of ideas connected to these signs. Thus, the sign for HEAD 𒊕 – a simple depiction of a human’s head – may be expanded to the sign MOUTH 𒅗 by adding marking strokes on the mouth area of that picture, thus creating a new, derived sign. The sign MOUTH 𒅗 could then be used – and was used throughout the history of cuneiform writing – not only to denote MOUTH, but also WORD, TO SPEAK and other referents connected conceptually with MOUTH. Furthermore, in combining this MOUTH sign with the depiction of RATION VESSEL, the derived (new) sign could express the semantic concepts of EATING and CONSUMPTION. Conceptual or semasiographic (ideographic) writing is undoubtedly grounded in the polysemy of the written signs. It should also be noted that in the earliest stages of the script no unequivocal evidence for the existence of phonophores has yet been found.¹

To this day, the notion that writing has to be a fully fledged system to render speech prevails in scholarly literature (Boltz, 1994: 8; Falkenstein, 1936: 31; Gelb, 1963: 65; Handel, 2019: 41; the entire situation is reviewed in Selz, 2022). It has been repeatedly noted, however, that semasiographic writing dominates the transmission of Sumerian texts;² a relatively clear linguistic system representing the speech cannot be traced prior to the end of the 3rd millennium BCE, roughly one millennium after the invention of the script. However, a conceptual or ideographic writing system is connected to speech, albeit not necessarily to a specific language. Since the earliest written documents were unearthed in the southern Mesopotamian city of Uruk – where the later literary tradition in fact locates the invention of writing – the issue of the language behind these documents has been addressed repeatedly and has generated much controversy.

Scholarly discussion was not limited to the ideographic and conceptual hypotheses; much effort was also made to identify and reconstruct the language behind the earliest Uruk texts. Throughout the history of cuneiform writing, the principle of homophony and homoiophony (similar phonetic shape) played a salient role, especially in Sumerian, a language with a large number of homonymous words.³ Cuneiform script exhibits the rebus-principle, which means that a graph linked to a specific word is also used to write quite different words showing the same or similar phonetic shape. This phonetization of signs became necessary in order to render all sorts of names and titles.

By analysing these phonetic components, scholars made several attempts to identify the language that the script was being used to encode. While some of these lead to dead ends, most scholars – with the notable exception of Englund – are presently convinced that speakers of Sumerian made some contribution to the moulding of this early script.⁴ Sumerian is an agglutinative language without inflection and rather narrow word boundaries. Therefore, it is extremely well-suited for the application of the rebus-principle. Be this as it may, the fact that the early Mesopotamian writing system is deeply grounded in two opposing principles – polysemy and homophony – can hardly be questioned.

The beginnings of semantic classification in the earliest texts from Uruk

Roughly 85% of the earliest Uruk texts originate in administrative contexts, most likely documenting the expenditures (specifically distributions), income and inventories (treasuring) of staple goods controlled by specific offices or persons. However, 15% of the texts are usually labelled ‘lexical lists’, and they have proven to be a salient means for understanding the semantics behind the administrative documents.

The most obvious purpose of the Uruk lexical lists was – as also in later tradition – to make the apprentice scribe familiar with the system. However, despite the fact that they show no particular concern with ‘metaphysical’ issues, the lists provide us with important insights into the Mesopotamian worldview. The world represented in these lists is organized in various themes, all reflecting the contemporary environment. Most importantly, these lists were transmitted to the later periods, although with notable changes.⁵ Consequently, scholars of the ancient Near East call these texts ‘thematic lists’; their organizational scheme we have tabulated repeatedly points to their rather persistent scribal tradition.⁶

These thematic lists can therefore also be taken as an expression of a Mesopotamian endeavour to organize and categorize their world: they are, in the strict sense, classificational. A more detailed analysis of the individual lists will further reveal classificational hierarchies and one might even say that they explicitly deal with the semantics of the signs and words. Most likely this attitude also affected sign formation; of special interest here are the so-called frame signs, a number of signs in which other specifying signs are inscribed.⁷ In these frame signs, the frame might convey semantic information of the classifier type.⁸ As an example, we choose here the sign EZEN/M (𒂡), which in the later tradition conveys the meaning FESTIVAL.⁹ The word ‘festival’ entered Akkadian as a loan-word, isinnu. Pictographically, the sign represents a central place with four or two gateways , and from the original meaning MEETING PLACE it acquired, by semantic extension, the meaning FESTIVAL.¹⁰ In the earliest texts, however, in conjunction with the various inscribed signs, it functions sometimes as a kind of classifier for geographical names, in much the same ways as later the (37) [place name] classifier (𒆠) (which replaced it).¹¹ Necessary specification then was achieved by insertion of an additional miniature sign (EN) resulting in the form (ZATU 151)– that is, EZEMxEN (Uruk IV period): the combined sign most probably stands for ‘festival’ and the function of the inscribed sign EN is either as semantic or phonetic¹² indicator.

If this observation is correct, a closer look on such frame or matrix signs is wanted; besides classifiers based on their identification by various scholars in the ‘classical’ standard Sumerian script, we may reckon with further means of classification, at least in the earlier phases of cuneiform. There, frame or matrix signs play an important role. Some of them may have a graphic semantic function that so far¹³ was rarely considered.¹⁴ Such frame or matrix signs may have the function of a semantic indicator.

In addition to the just discussed EZEM, we would like to add a few remarks on the DUG. The sign is differentiated into DUG_a and DUG_b as follows: (ZATU 88). In fact DUG_b has a shape quite similar to the sign UKKIN that, however, has to be kept apart:

(ZATU 580).¹⁵

DUG_b is clearly the forerunner of the later 𒂁 ‘(clay) pot’; already Green and Nissen (1987: 189) observed that DUG ‘with inscribed sign functions as determinative’; in other words, they suggested a connection between a frame sign and the later classifier [pot].¹⁶ The meaning and consequently the reading, however, are not always certain: thus, (cf. ZATU 97) is DUG_bxGA ([POT]x[MILK]) and (cf. ZATU 118) is DUG_bxŠE ([POT]x[BARLEY]); they might semantically refer to ‘milk pot’ or ‘barley pot’, or to ‘pot of milk/barley’. Therefore, we cannot (yet) establish a reading. Controversially, in the related sign DUG_bxKASKAL it was suggested that /kaskal/ 𒆜 ‘road’, as with the value /kas/, is here a phonetic indicator and that the combined sign may consequently stand for /kaš/ ‘beer’. As far as we can see, however, a semantic interpretation ‘travel vessel’ is by no means excluded.

The signs GA_2a and GA_2b (ZATU 162) originally depict boxes of some sort. Both forms are attested in Uruk III-IV periods with a greater variety of inscribed miniature signs: thus, in GA_2axEN or MEN (ZATU 360), later read /men_x/ 𒃉 and /men/ 𒃞 – the word designating a type of crown – the inscribed sign EN may be either a phonetic indicator /en/ or have a semantic function, for example, [BOX] (GA₂)x[LORD](EN). Similarly in AMA_a, analysed as GA_2bxAN/AM₆ , and also written AMA_b (ZATU 28), both are later 𒂼 and rarely 𒆾 = ama₆(?). Later, the compounded sign has two readings /ama/ ‘mother’ and /daĝal/ ‘wide’. It is only for the first word that we may suggest that AN/AM₆ is a phonetic indicator, but also a semantic interpretation [BOX]x[DIVINE] is not excluded¹⁷ and may eventually explain why the sign represented the notion ‘wide’.

A little later, in the Early Dynastic periods, we find the sign GA_2axNUN , also written ĝa₂-nun, for the word ĝanun ‘storehouse’. It is tempting to suggest that signs like GA_2axGUD , GA_2axMAŠ (ZATU 176), GA_2axKU₆ (ZATU 173) and even GA_2axKUG (ZATU 172) compounds designate containers or depositories for GUD ‘oxen’, MAŠ ‘goats’, KU₆ ‘fish’ and even ‘shining metal’ KUG. Whether this was always mirrored linguistically – for example, by nominal compounds– remains uncertain. But this assumption can be corroborated by a word like /araḫ/, written GA₂xŠE [BOX] x[BARLEY] (= araḫ₃) and E₂xŠE [HOUSE]x[BARLEY] (= araḫ_x),¹⁸ writings that suggest that the word designates a ‘barley storage facility’ and has phonetically ‘house’ (ʾa₃/e₂) as its first element. Interestingly, the compound writing e₂-us₂-gid₂-da, literally ‘house with long corridors’, is often understood as a DIRI writing¹⁹ with the reading araḫ₄.²⁰

Therefore, all types of semantic classifiers may originate as elements of nominal compounding, notwithstanding the observation that they increasingly became purely graphic elements in the standard form of cuneiform scripts.

Determinatives/classificators/classifiers: The situation in ‘classical’ Sumerian texts

Modern grammars of Sumerian mention several classes of cuneiform signs: besides (a) logograms (or semasiograms, as we prefer) these are (b) syllabograms (derived via the rebus-principle from (a)) and a special sub-class labelled phonetic indicators, referring to the pronunciation of a semasiogram in a given context, (c) number signs (including ‘combined notations of measure + number’) and, as a separate class, (d) ‘determinatives’.²¹ In ancient Near Eastern studies, the term ‘determinative’ is used for allegedly unpronounced glyphs, ‘which precede or follow words or names in order to specify them as belonging to semantic groups’ (Edzard, 2003: 9). It is remarkable that this term was introduced into cuneiform studies in 1847 by Edward Hincks,²² who was working in the decipherment of Egyptian and cuneiform scripts, and modelled apparently after the term Champollion coined in 1836 in his study of ancient Egyptian (see Goldwasser, 2006: 17–20). However, due to the chronological and typological diversities of cuneiform texts, the set of what scholars consider ‘determinatives’ is rather ill-defined. Considering the various accounts of renowned scholars (specifically Borger, 1978; Edzard, 2003; Ellermeier, 1979; Falkenstein, 1949, 1964; Foxvog, 2016; Huehnergard, 1996 [1998]; Labat, 1988 [1948]), we produced in 2017 a ‘consolidated list of cuneiform classifiers’, the shortcomings of which – most stem from a dearth of diverse sources – were explicitly acknowledged (see Selz et al., 2018: 287–305).

The Selz consolidated list of cuneiform classifiers

The list contains 50 items and is as follows (the position of the classified item is here marked by a hyphen (–))²³ (based on Selz et al., 2018):

1. –*didli (HAL) [several]²⁴ 𒀸𒀸; 2. *kuš (SU)– [hide/leather] 𒋢; 3. *diĝir (AN)– [deity] 𒀭; 4. uru/eri– [city] 𒌷; 5. iti/itu(d)– [month] 𒌗; 6. –*mušen (HU) [bird] 𒄷; 7. gi– [reed] 𒄀; 8. gada– [linen] 𒃰; 9. sa– [bundle] 𒊓; 10. gana₂/aša₅–, also –iku [field(-measure)] 𒃷; 11. saĝ– [head/slave] 𒊕; 11a. ma₂– [ship] 𒈣; 12. mul– [star/constellation] 𒀯;13. –urud(u/a)(–) [copper] 𒍏; 14. uzu– [flesh/meat] 𒍜; 15. anše– [donkey/equid] 𒀲; 16. kaš– [beer /alcoholic drinks] 𒁉; 17. šim– [aromatics] 𒋆; 18. na₄– [stone] 𒉌𒌓; 19. ĝiš/ĝeš– [tree/wood] 𒄑; 20. gu₄/gud– [ox] 𒄞; 21. –ta-am₃ [each] 𒋫𒀀𒀭; 22. dug– [pot] 𒂁; 23. u₂– [grass] 𒌑; 24. e₂– [house] 𒂍; 25. lu₂– [person] 𒇽; 26. –nisig [greenary/vegetable] 𒊬; 27. kur– [mountain] 𒆳; 28. še– [barley/cereals] 𒊺; 29. mul₂ (TE)– [star/constellation] (see 12.) 𒋼; 30. –zabar [bronze] 𒌓𒅗𒁇/ 𒆉𒁇; 31. –hi-a [mixed] 𒄭𒀀; 32. IM/tu₁₅– [wind] 𒅎; 33. im– [clay] 𒅎; 34. –kam, –kam₂ [of it, marks ordinalia] 𒄰, 𒆚; 35. ul– [bud/flower](?) 𒌌, doubtful; 36. gig– [illness] 𒍼; 37. –ki [place] 𒆠; 38. *diš–, aš– [one] 𒁹, 𒀸; 39. –meš [plural < ‘they are’] 𒈨𒌍; 40. tug₂– [textile/garment] 𒌆; 41. zid₂– [flour] 𒍥; 42. tukul– [weapon] 𒆪; 43. udu– [sheep] 𒇻; 44. sik(i)₂– [wool] 𒋠; 45. mi₂/munus– [female] 𒊩; 46. –min [two] 𒈫; 47. –am₃ [it is] 𒀀𒀭; 48. id₂/i₇– [river] 𒀀𒇉; 49. ninda/inda₂– [bread] 𒃻; 50. –ku₆ [fish] 𒄩.

An evaluation of this list showed that 18 core items were identified by all or all but one of the authors, and can be considered as the constitutive set of Sumerian determinatives – that is, classifiers (Selz et al., 2018: 304). Based on the opinions of several prominent scholars, the thematic organization of the cuneiform classifier system reveals the picture shown in Table I, typographically differentiated in these ways: BOLD FACE acknowledged as ‘determinative’ by all authors; BOLD FACE ITALIC by all but one and UNDERLINE indicates that the classifier is a major topic of early lexical lists. ln addition, the position of the classifier is indicated by a hyphen before or after the classified item.

Table 1.

Thematic organization of the Sumerian classifier system according to the consolidated list (Selz et al., 2018: 308).

SORTAL		MENSURAL
ANIMATE	INANIMATE	Unitizers/plural – one, unit (38) individual things– (1), (47) –head (11) distributive– (21) dualis– (46) plural– (31), (39) Mensurals (arrangements and containers) –bundle (9) – jar, jug, pitcher (22)
Human and deity – deity (3) – human, male (25) – human, female (45) –human (38)–illness (36) Animals Quadrupeds – bovine (20) – sheep (43) – donkey (15) More animals fish – (50) bird (and insect)– (6) Body parts of humans and animals – flesh (14) – skin/leather/hide ( 2)
Vegetal –wood/lumber/tree (19) –reed (7) plant/grass – (23)– barley, cereal (28) vegetables – (26) aromatic, perfume–(17) Manufactured – jar, jug, pitcher (22) From vegetal –flax linen (8) Never used for the plant! – cloth/textiles (40) – beer (16)–flour (41) – bread (49) –ship (11a(?)) From animal – animal hides/leather (2 ) –wool (44)– cloth/textiles (40)
	Mineral world (and man-made objects) – stone (18)–clay (33) (–)copper– (13) bronze– (30) –weapon (42) Natural elements – wind (32) – star 1 (older) (12), (35) –star 2 (younger) (29); cf. (35) Spatial place– (37) – mountain/land (27) –river (48) Man-made spaces – field/piece of land (10) – house (/container) (24) – city ( 4) garden plot/vegetables – (26) Time –month (5)

With few exceptions – probably later inventions – we can observe that nearly all classifiers are sortal classifiers. In 2017 and 2021, we provided a structured table on the organization of the standard Sumerian classifiers.²⁵ When we ignore here (21) [distributive]–, (46) [dual]–, (31) and (39) [plural]– and (47) –[unitizer], which are actually Sumerian grammatical morphemes and not used as ‘classifiers’ in classical Sumerian contexts, we see that all these classifiers appear as free morphemes having lexeme status as autonomous words. They function as sortal classifiers with the exception of (9), (11), (22) and (38), which might compare to mensural classifiers. Therefore, whenever they precede or follow another word in order to classify it, they follow the pattern of Sumerian nominal compounding; more precisely, they can be described as appositive nominal compounds.²⁶ As Brigitte Bauer states: ‘the primacy of classification in man’s thinking is crucial to the classifying function of nominal apposition’ (Bauer, 2017: 35).²⁷ The lexical lists, as well as our list of classifiers, show a profound engagement with taxonomies – classificational endeavours – which provide us with an unsurpassed insight into the organization of the ancient world (see Selz et al., 2018; also Goldwasser, 2003; Goldwasser and Grinevald, 2012). These classifiers may, of course, refer to a semantic class only graphically. However, the contention that in any case they ‘can be proven not to have been pronounced’ (Edzard, 2003: 9; similarly, Civil, 2004: 3–4²⁸) is in our opinion quite imprecise and indeed incorrect. Even Civil explicitly noted the fact that ‘in some cases the classificators were in fact parts of nominal compounds and were pronounced, while a true classificator was never pronounced’. Be this as it may, these classifiers have to be studied as a linguistic phenomenon.²⁹

The relationship between the most common – and widely accepted – classifiers and the lexical tradition demonstrates the salient taxonomic and classificational reasoning of ancient Mesopotamia: ‘It is quite intriguing that of 13 “themes” of the so-called “thematic lists” a minimum of 10 relate to the later so-called “determinatives”’ (Selz, 2018: 415). The ‘determinatives’ do in fact classify (see already Farber, 1991: 83 and cf. Selz, 2018: 415), so henceforth, we should perhaps better use the term ‘(graphic) classifiers’. While the majority of scholars insist that these classifiers are purely graphical means to attribute a word to a semantic class (cf. above Edzard, 2003: 9; Civil, 2004: 3–4), the hypothesis that the Sumerian system of ‘classifiers’ or ‘semantic script determinatives’ is grounded in language was already suggested by Gadd (1924: 13): ‘It is by no means certain that their use (“determinatives”) is merely a device of writing. On the contrary, it is a strong probability that they were in most cases actually pronounced.’³⁰ Bauer (2017: 67) summarizes the situation: ‘There is a remarkable lack of motivation of why script determinatives allegedly were not pronounced or – conversely – why that assumption is doubtful.’

As all Sumerian sortal classifiers are common nouns (nomina substantiva), it is necessary to study their linguistic function within the frame of noun-compounding. It is suggested here that the study of Sumerian noun-compounding and especially appositive compounding may illustrate what happened to the classifiers and elucidate their semantic functions. The vast majority of Sumerian nominal compounds (noun compounding) are left-headed.³¹ This head is then semantically modified, specified or restricted by the addition of an attribute, either a nomen substantivum or adjectivum,³² and might also be a deverbal (participle). We note that what is commonly termed classifier/determinative – the pre-positioned generic or common noun – corresponds to the hypernym in non-coordinate nominal apposition.³³ In deverbal modifiers, the modifier itself can be a compound, which, however, look like right-headed compounds according to the Sumerian verbal syntax. And further, there exist countless cases where the head of such a nominal compound is not written; these occur most commonly in profession names, but also appear in other items.³⁴ However, many cases are attested in which the head is expressed, and in post Old-Babylonian texts this manner of writing became dominant. In other words, we suggest that the pronunciation of the pre-positioned classifiers was a possibility, while in numerous cases they certainly remained unpronounced (and as the long tradition of cuneiform developed, increasingly so).

The situation of the post-positioned classifier is a bit different. There are only three major post-positioned classifiers: (6) [bird], (50) [fish] and (37) [place name], the latter of which only appears at the end of the Early Dynastic period (ED IIIb, c. 2350 BCE). For a fourth, (26) [vegetable], we could not find an attestation prior to the Old Babylonian period. With respect to the classifier (13) [copper], the situation remains unclear:³⁵ the post-position is attested since ED IIIb (c. 2350 BCE), while in the Gudea period (before the 21st century BCE), changing positions are attested. This parallels the case of (30) [bronze], which in the dominant post-position appears towards the end of the Early Dynastic period (ED IIIa, c. 2450 BCE) for reasons not clear to us (but see below).³⁶ It is suggested below that the post-positioned [copper] and [bronze] classifiers are not material classifiers in the strict sense, – that is, that they are specifications³⁷ used like adjectives and therefore do only in a limited way belong to the system of classifiers discussed here.

If we accept this, we are left with only four post-positioned classifiers: [fish], [bird] and the somewhat younger³⁸ [place name] and [vegetable]. These post-positioned classifiers follow a well-established appositional pattern where a proper name is specified (semantically restricted) by the ensuing appositive noun in compound-final position.³⁹ Names of [fish] and [bird] seem to be treated like proper nouns⁴⁰ similar to place names followed by the classifier (37) [place name] and, we suggest, for [vegetable] a similar function seems likely.

Therefore, we suggest that the different position of the classifiers can be explained best within a hypernym-hyponym grid – that is, as a NOUN-APPOSITION chain. Whether or not they have become purely graphic indicators by the ‘classical’ Sumerian period, classifiers are linguistic phenomena, not mere morpho-graphemic aids. We hope that a comparison with other noun classifier languages as envisaged in the scope of the iClassifier project may yield additional insights (see Selz et al., 2018, especially 306–307; for the hieroglyphic classifiers, see also Goldwasser, 2006).

The iClassifier project and its application to Sumerian texts (preliminary report)

The starting point

So far, we have hopefully demonstrated the importance of Sumerian semantic nominal classifiers and the problems they precipitate in identifying the connected problems, especially the difficulties in identifying them and understanding their semantic functions. We turn now to a preliminary account of classifiers gleaned from the implementation of the Jerusalem-based iClassifier project. The greater benefit scholarly transcriptions of cuneiform texts offer is that they must account for every glyph or graph of the original document. We further possess the ePSD and ePSD2 (electronic Pennsylvania Sumerian Dictionary), a highly advanced electronic tool for lexical research, somewhat compensating for the lamentable fact that a stable Sumerian Dictionary is not yet in sight.⁴¹ In the ePSD2, all classifiers are – according to good Assyriological practice – rendered in superscript, which opens a nice window for retrieving all presumed classifiers from this database. An additional advantage the ePSD2 offers is the ability to link to other databases incorporated into each entry, above all the CDLI (Cuneiform Digital Library Initiative), DCCLT (Digital Corpus of Cuneiform Lexical Texts), BDTNS (Database of Neo-Sumerian Texts) and ETCSL (Electronic Text Corpus of Sumerian Literature), thereby providing references to the individual original sources. Therefore, we hope that in the future we will be able to trace all collected information on a given lemma and its tokens to the original sources, to photos as well as to copies and transliterations of an individual text.

When Bo Zhang collected and entered the information from ePSD2 into the iClassifier database, we immediately became confronted with three major problems:

Besides classifiers/determinatives, by academic tradition phonetic indicators (also termed phonetic classifiers or phonophores) are also given in superscript. For Sumerian, such phonetic indicators are simple reading aids in order to disambiguate linguistically the various semasiograms. However, in order to make our data set compatible with other projects with similar phenomena, we retained these phonetic indicators, but programmed the query system to allow for their exclusion.

Semantic nominal classifiers are used not only for classifying simple common or proper nouns; they may also classify an entire nominal compound, or a single element of a compound which sometimes leads to multiple classification – for example, a divine name bearing the classifier [divinity] often contains a toponym bearing the classifier for [place name]. In such cases, the description of a classifier’s function is reserved to commenting; in other cases, for several nouns given in superscript, their identification as classifiers remains doubtful.

An additional problem already addressed is that the existing databases, including ePSD2, sometimes make use of rather diverging transcriptions and what was perceived as a (mute) classifier or part of a pronounced nominal compound is often varying and inconsistent.

Thus, for a start we simply follow the transliteration modi of ePSD2. Therefore, all signs shown in superscript are considered as (possibly) relevant for our topic; only in a second step do we distinguish between phonophores and lexemes considered to represent classifiers. The result of this simplified use of the ePSD2 transcriptions is that they include a larger number of words, specifically material qualification, attested in adjectival position. Most of these alleged classifiers are in fact part of a NOUN + NOUN compound.⁴²

Data input – current state

So far, Bo Zhang has entered nearly 2800 lemmata, according to the normalized transcription of ePSD2. We consider only those lemmata that show at least one token with an identifiable classifier or phonetic indicator.⁴³ Tokens in our project are defined as script forms (transliterations as well as the basic original data sources can be reached via ePSD2). Statistics as of 12/04/2023 run as follows with almost double figures when compared with Selz (with Bo Zhang) (2021b) (nonetheless, the total number of qualified lemmata turned out to be even larger than we had estimated in 2021, and now we estimate that it may go beyond 6000 entries):

Number of script variations (tokens):

9297

Number of lemmas: 2810

Number of tokens by classifier count

With one classifier: 4767

With two classifiers: 454

With three classifiers: 11

Without a classifier: 4065

This means that the database presently (12 April 2023) contains 4065 tokens (script variations) without a classifier,⁴⁴ 4767 tokens with one classifier, 454 with two classifiers and 11 tokens with three classifiers.

All tokens (script forms) of a selected lemma, either with or without a classifier, are entered into the database (provided the lemma has at least one token with a classifier). A token presently refers to the transliterated form of a lemma, and does not necessarily reflect the transcribed (phonetically reconstructed) form of the word. In the add/edit tokens interface, we enter the following information:

Choose the lemma to which a token belongs from the list of lemmas already put in.

Choose ‘token type’. Presently, we register only simple tokens.

Input the cuneiform writing form of a token in the field Token – for example, (= bulug).

In the field Token with markup,

for an un-classified token (i.e. a token without a classifier), we copy its cuneiform attestations, with no markup, in this field:

for a classified token, tildes are used to mark its classifier(s): ∼ ∼ (=^urudbulug).

In the sub-field of Classifiers, choose Classifier level; we simply distinguish between lexical and phonetic.

Under Phonetic reconstruction, we offer the normalized transcription– that is, the phonetic representation of a lemma as in ePSD2 – for example, bulug.

Under Transliteration, we enter the transliteration of a given token– that is, its graphical representation according to Assyriological conventions. Braces are used to mark up its classifier(s) (or phonetic indicators) – for example, {urud} bulug.

Under Classification status, we choose between ‘classified’ and ‘not classified’.

Under Comments, we always enter the numbers assigned to the script forms (token) according to the sequential order of script forms of a lemma in ePSD2; in the future, we hope to connect this with the respective HYPERLINKS provided by ePSD2. Thus, eventually a chronological distribution will become accessible directly, adopting the grid used by ePSD2. Periods are: PC (Proto Cuneiform) – ED (Early Dynastic) IIIa – ED (Early Dynastic) IIIb – Ebla – OAkk (Old Akkadian) – Lag II (Lagaš II dynasty) – Ur III – OB (Old Babylonian) – Post-OB (Post-Old Babylonian) – (unknown).

Other relevant observations or discussions may be also included here – for example, the semantic range of a lemma, its function in nominal compounding or dialectic variations.⁴⁵

Classifier list by iClassifier

Presently, we count 46 semantic⁴⁶ classifiers.⁴⁷ When they are arranged⁴⁸ in a graph (Figure 1), the list of classifiers exhibits a long-tailed distribution.⁴⁹ The meaning is given to the right of the classifier.

Figure 1.

The classifier list according to iClassifier.

The most attested classifiers are [place], [wood] and [divinity]. In comparison to the standard set provided further above we have 𒉾- ‘mouse’, 𒊩𒆳- ‘female slave’, 𒉅- ‘locust’, 𒋢𒁷- ‘bat’, 𒈲- ‘snake’ and 𒂵- ‘milk’ as new classifiers. Besides showing the productivity of the system, their functions require further discussion as do the ‘missing’ classifier numbers 29 𒀯 [constellation], 35 𒌌 [bud], doubtful or 36 𒍼 [illness] from our standard set. In sum, such alterations demonstrate that the system is alive throughout several millennia.

Classifier axis: Classifier reports on a single classifier: The [hide] classifier

By clicking on the classifier, we receive the following information, with kuš [hide, leather] as an example.

Figure 2 shows all lemmata attested with this [hide] classifier. The hue of the node represents the percentage of occurrences of a certain lemma with this classifier. For example, the lemma ĝirikur ‘drum’ has a dark (red) hue, representing that it occurs with this classifier in all of its attested script variations (4 out of 4 forms). A light colour represents that the classifier occurs in fewer script forms with a lemma – for example, garig ‘comb’ (1 out of 12 script forms).

Figures 2-3. Figure 2 (left).

A map with all lemmata attested with the kuš [hide] classifier. Figure 3 (right). List of exemplary script forms (tokens) with the kuš [hide] classifier.

The thickness of the (blue) lines between the classifier and each host lemma represents how many script variations a lemma has with a certain classifier.

We can see that the classifier kuš 𒋢 [hide] (Akk. mašku; zumru) is – according to ePSD2 – attested with the following lemmata.

a [arm] (‘a strap’); A₂ (‘a weapon or a leather holder for a weapon’);⁵⁰ aguhum [clothing]; aĝala [sack]; akar [armor]; ala [manacles]; ala [drum]; ama'usan [∼whip] (‘part of a whip’); antibal [sign] (‘kind of leather bag’); asi [strap]; balaĝ [instrument]; balaĝdi [instrument]; bar [outside] (‘skin’); baranse [goad]; bareda [strap]; barsig [garment]; barutum [object] (‘kind of quiver’); da [line] (‘shoe strap’); daba [strap]; dabasin [object] (‘tarpaulin?’); dabihum [slaughterer]; dagsi [hook]; dala [thorn]; daluš [sling]; damga [∼plow] (‘part of a plow’); dubšum [object]; duggan [bag]; duggantug [BAG]; duksium [shield]; duggantibala'a [bag] + [sign] (‘bag for wooden signs?’);⁵¹ dusi'a [stone] (‘turquoise?’, > ‘greenish leather?’); e [leather strip];⁵² ellag [bead] (‘leather string for beads?)’; ellamkuš [bladder]; eme [tongue]; esir [shoe]; esir [fish]; emarru [quiver]; eme [tongue]; es [tent]; ezulumak [quiver]; gababum [shield].

We can also observe that the several script variations (tokens) for /esir/ ‘sandal’, /emarru/ ‘quiver’ and /aĝala/ ‘sack’ often receive the [hide] classifier, but queries for each lemma show classification variants: /esir/ ‘shoe, sandals’ is also used once with the urud– [copper] classifier. The word /emarru/ ‘quiver’ also has the ĝeš– [wood] and /aĝala/ (or /aŋala/) the –tug₂ [cloth] classifier_.⁵³ The word /antibal/ ‘sign’ may also take the [wood] or [copper] classifier, indicating that ‘signs’ may be made of wooden, copper and/or leather parts.

ln the case of /esir/ ‘a type of fish’ (certainly a flatfish of the order pleuronectidae) and lexically differentiated from /esir/ ‘shoe sandals’, the use of the [hide] classifier underlines the metaphoric origin of the fish's name, but has no other semantic function. The fish name normally occurs, of course, with the [fish] classifier. The word /esir/ also has the meaning ‘bitumen’ and ‘street’, both attested without a classifier.⁵⁴

The list of items attested with the [hide] classifier demonstrates beyond doubt that, in many cases, the [hide] classifier is part of an (original) nominal compound. Nevertheless, in numerous cases it is difficult or even impossible to decide whether [hide] is part of noun-formation. kuš-balaĝ(-di) or ĝeš-balaĝ(-di) refers to the material from which harps or drums are made⁵⁵ and mul-balaĝ refers to the eponymous constellation (mul), ‘The Drum’. Further, balaĝ(-di) is not only used for nomina instrumenti, it also functions as nomen agentis ‘(the one) who plays the harp/drum’> ‘lamentation singer’. In other cases, it seems likely that [hide/leather] was originally not silent, as in kuš-e-sir₂ or kuš-esir_3-5 [HIDE] + [STREET] ‘shoe’, which is thus distinguished from the allegedly homophonous /esir/ ‘street’.

ln a separate subfield exemplars of script variations – our tokens – attested with the [hide] classifier are listed in Figure 3.

The frequency of tokens (i.e. script forms) for a specific lemma is provided by the statistics given in Figure 4.

Figure 4.

Percentages of occurrence of each lemma with the kuš [hide] classifier in all its script variations.

Lemma axis: All classifiers attested with a specific lemma: esir ‘sandal’

Within the category, we present a sample lemma /esir/ ‘sandal’.

Figure 5 produced by iClassifier presents the first three tokens with zero classification, 10 tokens showing the [hide] classifier variation, and one showing the [copper] classifier in initial position (classifiers are coloured in red). Figure 6 shows the distribution with the lemma esir ‘sandal’: we see two classifiers. Most examples, when classified, show the [hide] classifier. A single example shows the [copper] classifier.

Figure 5-6. Figure 5 (left).

List of all script variations of the lemma /esir/ (with and without a classifier). Figure 6 (right). Classifiers attested with the specific lemma /esir/ ‘sandal’. Thickness of the blue line represents the number of examples.

The classified Cuneiform World NETWORK MAP

The iClassifier map. Concentrating on the phenomenon of unpronounced (mute) cuneiform classifiers, the iClassifier database produces a dynamic interactive map that provides important information on how the ancient peoples classified their world. To be sure, this map does not and will not cover all taxonomic or classificational phenomena,⁵⁶ which will continue to be uncovered by future research. However, even when restricting ourselves to the allegedly ‘mute’ classifier, we obtain an important basis for any comparison with similar phenomena in other writing systems – for example, Ancient Egyptian, Chinese or even Luwian and Maya scripts.

All the information entered into the iClassifier digital tool is summarized in a dynamic network classifier map (Figure 7). This map arranges all classified items (lemmata) with their classifiers. The frequency of tokens (script variations) is indicated by the thickness of the connecting lines. In the case of multiple or changing classifiers a lemma is, of course, connected to all classifiers attested with it. Although the frequency refers so far only to the script variations – our tokens – this representation offers a quick, intuitive understanding of which classifiers are ‘core’ and which are ‘fringe’.

Figure 7.

A classifier-based network of Sumerian, according to the uppercase marking of classifiers (aka determinatives) in the ePSD2 lemma list, after Selz (2021b), updated (2023), includes 2810 lemmata, and 9297 written forms, digitized by Bo Zhang in iClassifier ©Goldwasser/Harel/Nikolaev. The Sumerian iClassifier, Conceptualization: Gebhard Selz. Data input: Bo Zhang, ArchaeoMind Lab (https://archaeomind.huji.ac.il).⁵⁹

At the fringes of the closer intertwined core sections, we see several rare and isolated classifiers. These are, for example, the gud [ox] classifier 𒄞, which only precedes the words /alim/ ‘bison’ and /am/ ‘bull’. The classifier /zid/ (zid₂) [flour] 𒍥 occurs only with the flour types /baba/, /barsi/, /edadi/ and /eša/. The /kaš/ [beer] 𒁉 classifier is attested with  /eblak/ ‘a Syrian(?) beer’, /dida imĝaĝa/ ‘sweet beer’ and /dida/ ‘beer wort’; this latter word also appears with the /dug/ [pot/vessel] 𒂁 classifier.

After the [divinity] classifier AN 𒀭, the most widely used classifier is ĝeš- [wood] 𒄑, which demonstrates a certain level of interchangeability with other material classifiers such as gi- [reed] 𒄀, u₂- [grass/plant] 𒌑, -nisig [vegetable] 𒊬, na₄- [stone] 𒉌𒌓, urud- [copper] 𒍏, but also dug- [pot/vessel] 𒂁, which stresses functionality (see Figure 8).

Figure 8.

Classifier map of ĝeš- [wood] and its relation to other material classifiers.

Quite expectedly, the -ki [place] 𒆠 and diĝir- [divinity] 𒀭 classifiers possess a large number of tokens, as a result of the larger number of place or divinity names included in the ePSD2 database (Figure 9). What might be surprising, however, is the number of tokens attested with both classifiers (indicated by the thick (red) connecting line), which evidently stresses the significance of local hypostases of the divine. The thinner (red) lines linking the na₄- [stone] classifier 𒉌𒌓 or the -mušen [bird] classifier 𒄷 with divinities allude to the representation of divinities either as statues or as bird-like emblems. The section shown here also displays, in a concise manner, existing links to river names and celestial constellations.

Figure 9.

Classifier map for -ki [place] and diĝir- [divinity] classifiers.

Conclusions and perspectives

Even at this preliminary stage, the relevance of our endeavours seems apparent: cuneiform classifiers usually attribute an item to a semantic class. The items – common nouns as well as proper nouns – stand in an appositive hyponymic relation to their classifiers. Our database offers unprecedented insights into the emic classification– that is, the ways in which these peoples conceived of their world and organized their mental representations.

Frequent uses of the various classifiers underline their emic mental salience. More specifically, this provides important insights into the taxonomic efforts of cuneiform scholars. In the future, when our database will account for the chronological and geographical distribution of these efforts, the perception of the evolution of cuneiform-linked Mesopotamian thinking will certainly become finer and more precise. Classification is, of course, a perpetual living process. As indicated above, a broader view of nominal classification should include linguistic features of nominal compounding, as well as additional graphic features attested in semasiographic compounding.⁵⁷ Subsequently, the evolution of the system of mute cuneiform classifiers can be described in more detail and better compared with related scripts. By now, it is already evident that the system of ‘determinatives’ evolved at the intersection between linguistic and mere graphic means of classification. Therefore, the origin of (the later unpronounced) cuneiform classifiers should be studied from the point of nominal and graphic compounding. In this vein, we hope to contribute to an improved understanding of cuneiform sources.

The data collected so far reflects the state of the art as presented in ePSD2.⁵⁸ It therefore overemphasizes the ancient lexical sources, and thus insufficiently reflects what might be termed ‘folk taxonomy’. Likewise, a genre-specific distribution of the classificational data is not yet in sight. However, as Mesopotamian lexical lists are the best and most important source for any sketch of the scientific reasoning of the ancient scholars, the value of our data collection is beyond doubt. It opens a window into the ways the Mesopotamians conceived of their worlds.

Footnotes

Acknowledgements

We are most grateful to Orly Goldwasser for including us into the iClassifier project and for numerous stimulating discussions. We are also obliged to Thedore Motzkin, who critically commented on this article and polished our English, and last but not least the operators behind the project, Dmitry Nikolaev and Haleli Harel (who also helped with the figures); see now Harel et al., 2023.

Authors’ contributions

Gebhard J. Selz: conceptualization, data curation, methodology, project administration, supervision, writing – original draft; Bo Zhang: data curation, investigation, writing – review and editing.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

ORCID iD

Bo Zhang

Notes

References

D’Anna

Jauß

Johnson

(2016) Food and urbanization: Material and textual perspectives on alimentary practice in early Mesopotamia, Introduction. Origini: Rivista di Preistoria e Protostoria della Civiltà Antiche 37: 8–14.

Attinger

(2021) Glossaire sumérien-français principalement des textes littéraires paléobabyloniens. Wiesbaden: Harrassowitz .

Besnier

(2021) From Sumer to IHAC: ‘Home(s) of the Fish’ – some thoughts on gathering (classifying) fishes. In: Günther

Horowitz

Widell

(eds) Of Rabid Dogs, Hunchbacked Oxen, and Infertile Goats in Ancient Mesopotamia: Studies Presented to Wu Yuhong on the Occasion of his 70th Birthday. (Supplements to the Journal of Ancient Civilizations, vol. 7). Changchun: Institute for the History of Ancient Civilizations, pp. 7–38.

Bauer

BLM

(2017) Nominal Apposition in Indo-European. Boston/Berlin: De Gruyter .

Boltz

(1994) The Origin and Early Development of the Chinese Writing System. New Haven, CT: American Oriental Society .

Borger

(1978) Assyrisch-Babylonische Zeichenliste. Kevelaer: Neukirchen & Vluyn .

Civil

(1969) The Series lú = ša and Related Texts. Materials for the Sumerian Lexicon XII. Roma: Pontificium Institutum Biblicum .

Civil

(2004) The Series DIRI = (w)atrum. Materials for the Sumerian Lexicon XV. Roma: Pontificium Institutum Biblicum .

Civil

(2008) The Early Dynastic Practical Vocabulary A (Archaic HAR–ra A) Archivi Reali di Ebla Studi IV. Roma: Missione Archeologica Italiana in Siria .

10.

Civil

(2010) The Lexical Texts in the Schoeyen Collection. Bethesda (MA): CDL .

11.

Cooper

(2004) Babylonian beginnings: The origin of the cuneiform writing system in comparative perspective. In: Houston

(ed.) The First Writing: Script Invention as History and Process. Cambridge: Cambridge University Press , pp. 71–99.

12.

Deimel

(1924) Šumerische Grammatik der archaistischen Texte. Roma: Pontificium Institutum Biblicum .

13.

Edzard

(2003) Sumerian Grammar. Leiden: Brill .

14.

Ellermeier

(1979) Sumerisches Glossar. Band 1, Teil 1. Nürten-Hardenberg bei Göttingen: Selbstverlag .

15.

Englund

(2009) The smell of the cage. Cuneiform Digital Library Journal 2009(4): pp. 1–27.

16.

Falkenstein

(1936) Archaische Texte aus Uruk . Berlin/Leipzig: Deutsche Forschungsgemeinschaft Kommissionsverlag O. Harrassowitz.

17.

Falkenstein

(1949) Grammatik der Sprache Gudeas von Lagaš I. Schrift- und Formenlehre. Roma: Pontificium Institutum Biblicum .

18.

Falkenstein

(1964) Das Sumerische. Handbuch der Orientalistik: Der Nahe und der Mittlere Osten. Zweiter Band: Keilschriftforschung und Alte Geschichte Vorderasiens. Leiden: Brill .

19.

Farber

(1991) Konkret, Kollektive, Abstrakt? Aula Orientalis 9 (Festschrift M. Civil): pp. 81–90.

20.

Foxvog

(2016) Introduction to Sumerian Grammar (Cuneiform Digital Library Preprints). Available at: http://cdli.ucla.edu/pubs/cdlp/cdlp0002_20160104.pdf (accessed October 2016).

21.

Gadd

(1924) A Sumerian Reading-Book. Oxford: Oxford University Press .

22.

Gelb

(1963 [1952]) A Study of Writing. Chicago: University of Chicago Press .

23.

Goldwasser

(2003) The building block of a conceptual category. In: Morra

Bazzanella

(eds) Philosophers and Hieroglyphs. Torino: Rosenberg & Sellier , pp. 219–236.

24.

Goldwasser

(2006) A comparison between classifier languages and classifier scripts: The case of ancient Egyptian. In: Goldenberg

Shisha-Halevy

(eds) Egyptian, Semitic and General Grammar: Jerusalem: Magnes Press , pp. 16–39.

25.

Goldwasser

Grinevald

(2012) What are ‘Determinatives’ good for? In: Polis

Grossman

Winand

(eds) Lexical Semantics in Ancient Egyptian. Hamburg : Widmaier, pp. 17–53.

26.

Green

Nissen

(1987) Zeichenliste der archaischen Texte aus Uruk. Berlin: Gebr. Mann .

27.

Grinevald

(2012) Basics on Classifier Systems, appendix to What are ‘Determinatives’ good for? In: Polis

Grossman

Winand

(eds) Lexical Semantics in Ancient Egyptian. Hamburg : Widmaier, pp. 46–53.

28.

Grinevald

(2015) Classifiers, Linguistics of. In: Wright

(ed.) International Encyclopedia of the Social & Behavioral Sciences, 2nd ed., Vol. 3. Oxford: Elsevier , pp. 811–818.

29.

Hackstein

(2010) Apposition and Nominal Classification in Indo-European and Beyond. Wien: Verlag der Österreichischen Akademie der Wissenschaften .

30.

Handel

(2016) Does Xŭ Shèn’s Huìyì category reflect historical reality? An argument for the existence of compound graphs lacking phonophorics. In: Ting

(eds) New Horizons in the Study of Chinese: Dialectology, Grammar, and Philology. Studies in Honor of Professor Anne Yue. Hong Kong, pp. 581–598.

31.

Handel

(2019) Sinography: The Borrowing and Adaptation of the Chinese Script. Leiden/ Boston: Brill .

32.

Harel

(in press) Comparing networks of semantic categories: Digitizing graphemic classifiers in ancient complex scripts using the iClassifier research platform. In: Chantrain

(ed.) Language, Semantics and Cognition: Saying and Conceptualizing the World from Ancient Egypt to Modern Times (Yale Egyptological Studies 14). New Haven , CT: Yale Egyptological Seminar, pp. 109–132.

(2023). Mapping the ancient mind: iClassifier, a new platform for systematic analysis of classifiers in Egyptian and beyond. In: Roberson

Lucarelli

Vinson

(eds) Ancient Egypt and New Technology: The Present and Future of Computer Visualization, Virtual Reality and other Digital Humanities in Egyptology. 29–30 March 2019 at Indiana University, Bloomington. Leiden: Brill , pp. 130–158.

34.

Hincks

(1846a) On the first and second kinds of Persepolitan writing. Transactions of the Royal Irish Academy 21: pp. 114–131.

35.

Hincks

(1846b) On the three kinds of Persepolitan writing, and on the Babylonian lapidary characters. Transactions of the Royal Irish Academy 21: pp. 233–241.

36.

Hincks

(1846c) On the third Persepolitan writing, and on the mode of expressing numerals in cuneatic characters. Transactions of the Royal Irish Academy 21: pp. 249–256.

37.

Hincks

(1847) On the inscriptions at Van. The Journal of the Royal Asiatic Society of Great Britain and Ireland 9: pp. 387–449.

38.

Houston

(ed) (2004) The First Writing: Script Invention as History and Process. Cambridge: Cambridge University Press .

39.

Huehnergard

(1996 [1998]) A Grammar of Akkadian. Atlanta, GA: Scholars Press .

40.

Jagersma

(2010) A Descriptive Grammar of Sumerian. Leiden : Scholarly Publications Leiden University.

41.

Johnson

(2019) Meat distribution in late Uruk diacritical feasts: Second-order bookkeeping techniques and their institutional context in late fourth millennium bce Mesopotamia. In: Renn

Schemmel

(eds) Culture and Cognition: Essays in Honor of Peter Damerow. (Max Planck Research Library for the History and Development of Knowledge, Proceedings 11). Berlin: Edition Open Access , Max Planck Institute for the History of Science, pp. 75–86.

42.

Keetman

(2020) Sumerisch auf Tafeln der Schriftstufe Uruk III. In: Arkhipov

Kogan

Koslova

(eds) The Third Millennium. Studies in Early Mesopotamia and Syria in Honor of Walter Sommerfeld and Manfred Krebernik (Cuneiform Monographs 50). Leiden: Brill , pp. 341–376.

43.

Krebernik

(2002) Zur Struktur und Geschichte des ältesten sumerischen Onomastikons. In: Streck

Weninger

(eds) Altorientalische und semitische Onomastik. (Alter Orient und Altes Testament 296). Münster: Ugarit, pp. 1–74.

44.

Krebernik

(2007) Zur Entwicklung des Sprachbewusstseins im Alten Orient. In: Wilcke

(ed.) Das geistige Erfassen der Welt im Alten Orient. Wiesbaden: Harrassowitz , pp. 39–61.

45.

Krebernik

(2013) Die frühe Keilschrift und ihr Verhältnis zur Sprache. In: Crüsemann

Van Ess

Salje

Hilgert

(eds) Uruk: 5000 Jahre Megacity Begleitband zur Ausstellung ‘Uruk-5000 Jahre Megacity’ im Pergamonmuseum-Staatliche Museen zu Berlin, in den Reiss-Engelhorn-Museen Mannheim. Petersberg Allemagne: Michael Imhof Verlag , pp. 187–193. [Engl. edn 2019, 167–173].

46.

Labat

(1988 [1948]) Manuel d’epigraphie Akkadienne. Paris: P. Geuthner .

47.

Marchesi

Marchetti

(2011) Royal Statuary of Early Dynastic Mesopotamia. Winona Lake, Indiana: Eisenbrauns .

48.

Michalowski

(1993) Tokenism. American Anthropologist 95: pp. 996–999.

(1993) Archaic Bookkeeping. Chicago: University of Chicago Press . [Engl. edn of Nissen HJ, Damerow P and Englund RK (1990) Frühe Schrift und Techniken der Wirtschaftsverwaltung im Alten Orient. Bad Salzdetfurth: Franzbecker].

50.

Parpola

(2016, 2022). Etymological Dictionary of the Sumerian Language. Part 1 Lexical evidence, Part 2 Semantic analysis and indices, Part 3 Linguistic analysis. Addenda and corrigenda. (Publications of the Foundation for Finnish Assyriological Research 16/1–3). University Park, PA: Pennsylvania State University Press .

51.

Schmandt-Besserat

(1992) Before Writing. Austin TX: University of Texas Press .

52.

Schmandt-Besserat

(1996) How Writing Came About. Austin TX: University of Texas Press .

53.

Selz

(2018) A Mesopotamian path to abstraction? On Sumerian ‘ontologies’. In: Rollinger R (ed.) Conceptualizing Past, Present and Future (Melammu Symposia 9). Münster: Ugarit Verlag , pp. 409–433.

54.

Selz

(2020a) The Uruk phenomenon. In: Radner

Moeller

Potts

(eds) The Oxford History of the Ancient Near East Volume 1. From the Beginnings to Old Kingdom Egypt and the Dynasty of Akkad. New York: Oxford University Press , pp. 163–244.

55.

Selz

(2020b) Sumerian compound nouns, typological considerations. In: Benvenuto

et al. (eds) Word Formation, Grammar and Lexicology: Contrastive, Multilingual and Comparative-Historical Perspectives (Indogermanische Textlinguistik, Poetik und Stilistik 4). Hamburg : Baar Verlag, pp. 319–347.

56.

Selz

(2021a) The puzzling Logogram: Writing and reasoning in Early Mesopotamia. In: Gabriel

Overmann

Payne

(eds) Signs – Sounds – Semantics: Nature and Transformation of Writing Systems in the Ancient Near East. (Wiener Offene Orientalisitk 13). Münster: Ugarit Verlag , pp. 27–47.

57.

Selz

(2021b) Appositive semantic classification in Sumerian cuneiform and the implementation of iClassifier. Ash-Sharq 5: pp. 142–171.

58.

Selz

(2022) Beyond speech: Advocating a non-logocentric view on the evolution of cuneiform writing. In: Wengrow

(ed.) Image, Thought, and the Making of Social Worlds (Freiburger Studien zur Archäologie & Visuellen Kultur 3). Heidelberg: Propylaeum , pp. 213–249.

59.

Selz

(in press a [2015]) The World Behind the Words: Reflections on Concepts Crossing Linguistic Boundaries. Pre-print [2015] at academia.edu.

60.

Selz

(in press b) Sumerian word formation: The issue of Sumerian nominal compounds. In: Gundacker

(ed.) Componendo composito. Lingua Aegyptia Studia Monographica. Hamburg: Widmaier Verlag .

61.

Selz

(in preparation) How meaning evolves – Proto-cuneiform signs: considerations on their origins at the intersection of visual and oral communication.

(2018) The question of Sumerian ‘Determinatives’: Inventory, classifier analysis, and comparison to Egyptian classifiers from the linguistic perspective of noun classification. Lingua Aegyptia. Journal of Egyptian Language 25: pp. 281–344.

63.

Schretter

(2000) Zu den Nominalkomposita des Sumerischen. In: Graziani S (ed.) Studi sul Vicino Oriente Antico dedicati alla memoria di Luigi Cagni. Vol. 2. (Istituto Universitario Orientale. Dipartimento di Studi Asiatici. Series minor 61). Napoli: Istituto Universitario Orientale , pp. 933–952.

64.

Steinkeller

(1995) Review of Green & Nissen 1987. Bibliotheca Orientalis 52: pp. 689–713.

65.

Veldhuis

(2006) How did they learn cuneiform? In: Michalowski

Veldhuis

(eds) Approaches to Sumerian Literature: Studies in Honour of Stip (H. L. J. Vanstiphout). Leiden: Brill , pp. 181–200.

66.

Veldhuis

(2014) History of the Cuneiform Lexical Tradition. (Guides to the Mesopotamian Textual Record, Vol. 6). München: Ugarit-Verlag .

67.

Wagensonner

(2009) Non-textuality in the Ancient Near East: Some glimpses from the Mesopotamian field. In: Andrássy

Budka

Kammerzell

(eds) Non-Textual Marking Systems, Writing and Pseudo-Script from Prehistory to Modern Times. Göttingen: Seminar für Ägyptologie und Koptologie , pp. 33–67.

68.

Wagensonner

(2010) Early lexical lists revisited: Structures and classification as a mnemonic device. In: Kogan

(ed.) Language in the Ancient Near East. Winona Lake IN: Eisenbrauns , pp. 285–310.

69.

Wagensonner

(2016) Die frühen lexikalischen Texte und ihr Aufbau: zu den Archaischen und frühdynastischen Wortlisten, der Anordnung ihrer Einträge und den Klassifikationssystemen in den frühen Phasen der Keilschrift. Unpublished PhD dissertation, University of Vienna, Austria.

70.

Wagensonner

(2021) Classifiers between Euphrates and Tigris. On development and use of noun categorization in cuneiform script. In: Gabriel

Overmann

Payne

(eds) Signs – Sounds – Semantics: Nature and Transformation of Writing Systems in the Ancient Near East (Wiener Offene Orientalisitk 13). Münster: Ugarit Verlag , pp. 171–212.

71.

Woods

(ed.) (2010) Visible Language: Inventions of Writing in the Ancient Middle East and Beyond. (Oriental Institute Publication 32). Chicago: The Oriental Institute .

72.

ZATU = Green

Nissen

(1987) Zeichenliste der archaischen Texte aus Uruk. Berlin: Gebr. Mann .

Classification in Sumerian cuneiform and the implementation of iClassifier

Abstract

Keywords

Introduction

The beginnings of semantic classification in the earliest texts from Uruk

Determinatives/classificators/classifiers: The situation in ‘classical’ Sumerian texts

The Selz consolidated list of cuneiform classifiers

The iClassifier project and its application to Sumerian texts (preliminary report)

The starting point

Data input – current state

Classifier list by iClassifier

Classifier axis: Classifier reports on a single classifier: The [hide] classifier

Lemma axis: All classifiers attested with a specific lemma: esir ‘sandal’

The classified Cuneiform World NETWORK MAP

Conclusions and perspectives

Footnotes

Acknowledgements

Authors’ contributions

Declaration of conflicting interests

Funding

ORCID iD

Notes

References