Classification of nouns in the world’s languages

In most languages, a noun belongs to a closed group of words with a common grammatical fate. Linguists call such a group a gender, class, or animacy category. Membership in a group is manifested not in the word itself, but in its surroundings: the article, adjective, numeral, pronoun, and verb change form depending on the group to which the noun belongs. This dependence serves as the main criterion: classification exists only to the extent that it is visible in agreement.

The scale of the phenomenon is enormous. The Niger-Congo family comprises approximately 1,500 languages, and almost all of them are characterized by noun classes. Old English distinguished three genders with full agreement; by the end of Middle English, the system had disintegrated, and today only the pronouns he, she, and it remain. Between these extremes lie dozens of intermediate solutions that language offers to a single problem: how to store information about the type of object within a word.

Approval criterion

The linguistic tradition, rooted in Peter Matthews’s definition, is simple: gender is a system in which the class of a noun is reflected in the forms of its syntactically related elements. The Spanish adjective, the Russian participle, the Swahili demonstrative pronominal — all read the noun’s label and adapt to it. Agreement extends throughout the sentence, and this distinguishes gender from a simple dictionary label.

Classifiers work differently. The classifier word stands next to the numeral and designates the type of object being counted, while the form of the other members of the phrase remains unchanged. In languages with classifiers, their number runs into the hundreds, while gender systems typically fit into two or three classes. In her 2000 monograph, Alexandra Eikenwald proposed a continuum between these extremes: from loose, context-sensitive nominal classifiers to rigid gender systems with rigid agreement.

Three genders: the Indo-European standard

German, Latin, Ancient Greek, and Sanskrit divide nouns into masculine, feminine, and neuter genders. Russian inherited this system almost unchanged. The Romance languages, transitioning from Latin, lost the neuter gender: Spanish, Italian, and French make do with two. French has gone the furthest — it doesn’t even have a distinction of number in the participle.

The semantics in these systems are fuzzy. Spanish words ending in -o are more often masculine, words ending in -a are more often feminine, and this rule covers most of the vocabulary; counterexamples like el día ("the day," masculine) or la mano ("the hand," feminine) must be learned separately. German is predictable elsewhere: any diminutive suffixes -chen and -lein automatically assign neuter gender, so Mädchen ("girl") is grammatically neither masculine nor feminine. The suffix here is stronger than the natural gender of the referent.

Classical grammarians call such words hybrids. In Greek, paidion ("child") behaves the same way for the same reason: the diminutive is assigned to the neuter gender regardless of whether the speaker is a boy or a girl. In modern German, Fräulein no longer means "little woman" — the suffix has faded semantically, leaving behind only "unmarried," while the gender remains the same. Grammar is more conservative than meaning.

Genderless: Algonquian languages

Ojibwe, an Anishinaabe language spoken around the Great Lakes, distinguishes two classes: animate and inanimate. There is no gender distinction here: the same adjective and verb forms are used for both male and female subjects, and the choice of form depends on the animacy. This class includes people, spirits, mythological creatures, animals, birds, fish, insects, and trees.

The class is pronounced at the end of the plural. Animate nouns take -g, inanimate ones -n. The word mitig exists in both forms: the animate mitigoog means "tree," while the inanimate mitigoon means "stick." The difference between a whole plant and a fragment is expressed not by a dictionary, but by the agreement class.

The class boundary is noticeably broader than biology. Stones in Ojibwe are animate because tradition ascribes a spirit to them. Cars, potatoes, and apples are freely placed in the animate class by native speakers. The gender of drawings depends on dialect: words for money and bread fluctuate between classes among different groups of speakers. The system remains vibrant precisely because decisions within it are made by the community, not by a dictionary.

Reasonable and unreasonable: the Dravidian variant

Tamil and related South Asian languages construct the highest level of division based on rationality. The rational class includes humans and deities; everything else — animals, plants, and objects — forms the nonrational class. Within the rational class, there are masculine, feminine, and collective subclasses; animals and inanimate objects are not included in this hierarchy at all.

Comparative data shows that the scheme was not uniform across the entire family. Three types of gender systems have been reconstructed in the Protem Dravidian languages. The southern branch distinguishes masculine, feminine, and neuter genders in the singular, and in the plural, draws a distinction between humans and all others. The central branch, with the exception of Telugu, divides genders differently: male humans versus all others. In Telugu and the northern languages, the singular retains the distinction by masculine gender, while the plural distinguishes by human gender.

The system’s operation is clearly visible in pronouns. In Telugu, waadu ("he") is opposed to adi ("she, it"), and in the plural, waaru refers only to people. Gender is fixed in the pronominal system here, and researchers of Dravidian morphology see agreement within noun phrases as a rudimentary classificatory structure on top of the older gender structure.

Dust, prefix and agreement: Niger-Congo

The Niger – Congo family is the largest language family on the planet: approximately 1,500 languages spoken by approximately 459 million people. Traces of noun classes are found in all branches of the family except Mande, Ijoid, Dogon, and Katla-Rashad. The core of the system consists of the Bantu languages of East and Southern Africa.

Swahili provides a textbook example. Each noun receives a class prefix, and the same class is repeated in dependent words: agreement encompasses adjectives, numerals, demonstratives, and verb markers. Classes 1 and 2 are people: mtu "person," watu "people." Classes 3 and 4 are trees and plants: mti "tree," miti "trees." Classes 7 and 8 represent objects like kiatu "shoes," viatu "shoes."

There’s a catch here. Odd classes are singular, even classes are plural, and the plural is formed by moving to the next class above. The class 11 noun ubao ("board") can take pluralization from class 10, and class 14 ukweli ("truth") doesn’t decline for number at all like the mass class. Number in Bantu languages lives within the class system, and these two grammatical meanings are morphologically intertwined.

The number of classes varies. A typical Bantu language contains 15 to 18 classes; according to Kreyssel’s calculations, agreement classes typically number 10 to 15, while the Atlantic languages of West Africa reach up to 30. For Proto-Bantu, 18 agreement and 16 formal classes are reconstructed, while Ganda retains about ten. In Proto-Niger-Congo, class 1 marked the singular person, class 2 the plural: the oldest opposition in the family is for humanness.

Fula (Fulfulde) demonstrates the other pole of the same idea. Classes are marked by suffixes, and there are 24 to 26 of them, depending on the dialect. The first two classes are personal, classes 3–6 are diminutive, 7–8 are augmentative, and the rest are neutral. The suffix suggests the type of referent: -ki for trees and plants, -nge for large animals, -ndu for round objects and dwellings. Nagge means "cow," suudu means "house," and lekki means "tree."

The pronoun layer of Fula is revealing. Man and woman belong to the same class, so the English "he" and "she" are translated by the same pronoun. Gender is expressed here by adjectives, while class morphology is occupied with something else — the distribution of worldly objects into grammatical categories. Gender and noun class turned out to be independent dimensions, and Fula chose the latter.

The Moon, Women and Fire: Dyrbal

The Australian language Dyrbal from Queensland is famous for its four classes, which Robert Dixon described in the 1960s. Class I encompasses men and most animate objects, including the moon: bayi yara – "man," bayi gagara – "moon." Class II collects women, water, fire, war, danger, and exceptional animals: balan jugumbil – "woman," balan buni – "fire." Class III contains edible fruits and vegetables, and Class IV contains everything that doesn’t fit into the first three.

The distribution is based on mythology. The moon and the sun are linked by a story of spouses, with the sun being a woman from class II, and the grammar preserves this story. Birds are placed in class II as dangerous, exceptional creatures. For a native speaker familiar with the traditions, each distribution is logical; to an outsider, it appears arbitrary.

Formally, class is manifested in noun markers and demonstratives. Katherine Plaster and Maria Polinsky emphasize that the four classes are realized in agreement with a demonstrative, a "noun marker." This is the same mechanism as in Swahili or Ojibwe, only with a different semantic vocabulary.

Counting and Form: East Asian Classifiers

In Mandarin Chinese, a numeral is not directly connected to a noun. A classifier is required between them: sān běn shū literally means "three-piece-bound-books," where běn refers to bound volumes. Three dogs are sān zhī māo, with the classifier zhī for animals; three rivers are sān tiáo hé, with tiáo for long, thin objects; three sheets of paper are sān zhāng zhǐ, with zhāng for flat surfaces.

The vocabulary of classifiers is more open than the vocabulary of genders. There are about 150 classifiers in common use in Mandarin, with over a hundred more used in specialized and literary speech. The universal gè serves as a fallback when a precise classifier is unavailable. Japanese is called a language of counters: animals are counted as san-biki, people as yo-nin, and long objects as go-hon, with the choice depending on the type and shape of the object being counted.

The same range yields variations. Korean, Vietnamese, and Thai require a classifier for numerals; Bengali, Assamese, Persian, Austronesian, and Mayan languages have developed their own similar systems. The bookish tradition of Classical Chinese dispensed with mandatory classifiers, but the need for them has grown in modern varieties.

The fundamental difference from gender is locality. A classifier doesn’t change the form of other words in a phrase; it stands in its place next to a numeral and remains silent in other positions. A Chinese noun is neutral in number: shū means "book" and "books" without distinction. The very act of counting requires distinguishing a single unit from a mass, and this work is performed by a classifier, not a suffix.

Ring around the root: Berber languages

The Afro-Asiatic family maintains binary gender as a rigid category. Masculine gender is usually unmarked externally; feminine gender is marked. In Semitic languages, the marker appears on the right edge of the word: Arabic tˤifl "boy" versus tˤifl-a "girl."

The Berber languages added a second marker. In Tashelhit, female is indicated by the t- marker at the beginning of a word and -t at the end: afrux "boy," t-afrux-t "girl." Every feminine noun ends with a vowel or t; feminine words do not allow other consonants in this position. The marking is regular: the prefix is always t-.

Morphologists debate the gender marker’s place. Some believe the left marker is the primary one, with the ring structure secondary. Faust and Lahrouchi discuss why t appears both before and after the root, and what dictates the prohibitions on its combination with other markers. For an outside observer, the bottom line is more important: gender is written on both sides of the word, simultaneously on both sides of the root.

Comparison of strategies

The numbers in the table are guidelines, not dogma: the counting of Swahili classes differs among different authors, and seven classes cover most of everyday speech.

Living and Erasing Systems

Class systems wear out and are rebuilt. Bantu languages have historically reduced their class set to ten, as has happened in a number of modern languages while retaining the ancient number system in Ganda. Kreyssel draws attention to morphological opacity: class and number markers in Niger-Congo have merged into indecomposable portmanteaus, which can no longer be divided into "gender" and "number" even under the most careful analysis.

Coordinative markers are adopted into borrowings in various ways. German diminutives hold the neuter gender with an iron fist even when the suffix has lost its diminutive meaning; intimate forms ending in -i remain tied to the idea of small and childish. The Ojibway dictionary allows for gender fluctuations in words for the same thing in neighboring dialects. Dyrbal logic of the fourth class is read only with a myth in hand.

The choice of strategy remains with the language. Spanish preserves two genders in the article, Swahili arranges the world into twenty shelves with verbal echoes, Mandarin takes the shape of an object when counting, Ojibwe asks about a spirit, and Fula asks whether a dwelling is round. In the 19th century, von Steinthal debated the need for such luxury; there is no single formula for the answer, only a catalog of solutions, each of which works for its own speakers.