Introduction: Graphemic classifiers in complex script systems

Abstract

This article explores the role of unpronounced semantic classifiers, also known as graphemic classifiers or determinatives, in three ancient complex scripts: Egyptian, Chinese and Sumerian. These classifiers are silent hieroglyphs, Chinese characters or cuneiform signs that are combined with other signs that carry phonetic information to form a complete written representation of a word. While these classifiers are written and visible, they are not pronounced. They add silent, motivated semantic information related to the meaning of the word. These classifiers can be found in various positions within words, reflecting cultural and referential information. Classifier studies, in general, have gained significant interest at the intersection of linguistic typology, cognitive linguistics, semiotics of scripts and neuroscience. The research field examines classifiers in oral languages, signed languages and complex scripts, emphasizing that regardless of modality they reflect a shared cognitive effort to organize knowledge. It is our hope that the scholarly contributions in this issue will open up a new chapter in classifier studies and in comparative script analysis. Theoretical and analytical work undertaken in the last few decades has been done primarily by individual researchers specializing in one language or script. Our approach combines large-scale corpus data with comparative script analysis carried out by teams of collaborators who can contribute specialized expertise in different ancient writing systems. The research possibilities opened by our newly developed digital tool iClassifier are presented in detail in the other contributions in this issue. This work has laid a strong comparative foundation that we can now build on, to develop new insights into the early history of script development and the commonalities and differences among ancient cultural conceptualizations of the world.

This short contribution serves as an introduction to a special issue of Journal of Chinese Writing Systems that deals with a unique script function, the semantic classifier, shared by complex scripts of the past and present (e.g., ancient Chinese scripts, ancient Egyptian scripts, cuneiform scripts, Anatolian hieroglyphs, Mayan glyphs, modern Chinese scripts, Japanese scripts). Signs in complex scripts may have three basic functions: logograms [sound + meaning], phonograms [sound only] and (semantic) classifiers [meaning, but no sound]. Figure 1 illustrates the main different functions a single sign may take in complex scripts. The examples are from ancient Egyptian and ancient Chinese. ¹

Figure 1.

The signs “house” and “tree” in three possible functions. When functioning as semantic classifiers they are unpronounced. All signs in classifier function are colored in amber throughout the article. The signs [legs/motion] and [water] (in the second row) also function as classifiers for the words "go out" in Egyptian and "wash" in Chinese.

In this volume, we deal with ancient Egyptian hieroglyphs, Sumerian cuneiform and ancient Chinese characters, hoping to expand the scope to other scripts in the future. Ancient Egypt, Ancient Sumer and Ancient China are three great civilizations of the past, removed from one another both geographically and culturally. Nevertheless, the scripts they used are remarkably similar in key semiotic respects, despite their (apparently) independent origins and differences in form.

As mentioned above, this issue concentrates on the unpronounced semantic classifiers (graphemic classifiers) in scripts. This phenomenon is known in Egyptology and Assyriology as “determinatives” or “classifiers” and in Chinese studies as “semantic components,” “semantic determinatives,” or “semantic classifiers” (Handel, 2015; Polis, 2018; Selz et al., 2017). We use the term “classifier” as a comparative concept, referring to the signs that play this role in the respective scripts. Classifiers are unpronounced hieroglyphs, Chinese characters or cuneiform signs that combine with other signs that present phonetic information (the phonetic part) to represent a word altogether. The signs that perform the function of “classifier” are written and seen, yet they are not heard. They add silent, motivated semantic information for the reader, always related to the host word at the time of character/word creation (Selz et al., 2017, Egyptian and Sumerian scripts; Boltz, 1994; Chen, 2016; Handel, 2023 and Xu in this issue, Chinese scripts). For example, the word “younger sister” is written in ancient Chinese by the compound graph (mèi, Old Chinese *mə̂, Yin Zhou Jinwen Jicheng (Compendium of Yin and Zhou Bronze Inscriptions) No. 4589_Songgongluanfu). The right-side component 未 (wèi, Old Chinese *məs, “not yet”) is the phonetic element. ² The left-side component is an unpronounced semantic classifier adding to the phonetic part the silent information [woman]. The word mwt “mother” in Egyptian, also shows as phonetic part mwt , and an unpronounced semantic classifier [woman] at the end of the word. In Sumerian, we may compare the word šag4-zu “midwife,” where the phonetic part šag4-zu follows the unpronounced semantic classifier [woman]. Graphemic classifiers can be found in pre-position and post-position (Sumerian), post-position (Egyptian) or interchanging positions in Chinese scripts (see recently Myers, 2019).

These three complex scripts reflect through the silent semantic classifiers that younger sisters, mothers and midwives in the respective cultures are prototypically women.

Classifier studies in general

Classifier studies in languages and scripts is a relatively new research field in linguistics and cognitive linguistics. Presently, the field enjoys considerable interest. It is located at the crossroads of linguistic typology (Senft, 2000), cognitive linguistics (Lakoff, 1987), semiotics of scripts (Polis and Rosmorduc, 2015) and lately neurosciences (Kemmerer, 2016, 2017; Goldwasser, forthcoming). Classification may surface in various linguistic manifestations: in the lexicon, grammar or script (Bauer, 2017; Bisang, 2017; Grinevald, 2015). Classifiers, which are part of this general human practice of classification, appear in oral language systems as pronounced morphemes (for example, the numeral classifiers found in many languages, including Chinese), in signed languages as gestures (Brentari, 2023; Grinevald, 2004; Lincke and Kutscher, 2012), and in complex scripts as unpronounced graphemes (see Chen, 2016; Handel, 2023; Payne, 2017; Selz et al., 2017, Anatolian hieroglyphs). Whatever the linguistic modality in which they are present, they manifest the same cognitive effort to organize the world. Early on, classifier studies were concerned only with spoken (pronounced) classifiers, a widespread phenomenon in world languages (Aikhenvald, 2000; Grinevald, 2004 ³ ) that is well attested in modern spoken Chinese (Sybesma, 2015; Zhang, 2013). Denny (1976) was the first to promote in contemporary linguistics the classifiers’ semantic value in his influential article, “What are noun classifiers good for?” Craig’s seminal edited book Noun Classes and Categorization (1986) and Lakoff’s Women, Fire, and Dangerous Things (1987) established classifier studies as a discipline. It was in Craig’s book that the linguist Noel Rude (1986) suggested for the first time that “determinatives” in cuneiform and hieroglyphic scripts are also classifiers. Later, the role of classifiers as “referent tracking” devices was established by Croft (1994). The first two decades of the 20th and the beginning of the 21st century saw lively discussions of pronounced classifiers in various languages and theoretical analyses of classifiers’ sorts in different cultures and languages (e.g., Aikhenvald, 2000, 2021; Allassonnière-Tang et al., 2021; Bauer, 2017; Bisang, 2017; Bisang and Wu, 2018; Contini-Morava and Kilarski, 2013; Grinevald, 2004, 2015; Senft, 2000), as well as some assessments of their cognitive role across the world’s languages (Imai and Saalbach, 2010; Saalbach and Imai, 2007).

Classifier research in Egyptology

The study of the graphemic classifiers has been an established research topic in Egyptology only during the past two decades (for earlier research, see Goldwasser, 1995, 1999; Kammerzell, 1999). These classifiers—traditionally called determinatives—were known since the days of Champollion, who identified their role in the script and allotted them special attention in his Grammaire égyptienne (1836). Interestingly, Champollion studied Chinese, exchanged letters with the Sinologist Abel-Rémusat, and adopted the term “determinative” from the famous early French Sinologist de Guignes (Lefebvre, 1955: 18, note 6; Madrigal, 2022: 53–54). In the decades following Champollion’s decipherment, Egyptology regarded the determinatives as a mere scribal phenomenon born out of the need to disambiguate similar written forms and signal word boundaries. The motivated semantic information they carry was acknowledged but mostly discussed only when relevant to a specific lexical meaning (with some exceptions, see Wiesmann, 1922; Te Velde, 1977). However, they were never treated as a linguistic system. In her 2002 book Prophets, Lovers and Giraffes: Wor(l)d Classification in Ancient Egypt, Goldwasser suggested seeing the determinatives as a rule-governed graphemic classifier system (Goldwasser, 2002, 2006). Since then, more than a few case studies were carried out on different classifiers (e.g., Allon, 2007; Chantrain, 2014, 2021; David, 2000; Goldwasser, 2010; Lincke, 2011; Shalomi-Hen, 2000, 2006, 2008; Thuault, 2020 ⁴ ). In the last decade, determinatives were established as a system of unpronounced graphemic classifiers parallel in many ways in their use to pronounced classifiers in spoken languages. (Chen, 2016; Goldwasser, 2022; Goldwasser and Grinevald, 2012; Handel, 2023; Kammerzell, 2015; Lincke and Kammerzell, 2012; Selz et al., 2017). Various other issues in the Egyptian classifier system were discussed in Werning (2011), Lincke (2011), Grossman and Polis (2012), Polis and Winand (2015), Polis and Rosmorduc (2015), Kammerzell (2015), Pommerening (2017), Pommerening and Bisang (2017), Polis (2018), Winand (2019) and Winand and Stella (2013). The graphemic system of Egyptian classifiers was compared to classifiers in sign language (Lincke and Kutscher, 2012).

Classifier research in Sumerology

Clearly influenced by Champollion’s description of “determinatives,” Edward Hincks introduced in 1847 “determinatives” to designate unpronounced signs—semantic classifiers—for texts written in cuneiform (Selz et al., 2017). Sumerian, the language of the earliest Mesopotamian documents, shows many such classifiers, even while their precise number is still disputed. Based on earlier accounts, Selz compiled a new list of semantic classifiers (Selz, 2021; Selz et al., 2017; Selz and Zhang, 2024). Selz described the semiotic role the semantic classifiers play in Sumerian when compared to Egyptian. He also advocates for the Sumerian system’s possible pronounced origin, which leans strongly on various compounding processes (Selz, 2021).

Classifier research in Sinology

The modern Chinese spoken languages (of which Mandarin is the most thoroughly analyzed) include a developed system of pronounced numeral classifiers (Sybesma, 2015; Zhang, 2013). These classifiers are not part of our discussion here. Graphemic classifiers have been present in the Chinese script from its earliest attested uses, while the spoken language did not become a classifier language until the transition between the Old Chinese and Middle Chinese stages—more than a millennium after the invention of the script (Peyraube, 1998).

In the Chinese script, the long-recognized components termed “bùshǒu” ⁵ were employed as a grouping mechanism in the character dictionary Shuō wén jiě zì 說文解字, compiled by Xu Shen (Xŭ Shèn 許慎, c. 58–147 ce). Xu Shen called them bùshǒu 部首, literally “group heads,” because each was placed at the start of the section of characters containing it. Xu Shen explicitly identified all characters listed under a bùshǒu as “derived from” it, that is, having some kind of semantic relationship to it. These bùshǒu were termed classifiers by Boltz (1994: 147). Boltz (1994, 1996) presented examples of their diachronic development from the earliest oracle bone inscriptions up to modern Chinese and explicitly drew parallels to their use in other complex script systems such as cuneiform, Mayan hieroglyphs, Egyptian hieroglyphs, and so forth. ⁶ More recently, Handel (2023) studied the graphemic classifiers in the modern Chinese script, referring to their functional resemblance to Egyptian graphemic classifiers.

Because of their classificatory function in his dictionary, we can refer to Xu Shen’s bùshǒu as “class heads” or lexicographic classifiers in English. However, not all of Xu Shen’s 540 bùshǒu are considered by all modern scholars to have semantic function when they occur as a component in the characters listed under them.

Li (1996) writes that only 374 bùshǒu could be analyzed in a function similar to semantic classifiers (yìfú 義符). According to him, the rest of the bùshǒu are activated as logograms and phonograms, and so forth. Despite the fact that Xu Shen’s bùshǒu are not the same as the set of semantic classifiers that our contemporary analysis reveals in the script, the 540 bùshǒu of Shuō wén jiě zì remain the starting point of analysis of semantic classifiers for many modern scholars. Xiang (1987) suggested in a pioneering short article that semantic components (i.e., semantic classifiers) within a Chinese character could stand in four kinds of relationship to the meaning of the whole character’s host word. These four kinds of relationship can be compared to Goldwasser’s later analysis of classifiers in the hieroglyphic script (2002: 15–18). Xiang identified the relationships: biāolèi xíngfú 標類形符 (parallel to taxonomic classifier), xiāngguān xíngfú 相關形符 (parallel to metonymic classifier), bǐkuàng xíngfú 比況形符 (parallel to metaphorical classifier) and biāoyì xíngfú 標義形符 (somewhat like a repeater classifier). ⁷

In her book, F Chen (2006) conducted pilot research on 38 semantic components (classifiers) from the perspective of structural semantics and cognitive semantics. She also discussed prototype theory and cognitive models such as metonymic and metaphoric models by giving examples from Shuō wén jiě zì.

In the last decade, YS Chen (2016) offers in an article the first comparative analysis of Chinese “semantic determinatives” and Egyptian classifiers/determinatives using the tools developed by Goldwasser. Based on examples taken from Xu Shen’s dictionary Shuō wén jiě zì, he highlights the semiotic similarity of the Chinese and Egyptian semantic classifiers. He compares a few object categories of the two scripts: [water body (canal)] and [water body], [tree] ⁸ and [grain], [precious metal] and [precious stone]. He continues into the event categories and [force/effort], [movement] and [movement/motion]. He also signals classifiers that tend to stand in metomymic relationship to the meaning of their host words such as the classifier [fire/hot] in Chinese and the classifier [knife/sharp] in Egyptian. He ends his innovative article with an interesting analysis of the character “woman” that when used as a classifier not only can mean [woman] but may also embrace a category of [beauty & grace and bad quality].

Wang (2015) takes a corpus linguistic approach when he examines the repertoire of words that take the [negative] classifier in a single long literary text from the Middle Kingdom. He presents a comparison to the Chinese category [dog] (this time not from a single text) that presents similar negative evaluation. In all cases above, in Egyptian and Chinese, we see categories that present transparent motivated semantic relations between the host word and the semantic classifier.

Some comparative facts on the graphemic classifier systems in Egyptian, 
Sumerian and Chinese

Egyptian

Type of host : The classifiers classify nouns, pronouns, verbs, deverbals, adverbs and compounds.

Position : Always in post-position.

Multi-classification (a single word may take more than one classifier simultaneously): Widespread, 1–5(!) classifiers.

Classifier alternation (substituting classifiers or zero classification in the same word): Possible, common, may depend on materiality considerations, script variation or idiosyncratic choices—possible in hieroglyphs (pictorial version of the script) or hieratic (cursive version of the script).

Number of classifiers in the system : The Egyptian classifier system shows a clear tendency to increase in number from the Old Kingdom to the Middle Kingdom and the 18th Dynasty (c. 2600–1300 bce). Gardiner (1957: 32–35) compiled a list of “generic classifiers” that shows, during the classical period (Middle Kingdom and 18th Dynasty), approximately 100 classifiers that classify more than three words (Winand and Stella, 2013: 129; Goldwasser and Soler, 2024). Gardiner’s list is based on original Middle Kingdom and 18th Dynasty sources. The later Ramesside Period (c. 1300–1100 bce) seems to show a decrease in the number of classifiers in the cursive hieratic scripts (Chantrain, 2014; Kammerzell, 2015). However, the lapidary hieroglyphic script during the Ramesside Period (especially in historical texts) shows the opposite tendency: the creation of new original classifiers. Demotic, the new cursive script (based on hieratic) developed in Egypt from the 7th century bce on, shows a reduction in classifier number and a more restricted use. When the Egyptians started to write their language in the new script called Coptic (from 1st century ce)—an alphabetic script based on the Greek alphabet—the classifier phenomenon vanished like a fata morgana, leaving no trace in the new alphabetic script system (Grossman and Richter, 2015).

Type of source used for classifier studies : Original sources. Research has been conducted on certain classifiers and on certain words, texts and corpora.

Sumerian

Type of host : The classifiers classify nouns only, including compounds. The system is a strict noun classifier system, and strongly related to compounding processes.

Position : Most classifiers are in pre-position, a few are in post-position.

Multi-classification : Rare, though possible, with up to two classifiers.

Classifier alternation in a single word : Possible, common.

Number of classifiers in the system : The Selz Consolidated List (see Selz et al., 2017) contains 52 classifiers. For a new discussion see Selz and Zhang, 2024.

Type of source used for classifier studies : Original sources, from the Digital Dictionary ePSD. The dictionary aims to present all occurrences in their contexts, marking what the compilers considered as semantic classifiers in upper case transcription (see discussion in Selz, 2021).

Chinese

Type of host : All words.

Position : Position varies according to the particular classifier and the tendency to arrange components into an overall balanced, rectangular shape. Most classifiers appear at the top or left side of characters, but some classifiers typically appear at the bottom, the right or surrounding other components.

Multi-classification : Rare. Sometimes characters appear to have more than one classifier, but this is usually the result of iterative compounding. The embedded classifiers are non-functional at the top level (see discussion in Xu, 2024).

Classifier alternation in the written form of a single word : Chinese writing does not allow the separation of the phonetic part and the semantic part. Each compound character is regarded as an inseparable unit, even though it is clearly recognized as composed of two separate components that are different semiotically. However, during the Warring States Period (475–221 bce), alternation was common. The same word might be written with a character having no classifier at all (only a phonetic part) or alternating semantic classifiers (e.g., Xu, 2024).

Number of classifiers : The dictionary Shuō wén jiě zì compiled by Xu Shen c. 100 ce arranges characters under 540 lexicographic components, which were understood by Xu Shen as taxonomic components (Bottéro and Harbsmeier, 2016). Modern scholars do not necessarily agree that all of these components were functioning as semantic classifiers (see above). Texts of the Warring States Period show around 100 characters (in the studied corpus alone, see Xu, 2024) that function as semantic classifiers in compound characters. Modern Chinese dictionaries have around 200 lexicographic components, but not all are semantic classifiers and relatively few of them are still used when creating new characters (Handel, 2023).

Type of source used for classifier studies : Discussions of Xu Shen’s Shuō wén jiě zì and original sources. Limited corpus-based research has been conducted on certain semantic components, words, texts, corpora or specific genres (see Xu, 2024).

Chinese and Egyptian—Three comparative notes on sign function 
and classification

There are many different typologies of Chinese character structure, most of which take as their starting point the six-type classification in Shuō wén jiě zì (for an example of an innovative typology, see Behr, 2010).

The character types (or word types in Egyptian) relevant to our discussion are those of the compound character/word type. We discuss below two types of compound characters in Chinese:

1. Semantic-Semantic (SS) compound graphs (Logogram + Logogram)

In Semantic-Semantic (SS) characters (traditionally termed huìyì 會意), the two components are activated semantically to represent a third, different meaning (Handel, 2016).

 An example of a SS character in Chinese ⁹ is:

 安 (ān, Old Chinese *ʔân). The Shuō wén jiě zì entry reads: “‘to be stable’. From [semantic component] 女 ‘woman’ and [semantic component] 宀 ‘house’.”

 While Xu Shen and his contemporaries recognized SS compounds as an important category of Chinese character types, the historical existence of this type in the early development of the script is a matter of debate (Boltz, 1994). Even if one rejects Boltz’s argument that the graph 安 contains a “crypto-phonetic” component, that is, a component whose original phonetic function was no longer apparent to Xu Shen and remains invisible to us, from the synchronic perspective of Xu Shen this is indeed an SS compound.

 On the other hand, even if not entertaining the crypto-phonetic hypothesis, one might still argue that the above graph 安, looked upon from a diachronic perspective, did not originally combine two distinct graphs for their individual semantic functions. It could be viewed as a pictographic representation of a single scene (a woman inside a house) whose elements only later became graphically atomized as the script developed.

 A somewhat similar, though rare, phenomenon can be observed in some cases in the very early Egyptian script. For example, the word nbi “swim” is written with the pictorial logogram , and in the earliest examples from the Archaic Period, the man clearly swims in the water (Kahl, 1994: 430, note 64). Somewhat later, we see the tendency to extract the image of the man and he may be shown above the water (after Regulski, 2010: 336–337), creating a kind of an Egyptian SS: “man” + “water pool” = “swim.” In the Pyramid Texts (a few hundred years later), the logogram of the swimming man in the water is kept, but it functions already as a repeater classifier for a phonetic part that precedes it. However, when we meet安, or and later , the components behave as graphs in the script system, bound to the rules of script and not picture, by relative size, position, use, and so forth (Goldwasser, 2016; Vernus, 2020).

 As mentioned above, Xu Shen did not distinguish between semantic classifiers as we understand the term and the bùshǒu 部首 under which he grouped graphs as the primary organizing principle of his dictionary. Xu Shen understood 宀 “house” to be a semantic component in 安, and he accordingly placed the character 安 in the 宀 section of the dictionary, even though the “house” element does not classify the word “to be stable” in a taxonomic sense. Unlike in the SP compounds described below, it cannot be viewed as a theoretically superfluous component whose removal from the graph would not prevent a reader from understanding the word represented by the graph through its phonetic reference alone. The SS combinations are very rare in the Egyptian script system (see below) and somewhat more common (though still low in number) in the Chinese script (Handel, 2016).

2. Semantic-Phonetic (SP) compound graphs (Semantic Classifier + Phonogram)

This is the most commonly occurring type of graph in the Chinese writing system of all periods. An example of a Semantic-Phonetic (SP) graph is , writing the word meaning “guest”:

  (kè, Old Chinese *khrâk). The Shuō wén jiě zì entry reads: “‘to entrust’. From [semantic component] ‘house’; 各 (gè, Old Chinese *krak, ‘go to’) is the [phonetic component].”

 Like the example given earlier of an SS compound 安, Xu Shen places this character in the 宀 section of the dictionary. In this case, however, we can understand the “house” element as a semantic classifier. Generally speaking, the S part of any SP graph functions as a classifier in this way. The difference with an SS compound like 安 can also be seen in the fact that the phonetic component 各 could theoretically, consistent with the logic of the script, be used without the classifier to represent the word “guest.” ¹⁰ Words written by a phonetic component alone (termed “phonetic loan,” “loaned phonogram,” or “rebus” characters) are well known in ancient Chinese scripts (see Xu, 2024).

 In Egyptian, SP (or more accurately PS, as the classifier is always in the post-position) constructions are very common. Such an example is the word mentioned above. The two first hieroglyphs (reading from left to right) contain the full phonetic information mwt. The is not pronounced and classifies the word into the super category [woman].

3. Semantic-(Semantic + phonetic) S(Sp) graphs

Although not one of the basic types of graphs identified by Xu Shen, a significant number of compound graphs could be analyzed as having properties consistent with both the SS and SP categories. Xu Shen sometimes identifies characters as simultaneously belonging to both types. In the S(Sp) character, the S component is generally an unpronounced classifier. The Sp component is labeled by Xu Shen as a semantic component that is “also a phonetic.” For example, the word ān (Old Chinese *ʔân) meaning “saddle” is written with compound graph , analyzed in Shuō wén jiě zì as “from [semantic component] ‘leather’ (classifier) and from [semantic component] ‘to be stable’; 安 (ān, Old Chinese *ʔân) is also the phonetic [component].” The character is listed under the [leather] bùshǒu. As we have seen above, 安 is itself a composite SS character, carrying the phonetic value ān, functioning here as the phonetic part of the composite . The 宀 “house” does not function here as a classifier on the top level of the word.

 From our modern perspective, we can see that S(Sp) graphs represent words that are etymologically related. The Old Chinese word for “saddle” is derived from the word for “to be stable.” For this reason, the word for “saddle” has both semantic and phonetic similarity to the word for “to be stable.”

 It is worth pointing out that in an SP graph, one of the components may itself be an SS, SP, or S(Sp) graph in origin. This is because once a compound graph is conventionally associated with the spoken word it represents, it becomes available for use as a phonetic or (rarely) semantic component without regard to its internal structure.

 A rare example of a somewhat similar compound S(Sp) or more precisely (Sp)S graph in Egyptian ¹¹ could be ini “bring,” built of the compounding of two logograms nw “vessel” and iw “come.” The compounded hieroglyph creates a kind of SS compound that provides the new meaning “bring” = “vessel” + “come” (see the definition “pictorially meaningful,” Stauder, 2020: 882). In , the hieroglyph nw is a logogram that carries at once the semantic information “vessel” and the phonetic information nw. As the Egyptian script system presents mostly consonants, when enters the compounded hieroglyph with the meaning “bring,” it also keeps its phonetic meaning (phonogram), additionally providing the consonant “n” for “ini.” Thus may be perceived as a sort of Egyptian (Sp)S compound graph, in which the basic meaning is created by the semantic combination of the two logograms “vessel” +  “come.” Yet, is a kind of Sp, as phonetic information is also coded in it. Due to the high iconicity of the Egyptian script, the pictorial combination creates a new, slightly surrealistic, image of an “arriving vessel” that seems to embody in the pictorial the meaning of the hieroglyphic sign, including the semantic property “movement” which is included in “bring.” A full iconic realization of the hieroglyph is offered by the Egyptians themselves. In the Late Period, almost 3000 years later, we find the rare variation for the same word, where the more concrete image clearly shows that the vessel is the prototypical item that is brought (Goldwasser, 1995: 23; Goldwasser and Grinevald, 2012: 39. For the earliest occurrences of , see Regulski, 2010: 207, 702–703).

 Another sort of S(Sp), or more precisely, what could be described as (Ps)S because of the post-position of the classifier in Egyptian, is the word “fowler,” which is made of two pictorial graphs that are not compounded. The word is read from left to right. The first hieroglyph “boat” is a phonogram carrying all the necessary phonological information , functioning in a parallel fashion to a Chinese phonetic component. The hieroglyph functions as the semantic component, disambiguating the possible other meanings of and classifying the fowler into the generic category [man]. Yet, the hieroglyph provides simultaneously additional semantic information, as the fowling action is done on a boat (see Figure 2), and thus we face here a sort of (Ps)S spelling of the word.

Figure 2.

A fowler in the Tomb of Nebamun, 18th Dynasty (image courtesy of British Museum, no. EA37977, © The Trustees of the British Museum).

Classifiers as maps of knowledge organization

Our central hypothesis is that each semantic classifier signals a “category head” in the knowledge organization in each culture (Lakoff, 1987; Goldwasser, 2002, 2022, forthcoming; Handel, 2023). Suppose we collect all words that are written using a particular classifier; this collection will provide us with members of a category in the ancient culture’s “collective mind,” reflecting the rich palimpsest of different conceptualizations during different periods and places. Preliminary studies conducted in Egyptian, Sumerian and Chinese in the past two decades have confirmed this hypothesis (e.g., Allon, 2007; Chen, 2016; Goldwasser, 2002, 2010, 2023; Selz et al., 2017; Stauder, 2020). Recently, Handel (2023) has undertaken a preliminary corpus-based model study on classifier categories in the modern Chinese script. Multi-classification and classifier alternation signal overlapping and interconnections between categories, as shown in the examples mapped in other articles in this issue.

Graphemic classifiers represent, abstractly, a set of conceptual categories. Their semantic values in the script are derived from a combination of cultural-cognitive classification of the world and the structural systematicity of their use in the inherited writing system. We focus on both the similarities and the differences in how cultures organize their universe according to the scripts’ respective classifier systems.

New research methods used in the study of semantic classifiers

In order to create “mind-maps” of cultures according to the classifier systems, and as the corpus of classified words in texts in Chinese, Egyptian and Sumerian amounts to millions (i.e., “big data”), we have built in the ArchaeoMind lab (https://archaeomind.huji.ac.il/) a digital research tool for the collection, marking and analyses of classifiers. We start from a defined corpus of ancient texts in each script and assemble our lemmas (and tokens) from them. After marking the classifiers by tildes, the words are filtered by our digital tool iClassifier (©Goldwasser/Harel/Nikolaev). ¹² iClassifier is designed for a context-sensitive analysis of classification in scripts and languages. It provides maps of categories, their interconnections and their members. It signals prototypical members, members intersections and categories inner structures (center vs. edge, etc.; see 2024 Harel et al.; Goldwasser and Soler; Selz and Zhang; Xu). Classifiers are analyzed in accordance with their semantic relations to their host word: taxonomic, schematic, or metaphoric. In the case of verb classifiers (only Chinese and Egyptian), semantic roles which are represented by the classifiers are identified (after Kammerzell, 2015: 1400). Some similarities in verb classification in Chinese and Egyptian can be observed. An example of a verb classifier in ancient Chinese script is [road + foot/movement], which appears in verbs of motion such as (fù, Old Chinese *buk, “return,” Guodian bamboo manuscripts_Taiyishengsui 2, 9). The iconic meaning of 辵 is “foot walking on a road.” This sign is in origin composed of two parts: 彳 “road” and 止 “foot.” The part above is 彳, which in turn is in origin an abbreviated form (left part) of the sign 行 “crossroads, intersecting roads” (Guodian bamboo manuscripts_Laozi C 9, 7). The part below is a pictogram of a foot, which was also used as the logogram 止 (zhǐ, Old Chinese *təʔ, “foot,” Guodian bamboo manuscripts_Liude 26, 7). The character 辵 is generally used as a semantic classifier in SP compounds rather than a logogram on its own (see Xu, 2024). In Egyptian, such a combination is known in the verb “go upwards” that may take the [road] classifier together with the [movement] classifier. In both scripts, the [road] specifies the prototypical semantic role location of the actions movement of “return” and “go upwards,” and the [leg/walking legs] represents the prototypical semantic role of tool and/or experiencer as well as classifying both actions into the category of [motion] (see Lincke and Kammerzell, 2012). The category is one of the largest (by number of members) categories in the Egyptian script. However, the largest categories in Egyptian and in ancient Chinese script are the two comparable categories [senses/emotions/speech] in Egyptian and the [heart/senses & emotions] category in Chinese (Goldwasser, 1995; Goldwasser and Soler, 2024; Winand and Stella, 2013: 129; Xu, 2024).

The research possibilities opened by iClassifier are presented in more detail in the other contributions in this issue. It is our hope that the scholarly contributions in this volume will help to open up a new chapter in classifier studies and in comparative script analysis. Theoretical and analytical work undertaken in the last few decades has been done primarily by individual researchers specializing in one language or script. Our approach combines large-scale corpus data with comparative script analysis carried out by teams of collaborators who can contribute specialized expertise in different ancient writing systems. This work has laid a strong comparative foundation that we can now build on, to develop new insights into the early history of script development and the commonalities and differences among ancient cultural conceptualizations of the world.