In general, there are three main ways by which natural languages categorize nouns into noun classes:
Usually, a combination of the three types of criteria is used, though one is more prevalent.
Noun classes form a system of grammatical agreement. The fact that a noun belongs to a given class may imply the presence of:
Modern English expresses noun classes through the third person singular personal pronouns he (male person), she (female person), and it (object, abstraction, or animal), and their other inflected forms. The choice between the relative pronoun who (persons) and which (non-persons) may also be considered a way of categorizing nouns into noun classes. A few nouns also exhibit vestigial noun classes, such as actress, where the suffix -ess added to actor denotes a female person. This type of noun affixation is not very frequent in English, but quite common in languages which have the true grammatical gender, including most of the Indo-European family, to which English belongs.
When noun class is expressed on other parts of speech, besides nouns and pronouns, the language is said to have grammatical gender.
In languages without inflectional noun classes, nouns may still be extensively categorized by independent particles called noun classifiers.
Common criteria that define noun classes include:
A more or less discernible correlation between noun class and the shape of the respective object is found in some languages, even in the Indo-European family.
The Ojibwe language and other members of the Algonquian languages distinguish between animate and inanimate classes. Some sources argue that the distinction is between things which are powerful and things which are not. All living things, as well as sacred things and things connected to the Earth are considered powerful and belong to the animate class. Still, the assignment is somewhat arbitrary, as "raspberry" is animate, but "strawberry" is inanimate.
In Navajo (Southern Athabaskan) nouns are classified according to their animacy, shape, and consistency. Morphologically, however, the distinctions are not expressed on the nouns themselves, but on the verbs of which the nouns are the subject or direct object. For example, in the sentence Shi’éé’ tsásk’eh bikáa’gi dah siłtsooz "My shirt is lying on the bed", the verb siłtsooz "lies" is used because the subject shi’éé’ "my shirt" is a flat, flexible object. In the sentence Siziiz tsásk’eh bikáa’gi dah silá "My belt is lying on the bed", the verb silá "lies" is used because the subject siziiz "my belt" is a slender, flexible object. See Navajo language: Classificatory Verbs for more discussion.
Koyukon (Northern Athabaskan) has a more intricate system of classification. Like Navajo, it has classificatory verb stems that classify nouns according to animacy, shape, and consistency. However, in addition to these verb stems, Koyukon verbs have what are called gender prefixes that further classify nouns. That is, Koyukon has two different systems that classify nouns: (a) a classificatory verb system and (b) a gender system. To illustrate, the verb stem -tonh is used for enclosed objects. When -tonh is combined with different gender prefixes, it can result in daaltonh which refers to objects enclosed in boxes or etltonh which refers to objects enclosed in bags.
The Dyirbal language is well known for its system of four noun classes, which tend to be divided along the following semantic lines:
The class usually labeled "feminine", for instance, includes the word for fire and nouns relating to fire, as well as all dangerous creatures and phenomena. This inspired the title of the George Lakoff book Women, Fire and Dangerous Things (ISBN 0-226-46804-6).
The Ngangikurrunggurr language has noun classes reserved for canines, and hunting weapons, and the Anindilyakwa language has a noun class for things that reflect light. The Diyari language distinguishes only between female and other objects. Perhaps the most noun classes in any Australian language are found in Yanyuwa, which has 16 noun classes.
The indigenous Australian language Yanyula has 15 noun classes, including nouns associated with food, trees and abstractions, in addition to separate classes for men and masculine things, women and feminine things. In the men's dialect, the classes for men and for masculine things have simplified to a single class, marked the same way as the women's dialect marker reserved exclusively for men.
Some members of the Northwest Caucasian family, and almost all of the Northeast Caucasian languages, manifest noun class. In the Northeast Caucasian family, only Lezgian, Udi, and Aghul do not have noun classes. Some languages have only two classes, while the Bats language has eight. The most widespread system, however, has four classes: male, female, animate beings and certain objects, and finally a class for the remaining nouns. The Andi language has a noun class reserved for insects.
Among Northwest Caucasian languages, Abkhaz shows a human male/human female/non-human distinction. Ubykh shows some inflections along the same lines, but only in some instances, and in some of these instances inflection for noun class is not even obligatory.
In all Caucasian languages that manifest class, it is not marked on the noun itself but on the dependent verbs, adjectives, pronouns and prepositions.
According to Carl Meinhof, the Bantu languages have a total of 22 noun classes called nominal classes (this notion was introduced by W.H.J.Bleek). While no single language is known to express all of them, most of them have at least 10 noun classes. For example, by Meinhof's numbering, Shona has 20 classes, Swahili has 15, Sesotho has 18 and Luganda has 17.
Specialists in Bantu emphasize that there is a clear difference between genders (such as known from Afro-Asiatic and Indo-European) and nominal classes (such as known from Niger-Congo). Languages with nominal classes divide nouns formally on the base of hyperonomic meanings. The category of nominal class replaces not only the category of gender, but also the categories of number and case.
Critics of the Meinhof's approach notice that his numbering system of nominal classes counts singular and plural numbers of the same noun as belonging to separate classes. This seems to them to be inconsistent with the way other languages are traditionally considered, where number is orthogonal to gender (according to the critics, a Meinhof-style analysis would give Ancient Greek 9 genders). If one follows broader linguistic tradition and counts singular and plural as belonging to the same class, then Swahili has 8 or 9 noun classes, Sesotho has 11 and Luganda has 10.
The Meinhof numbering tends to be used in scientific works dealing with comparisons of different Bantu languages. For instance, in Swahili the word rafiki ‘friend’ belongs to the class 9 and its "plural form" is marafiki of the class 6, even if most nouns of the 9 class have the plural of the class 10. For this reason, noun classes are often referred to by combining their singular and plural forms, e.g., rafiki would be classified as "9/6", indicating that it takes class 9 in the singular, and class 6 in the plural.
However not all Bantu languages have these exceptions. In Luganda each singular class has a corresponding plural class (apart from one class which has no singular–plural distinction; also some plural classes correspond to more than one singular class) and there are no exceptions as there are in Swahili. For this reason Luganda linguists use the orthogonal numbering system when discussing Luganda grammar (other than in the context of Bantu comparative linguistics), giving the 10 traditional noun classes of that language.
The distinction between genders and nominal classes is blurred still further by Indo-European languages that have nouns that behave like Swahili's rafiki. Italian, for example, has a group of nouns deriving from Latin neuter nouns that acts as masculine in the singular but feminine in the plural: il braccio/le braccia; l'uovo/le uova. (These nouns are still placed in a neuter gender of their own by some grammarians.)
Here is a complete list of nominal classes in Swahili:
|Class number||Prefix||Typical meaning|
|1||m-, mw-, mu-||singular: persons|
|2||wa-, w-||plural: persons (a plural counterpart of class 1)|
|3||m-, mw-, mu-||singular: plants|
|4||mi-, my-||plural: plants (a plural counterpart of class 3)|
|5||ji-, j-, Ø-||singular: fruits|
|6||ma-, m-||plural: fruits (a plural counterpart of class 5, 9, 11, seldom 1)|
|7||ki-, ch-||singular: things|
|8||vi-, vy-||plural: things (a plural counterpart of class 7)|
|9||n-, ny-, m-, Ø-||singular: animals, things|
|10||n-, ny-, m-, Ø-||plural: animals, things (a plural counterpart of class 9 and 11)|
|11||u-, w-, uw-||singular: no clear semantics|
|15||ku-, kw-||verbal nouns|
|16||pa-||locative meanings: close to something|
|17||ku-||indefinite locative or directive meaning|
|18||mu-, m-||locative meanings: inside something|
Class prefixes appear also on adjectives and verbs, e.g.:
The class markers which appear on the adjectives and verbs may differ from the noun prefixes:
In this example, the verbal prefix a- and the pronominal prefix wa- are in concordance with the noun prefix m-: they all express class 1 despite of their different forms.
The Zande language distinguishes four noun classes:
There are about 80 inanimate nouns which are in the animate class, including nouns denoting heavenly objects (moon, rainbow), metal objects (hammer, ring), edible plants (sweet potato, pea), and non-metallic objects (whistle, ball). Many of the exceptions have a round shape, and some can be explained by the role they play in Zande mythology.
In languages with genders, the gender is a selective category for noun. It means that all nouns must be assigned to a gender, and thus all nouns may be divided into groups, considering their gender. For instance, the Polish word ręcznik ‘towel’ is of masculine gender, which includes words for male human beings and animals; encyklopedia ‘encyclopaedia’ is of feminine gender, which includes words for female human beings and animals; and krzesło ‘chair’ is of neuter gender. The word "gender" derives from Latin genus, which is also the root of genre, and originally meant "kind", so it does not necessarily have a sexual meaning. For instance, in Swedish nouns are either of common or neuter gender; words for both males and females are assigned to common.
A language has grammatical gender when changes in the gender of a noun necessarily induce morphological changes in adjectives and other parts of speech (such as verbs) that refer to that noun. For adjective and some other inflecting words, gender is an inflected category. It means that (in languages with genders) adjectives are inflected by genders, or change their forms depending on gender of the noun to which they refer. In yet other words, when a noun belongs to a certain gender, other parts of speech that refer to that noun have to be inflected to be in the same class. These obligatory changes are called gender agreement.
In Polish, the adjective which means ‘big, large’ has three forms (in nominative singular), one for masculine, one for feminine, and one for neuter gender: duży ręcznik ‘big towel’, duża encyklopedia ‘big encyclopaedia’, duże krzesło ‘big chair’.
Some languages, such as Japanese, Chinese and the Tai languages, have elaborate systems of particles which classify nouns based on shape and function, but are free morphemes rather than affixes. Because the classes defined by these classifying words are not generally distinguished in other contexts, many if not most linguists take the view that they do not create grammatical genders.