Printer Friendly

Referential hierarchies in three-participant constructions in Vera'a.

1. Introduction

1.1 Database and grammatical description

This study is based on a corpus of the Vera'a language compiled by the author as part of a DoBeS Documentation Project. It draws exclusively on data from a text corpus of mainly narrative, but also some procedural and descriptive, texts. This means that it does without elicitations of all potentially occurring structural variants in terms of argument structure and referentiality / animacy properties in favor of data from more natural text production. As for the latter type of data, firstly they seem to represent more accurately the actual linguistic behavior of speakers, and secondly allow for contextualization and quantification. Especially the former point is highly relevant for the present study.

Some aspects of the structure of Vera'a have been analysed in various works by Alex Francois, for instance Francois (2005, 2007, 2009). The first modern descriptive account of the Vera'a language is presented in Schnell (2011); the basic structural features of the language will be outline in Section 2.

1.2 Basic concepts and outline of paper

This paper seeks to determine how inherent (as opposed to relational semantic role) features of arguments impact on the structure of Vera'a three-participant constructions. The features to be considered here are by and large the ones comprised in feature hierarchies variously labeled as "(extended) animacy", "person", "empathy", "topicality", "Silverstein's", etc. hierarchies in the literature (cf. Siewierska 2004:149; Croft 2004:132). I will use the term referential hierarchies here to refer to the different hierarchies discussed by Siewierska (2004:148ff.), listed in (1):
(1) a. Person Hierarchy:        1st > 2nd > 3rd

    b. Nominal Hierarchy:       pronoun > noun

    c. Animacy Hierarchy:       human > animate >
                                inanimate > abstract

    d. Referential Hierarchy:   definite > indefinite
                                specific > non-specific

    e. Focus Hierarchy:         not in focus > in focus

As for the first three hierarchies in (1a)-(1c), I will not refer to the combination of these labeled "Extended Animacy Hierarchy" by Croft (2003:130), but take these smaller hierarchies as points of reference for relating the Vera'a facts to those observed in other languages. Also, I assume that these hierarchies should not be taken as directly reflecting the grammatical categories of any given language but instead be understood as comparative concepts, in the sense of Haspelmath (2010), to which the categories determined for Vera'a may be related for the purpose of cross-linguistic comparison. The notion of definiteness is understood here as a property of referential expressions that signals the identifiability and activation of discourse participants (cf. Lambrecht 1994:77ff.). The term focus refers to a relational pragmatic category. As a detailed treatment of pragmatic relations is beyond the scope of this paper, I will use the term here to refer to an element of discourse that is in some sense unpredictable and not recoverable for the hearer with respect to the proposition of a given sentence (cf. Lambrecht 1994:207). In this sense, marking something as focal is not to be equated with marking it as new information or as indefinite; identifiable and activated discourse referents may also be focal in certain contexts (Lambrecht 1994:210f.).

In Section 2 I outline the basic features of verbal clauses and argument encoding in Vera'a and referential hierarchy effects in monotransitive clauses. In Section 3 I present the basic types of Vera'a three-participant constructions and their semantic properties. Section 4 investigates prepositional constructions and Section 5 possessive-like constructions. In Section 6 I give a short summary of the main findings.

2. Vera'a Clause Structure

2.1 Verbal clauses and argument encoding

In Vera'a, care argument relations in verbal clauses are encoded, core argument relations are encoded solely by means of their position relative to the verbal predicate, henceforth called verb complex (VC) and rendered in bold face characters in all examples. Argument alignment is nominative-accusative, so that S and A (subject) arguments precede the verb complex and P (object) arguments--where expressed by lexical NPs--immediately follow it:
(2a)   [[n   'ama'].sup.S]   ne    van   ma

       ART   spirit          TAM   go    hither (2)

       'Then the spirit came.' ASMS.048

(2b)   [[di].sup.A]   =m    vus   [[en   qo    ga    mew].sup.P]

       3SG            TAM   kill   ART   pig   TAM   be.many

       'He killed many pigs.' JJWR.017

Oblique argument relations and adjuncts are expressed by prepositional phrases or a range of adverbs. With the exception of dative PPs (see Section 4.2), oblique arguments and adjuncts must be placed at the right periphery of the clause, following the object NP where present:
(3a)   [[gidu].sup.A]   me    vil      [[en nak].sup.P]

       1DL.INCL         TAM   tie.up   ART canoe

       [wal'i        'ekene]   [le   =n     womomo']

       immediately   here      LOC   =ART   bush

       'We (DL) will tie together the canoe right here in
       the bush.' JJQ.083

(3b)   [[di].sup.A]   ga    'ivin   [[en 'ili-gi].sup.P]

       3SG            TAM   bury     ART egg-3SG

       [[le =n lolo- qe'an].sup.G]

       LOC =ART inside- ground

       'It [the incubator bird] buries its eggs in the
       ground.' GAML.056

Some temporal adjuncts and left-dislocated topics may precede the subject, but these structures are not immediately relevant for the present study (cf. Schnell 2011:32ff.). There are four basic types of referential expression in Vera'a, namely bare pronouns and pronominal NPs (cf. 2.2 below) that may occur without an article; common NPs introduced by the common article ; and personal NPs introduced by the personal article e. Pronominal NPs are headed by a personal pronoun, common NPs by a common noun, and personal NPs by a personal name or one of the two personal nouns raga 'people (PL)' and ruwa 'two people (DL)' which are used to mark number with certain types of nouns (cf. Schnell 2011:83ff.).

In the examples above the verb complex exhibits its minimal structure, consisting of an initial particle expressing tense/aspect/mood (TAM) (3) categories and a single verb only. However, typical for an Oceanic language, the verb complex in Vera'a may show considerably more complex structures consisting of several verbs (nuclear-layer serial verb constructions (SVCs)), additional adverbs and directional particles (DIR):
(4a)   [[di].sup.S]   ne    [[van kal kel].sup.SVC]

       3SG            TAM   go move.upwards return

       [sar.sup.DIR]   [ma.sup.DIR]   va'ane

       bushwards       hither         now

       'And then he came up to the bush again.' 1.TNU.041

(4b)   [[di].sup.A]   ne    [[van din].sup.SVC]

       3SG            TAM   go reach

       [ma.sup.DIR]   [[=n 'uvu- woqe'enge ne vo-wal].sup.P]

       hither         ART base- tree LIG CARD-one


       'And when it [the incubator bird] comes across a tree...'

Examples (4a) and (4b) demonstrate that the directional particle ma 'hither' can be considered the right boundary of a VC: object NPs or PPs have to follow ma and only TAM particles, verbs and a restricted class of adverbs and other directional may precede it (Schnell 2011:93ff.). Furthermore, SVC formation may alter the valence of the VC, so that in (4b) the second verb (V2) din 'reach' licenses a P argument expressing the semantic role of a Goal. Typically, Theme arguments are licensed by a V2 'o 'carry', and this will be relevant for the discussion of three-participant constructions below.

2.2 Referential hierarchy effects in two-participant constructions

The most obvious referential hierarchy effect in two-participant constructions triggers the expression of P arguments. Firstly, while pronominal subjects occur in the same syntactic position as their lexical counterparts, namely immediately preceding the VC (cf. examples (3) and (4) above), bare object pronouns are incorporated into the VC, occurring before the directional particle ma 'hither', rather than on clause level:
(5)   sow    'ama'    ene   ne     lisir    mog

      DISC   spirit   DEM   TAM1   follow   constantly

      'i     dir   ma

      TAM2   3PL   hither

      'Alright, then the spirit constantly followed them
      coming hither.' JJQ.476

Pronouns may alternatively form a pronominal NP which typically consists of the pronominal head and a demonstrative. A pronominal NP has to occur in the post-VC slot like a common (cf. (4b)) or personal (cf. (6b)) NP:
(6a)   Di    =m     van   'enene     ma

       3SG   =TAM   go    directly   hither

       [[di segen].sup.P]   [...]
       3SG DEM

       '[She paved her way through the dancers.]
       She came straight up to him here...' ISWM.340

(6b)   [...]   van   din     ma       [[e raga agene].sup.P]

               Go    reach   hither   ART people here

       '[This man] [...] came across these people here.'

In (6a), the pronominal NP 'he (the one) here' bears P function licensed by the transitive V2 enene 'directly' and occurs on clause level rather than VC-internally. Hence, pronominal NPs pattern with common and personal NPs in occupying the post-VC slot when functioning as P arguments, contrasting with incorporated pronominal P arguments. This distribution of P arguments can be roughly related to Siewierska's (2003:149) Nominal Hierarchy, so that the VC-internal P position is restricted to bare pronouns contrasting with other more complex nominal expressions.

The choice of pronominal expression is determined by the Focus Hierarchy, so that focal elements are realized as NPs rather than bare pronouns. Hence, the pronominal NP in (6a) is employed instead of a bare pronoun to signal that the referent in question has to be selected from a set of possible given referents (the other dancers). In this sense, it bears the relation of "contrastive focus" (cf. Lambrecht 1994:286). The deictic properties of the demonstrative serve to correctly identify the intended referent. Hence, bare non-focal pronouns are incorporated into the VC, and more complex focal NPs have to occur on clause level following the VC.

Another type of variation is triggered by the animacy of the P argument. Where a P argument is given and activated it will be realized differently depending on its animacy: while human Ps are granted pronominal realization in these circumstances, as in the examples discussed above, their non-human counterparts tend to be left implicit, as in the following example of a small stretch of discourse where zero objects are rendered as '[[empty set].sup.P]:
(7)   n     qetqet    men    di    ne    le

      ART   bird.sp   bird   3SG   TAM   take

      [sub.j][[=n qi'i- 'ama' ane].sup.P],   ne    ot    wal

      =ART head- corpse DEM                  TAM   put   straight

      sag       sa        le    =n     qi'i-gi,   [...]  ale

      upwards   exactly   LOC   =ART   head-3SG          INTERJ

      duru   =k     van   'o      [sub.j][[empty set].sup.P]

      3DL    =TAM   go    carry   hither

      ma       va'ane,   Van   'o     [sub.j][[empty set].sup.P]

      hither   now       Go    carry

      ma       le    =n     lolo     nime

      hither   LOC   =ART   inside   house

      'The qetqet bird, it took the head of the dead body, put (it)
      right up onto its head. [...] Alright, then they [i.e. the two
      birds] brought (it), brought (it) into the house.' ASMW.146-150

This pattern does not represent a categorical rule in Vera'a, and preliminary observations from procedural texts suggest that the pattern may be slightly different in text types with inanimate discourse topics. However, a quantitative investigation of narrative texts (cf. Haig et al. 2011) annotated following "GRAID (4)" glossing conventions (Schnell & Haig 2011) reveals that this tendency for non-human objects to be left implicit contrasts significantly with the realization of human and non-human S and A arguments which show an unexpectedly high degree of overt realization irrespective of animacy features (cf. Schnell 2011b). Haig et al. (2011) also find that P arguments are significantly more frequently inanimate than animate, and pronominal reference is thus avoided for P arguments in Vera'a (cf. Genetti & Crain 2003 on the same pattern observed in Nepali).

To sum up, the expression of P arguments is related to the Referentiality Hierarchy, the Animacy Hierarchy, the Nominal Hierarchy, and the Focus Hierarchy. Where a discourse participant is identifiable and activated, the form of reference will depend on its animacy, so that animate referents are expressed by a bare pronoun, while inanimates tend to be left implicit (zero anaphora). For animate referents, a pronominal NP is used instead of a bare pronoun where the referent is focal. Unidentifiable discourse participants are usually referred to by a lexical NP, while identifiable but inactive ones may be referred to by a lexical or pronominal NP. All P arguments in the form of a NP occur in a position immediately following the verb complex, and bare pronouns are incorporated into the VC. It is worth noting that this interaction of parameters could hardly be rendered in terms of the combined hierarchy, i.e. the Extended Animacy Hierarchy or Silverstein's hierarchy (cf. Croft 2003:130; Silverstein 1976).

3. Means of Expressing Three-Participant Events in Vera'a

3.1 Three-participant events and constructions

The following three-participant events will be considered in the current paper:

a) Physical transfer events involving Agent, Theme and Recipient

b) Physical transfer events involving an inanimate Goal instead of an animate Recipient

c) Mental transfer events involving an Addressee in the place of a Goal/Recipient

d) Events of creation or obtaining of an item for an intended possessor, hence involving the semantic role of Beneficiary

Following Malchukov et al. (2010:1), I consider physical transfer events to be the prototypical three-participant event. Most typically, physical transfer is designated by verbs like English give, send or hand and equivalents thereof in other languages, and the corresponding constructions can be regarded as typical three-participant constructions in any given language. For Vera'a, those constructions are considered prototypical three-participant constructions that involve the verb le, roughly 'give', the semantics of which will be discussed in the following section. The event types in b)-d) above are those that can be expressed with the same types of construction.

In the following section, I will first discuss the semantics and constructional properties of the verb le which is typically used to express physical transfer. I will then outline the possible constructions that are compatible with this meaning of le. These constructions are taken as the point of departure for the analysis of three-participant constructions in Vera'a. In the remainder of this paper, they will be examined in terms of i) structural variation attributed to referential hierarchy effects, and ii) their accessibility for other verbs expressing different types of events.

3.2 The meaning of the verb le

The Vera'a verb le occurs in two different types of contexts illustrated by the following two examples:
(8a)   di   ne    le     =n     qe'e   me    di    ane

       3SG  TAM   give   =ART   taro   DAT   3SG   DEM

       'Then he gave him a taro.' ASMS.106

(8b)   [...]  n     gie    ane   =m     salma,    'uvusmel

              ART   kava   DEM   =TAM   prepare   chief

       di    ne    le,    di    ne    in

       3SG   TAM   take   3SG   TAM   drink

       '[...] and when the kava is ready, the chief will
       take [it] and drink [it].' JJK.016

In (8a), the verb le translates as 'give' and the event expressed involves three participants with the roles Agent, Theme and Recipient. In (8b), the same verb le is rendered as 'take' in the English translation, and here the event expressed involves only two participants, namely an Agent and a Theme. In both cases, however, the crucial point is that the Theme changes its location, and the semantic analysis I am proposing for le is that both uses should be attributed to a single meaning, namely 'change of location' or CAUSED MOTION in the words of Rappaport Hovav & Levin (2008); and this meaning is compatible with both types of participant and semantic role constellations. Only the 'give' sense and respective constructions will, however, be relevant in what follows.

That Vera'a le has a CAUSED MOTION rather than a CAUSED POSSESSION component as its core lexical meaning (cf. Rappaport Hovav & Levin 2008) is evident from examples like the following:
(9)   No    =k     van   no    =k     le     lu     di    ma

      1SG   =TAM   go    1SG   =TAM   move   over   3SG   hither

      '"I'll go and bring her here." [To show his wife who had
      been hiding out in the bush to his parents.]' JSV.096

Example (9) provides three main pieces of evidence in favor of a CAUSED MOTION rather than a CAUSED POSSESSION event schema associated with the meaning of le (cf. Rappaport Hovav & Levin (2008:137ff.) for analogous diagnostics against a CAUSED MOTION meaning of English give): 1. the event expressed in example (9) clearly involves a Source, namely the place where the speaker is going to pick up his wife; 2. the presence of a Source implies the presence of a Path component; and in fact a Path role is explicitly mentioned in the form of the adverb lu 'through, over, from... to...' which serves to express that a Theme is moving / being moved across a distance or some obstacle. The Theme is animate here and at this point the purpose of the transfer from the bush to the house is merely to present the young woman to the speaker's parents (5). In sum, the clause in (9) designates an event of CAUSED MOTION and does not entail the establishment of a possessive relationship of any kind. An additional piece of evidence for the CAUSED MOTION rather than CAUSED POSSESSION meaning of le comes from compounds like le kal 'lift up' ; it is also supported by examples where le takes an inanimate Goal argument, as in (10b) below. Consequently, le cannot be assumed to contain a CAUSED POSSESSION root as part of its lexical meaning. As will be demonstrated below, the notion of CAUSED POSSESSION is conveyed by means of a possessive-like "adnominal" three-participant construction which is compatible with a number of different verbs ("adnominal" is adopted from Margetts & Austin (2007:426ff.) who call such possessive constructions in three-participant events an 'adnominal strategy'). Where le is used in such abnominal constructions it merely signals that the possessive relationship is brought about by an act of transfer.

3.3 Three-participant strategies

The verb le occurs in the following types of construction when expressing physical transfer to a third participant:
(10a)   ale    di    ne    le     kel      suw

        DISC   3SG   TAM   give   return   thither

        [[en tete ane].sup.T]   [[me-n   e     'ama'

        ART infant DEM          DAT-CS   ART   devil



        'Alright, then she handed the baby back to the
        devil.' ASB.067

(10b)   dir'ol   le     ba'a     [[di].sup.T]

        3TL      take   inside   3SG

        sar   [[le =n lolo nime].sup.GOAL]
        in    LOC =ART inside house

        'Then they brought her into the house.' ASMW.083

(10c)   Nik   e     le     =n     go-k           en    va'al

        2SG   TAM   give   =ART   POSS.EAT-1SG   ART   banana

        'Give me a banana [to eat].' JJQ.206

Two basic construction types can be distinguished on the basis of these data: example (10a) and (10b) represent the prepositional construction where the Recipient or Goal argument is marked by means of prepositional flagging. The second type, the adnominal construction, is illustrated by example (10c): here, the Recipient is encoded by means of a so-called possessive classifier (cf. 5.1 below). As will be outlined below, possessive classifiers specify the kind of relationship between possessor and possessum expressed in so-called indirect possessive constructions. Where these possessive constructions occur with le--and a number of other verbs- -the possessor may be understood as a Recipient coming into the possession of the possessed noun's referent.

The meaning expressed by the prepositional construction is clearly compatible with that of CAUSED MOTION and does not entail a CAUSED POSSESSION component as it is accessible for inanimate Goals which cannot be said to enter a possessive relation as a result of the transfer event. Even with dative PPs, a CAUSED POSSESSION reading seems to be inferred from the context rather than entailed. This is for instance the case in (11a) below, where the Recipient expressed by the dative PP comes into the possession of a cow. In contrast, in (10a) above, no possessive relationship is established as a result of the transfer. The adnominal three- participant construction, on the other hand, clearly designates events of CAUSED POSSESSION in all instances in the corpus, with the type of possessive relationship being specified by the possessive classifier deployed. The adnominal construction will be discussed in greater detail in section 5.

Both construction types represent indirective alignment types where the Theme argument is encoded in the same way as a monotransitive P argument entering a core argument relation, and the Recipient/Goal is realized as an oblique argument that differs in marking from P arguments (cf. Malchukov et al. 2010:3; Haspelmath 2005:2). The directional particles ma 'hither' and suw(o) 'thither' designate the movement direction of a Theme either towards the deictic centre or away from it. They represent another three-participant strategy that apparently contributes a CAUSED MOTION rather than a CAUSED POSSESSION meaning component. In this way these two elements may be used to add a Goal role to the event expressed, for instance in (9) above.

Frequently, however, these directionals co-occur with a referential expression of the Recipient, as is the case in the following example (cf. also (10a) above):
(11a)   Nik   e     le     ma       [[me no].sup.R]

        1SG   TAM   give   hither   DAT 1SG

        [[=n buluk ane].sup.T]   [...]

        =ART cattle DEM

        'You give me that cow [so that it will be mine],
        [...].' 1.NO.026

(11b)   ale      ne    le     suw       [[me di].sup.R]

        INTERJ   TAM   give   thither   DAT 3SG

        [[=n qe'e ne vowal].sup.T]

        =ART taro LIG one

        'Alright, then [she] gave him a taro.' ASMS.97

In the following two sections I will examine the prepositional and the adnominal construction in terms of the range of three-participant events they express and in terms of referential hierarchy effects.

4. The Prepositional Three-Participant Construction

4.1 Animacy and event types

Prepositional three-participant construction in Vera'a show a cross- linguistically widely attested "Differential R Marking" pattern (Kittila 2008:248ff.): inanimate Goals are marked by the locative preposition while animate Recipients are marked with the dative preposition when occurring in a three-participant construction with the transfer verb le.
(12a)   dirol   le      ba'a     [[di].sup.T]  sar

        3TL     bring   inside   3SG           in

        [[le =n lolo nime].sup.G]

        LOC =ART inside house

        'Then they brought her into the house.' ASMW.083

(12b)   Di    ne    le     [[=n sis].sup.THEME]

        3SG   TAM   give   =ART breast

        [[me =n tete ane].sup.GOAL]

        DAT =ART infant  DEM

        '(Then) she gave the breast to the baby [= breastfed
        the baby].' ASB.055

Besides marking Goals in three-participant constructions expressing transfer, the locative preposition is also used to mark Goal arguments in intransitive and monotransitive constructions or locative adjuncts, and to express Locations in non-dynamic states of affairs (cf. Schnell 2011:158ff.); the difference between the roles Goal and Location is not explicitly marked but only inferred from the semantic valence of the verbal predicate and context. The preposition le is also employed to express some Instruments. In the following example, the inanimate Patient argument is left implicit:
(13)   'a    =n     kele-gi     di    ne    lies

       LOC   =ART   after-3SG   3SG   TAM   rinse

       [[le   =n     do-     nir].sup.INSTR]

       LOC    =ART   leaf-   tree.sp

       'After this he will rinse [it; i.e. the kava]
       with a nir leaf.' JJKP.004

The use of le for Instruments demonstrates its close associations with 'additional' inanimate event participants. This marking of Instruments appears to be nevertheless odd in a cross-linguistic perspective, and in fact seems to be restricted to a limited set of Instruments in Vera'a, namely those that come close to a 'locative' participant in a broader sense: in (13) the leaf is used as a rinse for the kava, and it can thus be regarded as fulfilling a type of path role. The locative marking may thus be interpreted as a means of expressing a special type of locative role.

The dative preposition me is employed to express the Addressee of verba dicendi and other "mental transfer verbs" (cf. Malchukov et al. 2010:2), for instance the verbs 'aram 'tell, inform' and kaka 'tell, narrate':
(14a)   no    me    'aram    [[me   kumru].sup.ADDR]

        1SG   TAM   inform   DAT    2DL

        [[=n   maru-mruo].sup.T]

        =ART   uncle-2DL

        'I will point out your uncle to you.'HHAK.166

(14b)   Di    ne    kaka      [[me  dire].sup.ADDR]

        3SG   TAM   narrate   DAT   3PL

        [[=n   'ere   kaka    rekse   sivie   di    =m

        =ART    PL    story   like    how     3SG   =TAM

        van   [...]].sup.T]


        '(Then) he told them the stories about how he went
        (... and how he got back home).' ISAM.089

This use of me underscores its high affinity to animate participants in the role of a Goal or Recipient, as the Addressee of verba dicendi is necessarily animate: informing someone about something, namely about who 'their' uncle is in (14a) and about the story in (14b), is construed as a transfer of knowledge from one person to other persons.

It was already mentioned above that Theme arguments in three-participant constructions pattern with monotransitive P arguments. And animacy has the very same effect on thes Theme arguments as it has on monotransitive P arguments: where it is identifiable and activated, it takes the form of a bare pronoun that is incorporated into the VC if it has an animate referent. The only examples with an animate Theme in CAUSED MOTION events found in the corpus involve inanimate Goals marked by the locative preposition le:
(15a)   [[dire].sup.A]   =m     van   'o      kal

        3PL              =TAM   go    carry   move.upwards

        [[kamam].sup.T]   ma       [[le   =n     Vera'a].sup.G]

        1PL.EXCL          hither   LOC    =ART   Vera'a

        'They brought us up here to Vera'a.' GMV.020

(15b)   [[dire].sup.A]   =k     qere'   ba'a     [[di].sup.THEME]

        3PL              =TAM   push    inside   3SG

        sar   [[le   =n     mo-gi           =n     nime].sup.G]

        in    LOC    =ART   CLF.HOUSE-3SG   =ART   house

        'They jostled him into his house.' ISWM.171

Where the Theme is inanimate, it is left implicit:
(16a)   dire   =k     <eh> (0.4)   mom   le    =n     lolo

        3PL    =TAM   HES          put   LOC   =ART   inside

        qoro   ve'    [...]

        hole   rock

        'Then they put [it; leaf of plant] into the hole in
        the rock [where they make rain].' GAQ.007

(16b)   Di    ne    lele       suw       me    =n

        3SG   TAM   RED:give   thither   DAT   =ART



        '[He pinched off some Malay apples.] Then he gave [them,
        i.e. Malay apples] to his nephew.' JSV.069

Thus, in examples (15a) and (15b), the Theme is non-focal and thus realized as a bare pronoun incorporated into the VC. This contrasts with focal lexical Theme NPs which follow the VC and precede the locative PP expressing the Goal (cf. (3) above). There is no example in the current corpus where a pronominal focal Theme NP occurs together with a locative Goal PP; the expected position of such a pronominal NP would be between the VC and the locative PP, as the latter always have to occur at the right periphery of the clause. Further data will reveal whether this prediction is borne out.

In most examples examined thus far, the preposition takes a lexical NP complement and the PP occurs following an object NP where present. While locative le PPs are indeed restricted to lexical NP complements and post-object position, dative PPs allow for pronominal and personal in addition to lexical NP complements and show word order variation, depending on the referential properties and pragmatic role of the Recipient argument. This will be discussed in the following section.

4.2 Word order variation and referential form

As opposed to locative le, dative me may take personal or pronominal NPs, and bare pronouns as its complement in addition to lexical NPs. Depending on the type of complement, the PP occupies different syntactic positions: If the complement of the preposition is a lexical NP, the PP follows the object NP, as in (12b) above, repeated here as (17a). The same is true for personal NP complements which are accommodated by a construct suffix attached to me, as in (17b):
(17a)   Di    ne    le     [[=n   sis].sup.T]   [[me   =n

        3SG   TAM   give   =ART   breast         DAT   =ART

        tete     ane].sup.G]

        Infant   DEM

        '(Then) she gave the breast to the baby [= breastfed
        the baby].' ASB.055

(17b)   ale    di    ne    le     kel      suw       [[en

        DISC   3SG   TAM   give   return   thither   ART

        tete     ane].sup.T]   [[me-n   e     'ama'   ane].sup.R]

        infant   DEM           DAT-CS   ART   devil   DEM

        'Alright, then she gave the baby back to the devil.'

Where the complement of is a bare pronoun, it may precede the object NP immediately following the verb complex:
(18a)   nik   e     le     ma       [[me   no].sup.R]

        2SG   TAM   give   hither   DAT    1SG

        [[=n   buluk    ane].sup.T]   wo

        =ART   cattle   DEM           and

        no    =k     le     [[me   nike].sup.R]

        1SG   =TAM   give   DAT    2SG

        [[=n   seven bin].sup.T]

        =ART   seven beans

        'You give me that cow, and I will give the seven
        beans.' 1.NO.026

(18b)   ale      ne    le     suw       [[me   di].sup.R]

        INTERJ   TAM   give   thither   DAT    3SG

        [[=n   qe'e   ne    vowal].sup.T]

        =ART   taro   LIG   one

        'Alright, then [she] gave him a taro.' ASMS.97

However, a dative PP with a bare pronoun as its complement may also follow the Theme argument, as in (19a), and where the complement of is a pronominal NP rather than a bare pronoun, it has to follow the Theme argument, as in (19b). Here again, pronominal NPs pattern with lexical and personal NPs.
(19a)   di    ne    le     kel      suw       [[en

        3SG   TAM   give   return   thither   ART

        tete].sup.T]   [[me   di].sup.R],

        infant         DAT    3SG

        ne    van   qel         kel      row

        TAM   go    downwards   return   seawards

        '[She was about to go down to the sea again,
        and so the devil said: "Give be the baby

        Then she gave the baby back to her and went
        down to the sea again.' ASB.060

(19b)   Di    ne    le     [[=n   qe'e].sup.T]   [[me   di

        3SG   TAM   give   =ART   taro           DAT    3SG

        ane].sup.R]   ne    Van   sur    va'ane

        DEM           TAM   go    down   now

        'Then she gave the taro tuber to HIM and HE went
        down.' ASMS.106

While the selection of referential form of the complement of me, i.e. a lexical, personal, pronominal NP or a bare pronoun, is generally determined by a number of factors related to referential properties, as outlined for P arguments in section 2.2 above, the choice between a bare pronoun, as in (19a), and a pronominal NP, as in (19b), is driven by pragmatic factors: the more complex NP di ane 'him', consisting of a pronominal head di and the demonstrative ane, is employed here in order to foreground the relevant discourse participant so that it is activated and accessible for the implicit subject relation it bears in the subsequent clause. This becomes clear in comparison with (19a) where the subject of the subsequent clause is the same as that of the first clause. In (19b) it is the referent of di ane 'him' that is understood as the subject of the subsequent clause. Subjects are always topical in Vera'a, and the more complex PP construction is employed here to signal the shift of topic or to re-establish this participant as a topic in the following clause. I classify this as a focus function in the sense that the hearer's attention is directed to this participant, so it can be treated as topic in the subsequent discourse.

The available complement types for the prepositions le and me and their respective associations with the two word order patterns are summarized in Table 1. As can be seen from the table, only dative PPs with pronominal complements are amenable of occurring in a position between verb complex and Theme NP. Hence, the unmarked position for Goal and Recipient arguments is the right-most position in the clause. This pattern can be regarded as being motivated by principles of iconicity, so that the conceptual structure of a transfer event has its mirror image in linguistic structure: the Goal or Recipient participant, the endpoint of a transfer, is expressed after the Agent and the Theme in the linear structure of the clause. Furthermore, recall from section 2.2 that word order is quite rigid in Vera'a and in fact the only means of encoding the core argument relations S, A and P; oblique arguments and adjuncts usually occur in clause-final position. The order VC NP PP is thus the default order for the expression of a transfer event in Vera'a.

Table 1: Word order and form of complement of le and i

Given that pronominal dative PPs may occur in clause-final position, why do they also occur in a position immediately following the VC, contrary to principles of iconicity of conceptual structure and of rigid word order rules? Apparently, this question cannot be answered in terms of a rigorous categorical rule, as becomes clear from the word order variation observed with bare pronoun PPs (compare examples in (18) with example (19a)). This is so either because the corpus data currently available is not ultimately decisive in this regard, or because the varying orders are indeed conditioned by probabilistic rather than categorical rules, similar to those observed by Bresnan et al. (2007) for the dative alternation in English. If so, a number of factors may interact in motivating one of the two word order patterns available to pronominal PPs; these factors are summarized in Table 2. Note that in this table only the combination of lexical Theme arguments and pronominal Recipient arguments is considered, reflecting the structures illustrated in (18) and (19a); there are no examples in the corpus where both the Theme and Recipient argument are pronominal.

The factors relevant for pronominal dative PP placement are related to the Animacy, Referential and Focus Hierarchy. The examples in (18) represent the VC PP NP order. Here, the lexical Theme is non-human, definite (= n buluk ane 'that cow' in (18a)) or indefinite (=n qe'e ne vowal 'a (piece of) taro / some taro' in (18b)), and focal, as in all three clauses the attention of the hearer is directed to the Theme participant, and all clauses inform the hearer about the Theme, presupposing Agent and Recipient. In contrast, the pronominal Recipient is human, always definite, and non-focal. Hence, except for the definiteness of the Theme argument in (18a), the Recipient argument ranks higher on all hierarchies than the Theme argument. Example (19a) represents the structure VC NP PP. In this case, the Theme and the Recipient argument rank equally on all three hierarchies. However, the classification of the Recipient argument as human is slightly unclear, as the pronominal Recipient argument refers to a spirit rather than a human being. Although spirits are in many respects treated like human beings in the Vera'a culture and language, this example may suggest that finer-grained distinctions are nevertheless relevant. Also, the spirit is construed in the story as the antagonist of the child's mother who in turn is the protagonist or 'hero', and hence the clearly greater empathy with the former may also be of relevance here.

In sum, the placement of pronominal Recipient PPs seems to be influenced by three different hierarchies. The order VC NP PP is the default order, and the high ranking of the Recipient on the Referential Hierarchy and the Focus Hierarchy, and the low ranking of the Theme argument on the Animacy Hierarchy, apparently pull towards the more marked order VC PP NP. More corpus data are expected to reveal whether definiteness may indeed be of lesser importance, as is suggested by the first clause in (18a), or humanness of the Recipient argument is indeed of greater importance, as is suggested by (19).

What seems typologically remarkable about this pattern found in Vera'a is that bare word order change--without alignment change or construction split (Malchukov et al. 2010)-- is possible although word order is rigid and the marking of Recipient argument is done by means of prepositional flagging rather than NP case marking (cf. Rappaport Hovav & Levin 2008:160ff. for a comparison of English and Dutch with German in this regard).

5. The Adnominal Construction

The second type of three-participant construction in Vera'a draws on the structure of so-called indirect possessive constructions and is thus labeled "adnominal" here (cf. Margetts & Austin 2007:426ff.). The deployment of possessive structures in three-participant constructions has long been observed for many Oceanic languages (Lichtenberk 2002; Margetts 2002, 2004; Song 1997, 2005, 2007). Before turning our attention to adnominal three-participant constructions, I will briefly outline the system of possessive constructions in Vera'a. It will become clear that referential hierarchies have an effect on indirect possessive constructions (cf. Section 5.1); and this effect is carried over to the adnominal three-participant construction (cf. section 5.2).

5.1 Direct and indirect possessive constructions in Vera'a

Typical for an Oceanic language of Melanesia, Vera'a has two basic types of possessive construction, namely a direct and indirect construction (cf. Schnell 2011:117ff.), exemplified by (20) and (21), respectively, with pronominal, personal and lexical NP possessors (The possessor is rendered in bold face.):
(20a)   itok,     rem     'a    =n     kolo-k

        alright   climb   LOC   =ART   back-1SG

        'Alright, climb onto my back.' ISAM.056

(20b)   maranaga   'ale    =n     'o'ow    di    e

        chief      ASSOC   =ART   growth   3SG   ART

        gunu-n      e     Noileen

        spouse-CS   ART   N.

        'The chief for development, that's Noileen's
        husband.' GMV.061

(20c)   no    =k     ron    en    nelno   'ansara

        1SG   =TAM   hear   ART   voice   person

        'But now I hear a human voice/the voice of a
        person.' MVBW.071

(21a)   [...] =k   van   kal       ma       le    =n

        =TAM       go    upwards   hither   LOC   =ART

        qe'an    go-ruo

        ground   POSS.EAT-3DL

        'They (DL) decided to move up here onto their
        ground.' GMV.025

(21b)   lumgav      e     ne    vigir     en    gie   DEM   TAM   squeeze   ART   kava

        mo-n            e     'uvusmel

        POSS.DRINK-LK   ART   high.chief

        'The young man is squeezing the high chiefs kava
        /the kava for the high chief now.' JJKP.001

(21c)   di    ne    dada       =n     gie    mo

        3SG   TAM   RED:make   =ART   kava   POSS.DRINK



        'And now he is preparing the high chiefs kava/the
        kava for the high chief.' JJKT.001

In direct possessive constructions (henceforth DPCs), the possessor is attached directly to the possessed noun, either in form of a pronominal possessive suffix, a linking suffix plus adjacent personal NP, or an adjacent common noun. DPCs typically express inalienable relationships between possessum and possessor, for instance kin or part-whole relations. In indirect possessive constructions (henceforth IPCs), the possessor is attached to a so-called possessive classifier instead of the possessum noun itself. The possessive classifier specifies the type of possessive relationship (cf. Lichtenberk 1983). In DPCs, the possessor may have an inanimate referent, for instance when referring to a part of a plant, while in IPCs, possessors are restricted to animate referents (Schnell 2011:124ff., 131ff.). This restriction on possessors in IPCs apparently reflects the alienability of the expressed relationship which entails--though to varying degrees--control on behalf of the possessor, hence in turn entailing its animacy. Indirect possessive constructions are the ones employed in the adnominal construction (cf. (21b) and (21c) above where a benefactive reading is implied by the IPC), and I will outline some of their properties in some more detail in the remainder of this section.

Two types of IPCs can be distinguished, depending again on referential properties of the possessor. In one type of IPC, the possessive classifier together with the possessor follows the possessed noun, as in examples (21) above. In the other type of IPC, the possessive classifier plus possessor precedes the possessed noun. The first type of IPC will be labeled [-CONTROL] IPC and the latter [+CONTROL] IPC, because they are associated with different degrees of control over the possessive relationship, as will be outlined immediately below. While [-CONTROL] IPCs allow for all types of possessors, i.e. pronominal, personal and lexical, [+CONTROL] IPCs are restricted to pronominal possessors, occurring in the form of a pronominal possessive suffix.

Table 3 gives an overview of the different types of possessive construction in Vera'a. As can be seen from the schema of the NP structures involved, [-CONTROL] constructions are simple NPs with the possessive classifier plus possessor following the head noun. In contrast, [+CONTROL] IPCs consist of two conjoined NPs; the first one headed by a suffixed classifier, and the second one by the possessum noun. Both are introduced by the common article (e)n. Their conjunction into one complex NP is evidenced by their possible occurrence in one slot in the clause--where it could be replaced by a simple NP or a pronoun--and the fact that only the second NP allows for further modifiers or determiners.

Examples of [-CONTROL] IPCs and [+CONTROL] IPCs are given in (22a) and (22b):
(22a)   mul   din     kel      en    gamal   mu-gi

        go    reach   return   ART   house   POSS.GEN-3SG

        ne    mi'ir

        TAM   sleep

        'Came back to his house and went to sleep.' JJQ.344

(22b)   [...]   si   ne    kal     ma       le    =n

                or   TAM   enter   hither   LOC   =ART

        mu-k           en    gamal

        POSS.GEN-1SG   ART   men'

      '[I do not allow that anybody comes here and talks to
      me] or will come into my house.' ISWM.251

At the face of it, the two types of construction seem to be semantically interchangeable in these contexts. However, in (22a) the [-CONTROL] construction is employed, and this merely expresses that the house is owned by the possessor. In (22b), the use of the [+CONTROL] construction is used to express that the possessor controls the access to the inner of the house where he will perform a number of magic rituals in the context of this story.

Hence, the [-CONTROL] construction is used to express a possessive relationship that is construed as a given fact and left unchanged in the respective discourse context, as in (23a). In contrast, the [+CONTROL] construction is also employed to express that a specific possessive relationship is yet to be established, as in (23b):
(23a)   duru   =m     da   so      =k     van   kal

        3DL    =TAM   do   PROSP   =TAM   go    upwards

        ma       le    =n     qe'an    go-ruo

        hither   LOC   =ART   ground   POSS.EAT-3DL

        'They (DL) decided to come up here onto their
        ground [that they eat from].' GMV.025

(23b)   "Maranaga,   go-de               =n     kel   bigbig

        chief        POSS.EAT-1PL.INCL   =ART   big   'meat'

        row           e."   dem

        [Then they went up to the village to tell the chief.]

        '"Chief, there is a big [piece of] meat for us (to eat)
        down at the sea.

        [We have already tied it up (i.e. a turtle)."' GAQG.059

In (23a), the possession of the ground is a given fact in the context, and hence the [-CONTROL] construction is employed here. In example (23b), the context suggests that the referents of the possessive suffix are not yet in the possession of the 'meat' and hence the possessive relationship is rendered here as merely designated. The term "designated" will be used where a possessive relationship is not yet existent, but merely anticipated at a given point in discourse.

A further aspect of the two types of indirect possessive construction seems to be relevant here: in [-CONTROL] constructions, where the possessive classifier with the possessor occurs in immediate post-head position, it functions as a modifier of the head noun. Here, the (definite) personal pronoun suffix attached to the possessive classifier has the effect of marking the possessive NP as definite. This observation is supported by the fact that in the contexts of both (22a) and (23a), the 'house' and the 'ground', respectively, are given, hence identifiable to the hearer. In [+CONTROL] constructions, the second NP is always unmarked for definiteness, as in (22b) and (23b), and hence the referent of that NP may be identifiable to the hearer or not. That both interpretations are possible is supported by the contexts of these two examples: in (22b), the referent of the NP is identifiable to the hearer--in fact the hearers are standing in front of the 'house'. In (23b), the 'turtle' is mentioned to the chief for the first time; hence it is not identifiable to the hearer. Indeed, the fact that possessum NPs in [+CONTROL] constructions are left unmarked with regards to definiteness also allows for a non-specific reading of these NPs, and this will be relevant for the discussion below. It seems that these pragmatic features of indirect possessive constructions contribute to a possessive (proper) vs. designative reading of the two types of construction.

Crucially, this distinction is only available where the possessor is an identifiable and activated referent realized by pronominal form. These aspects of possessive constructions in Vera'a are carried over to three-participant constructions involving the same possessive structures, and this will be discussed in the following section.

5.2 The adnominal three-participant construction

Similar to other Oceanic languages (Lichtenberk 2002; Margetts 2002, 2004; Song 1997, 2005, 2007), IPCs in Vera'a are employed to express three-participant events. This was illustrated in example (10c) in Section 3 above where the verb le occurs with an indirect possessive construction and the clauses express the transfer of an item to a Recipient. Example (10c) is repeated below as (24):
(24)   Nik   e     le     =n     go-k           en    va'al

       2SG   TAM   give   =ART   POSS.EAT-1SG   ART   banana

       'Give me some bananas [to eat].' JJQ.206

As opposed to the prepositional construction, the adnominal construction is clearly associated with a CAUSED POSSESSION event schema in the sense of Rappaport Hovav & Levin (2008): in example (24), the Recipient argument is construed as 'possessing' the bananas as a result of the transfer, and even the type of possessive relation is specified by the respective possessive classifier go- which signals that a possessor has something in order to eat it. As outlined in the preceding section, it is the [+CONTROL] construction that triggers a designative possessor reading, and the deployment of le as the verbal predicate merely signals that the new possessive relationship is established via transfer of the Theme rather than its creation or otherwise obtaining.

In cases where the verb le is used in the sense of 'take' rather than 'give', the reading of the possessive construction depends on whether it is a [-CONTROL] or [+CONTROL] construction in the same vein as outlined above:
(25a)   dir   =k     le     =n     nak     su-suo

        3PL   =TAM   take   =ART   canoe   RED-paddle

        mu-gi          =k     suo      den

        POSS.GEN-3SG   =TAM   paddle   away

        'They took his canoe and paddled away.' JJQ.169

(25b)   di    =m     le     =n     mu-gi          =n     vus   [...]

        3SG   =TAM   take   =ART   POSS.GEN-3SG   =ART   bow

        'He took his bow [and went off (with it).].' ISWM.091

In example (25a), the possessive relationship is presented as given and of no further concern at this point in the discourse--the canoe is still the property of the possessor. In (25b), on the other hand, the possessive relationship is presented as something that will be of relevance for the subsequent discourse: the ownership itself can be regarded as being established prior to the event of taking the bow, and what is emphasized here is that the possessor can use the bow after having taken it.

Where [+CONTROL] IPCs occur as objects of verbs of creation, again the possessive relationship is understood as being designated rather than already established, so that the possessor expresses the semantic role of a Beneficiary. In these cases, the designated possessive relationship is brought about through the creation of an entity rather than its transfer:
(26a)   nik   e     da     =n     mo-k

        2SG   TAM   make   =ART   POSS.DRINK-1SG

        en    gie    le    =n     gamal

        ART   kava   LOC   =ART   mens'.house

        '[In former times,] you would have prepared
        my kava/kava for me in the men's house.'

(26b)   nik   e     'es     en    mu-madu

        2SG   TAM   carve   ART   POSS.GEN-1DL.EXCL

        =n     vus

        =ART   bow

        'You cut our bows/bows for us.' PAWW.013

With a lexical or personal NP possessor, only a [-CONTROL] construction is available, and this may nevertheless have a CAUSED POSSESSION reading when occurring as objects of verbs of creation:
(27a)   lumgav      e     ne    vigir     en    gie   DEM   TAM   squeeze   ART   kava

        mo-n            e     'uvusmel

        POSS.DRINK-LK   ART   high.chief

        'The young man is squeezing the kava for the
        high chief.'JJKP.001

(27b)   di    ne    dada       =n     gie    mo           'uvusmel

        3SG   TAM   RED:make   =ART   kava   POSS.DRINK   high.chief

        'And now he is preparing kava for the high chief.' JJKT.001

Verbs of creation do not occur with [-CONTROL] constructions with pronominal possessors. However, where [-CONTROL] constructions occur with pronominal possessors and function as objects of verbs of dispossession, like bol 'steal', the possessor refers to the original owner from whom the canoe is stolen rather than the ones who take the canoe. This means that the possessive relationship is not altered here during the event designated by the clause:
     dir   ga  moros    dir   =m     bol   en   nak    mu-gi
     3PL  TAM  want     3PL   =TAM   steel ART  canoe  POSS.GEN-3SG
     'They wanted to steal his canoe.' JJQ.134

To summarize, the adnominal three-participant construction is undoubtedly associated with a CAUSED POSSESSION event schema. It is restricted to [+CONTROL] constructions. A caused possession reading may arise with [-CONTROL] constructions with personal or lexical possessors when occurring as objects of verbs of creation, and is excluded with [-CONTROL] constructions with pronominal possessors. The semantics of possessive constructions is fully intact in adnominal three-participant constructions; different types of possessive classifiers may be employed to specify the nature of the resulting possessive relationship.

It appears that in Vera'a, the CAUSED POSSESSION reading of clauses with possessive constructions as objects does not fully depend on the nature of the predicate they occur with, as is suggested in the literature on possessive-like benefactive constructions in Oceanic (cf. Lichtenberk 2002; Margetts 2004; Song 2005, 2007). Rather, such a reading is already present in the type of possessive construction employed and the mere result of referential hierarchy effects, namely of the Nominal Hierarchy that permits pronominal possessors in a pre-head position. This opens up the possibility to move the possessor outside of the NP, and hence where the possessum NP follows the possessor in a [+CONTROL] construction, it is left unmarked with regards to definiteness, allowing for an indefinite or even non-specific reading.

5.3 Pronominal Beneficiary incorporation

Departing from the structures examined in section 5.2, which involve a [+CONTROL] IPC in object position, Vera'a has developed a specialized Recipient/Benefactive construction in which a suffixed possessive classifier is incorporated into the verb complex. Consider the following example:
(29)   Nik   e     le     go-k           lik

       2SG   TAM   give   POSS.EAT-1SG   more

       ma       =n    'ilise!

       hither   =ART   sea.almond

       'Give me [some] more [almonds]!' ISGG.E.024

In (29) the possessive suffix on the classifier bears the semantic role of a Recipient in much the same way as it does in adnominal three-participant constructions discussed in section 5.2. In contrast to the adnominal construction, however, the suffixed classifier is not a NP-internal constituent here, as is witnessed by the absence of an article. Instead, it occurs in a VC-internal position before the directional particle ma and the adverb lik 'more'. The Theme NP follows the VC in object position. The same type of construction occurs with verbs of obtaining and creation:
(30a)   nik   me    bis         go-k           ma

        2SG   TAM   POSS.EAT-1SG   hither

        'Pinch off some [to eat] for me.' JSV.066

(30b)   N     ankol   mu-ru          ane   ne    'es

        ART   uncle   POSS.GEN-3DL   DEM   TAM   carve

        mu-ru          =n     vus

        POSS.GEN-3DL   =ART   bow

        'And then their (DL) uncle cut bows for them
        (DL).' ANV.029

It is not clear at present what the functional difference is between this type of three-participant construction and the adnominal construction investigated in section 5.2, and whether there is one at all. Francois (2001:564) reports an analogous construction in the closely related language Mwotlap, and analyses it as bearing an "indefinite partitive" meaning. It seems, however, that in Vera'a this reading is present with the adnominal construction to a large extent. The VC-internal position of pronouns with the semantic role of a Recipient/Beneficiary may be a mere reflection of the tendency for pronouns to occur closer to or even within the verbal predicate, thus again resembling the structural properties of pronominal P arguments. The choice between the two types of construction may thus be a matter of free variation. Again it must be stressed that the resemblance with P-like properties is restricted to the position of the pronoun, with the alignment type not being altered: it is still indirective as the pronominal possessor is still an embedded constituent of the incorporated classifier construction. This is typologically interesting as in many cases where bound pronouns on verbs express a Recipient, this involves an alternation of alignment from indirective to secundative.

6. Summary

This paper investigates the connections between referential hierarchies and the constructional variation found with three-participant constructions in Vera'a. As for the prepositional construction, the associations between different values on the Animacy, the Nominal and the Focus Hierarchies, and formal characteristics of expressions for Goal and Recipient are rendered in Figure 1. Crucially, only a dative PP which has a non-focal bare pronoun as its complement and expresses an animate Recipient or Addressee may immediately follow the verb complex before the Theme NP. Dative PPs with other types of complement have to follow the Theme NP. The morphosyntactic realization of the Theme is, however, determined by similar factors as the realization of Recipient: non-focal bare pronouns bearing this role are incorporated into the verb complex like P arguments in monotransitive clauses. All other types of Theme or P arguments occur as NPs on clause level.


As for the adnominal three-participant construction involving an indirect possessive NP construction, non-pronominal possessors have to follow the possessed noun, and here a Recipient/Beneficiary reading is implied in certain contexts. Where the possessor is pronominal and follows the possessed noun, this reading is excluded. only pronominal possessors may, however, occur preceding the possessed noun. This type of construction may express that the possessive relationship is merely anticipated rather than already established and given, hence allowing for a designative reading. The entire clausal construction then expresses a three-participant event of CAUSED POSSESSION including a Recipient/Beneficiary role. In this latter type of adnominal construction, the possessor occurs next to the verb complex similar to dative PPs. Departing from this structure, such pronominal possessors are incorporated into the verb complex. Nevertheless, argument alignment is indirective in both types of construction and all constructional variants.

As for the choice between the two types of construction, the prepositional construction is associated with a CAUSED MOTION reading, and the adnominal construction with a caused possession reading. The verb le, the nearest equivalent to English 'give', is restricted to the expression of CAUSED MOTION events, and where it occurs in an adnominal construction, this means that the possessive relationship is brought about by a CAUSED MOTION event.


Andrews, Avery. 2007. The major functions of the noun phrase. Language Typology and Syntactic Description. 2ndEdition ed. by Timothy Shopen, 132-223. Cambridge: Cambridge University Press.

Bresnan, Joan, Anna Cueni, Tatjana Nikitina, and Harald Baayen. 2007. Predicting the dative alternation. Cognitive Foundations of Interpretation, ed. by Gerlof Bouma, Irene Kraemer, and Joost Zwarts, 69-94. Amsterdam: Royal Netherlands Academy of Science.

Croft, William. 2003. Typology and Universals. 2ndEdition. Cambridge: Cambridge University Press.

Francois, Alexandre. 2001. Contraintes de structures et liberte dans l'organisation du discours. Une description du mwotlap, langue oceanienne du Vanuatu. Doctoral thesis in Linguistics, Universite Paris-IV Sorbonne.

-----. 2005a. A typological overview of Mwotlap. Linguistic Typology 9/1. 115- 146.

-----. 2005b. Unraveling the history of the vowels of seventeen Northern Vanuatu languages. Oceanic Linguistics 44/2. 443-504.

-----. 2007. Noun articles in Torres and Banks languages: Conservation and Innovation. Language description, history and development: Linguistic indulgence in memory of Terry Crowley ed. by Jeff Siegl, John Lynch, and Diana Eades, 313-326. Amsterdam & New York: John Benjamins.

-----. 2009. Verbal spect and personal pronouns: The history of aorist markers in North Vanuatu. Austronesian historical linguistics and culture history: a festschrift for Bob Blust ed. by Andrew Pawley and Alexander Adelaar, 179-195. Canberra: Pacific Linguistics.

Genetti, Carol and Laura D. Crain. 2003. Beyond preferred argument structure: Sentences, pronouns and given referents in Nepali. Preferred Argument Structure: Grammar as Architecture for Function, ed. by John W. Du Bois, Lorraine E. Kumpf, and William J. Ashby, 197-245. Amsterdam & Philadelphia: John Benjamins.

Haig, Geoff and Stefan Schnell. 2011. Annotations using GRAID (Grammatical Relations and Animacy in Discourse). Introduction and guidelines for annotators. Version 6.0. Available at:

Haig, Geoff, Stefan Schnell, and Claudia Wegener. 2011. Comparing corpora from endangered language projects: Explorations in language typology based on original texts. Documenting endangered languages ed. by Geoffrey L.J. Haig, Nicole Nau, Stefan Schnell, and Claudia Wegener, 55-86. Berlin: Mouton de Gruyter..

Haspelmath, Martin. 2005. Argument Marking in Ditransitive Alignment Types. Linguistic Discovery 3/1.1-21.

-----. 2010. Comparative concepts and descriptive categories in cross-linguistic studies. Language 86.663-687.

Hyslop, Catriona. 2002. The Lolovoli dialect of North-East Ambae, Vanuatu. Canberra: Pacific Linguistics.

Kittila, Seppo. 2007. On the encoding of transitivity-related features in the indirect object. Functions of Language 14/1.149-164.

-----. 2008. Animacy effects on Differential Goal marking. Linguistic Typology 12/2.245-268.

Lambrecht, Knut. 1994. Information structure and sentence form. Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press.

Lichtenberk, Frantisek. 2002. The Possessive-Benefactive Connection. Oceanic Linguistics 41/2.439-474.

Malchukov, Andrej & Martin Haspelmath & Bernard Comrie. 2010. Ditransitive constrcutions: a typological overview. Ditransitive constructions: a comparative handbook ed. by Andrej

Malchukov, Martin Haspelmath, and Bernard Comrie, 1-64. Berlin and New York: Mouton de Gruyter.

Margetts, Anna. 1999. Valency and transitivity in Saliba, an Oceanic Language of Papua New Guinea. MPI Series in Psycholinguistics 12. Nijmegen.

-----. 2002. The linguistics encoding of three-participant events in Saliba. Studies in Language 26/3.613-636.

-----. 2004. From Implicature to Construction: Emergence of a Benefactive Construction in Ocenic. Oceanic Linguistics 43/2.445-468.

Margetts, Anna & Peter K. Austin. 2007. Three participant events in the languages of the world: towards a crosslinguistic typology. Linguistics 45/3.393-451.

Rappaport Hovav, Malka & Beth Levin. 2008. The English dative alternation: The case for verb sensitivity. Journal of Linguistics 44/1.129-167.

Schnell, Stefan. 2011a. A grammar of Vera'a, an Oceanic language of North Vanuatu. PhD Thesis. Christian-Albrechts-Universitat zu Kiel.

-----. 2011b. Pronominal reference in Vera'a discourse. Paper presented at the International workshop on the languages of Vanuatu, Kioloa, 21-23 october 2011.

Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press.

Silverstein, Michael. 1976. Hierarchy of features and ergativity. Grammatical categories in Australian languages, ed. by Robert M.W. Dixon, 112-171. Canberra: Australian Institute of Aboriginal Studies.

Song, Jae Jung. 1997. The history of Micronesian possessive classifiers and benefactive marking in Oceanic languages. Oceanic Linguistics 36/1.29-64.

-----. 2005. Grammaticalization and structural scope increase: possessive- classifier-based benefactive marking in oceanic languages. Linguistics 43/4.795-838.

-----. 2007. Getting three out of two. The development of a three-participant construction in Oceanic languages. Functions of Language 14/1.127-148.

Thieberger, Nicolas. 2006. The Benefactive Construction in South Efate. Oceanic Linguistics 45/2.297-310.

Author's Contact Information:

Stefan Schnell

Centre for Research on Language Diversity

La Trobe University

(1) The research presented in this paper was undertaken in connection with the DoBeS project "Documentation of Vera'a and Vures, the two surviving endangered languages of Vanua Lava, Vanuatu" funded by grant II/81 898 from the Volkswagen Stiftung. I am furthermore grateful to two anonymous reviewers for various comments that helped enormously to improve the paper. Finally, I'd like to thank the editor of this volume, Eva van Lier, for most valuable comments on linguistic analyses as well as suggestions concerning their presentation. I am of course responsible for all remaining errors.

(2) Morpheme glossing follows the Leipzig Glossing Rules. In addition, the following glosses and abbreviations are used: ADDR 'ADDRESSEE role'; ANT 'anticipatory marker'; CARD 'cardinal numeral prefix'; CS 'construct suffix'; DISC 'discourse particle'; G 'GOAL role'; HES 'hesitation': INSTR 'instrument'; INTERJ 'interjection'; LIG 'ligature'; POSS.DRINK 'possessive classifier specifying drink relation', POSS.EAT 'possessive classifier specifying food possession', POSS.GEN 'possessive classifier for underspecified possessive relations'; POSS.HOUSE 'possessive classifier specifying possession of a house'; POSS.VES 'possessive classifier specifying possession of a vessel'; PROSP 'prosepective marker'; R 'RECIPIENT role'; T 'THEME role'; VC 'verb complex'

(3) I gloss all TAM particles simply as TAM, as the function of TAM markers does not appear relevant to the topic of the current paper.

(4) GRAID (Grammatical Relations and Animacy in Discourse) is a system of annotation conventions that facilitates quantitative investigations of texts in terms of animacy and referentiality. GRAID glosses capture the referential form, animacy features and syntactic functions of referential expressions. The GRAID manual (Haig & Schnell 2011) comprises glossing conventions and general guidelines for annotators.

(5) Of course, as an anonymous reviewer points out, the woman will become a POSSESSUM of his parents in some sense, but I believe that this is not relevant at this very point in the story where the focus is clearly on presenting the wife to the parents for 'inspection', rather than 'giving' her to them as a daughter- in-law.
Table 1. Word order and form of complement of le
and me

                    VC NP PP        VC PP NP

complement of le.   common NP       NOT ATTESTED

complement of me.   bare pronoun    bare pronoun

                    pronominal NP
                    personal NP
                    common NP

Table 2: Factors influencing the placement of pronominal dative PPs

                  VC NP PP                 VC PP NP


Animacy       human       human (?)   human       non-human

Referential   definite    definite    definite    definite/
Hierarchy                                         indefinite

Focus         non-focal   non-focal   non-focal   focal

Table 3: Types of possessive construction with different types of

          DPC                   [-CONTROL] IPC     [+CONTROL] IPC

pron.     [[ART N.sup.POSS'M]   [[ART N.sup.       [[[ART CLF-PRO].
POSS'R    [-PRO.sup.POSS'R]]    POSS'M][CLF-PRO.   sup.POSS'R]
                                sup.POSS'R]]       [[ART N].sup.

pers.     [[ART N.sup.POSS'M]   [[ART N.sup.       NOT ATTESTED
POSS'R    -CS [[ART N].sup.     POSS'M] CLF-CS
          POSS'R]]              [[ART N].sup.

pex.      [[ART N.sup.POSS'M]   [[ART N.sup.       NOT ATTESTED
POSS'R    [N.sup.POSS'R]]       POSS'M] CLF
COPYRIGHT 2012 Dartmouth College Library
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2012 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Schnell, Stefan
Publication:Linguistic Discovery
Date:Sep 1, 2012
Previous Article:Referential hierarchies in three-participant constructions in Blackfoot: the effects of animacy, person, and specificity.
Next Article:Referential hierarchy effects in Yakkha three-argument constructions.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |