The Parameterization of Plural Markings in Classifier Languages

Although the presence and absence of classifiers is a major difference which draws a distinction between classifier languages and non-classifier languages, another significant difference can be found in the properties of plural markers. The distinct roles of plural markers in classifier languages are investigated in this paper. It will be shown that the plural markers in these languages mark more than the plurality of referents of nouns, unlike plural markers in non-classifier languages such as the –s in English. With regard to the fundamental nature of the plural marker in classifier languages, we raise the following questions: (i) are plural markers in classifier languages really optional?; (ii) if not, what kind of semantic or pragmatic function(s) do they perform?; and, (iii) in what way(s) are they different from plural markers in non-classifier languages? The goal of this paper is to answer these questions and to account for the distinct properties of plural markers in classifier languages. Our discussion of these issues is organized as follows: In sections 2.1 and 2.2, the properties of the Chinese plural marker –men and the Japanese plural marker –tachi are examined. In section 2.3, the distributional and semantic properties of Korean plural marker –tul are explored. In section 2.4, we discuss „the plurality split‟ phenomena and show how it can be applied to the micro-parametric differences of plural marking in Chinese, Japanese, and Korean. In section 2.5, the concepts of specificity and definiteness are reconsidered. It will be shown how these two closely-related notions are represented in languages. Section 3 concludes this paper.


Introduction
The occurrence of a classifier is required in enumerating nominal structures in classifier languages such as Chinese, Japanese, and Korean.Thus, quantity expressions cannot be directly combined with nouns in these languages.Such English expressions as three students, ten dogs are not acceptable in classifier languages, as contrastively illustrated by the Chinese examples in (1): (1) a. *san xuesheng three student b.sange xuesheng three -Cl student "three students" The plural markers of classifier languages display distinctive properties which are not shared with those of non-classifier languages.Plural markers in non-classifier languages are nothing more than morphological markers of the plurality of the entities/individuals denoted by nouns.Plural markers in classifier languages, on the other hand, are more than plural markers per se.Some semantic/pragmatic information other than the plurality is also loaded on plural markers in this type of language.

Plural marker -men in Chinese
As the gloss in (2) below illustrates, bare singular nouns in Chinese can be construed in four different ways depending on the context: (2) a. wo qu zhao haizi.I go find child "I will go find a child/ the child/ some children/ the children." The bare singular noun haizi can get, that is, one of the following readings: (i) indefinite singular, "a child"; (ii) definite singular, "the child"; (iii) indefinite plural, "some children"; (iv) definite plural, "the children".Two factors causing the four-way ambiguity are the number (singular/plural) and the (in)definiteness.If -men is attached to the bare singular noun, haizi, as in (3), the ambiguity disappears and the plural marked nominal haizi-men can only be construed as "the children": (3) wo qu zhao haizi-men.I go find child-PL "I will go find the children." As can be deduced from the fact that the -men marked noun gets the definite plural reading, the Chinese plural marker -men marks not only the plurality but also the definiteness of the noun to which it attaches.
Chinese -men shows different distributional properties from its supposed English counterpart, -s.The English plural marker is attached to any countable noun which denotes more than one entity/individual.As illustrated in (4), for example, the distribution of -men is much more limited than that of English -s.
(4) a. wo qu zhao haizi-men.I go find child-PL "I"ll go find the children."b. wo qu zhao gou (*-men) I go find dog -PL (Intended meaning: "I"ll go find the dogs.")c. wo qu zhao shu (*-men) I go find book -PL (Intended meaning: "I"ll go find the books." Note here that the Chinese plural marker -men cannot co-occur with non-human count nouns such as gou "dog" or shu "book".The attachment of -men to any of non-human nouns, as can be seen in ( 4b) and (4c), is not allowed.
In sum, the Chinese plural marker -men, which can be attached to only human nouns, marks both the plurality and the definiteness.

Plural marker -tachi in Japanese
Japanese plural marker -tachi shares most of distributional and semantic properties of Chinese -men that we have discussed so far; however, inspection reveals some interesting differences between these two plural markers.Let us first consider (5), which illustrates selectional properties of -tachi.
(5) a. gakuseitachi student -PL "the students" b. inutachi dog -PL "the dogs" c. *kurumatachi car -PL "the cars" As can be seen in ( 5), -tachi poses a restriction on the nature of the referents denoted by nouns to which -tachi attaches.It attaches to nouns whose referents are animate like gakusei "student" as in (5a) and inu "dog" as in (5b).As (5c) shows, -tachi cannot mark the plurality of inanimate nouns like kuruma "car".It is significant to note that the Chinese counterpart of (5b) is not allowed.Consideration of the so-called "animacy hierarchy" will help us to understand the selectional properties that Chinese -men and Japanese -tachi have with respect to their stem nouns.There have been proposed many different versions of the animacy hierarchy in the literature (Silverstein, 1976;Comrie, 1981; among many others).The animacy hierarchy shown below can be considered as an integrated version.Plural marking in classifier languages can be a good example which shows how such an extra-linguistic concept such as animacy interacts with linguistic phenomena.Plural markings in most classifier languages seem to be allowed to be affixed with stem Kansas Working Papers in Linguistics, Vol. 30 (2008), p. 161 nouns whose degree of "animacy" is relatively high.As pointed out in Comrie (1985), it depends on the language used how fine the distinctions are required to be for the animacy hierarchy.Some languages require less fine distinctions such as human vs. non-human or animate vs. inanimate, whereas some other languages require finer distinctions.As we have seen, the demarcating line for Chinese plural marking is between human vs. non-human nouns since only [+human] nouns are allowed to be morphologically pluralized.When it comes to Japanese plural marking, the demarcating line is drawn between animate vs. inanimate nouns since only animate nouns, which include human nouns, can be overtly plural-marked.English plural marking, however, has nothing to do with the animacy hierarchy since it is obligatory to plural-mark any countable nouns which denote more than one individual/entity.Table 1 schematically shows how the animacy hierarchy puts a restriction on the types of nouns to which morphological plural markings are allowed to be attached in Chinese and Japanese.

First-and second-person pronouns
Table 1: The effect of the Animacy Hierarchy on Chinese and Japanese plural marking.

Korean plural marker -tul
So far, we have compared Chinese and Japanese plural markers in their distributional and semantic/pragmatic properties.
In this section, we will examine the syntactic and semantic properties of the Korean plural marker -tul which has not drawn much attention in the literature in comparison with its Chinese and Japanese counterparts.For Korean plural markers, it seems to have been tacitly assumed that it can be characterized more or less like -men and -tachi.But, as will be shown, -tul has quite different properties which are not shared with Japanese and Chinese plural markers.Now let us turn our attention to the distribution of the Korean plural marker -tul.
(6) a. haksayng-tul-i manh-ta student-Pl-Nom many-Dec "There are many students."b. kay-tul-i manh-ta dog-Pl-Nom many-Dec "There are many dogs."c. cha-tul-i manh-ta car-Pl-Nom many-Dec "There are many cars."d. *mwul-tul-i manh-ta water-Pl-Nom many-Dec "There is much water." As illustrated in (6a) through (6c), -tul can be attached to any countable noun, irrespective of the relative degree of the animacy of the noun.The only case where it cannot occur is with a mass noun like mwul "water" as in (6d).-tul can mark the plurality of human nouns like haksayng "student", non-human/animate nouns like kay "dog", and inanimate nouns like cha "house".Although it is most natural to pluralize a noun denoting human beings, which has the highest degree of animacy, it is also allowed to pluralize inanimate nouns with -tul, as far as the nouns are countable.Let us recall the cases of Chinese and Japanese, where the lexical pluralization is allowed only for humanand animate-denoting nouns, respectively.Compared with these two languages, Korean -tul shows the broadest scope of plural marking.Table 2 summarizes the scope of pluralization of Chinese, Japanese, and Korean lexical nouns, along with English ones in terms of animacy features: According to Table 2, it seems that Korean plural marker shares the same distributional properties with those of the English counterpart.But, differently from English, the lexical marking of plurality is not always required in Korean.Bare singular nouns can refer to plural entities when a proper context is provided, as exemplified in ( 7).
(7) haksayng-i manh-ta student-Nom many-Dec "There are many students."(lit.Students are many.) Due to the semantic demand of the predicate manh-"many", the bare singular noun haksayng "student" must be interpreted as plural "students".If this is the case, a question immediately arises: are (6a) and ( 7) free variants?It has been tacitly assumed in the literature that nouns with and without a plural marker are not semantically distinguished in classifier languages.Hence, plural markers in classifier languages have been generally termed "facultative plural markers" to convey the idea that they do not contribute to the plurality of nouns.But, as was the case for Chinese and Japanese plural markers, Korean plural marker -tul is not facultative at all.-tul-marked nouns are not interchangeable with bare nouns in Korean, as can be seen in the following example.As can be seen in ( 8), plural-marked nominals cannot be used as a predicate.This example clearly shows that -tul is not like English -s, the sole function of which is to mark the plurality of nouns.On top of that, -tul-marked nouns cannot be used as a generic, whereas bare nouns can, as contrasted in ( 9).The only difference between (9a) and ( 9b) is the absence and presence of the plural marker -tul.The bare noun kay "dog" of (9a) is generically construed, but the pluralmarked noun kaytul "dogs" of (9b) cannot be generically construed.If -tul marks the plurality only, the contrast illustrated in ( 9) is rather surprising.This example suggests that -tul is more than a plural marker.
Considering that -tul-marked nominals cannot be used either as a predicate or as a generic, we wonder whether the Korean plural marker is also involved in marking definiteness like Chinese and Japanese plural markers.Our answer to this question is negative.It will be shown that treating -tul as a definiteness plural marker is too strong.

Plural Marker Copying
Korean plural marker -tul has a very puzzling property that has not been discussed in the literature.It can be attached to the end of any constituents within the predicate under a certain context, as (10) illustrates: (10) a. haksayng-tul-i wundongcang-eyse-tul yelsimhi taliko -issta.
Student -Pl-Nom playground -Loc hard run -Prog-Dec-Pl "Students are running hard at the playground." Note that the plural marker -tul is attached to the locative phrase wundongcang-eyse "at the playground" in (10a), to the adverb yelsimhi "hard" in (10b), to the verb phrase taliko "run" in (10c), and even at the end of the sentence as in (10d).Unlike -tul which is realized onto nominal categories, copied plural markers (henceforth, CPM), as exemplified in (10), do not assign any meaning of plurality to a constituent to which it is attached.Kuh (1986) dubs this phenomenon "Plural Marker Copying" to represent his idea that the plural feature of the subject NP is copied onto other categories in Korean.It is beyond the scope of the present work to account for the semantic/pragmatic properties of the CPM.Rather, our immediate concern is on the structural requirement which licenses the occurrence of the CPM.It has been suggested that the plural feature of subject NP licenses the CPM (Kuh, 1986).In the case of ( 10), CPMs are licensed, since the subject, haksayng-tul (student-PL) "students", is plural.However, if the plurality of subject nominals is the only necessary condition for licensing CPMs, we are faced with a problem.As is well-known, Korean bare (singular) nouns can be construed to be plural as well, if an appropriate context is provided.When a bare noun haksayng "student" occurs as an argument of a verb, moyko-"gather", which requires its argument to be plural, haksayng must be construed to be plural.Let us consider (11).As the English glosses clearly show in (11), the bare noun haksayng "student" gets a singular reading in (11a) and a plural reading in (11b), depending on its predicates: taliko-"run" and moyko-"gather", respectively.If the plurality of subject nouns is the sufficient condition to license CPM, as Kuh (1986) argues, we predict CPMs to be allowed to occur in (11b), since the subject NP, haksayng "student" is construed to be plural.But, our prediction is not born out, as illustrated in (12).As can be seen (12b) and (12c), the plurality of subject NPs alone cannot license the occurrence of CPM.The plural subject NPs must be -tul-marked to allow the occurrence of CPM, as illustrated in (13).However, if -tul marks only the plurality of its attached noun, as has been assumed in the literature, the contrast between ( 12) and ( 13) is surprising.As far as the plurality of the subject NP is concerned, ( 12) and ( 13) are equivalent.-tul must be more than a plural marker.Keeping this in mind, let us consider the following example, where the subject NP is left out, but CPM occurs.
(14) (Situation) the husband returns home late from work and asks his wife whether she and the kids have had dinner or not.Husband: cenyek-tul mek-ess-na?Dinner -Pl eat -Past -Q "Did you (including the kids) eat dinner?"Wife: ta-tul mekess-eyo.All -Pl eat -Past -Hon "We all ate dinner."(Moon 1995: 358) Although the husband does not utter a subject overtly, the wife knows that he is talking about the kids and herself.And, in this case, the CPM can occur without an overt subject.The occurrence of CPM seems to be licensed by plural and specific subjects, whether they are overt or not.Differently put, the occurrence of the CPM indicates that its subject must be not only plural but also specific.Let us go back to the case of ( 12) and ( 13).
The reason that only -tul-marked nouns license CPM in ( 14) is that -tul marks not only plurality but also specificity of the noun it attaches to.On the other hand, bare plural nouns without -tul in (12) do not satisfy the two-fold condition for licensing CPMs, since they are plural, but not specific.

Specificity of Korean plural marker
In this section, we will examine the concept of specificity and support our claim that -tul is a specificity plural marker with more evidence.
Although it is still a matter of great controversy in the literature how to define the concept of the specificity, it has been generally assumed that a specific expression is used "when the speaker refers to a particular entity in the universe of discourse, which may be identifiable or non-identifiable."Let us recall that the "identifiability" is understood in this paper as denoting a pragmatic concept of the definiteness.Therefore, the aforementioned characterization of the specificity implies that a specific expression can be definite or indefinite.As summarized in Table 3, definiteness and specificity can be characterized according to the "identifiability criteria" of the speaker and hearer in the discourse (Von Heusinger, 2002: 3: The "identifiability" criteria for definiteness and specificity According to Table 3, definite expressions are used when the referent is identifiable to both the speaker and the hearer; indefinite specific expressions are used when the referent is identifiable only to the speaker; and indefinite non-specific ones are used when the referent is identifiable to neither the speaker nor the hearer.Since indefinites can be used for referring specific or non-specific referents 1 , an ambiguity can be created in a sentence containing an indefinite nominal as given in ( 15): (15) Nobuko wants to marry a native speaker of Ainu.(Haspelmath, 1997: 37) The indefinite expression "a native speaker of Ainu" gets a specific reading if the speaker knows who the Ainu native speaker is, but it gets a non-specific reading if the speaker wants to inform the hearer of Nobuko"s wish to marry an Ainu native speaker.Thus, ( 15) can be dis-ambiguated if more information about the identity of the Ainu speaker is provided as in ( 16): (16) Nobuko wants to marry a native speaker of Ainu.
a. … She fell in love with him during fieldwork sessions.(specific) b. … because she is Ainu herself, and she wants her children to acquire her ancestors" language.(non-specific) The ambiguous reading of indefinite nominals was also noted by Quine (1953), who claimed that "the reference status of nominals" is closely related to "the propositional modalities under whose scope they fall".According to him, the ambiguity observed in ( 15) is due to the scope of the irrealis verb, "want", under which the indefinite nominal, "a native speaker of Ainu" fall.If the speaker has "a referential commitment", we get the specific reading as in (16a).If the speaker is not referentially committed, we get the nonspecific reading as in (16b).
As Givón (2001) says, "all lexical verbs carry some inherent modality."( 17) illustrates how such inherent modalities of lexical verbs fix the interpretation of indefinite nouns: 1 Semantic definitions of specific and nonspecific indefinites are as follows: a. Specific indefinites A sentence of the form [ar ζ] ξ expresses a proposition only in those utterance contexts c where the speaker intends to refer to exactly one individual a and a is ζ in c.When this condition is fulfilled, [ar ζ] ξ expresses that proposition which is true at an index i if a is ξ at i and false otherwise (Heim, 1991:30).b.Non-specific indefinites A sentence of the form [a ζ] ξ expresses that proposition which is true if there is at least one individual which is both ζ and ξ, and false otherwise.(Heim, 1991:26) (17) a.He has a dog.
(i) > a particular dog (specific) (ii) > any dog (non-specific) c.He lacks a dog.
(i) *> a particular dog (specific) (ii) > any dog (non-specific) (Givón, 2001: 442) Under the "inherently presuppositional" verb "have" as in (17a), the indefinite nominal, "a dog" gets the specific reading.When the same indefinite nominal occurs with such an inherently irrealis verb as "want" as in (17b), it can be interpreted either specifically or non-specifically.(17c) shows that the indefinite nominal should get the non-specific reading under the scope of the inherently negative verb, "lack".
As can be seen in ( 17b), the specific/non-specific contrast is not grammatically marked in English.In Korean, on the other hand, the specificity can be marked through the presence of its plural marker -tul.Let us consider the following example: (18a) illustrates that the bare singular noun kae "dog" can be construed in a singular or plural way.Whether it gets a singular reading or a plural reading, this bare nominal is interpreted in a non-specific way only, unlike English.When the plural marker -tul is suffixed to the bare noun as in (18b), however, it is construed to be plural specific, i.e, "certain dogs".The reason that this sentence sounds a bit infelicitous is due to the absence of a proper context.The unnaturalness disappears, however, if some additional information is provided regarding the dogs which Mary wants to raise.

'The Plurality Split' revisited
Earlier in this paper, it has been pointed out that it seems to be too strong to characterize languages with such binary features as [±classifier] and [±(obligatory) plural marking].There have been not a few attempts to show that a negative correlation exists between the presence of classifiers and (obligatory) plural markers: if a language requires the occurrence of classifiers in numeral expressions, plural markings on nouns are not obligatory.Although such a typological generalization is not without a grain of truth, it does not reflect the actual linguistic phenomena that we observe in human languages.As we have examined in this paper, Chinese, Japanese, and Korean, all of which are considered to be typical classifier languages, do not fit into the generalization.For certain types of nominals, the plural marking is required in these languages.Smith-Stark (1974) observed that the number distinctions are closely related to the properties of nominals; specifically, the animacy properties.Dubbing this phenomenon "plurality splits", he explains as follows (Smith-Stark, 1974: 657): One can say that plurality splits a language in that it is a significant opposition for certain categories but irrelevant for others.In particular, it splits the category of noun such that for some nouns, plurality is distinguished from the singular, while for others the distinction may be irrelevant.That is, it becomes neutralized.Such a split may occur with respect to any of the mechanisms used to mark plurality, of which verb-argument concord, noun-modifier concord, direct marking of a noun, and direct marking of the noun phrase seem to be the four principle types.Where any one of the mechanisms for expressing plurality is neutralized for a subset of nouns, I will say that a split has occurred.
Based on the observation that we made through the cases of Chinese, Japanese, and Korean plural markings, we propose to incorporate the referential features such as [±definite] and [±specific] into the splitting feature hierarchy.The pertinent feature for splitting plurality in Korean is [±specific].Thus, the plural marking is required for [+specific] nouns, while it is not required for [-specific] ones.In Chinese and Japanese, [±definite] is the splitting feature.For these languages, plural markings are not optional for [+definite] nouns, while optional for [-definite] nouns.The parameterized splitting features for Chinese, Japanese, and Korean nominals are summarized as Table 4.

Specificity and Definiteness
The notions of specificity and definiteness are closely related.At least, they are discourse related, since both of them are used in the discourse on the ground that the referential individuals/entities are identifiable to the speaker.If the referents are also assumed to be identifiable to the hearer, definite expressions are employed.Otherwise, specific expressions are used.In this section, it will be shown how these two referential notions are reflected in languages.

Two types of article system
In the case of English, the distinction between definite and indefinite is overtly marked through two different articles, the and a(n).But, specifics and non-specifics are not lexically distinguished in English, since both specific and non-specific indefinites are expressed as a(n) NP.
To the contrary, the Samoan article system overtly makes a distinction between specific and non-specifics.As reported in Lyons (1999), Samoan has two articles which are distinguished by the specificity: le for specifics and se for non-specifics.Unlike English, the definite/indefinite distinction is not lexically marked in this language.Samoan specifics marked with le can be either definite or indefinite, while English indefinites marked with a(n) can be either specific or non-specific.
How to organize articles in two-article languages with respect to definiteness and specificity seems to subject to parameterization.Based on this observation, Ionin et al. (2003) proposed the so-called "the article choice parameter," which is defined as ( 19): (19) The article choice parameter (Ionin, Ko and Wexler, 2003: 4) A language which has two articles distinguishes them as follows: The Definiteness Setting: Articles are distinguished on the bases of definiteness.
The Specificity Setting: Articles are distinguished on the bases of specificity.
The article choice parameter is governed by the two factors: definiteness and specificity, among which the articles of English-type languages are divided by the first factor and the article of Samoan-type languages by the second factor.

Two Types of Plural Marking System
We have seen how definiteness and specificity can be parameterized in the organization of articles in languages like English and Samoan in the previous section.Keeping this discussion in mind, let us consider how definiteness and specificity play roles in the parameterization of plural markings in article-less languages like Chinese, Japanese and Korean.As discussed before, plural markers in these languages mark more than the plurality of referents of the nouns they attach to, unlike English plural marker -(e)s.
They mark not only plurality but also definiteness/specificity of their base nouns.Plural markers in Chinese and Japanese mark definiteness and the Korean plural marker marks specificity.An interesting pattern emerges in the article-less languages and the article languages with respect to marking definiteness and specificity.
Adopting Ionin et. al (2003)"s article choice parameter, we also suggest that plural markings in Chinese, Japanese and Korean are subject to parameterization which gives rise to two types of setting: the definiteness setting and the specificity setting.The working definition of the plural marking parameter is as follows: (20) The plural marking parameter The plural markers of classifier languages are characterized as follows: The Definiteness Setting: Plural markers mark definiteness.
The Specificity Setting: Plural markers mark specificity.
According to the plural marking parameter, Chinese and Japanese plural markers fall under the definiteness setting, and the Korean plural marker falls under the specificity setting.Overt plural markings are required for definite plural nouns in Chinese and Japanese, and specific plural nouns in Korean.( 54) illustrates the plural marking parameter of these three classifier languages in terms of two binary features, [±definite] and [±specific].
The definiteness setting (Chinese/Japanese Case) +definite -definite +specific -men/-tachi  -specific -men/-tachi  The Specificity setting (Korean Case) +definite -definite +specific -tul -tul -specific   Table 6: Two types of parameter setting of the plural marking parameter When Ionin et al."s article choice parameter and the plural marking parameter are compared, another potential linguistic parameter emerges: markings of definiteness and specificity become relevant for singular nouns in languages with articles, while they become relevant for plural nouns in article-less languages.

Conclusion
We started our discussion of plural marking in classifier languages by asking the validity of the generally-assumed correlation between the presence of classifiers and the putative plural marking.What has been shown in this paper is that plural marking is not "facultative" in classifier languages such as Korean, Chinese, and Japanese.Plural markers in these languages not only mark plurality, but also indicate some semantic functions.In this regard, plural markers in classifier languages are different from the English plural marker -s, whose only function is to signal the plurality of referents that a noun denotes.
The distinct properties of plural marking in classifier languages have been considered in relation to the "plurality split" phenomenon: plural marking for some referents is obligatory and for other referents it is optional.There exists a demarcating feature which splits the requirement of obligatory plurality in plurality split languages.We have claimed that such features splitting plurality are subject to parameterization.As has been shown, the plurality splitting feature of Korean is [±specific], and that of Chinese and Japanese is [±definite].Differently put, plural marking is not optional for specific plural nominals in Korean and definite plural ones in Chinese and Japanese.
Based on the findings we made in this paper, several predications can be made with respect to the acquisition of English as a second language (L-2 English).It is a well known fact that Chinese, Japanese and Korean speakers have difficulties in learning the usage of English articles.It seems that it is not only because these languages simply lack articles, but also because the levels of the definiteness/specificity marking are different.Secondly, we can also predict that Korean speakers have more difficulties mastering English articles, compared to Chinese and Japanese speakers.Korean and Samoan can be grouped together as specificity marking languages, and Chinese, Japanese, and English as definiteness marking languages.To Korean speakers to whom the (non)specificity distinction is more prominent than the (non)-definiteness, it should be difficult to learn the English article system which is arranged by the (non)-definiteness.

Table 2 :
Scope of pluralization in Chinese, Japanese, Korean, and English countable nouns

Table 5 :
Two types of parameter setting of the article choice