Semantic classifiers (determinatives) and categorization in the ancient Egyptian writing system: Rules,list of classifiers,and studies by iClassifier on the Story of Sinuhe

Abstract

In the last two decades, we have extensively explored the semantic classifiers in ancient Egyptian scripts, showing how they encode the world from two complementary perspectives: universal cognitive tendencies of classification along with Egyptian society's categorization of the world. Our central hypothesis is that each graphemic classifier in the Egyptian writing system heads a conceptual category. The assemblage of words classified by a particular classifier presents us with a dynamic map of an emic category in the mind of a culture. Classifiers in the Egyptian script allow us to trace central and marginal members of conceptual categories, spot interrelations and overlaps between categories, observe diachronic developments and changes, and discover incompatibility of categories. The number of classifier occurrences in complex script systems amounts to millions (i.e., big data). In the last decade, we developed in the ArchaeoMind Lab the digital tool iClassifier ¹ (© Goldwasser, Harel, and Nikolaev) designed to check our hypothesis on a large scope of texts from different periods and different scripts. We mark classifiers in imported digitalized corpora—in ancient Chinese, ancient Egyptian, and Sumerian cuneiform. This contribution introduces the reader to the rules of the classifier system in the Egyptian script and presents new results from a pilot project on a genuine Egyptian corpus (variants of the Story of Sinuhe).

Introduction

The topic of this article is the phenomenon of semantic classifiers (or determinatives) that make up an important part of the hieroglyphic writing system. The ancient Egyptian script is a complex writing system actively used for over 3000 years (Gardiner, 1957; Polis, 2023). It is made of signs that can take three basic semiotic functions—logogram, phonogram, and classifier. The term classifier refers to the specific function many hieroglyphs perform in the ancient Egyptian script. The same hieroglyph that functions as a classifier may perform other graphemic functions. When playing the role of classifier, the hieroglyph is unpronounced. Classifiers are seen but not heard. They refer to concepts and not to words in the lexicon. Egyptian classifiers always appear in the end position in a word or compound. The example in Table 1 illustrates the same hieroglyph in the three possible semiotic functions (all classifiers are colored in amber in this article).

Table 1.

Graphemic functions of Egyptian hieroglyphs. The sign

2	Logogram – pictorial and phonetic signifieds coincide	zꜣ “duck”
	Phonogram – phonetic signified only, pictorial signified discarded	zꜣ “son”[man]
	Classifier – only pictorial signified, phonetic signified discarded	bjk “falcon”[ bird ]

These three basic functions can be subdivided into more specific uses (Polis, 2018). Logograms include hieroglyphs of high iconicity ³ (e.g., “duck,” “crocodile”) and icons of lower iconicity (e.g., “house”). The same icon can function as a phonogram, as in the example (here the seated man is the classifier, classifying the “son” to the category [man]). When functioning as phonograms, uniconsonantal hieroglyphs (hieroglyphs that refer to single consonants) ⁴ are frequently used in the semiotic role of phonetic complements/interpretants. They offer redundant phonological information that leads the reader to the correct phonological path of reading (Kaplony, 1966; Goldwasser, forthcoming). For example, the word ḥqꜣ “ruler” may be written logographically by the hieroglyph , which is a triliteral sign. The icon represents the ruler metonymically by his prototypical scepter. However, the word could also be written ḥqꜣ, with two redundant, unpronounced uniconsonantal hieroglyphs, q and ꜣ, that are phonetic complements (Faulkner, 1962: 178). Biliteral and triliteral phonograms are more rarely activated in this role. ⁵

When functioning as a classifier, the hieroglyphic icon loses all phonetic realization and carries semantic information alone. In most cases, the classifier functions as a representative of a whole category, and as such, its meaning is extended considerably. Such is the case of “falcon” (e.g., CT ⁶ I 97), where the duck hieroglyph gets the extended semantic meaning [bird] and comes to represent the category to which the falcon belongs according to the Egyptians (Goldwasser 2002: 19–20).

Hieroglyphs that function as classifiers are always attached to a host lemma, compound, or root. Semantic classifiers always show a motivated relationship with their host. We think that the term classifier should replace the old term determinative coined by Champollion (1836) himself, the decipherer of the ancient Egyptian hieroglyphic script. ⁷

Efforts of classification and categorization are visible in all human activities. In language, classification surfaces at different linguistic levels: lexicon, grammar, and script. Classifiers are part of this effort. They appear in many oral language systems ⁸ (as pronounced morphemes), in signed languages (as gestures), and in complex scripts (as unpronounced graphemes). Whatever the linguistic medium in which they are present, they are manifestations of the same cognitive effort. Therefore, when appearing in writing systems, they are not simply representative of phenomena particular only to scripts.

Graphemic classifiers, despite the traditional argument, are not mere graphic devices activated to clear up script ambiguities or mark the boundary of a word. As will be demonstrated, referring to this function as classifier instead of determinative is not mere relabeling. Consequently, classifiers must be investigated from a typological perspective, as a system, open to comparison with other scripts and languages. Indeed, once Egyptian classifiers are put in the framework of modern classifier studies, we can better understand, explain, and even predict their behavior.

Almost all parts of speech in the Egyptian writing system could take a graphemic classifier. They appear as a rule after the host word. The classifier's appearance with a lexeme is not compulsory and may be subjected to scribal traditions, materiality, idiosyncratic choices, space limitations, and diachronic changes. Script variations (Hieroglyphs, Cursive [linear] Hieroglyphs, Hieratic, and Demotic) show different classification tendencies and different diachronic developments (e.g., Chantrain, 2014a, 2021). Finally, we must consider the unusual phenomenon of multi-classification in the Egyptian writing system: a single host may take one to four or even five classifiers and in another text not be classified at all, as Egyptian does not have a canonical spelling for words. ⁹

In the first section of this article, we briefly review some of the essential characteristics of the Egyptian graphemic classifier system, as far as space limitation allows us. First, we present the possible parts of speech that can function as hosts. Then, we exemplify the identification of most classifiers as prototypes of the conceptual categories they represent. Finally, semantic relations between a classifier and its host are presented, and alternating classification and multi-classification and their possible reasons are shortly discussed. Encyclopedic versus pragmatic classification and the emergence of semantic roles in verbal classification are also discussed. This section ends with a short mention of indoctrination by pictorial models in the script. The second part of the article is a pilot corpus linguistic study on the Story of Sinuhe, a prominent literary text of ancient Egypt. It is a data-based statistical effort to test our theory of classification. We imported the main versions of the text (courtesy of the Thesaurus Linguae Aegyptiae, see corpus credits) into our digital research tool iClassifier. We proceeded then to analyze the data with this digital tool. Some of the results are presented on network maps and tables. The last section presents the repertoire of classifiers that appears in Sinuhe in comparison to general lists of Middle Kingdom classifiers.

1 The uses of classifiers in the Egyptian script system

1.1 Classifier position and word class of host

Classifiers classify nouns, verbs, deverbals, adverbs, and even pronouns. When classifying a noun or a verbal root, classifiers appear as a rule before suffixes and grammatical markers, which follow the classifier.

Nouns: šrj ^[childhood] “male child.” In the female counterpart šrj.t “female child” the classifier still appears after the root in most cases before the female ending t. The uniconsonantal hieroglyph t is the female marker of nouns and comes at the end of the word. In the above example we see that the classifier is attached to the root before the . However, sometimes the classifier may also appear after the female marker t, e.g., . In this example classifies the female child, into the [childhood] category.

Verbs: The classifier usually comes after the root before suffixal pronouns, and before the past marker n, e.g., sḫr^[fall].n = f “he has overleapt” (anterior = 3sg.m), classifier [prostrated man], or šm^[motion].kwj “I went” (stative:1.sg), with the classifier [motion] after the root šm.

Deverbals: qd^{[wall- man]} “builder,” deverbal from the verb “build.” The classifier [wall] stands in schematic (metonymic) relations to the root “build” representing a prototypical patient of the act (Kammerzell 2015: 1400). The second classifier stands in taxonomic relation to the host, classifying “builder” into the [man] category. It is also a pictorial nominalizer, suggesting that this is a deverbal “builder” (Lincke and Kammerzell, 2012).

Adverbs: dj ^[road] “here,” classifier [road, distance].

Pronouns: Personal pronouns could often accept classifiers: jnk^[divine] “I” (1sg.m.^[divine]) ^{10 .} The classifier [god/divine] is a pragmatic classifier here (see below), signaling that the referent for the pronoun “I,” in this case, is a god or the divine king. Possessive pronouns may also show pragmatic classifiers, especially in the New Kingdom, e.g., pꜣy.j^[man] “my,” specifying the owner's gender. Demonstrative pronouns typically do not take a classifier. However, the demonstratives pfꜣ^[road], tfꜣ^[road], and nfꜣ^[road] that carry the meaning “that,” tend to take the [road, distance] classifier. These classified demonstratives may allude to the “distance” of the deictic term “that” compared to “this,” i.e., “distal versus proximal,” respectively.

1.2 Cultural power struggles within the script: How is a hieroglyph chosen to represent a whole category?

Due to the script's high iconicity, we can observe that the signs that function as classifiers in the script are mainly created from cultural prototypes. At a particular moment, the script draws on prototypical members of a particular category to stand as the pictorial “best example” of a certain word, i.e., to represent the iconically diverse referents or signifieds of the word in the real world (Goldwasser 1995: 32–33). For example, one chosen dog, the greyhound, plays the role of signifier in the script for all kinds of dogs (see Goldwasser, 2002: 91–108, 24*–25*). This dog was the breed that the king and the elite owned. Humanity is represented in the hieroglyphic script by a young, slim Egyptian man, as we can see clearly in a detailed example of the [man] classifier —a hieroglyph from a queen’s tomb in the New Kingdom. The correct “best exemplar” of [man] is Egyptian.

Another prototype that reflects its cultural background is the duck (see Table 1), which was originally the logogram for “duck.” It can also assume the role of classifier representing the generic category [bird] referring to all birds, including falcon and ostrich. The classifier extends later into an even more encompassing category [winged ones] to include members such as mosquitos, locusts, etc. Although not a prototypical “bird” to the modern viewer, the duck was an important bird in Egyptian culture. Egypt being the “gift of the Nile,” this common waterfowl was conceived as the most prominent type of bird. The duck was a “good bird” and probably promoted to its role of representative of all birds in the script because of its strong positive associations with offerings, food abundance, etc. (Goldwasser, 2002: 19–20; Goldwasser and Grinevald, 2012: 34). Birds with negative connotations (such as the sparrow that eats the seeds of grain in the fields) were, in all likelihood, unable to assume a similar status despite their abundance. Instead, the sparrow became the classifier of the generic category [negative] from the Middle Kingdom on (Kammerzell, 2015: 1407; for the prototypical birds in the Western civilization, see Lakoff 1987: 41–45; Aitchison 1992: 72).

A good example of the cultural pressures at work behind the choice of a prototype is the history of the all-important generic category [god/divine], which projected two main classifiers. The politically most powerful group in early Egyptian history was capable of establishing its god, “Horus on the standard” as the “god par excellence” and, therefore, as the [divine] classifier in the script. The Horus falcon was the god that was also established as the pictorial prototypical image of the divine king very early (on the importance of the falcon god, see Hornung, 1982: 39; Shalomi-Hen 2006: 13–38). A few centuries later, a new [divine] classifier made its appearance——a seated male human figure wearing a beard. The new classifier probably emerged from circles championing the power of the anthropomorphic god Osiris (Shalomi-Hen, 2006: 71–95). The script mirrors the struggle between the crucially different conceptualizations of the divine, the zoomorphic, and the anthropomorphic (Hornung 1982: 33–66; 100–109). This negotiation was never decided in the Egyptian culture, as the two classifiers continue to appear side by side during the whole life span of the script (Goldwasser, 2002: 19–24, and discussion below). Differently from Egyptian, in the Sumerian writing system, the classifier [divine] is a star signaling a stellar prototype (Selz et al., 2017: 290). In Chinese, it is a sacred religious item, 示 a “sacred tablet” that represents a similar notion (Xu, 2024, and our discussion in Section 2.3).

One important exception to the observation that classifiers in the Egyptian script always represent prototypes of generic categories is the widespread classifier [hide&tail]. Probably due to the extreme versatility of animals’ forms, a shared schematic quality was chosen to represent this major category—“having hide and tail,” i.e., quadrupeds and later also other animals. The hieroglyph depicts the hide and tail of a dead leopard. Interestingly, the Egyptian lexicon has no word parallel to the English “animal.” However, this covert category of the Egyptian mind surfaces in the script system (see detailed discussion in Goldwasser, 2002: 57–89; 2023b: 129–148 with Figure 5.5.).

1.3 Semantic relations between classifiers and their host words

Semantic classifiers may stand in taxonomic or schematic relations to host words (for terminology, see Goldwasser 2002: 25–38). In all cases, relations between classifier and host are motivated. For example, in ḫꜣr.t^[hair-woman] “widow,” the host word ḫꜣr.t is classified by the schematic classifier [hair], a central element in the ritual of mourning (Goldwasser 2023b: 126–127 with Figure 5.2.), and the (generic) taxonomic classifier [woman], as “widow” is “an example of” the category [woman]. Another example shows an alternative classification of the same word ḫꜣr.t^{[negative-woman]}. Here, instead of the [hair] classifier, the [negative] classifier appears. The classifier reflects, in this case, an emic social judgment in the culture. We know from Egyptian texts that widows were vulnerable and of low social status in ancient Egyptian society. Therefore, in the second spelling, the widow is classified into the taxonomic category [negative] and into the generic taxonomic category [woman].

1.4 Multi-classification, classifier order, and alternative classification

Cases of multi-classification (2–5 classifiers) appear mostly in verbs and deverbals (but also with nouns). In almost all cases, the order of classifiers is rule-governed (Goldwasser and Grinevald 2012: 33–35). As we have already seen in the examples above, classifiers that stand in schematic (metonymic) relation to the host word would precede taxonomic classifiers.

In the word mḥw^{[ship-fish-man]} “hunter,” the two first schematic classifiers refer to different components of the water hunt. The first is the vehicle, “ship,” and the second is the patient of the act, “fish.” The last taxonomic classifier informs us that the hunter is “a kind of” [man]. Verbs and deverbals could take up to five classifiers. An alluring example is the deverbal wḥꜣ.(w)^{[bird-fish-action-man]} “fowler.” ¹¹ In this example, wḥꜣ.(w) “fowler” shows four(!) unpronounced graphemic classifiers. The first two classifiers, [bird] and [fish], stand in schematic relationship to the host, being the patients of the fowling action. The last two classifiers, [action] and [man], stand in taxonomic relation to the host and represent superordinate, generic categories. In the case of deverbals, the generic category of the verb (in this case [action]), almost always precedes the category of the deverbal noun—here [man]. ¹² The fowler is “an example of” these two generic categories. This somewhat exaggerated list of classifiers for a single word keeps the order of schematic before taxonomic in the classification arena. In a recent article, Goldwasser (forthcoming) suggested that this “classifier excess” is related to priming practices that speed the reading process by presenting the most fitting categories/hubs in the brain to the reader.

1.5 Verb classification—exposing semantic roles

In the case of verbs, a variety of semantic roles may be exposed by classifier choice (Kammerzell, 2015). A typical example is the verb swr^{[water-senses&emotions]} “drink” that first presents the prototypical patient of the act of drinking, namely “water” (schematic relation, water being “part of” the action of drinking). The second classifier [senses&emotions] (iconically representing a man with his hand touching his mouth) refers to a major generic taxonomic category of the script “everything that comes in and out of the body container,” i.e., speech, food, emotions, thoughts, etc. (Goldwasser, 2005, and the discussion in Section 2.2.2). In another verb, znb^{[wall-motion]} “overleap,” the first schematic classifier [wall] shows a prototypical patient of the overleaping action. The final classifier shows the generic taxonomic category [motion], to which the verb belongs in the Egyptian script. Even though the classifier iconically represents horizontal movement, this extensive “event category” of the script also includes vertical movements, such as ꜥḥꜥ ^[motion] “to stand up.”

The verb nꜥ ^[boat] “sail” is classified by the [boat] classifier, the vehicle of the action. Interestingly, a word meaning “be without a boat” jwj^[boat] also takes the —the “missing vehicle” classifier. In Egypt, where the Nile is central, “boatless” means “helpless.” As a rule, lemmas that refer to deficiencies are put into the category of the “missing ability,” e.g., šp^[eye] “blind” is classified by the [eye] classifier (Goldwasser, 1995: 92–93).

1.6 Repeater classifiers

Repeater classifiers are a widespread phenomenon in the Egyptian script. They do not “classify” in the full sense of the English word. However, they are an integral part of all systems of classifier languages and classifier scripts worldwide (Senft, 2000). They represent a specific type of taxonomic classification. In Jakaltek, a Mayan language that shows a system of spoken noun classifiers, we find: ix ix = ^[woman]n:woman, which expresses simply “woman,” while ix Malin = ^[woman]pn:Maline, which tells us that the feminine proper name “Maline” belongs to the category [woman] (Goldwasser and Grinevald, 2012, Goldwasser and Handel, 2024; in Jakaltek classifiers are always in pre-position).

In Egyptian, the large taxonomic category [woman] includes nouns, deverbals, and proper names and pronouns referring to women (see the word “widow” above). However, in the word ḥm.t^[woman] “woman,” the hieroglyph is an example of a repeater classifier, repeating the information that a woman is part of the category [woman], as in the above example from Jakaltek. The classifier is also found with all female personal names, as in Jakaltek, activated as a generic taxonomic classifier, e.g., tꜣ-mj.t^[woman]= “(She)The cat”^[woman] (proper name). Repeater classifiers in the script may easily be exchanged with a generic taxonomic classifier, e.g., a repeater in ṯsm^[dog] “dog^” versus ṯsm ^[hide&tail] “dog.”

1.7 Encyclopedic versus pragmatic classification

In most cases, classifiers refer to encyclopedic knowledge. Still, we also find pragmatic classification referring to a particular referent in a specific text, which differs from the encyclopedic classifier. The word ꜥw.t means “livestock” and normally gets a prototypical animal as a classifier. However, in one particular context in the Coffin Texts, it shows the referential classifiers ^[people] (CT I 155). Here, “livestock” is temporarily moved by the classifiers to the category [men], evoking the metaphor humans are god's cattle, which is indeed spelled out lexically in Egyptian when referring to people as the jꜣw.t špss “noble livestock” in a Middle Kingdom literary text (Goldwasser, 1995: 93, n.59; 2005 for metaphoric classification).

1.8 Classifiers in a supplementary grammatical role

Classifiers add morphological transparency to the Egyptian script system and testify to an emic Egyptian “grammatical awareness.”

As we have seen above, human classifiers, e.g., or , appear as classifiers in deverbals, where the generic taxonomic category [man/woman] marks the word as a derivational form (see the discussion of qd “builder,” mḥw “hunter,” and wḥꜥ “fowler” above). In such cases, the human classifiers represent nominalization in addition to their role as markers of a generic taxonomic class. Classifiers may also mark the different semantic parts in a compound, e.g., Sbk-Rꜥ(w) the name of the god “Sobek-Re,” a combined god-crocodile and sun god. The crocodile god Sbk “Sobek” is classified by the prototypical [divine] classifier, the “falcon on the standard.” In the second part of the compound, the name of the sun god Rꜥ “Re” is written by phonograms and then classified by the repeater classifier [sun]. After the sun classifier appears a second anthropomorphic [divine] classifier in the form of the human god . In this example, the last classifier, the prototypical seated human god , may classify the second divine element “Re,” or better classify the whole compound, “Sobek-Re,” into the [divine] category. Here the prototypical human image refers to the abstract idea of the [divine], as neither the crocodile nor the sun could be depicted as humans.

1.9 Classifiers as tools of cultural indoctrination in hieroglyphic inscriptions

Classifiers are a rare mine of emic information in all cultures, mostly reflecting the “collective culturemes” of elite members of a specific culture. However, compared with other writing systems, they could also serve as pictorial indoctrination tools of the power-holders in ancient Egypt due to their very detailed degree of iconicity. For example, when the word ꜥꜣmw^{[foreign-captive] 13 “}Asiatic,” ¹⁴ here reproduced in the original hieroglyphs (read from right to left), gets the classifier of a bound Asiatic enemy , it a priori classifies all Asiatics, even those who come in peace to Egypt, as potential state enemies (Goldwasser, 2023a: 199).

2 A pilot corpus-linguistic study on a sample text by iClassifier: Classifiers in the Story of Sinuhe

This section of the article attempts to test to what extent the theoretical framework presented above manifests itself in a “real” text. Susana Soler marked and analyzed the classifiers in the imported corpus of texts (courtesy of TLA; see corpus credits). Due to space considerations, only a few issues and results are presented and discussed in the following.

2.1 The sample corpus: The Story of Sinuhe

The Story of Sinuhe ¹⁵ is a masterpiece of ancient Egyptian literature, highly appreciated by the Egyptians themselves. It was written, rewritten, and copied from the Middle Kingdom to the New Kingdom ¹⁶ (Parkinson, 1997, 2009; Winand, 2014). It has reached us in several preserved copies on papyri and ostraca. This textual transmission allows a study of variation and diachrony in the text, including in its classifier system. The versions ¹⁷ of Sinuhe used for our analysis are four Middle Kingdom (henceforth MK) copies: ¹⁸ Papyrus Berlin P 3022 and Papyrus Amherst fragments m-q (B), Papyrus Buenos Aires (BA), Papyrus UC 32773 (Harageh 1) from el-Fayum (H), Papyrus Berlin P 10499 from Thebes West (R); and six New Kingdom (henceforth NK) copies: Ostracon Senenmut 149 recto (TT 71) (S), Ostracon Ashmolean Museum 1945.40 from Deir el-Medina (AOS), Ostracon BM 5629 from Thebes West (L), Ostracon Borchardt (also known as Uni Zürich 1892) from Biban el-Muluk (Bdt), Papyrus Moscow 4657 (Golenischeff) (G), and Ostracon Cairo CG 25216 from the tomb of Sennedjem TT 1 (C). The whole corpus is written in the hieratic script, mainly used on papyri and ostraca in ancient Egypt. This script variety is a cursive variation of the pictorial hieroglyphs, showing a much lower level of iconicity. Papyri and ostraca, the most common supports of hieratic texts, enhance some particular classification tendencies. Yet, this script variation preserves all the semiotic and grammatical rules of the lapidary hieroglyphs. ¹⁹

Our current corpus encompasses 7733 tokens representing 1068 different lemmas. ²⁰ Approximately 45% of the tokens present semantic or pragmatic classifiers (excluding personal pronouns). ²¹ All results presented below were obtained by using the digital research platform iClassifier (© Goldwasser, Harel, Nikolaev). This tool enables in-depth research on classifiers from many angles (see Harel et al., 2024). A collaboration with the digital dictionary and textual databank Thesaurus Linguae Aegyptiae has supplied the digitized texts for the corpus (lemmas, transliteration, translation, etc.). ²²

2.2 Methods

Due to space limitations, we discuss only some of the research methods in the following.

2.2.1 Network maps: The general vista

In recent years, the social sciences methodology known as network analysis has been applied to many disciplines, including linguistics. Therefore, the lexicon has been read as a network ²³ structured by different types of connections. As demonstrated in the first part of this article, classifiers perform significant roles in the lexicon, reflecting emic semantic networks that encode the categorization patterns modeling the ancient Egyptian view of the world.

Figure 1.

(a) Middle Kingdom network map of classifiers in Sinuhe.© iClassifier, Susana Soler. Data imported from TLA. (Enlarge for details.) (b) Close-up of [divine] classifiers in Middle Kingdom network map of classifiers in Sinuhe. © iClassifier, Susana Soler. Data imported from TLA.

iClassifier is designed to enable semantic network analysis according to the classifiers in a script (Harel et al., 2023). It presents the visualization of categories and their members in classification patterns, including semantic relations between hosts and classifiers and intra-connections among classifiers (for Chinese mapping by iClassifier, compare Xu, 2024). In network maps generated by iClassifier (e.g., Figure 1(a)), the blue lines connect the classifiers to all the lemmas that use them, and the red lines connect the classifiers that co-occur in a word, i.e., visualizing the multiple classification cases of a few classifiers in a single lemma. The thicker the lines are, the more frequent these connections are, revealing the consistency of classification for a given lemma with a classifier (blue). Classifier clusters formed by a combination of classifiers are also clearly visible (red). For example, in the MK network map of Sinuhe (Figure 1(a)), a group of signs connected by thick red lines forms an example of multiple classification described above. What we learn from the map is that already in the MK, a classifier cluster that originates in a reoccurring composite classifier is created: [man] [woman] [foreign] and often also [foreign land]. ²⁴ These signs commonly appear in a repetitive combination in ethnonyms or collective nouns relating to foreign groups of people (e.g., ṯmḥj.w^{[foreign-man-woman-foreign land] “}Libyans” [R 12]). This result coincides well with the content of the story (viz., the wars of the king in Libya and the wanderings of the Egyptian dignitary Sinuhe in foreign lands among foreign people, see note 14). However, as we see on the map, the sign , which connects ethnicity and “foreign land,” is not a steady part of this compounded classification in Sinuhe until the New Kingdom (see Figure 7), when it will also show the more usual arrangement of signs of this cluster of classifiers, e.g., ṯmḥw “Libyans” [C 2] or štt.tjw “Asiatics” [AOS 34].

A broad look at the map in Figure 1(a) immediately reveals the central categories in the MK texts of Sinuhe—Y1 ²⁵ [writing/abstract/default ²⁶ ], A1 [man], D54 [motion], and A2 [senses&emotions], which will be discussed next. Also prominent are A24 and D36, ²⁷ embracing the semantic category [action]. Mid-sized categories such as P1[boat, travel], N23 [land], G37 [negative], N5 [sun/time], and O1 [habitat] are also well represented. The map also clearly shows the incompatibility of some classifier categories, e.g., , , and do not seem to share any lemma in Sinuhe. Other small categories with few members appear to be “islands” disconnected from the rest.

2.2.2 Frequency bar chart: “Long-tail” distribution

Another sort of data visualization is the frequency bar chart, ²⁸ which clearly shows the distribution of classifiers in lemmas and tokens. Figure 2(a) displays the classifiers in MK versions of Sinuhe by frequency of lemmas and tokens. The resulting graph presents what is called a “long-tail” distribution, ²⁹ meaning that a few classifiers concentrate most classified words.

Figure 2(b) shows a detail of the previous graph showing the first 40 classifiers out of 122 counted in Figure 2(a). This group, which makes up about one-third of the list, contains classifiers that occur with 100 to four lemmas each (blue column). All remaining classifiers outside this graph are attested with only three to one lemmas.

The number of lemmas ascribed to a single classifier reflects the centrality of the specific classifier category as it is represented in a certain text. Alternatively, tokens appearing repetitively with a certain lemma reflect the centrality of certain words and topics in a specific corpus. The semantic network map of MK versions of Sinuhe (Figure 1(a)) had already revealed the central categories of this corpus: [writing/abstract/default], [man], [motion], and [senses&emotions]. Following are some remarks on the main categories:

Figure 2.

(a) Graph showing all the classifiers (122) in Middle Kingdom versions of Sinuhe by frequency of lemmas and tokens © iClassifier, Susana Soler. Data imported from TLA. (b) Detail with the 40 first classifiers of (a), Middle Kingdom © iClassifier, Susana Soler. Data imported from TLA. (Enlarge for details.)

Y1 [writing/abstract/default] (100 lemmas/195 tokens).‏ The hieroglyph depicts a rolled and tied papyrus scroll. While probably the most frequent classifier in the Egyptian script, it still represents a challenge for Egyptologists. It has been considered as head of the category [writing] in the Pyramid Texts (Wiesmann, 1922). Later, as the category was extensively extended, it was dubbed “abstract” (see Gardiner, 1957: 533 inter alia, and Goldwasser, 1995: 24 n.68, 88). A new analysis of the meaning of this classifier as a “residual” classifier was suggested by Kammerzell (2015). We presently call this category [default] category (see note 23). In MK Sinuhe, there is a small group of lemmas fitting into the category [writing] such as wḏw.w “command, decree” [B 178], ṯꜣz “saying, utterance” [B 184] or zẖꜣ.w “writing, record” [B 311], while the large majority of lemmas should be already assigned to the extended category of [abstract/default]. ³⁰ The most attested lemma that takes in all its occurrences in Sinuhe is rḫ “know.” ³¹ Interestingly, this lemma is always classified in the text with a single classifier . The high number of occurrences of the verb rḫ “to know” reflects the many mental and cognitive transpositions that the hero of the story goes through.

A1 [man] (93 lemmas/210 tokens) reflects the central role of human actors in Sinuhe, both Egyptians and Asiatics. Various human professions and qualities play an essential role in the evolving plot of this literary text, and as already noted in the first part of this article, human classifiers also reflect the grammatical role of nominalizers (Lincke and Kammerzell, 2012: 97), e.g., jb “the one who is thirsty” [B 94]. The most attested lemma in the category is the compounded lemma bꜣk jm “thy humble servant” ³² (14 occurrences). This compound, literally meaning “servant there,” is used by Sinuhe to refer to himself when addressing the king. In the example above, the classifier appears in the midst of the compound after the word “servant.” In New Kingdom versions, we see that the compound is now understood as a single linguistic unit (noun + adverb) referring as a whole to the human speaker, with the repetition of the classifier at the end of the compound, i.e., [AOS verso 4] (on classifiers in compounds, see Section 1.7). Another highly attested lemma with the [man] classifier is [B 17] and [R43] sṯ.tjw “ Asiatics” (10 occurrences, referring to the ethnic group, no location classifier). In the last examples, the number of tokens reflects firstly the centrality of the discourse with the king and the pivotal role of the Asiatics in the plot. Besides the counted semantic classifiers , there are also about 300 tokens consisting of 1st person pronouns showing as a pragmatic/grammatical classifier (see note 9).

D54 [motion] (81 lemmas/178 tokens) is the most relevant category of action in the narrative. Sinuhe’s main plot develops through his journey on land to the Levant from Egypt. The most prominent taxonomic members of the [motion] category in MK Sinuhe are verbs like šmi̯ “to go” (10 occurrences), ꜥḥꜥ “to stand” (9 occurrences), or hꜣb “to send” (8 occurrences). In these verbs, we witness a 100% occurrence of the classifier. Figure 3 displays a query for centrality in the use of this classifier in MK versions. In this graph, the differences in the shade of red mark whether a specific lemma is always classified by or seldom classified by it. ³³ The values of the shades of color are clarified with percentages below the graph. Thus, the darker red circles mean that the lemma is always or very frequently classified by as in hꜣb “to send,” and a light red circle indicates that the lemma is only rarely classified by it. One of the examples is the verb dmj “to touch,” which is only once classified by (1 out of 4 examples in the MK corpus in Figure 3, and 1 out of 7 in all versions, see Figure 4). It occurs in a sentence where dmj has the meaning of “arrive/approach (to a destination)”: “(After) I had given the way to my feet to the north, I came near (, lit. “touched”) the Walls of the Ruler” [R 42]. The lemmas in Figure 3 also includes compounds and collocations in which the classifier can appear in the middle, after the word (mostly a verb) expressing [motion]: e.g., ʿḥʿ-ỉb “standing-minded” ³⁴ [B 57] (after Stauder 2023: 287–291). This means that the compound form is not lexicalized and can even allow other components to break the compound, such as in wḏb = f ꜥ, where the personal suffix pronoun f (“he”) is in the midst of the compound wḏb (ꜥ) [B 163] “turn the hand away; desist” [B 163]. ³⁵

Figure 3.

The classifier in Middle Kingdom Sinuhe. Edge by centrality rank © iClassifier, Susana Soler. Data imported from TLA. (Enlarge for details.)

Figure 4.

Classifier alternation graph and co-occurrence table for dmj “touch,” MK and NK. © iClassifier, Susana Soler. Data imported from TLA.

Figure 5.

The classifier in Middle Kingdom Sinuhe. Edge by no. of examples © iClassifier, Susana Soler. Data imported from TLA.

Table 2.

Sample of lemma centrality rank statistics for Figure 3. ³⁶

lemma	translation	counts
šmi̯	go	10/10
ꜥḥꜥ	stand	9 / 9
hꜣb	send	8 / 8
šms	follow	3 / 3
zš	spread	3 / 3
wpw.tj	messenger	2 / 2
m-ḫt	behind	2 / 2
zbi̯	go	5 / 7
swꜣi̯	pass	4 / 7
pḥ	reach	4 / 7
ꜣs	hasten	2 / 4
sjni̯	wait	1 / 2
ḥzi̯	turn back	1 / 2
ẖn	approach	1 / 2
wpw.t	message	2 / 6
ptpt	tread	1 / 3
zꜣi̯	linger	1 / 3
šms.w	follower	1 / 3
dmj	touch	1 / 4
jwi̯ *(see note 30)	come	5 / 19
wꜥr.t	flight	2 / 10

A2 [senses&emotions] (71 lemmas/146 tokens). The sign plays the role of an extended generic classifier, which covers three semantic fields: [eating, drinking], [speech], and [thoughts & feelings]. ³⁷ According to the Egyptian classification system, food, drink, voice, thoughts, and emotions reside in the “body container” (Goldwasser, 2005: 111; 2006a: 476). The conceptual metaphor [the body is a container], ³⁸ which embodies this idea, is one of the central metaphors that construct human behavior until today in many cultures (Lakoff and Johnson, 1980: 46). More than any other Egyptian literary text of the Middle Kingdom, the protagonist in Sinuhe is concerned with his own decisions and impulses that shaped his fate, such as running away from Egypt, settling abroad with the Asiatics, his undying feelings towards the king of Egypt, and finally the decision to leave everything behind and go back to Egypt. All along the story, [thoughts & feelings] govern the plot. Figure 5 presents a graph showing all the lemmas classified by in MK versions of Sinuhe. This is another type of data visualization in iClassifier that informs about the centrality of a lemma in a certain classifier category in the text by displaying a wider connecting line that depends on the number of tokens for a given lemma. As a result, we can see that snḏ “fear,” mrj “love” ³⁹ (with the nominal form mrw.t), and kꜣj “to think about, plan” are central in the text. They are the lemmas with the greatest number of tokens for this classifier in Sinuhe. However, the largest number of lemmas classified by belongs to the semantic field of [speech].

Finally, the semantic field [eating, drinking, hunger, thirst] is present in the physical needs and states mentioned in the text: e.g., wnm.t “food” ⁴⁰ [B 104], jb.t “thirst” [B 21], or even ḫnm “to breathe” [B 234]. We suggest that the last verb is an example of the broader scope of this category—the body can be filled not only with edibles and beverages but also with air and scents.

2.3 Some remarks on diachrony in Sinuhe: The repertoire of classifiers and the [divine] category

The first effort to discuss the diachrony of classifiers in Sinuhe was made by Lincke (in Polis, 2023: 193). She published a table (Figure 6) presenting a diachronic comparison between the 20 most frequent classifiers in the two best-preserved copies of Sinuhe: Middle Kingdom (B) and New Kingdom (AOS). The four largest categories described in the previous chapter keep the top occurrence rate in the New Kingdom version of Sinuhe, as we see now in Lincke’s table.

A clear change in what she describes as the “number of attestations” could be observed. The classifier shows more attestations in the NK (compare Kammerzell, 2015). On the other hand, the classifier has somewhat fewer attestations, augments in number, and the classifier is rather similar in the number of attestations in the NK and MK versions.

According to Lincke’s table, a remarkable diachronic change in the use of classifiers can be observed in the [divine] category in Sinuhe (see also the discussion in Section 1.8). As pointed out in her table, the classifier that was used in MK (version B) disappears completely. The use of the classifier increases significantly in the NK (version AOS).

In the NK network map of Sinuhe (Figure 7) created by iClassifier, we can clearly follow these changes in detail. The [divine] classifier , which appeared only in MK version B, has disappeared ⁴¹ (compare the MK map Figure 1(a), including a close-up of Figure 1(b)). The Horus classifier , the first prototypical member of the [divine] category (see Section 1.2), becomes the category’s most frequent classifier in the NK. Moreover, the distinctive diachronic feature from MK to NK in Sinuhe consists of the emphatic insistence on categorizing the ruler as [divine]. ⁴² For example, the compound ms.w-nswt “royal children/escort,” written in the MK versions [B 197] or [B 264], takes in all NK examples the additional classifier after the word nswt “king, royal,” e.g., [AOS 9] (see the map in Figure 7, including a close-up).

Figure 6.

After Lincke, in Polis (2023). Guide to the Writings from Ancient Egypt, p. 193.

An enlightening example of the classification of the king as [divine] in the NK is found in the sentence zꜣ (=f) smsw m ḥr,j jr,j nṯr-nfr Z-n-wsr.t “(His) son^[divine], the eldest^[old-divine], was its commander-in-chief, the perfect^[divine] god^[divine] Senwosret” [C2, NK, Ramesside]. Here, the son “zꜣ” (the eldest son) is classified pragmatically as [divine]. We know that Senwosret (the son) was crowned coregent with his father Amenemhat I in the year 20 of his reign—so he is the king’s son and a king simultaneously. This situation is reflected in the pragmatic classification in this sentence. Senwosret also takes the common epithet for the Egyptian kings, “perfect god,” which is also classified with .

Finally, the categorization as [divine] is also extended to other members of the semantic field of kingship in the New Kingdom, for example, the ḥm.t-nswt “king’s wife” [AOS 2] (see discussion below), or ꜥḥ.t “palace” [AOS 64] versus MK [B 166].

The female divine classifier I12 is not included in Lincke’s list. Here, a cobra goddess was selected to classify all goddesses, including anthropomorphic ones. The classifier is well known in the MK versions of Sinuhe, mostly with names of goddesses (e.g., Sḫm.t “Sakhmet” (a lioness goddess) [B 45]). Interestingly, Sakhmet is also known in the Coffin Texts with this classifier, although the sign is barely attested with a classifier function in these funerary texts (Shalomi-Hen 2008: 187 n.17). The category [divine-female] is extended in the NK manuscripts of Sinuhe to include additional goddesses and royal females, for example, the name of the goddess Ḥw.t-ḥr.w “Hathor” [AOS verso 23], usually manifested in a human or cow form. Her compounded name, meaning the “Chapel of Horus,” is classified by , while the word ḥw.t “chapel” that makes the first part of her name correctly gets the [house/habitat] classifier. Other examples are Nw.t “Nut” the sky goddess [AOS verso 25] or nb.t-tꜣ.wj “mistress of the two lands” = the queen [AOS verso 56]. In the title ḥm.t-nswt “king’s wife” [AOS 2], the first word within the cartouche, “king,” is classified by and the word “wife”—being the queen, here the divine wife—is classified by , while the whole compounded female royal title closed in the royal cartouche is again classified into the [divine] category. Female goddesses get full access to the (originally male) [divine] category already in the Old Kingdom Pyramid Texts but only in the later pyramid of Teti. ⁴³

There is another hieroglyph that functions as a generic taxonomic classifier for the [divine] category in Sinuhe: R8. This sign, commonly used as a logogram ⁴⁴ for the word nṯr , may be used as an unpronounced classifier in the names of different gods, and it is well documented as a category-head of [divine] interchanging with and already in the MK Coffin Texts ⁴⁵ (see discussions in Goldwasser, 2002: 113–114; Shalomi-Hen, 2008: 185).

Indeed, in Sinuhe, the classifier is known in the MK manuscript [B 207] with the single lemma psḏ.t “Ennead” as a triple repeater classifier to indicate the plurality of nine gods. In the only parallel for this word from the NK [AOS verso 24], we see the use of as a triple repeater that now gets the additional generic [divine] classifier at the end of the word.

Figure 7.

New Kingdom network map of classifiers in Sinuhe © iClassifier, Susana Soler.

In the case of the hieroglyph , the image chosen to represent the idea of the divine is not “an example of” the divine, a specific god, such as , , or , but a metonymic representation of the divinity, i.e., the flagged pole that was marking temple buildings in ancient Egypt (Hornung, 1982: 33–66, Goldwasser, 2006b: 274).

The classifier may be compared to the semantic classifier 示, literally “sacred tablet,” in the ancient Chinese script (see Section 1.1 and Xu, 2024). In ancient Chinese culture, the sacred tablet, like the flagged pole in ancient Egypt, is a necessary object connected to the veneration of divine ancestors and comes to represent them by metonymic relations.

The use of [divine] classifiers in Sinuhe also reflects the early monotheism developed in Egypt during the Amarna period (c. 1352–1323 bc), as well as the seamless correlation between the classifiers and cultural changes in Egypt. This conceptual transformation is reflected in the spelling of the word Jtn-Aten, “sun disk,” which later became the name of the monotheistic god of the Amarna period in the New Kingdom. ⁴⁶ However, the Aten was classified during the Middle Kingdom in all Egyptian texts only by the sun disk as it was not yet conceptualized as a full-fledged god. Accordingly, the lemma Jtn “sun disk” appears in the MK versions of Sinuhe only with the sun classifier (e.g., [R 7]). At the beginning of the New Kingdom, during the 18th Dynasty before Amarna, the Jtn-Aten sun disk is officially made a member of the Egyptian Pantheon and may get like all other Egyptian gods the divine classifiers , , 
and (e.g., pEbers XCII,14). A surprise awaits the student of the Egyptian culture when, during the Amarna period, the Aten becomes the single official god. When the Aten is proclaimed to be the one and only god during the Amarna period, it is written systematically without any [divine] classifier, keeping only the iconic “sun” classifier. However, a closer look reveals an astonishing correlation between the hieroglyphic script's spelling and the abrupt changes in beliefs in the Amarna time. As the revolutionary decrees of King Akhenaton in Amarna officially erase all the other gods, Jtn cannot take any [divine] classifier. A [divine] category implies, a priori, a superordinate generic term with at least a few “examples of” members in its category (see Sections 1.1 and 1.2). As the Aten is the only official god during this particular period in the history of Egypt, it cannot be a member, “an example of,” any generic [divine] category. On the level of official theology, the category [divine] stopped to exist in this period, as there is only one single god.

Immediately after the Amarna period, the New Kingdom versions of Sinuhe show that the Aten has already become again a steady member of the revived [divine] category as he now gets the additional [divine] classifier (e.g., [Sinuhe NK version AOS versus 35]). The cultural and religious changes are again accurately reflected in the script. According to the classifier system, after a short interval of less than 20 years, Egyptian religion returned to full-fledged polytheism, which is clearly reflected also in texts and all sources of material culture of the time. The category [divine] is installed again on its many former members, and the Aten became “one of the many.”

2.4 Tracing alternative classification

As indicated previously (Section 1.3), words may appear in different spellings with different classifiers and with more than one classifier in Egyptian. Classifiers may alternate during the same period due to the extension of meanings and be assigned to different categories. Diachrony also exposes changes in classification and classifiers combinations; see the following example of the verb dmj “touch,” discussed above with Figure 4.

The prominent classifier in the MK is [arm-related movement/action]. One example in the context of “touch ground, reach” () gets a referential classifier—the generic [motion] classifier (see Section 2.2.2). In the NK, the Sinuhe texts show the [default] classifier and once a double classification with the more generic [action].

3 Classifier lists: How many classifiers does the Egyptian script use?

Gardiner created the primary sign list of the most common generic classifiers in his grammar book (1957: 31–33; see Table 3). His list contains 93 hieroglyphs that may function as classifiers. The list is principally organized according to the order of hieroglyphs in his full sign list, which appears at the end of his grammar book (1957: 439–548). The corpus of texts that Gardiner used for creating his general list dates mainly to the Middle Kingdom and the 18th Dynasty. It includes around 743 hieroglyphs, with specifications of their possible semiotic functions and a few examples of each use. It is in no way a comprehensive list of all known hieroglyphs. A complete digital list of hieroglyphs of all periods and their possible functions is currently built in the online digital project Thot Sign List (TSL) http://thotsignlist.org, headed by Stéphane Polis. ⁴⁷ Table 3 presents Gardiner’s classifier list in conjunction with the codes of the hieroglyphs from his general sign list by the end of his book.

Table 3.

Gardiner’s list of common “generic determinatives” with added code numbers from his general list of hieroglyphs. Order and category description (“names”) according to his publication. Read the list from left to right.

Gardiner’s number according to sign-list	Sign	Meaning after Gardiner	Gardiner’s number according to sign-list	Sign	Meaning after Gardiner
A1		[man, person]	A2		[eat, drink, speak, think, feel]
A7		[weary, weak]	A9		[lift, carry]
A13		[enemy, foreigner]	A14		[enemy, death]
A17		[child, young]	A19		[old man, old, lean upon]
A21		[official, man in authority]	A24		[force, effort]
A28		[high, rejoice, support]	A30		[praise, supplicate]
A40		[god, king]	A41/A42		[king]
A50 (dyn.XVIII A51/A52)		[exalted person, the dead]	A53		[mummy, likeness, shape]
A54/A55		[lie down, death, bury]	B1		[woman]
A1 + B1		[people]	D1		[head, nod, throttle]
D3		[hair, mourn, forlorn]	D4		[eye, see, actions of the eye]
D5		[actions or conditions of the eye]	D19/D20		[nose, smell, joy, contempt]
D32		[envelope, embrace]	D36		Substitute in hieratic for D40 and other arm signs, less often in hieroglyphic script
D39		[offer, present]	D40		[force, effort] interchangeable with A24
D41		[arm, bend arm, cease]	D53		[phallus, beget, urinate]
D54		[walk, run]	D55		[move backwards]
D56		[leg, foot, actions of foot]	E1/E3		[cattle]
E20		[savage, typhonian]	F18		[tooth, actions of teeth]
F21		[ear, states or activities of ear]	F27		[skin, mammal]
F51		[limb, flesh]	G7		[god, king]
G37		[small, bad, weak]	G38		[bird, insect]
I12/I13		[goddess, queen]	I14		[snake, worm]
K5		[fish]	M1		[tree]
M2		[plant, flower]	M3		[wood, tree]
M33/M33A		[grain]	M43		[vine, fruit, garden]
N1		[sky, above]	N2		[night, darkness]
N5		[sun, light, time]	N14		[star]
N21		[land] often replaces N23	N23		[irrigated land]
N25		[desert, foreign country]	N31		[road, travel, position]
N33A		[sand, minerals, pellets]	N34		[copper, bronze]
N35A		[water, liquid, actions connected with water]	N36/N37		[sheet of water] less often N37
O1		[house, building]	O31		[door, open]
O39		[stone]	O49		[town, village, egypt]
P1		[boat, ship, navigation]	P3		[sacred bark]
P5		[air, wind, sail]	Q6		[box, coffin]
Q7		[fire, heat, cook]	S28		[clothe, linen]
T14		[foreign] country or person	T30		[knife, cut]
U7		[hoe, cultivate, hack up]	U9		[corn]
V1		[rope, actions with cord or rope]	V10		[royal name, king]
V12		[bind, document]	V19		[shrine, palanquin, mat]
W3		[festival]	W10		[cup]
W22/W24		[pot, vessel, beverages] less accurately W24	W23		[vessel, anoint]
X2		[bread, cake]	X4/N18		[loaf, cake, offering]
Y1/Y2		[book, writing, abstract] Y2 older form	Z1		[one], the object depicted
Z2		[several, plural]	Z5		Substitute for signs difficult to draw (mostly hieratic)
Z9		[break, divide, cross]	Aa2		[tumours, odours, disease]
Aa3		[bodily discharges]

The main shortcoming of Gardiner’s “list of generic determinatives” is that it presents together without any remark “determinatives” that head large-scale categories comprising dozens of members (e.g., [senses&emotions] and [motion], see below) side by side with marginal categories, such as [carry, work], which include a few members only.

Winand and Stella (2013), who published a new classifier list, took a big step forward in classifier studies in Egyptology. Their book Lexique du Moyen Égyptien comprises around 2500 words selected by authors’ criteria, mainly from Faulkner’s Dictionary (1962) and Gardiner’s glossary (included in his Grammar). The chronological period covered by the Lexique starts from the First Intermediate period to the 18th Dynasty (Middle Egyptian and “Traditional Egyptian”), excluding vocabulary from the Coffin Texts. Their book offers a much more extended classifier list encompassing 343 hieroglyphs with the necessary inventory of words that make each category according to their collection, i.e., presenting category members and, thus, categories’ size in their corpus. However, the members (lemmas) in the categories are not sourced. In their list, 202 classifiers take only one or two lemmas. In an additional list, they signal out the largest classifier categories of the script based on their sources (see Table 4). This small list of classifiers that can occur in 20 or more words is what they consider to be the list of “generic classifiers.”

We observe that the three largest categories in their corpus are in this order: [writing/abstract/default], 
 [senses&emotions], and [motion]. If we consider the classifiers and as representing together the category [action] and its variations, these two classifiers would represent the fourth-largest semantic category. The wide use of both last classifiers echoes the importance of physical and even state power in official ancient Egypt (e.g., ḫrp “to govern” and dꜣr “to control”), as it is mirrored in the records of the writing elite.

Table 4.

Classificateurs génériques after Winand and Stella (2013: 129). * Bold numbers mark the largest categories (added by Goldwasser-Soler). Blue marks the categories and discussed above.

A1		61	M2		24	P1		22
A2		129	M3		25	Q7		31
A17		20	N5		37	S28		24
A24		58	N23		25	T14		26
D19		20	N25		26	V1		32
D40		49	N31		20	W22		21
D41		25	N33		30	Y1		145
D54		109	N35A		28	Z9		49
G37		87	O1		51	Aa2		24

The classifier embracing the category [negative/small] also has a strong presence in their list, with a large number of lemmas referring to small, young, and negative actions and feelings, illnesses, and not a few negative social judgments, e.g., “poor” (see the example of ḫꜣr.t “widow,” Section 1.2 and ḏr “tomb in a foreign land,” Figure 8). The [man] category is also very visible in Winand and Stella’s list. ⁴⁸ Finally, the classifier [habitat] is a prominent category that embraces human abodes (and their parts) as well as animals’ habitats, such as “stable” or “nest” (Goldwasser, 2005: 96–97, compare the network maps Figures 1(a) and 7).

Surprisingly, in our small, studied corpus of Sinuhe texts of the MK, most classifiers of the basic Gardiner list of classifiers (Table 3) are present. The absence of a few members of the Gardiner list that head categories of medium or small scale may be explained by the range of vocabulary used for the literary text. For example, the classifiers V19 [basketry] and M33 [grain], which usually relate to administrative contexts, are absent from our literary text. In the broad picture, the centrality rate of the classifiers in Sinuhe shows results that are very similar to those of Winand and Stella (see Figure 2(b) and Table 4). Yet, Sinuhe shows an unusually large category of [man], with 93 lemmas and 210 tokens. The [woman] category is also more extensive than the representation of this category in Winand and Stella’s book (see note 43), showing 26 lemmas. These results reflect the central role of humane collective nouns, what Winand and Stella call [collectivitÉ humaine] and [collectivitÉ]. ⁴⁹

The most prominent tokens of the collective nouns including the classifier are wḥyt “tribe” and ms.w-nswt “royal children, royal escort.” In all MK examples in Sinuhe the word tribe takes additionally the [foreign] classifier marking “tribe” in Sinuhe as a specific non-Egyptian social organization (e.g., wḥyt [B 86], [R 52]).

On the other hand, some classifiers used in Sinuhe are missing in Gardiner’s list of classifiers. The hieroglyph O36 [wall] is noticeably absent from Gardiner’s list (Table 3) but correctly listed by Winand and Stella (2013: 161). When functioning as a classifier, heads a medium category in the script system that includes a few taxonomic relations such as “types of walls” as well as a set of schematic roles of “types of buildings” (e.g., jnb “fortification” [R 141]) of which the wall is an essential part (metonymic relation).

Figure 8.

Classifier alternation graph for ḏr, MK and NK.

The classifier is used in a MK version of Sinuhe [B 198] in the word ḏr “grave in a foreign land,” which may refer to the typical Canaanite tomb where a rectangular brick or stone wall enclosed the dead. ⁵⁰ This is a further example of a metonymic relation with the “host” word since the “wall” is a part of the “grave.” In a personal letter addressed to Sinuhe in exile, the Egyptian king says “(it cannot happen) that you will die in the desert! Asiatics will not bury you! You will not be put in the hide of a ram as they make your grave [ḏr]” [B 197–198]. ⁵¹ Remarkably, the classification strategy for ḏr in the same passage of NK version of the text (AOS) presents a radical change (Figure 8). Here, ḏr [AOS verso 18] is classified into the taxonomic category [negative], hence presenting an emic Egyptian social judgment as in the case of the widow and the Asiatic discussed previously (Sections 1.2 and 1.8). Moreover, the [negative] classifier is preceded by the classifier . By metaphorical extension, the category [fish] also refers to [bad smell]. ⁵² Thus, a grave outside Egypt is not only categorized here as [negative] but also pejoratively as something with a bad smell, as the fish is the prototype of “things that stink” for the Egyptians (Goldwasser 2005: 106). The preparation for the ideal afterlife in Egypt centered on constructing a tomb of typical Egyptian architecture and providing correct funerary equipment. It was an essential requirement in the culture and belief system of the elite. The Egyptians, who held their country in the highest esteem, would have abjured the prospect of a burial in a foreign land even more if it did not conform to the Egyptian ritual (i.e., the use of hide wrappings instead of linen bandages coating the mummified body of the deceased).

Once more, when, after long years, an old Sinuhe presents himself in front of the king, the ruler first criticizes him for his reckless disregard of his funerary rites: “Your wanderings have taken a toll on you, Old Man! ⁵³ You have reached [old age]. This is no small matter, the burial of your body! Your burial ⁵⁴ will [not] be by the foreigners! Your grave [ḏr] was not made (by them)” [AOS verso 47–48]. ⁵⁵ This sentence contains the third and last example of ḏr, [AOS verso 48] in Sinuhe, this time only in the text's NK version (AOS). In this last occurrence, ḏr repeats the classifier, thus keeping the “foreign tomb” consistently in the [negative] taxonomic category. In conclusion, the “happy ending” of Sinuhe meant our hero's return home and the royal gift of an adequate burial in Egypt. This elevation of “Egyptianity” versus “foreignness” is very prominent in the Story of Sinuhe.

Finally, when employed in verbs, the classifier plays the semantic role of a prototypical patient as in znb “overleap” [R 141], nicely exemplifying this semantic role (see Section 1.4).

Another “missing classifier” in both lists (Tables 2 and 3) is B56 [divine-female] in a human manifestation (Shalomi-Hen 2008: 186). This sign, showing a woman with the royal cobra on her forehead , is known to classify goddesses also in MK hieroglyphic inscriptions. The above depiction of the iconic hieroglyphic sign comes from the niche of the tomb of Sarenput II in Aswan from the 12th Dynasty (MK), where it occurs in the name of the goddess St.t “Satet.” In Sinuhe, the classifier appears only in the MK text [R] with the compound lemmas ḥm.t-nsw.t “king's wife,” ḥn.wt-tꜣ “mistress of the land,” jr.jt-pꜥ.t “hereditary princess,” wr.t-ḥzw.t “great of praise (the queen),” and the name of queen Nfr.w “Neferu.” Therefore, the Egyptian queen is classified in this text version [R] as a [divine-female]. The situation is different in the other versions. For example, the spelling of the lemma ḥn.wt-tꜣ that shows in text [R 194] the classifier, appears in the other major Middle Kingdom papyrus [B 166] as with the classifier , the [divine-female (non-human)]. This exchange presents the “classifier competition” in the Middle Kingdom between two [divine-female] prototypes: the anthropomorphic goddess and the cobra (zoomorphic) goddess, which parallels the struggle of the [divine-male] prototypes discussed in the first part of the article (see 1.2). Finally, in the New Kingdom version [AOS 63], the same word ḥn.wt-tꜣ presents the generic [divine] (originally male) classifier. The NK versions of Sinuhe show the convergence of the gods, male and female, into a single overreaching concept with a neutralized gender.

Due to space considerations, we could comment only briefly here on the topic of the classifier repertoire. In the future, the Egyptian lists will be compared with Chinese lists of classifiers, such as the one compiled by Xu (2024) for the corpus of Guodian texts. ⁵⁶

To sum up, the relatively small hieratic corpus from the Middle Kingdom (i.e., MK versions of Sinuhe) chosen by us for our pilot research in corpus linguistics shows a few very clear and fundamental results:

The small corpus of MK versions of Sinuhe presents almost all the repertoire of generic classifiers compiled by Gardiner.

Albeit some small differences related to the story's plot, the most frequent classifiers are the same classifiers singled out by Winand and Stella as the most central classifiers. Thus, these seem to be the main categories or hubs of knowledge organization in the Egyptian mental organization during the Middle Kingdom. These hubs organize many concepts that materialize in different words and are connected through fringe members to other categories (see the network map Figure 1(a)).

All semantic classifiers and their host words show clearly motivated relationships. All classifier variations and diachronic changes can be matched with referential reasons or cultural changes known from other sources.

4 Conclusions

From the theoretical point of view, this state of affairs reiterates the conclusion that classifiers are a system in the Egyptian script. As every Egyptian text shows a consistent repertoire of grammatical forms obeying a defined set of grammatical and syntactic rules of the period, we can expect that texts would also show a consistent repertoire of classifiers that are a rule-governed system in the script, as was suggested in the first part of the article. This hypothesis should be verified in future corpus linguistic studies of classifiers. The classifier phenomenon belongs to the cognitive-linguistic space of language (Lakoff, 1987). Classifiers dissect the universe of knowledge and vocabulary into categories, offering faster access to word meanings (Kemmerer, 2016; Goldwasser, forthcoming). The system is deeply rooted, on the one hand, in the cognitive abilities of the human mind and, on the other, in the layered cultural knowledge acquired over generations in every single culture.