Chapter Sex-based and Non-sex-based Gender Systems

by Greville G. Corbett

Linguistic gender systems are frequently linked to biological sex. This is not the only possibility; alternatives occur, particularly in some of the larger gender systems. 

1. Defining the values

We saw in Chapter 30 how, in many languages, nouns may be divided into groups according to the agreements they take, even when we control for other factors such as number and case. We should then ask whether these groups are arbitrary. The answer is that there is always a semantic “core” to the system. That is, there is an overlap between the nouns which take a particular set of agreements and some semantic feature. (This overlap may be greater or smaller, as we shall see in Chapter 32.) In the familiar systems such as French and German, and indeed in the majority, the link is to biological sex. This was illustrated from Russian, in Chapter 30, and can be seen again in Bininj Gun-Wok (Gunwinygic; northern Australia; Evans et al. 2002: table 1). Here we see agreement of the adjective in gender in the Kunwinjku dialect of the language (older speakers):

(1) Bininj Gun-Wok 








‘good man’ 








‘good woman’ 








‘good cheeky yam’ 








‘good water’ 

Nouns denoting males, like bininj  ‘man’, are found in gender ‘I’, and take the ‘I’ agreement marker na-. Male sex forms the semantic core of this gender, but there are other nouns found in it too; for example, it is the default gender for non-human animates. Nouns denoting females, like daluk  ‘woman’, are found in gender ‘II’, and take the ‘II’ agreement marker ngal-. Again, though female sex forms the semantic core of this gender, there are other nouns in it too, mainly denoting lower animates.

Such systems present considerable variety, as we shall see. However, not all gender systems are sex-based. Consider these examples from Maasina Fulfulde (Atlantic; Breedveld 1995: 295): 

(2) Maasina Fulfulde 








‘the person’ 







‘the elephant’ 








‘the boat’ 







‘the tree’ 

Fulfulde/Fula has around twenty genders (dialects vary: Guinean Fula, the dialect recorded on the map, is particulary rich); the singular forms of some genders are illustrated here. They are usually labelled by capitals, according to the form of the article they take (e.g. 'o or ba); adjectives, demonstratives, numerals and pronouns also show similar agreement. We can see that the form of the article changes to mark agreement with the noun, as in French or German. The noun has a class suffix, which may be similar in form to the article, or it may be in conflict with it (Breedveld 1995: 296), in which case, of course, it is the agreement which tells us the gender. The gender 'o/'ba (that is, nouns taking 'o in the singular and 'ba in the plural) includes dɛbbɔ  ‘woman’, and indeed a whole range of nouns denoting humans. Thus it has a clear semantic core; however, sex is not a part of it: nouns denoting human males and human females are found in the same gender. Other genders overlap with semantic categories to a greater or lesser degree; thus together with ‘elephant’ we find other large animals, wild and domesticated, and the word for ‘field’. Sex plays a minor role in the Fulfulde ndi gender, where according to Breedveld (1995: 329-330) there are nouns denoting signs of wealth, seeds, slave names and domestic male animals; note, however, that not all domestic male animals are in this gender (for example, ‘male donkey’ belongs with ‘elephant’ (1995: 411)).

There are various types of gender systems where biological sex is not the semantic core. These are all based on some notion of animacy. When required there is no problem about expressing distinctions based on sex in such languages, by the use of separate lexical items (whether based on different roots or derivationally linked) or by qualifying items with the meaning ‘male/female’. The use of the term gender is still fully appropriate for systems based on animacy, because structurally the systems are fully comparable. Note too that gender derives etymologically from Latin genus, via Old French gendre, and originally meant ‘kind’ or ‘sort’.

The values for the map, which is based on the same sample of 256 languages as Map 30A, are as follows:

Values of Map 31A. Sex-based and Non-sex-based Gender Systems
No gender system 145
Sex-based gender system 84
Non-sex-based gender system 28

2. Variety in sex-based systems

Though a major division is between sex-based and non-sexed-based systems, we should not neglect the variety within these groupings.  

First there is variety in how closely the grammatical gender system relates to the corresponding semantic category. At one end of the scale, in sex-based systems the genders may match the semantic category almost completely. Examples can be found in Dravidian languages, where for instance in Tamil, it is almost correct to say that nouns denoting male humans are masculine, and masculine nouns denote male humans. There is more to be said, since this gender also includes male deities, but the statement captures the essence of the system. Compare this with the very different system of a typical Indo-European language like French or Russian, where it is also correct to say that nouns denoting males are typically masculine, but where the masculine nouns include a large proportion of nouns which do not denote males. We may also find exceptions where nouns denoting humans of one sex are found in the ‘wrong’ gender. Typically, however, these nouns are not fully in that gender but take agreements of more than one type and are "hybrid nouns" (Corbett 1991: 176-181).

Second, these languages may set the threshold for "sex differentiability" at different points. Sex distinctions extend to insects and plants, but no language has been reported as including reference to their biological sex within a grammatical system. Humans are most interested in the sex of other humans, and the threshold may well be set here. This is the case for Tamil, where both ‘bull’ and ‘cow’ are in the neuter gender. Similarly in Tidore (West Papuan; North Moluccas, Indonesia), a rooster is treated as grammatically neuter, as is a pregnant goat (van Staden 2000: 77-78 and personal communication). Other languages set the threshold lower. In Russian, sex-differentiability extends to creatures whose sex matters to humans (that is, primarily animals which humans breed), or where the difference is striking (as with lions); nouns which fall below the threshold may be in any gender (thus ‘shark’ is feminine and ‘dolphin’ is masculine).  

There are some curious effects of status, usually within the feminine gender. Lak (Daghestanian; central Daghestan highlands) has four genders, in broad outline: male rationals (I), female rationals (II), other animates (III), though this has other members too, including many inanimates, and a residue gender (IV), which also includes a few animates. There was an important exception, namely duš  ‘girl, daughter’, a member of gender III instead of the expected gender II. Gender III agreements became a sign of politeness when addressing young women (Xajdakov 1963: 49-50), particularly those earning their own living, and nouns denoting them have been transferred to gender III. This usage has extended so that now gender III agreement forms are appropriate for any woman outside the immediate family. Within the family, older women such as ninu  ‘mother’ and amu  ‘grandmother’ are addressed and referred to using gender II forms. Thus gender II is semantically restricted and is left with extremely few nouns in it. Something comparable has happened in Konkani (Indo-European; west coast of India; Miranda 1975: 208-13), where the word for ‘girl’ was neuter. Where human referents are concerned, the neuter has become the gender for young females (or those relatively younger from the speaker’s standpoint), while the feminine is for old, or relatively older, females. A similar change in the core meaning of genders has occurred in some southern Polish dialects (Zaręba 1984-85). In several of these dialects, nouns denoting girls and unmarried women (irrespective of age), and including hypocoristics, are of neuter gender. Neuter agreements are employed when unmarried women are addressed, and they use them for self-reference. In a smaller area, to the south-west of Kraków, instead of the neuter the masculine is used. In both types of dialect, the feminine is used for married women. The change from neuter or masculine to feminine for a particular woman occurs immediately after the church wedding ceremony; the communities involved are small, and so there is no difficulty about knowing who is married and who is not (A. Zaręba, p.c.). The meaning of the feminine has changed in both dialect types, being restricted now to denote married women. (Feminine nouns which are not semantically motivated also remain feminine.) For further details on all these, and suggestions as to how they have arisen, see Corbett (1991: 24-26, 99-101).

3. Variety in non-sex-based systems

These systems are all based on some type of animacy. The first source of variation is the threshold for differentiability. One possible distinction is human versus non-human, as we saw in Fulfulde. This is found more widely in Niger-Congo; a clear account of these systems can be found in Welmers (1973: 159-183). Several Bantu languages have lowered the threshold, so that animals go together with humans to give an animate gender (this change in languages of the coastal area of Kenya and northern Tanzania is documented in Wald 1975). The other main area for non-sex-based gender systems is the Algonquian family of Canada and the northern United States. These two-gender systems distinguish animate and inanimate. As with sex-based systems, a source of variation is the type of “leaks” that occur into semantically motivated genders. In Algonquian there are various examples of apparent non-animates which are treated as animates. For example, in Eastern Ojibwa, nouns denoting persons, animals, spirits and trees are animate: enini  ‘man’, enim  ‘dog’, menito:  ‘manitou’, mettikumi:šš  ‘oak’ (Bloomfield 1957: 31-32). Others are inanimate. But several nouns are unexpectedly animate, including: enank  ‘star’, meskomin  ‘raspberry’, ekkikk  ‘kettle’. Various approaches to this unexpected deviancy in animacy could be taken. One might simply treat them as lexical exceptions; they are few in number, and such small groups of exceptions are often found in gender systems. Others have suggested that animacy is only a part of the explanation, and that the system must be seen in terms of a different world view, in which "power" is the dominant element, and where nouns treated as animate have at some point been viewed as denoting powerful entitities. A proponent of this view is Black-Rogers (1982); see Corbett (1991: 20-24) for further data and sources.

A second source of variation is the nature of the semantic basis for the other genders. In larger systems, as in Niger-Congo, the non-human or non-animate genders (of which there may be several) may in turn be semantically justified to a greater or lesser extent. This is an ongoing debate, in that some see the apparently mixed collection of nouns in certain genders as reflecting some conceptual cohesion, while others treat them as relics of earlier systems, now linked by formal rather than semantic similarity. 

4. Geographical distribution

The picture is relatively clear. Sex-based systems are found in almost all areas where there is gender. Of the 112 languages with gender in the sample, three quarters (84) have sex-based systems. The main non-sex-based area is covered by the extensive Niger-Congo family in western, central and southern Africa, which contributes 17 of the 28 non-sex-based systems. Most of these have five or more genders, but Grebo and Koromfe have three. The other substantial non-sex-based area is that of the Algonquian family of North America, reflected in our sample by Plains Cree, Ojibwa and Passamaquoddy-Maliseet. Elsewhere there is Ju|'hoan, representing Khoisan languages from southern Africa. In Austro-Asiatic, the languages in our sample are Mundari and Nicobarese. In Australia there is Wardaman. Two Carib languages (Hixkaryana and Macushi) are both of this type, as is Lealao Chinantec (Oto-Manguean; Mexico). The wide scatter of these languages shows that animacy is a viable basis for gender systems. Nevertheless, it is overshadowed by sex-based-systems.

5. Theoretical implications

The existence of systems which are structurally similar, but which have different semantic content, offers exciting perspectives for psycholinguistic research. The longevity of the non-sex-based systems within Niger-Congo has implications for historical linguistics, since it shows that a minority system can maintain itself and develop in a variety of ways without adopting the majority scheme. Finally, for those interested in language and gender in the sociological sense, these direct reflections of biological sex in many languages, with the "control group" of languages with similar gender structures but without the sex component, should provide a valuable source of data.