Linguistics Of Classifiers Research Paper

Classifier studies became of interest to general linguists in the 1970s, following proposals to capture the universal semantic properties of classifier systems. Adams and Conklin (1973), Denny (1976), Allan (1977) have provided the framework for many of the subsequent descriptions and discussions, and can be considered as classics of the field. Adams and Conklin were the first to wade through much comparative data and claim the existence of some universal semantic properties, based primarily on data from Asian numeral classifier systems. They established the primacy of three basic shapes, which are semantically combinations of one of the major dimensional outlines of objects (1D, 2D, 3D) with a secondary characteristic of consistency and or size. This combination is directly inherited from the most common lexical sources of a basic set of classifiers, which are the primary elements of the physical world being handled for the survival of human communities (Table 2).

Denny (1976) is the work of a psychologist handling data secondhand. He offers the appealing proposal that the semantic traits of classifiers may be organized into three kinds, those of ‘social, physical and functional interaction,’ assuming that what classifiers are ‘good for’ is to signal how humans interact with the world. Under social interaction he places interaction with animate entities of our world, principally fellow human beings, classified by sex, social rank, or other categorization schema, as well as other entities such as divinities and other powers specific to a culture. In the physical interaction realm, objects of the world are classified along certain parameters linked to their nature as manipulable and manipulated objects, principally the parameter of shape. Finally, in the functional interaction realm, entities of the world are classified by the use to which they are put, such as items of clothing, hunting or fishing, transportation, for instance.

Allan (1977) is a first typological of study of so-called classifiers, based on a broad data base of fifty classifier languages. Although the reliability of the data is variable and different types of nominal classification systems are lumped together under the label of ‘classifiers,’ there is still remarkable overlap between his seven ‘categories of classification’ (material, shape, consistency, size, location, arrangement, and quanta) and Denny’s three. Two of Allan’s original statements are of particular interest for later discussions on the nature and purpose of classifier categorization; they are the fact of a total absence of color classifiers, and the constraint that the characteristics denoted by the categories of classification be perceivable by more than one of the senses alone, such as sight and touch, where sight means primarily perception of shape. It is worth noting here that to arrive at the kind of statements of universals found in the above mentioned publications meant wading through vast amounts of data from fieldwork often containing large sets of classifiers, and interpreting their semantics through approximate translations, in order to identify those universal characteristics.

Besides varying as to the semantics of the individual classifying elements, systems of classifiers vary greatly as to the number and the specificity of the classes around which the systems seem to be organized. The classes headed by classifiers can vary from very simple to very complex; they can be small or large, homogeneous or extremely heterogeneous. Homogeneous classes are those with transparent semantic motivation, while heterogeneous classes are usually considered to be composed of a core set of prototype elements to which others have been added through various means of extension. Therefore, within the literature on classifier systems one finds different labels to indicate the nature of the classes themselves. One talks, for instance, of specific, general, unique and repeater classifiers. Specific classifiers are the most common type. The classes they head are built around prototypical exemplars, with incorporation of other elements by any number of types of extensions of the class. One of the most notorious examples in the literature of a specific classifier heading a very heterogeneous class is the case of the Japanese numeral classifier hon, used prototypically for long, thin objects.

General classifiers, as their label indicates, are largely desemanticized and head large heterogeneous classes with no distinct semantic motivation. Large Asian numeral classifier systems are known to have general classifiers. At the opposite end, unique classifiers head classes of just one element. One finds in the literature examples of unique classifiers for certain animals, for instance, such as the elephant or the tiger, or even the dog, generally interpreted a posteriori as highlighting some cultural item of particular significance. Finally, the term repeater refers to classifiers that are homophonous with a noun while being either unique or specific (Table 3). The existence of repeaters is what makes for the openendedness of some classifier systems.

The semantic studies of the 1970s had a tendency to overlook the existence of different types of systems, often lumping together classifier systems with other systems. Towards the end of the twentieth century, attention has been given to the fact that classifier systems are one type of nominal classification system among several others, as argued in Grinevald (2000), in contrast to the position taken by Aikhenvald (1999). The non-lumping position argues that classifier systems are intermediate systems in a continuum of nominal classification systems that range from lexical to morphosyntactic systems. At the lexical end they may be distinguished from two types of systems, the measure terms and class terms systems, with which they are often either confused or consciously lumped. At the grammatical end, they are widely considered as distinct from the gender systems and the noun class systems, which are both essentially grammaticalized concordial systems.

All languages have lexical sets of measure terms, the expression ‘measure terms’ being used here as a cover term for what are strictly speaking measures and for types of arrangements. Examples of English measure terms include actual measure terms such as a glass of water, a pound of sugar, a slice of bread, a sheet of paper, and arrangements such as a pile of books, a group of children, a line of cars. Class terms are sets of lexical items used in lexicogenesis; they participate in compounding processes of word formation that are functionally equivalent to derivational processes. The English class terms ‘-berry’ (as in strawberry, blueberry, boysenberry, goodberry, loganberry), ‘-tree’ (as in apple tree, banana tree, cherry tree), or even ‘-man’ (as in mailman, policeman, garbage man) are the functional equivalent of French derivational suffixes such as ‘-ier’ (as in pommier ‘apple tree,’ bananier ‘banana tree,’ cerisier ‘cherry tree’) or ‘-ier /-eur’ (as in facteur ‘mailman,’ policier ‘policeman,’ eboueur ‘garbage man’).

At the grammatical end of the continuum of nominal classification systems are the gender systems found in Indo–European languages and the noun class systems of Bantu languages. Note in the examples in Table 4 the ubiquitous presence of noun class markers (here of class 5) on nouns, adjectives, and demonstratives, and in the verb as pronominal clitics. The different degrees of grammaticalization of gender noun class systems and classifier systems are generally assessed according to the set of criteria listed in Table 5.

Beyond distinguishing classifier systems from other systems of nominal classification it is important to acknowledge the existence of several subsystems of classifiers, which are usually identified and labeled primarily by their morphosyntactic locus. The best known and documented types are the ones found as elements of the noun phrase itself, such as the numeral classifiers (numeral+CL) used in quantifying expressions; the noun classifiers (CL noun) so-called for appearing with a bare noun, not linked to the expression of quantification or possession; and the genitival or possessive classifiers ( poss+CL) that are part of possessive constructions. The verb forms are the locus of two possible systems of classification: the verbal classifiers (verb-CL) which belong to the systems of nominal classification and the lesser known verb classifiers, which actually classify types of verbs rather than nominal arguments.

Noun classifiers; Jakaltek-Popti’ (Craig 1986, p. 264)

xil                     naj                   xuwan              no7                  lab’a

saw                   CL                   John                 CL                   snake

‘(man) John saw the (animal) snake’

Numeral classifiers; Ponapean (Rehg 1981, p. 130)

Pwihk riemen                            ‘two pigs’

pig 2+CL: animate

tuhke rioapwoat             ‘two trees’

tree 2+CL: long

Genitive classifiers; Ponapean (Rehg 1981, p. 184)

kene-i mwenge                          ‘my(edible) food’

CL-GEN.1 food

were-i pwoht                             ‘my(transport) boat’

CL-GEN.1 boat

Verbal Classifiers; Cayuga (Mithun 1986, pp. 386–8)

1. ohon’atatke: ak-hon’at-a:k

it-potato-rotten past.I-CL-eat ‘I

(potato)ate a rotten potato’

1. so:wa:s akh-nahskw-ae’

dog I-CL:domestic.animal-have

‘I have a ( pet) dog’

1. skitu ake’-treh-tae’

skidoo I-CL: vehicle-have

‘I have a car’

Gunwinggu (Oates 1964

in Mithun 1986, p. 389)

1. gugu ga-bo:-mangan

water it-CL: liquid-fall

‘water is falling’

The major argument to prove the existence of different types of classifiers is the co-occurrence of several independent systems in the same language. These are systems with different inventories of classifier morphemes, different semantics and different morphosyntactic loci, such as the coexisting numeral and possessive classifier systems of Micronesian languages like Ponapean, for instance. Although much remains to be done in terms of the study of the semantics of classifiers, preliminary exploration of a correlation between the major morphosyntactic types of classifiers known and their semantic profiles point to the following alignment (Grinevald 2000).

Shape seems to be the dominant semantic parameter in numeral classifier systems, while function is the major semantic parameter of genitival classifier systems, and material the major one of noun classifier systems, in the following pattern:

(a) numeral classifiers physical categories:

two-ROUND oranges;

three-LONG RIGID pencils;

four-FLAT FLEXIBLE blankets

(b) genitive classifiers functional categories

my-EDIBLE food;

his-DRINKABLE potion;

their-TRANSPORT canoe

(c) noun classifiers material essence categories

an ANIMAL deer;

the ROCK cave;

MAN musician

The claim that there exist different types of classifiers raises two questions about the function of classifiers: one is the unavoidable one about the common function of classifiers in general in the languages that avail themselves of such systems. The other arises from the identification of different types of classifier systems and concerns the distinct functions that those different types of classifier fulfill, in view of their different morphosyntactic loci and semantic profiles. It has to be noted that when the issue of the function of classifier systems has been addressed in the literature, it has generally been, admittedly or not, from the perspective of numeral classifiers only, without regard for the variety of classifier types. In this context, (numeral) classifiers have been seen as markers of individuation or unitizing that operate in languages in which the semantics of nouns is taken to be essentially equivalent to that of mass or concept nouns of Indo–European languages.

One proposal dealing with the distinct functions of the various types of classifier systems has been an analysis of noun phrases as layered structures parallel to verbal layered structures in which different types of operators are found. In this framework numeral classifiers were considered to be quantification operators with semantics that appealed to the handling of items to be counted, hence primarily shape and physical characteristics. It was further argued that possessive classifiers were localizing operators, and their semantics linked to the function of the items appropriated, while the noun classifiers were quality operators and as such appealed to the material or essence of the items that appeared as arguments of discourse (Grinevald 2000). Much remains to be done to document the variety of classifier systems well enough to be able to address this issue comprehensively.

One of the major challenges of classifier studies is that the essentially intermediate nature of classifier systems, as secondary linguistic systems halfway between lexicon and grammar, means great variability of the systems. Acknowledging the need to take into account the inherent dynamics of classifier systems makes the descriptive task both more onerous and more productive if comparative and typological work is to proceed properly. The need is felt to include a number of dynamic variables to handle the description of specific classifier systems, by attending to their place in the overall grammar of the language within which they develop and are used. These variables include:

(a) their degree of grammaticalization: within each subtype of classifier system, one can identify systems at different stages of grammaticalization. For instance, incipient systems of noun classifiers can be found on the Australian continent next to well-established ones, meanwhile the numeral classifiers of the Chibchan languages of Central America are much more grammaticalized than those of Asian languages.

(b) the age of the system: some systems are very old, e.g., the Chinese system of numeral classifiers, while others can be argued to be only several centuries old, like the q anjob alan-Mayan noun classifiers.

(c) the productivity of a classifier system must be considered too, independently of its age. For instance the Thai numeral classifier system, which is very old, is also very productive: it is open and adapting to the language of modern life, while the noun classifier system of Jakaltek-Popti , which is not very old, seems frozen and unable to cope with the classification of modern imports and products.

(d) the particular classifier system needs to be assessed in the context of the common phenomenon of areal spread of such systems. The spread can operate either through the actual borrowing of a system, morphology included, as was the case with the expansion of the original Chinese numeral classifier system into its surrounding regions, or through the borrowing of the idea and motivation for the development of such systems, as seems to have taken place between the q`anjob`alan languages of Guatemala and their neighboring Mamean languages.

While early propositions of matching classifier systems with morphological types of languages were not enlightening, there is indeed a tendency for different types of nominal classification systems in general, and of different types of classifiers in particular, to distribute themselves in clusters around the world. For instance, gender systems are a widespread phenomenon in Indo–European languages, while noun class systems were originally mostly known from Bantu languages. In addition, they have been described for languages of Australia and Papua New Guinea, and are perhaps more widespread than previously recognized in Amazonia.

Turning our attention to classifier subtypes, numeral classifiers are best known for their presence in South East Asian languages, but have also been identified in America, in particular Mesoamerica. Noun classifiers appear to be a rare type, mostly identified in Mesoamerica and Australia, while possessive classifiers are the hallmark of Micronesian languages, although they are also found in various parts of America. As for verbal classifiers, they have been documented for North American languages and for signed languages, though it is sometimes difficult to establish how segmentable and identifiable the actual classifier morphemes are in verbal predicates.

Classifier studies raise difficult methodological issues. There is the first-degree challenge of the fieldwork to be done to produce descriptions of these systems. Fieldwork faces the problem of the ethnocentrism of the semantic analysis. This will continue as long as much of this work is done through translation, and is still largely carried out by linguists who are native speakers of languages where the phenomenon does not exist. ( Work done on South East Asian languages is to some extent an exception to this problem.) For instance, it is often difficult to say whether the semantics of a classifier is one of the strictly physical characteristic of shape or one of function, since certain shapes naturally lend themselves to certain functions. Are objects hollowed or made to be concave to be considered for their shape, as round hollow objects, or for their function, as recipients and containers? The basic issue is, of course, whether the right questions are being asked of these systems in the first place. There is the further challenge of the fundamental lexico-grammatical nature of such systems, with their common open-endedness and subtle discourse functioning that require extensive studies all too rare in the descriptive tradition. And there is the second-degree challenge facing linguists working on secondhand data of uncertain reliability and common incompleteness, particularly in terms of the pervasive dynamics and complex internal typology of such systems, of the kind introduced in this research paper.

A number of publications are shaping the field of classifier studies in the context of the wider discipline of nominal classification. Craig (1986) meant to begin to confront various approaches to the study of nominal classification systems and included articles on the acquisition, historical development, discourse function, and semantics of classifier systems. Senft (2000) is a more recent collection (based on a 1993 working conference) where the issues of typology, grammaticalization, and function of classifiers are further elaborated. Aikhenvald (1999) is a substantial monograph which attests to the importance of Amazonian data, for the richness of its classification systems and the challenges they pose to the proposed typology, and which underlines in general the extreme complexity of the inter-relation of systems in many languages. Sands (1995) is a useful survey of nominal classification systems in Australia which reveals two interesting phenomena: one is the parallel development, in different languages, of concordial noun class systems and noun classifier systems out of the same lexical material of generic nouns; and the other is the documentation of the various stages of the evolution of noun classifier systems from the discourse sensitive use of generic nouns and through the increasingly frequent collocation of generics and nouns in classifier constructions. Bisang (1996, 1999) provides overviews of the grammaticalization dynamics through which the classifier systems of East and South East Asian languages arose, within the areal typological frame called for by the language contact situation of the region.

In terms of an agenda for the development of classifier studies in the twenty-first century, the work is proceeding on two fronts. On the linguistic fieldwork front, there is still an enormous need for more comprehensive descriptions. And, given the fact that many systems worth investigating are to be found in the languages of Amazonia, Australia, and Papua New Guinea, the challenging nature of this fieldwork has to be kept in mind. Work on nominal classification processes in signed languages is also under way; a collective volume on that topic in a cross-linguistic perspective is scheduled to appear following a working conference in 2000 (Emmorey in press).

A better understanding of the general phenomenon of classifiers should also emerge from the confrontation of what is now known of nominal classification systems with the much less known phenomenon of verb classification. This phenomenon has been described for some Australian languages, but has not come fully to the attention of linguists interested in nominal classification. However, the identification of a similar phenomenon in South American languages such as the Barbacoan languages of Ecuador and its ongoing description should facilitate further exploration of the parallels between the organization of nominal and verbal linguistic expressions and the similarity in the function of their operators, in particular that of their respective classifier systems.

On the more theoretical front, various debates are open and in need of further consideration. They include taking a position on the a priori lumping or not lumping together of most nominal classification systems. Lumping means subsuming, under the label of classifiers, cases of class terms, measure terms, noun classes, as well as classifiers. This position is defensible in terms of how the new data collected in the field, particularly data on languages until recently never described, appear startling if not overwhelming because of overlapping layers of elements of supposedly various types of systems. Alternatively one can opt to tease apart different types of classification systems using as a reference clear cases of classifier systems with the characteristics given above in Table 5.

To handle the cases of data overlaps one can proceed with a number of tools from a functional-typological approach to the study of language. One is the notion of prototype, with its accompanying concept of fuzzy boundaries; this allows for some systems to be intermediate between two systems, such as noun class and classifier. Another is the notion of the grammaticalization dynamics, inter-and intra-types of nominal classification systems, allowing in particular for variations along this parameter within the same type, which can change an analysis of multiple classifier systems to one of incipient noun class system for instance. A third is the notion that several systems may indeed co-exist, the original system and another one evolved in part from it, resulting in homophonous morphemes belonging to various systems, such as class terms and classifiers.

Another major issue due for more debate in the twenty-first century is the issue about how much the linguistic phenomenon of classifiers is linked to categorization of referents in the world linguistic classification of nouns. This has been best articulated by Lucy (1992), while Foley (1997, Chap. 12) provides a good summary of recent experimental work on the cognitive impact of classifiers in categorization. It is clear that before constructing experimental studies that use classifiers in the search for the nature of the links that hold between language and cognition, an assessment of the degree of grammaticalization of the systems is needed. Ongoing discussions of so-called classifiers in signed languages underline also their fundamental discourse functions of referent identification and referent tracking, and the likelihood that the term ‘classifiers,’ now well established in the literature, both in its narrower and wider scope, may very well be a misnomer which delays a better grasp of their role in language.

Bibliography:

1. Adams K L, Conklin N F 1973 Towards a theory of natural classification. Papers from the Regional Meeting of the Chicago Linguistic Society 9: 1–10
2. Aikhenvald A 1999 Classifiers. A Typology of Noun Categorization De ices. Clarendon Press, Oxford, UK
3. Allan K 1977 Classifiers. Language 53: 285–310
4. Berlin B 1965 Tzeltal Numeral Classifiers. Mouton, The Hague
5. Bisang W 1996 Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in East and Mainland South East Asian languages. Studies in Language 20–3: 519–97
6. Bisang W 1999 Classifiers in East and Southeast Asian languages: Counting and beyond. In: Gvozdanovic J (ed.) Numeral Types and Changes Worldwide. Mouton de Gruyter, Berlin
7. Craig C G (ed.) 1986 Noun Classes and Categorization. John Benjamins, Amsterdam
8. de Leon L 1988 Noun and numeral classifiers in Mixtec and Tzotzil: A referential view. Ph.D thesis, University of Sussex, UK
9. Denny J P 1976 What are noun classifiers good for? Papers from the Regional Meeting of the Chicago Linguistic Society 12: 122–32
10. Dixon R M W 1982 Noun classifiers and noun classes. In: Dixon R M W (ed.) Where Ha e all the Adjectives Gone? And Other Essays on Semantics and Syntax. Mouton, The Hague, pp. 211–33
11. Dixon R M W 1986 Noun classes and noun classification in typological perspective. In: Craig C G (ed.) Noun Classification and Categorization. John Benjamins, Amsterdam, pp. 105–12
12. Emmorey K (ed.) in press Perspectives on Classifier Constraints in Sign Languages. Erlbaum, Mahwah, NJ
13. Foley W 1997 Anthropological Linguistics: An Introduction. Blackwell, Oxford, UK
14. Greenberg J H 1978 How does a language acquire gender markers? In: Greenberg J H (ed.) Universals of Human Language, Vol. III. Stanford University Press, Stanford, CA, pp. 47–82
15. Grinevald C 2000 A morphosyntactic typology of classifiers. In: Senft (ed.) Nominal Classification. Cambridge University Press, Cambridge, UK, pp. 50–92
16. Lucy J 1992 Grammatical Categories and Cognition. Cambridge University Press, Cambridge, UK
17. Matsumoto Y 1993 Japanese numeral classifiers: A study of semantic categories and lexical organization. Linguistics 31: 667–713
18. Mithun M 1986 The convergence of noun classification systems. In: Craig C G (ed.) Noun Classification and Categorization. John Benjamins, Amsterdam, pp. 379–97
19. Oates L 1964 A tentative description of the Gunwinggu language. Oceania Linguistic Monographs 10, University of Sydney, Australia
20. Rehg K 1981 Ponapean Reference Grammar. University Press of H7awaii, Honolulu, HI
21. Sands K 1995 Nominal classification in Australia. Anthropological Linguistics 37: 247–346
22. Senft G 2000 Systems of Nominal Classification. Cambridge University Press, Cambridge, UK