Existential, Possessive and Locative Constructions in Yami Language: A comparative study to Formosan languages ∗

： This study reports on Yami existential, possessive and locative constructions, regarding the issues on (i) two ‘HAVE’ structures, (ii) Definiteness Effect (DE) and (iii) an analysis of small clause, in terms of a formal syntactic approach. Compared with Formosan languages, Yami syntactically shows the differences based on two ‘HAVE’ structures. Besides, it is also proved that Definiteness Effect in Yami differs from that in other Formosan languages, such as Bunun language. Finally, this study also argues that the existential verb amian/abo ‘exist/not exist’ in Yami selects a small clause as its complement, representing a relative clause.


Related issues in language
Existential, possessive and locative constructions have been extensively reported in details in various languages, such as English, Mandarin Chinese, and Formosan languages, from different perspectives, for instance, syntactic, semantics, cognitive and so forth (Lyons, 1967;Kuno, 1971;Huang, 1987;Zeitoun et al, 1999;Wu, 2009;among others). However, there are few studies concerning existential, possessive and locative constructions of Yami language, spoken in Orchid Island, and viewed as a Batanic branch of the Philippine languages, which is one of Malayo-Polynesian languages of the Austronesian language family. That is one of the reasons why the studies of Formosan languages rarely report on Yami's existential, possessive and locative constructions. Yami language is a verb-initial language, and it has flexible word order variations (Rau & Dong, 2006;Blust, 2009;Lapolla & Poa, 2006) to manifest the different constructions by its case markers as well. Yami language ① also shows its language-specific properties in existential, possessive and locative constructions, for instance, the existential construction differs from the locative construction in its word order and case markers.
There are two relevant studies, concerning existential, possessive and locative constructions in Formosan languages (the subgroup of Austronesian language family): Zeitoun et al (1999) and Wu (2009).

Previous research
Zeitoun et al point out that the existential verb (AFFIRM)/negator (NEG) ② in several Formosan languages, which shares the same form with that of possession rather than the locative verb. Then, the existential verb/negator, in some others, shares the same form with the locative and possessive ones. Another situation is that the existential verb, in still others, shares the same form with the locative and possessive ones while the existential negator shares the form with the possessive one, not that of location.
In addition, Zeitoun et al mention two 'HAVE' structures in existential, possessive and locative constructions of Formosan languages. Indeed, Freeze (1991& 1992 notes that two 'HAVE' structures are found cross-linguistically, and they are (a) Y has X and (b) Y's X exists. Several Formosan languages display the first type only, such as Saisiyat and Paran Seediq, while most Formosan languages only display the second, for instance, Amis and Paiwan. Four Formosan languages display both types, namely Bunun, Kavalan, Puyuma and Mantauran Rukai (discussed in section 3.2).
In part of Definiteness Effect (hereafter DE), Zeitoun et al point out that four Formosan languages, Bunun, Kavalan, Puyuma and Mantauran Rukai included, have nominal case marking systems which involve a common referential/nonreferential distinction affecting both the case markers permitted in such constructions and their interpretations (discussed in section 4 ).
Finally, Zeitoun et al point out that word order variations in existential, possessive and locative constructions, and Zeitoun et al further note that it has been shown crosslinguistically that existential and locative constructions syntactically differ in terms of word order: locative phrases (LPs)/prepositional phrases (PPs) follow existential verbs in locative constructions whereas LPs/PPs will be in the sentence-final position in existential constructions. Wu (2009) demonstrates Isbukun Bunun existential constructions. Isbukun is one of the dialects of Bunun in Taiwan. Wu first illustrates how peculiar morphological and syntactic properties of the theme NPs (=DPs) in Isbukun Bunun existential constructions can be derived by means of an expletive account. Thus, Isbukun Bunun existential constructions which Wu investigates include structural Case marking, the emergence of definiteness restrictions, and the resistance of theme NPs (=DPs) to A'-extraction. Secondly, the study also argues for the analysis of a small clause of existential constructions in Isbukun Bunun, based on its word orders, island effects, the comparison with possessive constructions, and the coordination facts. Wu argues that Bunun existential verb selects a small clause as its sole internal argument.
Moreover, Wu investigates that DE plays an influential role in existential, possessive and locative constructions in Isbukun Bunun. She claims that definite determiners, such as -a/an 'that/this', adaiza 'that', sikaupa 'this' and proper names, cannot occur in existential, possessive and locative constructions (discussed in section 4 as well). Finally, she contends that a Bunun existential verb/negator selects a small clause as its complement from a formal-syntactic analysis (see also in Stowell, 1981;Safir, 1982;Paul, 2000).
In other words, Wu provides a description of the general properties in Bunun existential, possessive and locative constructions and a possible characterization of how these properties are derived according to the theory of grammar. Besides, she shows that the theme NP (=DP) in Bunun existential, possessive and locative constructions is not the pivot but rather stays within the VP and constitutes some structural relationship with a covert expletive. Moreover, Wu argues that existential, possessive and locative constructions in Isbukun Bunun cannot be analyzed along the lines of the complex NP /DP account nor the NP-XP analysis while they must receive a clausal treatment. Rau & Dong (2006, 2018 briefly provide a descriptive picture of Yami existential, possessive and locative constructions, as shown in (1) Based on Zeitoun et al (1999), Wu (2009) and Rau & Dong (2006, 2018, the former two report in details on existential, possessive and locative constructions in Formosan languages, spoken in Taiwan, whereas Rau & Dong briefly delineate the background of Yami existential, possessive and locative constructions. However, there are fewer previous studies concerning existential, possessive and locative constructions in Yami language in terms of a formal syntactic analysis. 1.3 This study It was discussed by Zeitoun et al (1999) that even though existential and possessive constructions may syntactically or semantically differ from the locative constructions in Formosan languages, the existential and possessive constructions in Yami language however share the same existential verbs/negators with the locative construction. Rau & Dong (2000& 2013 and Rau (2005) demonstrate that the existential, possessive and locative verb is amian/mian 'exist' while the existential, possessive and locative negator is abo 'not exist'. The examples of Yami existential, possessive and locative constructions are demonstrated respectively as follows.
(2) a. Ya m-ian so aviik. (Existential) The locative construction in (2c) shows that LP/PP moves to follow the existential verb mian 'exist' at the position between the VP and the DP, which syntactically differs from the existential construction by means of the word order variation. Also, the Case marking can make the distinction between the existential and locative constructions. The DP aviik 'pig' is marked by an oblique marker in the existential construction whereas that is marked by a nominative marker in the locative construction.
The current study thus attempts to investigate existential, possessive and locative constructions of Yami language from the concepts of syntax and semantics.

Two 'HAVE' structures
Two 'HAVE' structures can be divided into (a) Y has X, (b) Y's X exists, and (c) both. Zeitoun et al (1999) mention that several Formosan languages only display the first 'HAVE' structure, such as Saisiyat and (Paran) Seediq, and the existential and possessive verb functions as a two-place predicate, with theme/possessed entity marked by an accusative marker and the locative phrase/possessor by a nominative marker. The Paran Seediq existential sentences, cited from Zeitoun et al (1999), are illustrated in (3). On the other hand, Zeitoun et al. also contend that parts of Formosan languages manifest the second 'HAVE' structure only, such as (Truku) Seediq, Amis, Atayal, Paiwan, (Labuan) Rukai, which resembles a one-place predicate with a theme/possessed entity marked by a nominative marker and the possessor being realized by a genitive pronoun attached to the theme argument. The examples of Central Amis and Northern Paiwan, cited from Zeitoun et al. (1999), are shown respectively as follows: (4)  Moreover, Zeitoun et al (1999) delineate that Bunun, Kavalan, Puyuma and Mantauran Rukai show both 'HAVE' structures: (a) Y has X as well as (b) Y's X exists. The existential sentences of Bunun and Mantauran Rukai are also illustrated in (5) and (6). In (5a), as the nominative DP precedes the DP Ɂuvað 'child', it shows that the DP Ɂuvað 'child' is the possessed and the nominative DP Ɂinak 'I' is the possessor. (5a) indicates the second type; by contrast, in (5b), the DP -ik 'I' encliticize to the existential verb with a locative focus, and it means the DP Ɂuvað 'child' is located at the possessor, giving an existential reading.
As a matter of fact, Yami also demonstrates both two 'HAVE' structures. Rau & Dong (2000, 2013, 2018 claim that the possessive constructions in Yami language can show both (a) Y has X as well as (b)  Also, Chang (2000) mentions that an animate subject cannot occur in a possessive construction to be the subject (Spec of TP/VP/vP), so it syntactically differs from an existential construction. Semantically, (7b) in Yami moreover implies that the meaning is analogous to (6a) that in Mantauran Rukai, which indicates that 'I have money, and money is here (or on the speaker)', whereas (6b) and (7a) both point out that 'I have money (but it exists somewhere).' Besides, syntactically, Yami addresses Y has X structure by utilizing the obligatory marker, akin to what Mantauran Rukai does, while Isbukun Bunun is represented by LF. Moreover, in Y's X exists structure, besides, Yami syntactically represents it by means of the nominal marker.  Based on (8) and (9), the sentences show that DE plays a key role in (8c), (8e), (9c) and (9d), and makes the sentences ungrammatical in existential constructions. According to a formal semantic account, these definite determiners, such as every, the, that, this and seven, are not intersectivity quantifiers, so they cannot exist in existential constructions (Keenan, 1987;Wu, 2013).

Definiteness effects
Additionally, Wu (2009) demonstrates in detail that DE also plays an influential role in Isbukun Bunun existential constructions, which means that non-intersectivity quantifiers are not grammatically allowed to occur in existential constructions. The examples of Isbukun Bunun existential constructions are shown below.
(10) a. Aiza tasa ahil sia pankaka-cia. (Wu, 2009)  The form above demonstrates a piece of strong evidence that definite determiners are not allowed in existential constructions in support of the claim of Keenan (1987). Yami, however, shows a different way of representing generic meanings by means of the construction of amian o, which indicates the old information rather than new one. The examples of Yami existential sentences are shown, respectively ① : (11a) has an indefinite reading due to the oblique marker o while (11b) gives a definite reading owing to the nominative marker so, that is to say, DE does not play any significant part in an existential construction. However, seeing that (12a) and (12b) are not acceptable, the free definite determiners ya 'this' and ori 'that' do play fairly evident roles in the existential constructions. The findings are in support of the claim of Wu (2009).
Conversely, the existential negator abo 'not exist' only permits the nominative marker o, which indicates a definite reading, and thus (13b) is grammatical in a Yami existential construction whereas (13a), an indefinite reading, is not. Last but not least, the proper names in (14) are allowed to occur in a Yami existential construction as well, a clear-referential reading. Therefore, it appears that DE does not significantly work in the existential constructions.
To be brief, it can be rationally argued that the definite determiners, such as ori 'that' and ya 'this', are not grammatically allowed.

An analysis of small clause
The existential, possessive and locative verb amian 'exist/have' in Yami language is semantically analogous to English existential exist/have and Chinese cúnzài 'exist' whereas it syntactically functions as English expletive there (be)/have, Chinese yǒu 'have' and Taiwan Southern Min ū 'have', which can select a small clause (SC) as its secondary predicate (Stowell, 1981;Safir, 1982;Huang, 1987;Paul, 2000;Tang, 2010;and Zheng, 2013 Based on (15), the sentences manifest that these three existential verbs can select a secondary predicate as their complements; in other words, what secondary-predicate phrases that the existential verbs select can be syntactically diagnosed as small clauses, that is to say, analyzed as a [α DP XP] construction in support of the claim of Stowell, 1981;Safir, 1982;Huang, 1987 andPaul, 2000. A small clause has a subject-predicate of a clause, but that lacks tense, neither finitive nor infinitive. The small clause is closely related to the phenomena of raising-to-object, exceptional case-marking (ECM), accusativus cum infinitivo, and object control, which all indicate the analysis of a mono-clause (Haegeman, 1991;Grano, 2012). According to ECM, the small clause actually denotes an interpretation where the subject of an embedded infinitival verb seems to show in the superordinate clause. Additionally, it is unexpectedly marked with an object case morphologically if it is a pronoun in English.
All in all, an SC involves an embedded clause (or bi-clause), relative clauses and mono-clauses. Yami DP/NP will be thus marked by an oblique marker, which supports the above claim from the examples in this study.
Furthermore, in Austronesian languages, Keenan (1976) claims that the postverbal material is structurally similar to a complex NP, and it thus can be treated on a par with a relative clause with the relative pronoun and/or complementizer omitted, as shown in (16) ① .
Notice that in Malagaw, an NP precedes the modifying clause (Paul, 2000). Besides, Wu (2009) states that the existential verb aiza 'exist' in Isbukun Bunun takes a small clause as its complement, and the simple/complex sentences are illustrated below: (17)  The phraseβ in (17b) is syntactically analyzed as a small clause, and the syntactic tree is given in (18a), in contrast with (17a), as given in (18b) What phrases the linker a chooses in a Yami existential construction can be diagnosed as a relative clause. Chang & Rau (2017) claim that Yami relative clauses are connected to the head nouns by the linker a. The head noun hence can either follow the relative clause or precede it. In addition, Rau & Dong (2006) claim that if a relative clause precedes the head noun, it restricts the head noun. However, when it follows a noun, it describes the characteristics of the head noun, as shown in (19) As a relative clause follows the head noun, it will result in a relative-clause reading. The current study hence claims that the existential verb amian/abo 'exist/not exist' selects a small clause, as illustrated in (20) Based on the syntactic analysis of Bunun existential sentences, this study also motivates the structure for Yami existential constructions as presented in (22b) , and English language is a non null-subject language whose grammar forbids an independent clause to lack an explicit subject. On the contrary, Mandarin Chinese is a null-subject language that permits an explicit subject and therefore the expletive subject in Yami is not obligatory to occur in the Specifier of TP in a syntactic structure. Besides, Extended Project Principle (EPP) is another factor to manifest that Yami language is a null subject language, and there is an expletive pro in a Yami existential construction. EEP points out the obligatoriness of subjects, proposed by Chomsky (1995). Since Yami does not bear any EPP feature, it further indicates that it does not obligatorily require the Spec of vP to move to the Spec of TP to be the subject. It points out that there exists a pro expletive in Yami existential, possessive and locative constructions. Besides, another major motivation for claiming that the existential verb amian 'exist' selects a small clause is Huang's (1987) mentions of the existential verb taking an embedded clause, like Mandarin Chinese yǒu 'have/exist', which belongs to an SC as well. Therefore, it is argued that the existential verb amian 'exist' in Yami existential, possessive and locative constructions does choose a small clause.

Conclusion
I have delineated a picture of existential, possessive and locative constructions on three issues: (i) two 'HAVE' structures, (ii) Definiteness Effect (DE), as well as (iii) an analysis of small clause. Therefore, the contribution of this research is the argument that Yami language has two 'HAVE' structures and DE only plays a key role with respect to the free definiteness determiners, such as ori 'that' and yi 'this'. Additionally, the study has argued that E, P and L sentences in Yami language cannot be properly analyzed in the line with the complex NP/DP account nor the NP/DP-XP analysis. Rather, a clausal treatment is more effective. Hence, what syntactic constituent the existential verb/negator amian/abo 'exist/not exist' selects is diagnosed as a small clause, and the small clause the existential verb/negator amian/abo 'exist/not exist' selects is the relative clause.