Aikhenvald, Alexandra Y.;
Classifiers: A Typology of Noun Categorization Devices
Oxford University Press, 2003, 535 pages
ISBN 019926466X, 9780199264667
topics: | linguistics | classifier |
This impressive typological study is based on a comparison of what knowledge may be available from the study of 500 languages. The writing is rather dry, but for anyone interested in the semantics of gender or classifier classes, this work holds a wealth of information on similarity and metaphorical origins of classifier systems. What holds my nterests are these semantic aspects which are dealt with at length and comprises a fascinating study of how languages create morphological and agreement structures to classify objects according to categories that often die, leaving the categories inexplicable. While subdued in its discussion of metaphor (compare, e.g. George Lakoff's Women, fire and dangerous things, 1987)) the text highlights many of the same issues, but from a more cross-linguistic perspective.
Almost all languages have some grammatical means for the linguistic categorization of nouns and nominals. The term 'classifiers' will be used here as an umbrella label for a wide range of noun categorization devices. Different types of classifier can be distinguished by their grammatical status, degree of grammaticalization, conditions for use, meaning, kinds of origin, mode of acquisition, and tendencies towards loss.
Classifiers come in different guises.
Some languages have grammatical agreement classes, based on such core semantic characteristics as animacy, sex, or humanness. These are called Noun classes, or genders. The number of noun classes varies-from two, as in Portuguese (examples below), to ten, as in Bantu, or even to several dozen, as in some South American languages. Examples 1.1 and 1.2, from Portuguese, illustrate masculine and feminine genders which are marked on the noun itself and on the accompanying article and adjective. 1.1 o menin-o bonito ART:MASC.SG child-MASC.SG beautiful-MASC.SG 'the beautiful boy' 1.2 a menin-a bonita ART:FEM.SG child-FEM.SG beautiful-FEM.SG 'the beautiful girl'
A classifier can just categorize the noun by itself, as in the following example from Yidiny, an Australian language (Dixon 1982: 192 ff.). This is a NOUN CLASSIFIER. 1.3 bama waguja CL-PERSON man 'a man'
Other languages have special morphemes which only appear next to a numeral, or a quantifier. They may categorize the referent of a noun in terms of its animacy, shape, and other inherent properties. These are NUMERAL CLASSIFIERS. The way they are used is exemplified with a shopping list in Japanese (Rie Hasada 1995) given in Table 1.1. TABLE 1.1. Shopping list in Japanese ----------------------------------------------------------------- Shopping list Numeral Classifier Meaning of classifier ----------------------------------------------------------------- kyuuri (cucumber) hachi (8) -hon CL:ELONGATED nasu (eggplant) nana (7) -ko CL:SMALL.EQUIDIMENSIONAL hamu (ham) juu (10) -mai CL:SHEETLIKE -----------------------------------------------------------------
A special morpheme in a possessive construction may characterize the way in which the referent of a possessed noun relates to that of the possessor. This is illustrated in 1.5 and 1.6, from Fijian, an Austronesian language (Lichtenberk 1983a: 157-8). Such morphemes, underlined in 1.5 and 1.6, are called RELATIONAL CLASSIFIERS. 1.5 na me-qu yaqona ART CL:DRINKABLE-my kava 'my kava' (which I intend to drink) 1.6 na no-qu yaqona ART CL:GENERAL-my kava 'my kava' (that I grew, or that I will sell)
VERBAL CLASSIFIERS appear on the verb, but they categorize a noun, which is typically in S (intransitive subject) or O (direct object) function, in terms of its shape, consistency, and animacy. Example 1.7, from Waris, a Papuan language (Brown 1981: 96), shows how the classifier put- 'round objects' is used with the verb 'get' to characterize its direct-object argument, 'coconut'. 1.7. sa ka-m put-ra-ho-o coconut lso-to VCL:ROUND-GET-BENEFACT-IMPERATIVE 'Give me a coconut' (lit. 'coconut to-me round.one-give') --6. rarer forms: Locative / Deictic classifiers-- LOCATIVE CLASSIFIERS: occur on locative adpositions from Pa1ikur, an Arawak language from Brazil. 'on your (vertical) arm' / 'on (branch-like) tree' DEICTIC CLASSIFIERS. associated with deictics and articles are called from Mandan, a Siouan language (Barron and Serzisko 1982: 99): 'this one (lying)' / 'this one (sitting)'
The term 'classifier systems' is used to denote a continuum of methods of noun categorization. Well-known systems, such as the lexical numeral classifiers of Southeast Asia, on the one hand, and the highly grammaticalized gender agreement classes of Indo-European languages, on the other, are the extremes of this continuum.
The only useful generalizations about language are inductive generalizations. Features which we think ought to be universal may be absent from the very next language that becomes accessible.... The fact that some features are, at any rate, widespread, is worthy of notice and calls for an explanation; when we have adequate data about many languages, we shall have to return to the problem of general grammar and to explain these similarities and divergences, but this study, when it comes, will not be speculative but inductive. Bloomfield, L. (1933). Language. Holt, Rinehart & Winston.
In Ket gender can vary only for inanimates (Dul'son 1968: 62 ff., Krejnovic 1961). Then, for instance, a growing tree is masculine, a cutdown tree is inanimate; an upright tree is masculine, and a tree with a curved trunk is feminine. In languages with shape-based genders spoken in the East Sepik region of New Guinea, change in (covert) gender of most inanimate nouns signals change in shape (see Bruce 1984, on Alamblak). In Manambu (Ndu family) nouns which denote male humans and higher animates and long and thin inanimate objects are masculine, while those which denote female humans and higher animates, and short and round objects, are feminine. Consequently, nema-de wi (big-MASC house) is used to refer to 'a big house' (which is not necessarily long, or high). 24 [FN 24 A smallish woman-like man can occasionally be treated as feminine, and a largeish woman can be treated as masculine. The use of a different gender is impossible when the shape cannot be changed (turtles are 'round' and always feminine), or when the 'masculinity' is culturally important. Descent is strictly patrilineal, and so the word gwalugw 'patrilineal clan' is masculine. Morphologically, gwal-ugw is the plural form of gwal which means 'father's child (female or male)' and 'father's father'.] Along similar lines, in Kxoe (Khoisan; Heine 1982a: 198) an inanimate noun stem can be allocated to masculine or feminine gender depending on its shape: masculine is associated with big, long, rectangular, and feminine with small, round, broad, e.g. ngu 'hut', ngu-ma 'big rectangular hut (masc.)'; ngu-he 'small, round hut (fem.)'. Semantic choices may be more complex. Turkana (East Nilotic: Dimmendaal 1983: 220) has three genders: masculine, feminine, and neuter. By changing the gender of an animate noun, specific reference can be made to a male, female or a young of the species: masc. e-gete 'male antelope' fem, a-gete' 'female antelope' neut. i-gete 'small antelope of either sex' A number of nouns referring to inanimates allow variable gender. Some have two forms — masculine and feminine, e.g. 'grass' and 'tree'. Masculine forms mean 'growing, green', and feminine forms mean 'dead, dry state', e.g. masc. ni-na 'green grass' fem. na-na' 'dry grass' Some inanimate nouns have three gender forms. The opposition of the three is by size, e.g. masc. e-mor-u 'rocky mountain, big stone' fem. a-mor-u 'hill stone' neut. i-mor-u 'pebble'
[this proto-Australian language provides the catchy title of Lakoff's seminal text; "women, fire and dangersous things" constitute the gender class CL2 as described by Dixon]. Semantic assignment can be more complex. Dyirbal (Dixon 1972: 306-12) has four classes. Three are associated with one or more basic concepts: gender CL1, bayi : male humans, non-human animates; e.g. men, kangaroos, possums, bats, most snakes, most fishes, some birds, most insects, the moon, storms, rainbows, boomerangs, some spears, etc. gender CL2, balan : female humans, water, fire, fighting; women, bandicoots, dogs, platypus, echidna, some snakes+fishes, most birds, fireflies, scorpions, criekets, the hairy mary grub, anything connected with water or fire, sun and stars, shields, some spears, some trees, etc. gender CL3, balam: non-flesh food. all edible fruit and the plants that bear them, tubers, ferns, honey, cigarettes, wine, cake. gender CL4, bala : a residue class, covering everything else. parts of the body, meat, bees, wind, yamsticks, some spears, most trees, grass, mud, stones, noises and language, etc. There are also two rules for 'transferring' gender membership. By the first, an object can be assigned to a gender by its mythological association rather than by its actual semantics. Birds are classed as feminine by mythological association since women's souls are believed to enter birds after death. The second transfer rule is that if a subset of a certain group of objects has a particular important property, e.g., being dangerous, it can be assigned to a different class from the other nouns in that group. Most trees without edible parts belong to gender 4, but stinging trees are placed in gender 2. Many languages allow variable noun class marking on a number of nouns like 'baby' and 'child', on some kinship terms and some inanimates. In Dyirbal, bimu is both 'father's elder sister' (taking a feminine gender marker) and 'father's elder brother' (masculine marker). Jaja 'baby' can be specified as either masculine or feminine. (Alternately, these may be viewed as different lexemes.) The degree of variability in agreement class assignment depends on the language. In Dyirbal, variable class assignment is restricted to sexd-ifferentiable animals; the Class 1/Class 2 correlation with male/female is obligatory for humans. Each name of an animal has a fixed class membership; however, exceptionally, noun class assignment can be changed to stress the sex of a particular animal, e.g. 'to point out that a certain dog is male bayi guda can be used' (Dixon 1982: 182). Usually, guda 'dog' belongs to Class 2 (Dixon 1982: 180), and so the 'unmarked' usage would be balan guda. Very occasionally, changing noun class can create a pragmatic effect. In Dyirbal, yara 'man' belongs to Class 1, and so would be referred to as bayi yara. However, Dixon (1982: 166) reports that a hermaphrodite was once jokingly referred to as balan yara, with a feminine Class 2 marker, pointing out his female characteristics. In this case, the manipulation of noun class realized in agreement has pragmatic, as well as semantic effect.
The choice of noun class agreement depends on what aspect of the noun is highlighted. Some northern Australian languages allow different agreement possibilities for certain nouns depending on the 'viewpoint' from which the referent is seen (Evans forthcoming: 108). In Gunwinjgu kukku 'water, drink' triggers Class 4 ('neuter') agreement when it is seen as a part of the landscape: 2.21 kun-ekke kukku kun-bo-gimuk CL4:NEUTER-that water CL4:NEUTER-CL:LIQUID-big 'That water (i.e. river) is big.' When considered a drink, it is assigned to Class 3 (which also covers vegetable food). 2.22 yun yi-bongu-n man-ih kukku don't 2SG-drink-NON.PAST CL3:NON.FLESH.FOOD-this.here water In Tamil, rational nouns (which include humans, gods, and other mythical beings) cannot be conjoined [combined with "and"] with irrational ones. p.56
In wiki: Tamil grammar (wiki), nouns (and pronouns) are classified into two super-classes (tiṇai) - the "rational" (uyartiṇai), and the "irrational" (aḵṟiṇai) - which include a total of five classes (paal, which literally means 'gender'). Humans and deities are classified as "rational", and all other nouns (animals, objects, abstract nouns) are classified as irrational. The "rational" nouns and pronouns belong to one of three classes (paal) - masculine singular, feminine singular, and rational plural. The plural form for rational nouns may be used as an honorific, gender-neutral, singular form. The "irrational" nouns and pronouns belong to one of two classes (paal) - irrational singular and irrational plural.[3] As the example in the table indicates, the paal is often indicated through suffixes. Rational (uyartiṇai) He who did ceytavaṉ sg masc rational āṇpāl She who did ceytavaḷ sg fem rational peṇpāl They who did ceytavar pl rational palarpāl Irrational (aḵṟiṇai) That which did ceytatu sg irrational oṉṟaṉpāl Those ones which did ceytavai pl irrational palaviṉpāl ]
Metaphors and metaphorical transfers (see the definition in Sapir 1977: 6) are important for the semantic organization of classes and the ways in which they get extended. This can be illustrated from the semantic organization of numeral classifiers in Burmese described by Becker (1975; 1986). The structure of underlying Burmese categorization is based on an interpretation of self-see Table 12.1. As Becker (1975: 118) puts it, 'the structure underlying classification starts with the self at the centre, divides the self into head and body, and then ranges objects at four distances from the self, associating them either with the head (metaphorically top, round) or with the body (metaphorically, bottom, straight)'. The system of categorization operates on 'applied metaphors'; thus 'head to the body' is the same as 'cup to saucer', and so on; the relationship above/below is basic for categorization. Moreover, since several of the classifiers are words for parts of a tree, one can say that 'the tree is the metaphor for the person'. BURMESE NUMERAL CLASSIFIERS for inanimate objects (Table 12.1) ----------------------------------------------------------------------------- 1st Orbit 2nd Orbit 3rd Orbit 4th Orbit Self: On self Near self Far from self part of self (alienable) (inalienable) ----------------------------------------------------------------------------- ywc' (head): pain: loun: sin: hair on head head-dress round, upper things: upper things which leaf posts have circular orbit: furniture sun cup nvers, sea script arrows needles chaun (body): kwin: cha': si: hair on body body dress flat, lower things: lower things which fingers, toes body ornaments boards move in straight teeth the: mats lines: pin: folded clothes saucer vehicles sticks palmleaf for writing hunted animals twigs le': horses pens instruments used in dupes the hand; swords; thwt: musical instruments; rivers puppets roads ----------------------------------------------------------------------------- [Semantically] "the classifier is selected from a kind of conceptual space which has been historically shaped." (Becker 1986: 337) Thus, in Burmese, 'unless one knows that the traditional Burmese pictorial map of the cosmos has man located on an island, from the centre of which flows a river in a spiral course to the sea, one may question why rivers and oceans are classified here along with arrows and needles, which move in circular orbits' (Becker 1975: 118). Rhetorical uses of classifiers are further extensions of metaphors, and 'the use of classifiers in Burmese ... is in part an art and not just a grammatical convention' (Becker 1975: 113). Gender distinctions can be used metaphorically in languages with no, or almost no, agreement gender. Ilocano (Austronesian: Rubino 1997: 75-6) does not have genders; the natural gender of humans may be distinguished lexically, e.g. lalaki 'boy', babai 'girl'. These lexemes are also used to distinguish natural gender of hybrid nouns, e.g. kabsat a lalaki 'brother', kabsat a babai 'sister'. They can also cooccur with inanimates, highlighting semantic features associated with a metaphorical extension of 'masculinity' or 'femininity'. For instance, basi a lalaki 'strong sugar cane wine' is associated with masculinity; and basi a babai 'sweet sugar cane wine' with femininity. Gender languages often manipulate gender in legends. Personification of nouns with inanimate or abstract referents follows the gender which is assigned by non-semantic rules. For instance, Russian smert' 'death' is feminine and so is personified as a woman in folk tales and cartoons (Rothstein 1973: 464). [FN. which came first, can't be said] Similarly, morte 'death', a feminine noun in Portuguese, is depicted as Dona Morte (Lady Death) in cartoons. Change of genders can have a stylistic effect. Some nouns of feminine gender, essentially hostile epithets, can be applied to human beings. According to Rothstein (1973: 464), in Russian it is more insulting to call a man dura (fool: feminine) than durak (fool: masculine) (cf. (C) in §11.2.1; on the use of masculine diminutives of feminine names to express affective jocularity in Polish, see Wierzbicka 1996: 398). p.313
Some systems accept metaphorical extensions more easily than othersthis happens because systems differ in their productivity. The productivity, or vitality, of a system is measured by its ability to accept and classify new members, and reanalyse and extend the semantic range of a noun categorization device over time. Thai has an old classifier system which has great vitality and productivity; the more productive a system is, the more metaphorical and other types of extensions it accepts. In contrast, the Jacaltec system of noun classifiers can be looked upon as frozen. New loanwords are simply left unclassified; the system does not accept semantic extensions either (Craig forthcoming). Frozen or non-productive systems do not have to be in decay, though they sometimes are, as is the case of many classifier systems changing due to language obsolescence and contact.
An instructive example of prototype-and-extension in a multiple classifier system comes from the classifier tua in Thai. Carpenter (1987: 45-6) gives the following explanation of the extension and chaining principles which apply to tua (semantic features annotated by Aikhenvald in small caps): [Tua] covers a wide, but not incoherent, range of things. The articles of clothing used with /tua/ are trousers, shirts, jackets, skirts, and, less often, dresses, underwear and bathing suits. The items of furniture most likely to be used with /tua/ are tables and chairs, and less often, dressers and beds. These assignments suggests that it is THE PRESENCE OF LIMBS, giving these objects a body-like shape, that causes speakers to classify them with /tua/. Originally, the classifier was used with animals, and the PROTOTYPE is probably some good QUADRUPEDAL ANIMAL, such as a dog or a water buffalo. Tables and chairs were included on the basis of SHAPE, either because of their general quadrupedal outline or perhaps because of the specific presence of limbs. Other kinds of furniture were then added because of their shared FUNCTION with tables and chairs. Trousers and shirts were also included because of their their wearer, or specifically because they have limbs. Other kinds of clothing were included because of their SHAPES. A letter (of the alphabet) in Thai is a compound, /tua nangseu/ 'body book', so a combination of shape and repetition of the generic compound head caused letters to be classified with /tua/. Numbers were included either on the basis of shape or by their shared function with letters. The structure of the category is schematized in Diagram 12.1 (Carpenter 1987: 46). Lines indicate extensions from a prototypical member to less prototypical ones (Carpenter 1987: 45-6; cf. Downing 1996: 101-2). Underwear, bathing suit \ Skirts Cabinets, dressers, beds \ / Trousers, shirts Tables, chairs \ / ANIMALS / | \ Ghosts Mannequins Letters (alphabet) \ Numbers Fig 2.1 Structure of the tua category in Thai Similarly, ghosts were included because of their similarity with the shape of a human body. Carpenter concludes: 'The internal structure of this category, then, clearly, mixes prototypes and chains, with strongest members being those closest to an animate quadruped, but some chaining based on similarities to non-prototypical members.'
Opacity in classifier assignment can be due to semantic extensions that are explainable but unpredictable. Table 12.2 illustrates human classification in Burmese based on social status and age (Becker 1975: 116) and its extensions to human attributes. ["Explainable but unpredictable": AM: but what part of language is "predictable"? this reveals the human preference for clear rules...]
A problem may arise in how to distinguish metaphors from polysemy. For instance, in Kilivila (Senft 1996: 19) a 'dinghy' can be referred to with a classifier 'child', as well as 'wooden', e.g. ma-gudi-na waga ke-kekita (this-ci.ctut.o-this canoe ci.woonsx-small) 'this small dinghy'. This can be interpreted in two ways: either as a metaphorical extension (child > small object), or a polysemy: child, any small creature, or object. We think that a decision should be made in terms of language-internal criteria on what is literal and what is metaphorical meaning, before there is a universally accepted 'serviceable' 'clarification of the distinction ... between conventional metaphor and systematic polysemy', to 'separate (even provisionally) the literal from the metaphorical ... we need to accommodate the fact that over time metaphorical expression-systems may lose their metaphorical duality and assume the status of literal meanings, as when body-part terms become locational adpositions' (Goddard 1996: 150). This change is somewhat similar to semantic processes which take place in grammaticalization. p.314 [AM: why is it so important to make this classification? why not a continuum between polysemy and metaphor? ] Similar metaphorical extensions are found in other languages. In Bugis, tau 'human classifier' is usually employed for counting people; however, classifier lise 'small spherical objects' can be used for counting people who died in a war, presumably, because the heads of dead enemies used to be cut off and people were counted by their heads (Sirk 1983: 63). Thus, the composition of a classifier category and its expansion may be hard to predict. Though it is often possible for an analyst to suggest a rationale for the inclusion of most members of a given category, this may have not been the actual rationale employed by speakers. There is also a certain degree of unpredictability associated with social and cultural conventions. p.315
Classifier categories are semantically heterogeneous. There have been a few attempts to present the semantic organization of classifiers in terms of taxonomic trees and binary oppositions (Denny 1979a; 1979b; see criticism by Downing 1996: 125); however a superordinate-subordinate approach has limited applicability to classifier systems. The reason why superordinacy relations are limited and are unlikely to involve all the classifiers lies in the coexistence of different and cross-cutting semantic rationales (kind and quality classifiers), and on different types of extension. Some referents are included in more than one class... Thus, a sword in Japanese can be referred to by a shape-based classifier hon 'long, slender object', by a kind classifier Juri 'sword', or by a functionbased quality classifier ten 'items in an inventory, works of art'. It is also impossible to make all shape-based classifiers superordinate to kind classifiers which possess appropriate shape, because kind classifiers may unite referents of different shapes (e.g. the verbal classifier -pit used for any irregular shape object in Palikur). Other problems with creating distinct taxonomies result from a generic classifier, such as Japanese tsu, which participates in all of them. There are hardly any hierarchical relations in the assignment of noun classes (genders), deictic or locative classifiers; hence the discrepancies between scientific and folk taxonomies and noun categorization devices. Assignment of relational classifiers implies a functional categorization of objects, with no clear-cut hierarchical relations. The few Japanese numeral classifiers which have superordinate-subordinate relation are given in Table 12.3. In all these cases 'the more general term [can] be used for all members of the category denoted by a more specific term' (Downing 1996: 126).
Hierarchy in (some) Japanese numeral classifiers Superordinate-subordinate pairs
Superordinate Subordinate 1. dai 'vehicles, furniture, machines' ki 'air vehicles', ryo 'train cars', taku 'tables, desks' 2. hen 'literary work' shu 'poems' (long), ku 'poems' (short) 3. hiki 'animals' too 'large animals' 4. heya 'rooms' shitsu 'rooms', ma 'Japanese style rooms' 5. nin 'human beings' mei 'human beings' (honorific) 6. hon 'long, thin objects' juri 'swords' 7. tsu 'inanimates' most inanimate kind-classifiers
The situation is different with generics. The generic-specific taxonomy can be nature-based, or functionally defined (produce, pets, food, prey). Evidence from Yidiny shows a hierarchical relation between the two (cooccurrence of the two, i.e. 'person' plus 'man', or 'person' plus 'woman'), which is by no means universal (see §3.2.1 and 3.4 above). Thus, a taxonomic approach to classifiers may be useful, but only in limited circumstances.
Metaphoric extensions of sex- and animacy-based noun classes and classifiers to inanimates are often linked to the socio-cultural stereotypes associated with sex. In a fascinating study of sex roles as revealed through gender reference, Mathiot (1979) showed how role images of males and females are realized in the use of personal pronouns. The use of the pronouns he and she observed with inanimate referents in American English was found to correlate with a number of stereotyped features-part of the inherent image and role image American men and women have of themselves, and of each other. The semantic opposition BEAUTIFUL / UGLY manifests men's conception of women's vs. men's appearance; and the semantic opposition manifesting men's conception of women's and men's achievement potential is INCOMPETENT/ COMPETENT. Thus a beautiful flower is referred to as 'she', and an ugly cactus as 'he' (Mathiot 1979: 18-19). In contrast, the inherent image and role image American women have of themselves and of men can be formulated in one semantic opposition: MATUREhNFANTILE (Mathiot 1979: 25). Regular po1ysemy of feminine and diminutive in Afroasiatic languages is often accounted for by the low and subdued status of women in traditional Afroasiatic speaking societies (Diakonoff 1988). This is another example of a correlation between gender and social status. In a few New Guinea languages with masculine and feminine genders, masculine is associated with culturally important roles, and feminine with insignificant things. This is the case in Angave, an Angan language (Speece n.d.: 111), and in Abu' Arapesh (OUo Nekitel, p.c.). Robert Conrad (p.c.) reports that in Felefita, an Arapesh language (Torricelli phylum), masculine can replace any other gender provided the object is sufficiently important. In coastal Arapesh languages, this results in the massive expansion of masculine gender, which has become the unmarked one. Semantic extension principles, such as the Myth-and-Belief principle and Important Property principles in Dyirbal (see Dixon 1982), are based on cultural intricacies. Once they become obsolescent, the classifier assignment becomes opaque. 347
The principles of noun class assignment can be affected by cultural obsolescence. Traditional Dyirbal had four semantically assigned noun classes (§2.4.1). Young People's Dyirbal has gradually simplified its noun class system so that it became similar to the way he, she, it are used in English. The noun class referring to 'non-flesh food' was lost. The scope of noun Class 2 was reduced and came to be reserved only for females (it used to include water, fire, and things associated with fighting). Gender assignment by mythical association was lost; exceptions became regularized; and the use of the residue class was expanded (Schmidt 1985). Dahalo (Cushitic) is losing ground to Swahili, and is accordingly losing the common Cushitic distinction of masculine and feminine gender (Dimmendaal p.c.). 390 Schmidt (1985: 156-7) describes a series of changes in rules for noun class assignment in Young People's Dyirbal. Mythological association as a basis for class membership is lost. In traditional times birds were believed to be spirits of dead human females, and consequently assigned to Class Il, 'feminine'. With the loss of this belief, speakers of Young People's Dyirbal treat birds as members of the 'animate' Class 1. [40] Another distinguishing feature of change in language obsolescence and creolization is its speed: drastic grammatical restructuring sometimes happens within a short timespan (see Schmidt 1985: 213, on Young People's Dyirbal); areal diffusion within linguistic areas is usually slower (Aikhenvald 1996a). FN 40 Other changes may be due to overall simplification of the system to make it more similar to gender in English. This may explain why concept association is also abandoned (Schmidt 1985: 157). Traditional Dyirbal assigned yarra 'fishing line' and barrban 'fish spear' to Class I, by association withjahu 'fish'; speakers of Young People's Dyirbal place these two words into Class IV. with other inanimates. p. 349 Dixon, R. M. W (1968). 'Noun Classes', Lingua 21: 104-25. __ (1972). The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. __ (ed.) (1976). Grammatical Categories in Australian Languages. Canberra: Australian Institute of Aboriginal Studies. __(1982). Where Have All the Adjectives Gone? and other essays in semantics and syntax. Berlin: Mouton. __(1986). 'Noun Classes and Noun Classification in Typological Perspective', in Craig (1986a: 105-12). Mathiot, M. (ed.) (1979). Ethnology: Boas, Sapir and Whorf Revisited. The Hague: Mouton. __ and Roberts, M. (1979). 'Sex Roles as Revealed through Referential Gender in American English', in Mathiot (1979: 1-47).