2.1 Definitions

In (1–6) below, we reproduce Stassen’s definitions of the Locational Possessive, With-Possessive, and Have-Possessive types and reproduce one example of each with Stassen’s (2009) glossing.

This type has been expanded in Stassen (2013) to include conjunctions expressing simultaneity between clauses.

2.2 General comments

A serious shortcoming of the definitions reproduced in § 2.1 is that they imply the universality of a grammatical relation ‘subject’, a postulate which was widely accepted some decades ago but is now rejected by many general linguists and typologists. However, it would not be difficult to replace them by more or less equivalent definitions formulated in terms of alignment relationships between predicative possession and other functional types of predication (locational predication, comitative predication, transitive predication). We will not discuss further Stassen’s account of the constructions he classifies as Locational Possessive, With-Possessive, or Have-Possessive, since our purpose is to discuss much more fundamental issues concerning the very recognition of his ‘Topic Possessive’ type as a fourth basic type of predicative possession in a synchronic typology of predicative possession. With a view to the questions that will be discussed in the remainder of this article, some remarks are nonetheless in order about the Have-Possessive type.

2.3 Three remarks on the Have-Possessive type

Our first remark on the Have-Possessive type is that ‘transitive predicate’ in the definition reproduced in (5) above must not be understood as ‘verb showing all the properties of prototypical transitive verbs’. The verbs found in constructions for which the consensus holds that they belong to this type, even those whose transitive origin is unquestionable, are rarely if ever perfectly canonical transitive verbs. A case in point is Spanish tener ‘have’, whose behaviour in differential object marking differs from that of typical transitive verbs (Creissels 2013). Consequently, our proposal is to reformulate the definition of the Have-Possessive type as indicated below:

The second remark is that predicators shared by predicative possession and existential predication (such as var in Turkish – example (7) and écho in Greek – example (8)) can be found in two types of situations that must be distinguished carefully. Turkish illustrates a situation where the possessive use of an existential/possessive predicator must be analysed as an instance of the Locational Possessive type of predicative possession (or its Genitive Possessive variant): in (7a), the possessee NP is in the Nominative case, and the possessor NP has coding properties identical to those of adnominal possessors while (7b) shows its relation to a plain existential clause that also has a locative adjunct.

By contrast, Greek illustrates a situation where the possessive use of an existential/possessive predicator is an instance of the Have-Possessive type: in (8a), the possessor NP is in the Nominative case, and the possessee NP in the Accusative case. Note that example (8b) shows the inherent ambiguity of écho constructions in Greek between existential and possessive interpretation, in spite of the fact that the analysis of possessive clauses with écho as belonging to the Have-Possessive type is absolutely uncontroversial.

The distinction between the situations illustrated by examples (7) and (8) is crucial for the discussion of the diachronic developments underlying the situation we analyse in Southeast Asian languages (see Section 6.4).

The third remark is that diachronic change in predicative possession does not necessarily result from the emergence of new predicative possession constructions due to an extension of the uses of locational/existential predication, or to semantic changes affecting verbs such as ‘take’, ‘hold’, ‘get’, or ‘bear’ (as widely attested, among others, in various branches of Indo-European). Diachronic change in predicative possession may also result from purely formal changes in constructions already expressing predicative possession. As rightly highlighted by Stassen (2009: 208–243), the have-drift, by which predicative possession constructions of other types tend to acquire characteristics of the Have-Possessive type, is a very common type of evolution. Maltese, analysed by Comrie (1989: 219–225), constitutes a well-known case of such an evolution.

5.1 Subject in Standard Mandarin

Most descriptions of the languages dealt with in this paper (including Mandarin Chinese) use the term ‘subject’ without really discussing its definition, but the way they use it is mostly consistent with a general definition according to which, in the valency frame of a verb, the subject is the argument showing a cluster of morphosyntactic properties shared by the agent of prototypical transitive verbs and the sole argument of semantically monovalent verbs. However, as discussed by Lu et al. (2015) for the case of Standard Mandarin, East and Southeast Asian languages share area-specific features such as lack of argument flagging and indexing, and extensive use of the ellipsis of NPs whose referents can be recovered from the discourse context. These make it difficult to base the recognition of subjects on reliable tests, and in particular, to distinguish subjects from topics.

As far as Standard Mandarin is concerned, an array of empirically-based studies using text counts has shown that AVO and SV are statistically by far the basic word orders, consequently contesting the notion of topic-prominence advocated by Li and Thompson (1976). Relevant text studies include Sun and Givón (1985), Wang (1988), Wang-Alibert (2005), Chen and Yuan (2000), and Huang (2013)inter alia. Consequently, the subject in Mandarin Chinese is typically preverbal but not necessarily clause-initial, unlike the topic, as we will see below. However, due to ellipsis (or coreferential deletion) of pronominal subjects, a noun phrase immediately preceding the verb is not necessarily a subject. Moreover, there is an important exception to the preverbal position of subjects, which is that the subject of intransitive verbs occurs postverbally in the presentative construction.

5.2 Topic-comment constructions in Standard Mandarin

In this section, we clarify our use of the terms ‘topic-comment construction’ and ‘dangling topic’ while giving a brief overview of the research in this domain for Standard Mandarin.

The topic-comment construction comprises several subtypes, whose unifying feature is a topic NP occurring in clause-initial position (the classic description being Chao 1968: 69, 95–104; see also Rygaloff 1971; Li & Thompson 1976). Furthermore, according to some researchers, there is a fundamental distinction between syntactically licensed and semantically licensed topics (Huang & Ting 2006).

Topic-comment constructions with syntactically licensed topics include, in particular, patient (or object) topicalisation and so-called double subject or double nominative constructions.

In patient topicalisation, an O-argument can be placed in clause-initial position as topic to produce a non-canonical construction type with O_topic-(A)-VP word order, either with an omitted agent, understood from the immediate context, (15a) or with the agent overt as in the O_topic AV structure in (15b). The topic NP may be reprised by a co-referential pronoun, as in (15c), or not at all, as in (15b).

Note that the first example is about a time of famine and poverty in China while the second is about contemporary avant-garde art: the narrator describes how an unknown English word has been stamped all over an animal’s body:

In the ‘double subject construction’, the relationship between the topic and the subject is generally one of possessor and possessed, if not of a whole and its part, the latter including the possibility of a group and its subset. The possessive relationship is typically one of inalienability (as argued in Chappell (1996)). In fact, Li & Thompson (1976: 480) treated this subtype as the prototypical topic-comment sentence. ^[5] In spite of this, S. Huang (2013: 111) shows, on the basis of a corpus of discourse data, that the double subject type of topic-comment structure is rare in spoken and written Mandarin, as opposed to SVO clauses.

There are also some other minor subtypes with idiom chunks in the comment or with universal quantifiers as the subject (see Shi 2000; Xu & Liu 2007 for examples and a discussion).

The semantically licensed topics, known as ‘dangling’ or ‘aboutness’ topics, are claimed to be particularly prominent in Mandarin Chinese, and for that reason are sometimes designated as ‘Chinese-style topics’ (Chafe 1976). The very notion of dangling topic is however rejected by some authors like Shi (2000), who argues that, even in Mandarin, topics are always syntactically licensed.

It is true that some types of topics for which a dangling-topic analysis has sometimes been proposed lend themselves to an analysis in terms of syntactic licensing. For example, it can be argued that, in (17), ‘this proposal’ is not really a dangling topic, since it can be linked to the argument structure of ‘objection’ (an objection is necessarily against something).

Theoretical frameworks which reject the possibility of the semantic licensing of topics do so by viewing utterances as the visible part of an underlying discourse including as many ‘invisible’ elements as necessary to enable the appropriate syntactic treatment. In such a perspective, the “complete discourse” for (18) could be (among many other possibilities) something like ‘As for that fire_i, fortunately the fire brigade came quickly [otherwise it_i would have killed many people]’.

We fundamentally agree with the dangling-topic analysis of constructions such as (18), according to which, in a generative account, the topic NP cannot be analysed as extracted or moved from other loci, and must be base-generated. The criteria for analysing a sequence ‘NP(topic) – Clause(comment)’ as involving a dangling topic is that the topic NP is neither coreferential with a resumptive pronoun in the clause, nor analysable as filling a gap in the expression of the argument structure of one of the terms of the clause it precedes.

In our analysis of predicative possession in Mandarin, we exclusively refer to constructions with a dangling topic, structurally independent from its following comment, since this is precisely the one involved in Stassen’s definition of the Topic Possessive type.

6.1 Possessive and existential yǒu in Mandarin

Like the other Mainland East and Southeast Asian languages analysed in the remainder of this article, Mandarin has a verb (yǒu) that can be used as a possessive or existential predicator. In some of the languages dealt with in the following sections, the existential/possessive verb is also used as a plain locational predicator, but this is not the case in Mandarin.

In its possessive use, Mandarin yǒu occurs in a frame that can be schematised as follows:

This construction can express ownership of material possessions, as in (19a), but also ‘ownership’ of less tangible objects such as ‘solution’ in (19b), or an illness in (19c), as expected of a possession verb (see Bally’s characterisation of ‘have’ verbs (1926), including French avoir). ^[7]

Furthermore, yǒu can be used for ‘possession’ of body parts, as in (20). ^[8]

As an existential predicator, yǒu ‘there is’ needs only one, typically, postverbal argument, in a construction that can be schematised as follows:

The existential construction has an important discourse function as a presentative, introducing new referents (see Li & Thompson 1981: 509–519). As has been well-established, new information in Mandarin typically occurs in postverbal position, as for yí-ge hěn zhùmíng de yí-ge zuòpǐn ‘a very famous work of art’ in (21) (see, for example, Chao 1968: 76–78):

It is not normally possible to place the single argument in the position preceding the existential verb yǒu ‘there is’, unless it has been fronted and topicalised in a listing construction. By contrast, locative prepositional phrases (zài)-noun_ground-postposition_spatial may be found in clause-initial position, as in (21) above. This PP slot is ‘optional’ in the sense that the referent of the ground (the location) may be known from the prior discourse, or from some kind of shared knowledge.

As a rule, in Mandarin, locative phrases are introduced by the preposition zài 在 ‘at, in’, but the preposition is omissible in clause-initial position of existential clauses. ^[10]

[(zài)-noun_Location-postposition_spatial]_ground yǒu noun_figure

Locative phrases preceding the locative-existential predicate yǒu, as seen in example (22) above, may also be readily postposed after the sole argument NP (the figure element, yì-qún háizi ‘a group of children’), as in (23), but in this case, the preposition zài ‘at, in’ is obligatory:

The question that arises here is to what extent the possessor NP in the possessive use of yǒu could be analysed as a kind of animate location preceding the existential verb in the same way as the locative phrase in (22), according to Lyons’ suggestion (1967: 393ff.). ^[12] The main reason for rejecting this analysis is that, in contrast to the available alloforms for the existential construction in (22) and (23), it is usually impossible for the possessor NP in the possessive construction to be moved into a locative PP at the end of the clause, since it is not a locative referent in the first place. Moreover, it would be nonsensical in effect for this function of yǒu ‘have’, when it is used to express the possession of illness or body parts, as in (19c) and (20) above. ^[13] This already suggests that, as discussed in the following sections, the possessive use of yǒu ‘have’ is not fundamentally an existential construction at all and that its possessor NP acts rather as an argument of a possessive predicator. Put differently, in this new analysis of ours, we neither view the possessor as a dangling topic nor as a kind of animate location in the possessive clauses with yǒu, thereby reclassifying it as a Have-Possessive.

6.2 Argument structure and syntactic properties of possessive yǒu in Standard Mandarin

In this section, we present a series of observations showing that, in the possessive clauses of Mandarin, the possessor NP cannot be analysed as invariably and obligatorily occupying a topic position. We successively examine the following points: ^[14]

1. the prosodic properties of the clause-initial NP

2. information structure, givenness, and the morphology of the initial noun phrase

3. interrogative constructions

4. ‘whoever’ constructions

5. headless relative clauses

6.3 Possessive yǒu as a non-canonical transitive verb

One might object to our analysis that possessive yǒu is not a canonical transitive verb. However, as has already been emphasised in Section 2.1.4, the verbs found in Have-Possessive constructions, even those whose transitive origin is unquestionable, are rarely if ever perfectly canonical transitive verbs. Consequently, the notion of Have-Possessive does not imply that possessive clauses have all the properties of prototypical transitive clauses, but only that the coding of the possessor and the possessee is similar to that of the agent and the patient in the basic transitive construction.

First of all, the verb yǒu may not occur in the S bǎ O VP construction, cf. (33b):

The object-marking construction is high on any transitivity scale (Hopper & Thompson 1980; Tsunoda 1985) and its requirement of both a referential and affected direct object is symptomatic of this. The behaviour of yǒu in this respect is reminiscent of Spanish tener, which does not strictly adhere to the rule of marking referential human NPs in object role by the preposition a (Creissels 2013).

A second and related restriction is that yǒu ‘have’ cannot be passivised, and this resembles the situation for non-canonical transitive verbs of possession in other languages such as English have, French avoir or German haben. ^[18]

Third, whereas most transitive verbs are negated by the general negator bù 不 in present and irrealis contexts, yǒu ‘have’ has its own specific negator, méi 没 (otherwise mainly used in past or perfective contexts to negate the presupposition that an event has taken place).

Here again, cross-linguistically, there are similar cases of ‘have’ verbs behaving differently from other verbs with respect to negation. For example, the negation of ‘have’ involves a suppletive form in Dagbani (Gur, Ghana) and in Dhassanac (Cushitic, Ethiopa & Kenya) (Dixon 2010, Vol 2: 299).

Finally, unlike transitive action verbs, yǒu ‘have’ does not generally allow perfective aspect marking. ^[19]

6.4 Diachronic development

Two kinds of diachronic change may lead to the use of the same predicator in existential predication and in predicative possession of the Have-Possessive type. The first one is the creation of an existential predicator via impersonalisation of a ‘have’ verb (i.e. development and grammaticalisation of the use of a ‘have’ verb with unspecified possessors), as attested in the history of French and other European languages such as Spanish, Alemannic, or Greek. The second one is that a Locational Possessive construction undergoes changes that eventually lead to its reanalysis as a Have-Possessive construction (have-drift), as attested by Maltese and other Arabic varieties (Comrie 1989: 219–225; Heine 1997: 99). In this latter scenario, the crucial move in the evolution is the development of a tendency to topicalise possessors and to drop the locative marking of such topicalised possessors, as can be observed for example in Japanese.

Documentation for the Sinitic family dates back to the Shang dynasty oracle bone inscriptions of 14th–11th centuries BC, known as the jiǎgǔwén 甲骨文, which are representative of Pre-Archaic Chinese. ^[21] Even in this early corpus of divinatory texts, both the existential and possessive interpretations of yǒu may be found, cf. Takashima (1996: 303–348) and Djamouri (1987).

Controversy abounds in the domain of Chinese historical linguistics as to whether the existential use preceded the possessive use of yǒu or vice versa. In the tradition of European scholarship on the Chinese language, a dominant view has been that existential ‘there is’ derives from possessive ‘have’, thus described by Downer (1959), Nivison (1977) and Karlgren (1933: 64) (see discussion in Takashima 1996: Vol I, 303–348). ^[22] In contrast to this, many scholars in China take the diametrically opposed view that the existential construction predated its possessive use (for example, Zhu 2010; Yao 2015). Still a third approach is seen in the work of von der Gabelentz on Classical Chinese (1881: 445–446) and in the Japanese tradition. Here, a more prudent standpoint is taken, viewing the possessive and existential uses as equally attested from the time of the earliest known inscriptions (Takashima 1996: Vol I, 307–309).

In a detailed analysis of the polysemy of yǒu in the Oracle Bone Inscriptions, Takashima indeed shows that both uses were possible as early as the Period I inscriptions and argues that they should, therefore, be considered separate constructions (1996: Vol I, 304–305, our transcriptions and glossing):

He proposes that the etymological origin of yǒu is ‘the right hand’, the source for a bifurcation into the two main meanings of ‘have in abundance in the right hand’ and ‘exist in abundance in the right hand’, a stage from which these eventually generalise into ‘have’ and ‘exist’. ^[23] Such an origin evidently evokes the notion of holding, as too another main transitive sense of this verb (that of ‘offer’), also found in this early period of Pre-Archaic Chinese.

Note that an interesting parallel can be found in Indo-European languages, since Latin habeo ‘have’ is cognate with Old Indian ghábhasti-ḥ ‘forearm’, their common etymon being a root reconstructed as *ghabh- ‘grasp’ (Creissels 1979: 642).

Additionally to the above, in a detailed study of the Oracle Bone Inscriptions, Djamouri (1987: 113–114) makes the highly relevant observation that in its existential use, the subject of yǒu is found only in postposed position: ∅-yǒu-Noun_subject. This, in itself, clearly distinguishes the existential use structurally from the possessive one, in which it typically occurs before the verb: Noun_subject -yǒu-Noun_object. ^[24]

The Bronze Inscriptions and early transmitted texts ^[25] provide a further important source of historical data for the following period of the Western Zhou (11^th–7^th centuries BC), corresponding to Early Archaic Chinese. The two main uses of yǒu continue to be found, according to Yao (2015: 59), not to mention those of ‘take possession’ and ‘be rich’ which Schuessler explicitly relates to the sense of ‘possessing’ (1988: 769–771) and, we can add, to ‘have in abundance’ identified by Takashima in the earlier period of Pre-Archaic.

In example (40) below, Yao explicitly translates yǒu as zhàn 佔 ‘occupy’ in Standard Mandarin. The dynamic action it connotes points to an earlier meaning of acquisition such as ‘hold’ or ‘grab’, suggesting that, historically, the direction of evolution may have been acquisition>possession>existence.

6.5 Conclusion to Section 6

In this section, we hope to have shown that the possessive use of Mandarin yǒu is not a case of a possessor encoded as a dangling topic in a topic-comment construction based on a locative-existential predicate, but rather a separate construction type, a Have-Possessive albeit potentially diachronically related to the former. In Mandarin clauses including a single NP in preverbal position, it is not immediately obvious whether this NP should be analysed as occupying the argumental position of subject or the position of a topic at the left periphery of the clause. Nonetheless, the observation regarding prosody suggests that the NPs preceding yǒu in possessive clauses are not to be uniformly analysed as occupying the topic position, and this is confirmed by the observation of their syntactic behaviour.

As regards the diachronic development, in spite of the fact that the existential use of yǒu is unquestionably very ancient, the historical data also attest uses of yǒu that, while no longer found in contemporary Mandarin, are much easier to explain as retentions of an original meaning such as ‘take’ or ‘hold in the hand’ than as a development from an existential meaning.

8.1 Introductory remarks

In the following sections, we present data supporting the view that the analysis developed above for Mandarin and other Sinitic languages also applies to the other Southeast Asian languages analysed by Stassen as having Topic Possessive constructions as their only available option. As illustrated in (49), like Sinitic languages, the languages in question have possessive clauses in which the possessor NP precedes a predicator also found in existential clauses, but shows no flagging that could justify a Locational Possessive analysis.

In fact, this polyfunctionality of existential/possessive predicators has been viewed as an important areal feature in earlier literature on this linguistic area, for example, in M. Clark (1989) who discusses data from languages belonging to four unrelated language families: Hmong (Hmong-Mien), Khmer and Vietnamese (both Mon-Khmer, Austroasiatic), Thai and Lao (both Tai-Kadai) and Mandarin (Sinitic, Sino-Tibetan). ^[28]

 The existential predicates which have the same form as the possessive predicates similarly allow locative phrases in either clause-initial or clause-final position, the latter being the unmarked position (Jarkey 2015: 204–205). These locative phrases are all overtly marked by spatial terms and so cannot be interpreted as some kind of subject, neither in (50a) nor (50b) from Hmong, exemplifying the two different uses of muaj ‘there.is’ and ‘have’ respectively.

As in Mandarin, the possessive construction includes no obvious indication that the possessor phrase occupies a preverbal argument position or a topic position at the left periphery of the clause. However, if possessor NPs in possessive clauses such as (49b) or (50b) above could only be analysed as dangling topics, there should exist an alternative construction expressing predicative possession with possessors that cannot be topicalised. None of the descriptions we have consulted mentions the existence of such an alternative construction, while some descriptions do mention features that provide additional evidence supporting a Have-Possessive analysis, in particular, prosody, distributives, right and left dislocation, headless relative clauses, as well as interrogatives.

8.2 Predicative possession in Hmong-Mien languages

The Hmong languages (Hmong-Mien family, known in China as Miao-Yao 苗瑶) are found as far north as the province of Hunan in central China and extend to the southwest through Guizhou, Sichuan and Yunnan provinces into Vietnam, Laos and Thailand.

Our sources are principally Clark (1989) and Jarkey (2015) for White Hmong spoken in Xieng Khoung and Luang Prabang provinces in Laos, as well as Sposato (2015) on Xong, also known as Western Xiangxi Miao, spoken in the northwest of Hunan province in China. ^[29]

Jarkey observes (2015: 43–44) that, in the “generic existential” use of the existential/possessive verb of White Hmong muaj, a non-referential or ‘dummy’ 3rd person pronoun nws is found in the clause-initial position usually reserved for subjects, whereas a generic noun occurs in postverbal position.

In the presentative construction, temporal and locative adpositional phrases may occur in clause-initial position, and in this case, the 3rd person pronoun is absent. The single argument in the form of an indefinite, specific NP occurs postverbally.

In contrast to the existential use of muaj, possessive clauses make use of a clause-initial NP:

Relevant to our discussion is also the possible diachronic relationship between muab ‘grasp with the hand, take’ and muaj ‘have’ (Jarkey 2015: 50), which formally differ only in their tone. This hypothesis is consistent with the presence of an expletive third person pronoun in existential constructions such as (51) above, ^[30] since such expletives are typically found in existential constructions that developed historically from an impersonal use of Have-Possessives as dey in African American English dey got (54), or es in Alemannic es hot – example (55).

In Xong, the cognate verb to White Hmong muaj is mex. Similar to Mandarin yǒu,mex is a defective verb in that it may only co-occur with certain aspect markers, such as the experiential, but not with completive or progressive ones (Sposato 2015: 497). It has three main meanings: (i) ‘to exist’ – example (56), (ii) ‘to have’ – example (57), and (iii) ‘to be wealthy’ – example (58).

In (57), given the context of childbirth, the dynamic interpretation of an acquisition for this verb of having appears to be uppermost, providing more clear evidence for the existence of a transitive valency for mex.

In example (58), mex is ambiguous between ‘be wealthy’ and ‘have’ according to Adam Sposato (pers.comm):

Expressing ‘be wealthy’ by means of a bivalent verb ‘have’ in a construction not including any overt possessee phrase is a quite natural semantic extension, also found in other Hmongic languages, in Eastern Kayah Li, in Lao and in Sinitic languages including Baoding Jilu Mandarin – even for the earlier period of Western Zhou Chinese (11^th – 7^th c.).

In addition to these uses, mex combines with a variety of nouns to form compounds that would be difficult to explain without resorting to the presence of a transitive valency in its semantic structure:

The same formation of compounds with cognate muaj ‘epp’ is also found in White Hmong: ^[31]

In sum, our analysis of the existential/possessive verbs of Sinitic languages also applies to Hmong languages. ^[32]

8.3 Predicative possession in Tai-Kadai languages

The Tai-Kadai language stock, also known as the ‘Kra-Dai’ family (Ostapirat 2000) is spread over southern China, Thailand, Laos, Myanmar (Burma) and Vietnam. ^[33]  Our sample includes data from eight Tai-Kadai languages representing its three primary divisions of Li/Hlai, Geyang/Kra and Kam-Tai (for details, see Appendix).

The largest and most complete descriptions of this family, not surprisingly, relate to the languages with official status, Thai and Lao. In the Kam-Tai branch of Tai-Kadai, to which both Thai and Lao belong, the existential/possessive verb is regularly mii or a closely similar form. In the other two small branches of Hlai and Geyang, the forms are respectively tsau² and aŋ³¹, the latter being the form in Judu Gelao.

Similarly to the case for Hmong languages, there appears to be a consensus among specialists that mii has both transitive and intransitive valencies and can be interpreted as possessive ‘have’ in the first instance or as existential ‘there is/are’ in the second. For example, for Standard Thai, Noss (1964: 124) describes mii ‘have’ as “a typical transitive verb” but one which can have the translation of ‘S exists, there is S’ when it is used with just one S argument. The same explicit claim of possessive mii being “a typical transitive verb” can be found in Sookgasem (2016: 46).

Furthermore, as illustrated in (61), the figure NP in the construction of existential mii is normally required to occur postverbally (for Thai, see Clark 1989; Noss 1964; Sookgasem 1992; Iwasaki & Ingkaphirom 2005; for Lao, see Enfield 2007: 157–158; and for Maonan, see; Lu 2008: 211).

The use of this intransitive structure with its postverbal subject NP is naturally linked to information structure, the new referent being typically realised as a morphologically indefinite NP. In Maonan (Kam-Sui), for example, Lu observes (2008: 212–213) that demonstratives such as ‘this’ and ‘that’ may not modify the postverbal subject of mɛ² ‘exist’.

The same situation applies in Lao (Enfield 2007: 157–161).

A closely related property of the intransitive existential structure, pointed out in several studies, is its typical function as a presentative, as in narratives, for example, whenever a new and major character is to be introduced.

In Maonan (Kam-Sui, Northern Guangxi), mɛ² in its existential use may be accompanied by an optional clause-initial NP coding locus or time, as we have seen in many other languages. Lu (2008: 211) describes this NP as an adjunct functioning as the topic. Similar to Mandarin zài, the locative preposition ȵa:u⁶ ‘be at’ is omissible in the clause-initial topic position, but obligatory when the locative adjunct occurs in post-verbal position:

The descriptions of Tai-Kadai languages that we have consulted do not really discuss the status of the NP preceding the existential/possessive verb in its possessive use, but some of them at least provide examples of possessive clauses whose meaning excludes analysing the possessor phrase as a dangling topic, for example in (64) and (65), where the possessor phrase is a distributive.

In Lao (as in Hmong), mii2 used as a one-place verb can also mean ‘be wealthy’ which as pointed out earlier, is necessarily related to the notion of ‘have’:

In Lao, Enfield (2007) provides an example of a construction with the possessor fronted in a kind of left-dislocation construction and reprised by the anaphoric 3rd person pronoun, man2:

In (67), haan1 khòòj5 ‘my goose’ occupies the clause-initial position of a ‘classical’ framed topic-comment construction where it is followed by a pause. The presence of a co-referent pronoun in the following two clauses supports the view that the possessive use of mii2 is not classifiable as a Topic Possessive, but rather belongs to the Have-Possessive type.

To conclude this section, we observe that Sookgasem (2016: 46) mentions an interesting difference between mii ‘have’ (which she characterises as “a regular transitive verb”) and mii ‘exist’ in nominalisation. In Thai, as illustrated by (68a), verb phrases (i.e. verbs possibly followed by objects and/or adjuncts but not preceded by a subject) can be nominalised by preposing kaan. In this respect, mii ‘have’ can be nominalised like any other transitive verb, and kaan mii N is interpreted as ‘for an unspecified possessor to have N’, as in (68b). By contrast, as indicated in (68c), the nominalisation of existential mii is “questionable” at best.

Although Sookgasem does not comment further on this point, one may imagine that the problem with ?kaan mii phaayú rxŋ? ‘existing a strong storm’ is that ‘kaan mii N’ evokes the <Possessor, Possessee> argument structure of transitive mii ‘have’, and storms are not among the types of entities to which the role of possessee in the construction of a ‘have’ verb can normally be assigned.

8.4 Predicative possession in Austroasiatic languages

Austroasiatic languages are distributed in a discontinuous fashion from Northeastern India to Myanmar (Burma), Southern China, the Malaysian peninsula and Indo-China, and extend as far southwest as the Nicobar Islands in the Andaman Sea. The best-known members of the family are Vietnamese and Khmer (Cambodian). Our sample of 14 languages includes these two national languages as well as a range of less well-known ones (for details, see the Appendix, and Sidwell 2015 for a phylogenetic classification). The languages of the Munda branch, spoken principally in Central and Eastern India, are not included in our survey, since they are outside the linguistic area in question and have unproblematic Locational Possessives (Stassen 2009: 311–313).

With the exception of the Munda branch, most of the Austroasiatic languages for which the relevant data are available pattern in a very similar manner to Sinitic, Hmong-Mien and Tai-Kadai languages: they use the same verb for both possession and existence, and in the possessive construction, possessor NPs are invariably unflagged. ^[34] Vietnamese may be used to illustrate these features, also replicated in Jing (Vietic, see Ouyang et al. 1984: 110–111); Bugan (Pakanic, see Li 2005: 100; Khmer, see Haiman 2011: 208)

Mon is however an exception, with an alternative construction in which the possessor NP is locative-marked. Interestingly, according to Jenny (2005: 83), the construction with an unflagged possessor NP (i.e. the construction typical of Mon-Khmer languages) was more common in Old Mon, and the development of the construction with a locative-marked possessor NP is probably the result of contact with Burmese.

Here again, the descriptions we have consulted do not really discuss the status of the NP preceding the existential/possessive verb in its possessive use, but some of them at least provide examples of possessive clauses whose meaning excludes the possibility of analysing the possessor phrase as occupying a topic position. In examples (70) to (72), the possessor NP is an interrogative in a ‘whoever’ construction, or a distributive. As pointed out earlier, this type of NP cannot act as a topic by definition.

Another piece of evidence against the Topic Possessive analysis is that in Khmer “nouns of agency are regularly formed by the syntactic device of prefixing neak ‘person’ to a verb or a noun” (Haiman 2011: 74), as illustrated by neak naenoam lit. ‘person advise’>‘advisor’, and ‘rich’ is revealingly expressed by neak mian lit. ‘person have’ (Haiman 2011: 288), which can only be understood with reference to the <possessor, possessee> argument structure of a bivalent ‘have’ verb.

We also found in Bon’s (2014) grammar of Stieng (Bahnaric) the following example, which can hardly be reconciled with a Topic Possessive analysis.

Crucially in this sentence, the possessor NP is found within a temporal subordinate clause, which is clearly not a position where dangling topics are expected to occur.

8.5 Predicative possession in Tibeto-Burman languages

A striking feature of a large number of Tibeto-Burman languages is the fact that they may have anywhere from three to six existential and possessive verbs, typically distinguished according to parameters such as animacy of the possessee and manner of existence. For example, Shixing/Shihu (Qiangic, Tibeto-Burman, Sino-Tibetan) has four existential/possessive verbs including ^LHjĩ for animate beings, ^LHdʑõ for inanimate entities, ^HLkʰuɜ for existence inside a location or a container and ^LHdzi for existence of an entity attached to a location (Chirkova 2009). ^[35]

As rightly observed by Stassen (2009: 316), in the Tibeto-Burman languages of Tibet, Nepal, north-eastern India, and northern Burma, “locational Possessives are without competition”. However, among the Tibeto-Burman languages geographically close to the language families examined in the previous sections, the situation typical of the East and Southeast Asian area can be observed, with possessive clauses involving the same predicators as existential clauses, and invariably unflagged possessor NPs. Four such languages are mentioned by Stassen (Eastern Kayah, Arleng Alam (a.k.a. Karbi), Lisu, and Lahu), and the same configuration is also described in grammars of Nuosu (Loloish, cf. Gerner 2013), Guiqiong (Qiangic, cf. Rao 2015) and Menya (Qiangic, cf. Gao 2015).

In the Tibeto-Burman languages in question, as was the case with the languages dealt with in the previous sections, the available descriptions do not discuss the status of possessor NPs as topics or subjects in the possessive use of the existential/possessive verb. Several of them provide, however, data which would essentially contradict the analysis of the possessor NP as a dangling topic in a topic-comment construction.

For example, Gerner’s description of Nuosu provides examples of possessive clauses in which the possessor is an interrogative or a distributive – example (74), and in example (75) from Guiqiong (Qiangic), the possessor NP is sɯ⁵⁵ ‘who’.

Furthermore, there is another verb jɛ̃⁵⁵ in Guiqiong expressing existence or possession of an inanimate entity, and the noun jɛ̃³³-wu⁵³ derived from this verb by suffixing the nominaliser -wu can be interpreted either as ‘rich person’, or ‘belongings’ (Rao 2015: 429–430). As already discussed above for several other languages, this derivation is difficult to reconcile with the hypothesis of a monovalent predicate of existence expressing possession in combination with a dangling topic.

Similarly, in Menya (a language belonging to another branch of Qiangic), the noun resulting from the suffixation of the nominaliser -mi to the existential/possessive verb ²ndʑy can be interpreted either as ‘person possessing something’ or ‘person present somewhere’ (Gao 2015: 417–418), which again points to a distinct argument structure in the two uses of this verb.

The case of Qiang (Qiangic) is particularly interesting. In the same way as the East and Southeast Asian languages whose existential and possessive constructions have been analysed above, Qiang uses the same verbs for possessive and existential predication (and also for locational predication). In their possessive use, possessor NPs are consistently left unflagged, and consequently cannot be analysed as assimilated to locative adjuncts.

There is, however, an important difference with the languages analysed so far: in Qiang, 1st and 2nd person subjects are in general overtly indexed on the verb. Unsurprisingly, in the locational-existential use of the existential/possessive verbs of Qiang, the argument indexed on the verb is the figure NP. If their possessive use were a case of the Topic Possessive construction, in possessive clauses, the existential/possessive verbs would be expected to index the possessee, but this is not what can be observed. As illustrated by the following examples from three distinct Qiang varieties, in the possessive use of these verbs, the argument indexed on the verb is the possessor, that is, 1SG ‘I’ in (76b) and not ‘younger brother’, and similarly 1SG ‘I’ in (77b) and (78b) and not ‘legs’.

Consequently, argument indexation in possessive clauses rules out the Topic Possessive analysis and forces us to analyse the verbs in question as polysemous verbs occurring in two distinct coding frames, depending on the meaning they express: as verbs of location-existence, they select an intransitive construction with the figure encoded as the subject, whereas as verbs of possession, they select a transitive construction with the possessor encoded as the subject.

In other words, in Qiang, argument indexation provides crucial support for the analysis already proposed above for the existential/possessive verbs of other languages in which, unlike Qiang, the situation is obscured by the lack of any morphological marking of core syntactic relations.

Incidentally, we observe that, in addition to existential/possessive verbs such as those illustrated above, pure possessive verbs are also attested in Qiang. Unsurprisingly, they also have a transitive coding frame characterised by indexation of the possessor NP. What is particularly interesting is that one of them, Longxi Qiang tsé, has exactly the same form as tsé ‘catch, hold’, which suggests that tsé ‘have’ arose from tsé ‘catch’ via the same kind of semantic evolution as the Indo-European ‘have’ verbs, and possibly White Hmong ‘have’ as well (cf. §7.1 above). Moreover, one of the existential/possessive verbs of Puxi Qiang (zə, cf. ex. (76)) has exactly the same form as a ‘take’ verb in Longxi Qiang. All this suggests that the process of creation of transitive verbs of possession from verbs such as ‘take’ or ‘catch’ has been active in the history of Qiang, and that the existential/possessive verbs found in this language may have resulted from a subsequent evolution of ‘have’ verbs.

A situation very different from that found in the languages examined so far, but equally interesting from our perspective, is attested in Burmese. Like the vast majority of Tibeto-Burman languages, Burmese has a Locational Possessive construction involving the existential verb transcribed as ɕí or ʃi₁ depending on the sources, and the locative postposition (or case-marker) hma (also transcribed as hma₂ depending on the transcription system).

According to Jenny and Hnin Tun (2016: 247), the locative case-marker may be dropped. The optional dropping of locative flagging with topicalised locative phrases is relatively common in the languages of the word, but interestingly, this phenomenon cannot explain all the details of the behaviour of hma in the possessive clauses of Burmese. Crucially, if the dropping of hma were to be interpreted as conditioned by the topicalisation of a locative adjunct, it should not be possible with an interrogative pronoun fulfilling the semantic role of possessor. However, the data kindly provided to us by San San Hnin Tun attest to this possibility:

The optional dropping of the locative marker even with an interrogative pronoun in the role of possessor can only be analysed as a transitional stage in a process of have-drift, since the variant of (80) in which the locative marker hma₂ is not attached to the interrogative pronoun implies the possibility of treating the coding frame of ʃi₁ as including an argumental position in which unflagged NPs are assigned the semantic role of possessor.