5.1.1 Simple Head–Modifier Relations

Simple head–modifier constructions may feature an adjective or adjective-like verb as modifier of a head noun, with no overt marking of the relation between the two phrasal constituents. Here are some examples, with head-initial structure:

The languages just cited also show head-initial order when the modifier is a demonstrative (such as ‘this’ or ‘yon’), Again, the modifier does not take any overt marker linking it to the head:

By contrast with languages of the families just illustrated, speakers of Sinitic languages position modifiers before their heads.

When demonstratives are used as modifiers of head nouns in Sinitic languages, a linking marker – usually a classifier appropriate to the noun being modified – typically appears between the demonstrative and the head noun:

This is mimicked by non-Sinitic languages whose speakers have been in intense contact with Sinitic language speakers. An example is Thailand Mien, whose speakers are recent arrivals in Thailand, having come within the last two centuries from China. In Thailand Mien, a Sinitic linking particle tei²⁴ links the preposed demonstrative na:i⁵⁵ ‘this’ to the nominal head kwa³³ ‘melon’:

Head-initial languages of MSEA tend not to include this kind of linker, which explicitly connects a modifier to its head, but there is a pattern suggestive of this in Jahai. The use of a third-person singular pronoun immediately after a head noun is a kind of definite article in Jahai:

When a demonstrative, such as ton ‘that’, is used, the pronoun/article is retained, linking the demonstrative modifier to its head:

Sometimes a language seems to allow both orders of head and modifier in simple noun phrases: head-initial and head-final. In Mulao, for example, the term for ‘saddle’ (for horses) consists of a head noun aːn¹ ‘saddle’ and a modifier ma⁴ ‘horse’, but these two elements can be ordered in both ways, head-initial and head-final:

The first example shows the native Tai head-initial ordering, as observed, for example, in the Lao term for saddle: qaan3 maa4 [saddle horse]. The second example shows the Sinitic head-final ordering, as observed in Mandarin: mǎ ’ān [horse saddle]. It seems likely that the term with Sinitic ordering is adopted as a unit, and need not be taken as evidence that Mulao has a productive process of head-final ordering in simple noun phrases.

Here is another case of a noun phrase likely calqued as a unit into Mulao. As in many other MSEA languages, the term for ‘train’ in Mulao literally means ‘fire vehicle’. Notice that the ordering of head and modifier in the Mulao expression follows the Sinitic pattern (cf. Mandarin huǒ chē [fire vehicle]), while the morphemes used are a combination of a Sinitic borrowing (tshja¹; cf. Mandarin chē) and a native Tai word (fi¹; cf. Thai fai ‘fire’):

Finally, there are cases in Mulao in which two orderings occur simultaneously: a Sinitic phrase, in which the modifier precedes the head, is used as a postposed modifier of a native Tai head:

In the Mulao case, alternative orderings of head and modifier do not convey differences in underlying meaning. Rather, they index different sources of structure into the language: a native structure, inherited from ancestor languages, versus a borrowed structure.

In a different kind of case, a language may show alternative orderings of head and modifier, where this ordering either conveys a meaning difference or is actually not what it seems. In Eastern Kayah Li, two orders of nominal head and modifier are observed (Solnit Reference Solnit1997: 187–188). When a nominal (e.g., the noun ‘chicken’) acts as a modifier of a noun (e.g., ‘egg’), it appears before the noun it modifies. Here is an example, with ‘chicken’ modifying ‘egg’:

By contrast, when the modifier is a verb or adjective, it appears after the noun it modifies:

These two structures can be combined in Eastern Kayah Li, such that a nominal head may have a nominal modifier occurring before it, and a verbal/adjectival modifier occurring after it:

Another kind of case is illustrated in the following two examples from Lao, which show what appear to be opposite head-modifier orderings, with only slightly different meanings:

These examples might appear to suggest that both head-initial and head-final structures are possible in Lao noun phrases. But Lao is strictly head-initial in its noun phrase structure. In fact, the second example shows the usual order: the difference is that what looks an adjectival/verbal modifier – in this structure, tom4, glossed as the verb ‘to boil’ – must in example (5.22) be analysed as the head of the phrase. It is one of a set of verbs of food preparation that can also be used as nouns referring to dishes that are prepared using that action/process (see Enfield Reference Enfield, Ameka, Dench and Evans2006b). So, the appearance that the language allows both head-initial and head-final structures is illusory. Both examples above are head-initial.

A similar phenomenon is described in Cantonese. Matthews and Yip (Reference Matthews and Yip1994: 49) note that ‘some compound nouns like food and animal terms have their head nouns in the initial position (see Y.-S. Cheung Reference Cheung1969)’, which contrasts with the usual Sinitic head-final ordering. Their examples of apparently head-initial ordering in Cantonese nominal structures suggest that the pattern is only found in specified semantic domains. A first set of examples relate to foods (as in the Lao case just discussed):

A second set of examples relate to domestic animals, where the sex of the animal in question is expressed in a postposed modifier, rather than the preposed modifier that is expected in a Sinitic language (compare Mandarin gōng-jī for ‘rooster’):

5.1.2 Relative Clause Constructions

Relative clauses allow speakers to modify head nouns in more complex and specific ways than by adding simple modifiers as described in the last section. A relative clause is a complex modifier that is derived from a full clause or sentence. To illustrate, suppose that we planted two fields of rice: one is upstream, the other is downstream. Now suppose I want to refer to just the first rice. When I say ‘rice that we planted upstream’, this structure is a noun phrase, consisting of a head noun ‘rice’ modified by a relative clause. One way of describing the modifying element ‘that we planted upstream’ is to say that it is derived from the sentence ‘We planted rice upstream’: the target referent element is moved (‘rice’ becomes head of the new noun phrase), and a relativizer ‘that’ is added, linking the head to the modifying phrase.

Most, if not all, languages provide ways for their speakers to use relative clause constructions for modifying head nouns in noun phrases. Languages will differ in a number of ways as to how this is done, including (1) the relative order of head and modifier in the phrase, and (2) the presence or absence of an explicit relativizing particle. (Note that in many languages the use of an explicit relativizing particle is optional, as it is in English: compare ‘rice that we planted upstream’ and ‘rice we planted upstream’, both of which are fine.)Footnote ¹

In most languages of MSEA – with the notable exception of Sinitic languages, and languages that have been in intensive contact with them – relative clauses have the following properties:

1. 1. The relative clause appears after its head.

2. 2. A marker of relativization may be used but is often optional, in which case the relative clause is simply used directly as a modifier.

3. 3. The set of grammatical roles played by the nominal which is modified by the relative clause is somewhat unrestricted.

4. 4. The head is positioned externally to the relative clause.

Data from White Hmong can be used to illustrate these points. In White Hmong, relative clauses are typically introduced by the relativizer uas. Here is an example:Footnote ²

In this example, the nominal head refers to a soldier. The source predication for the construction is something like ‘The soldier was watching the riverbank’. Note that the soldier’s grammatical role in this construction is transitive subject, the argument in A function in the clause. The verb ‘watching’ has two arguments, A being the actor, and O being the undergoer.

In many languages, it is also possible to relativize on the transitive object, the argument in O function, as in ‘The riverbank that the soldier was watching’. In examining patterns of accessibility to relativization in languages of the world, linguists make reference to an accessibility hierarchy, identified by Keenan and Comrie (Reference Keenan and Comrie1977), as shown in Figure 5.1.

The generalization represented by the hierarchy in Figure 5.1 is that if a language can relativize on a nominal in one of these roles, it can relativize on all the other roles higher in the hierarchy. MSEA languages are liberal in this regard, allowing relativization on virtually any kind of argument in a source clause or sentence.

For example, in White Hmong, the relativized argument can be of any type on the accessibility hierarchy, including adjuncts (Riddle Reference Riddle1993). Here are examples showing relativization on arguments running all the way down the hierarchy from A (illustrated in the previous example) to adjuncts:

White Hmong allows a resumptive pronoun to be used to indicate the argument that has been relativized upon (compare the constraint against doing this in English: we say the woman who spoke Hmong but not *the woman who she spoke Hmong):

Moving now to an example from an Austroasiatic language, in Jahai, relative clause modifiers ‘occur at the right-hand periphery of NPs’ and ‘are introduced by means of the relative-marking proclitic /k=/, which is attached to the first constituent of the relativized element’ (Burenhult Reference Burenhult2005: 138). In the following examples, we see different elements relativized, including subjects, objects, and indirect objects:

Semelai, closely related to Jahai, has similar relative clause constructions. In Semelai, the relativizer is the proclitic mə=, which ‘is often used to name a category’ (Kruspe Reference Kruspe2004: 201), as shown here:

The following examples show mə= used as a relativizer:

By contrast with the examples we have seen above, speakers of Sinitic languages put relative clauses before their head nouns. Sinitic relative clauses are linked to their head nouns by an overt marker. In Mandarin, for example, this marker is de, as shown in the following examples (relativizing on arguments in A, O, and IO roles, respectively):

Cantonese shows the same pattern, using a marker ge, a classifier with general meaning (Matthews and Yip Reference Matthews and Yip1994: 109). A resumptive pronoun is obligatory when an indirect object or object of preposition/coverb/comparison is relativized. It is optional when a direct object or possessor is relativized, and not required when a subject is relativized (Matthews and Yip Reference Matthews and Yip1994: 110–111).

Tibeto-Burman languages show a similar pattern to Sinitic, with the relative clause preceding the head noun. Here are two examples from Daai Chin, a Tibeto-Burman language of Western Burma (So-Hartmann Reference So-Hartmann2009: 134).

Similar examples from Hakka Lai (also a Tibeto-Burman language of Western Burma), show head-final structures in which the relativized arguments can be of varied grammatical roles. Relativization in Hakka Lai ‘requires a particular ablaut grade and an invariant relative clause particle’. There are two relativizers – the postposed linkers -tuu and -mii – which do not appear to differ in meaning, ‘except in terms of their potential relativization targets’, with -tuu having a lower frequency in texts (Peterson Reference Peterson, Thurgood and LaPolla2003: 422). Here are two examples:

We saw above that White Hmong has post-head relative clauses. However, in the Hmong-Mien language family more broadly, there is variation in head-modifier order in these constructions. The following examples from Thailand Mien and Aizhai Miao show pre-head relative clauses, with an explicit relativizing marker (ȵɛi³³ in Thailand Mien, naŋ⁴⁴ in Aizhai Miao), following the Sinitic pattern:

Similarly, in languages of the Tai family whose speakers have had intensive contact with Sinitic speakers, we see variation in head–modifier order in relative clause constructions. Sometimes this variation is observed within a single language. Pre-head relative clauses occur in Tai languages that are spoken in China. These structures, which go against the general head-initial patterning of noun phrases in Tai languages, may have been caused by contact with Sinitic languages, suggested by the fact that the morphological marker used in these head-final structures often resembles a Sinitic marker.

In an example from Xia’ao Zhuang, a head-final relative clause construction is marked by the relativizer ti³³, which resembles the pan-Sinitic de and cognate forms (cf. Mandarin de, Cantonese ge or more formally dik; Matthews and Yip Reference Matthews and Yip1994: 110–111):

In Jingxi Zhuang, different types coexist. In a first type, considered to be the main, native type, a head noun is followed by a modifying relative clause, without any linking marker (in some cases also being marked by a classifier placed before the head noun, as in this example):

In a second type, a recent innovation on the Sinitic model, the relative clause precedes the head noun and is linked to it by the relativizer ti5:

Similarly, in Hainan Cham, an Austronesian language that has also been in intensive contact with Sinitic languages, different orderings of modifier and head in relative clause constructions are observed. The following example has a verb–object phrase as modifier postposed to the nominal head, in a head-initial pattern. There is no marker of relativization:

Thurgood et al. (Reference Thurgood, Thurgood and Li2014) state that these ‘posthead relative clauses are inherited’, that is, they are native to Hainan Cham as an Austronesian language. The following example illustrates the ‘prehead relative clause’ which with its ‘distinctive, relative clause-final sa³³ (has) evolved under the influence of Mandarin’:

Thurgood et al.’s further comments are worth noting: ‘This is of interest not just as an example of language contact but as an instance of contact producing a typologically marked configuration: for SVO languages, Hainan Cham prehead relative clauses are typologically rare’ (Thurgood et al. Reference Thurgood, Thurgood and Li2014: 245). The same point applies to the situations in the Tai and Hmong-Mien languages discussed above.

5.1.3 Possessive and Part–Whole Constructions

All languages provide ways of expressing a relationship of possession between two entities, by modifying a head noun in terms of who or what it belongs to (see Seiler Reference Seiler1983, Heine Reference Heine1997, Aikhenvald and Dixon Reference Aikhenvald and Dixon2013, and references therein on the linguistic expression of possession and ownership in the world’s languages). In English, one way to do this is to use special possessive pronouns, as in my house or her brother. Another is to use the possessive suffix 's, as in John’s dog and that man’s hat. Yet another is to use a phrase built around the word of, as in the wings of a butterfly or the leaves of this tree.

MSEA languages do not have dedicated possessive pronouns comparable to English my, his, or its. The most common way to form a possessive construction in MSEA languages is simply to juxtapose the possessor (Pr) and the possessee (Pe) in a phrase, where the possessee is the head of the phrase, and where there is no explicit marking of the possessive relationship.Footnote ³

Chamic languages spoken in Vietnam show the MSEA pattern Pe–Pr. Here are examples from Northern Roglai and Phan Rang Cham:

A language’s pattern of ordering of elements in a possessive construction can be affected by a history of language contact, as we saw in the last section in relation to relative clause constructions. By contrast with the Chamic languages just mentioned, in Hainan Cham, spoken in intensive contact with Sinitic languages, possessor occurs before possessed, and an explicit marker of possession is used (the same as the one used in a relative clause construction) in a pattern influenced by Mandarin:

However, a different structure, with the opposite ordering Pe–Pr can be observed when the possessor is first person:

Similarly, Tai-Kadai languages in contact with Sinitic languages show both orderings of Pr and Pe in possessive constructions. Mulao is an example. ‘Although post-modification is the basic order for nouns being modified by a pronoun or another noun, pre-modification is not uncommon when the modifier indicates possession or location, in which case the structural particle kɔ is normally used’ (Wang and Zheng Reference Wang and Zheng1980: 67). This particle is likely borrowed from Sinitic, as is appropriate to its function in the Sinitic-modelled Pr–Pe construction.

Similarly to the Tai-Kadai situation, some Hmong-Mien languages show both a native head-initial pattern and a Sinitic-modelled head-final pattern in possessive constructions. In Aizhai Miao, the Sinitic Pr–Pe pattern often requires marking with the linking particle naŋ⁴⁴:

When both orders are possible for the same possessor–possessee pair, the head-initial structure does not allow the linker naŋ⁴⁴ to be used, while the head-final structure allows optional use of the linker naŋ⁴⁴:

In many MSEA languages, when a linker between possessor and possessee is optional, it tends to mark a semantic distinction. As Haiman (Reference Haiman1983: 793–795, Reference Haiman1985: 130–136) has argued, the presence of formal marking will tend to correlate with more alienable forms of possession.Footnote ⁴ This is reflected in the patterns of interpretation of possessive constructions in MSEA languages generally. For example, in Thailand Mien, a possessive particle ȵɛi³³ is used when the possessed is a physical object that is alienable, such as a towel:

By contrast, when the possessed noun is human or a dwelling closely related to human life, the particle ȵɛi³³ is not used:

The following example from Hmong shows how the presence of an explicit marker (a classifier appropriate to the possessed noun) can, through its marking of higher alienability, signal a distinction between two meanings of a word. In this case, the noun txǐ means ‘father’ when no explicit possessive particle is used, and it means ‘husband’ (more alienable than a father) when an explicit particle is used:

The same sort of pattern is observed in languages of the Tai and Austroasiatic language families, though with the reverse order of possessor and possessed. See Section 2.3.4, above, for examples.

5.2.1 Numeral Classifier Constructions

When speakers of many MSEA languages want to refer to things in specified quantities (e.g., ‘three houses’, ‘many houses’), they use numeral classifier constructions. A numeral classifier is a nominal form. It is either a dedicated type of word that exclusively serves the numeral classifier function, or it is a regular noun that can, perhaps secondarily, be used grammatically as a numeral classifier. Numeral classifiers occur contiguous to numerals and other expressions of quantity (Grinevald Reference Grinevald and Senft2000: 63–64, Aikhenvald Reference Aikhenvald2000: 98), and often, but not necessarily, together with the nouns that they modify. Numeral classifiers tend to be semantically general in meaning, and are selected for their appropriateness of fit with the main noun in the construction, that is, the noun which refers to whatever it is that is being enumerated or quantified (e.g., ‘house’ in the phrase ‘three houses’).

Two semantic types of numeral classifier are recognized. Mensural classifiers are much like familiar measure terms from English, such as ‘spoon’, ‘bag’, ‘bowl’, etc. as a way of quantifying an amount of some referent. Mensural classifiers convey how much of a quantifiable referent is being referred to (cf. two spoons of sugar, two cups of sugar, two kilos of sugar). Sortal classifiers do not have a ready equivalent in English. Rather than telling a hearer how much of a noun’s referent is being referred to, a sortal classifier conveys something about the general category of entity that the referent belongs to. So, for example, in Sgaw Karen, the sortal classifier thu² is used when counting or quantifying things that are rolled or curled or furled, such as a cigarette, banana leaf, or cloth; and the sortal numeral classifier dɨ², meaning ‘body’, is used when counting four-footed animals and similar entities, including buffaloes, porcupines, bamboo rats, dogs, elephants, statues, ghosts, and Buddhist monks (for Sgaw Karen examples, see Ratanakul Reference Ratanakul1998: 102–103).

Here are some illustrative examples of sortal classifier constructions, ranging across the major MSEA language families (with the classifier element in bold face):

The examples in (5.79–83) each show the classifier element occurring directly contiguous to a numeral, where that numeral quantifies the referent of a noun that is also included in the construction. The examples also show that MSEA languages vary in how they order the elements of a numeral classifier construction. Jones (Reference Jones1970: 3) noted that languages of the area fall into two groups with respect to the constituent order of numeral classifier constructions. In his Type I languages, the order of basic elements of the numeral classifier construction is numeral-classifier-nounFootnote ⁵ (as seen in the Hmong and Cantonese examples just above) (Table 5.1).

In Jones’s Type II languages, the order of basic elements of the numeral classifier construction is noun-numeral-classifier (as found in the Burmese, Thai, and Khmer examples just above); note that this ordering is different from Type I in having the noun at the beginning of the construction, but it is like Type I in having numeral-classifier in the same relative ordering (Table 5.2).

Numeral classifier systems often contain large inventories of nominals that can appear in the numeral classifier slot. For Lao, the set supplied in the dictionary of Lao by Alan Kerr (Reference Kerr1972) has more than eighty entries. For Thai, Haas similarly states that there are ‘some eighty or ninety’ classifiers in the language (Haas Reference Haas1942: 201). A non-academic source on Thai lists well over 300 classifiers.Footnote ⁶ These claims of many dozens if not hundreds of classifiers in these systems might be technically correct but they can be misleading. This is because only a small number of classifiers are used with any frequency. Such lists therefore show a classic long-tail distribution of token frequency among types, whereby speakers rely mostly on only a fraction of the available classifiers, while the majority of available classifiers are used only infrequently.

Let us consider the classifier system of Khmu. Suwilai (Reference Suwilai1987: 34) states that in Khmu, numeral classifiers ‘form a closed set of words and are required when most kinds of items are counted’. Table 5.3 presents a representative list of the more frequent numeral classifiers used in Khmu (Suwilai Reference Suwilai1987: 34, slightly adapted transcription and glossing).Footnote ⁷

We can note three important characteristics of the Khmu system. The first is that constituents in a numeral classifier construction canonically appear in the order noun-numeral-classifier (Jones’s Type II, above), as in the following examples:

Second, while there is a fixed, tight relation between the numeral and the classifier – in what we can call a quantifier phrase (numeral-classifier) – the syntactic relation between this phrase and the main noun is loose. So, as Jenny et al. (Reference Jenny, Weber, Weymuth, Jenny and Sidwell2015: 127) note, ‘the quantifier phrase may be separated from the head noun by verbal and adverbial constituents and appear at the end of the clause’. (See Enfield Reference Enfield2007: 120 for the same point for Lao.) Here is a Khmu example, in which the noun-classifier phrase mòːj tó ‘one classifier’ is separated from the main noun k.néɁ ‘rat’ by a prepositional phrase:

A third feature of the system is that many nouns can be used as classifiers for themselves, creating a phenomenon known as repeater classifiers (termed ‘self-classifiers’ in Suwilai Reference Suwilai1987). Here are two examples:

In the second example, the quantified nominal is a noun phrase consisting of a head (first) and modifier (second). In a repeater construction the classifier slot is not filled by the entire main noun phrase in this case, but only the head noun (t.lɔ̀ŋ ‘log’).

The Khmu situation is not representative of Austroasiatic languages more generally. Other Austroasiatic languages have simpler systems than this, and in many Austroasiatic languages, classifier constructions are optional or non-existent. In Semelai, the set of classifiers ‘is small and rigid, and their use is not obligatory’ (Kruspe Reference Kruspe2004: 206). Kruspe identifies two subsystems: a numeral classifier system consisting of terms mostly borrowed from Malay, and an indigenous system the terms of which can be used either as numeral classifiers or as nouns independently of the numeral classification function (Kruspe Reference Kruspe2004: 206). These subsystems are used for different functions. Malay loan numeral classifiers can also ‘be used to express relative dimension’ (Kruspe Reference Kruspe2004: 207). The primary function of indigenous classifiers is akin to a mensural classifier usage, but vaguer: they function ‘to express a subset of referents, rather than a precise quantity’ (Kruspe Reference Kruspe2004: 208).

Unlike Khmu, Semelai follows Jones’s Type I pattern, with the constituent order numeral-classifier-noun. Here are some examples, illustrating some of the functions of classifiers in Semelai, namely, providing anaphoric reference, expressing relative dimension, and expressing a subset of the referents, respectively:

Jahai, a neighbouring Aslian language, has a small inventory of classifiers (eleven attested), more than half of which are loans from Malay (Burenhult Reference Burenhult2005: 80–82). Only two classifiers are widely used. As in Semelai, the use of classifiers is optional when counting or quantifying. Classifier constructions are ‘pragmatically motivated references to and emphases and specifications of the noun they refer to rather than syntactically obligatory components of quantified NPs’ (Burenhult Reference Burenhult2005: 81). Classifiers can be used outside of quantifying contexts (as with Semelai) to provide anaphoric reference. A classifier ‘may be used to replace a noun which has been introduced earlier in the discourse or an implicit noun not previously introduced overtly’ (Burenhult Reference Burenhult2005: 80).

The order of constituents in a Jahai numeral classifier construction is numeral-classifier-noun (Jones’s Type I), as shown here:

As was seen in Khmu, the main noun is not tightly integrated in the construction: ‘a classifier and its modifying numeral seldom form a phrasal unit with the noun they refer to. Instead, the numeral and the classifier typically make up a separate noun phrase which is in some way detached from and syntactically opposed to the noun’ (Burenhult Reference Burenhult2005: 80). Here is an example in which the main noun ʔamɛŋ ‘siamang’ (a primate species) is separated from its quantifier phrase duwaʔ k<nʔ>mɔʔ ‘two classifier’:

This separability of head noun and classifier phrase is not exclusive to Jahai or Austroasiatic. Here is an example from Lao, showing a similar phenomenon:

Sinitic languages also follow Jones’s Type I order, but differ sharply from these Austroasiatic cases in that their classifiers are obligatory in usage, and come from much larger inventories. Cantonese has a set of over sixty numeral classifiers (Matthews and Yip Reference Matthews and Yip1994: 92), with various sub-types. They are obligatory not only when counting or quantifying something, but also for other functions, including when referring to a specific (as opposed to generic) object, when a noun follows the demonstratives nī ‘this’ and gó ‘that’, and when a noun occurs with bī ‘which’ or bīn…dōu ‘any’ (Matthews and Yip Reference Matthews and Yip1994: 93–94). Classifiers in these functions are assigned according to broad semantic principles, similarly to the Khmu situation described above. Here are some examples of Cantonese classifiers in situ:

The examples just cited show the generic or sortal function of classifiers. The following examples show the same kind of structure, where the classifier either groups sets of nominal referents into collectives (example 5.99), or defines measures of nominal referents in terms of their containers (example 5.100):

Mandarin has a similar system, with some 200 classifiers, divided into different classes, including those in Table 5.4.

Here are some examples according to this sub-classification in Mandarin:

In Hmong-Mien languages, numeral classifiers are found widely, in large and elaborate systems comparable to those just described for Sinitic languages. In research on Hmong-Mien languages, classifiers have sometimes been termed ‘selectors’ (see Lyman Reference Lyman1979).

Lyman (Reference Lyman1979: 95–100) gives a list of 167 numeral classifiers in Mong Njua, grouped into four types: unit selectors, grouping selectors, independent selectors, and ambifunctional selectors (see Lyman Reference Lyman1979: 20–21 for details). These appear in a construction of Jones’s Type I, with the constituent order quantifier-selector-noun, as shown in the following examples:

Numeral classifiers are used in Mong Njua not only for counting nouns, but also as quasi-definite articles (Lyman Reference Lyman1979: 94).

Other Hmong-Mien languages show similar patterns of structure and usage, with large inventories of numeral classifiers, following Jones’s Type I constituent-order pattern, and being obligatory in the languages when counting or quantifying things.

Examples from Thailand Mien show how the choice of numeral classifiers can be a mechanism for expressing politeness. In everyday reference, as shown in the following two examples, the same classifier – tau³⁵ – is used for people as for animals:

But when referring to a respected person, such as a guest, a special classifier – laːn⁵³ – is used:

Illustrating another frequent function of classifiers in MSEA languages, this example from Aizhai Miao shows the use of repetition of a numeral classifier to convey the meaning ‘each and every’:

This is also shown in these examples from Cantonese:

Turning now to Tai-Kadai languages, in Judu Gelao, spoken in China, the numeral classifier construction has the same Type I structure as just observed in Hmong-Mien languages. Here are some examples:

A point of contrast between the systems of Judu Gelao and the languages we have examined already is that when a demonstrative is used, it appears after the classifier-noun phrase, as shown here:

Xia’ao Zhuang, also spoken in China, has the same Type I constituent order in classifier constructions as is found in Sinitic languages. The language illustrates a common phenomenon in which there is a special constituent order when the number ‘one’ is involved. Look first at this example with ‘two’, which follow Jones’s Type I structure numeral-classifier-noun:

Now compare to this case involving ‘one’, where the numeral is moved to the end of the phrase, with the order classifier-noun-one (in the same way as we have just seen with the postposing of demonstratives in Judu Gelao):

As we saw in Judu Gelao, this same constructional ordering is observed when a demonstrative is used:

Other Tai languages – in particular, those spoken further south, such as Thai and Lao – show the fundamentally distinct ordering of Jones’s Type II pattern in their basic numeral classifier pattern: noun-numeral-classifier. Here are two examples from Lao (the second illustrating the use of a noun as a ‘repeater’ classifier):

Lao also illustrates a quirk of ordering in classifier phrases associated with the numeral meaning nùng1 ‘one’. When nùng1 ‘one’ is used to quantify something, the usual Jones Type II structure may be used, as follows:

But there is an alternative, available only for the numeral nùng1 ‘one’: the ordering of classifier and numeral can be reversed, in which case the classifier can be phonologically reduced, as shown in this re-rendering of example (5.124):

The Jones Type II system is also seen in Tibeto-Burman languages (in contrast to their Type-I-ordered Sinitic cousins). Lahu, for example, has a system of numeral classifiers of similar size and functional diversity as seen for the Sinitic languages mentioned above, though used in structures of the Type II noun-numeral-classifier order. The following examples show that, as is often the case in these sorts of systems, there is no one-to-one mapping of classifiers to nouns. Lahu speakers can often choose from among different classifiers for the same noun, when referring to certain familiar entities, such as pigs or houses:

Further examples from Lahu show some of the functional categories of numeral classifiers identified by Matisoff (Reference Matisoff1973a):

Reduplication of numeral classifiers in Lahu has a number of functions, as these examples illustrate:

Also in the Tibeto-Burman family, Burmese has an elaborate and well-studied system. Pe (Reference Pe1965) distinguishes three separate systems of ‘numeratives’ – classifiers, quantifiers and repeaters – defined as follows:Footnote ⁸

The Austronesian languages of MSEA also have numeral classifier constructions, though it is suggested that these may have been acquired through areal processes. In Hainan Cham, most of the attested mensural classifiers are borrowed from Mandarin. In addition, there is one native form, one Malay loan, and one form of unknown origin (Thurgood et al. Reference Thurgood, Thurgood and Li2014: 158). Speakers of Hainan Cham make use of classifiers of both the mensural and sortal type (Thurgood et al. Reference Thurgood, Thurgood and Li2014: 157). The constructions are of the Jones Type I structure (number-classifier-noun), like the Sinitic languages that Hainan Cham has been in intensive contact with.

The following example shows a mensural classifier construction in Hainan Cham:

Of the sortal classifiers in Hainan Cham, only two general classifiers are frequently used. This simple system can be said to sort referents into the categories animate versus inanimate. The classifier se⁵⁵/se²¹ is used with reference to ‘people, animals, birds’ and more generally for animates (Thurgood et al. Reference Thurgood, Thurgood and Li2014: 160), while pho²¹/pho⁵⁵ refers to ‘round objects, pieces’ and more generally for inanimates (Thurgood et al. Reference Thurgood, Thurgood and Li2014: 160).

5.2.2 Class Terms

In many MSEA languages, certain classes of entity, including especially members of natural-kind sets such as trees, fish, beans, many languages obligatorily use a class term as the head of the phrase.Footnote ⁹ Compare English and Lao: in English, we can say either ‘I planted a eucalyptus’ or ‘I planted a eucalyptus tree’, where the word ‘tree’ is optional (the same for many other species, including oak, fir, pine, ash, etc.). In Lao, however, the word ‘tree’ is obligatory. (Also note that in English often the head word can’t be used at all: ‘They had salmon (*fish) for dinner.’) Many MSEA languages feature class terms, in systems that have only partial overlap with their inventories of numeral classifiers, and with quite distinct syntactic organization.

Bisang (Reference Bisang1993, using the term ‘class nouns’) describes class terms in White Hmong. Table 5.5 shows sets of referring expressions consisting of two nouns in combination. In these expressions, the first noun acts as a class term, denoting the broader category to which the referent belongs (e.g., ‘tree’, ‘fruit’, ‘bird’, ‘fish’), and the second noun gives a more specific characterization of the referent (e.g., by denoting a species or variety).

Other languages of the same family show the same kinds of class terms. Here are two examples from Aizhai Miao, showing the use of du³⁵ ‘tree’ as a class noun in names for trees:

In Tai languages, the use of class terms is widely attested, for example as described by Delancey (Reference DeLancey and Craig1986). DeLancey re-analyses Saul’s (Reference Saul1964) description of the classification devices in Nung. Saul’s analysis focused not on the class terms but on the more-specific nominals to which the class terms attach. He termed these ‘Qualifiers’: ‘Qualifiers, like imposed quantifiers, designate a particular species of the nouns that they precede. The meanings are generally more qualitative than quantitative’ (Saul Reference Saul1964: 279). Here are two Nung examples, showing that the class terms (in bold) are distinct from the numeral classifiers (underlined):

Similar structures are observed in Mulao, as shown in the following examples:

For details on class terms in another Tai language, see Enfield (Reference Enfield2007: 146ff) on Lao.

Finally, we can note that class terms are also found in Sinitic languages, but with the opposite ordering of class term and specifying noun, as expected for that group. Following the head-final order of Sinitic grammar, languages like Mandarin position the class term after the specifying noun. Here are some examples:

1. 1. bǐ ‘pen, writing instrument’ (Chinese Academy of Social Sciences (CASS) 2012: 68)

   (5.144)

   máo-bǐ

   hair-ct.pen

   ‘Chinese writing brush’

   Mandarin | SN | CASS 2012: 875

   (5.145)

   fěn-bǐ

   powder-ct.pen

   ‘chalk’

   Mandarin | SN | CASS 2012: 384

2. 2. chē ‘car, wheeled land vehicle’ (CASS 2012: 155)

   (5.146)

   huǒ-chē

   fire-ct.car

   ‘train’

   Mandarin | SN | CASS 2012: 589

   (5.147)

   qì-chē

   vapour-ct.car

   ‘car’

   Mandarin | SN | CASS 2012: 1028

3. 3. guǒ ‘fruit’ (CASS 2012: 498)

   (5.148)

   píng-guǒ

   apple-ct.fruit

   ‘apple’

   Mandarin | SN | CASS 2012: 1004

   (5.149)

   wú-huā-guǒ

   neg-flower-ct.fruit

   ‘fig’

   Mandarin | SN | CASS 2012: 1374