Schnell, Stefan. 2014. Vera'a annotation notes. Implementation of GRAID annotations in Vera'a...
Transcript of Schnell, Stefan. 2014. Vera'a annotation notes. Implementation of GRAID annotations in Vera'a...
Multi-CAST veraa annotation notes
Multi-CAST veraa annotation notes 2014-11-06
Schnell, Stefan.Notes on the annotation of Vera’a (veraa)
Overview
This document outlines the mophosyntactic structure of Vera’a, and describes theimplementation of GRAID glossing conventions, as outlined in the GRAID7 Manual(Haig & Schnell 2014). Section 1 gives an overview of Vera’a basic morphosyntacticfeatures and the implementation of the core set of GRAID glosses. Section 2 dealswith the treatment of other finite and non-finite types, and Section 3 with that ofcomplex sentences. Morphological glosses and form paradigms are provided in theAppendix. The following table of contents is intended to serve as a quick referenceto individual passages.
Table of Contents
1 Basic structural features of Vera’a and GRAID glossing 31.1 Clause structure and syntactic functions . . . . . . . . . . . . . . . . 3
1.1.1 Verbal clauses and syntactic functions . . . . . . . . . . . . . 31.1.2 Non-verbal clauses and syntactic functions . . . . . . . . . . . 71.1.3 Other syntactic functions . . . . . . . . . . . . . . . . . . . . 9
1.2 Form of referential expressions . . . . . . . . . . . . . . . . . . . . . 101.2.1 Common NPs . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.2.2 Personal NPs . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.2.3 Locative NPs . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.2.4 Numeral phrases . . . . . . . . . . . . . . . . . . . . . . . . . 121.2.5 Pronominal expressions . . . . . . . . . . . . . . . . . . . . . 121.2.6 Further types of expression: demonstratives, interrogative
nouns, conjoined NPs, etc. . . . . . . . . . . . . . . . . . . . . 151.3 Animacy and person of referential expressions . . . . . . . . . . . . . 181.4 Other elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4.1 Adverbs and clause-level demonstrative forms . . . . . . . . . 181.4.2 Particles, conjunctions . . . . . . . . . . . . . . . . . . . . . . 19
2 Clausal constructions with special features 192.1 Negation and neutralisation of syntactic categories: <np:pred> vs.
<v:pred> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2 Finite and non-finite clause constructions: <v> vs. <vother> . . . . 20
2.2.1 Imperative constructions and <vother> . . . . . . . . . . . . 202.2.2 ga-construction: <vother> . . . . . . . . . . . . . . . . . . . 212.2.3 Head-tail construction and zero TAMP marking . . . . . . . 22
3 Complex sentences, direct speech, clause repetitions and complexpredicates 243.1 Complement clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.1 Syndetic complement clauses . . . . . . . . . . . . . . . . . . 243.1.2 Asyndetic complement clauses . . . . . . . . . . . . . . . . . 25
1 of 38
Multi-CAST veraa annotation notes
3.2 Adverbial clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.3 Relative clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 Syndetic relative clauses . . . . . . . . . . . . . . . . . . . . . 273.3.2 Asyndetic relative clauses . . . . . . . . . . . . . . . . . . . . 273.3.3 Function of relative clauses . . . . . . . . . . . . . . . . . . . 29
3.4 Direct speech . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.5 Predicate or clause repetition . . . . . . . . . . . . . . . . . . . . . . 303.6 Complex predicates versus clause chaining . . . . . . . . . . . . . . . 31
A Appendix 32A.1 Notes on morphological glossing . . . . . . . . . . . . . . . . . . . . . 32A.2 Tense, aspect, mood, polarity marking . . . . . . . . . . . . . . . . . 34A.3 Person markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35A.4 Demonstrative forms . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
References 37
2 of 38
Multi-CAST veraa annotation notes
1 Basic structural features of Vera’a and GRAIDglossing
Vera’a is an isolating language with grammatical a�xes being confined to possessivepronominal su�xes on bound nouns. Exponents of some TAM categories and thecommon NP article are enclitics and occur detached from their functional heads.Typically, phrases consist of at least two and often more words, e.g. TAM marker +verb; article + noun, and sub-constituents are all glossed as such. In the followingI outline the basic structure of the Vera’a language alongside their handling inGRAID annotations.
1.1 Clause structure and syntactic functions
Vera’a has two basic clause types, verbal and non-verbal ones. These have di↵erenttypes of predicate expressions, a verb complex (VC henceforth) in verbal ones andsome other type of phrase in non-verbal ones.
1.1.1 Verbal clauses and syntactic functions
A verbal clause need not have any argument expression, and may consist of onlythe VC functioning as the predicate, as in (1):
(1) veraa isam 032
##
netam2:3sg
lv-pro:s
marandaylightv:pred
’(And then) it became daylight.’
As in (1), the VC receives the form gloss <v>, regardless of whether its head worditself is unambiguously classified as a ’verb’. Core argument function S, A and P(in the sense of Andrews 2007) are encoded by the position of the NP or pronounrelative to the VC. S and A arguments precede the VC, P arguments occupy apost-verbal position, as in the following examples (2–4):
(2) veraa palaa 003
##
renewomannp.h:s
netam2:3sg
lv-pro.h:s
wotoqtoqopregnantv:pred
’(And then) the woman got pregnant.’
(3) veraa gabg 083
##
rov-rov’ered-close.toother
nik2sg
pro.2:a
etam2
lv
kurgnawv:pred
kirmobreakrv
=n=art
=ln
gakostalknp:p
wova’alpawpawrn
edem3
rn
”...would you almost have gnawed o↵ the stalk of the pawpaw (fruit)?”
3 of 38
Multi-CAST veraa annotation notes
(4) veraa asms.e 068
##
Mag’ieold.womannp.h:a
anedem1.a
rn
netam2:3sg
lv-pro.h:a
vesiraskv:pred
saemph
other
=n=art
=ln
’ama’devilnp.d:p
edem3
rnsoquot
other
’(And then) the old woman asked the devil:’
Where the S or A function is expressed by a pronoun, it will occupy the samepre-VC slot where lexical S or A arguments occur, as in (5–6):
(5) veraa iswm 208
##
dir3pl
pro.h:a
=m=tam1
=lv
vuskillv:pred
die3sg
pro.h:p
’They killed him.’
(6) veraa isam 061
##
duru3dl
pro.h:s
=k=tam2
=lv
kalenterv:pred
ba’aintorv
kelbackrv
sareinother:l
’Then the two went ashore again.’
Where the P function is expressed by a bare pronoun, this pronoun is incorporatedinto the VC, as in example (7) where 3SG di precedes the directional adverb sar’in(wards)’:
(7) veraa iswm 171
##
dir3pl
pro.h:a
=ek=tam2
=lv
qerepushv:pred
ba’aintorv
di3sg
rv pro.h:p rv
sarinadp
leloc
=ln
=n=art
ln-pro.h:possmo-giposs.house-3sg=ln
=n=art
np:g
nimehouse
’They pushed her into her house.’
It seems that in some cases, bare pronouns may also follow the VC; in other—probably most—cases, this question is not decidable in particular contexts, asin (5) above. Bare P pronouns are glossed for VC-internal position (by preposed<rv >) only in those cases where this is indicated by the presence of other VC-internal elements following it, as in (7).
Oblique arguments are encoded by means of prepositional flagging, and soare adjunct functions. Both occupy positions following the VC. Three types ofoblique arguments are considered in the GRAID manual (Haig & Schnell 2014:13f.),i.e. those expressing locations (<l>), goals (<g>), or some other semantic role((<obl>), and all three are also distinguished in Vera’a. Examples (8–9) showlocation and goal roles:
4 of 38
Multi-CAST veraa annotation notes
(8) veraa iswm 004
##
duru3dl
pro.h:s
gastat
lv
’ogstayv:pred
walexactlyrv
saemph
other
leloc
adp
=n=art
=ln
vono-nhome-csnp:l-np.h:poss
epers.art
rn
WowotW.rn np.h:poss
’aloc.sp
rn
NosN.rn
’The were living right up in Wowot’s home village at Nos.’
(9) veraa jjq 008
##
dir3pl
pro.h:s
=m=tam1
=lv
vangov:pred
kalupwardsrv
sarinlandrv
leloc
adp
=n=art
=ln
womomo’bushnp:g
’He went down to the reef...’
In all three examples, the same basically locative preposition le is the head ofthe PP expressing either a location or a goal, with more specific semantic roleinterpretations relating to di↵erences in verbal semantics and world knowledge. Adative preposition is used where location or goal are human participants. Examplesof human locations did not occur in GRAID corpora so far, but would be glossedas done for the following elicited example in (10):
(10) (observed/elicited)
##
babutother
=n=art
=ln
gaselknifenp:s
gastat
lv
’og’ogred:stayv:pred
me-ndat-cs
adp
epers.art
ln
JanetJ.np.h:l
’The knife is with Janet.’
Goal-like roles carried out by humans are recipients/beneficiaries and addressees,and are all glossed with <:g>, as shown in (11–13):
(11) veraa mvb 103
##
000.h:a
netam2:3sg
lv-pro.h:a
letransferv:pred
=n=art
=ln
biegbreadfruitnp:p
nenum.art
rn
vo-walnum-onern #
woandother
000.h:a
netam2:3sg
lv-pro.h:a
letransferv:pred
medat
adp
di3sg
pro.h:g
nenum.art
ln
vo-walnum-onenp:p
’...took a breadfruit and gave her one (as well).’
(12) veraa anv 026
##
duru3dl
pro.h:a
=k=tam2
=lv
..e..hesnc
sorwearv:pred
medat
adp
duru3dl
pro.h:g
=n=art
=ln
gogovclothesnp:p
’They put them clothes on.’
5 of 38
Multi-CAST veraa annotation notes
(13) veraa jjq 318
##
TumerenT.np.a:s ds
netam2:3sg
lv-pro.h:s
teksayv:pred
medat
adp
dire3pl
pro.h:g
’Then Tumeren said to them:...’
In three-participant constructions, word order may vary slightly, according toconsiderations of referentiality and animacy features of arguments (cf. Schnell2012), demonstrated by (11) and (12). Hence, NPs with P function may actuallyoccur following a dative (or ablative for that matter) PP; thus, P NPs are thosethat are not flagged by a preposition and occur in some post-VC position. Theroles of recipients or beneficiaries may also be expressed by possessive morphology,which is glossed <:poss> for possessor, as the specific reading as either possessoror recipient/beneficiary is a matter of inference rather than encoding.
In accordance with the GRAID7 Manual (Haig & Schnell 2014), no sharpdistinction is made between arguments and adjuncts. Thus, locative PPs as in(14) would also be receiving the <l> function gloss. The same holds for argu-ments/adjuncts expressing a goal (<:g>) or some other semantic role (<:obl>).
(14) veraa jjq 310
##
kamam1pl.ex:tam1
pro.1:s
mi’irsleepv:pred
leloc
adp
=n=art
=ln
qanrisovennp:l
’We slept in the stone oven.’
Other oblique arguments express a variety of semantic roles. In some instances,the choice of a particular preposition unambiguously encodes a particular semanticrole, for instance source being expressed by an ablative preposition in (15), whilein other instances verb semantics and context reading appear to play an importantpart, as in (16) where the instrument reading is not encoded as such by the locativepreposition:
(15) veraa jjq 346
##
000.h:s
manpfv
lv
kaluexitv:pred
denabl
adp
enart
ln
wiobamboonp:obl #
dir3pl
pro.h:s
manpfv
lv
’eqeldescendv:pred
’(They) had already come out of the bamboo, they had already come down.’
(16) veraa as1 102
##
babutother
di3sg
pro:s
gastat
lv
manamagicalv:pred
’idel
rv
leloc
adp
=n=art
=ln
rawhermaphrodite.pignp:obl
wuvaonlyother
’But it [i.e. some water] is magic only through a hermaphrodite pig.’
In accordance with the GRAID7 Manual, the glossing of oblique PP arguments aseither <l>, <g> or <obl> follows semantic role considerations rather than formalones. Thus, the locative PP in (16) is glossed as bearing <obl> rather than <l>or <g> function because it expresses the semantic role of an instrument.
6 of 38
Multi-CAST veraa annotation notes
Clear instances of circumstantial adjuncts are glossed for their form and receivethe function gloss <other>. This is typically the case with temporal PP or NPadjuncts, as in (17):
(17) veraa jjq 393
[...]
##
no1sg
pro.1:s
=m=tam1
=lv
vangov:pred
mahitherrv
leloc
adp
=n=art
=ln
qonnightnp:other
[...]
”...I came here last night, (but then where were you guys?)”
For clause-level adverbs and other types of one-word modifiers the gloss<other>is used, not further classifying form and function distinctly.
1.1.2 Non-verbal clauses and syntactic functions
The predicate of a non-verbal clause is a phrase of various types, but not a VC.These phrases are glossed for their form like arguments and take the function gloss<pred>:
(18) veraa mvb 009
##
nart
ln
kakastorynp:dt s
agenedem2
rn
di3sg
pro:s
=n=art
=ln
kakastorynp:pred
neleovoicern
vu’spiritrn
’This story here, it is a customary story [lit. a spirits’ voice].’
(19) veraa as1 083
##
babutother
kumru2dl
pro.2:s
’o’withadp
=n=art
=ln
wovinqacoconut.shellnp:pred
’But do you have a coconut shell with you?’
Thus, it is a NP in (18) and a PP in (19) that bear predicate function in the respec-tive non-verbal clauses. As shown in these two examples, the subject expression ina non-verbal clause is considered to have S function, glossed <:s>. As with verbalclauses, non-verbal clauses may not contain a subject relation at all, as in followingexamples (20–21):
(20) veraa isam 002
##
qondaynp:other
nenum.art
rn
vo-walnum-onern
’erepl
ln
’ansarpersonnp:predex
’aloc.sp
rn
LemerigL.rn np
’Once upon a time, (there were) the people of Lemerig.’
(21) veraa panr 010
[...]
##
=n=art
=ln
lanwindnp:a
vushitrn
mreallv
vushitv:pred
kamam1pl.ex
pro.1:p ##
ecc
other
=n=art
=ln
marfaminenp:predex
’... [when] a hurricane hits us and (when) (there is) famine.’
7 of 38
Multi-CAST veraa annotation notes
Such clauses are existential clauses, i.e. they express that an entity or state ofa↵airs exists or has come into being. The predicates of these clauses receive thefunction gloss <predex> for ’existential predicate’.
There are a number of other elements, neither NPs nor PPs, that may functionas predicate. All of these are glossed as <other:pred(ex)>. Examples are thequotative particle so that accommodates direct speech in the matrix clause, as in(22), a bare numeral, as in (23), or the existential bene ’there is’, as in (24):
(22) veraa as1 040
##
epers.art
ln
DolD.np.h:s
soquot
other:pred ##ds
onoother
no1sg
pro.1:s
manpfv
lv
qe’finishv:pred
[...]
’Dol said: ’No, I am done. (The kava has already made me drunk.)”
(23) veraa jjq 003
##
epers.art
ln
ragapeopleln
’i-’isi-ginsg-same.sex.sibl-3sgnp.h:s
sanwultenother:pred
walonern
deme??rn
nenum.art
rn
vo-ruonum-tworn
’His brothers were twelve.’ [i.e. ’He had twelve brothers.’]
(24) veraa gabg 043
[...]
##
siifother
=n=art
=ln
wova’alpawpawnp:s
beneexistother:predex ##
du1dl.in
wpro.1:a
=k=tam2
=lv
geneatv:pred
000:p
’... (and) if there are pawpaw fruits we will eat (them).’
Where existence is expressed by bene (or likewise non-existence/absence by itsnegative counterpart gitag), it receives the function gloss <:predex>, and the NPdenoting the entity that is said to exist is considered an S argument. Where theseexistential particles occur with a locative oblique argument, the clause may havelocational or existential semantics and in either case is glossed as in the examplesin (25–26):
(25) veraa jjq 338
##neg
dir3pl
pro.h:s
ne gitagnot.existother:pred
leloc
adp
=n=art
=ln
bo-reposs.bed-3plnp:l
’They are not in their beds.’
(26) veraa iswm 175
##
nart
ln
qoro-gihole-3sgnp:s
luwobigrn
beneexistother:predex
suweidownother:l
’It had a big hole at the bottom.’ (lit.: ’A big hole of it existed at thebottom.’)
The locative expressions in the two examples presumably have di↵erent statuses,but the di↵erence is not noted in GRAID annotations.
8 of 38
Multi-CAST veraa annotation notes
1.1.3 Other syntactic functions
Other syntactic functions comprise functions of dislocated and appositional expres-sions, as well as possessors.
Dislocated expressions receive the function gloss <:dt> for ’dislocated topic’,regardless of whether the pragmatic function of its referent is actually considered a’topic’ in the narrow sense. No distinction is made between left- and right-dislocatedexpressions. Dislocated expressions can have lexical or pronominal form. Examples:
(27) veraa gaqg 084
##
nart
ln
nuoturtlenp.d:dt a
di3sg
pro.d:a
=m=tam1
=lv
lanlanred:slap
v:pred
enart
ln
bini-gihand/arm-3sg
np:p-pro.d:poss
’And when Turtle had clapped his hands,...’
(28) veraa palaa 090
[...]
##
no1sg
pro.1:dt s
no1sg
pro.1:s
mefut
lv
sagsitv:pred
’irwurbehindrv
’[You two sit first,] and I, I will sit last one in the back (of the canoe).’
Where applicable, information on clause-internal function a dislocated phrasecorrelates with is added to the <dt> function gloss, for instance <:dt a> and<:dt s> in (27–28). I assume here that a left-dislocated expression may correlatewith an object function that receives zero expression within the clause, glossing it<:dt p>, and the clause-internal object as <0:p>. These are entirely analogous toinstances where the object function is expressed by a pronoun:
(29) veraa palaa 022
##
lavetfeastnp:dt p
vo-walnum-onern
anedem1.a
rn
dir3pl
pro.h:a
=em=tam1
=lv
gisholdv:pred 0:p
’This feast, they held (it).’
(30) veraa palaa 009
##
epers.art
ln
ruwatwo.peopleln
re-renensg-woman
np.h:dt p
anedem1.a
rn
duru3dl
pro.h:a
=m=tam1
=lv
dadov:pred
’o’withrv
duruo3dl
pro.h:p
’The two girls, they (i.e. their parents) looked after them.’
A further type of function distinguished for Vera’a is that of appositional ex-pressions. These are typically co-referent with the one they are juxtapposed toand that provide additional information on this referent. They receive the functiongloss <:appos>, as in (31). Treated in the same way here are so-called ’inclusoryconstructions’ where the justapposed expression is a non-singular pronoun that ispartially co-referent with the expression it is juxtapposed to, as in (32).
9 of 38
Multi-CAST veraa annotation notes
(31) veraa mvbw 020
[...]
##
di3sg
pro.h:a
=m=tam1
=lv
letransferv:pred
=n=art
=ln
ni’Ismallnp.h:p
’amanmanrn
’a...hesnc
’isimeresecond.bornnp.h:appos
anedem1.a
rn
[...]
’(When) she gave birth to the boy, the second born, [...]’
(32) veraa jjq 347
##
epers.art
ln
Qo’Q.np.h:s
dire3pl
pro.h:appos
manpfv
lv
’eqeldescendv:pred
’Qo’ and those with him had already hopped o↵.’
Appositional expressions are distinguished form coordinated phrases or othercomplex argument expressions, as outlined in Section 1.2 below.
The only NP-internal function noted in GRAID glossing of Vera’a texts is thatof possessors, glossed <:poss>. All other NP-internal expressions do not receive afunction gloss. Also, possessors are glossed only where they have a specific referent,excluding certain cases of compounding and modification.
1.2 Form of referential expressions
Vera’a has the following basic types of referential expressions:
common NPs
personal NPs
locative NPs
numeral phrases
pronominal NPs
free pronouns
bound person markers
adverbs
Table 1 summarises the glossing practices for each of these form types. Com-mon, personal, locative NPs and numeral phrases are all glossed <np>. Vera’aalso has ’pronominal NPs’ which are multi-word expressions consisting of a freepronoun plus further modifiers, and these are glossed as pronouns, <pro>. Forfree pronouns, a distinction between default and weak forms is occasionally (butat present not consistently) made. Under bound person markers, we mainly sum-marise possessive su�xes. An additional ’bound person marker’ is assumed to becontained in one TAM allomorph, namely ne which is a portmanteau morph forboth the TAM2 category as well as 3rd person singular of the S or A (i.e. subject)argument. The gloss <other> is used for all other types of referential or non-referential expressions. The following subsections provide a brief outline of eachform type.
10 of 38
Multi-CAST veraa annotation notes
form common
NP
personal
NP
locative
NP
numeral
phrase
pronom
inal
expressions
free
pronou
n
bnd.pers.
marker
adverbs,
dem
onstr.
gloss np np np np pro pro -pro otheralt. gloss pro wpro pro
Table 1: Form type–gloss correspondences of referential expressions inVera’a
1.2.1 Common NPs
Common NPs are introduced by the common article =n which may be omitted inclause initial position. Examples can be found in (29–32) above, and elsewhere inthis document. They are glossed as <np>.
1.2.2 Personal NPs
Personal NPs are introduced by the personal article e and most typically havepersonal names or one of a small class of other personal nouns as their heads, as inthe following examples in (33–34):
(33) veraa jjq 040
##
soprosp?
other
epers.art
ln
Qo’pers.name
np.h:s
netam2:3sg
lv-pro.h:s
mulogov:pred
’(And now) Qo’ was about to go home.’
(34) veraa palab 226
##
nart
ln
’ama’devilnp.d:a
manpfv
lv
kurdevourv:pred
saemph
other
epers.art
ln
ruwatwo.peopleln
ni-ni’i-duored-child-1dl.innp.h:p-pro.1:poss
edem3
rn
’The devil has already devoured our (dl) two children.’
In some instances, a personal pronoun takes a personal article and thus formsa personal NP. It will nonetheless be glossed <pro> rather than <np>:
(35) veraa anv 063
##
epers.art
ln
no1sg
pro.1:s
’ow’beforeother:pred
ow
’I am first (to jump).’
11 of 38
Multi-CAST veraa annotation notes
1.2.3 Locative NPs
Locative NPs di↵er from other NPs in that they are not introduced by an article.They are headed by a local noun, e.g. a place name, and commonly function as thecomplement of the specific locative preposition a. Locative NPs are simply glossed<np> in GRAID annotations. Two examples:
(36) veraa bsvh 006
##
sulfolknp.h:dt
di3sg
pro.h:s
gastat
lv
’ogstayv:pred
aloc.sp
adp
loseasidenp:l
’The people who lived down at the sea,...’
(37) veraa as1 003
##
babutother
duru3dl
pro.h:s
gastat
lv
’ogstayv:pred
aloc.sp
adp
LemerigLemerignp:l
’The two lived at Lemerig.’
1.2.4 Numeral phrases
Numeral phrases are optionally introduced by the numeral article ne and headedby a numeral word which consists of a fossilised numeral prefix and a numeral root.Where numeral phrases function as arguments on clause level, they are glossed asNPs, as in (38):
(38) veraa mvbw 111
##
nenum.art
ln
vo-walnum-onenp.h:s
netam2:3sg
lv-pro.h:s
vangov:pred
mahitherrv
’(Then) one (of them) came over.’
Where they function as modifiers in a NP, they are glossed as sub-constituents,both numeral article and numeral word receiving <rn>, as in example (20) above.
1.2.5 Pronominal expressions
As for person markers, four types are distinguished here for Vera’a. All of theseare glossed as <pro>, despite their structural di↵erences. Free pronouns functionas subjects, objects and complements of prepositions:
(39) veraa iswm 208
##
dir3pl
pro.h:a
=m=tam1
=lv
vuskillv:pred
die3sg
pro.h:p
’They killed him.’
12 of 38
Multi-CAST veraa annotation notes
singular dual trial/paucal plural
1st incl — (gi)du(o) (gi)do’ol (gi)de1st excl no ka(ma)du(o) ka(ma)m’ol ka(ma)m2nd nik(e) kumru(o) kimi’ol kimi3rd di(e) duru(o) dir’ol dir(e)
Table 2: Vera’a free personal pronouns
(40) veraa gabg 025
[...]
##
no1sg
pro.1:s
makimm
lv
’aramtellv:pred
entegwellrv
medat
adp
nike2sg
pro.2:g
’...and I will make it clear to you immediately.’
The paradigm of free pronouns is given in Table 2. Initial investigation ofsubject pronouns (Schnell 2010, 2011, 2012a, b) suggests that these pronouns aregrammaticalising into subject indexes, showing tendencies for a tighter morphologi-cal integration with subsequent TAMmarkers. This involves occasional reduction inform of 1st person non-singular pronouns through deletion of the first or second syl-lable, cf. Table 2; the reduced form is considered ’weak’ here and glossed—thoughnot entirely consistently at this stage—<wpro>. Thus, the following two glossingpractices can both be found in the current Vera’a Multi-CAST corpus:
(41) veraa gabg 044
##
du1dl.in
wpro.1:s
=k=tam2
=lv
geneatv:pred
qe’finishrv ##
gedu1dl.in
pro.1:s
makimmediatelylv
mulogov:pred
[pro.1:s]
’...and I will make it clear to you immediately.’
Where final vowel deletion occurs with pronouns, the forms are not counted as’weak’. Note that subject pronouns are essentially treated as free pronouns here.Their possibly intensifying closer integration with the VC is taken as a researchquestion to be tackled through analysis of GRAID-annotated texts rather than afact that feeds into the annotation.
Free pronouns may take further modifiers in Vera’a, and thus form a complexexpression which is termed pronominal NP here and glossed <pro>. Note thatmodifiers of such pronominal heads are glossed <rn>:
(42) veraa mvbw 127
##
kamam’ol1tl.ex
pro.1:s
birinwithrn
enart
rn
veve-maduomother-1dl.exrn np.h-pro.1:poss
’og-’ogored-stayv:pred
’We two, together with our (two) mother, will stay behind.’
13 of 38
Multi-CAST veraa annotation notes
singular dual trial/paucal plural
1st incl — -du(o) -do’ol -de1st excl -k -madu(o) -mam’ol -mam2nd -m -mru(o) -mi’ol -mi3rd -gi -ru(o) -r’ol -re
Table 3: Possessive (pronominal) su�xes in Vera’a
(43) veraa palaa 061
##
’eiinterj
other
kamadu1Ddl.ex
pro.1:s
anedem1.a
rn
=m=tam1
=lv
vangov:pred
mahitherrv
sirforadp
nik2sg
pro.2:g
anedem1.a
rn
’We two have come just for you.’
Possessive su�xes are glossed as bound person markers, <-pro>. Their paradigmis given in Table 3. The possessive su�x may attach directly to the possessed nounor to one of 8 possessive classifiers that either precede or follow the head noun.Possessive classifiers themselves are mostly glossed as sub-constituents, thus either<ln> or <rn>, yielding <ln-pro...> and <rn-pro...> respectively. Examples:
(44) veraa jjq 032
##
dir3pl
pro.h:a
=k=tam2
=lv
vilvilred:tiev:pred
=en=art
=ln
nakcanoenp:p
mu-reposs.gen-3pl
rn-pro.h:poss
’Then they tied up their canoes.’
(45) veraa hhak 071
##
000.h:a
letakev:pred
=n=art
=ln
ko-ruposs.ves-3dl
ln-pro.h:poss
=n=art
=ln
nakcanoenp:p
su-suored-paddlern
’[...] took their canoe [...]’
(46) veraa palab 226
##
nart
ln
’amadevilnp.d:a
manpfv
lv
kurdevourv:pred
saemph
other
eart
ln
ruwatwo.peopleln
ni-ni’i-duored-child-1dl.innp.h:p-pro.1:poss
edem3
rn
’The devil has already devoured our (dl) two children.’
Possessive classifiers may also function as the head of a common NP, and are inthese cases treated like any other directly possessed noun in this function. WhileVera’a does not have a full-fledged subject indexing system like many other Oceaniclanguages, the paradigm of the morpheme glossed ’TAM2’ here (labelled ’aorist’ byFrancois 2009) has a distinct forms for the 2nd and 3rd person singular, e and ne,respectively. This information is reflected in the GRAID annotation by treating neas a sub-constituent with a bound person marker:
14 of 38
Multi-CAST veraa annotation notes
(47) veraa palab 226
##
epers.art
ln
Qo’Q.np.h:s
netam2:3sg
lv-pro.h:s
vangov:pred
mahitherrv ##
000.h:a
netam2:3sg
lv-pro.h:a
revdragv:pred
surdownrv
enart
ln
[...]
As this ’bound person marker’ is the only possible bound form for S and A function,these can be quantified distinctly from other person markers in these functions bycounting <pro> and <-pro> separately.
1.2.6 Further types of expression: demonstratives, interrogativenouns, conjoined NPs, etc.
There are some further elements that potentially pose problems in terms of analysisand glossing of formal properties.
Oblique pronominal forms. Vera’a has two ’special pronominal’ forms thatare restricted to oblique argument functions, typically expressing a location or goal.Their form is rendered as <other> in GRAID, as in (48–49) below.
(48) veraa 1.tnu 012
##
mulgov:pred
mahitherrv
leloc
adp
=n=art
=ln
vunuvillagenp:g rn #rc
arel
other
dir3pl
pro.h:s
=s=sim
=lv
’ogstayv:pred
beneobl.pro
other:l
’...went to the village where they lived.’
(49) veraa jjq 117
##
epers.art
ln
Qo’Q.np.h:s
netam2:3sg
lv-pro.h:s
vangov:pred
mahitherrv
[...]
##
netam2:3sg
lv-pro.h:s
000.h:s
remclimbv:pred
rowseawardsrv
ranaobl.pro
other:g
’Qo’ came and dragged down his canoe, climbed onto it and...’
They are classified as locative adverbs and glossed <other> in terms of form asthey are not personal pronouns in the narrow sense.
Demonstratives. The demonstratives ne(’e) and gen can form a referential ex-pression and function as an argument. It always has deicic (“Deixis am Phantasma”in narratives) or discourse-deictic reference. They are always glossed <dem pro>,as in (50).
15 of 38
Multi-CAST veraa annotation notes
(50) veraa mvbw 079
##
nedem1
dem pro:s
=n=art
=ln
’erepl
ln
ni-ni’i-kred-small-1sgnp.h:pred-pro.1:poss
walonceother
’[Oh, people,] this is truly my kids (whose voices we are hearing).’
Almost all other demonstrative forms are derived from these two basic forms. Theyoccur either as sattelites in NPs, glossed simply <rn dem>, or as modifiers onclause level, thus glossed <other( dem)>. The addition of < dem> is not doneconsistently in these latter cases.
Interogative and indefinite expressions. Vera’a does not have interrog-ative or indefinite pronouns, and instead NPs headed by interrogative-indefinitenouns fulfil the respective functions. Examples:
(51) veraa mvbw 102
##neg
siifother
kumru2dl
pro.2:s
woandother
mi’irsleepv:pred
rosneg2
other ##
kumru2dl
pro.2:a
=m=tam1
=lv
ronhearv:pred
enart
lnsavawhatnp:p
[...]
’If you don’t sleep at night, what you hear...’
(52) veraa jjq 277
##
nike2sg
pro.2:s
epers.art
ln
sewhonp:pred
’Who are you?’
Complex NPs. In complex NPs the gloss for the entire phrase is aligned withthe first nominal head, and all other constituents to the right are glossed <rn>,with sporadic further specifications of form and animacy, for instance in (53). Onlypossessors are specified for their function, see (54).
(53) veraa iswm 179
##
ama-gifather-3sgnp.h:dt a
=n=art
=rn
veve-gimother-3sgrn np.h
duru3dl
pro.h:a
=k=tam2
=lv
siksearchv:pred
di3sg
pro.h:p
soquot
other
’His father and mother, they looked for him.’
16 of 38
Multi-CAST veraa annotation notes
(54) veraa palab 120
##
dinreachrv
mahitherrv
=n=art
=ln
nimehousenp:p
mo-nposs.house-csrn
epers.art
rn
’amaruofather-3dlrn np.h:poss
woandrn
=n=art
=rn
veve-ruomother-3dlrn np.h:poss
’[... ran] to the house of their father and mother.’
The preposition birin ’with’ can function as a co-ordinator on NP level, andthe modifier PP is in these cases treated as a sub-constituent, as in (55):
(55) veraa mvbw 127
##
kamabo’ol1tl.ex
pro.1:s
birinwithrn
enart
rn
veve-maduomother-1dl.exrn np.h-pro.1:poss
’og-’ogored-stayv:pred
’We two, together with our (two) mother, will stay behind.’
In cases where the co-ordination analysis is not clearly suggested by the syntacticdistribution—the pronoun and PP in (55) occupy a single pre-verbal slot—it istreated as an oblique PP on clause level expressing a comitative role, as in (??)below:
(56) veraa hhak 109
##
duru3dl
pro.h:a
=k=tam2
=lv
vangov:pred
gisholdrv
enart
ln
vusbownp:p
birincom
adp
enart
ln
’erepl
ln
wo’iqearrownp:obl
’Then they grabbed (their) bows together with the arrow [and went.]’
(57) veraa mvbw 096
[...]
##ds
000.1:a
momputv:pred
’del
rv
kumruo2dl
pro.2:p
birinwithadp
enart
ln
go-mruposs.eat-2dlln-pro.2:poss
=n=art
=ln
gengenfoodnp:obl
’[...] an (we) will take you together with your food.’
Pluralising particle ’ere. The pluralising particle ’ere typically occurs asa plurality-marking particle in NPs, as in (56) above, but is also used as a freeform with 2nd person non-singular reference in imperative constructions. Hereit occupies a slot following a possible 2nd person pronoun (cf. Section 2.2.1 onimperative constructions), and is glossed <other.2> in these instances too, as in(58).
(58) veraa hhak 144
##ds
babutother
000.2:s
’erepl
other.2:voc
supaddlevother:pred
kaluprv
kelbackrv
mahitherrv
’But you guys paddle back here and come ashore...’
Note that the function gloss <voc> for ’vocative’ is used here due to this functionof the particle in everyday communication outside imperative constructions. Noexample of this latter kind is attested in the current corpus.
17 of 38
Multi-CAST veraa annotation notes
1.3 Animacy and person of referential expressions
Referential expressions with human referents receive an animacy feature gloss <h>.Those with non-human referents that are anthropomorphised—typically capable ofspeech/thought, desires, planned actions—receive the feature gloss <d>. Thesenon-human referents are typically certain spirits and animals in customary fables.
Where inanimate objects—typically rocks, reefs, trees—turn into human-likesuper-natural beings (Vera’a: vu’ or ’ama’ ) in a narrative, the discourse referent inquestion is treated as inanimate as long as it does not appear as human-like, andas human-like where it appears as such. Example:
(59) veraa isam 023
[...]
##
socpl
other
=n=art
=ln
me’reefnp:dt s
[...] di3sg
pro:S
=m=tam1
=lv
vangov:pred
mahitherrv
rekselikeadp
=n=art
=ln’ansarapersonnp.h:obl
’...that the reef, it had become like a human being.’
(60) veraa isam 024
##
di3sg
pro.d:s
=m=tam1
=lv
rerecrane.neckv:pred
kalupwardsrv
’It craned upwards.’
As a rule, animacy features are assigned according to reference, not to classificationof nouns. Reference is here treated as including ’class/generic reference’, thus thegloss for ’ansara above includes <h>. Where the same noun is used to refer to aspirit, it is glossed <d>:
(61) veraa isam 036
[...]
##
’ansarapersonnp.d:s
leloc
rn
=n=art
=rn
me’reefrn np
netam2:3sg
lv-pro.d:s ds
tekspeakv:pred
medat
adp
die3sg
pro.h:g
soquot
other
’[...] then the person inside the reef said to him:...’
Numeral expressions or NPs headed by the place-holder noun ge ’thingy’ likewisereceive animacy glosses by type of reference.
1.4 Other elements
A number of other elements are only noted as such, and mostly glossed <other>.
1.4.1 Adverbs and clause-level demonstrative forms
Adverbs and demonstratives on clause level are simply glossed <other>. Fordemonstratives, additional tags are occasionally—but not entirely systematically—added, for instance <other dem1>.
Temporal adverbs functioning as frame-setting topic expressions are also simplyglossed<other>, no indication of this particular pragmatic function is noted in theirglossing.
18 of 38
Multi-CAST veraa annotation notes
1.4.2 Particles, conjunctions
Particles and conjunctions on clause level are also simply glossed <other> in mostinstances. This is also true for all instances of the emphatic particle sa whichprecedes or follows the phrase it marks; further research is required in order todetermine its nature.
Clause-connecting elements are all glossed <other>. This comprises underivedconjunctions like adversitive/theme-shifting ba, coordinative wo, the disjunction si,and subordinators like clause-combining e, relativizer a, or the complementizer soetc.
Also glossed <other> are words that appear to be de-verbal conjunctions, orig-inally involving complex sentence structures. Typical examples are da ’do’ andqe’ ’finish’ that occur clause-initially to mark causal or temporal relations betweensentences:
(62) veraa gaqg 024
##
dadoother
bewaternp:dt s
di3sg
pro:s
=m=tam1
=lv
kaluexitv:pred
mahitherrv
’[The rain became really heavy.] And so the water came out [and carriedaway the trunk I live in].’
In other cases, however, the structures involved seem to resemble complex sen-tences, the verb da ’do’ heading a VC, thus clearly forming a clause entering acomplex sentence structure. These are glossed as in (63).
(63) veraa iswm 128
##
so??other
=m=tam1
=lv
dadov:pred #cc
socpl
other
di3sg
pro.h:a
=m=tam1
=lv
remclimbv:pred
enart
ln
qansidenp:p
ve’rockrn np
ane’edem1.a
rn dem1
’And consequently he climbed up this rock wall.’
Thus, the first elements here are taken to form a matrix clause for a subsequentcomplement clause, licensed by the verb da. The matrix clause does not have aclearly referential subject in these instances, thus no zero argument is consideredfor glossing. Section 3.1 below provides more details on the glossing of complementclauses.
2 Clausal constructions with special features
In this section I discuss a number of clausal constructions that di↵er in some regardfrom the basic structures outlined above.
2.1 Negation and neutralisation of syntactic categories:<np:pred> vs. <v:pred>
Negation in Vera’a is expressed by a separate set of TAMP (tense-aspect-mood-polarity) markers. Crucially, a VC marked with a negated-set marker can contain
19 of 38
Multi-CAST veraa annotation notes
a nominal expression as its head that would in a�rmative clauses form a non-verbalpredicate, e.g. a NP. Compare the following two examples:
(64) veraa iswm 035
##neg
di3sg
pro.h:s
egen.neg1
lv
ka-kalured-exitv:pred
rosgen.neg2
rv
denabl
adp
enart
ln
nimehousenp:obl
’He didn’t leave the house.’
(65) veraa iswm 029
a.
##neg
nik2sg
pro.2:s
egen.neg1
lv
WowotW.v:pred
wuvaonlyrv
rosgen.neg2
rv
’You are not Wowot for no reason,...’
b.
##
babutother
nik2sg
pro.2:s
WowotW.np.h:pred
sirbecauseadp
enart
ln
savawhatnp:other
’...you are Wowot because of something, [namely...]’
Thus, while the predicate in (65b) is a NP headed by the personal name Wowot,the predicate of in (65a) is treated like a VC due to the presence of TAMP markingand receives the <v:pred> gloss like the negated VC in (64). Even pronouns can bethe head of negated VCs, and these are glossed likewise, though information aboutthe pronominal form is added as follows:
(66) veraa iswm 326
##ds.neg
onoother
di3sg
pro v:pred
rosgen.neg2
rv
’No. (That’s) not him.’
2.2 Finite and non-finite clause constructions:<v> vs. <vother>
Besides imperative clause constructions, there exist two potentially non-finite clauseconstructions. These are a) a type of head-tail construction, and b) a type ofpurposive clause construction, the so-called ’ga-construction’, which occurs as thecomplement of the purposive preposition ’alen.
2.2.1 Imperative constructions and <vother>
Orders, commands and similar speech acts may be expressed in Vera’a by an im-perative construction in which an otherwise verbal predicate does not take TAMPmarking. Lacking a finiteness feature, the predicate is glossed <vother:>. Overtsubject pronouns may nevertheless occur, and thus where no overt subject appearsa zero argument is glossed:
20 of 38
Multi-CAST veraa annotation notes
(67) veraa isam 025
##ds
nik2sg
pro.2:s
vangovother:pred
mahitherrv
leloc
adp
=n=art
=ln
kolo-kback-1sgnp:g-pro.1:poss
”Come onto my back, [and then we go].”
(68) veraa anv 060
##ds
000.2:s
damhandvother:pred
mulumlumslowrv
qe’ia.momentrv
’Keep swinging for now, [I’ll swing back, and then we go.]’
The same applies to non-singular subjects. Here, the pluralising particle ’ere occursadjacent to the verbal head. It is glossed <other.2> (cf. Section 1.2.6 above), andits function is rendered as vocative, <:voc>:
(69) veraa jjq 439
[...]
##ds
kimi2pl
pro.2:s
’erepl
other.2:voc
vrigrushvother
qeldescendrv
waloncerv
rowseawardsrv
aloc.sp
adp
loseasidenp:g
”[...] you guys run down to the sea [and look for a tree...]”
(70) veraa mvbw 098
##ds
000.2:a
’erepl
other.2:dt
geneatvother:pred
saemph
other
=n=art
=ln
gengenfoodnp:p
”You guys eat this food that...”
2.2.2 ga-construction: <vother>
A clause-like construction functions as the complement of the prepositions ’alen or’amen and expresses a state-of-a↵airs that is the purpose of the action expressedin the matrix clause. The predicate in these constructions takes the stative markerga and does not allow for overt realisation of the subject, thus no zero subject isnoted and the head of the predicate receives the <vother:pred> gloss.
(71) veraa jsu 121
a.
##
di3sg
pro.h:s
soprosp
other
netam2:3sg
lv-pro.h:s
kalenterv:pred
ba’intorv
leloc
adp
=n=art
=ln
qoroholenp:g
liecavern
’alenass
adp
’...he wanted to go into the opening of the cave...’
21 of 38
Multi-CAST veraa annotation notes
b.
##cc:other
gastat
lv
letakevother:pred
=n=art
=ln
ve’stonenp:p
tiktiksmallrn
’alenpurp
adp #cc:other
gastat
lvvangovother:pred
mahitherrv
’...in order to collect small stones to bring them and [smash open hiscanarium nuts.]’
2.2.3 Head-tail construction and zero TAMP marking
Vera’a seems to have a type of clause construction that resembles what has cometo be called ’head-tail linkage’ or ’head-tail construction’. A candidate for such aconstruction is the following:
(72) veraa jjq 074
a.
##
womarawrawSpidernp.d:s
netam2:3sg
lv-pro.d:s
’ogstayv:pred
’idelrv
’And so Spider stayed behind.’
b.
##
000.d:s
’ogstayv:pred
’idelrv
’Stayed behind,...’
c.
##
leloc
adp
=n=art
=ln
qonnightnp:other
anedem1
rn
womarawrawSpidernp.d:a
netam2:3sg
lv-pro.d:a
durhollowv:pred
000:p
’...and at night Spider began to hollow (it, i.e. the canoe).’
As in this example, the clause essentially repeats the state-of-a↵airs expressed inthe preceding one, leaving the same subject zero. No TAMP marking occurs. How-ever, a would-be ’tail-head’ construction is not straightforwardly distinguishablefrom other constructions with similar properties. For one thing, similar discourse-structuring functions are carried out by canonical finite constructions, as in (73b):
(73) veraa palaa 041
a.
##
dir3pl
pro.h:s
=ek=tam2
=lv
lak-lakared-dancev:pred
’Then they danced.’
b.
##
dir3pl
pro.h:s
=em=tam1
=lv
lak-lakared-dancev:pred
edem3
other
[...]
’[And as] [T]hey danced,...’
22 of 38
Multi-CAST veraa annotation notes
c.
##
duru3dl
pro.h:a
=k=tam2
=lv
’enseev:pred
mahitherrv
=n=art
=ln
lumgavyoung.mannp.h:p
nenum.art
rnvowalnum-onern
’...[- the two were hiding in the bush -] they (the two girls) spotted ayoung man.”
On the other hand, constructions lacking overt TAM marking also occur inother contexts, as in (74–76), which are clearly not tail-head linkages, but theexact finiteness status of which appears to be yet unclear:
(74) veraa jjq 241
##
dir3pl
pro.h:a
’enseev:pred
vag-’olord-threenp:p
na-gi??-3sgrn
’They saw the third one.’
(75) veraa mvbw 052
a.
##
000.h:s
’eqeldescendv:pred
suwdownrv
mahitherrv
’[...He climbed the tree, picked a few (fruits),] (then) came down,...’
b.
##
000.h:a
bulstonev:pred
munmunoshatterrv
’...smashed them open... ’
c.
##
000.h:a
letransferv:pred
000:p
medat
adp
di3sg
pro.h:g
’...smashed them open and gave (some) to him.’
(76) veraa mvbw 127
##
kamabo’ol1tl.ex
pro.1:s
birinwithrn
enart
ln
veve-maduomother-1dl.exrn np.h-pro.1:poss
’og-’ogored-stayv:pred
’We two, together with our (two) mother, will stay behind.’
Thus, in (74) we seem to be dealing with a ’normal’ non-embedded independentclause. Yet, no TAMP appears between subject pronoun and verb. The chainedclauses in (75) seem to resemble essentially the same type of structure, with thesubject being left zero. In (76), the subject is 1st person trial, and it may bepossible that we are dealing with a zero allomorph of the TAM2 morpheme. Theexact nature of these ”zero TAMP markers” is yet unclear, and therefore, it seems,head-tail linkages are not clearly identifiable.
The practical conclusion from these combined analytical uncertainties is thatwe treat constructions without an overt subject as in (73) and (75) both as zero sub-jects, not distinguishing between would-be head-tail linkages and clause chaining.
23 of 38
Multi-CAST veraa annotation notes
Again, systematic analyses of GRAID-annotated Vera’a corpora should eventuallyinform our analytic decision, rather than a premature analysis inform our glossingpractice.
3 Complex sentences, direct speech, clauserepetitions and complex predicates
This section deals with the treatment of combinations of clauses into larger units,i.e. complex sentences. I will discuss the glossing of complement clauses (Section3.1), adverbial clauses (3.2), relative clauses (3.3), embedded direct speech (3.4), thehandling of clause repetitions (3.5), and clause-chaining constructions (3.6) whichare distinct from complex predicates involving serial verbs.
3.1 Complement clauses
Complement clauses can be syndetic or asyndetic, the latter case obviously beingthe more problematic one.
3.1.1 Syndetic complement clauses
Syndetic complement clauses are clearly recognisable by the complementiser so thatintroduces them. They are glossed as in (77–78).
(77) veraa hb08
[...]
##
nik2sg
pro.2:s
gastat
lv
moroswantv:pred #cc
socpl
other
nik2sg
pro.2:s
etam2:2sg
lv-pro.2:s
galalaknowv:pred
’[...] (if) you want to know.’
(78) veraa jjq.e 162
##
nik2sg
pro.2:s
etam2
lv-pro.2:s
’enseev:pred #cc
socpl
other
=n=art
=ln
nawsaltwaternp:dt
di3sg
pro:s
=m=tam1
=lvmelebreakv:pred
vag-’olmult-threeother
’When you see that the waves broke three times, [...]’
The function of a complement clause is taken here as a unique function, as thestructures involved do not resemble those of ’regular’ NPs with P function in thesense of Andrews (2007:138↵.). Thus, no function gloss is added to the <#cc>gloss. Consequently, the other argument in the matrix clause bears S rather thanA function. Note that syndetic complement clauses can never have the function ofan S or A argument.
The complementizer so is related to the quotative so ’say’ and glossed <other>.In some instances complement clause constructions as discussed here can be hardto distinguish from direct speech, cf. Section 3.4 for details. Syndetic complement
24 of 38
Multi-CAST veraa annotation notes
clauses show typical clausal properties: their predicate is TAM-marked, all argu-ments can be expressed, and all non-core positions are available to the left of thecore, for instance the left-dislocated position, as witnessed by (78).
3.1.2 Asyndetic complement clauses
Asyndetic complement clauses lack a complementizer, but are fully verbal andunreduced. They contain a TAM-marked VC functioning as predicate, and thesubject may be realised overtly, but need not be. Examples:
(79) veraa mvb 087
##ds
no1sg
pro.1:s
gastat
lv
moroswantv:pred #ds cc
no1sg
pro.1:s
=k=tam2
=lv
kakatalkv:pred
birinwithadp
nike2sg
pro.2:obl
’I want to talk to you.’
(80) veraa bsvh 034
##
leloc
adp
=n=art
=ln
masogitimenp:other
di3sg
pro.h:s
gastat
lv
moroswantv:pred #cc
000.h:S
netam2:3sg
lv-pro.h:s
vrigorushv:pred
’When he wanted to run away...’
Complement clause constructions with moros ’want’ as matrix predicate are tobe distinguished from constructions where moros ’want’ occurs in a series with afollowing verb:
(81) veraa paww 072
##ds
nik2sg
pro.2:a
gastat
lv
moroswantv:pred
kurdevourrv
kamaduo1dl.ex
pro.1:p
’You want to eat us.’
This construction is analysed as a serial verb construction (SVC) here, rather than acomplex sentence where the matrix predicate would take a clausal complement, as inthe English translation. There is no evidence for subordination in this constructionin Vera’a, and the structure resembles exactly that of a SVC. Treatment of serialverb constructions is discussed in Section 3.6 below.
Another case of fuzzy boundaries between complement clause construction andother structures is represented by the set of examples in (82–83).
(82) veraa isv 084
##
di3sg
pro.h:s
=m=tam1
=lv
’enseev:pred #cc
’ow’owbeforeother
enart
ln
loso-gitesticles-3sgnp:s-pro.h:poss
gastat
lv
sagsitv:pred
leloc
adp
=n=art
=ln
me’emedoornp:l
’Then he saw that before his [a giant’s] testicles had been sitting in thedoor(way).’
25 of 38
Multi-CAST veraa annotation notes
(83) veraa mvb 081
##
di3sg
pro.h:s
netam2:3sg
lv-pro.h:s
’enseev:pred #cc
enart
ln
’nsarpersonnp.h:s
nenum.art
rn
vo-walnum-onern
netam2:3sg
lv-pro.h:s
vangov:pred
mahitherrv
’[...] then he saw a man coming up (to him).’
While in (82), we find a complement clause with a clear left boundary marked bythe left-most adverb, the construction in (83) could be analysed as subject-to-objectraising. Again, nothing in Vera’a grammar forces such an analysis, and thus thecomplement clause analysis seems to be preferable. A reversed type of structure isfound in the following set of examples:
(84) veraa palab 024
a.
##
di3sg
pro.d:a
netam2:3sg
lv-pro.d:a
dinflickv:pred
likmorerv
enart
ln
liecavenp:p
nenum.art
rn
vowalnum-onern
ane’edem1.a
rn dem1
’He flicked yet another one of those caves,...’
b.
##
=n=art
=ln
liecavenp:s
netam2:3sg
lv-pro:s
wakopenv:pred
’...and the cave opened.’
(85) veraa palab 055
##
di3sg
pro.h:a
netam2:3sg
lv-pro.h:a
dinflickv:pred
enart
ln
liecavenp:p
anedem1.a
rn ##
000:s
netam2:3sg
lv-pro:s
wakopenv:pred
’You want to eat us.’
As the first predicate in (85) does license a NP complement but not a clausalcomplement, the NP must be regarded as bearing P function. The following clausehas a zero S argument, as is clear from comparison with (84).
The practical conclusion thus is that we gloss complement clauses in cases wherethis type of complementation is licensed by the matrix predicate in question, andgloss clause chaining in other cases.
3.2 Adverbial clauses
The distinction between adverbial clauses and main clauses is not consistently rep-resented in GRAID annotations of Vera’a texts. Vera’a is strongly paratactic andclues pointing towards complex sentence structure are often restricted to prosodicfeatures. Adverbial clauses are glossed as such only where the occurrence of certainsubordinators at the beginning of a clause makes this clear, as in (86). However,even in these latter cases the annotation of adverbial clauses has not been doneconsistently so that they are often simply treated like independent clauses.
26 of 38
Multi-CAST veraa annotation notes
(86) veraa mvbw 024
##ac
’ow’beforeother
owabl
adp
denart
ln
enmother-3dlnp.h:s-pro.h:poss
veve-rutam2:3sg
lv-pro.h:s
nedeadv:pred ##
ma’ di3sg
pro.h:s=m=tam1
=lv
rusosickv:pred
’Before their mother died, she fell sick.’
Thus, in (86) the dependency of the two clauses is overtly marked, the first clauseis glossed as an adverbial clause.
3.3 Relative clauses
Relative clauses are usually considered for GRAID annotation in Vera’a. Theymay be syndetic or asyndetic. The relativizing strategy in both types of relativeclauses is gapping for core arguments, and the ’gap’ is considered a zero argumentin GRAID, <0(.x):y>.
3.3.1 Syndetic relative clauses
Two examples of a syndetic relative clauses are given in (87–88) together withGRAID glossing. The relativizer a is glossed <other>. Where a relative clauseappears centre-embedded, its end is marked by <%>, as in (88).
(87) veraa gabg 030
##
nart
ln
belelbasketnp:dt s rn #rc
arel
other
reksolikeother
=n=art
=ln
gosuworatnp.d:a
gastat
lv
gisholdv:pred
000:p %
di3sg
pro:s
=m=tam1
=lv
wurfullv:pred
nenenentirelyrv
’The basket that the rat took with him was full.’
(88) veraa anv 023
##
nart
ln
maru-nuncle-csnp.h:s
epers.art
rn
renewomanrn np.h:poss rn #rc
arel
other
000.h:s
=m=real
=lvma’deadv:pred
nedem
other
[...]
%
’So the uncle of that woman that had died said:...’
3.3.2 Asyndetic relative clauses
In asyndetic relative clauses, the relativised function is often the object, in whichcase the relativizing strategy is gapping. The respective ’gapped’ function is glossed<0>, as in (89).
27 of 38
Multi-CAST veraa annotation notes
(89) veraa isam 033
[...]
##
leloc
adp
=n=art
=ln
vunuoislandnp:g
nenum.art
rn
vo-walnum-onern rn #rc
dir3pl
pro.h:a
gastat
lv
ulcallv:pred
000:p
soquot
other
HiwHiwnp:other
’...at one island which is called Hiw...’
However, asyndetic relative clauses with relativised zero subjects do seem toexist. These ’reduced’ relative clauses usually contain a ga-marked VC as theirpredicate which is in turn headed by a stative verb expressing a property, as in(90). Formation of this type of relative clause in Vera’a—as in many other Oceaniclanguages—is a means of modification by property words that are formally verbsand cannot usually function as modifier just on their own. Their GRAID glossingtherefore does not reflect the relativisation structure but merely treats the stativemarker ga and the following verb as <rn>-glossed NP constituents, even in caseslike this where the final adverb va’a ’still’ provides some evidence of the clausalstatus of this construction.
(90) veraa iswm 360
##
duru3dl
pro.h:s
wunvaproabalyother
=m=tam1
=lv
ma’deadv:pred
’ekeneloc.dem1
other:l
leloc
adp
=n=art
=ln
’eyearnp:other
gastat
rnmewmanyrn
va’astillrn
’Probably they died there after many years,...’ (lit. ’...in years that are stillmany.’)
Like complement clauses discussed above, reduced relative clauses potentiallyinvolve structural ambiguity as well. Two elicited examples illustrate this:
(91) (elicited)
a.
##
nik2sg
pro.2:a
etam2:2sg
lv-pro.2:a
’enseev:pred
enart
ln
me’emedoornp:p
gastat
rn
wakopenrn
’You see an open door, [go in this door].’
b.
##
nik2sg
pro.2:s
etam2:2sg
lv-pro.2:s
’enseev:pred #cc
enart
ln
me’emedoornp:s
gastat
lv
wakopenv:pred
’You see (that) the door is open, [you may go in].’
Glossing decisions are made according to the context of the surrounding discourse,which involves among other things the specificity of the NP’s referent.
Similar structures of relative clauses with ’gapped’ subjects are found with thesimultaneous marker =s.
28 of 38
Multi-CAST veraa annotation notes
3.3.3 Function of relative clauses
Where relative clauses function as modifiers in NPs, they are glossed <rn #rc>,as they represent a constituent within the NP (see examples above). Vera’a alsohas headless relative clauses, and their respective function and animacy features oftheir referent are annotated in GRAID. Thus, in (92), a relative clause functionsas a P argument.
(92) veraa jjq 174
##
di3sg
pro.d:a
netam2:3sg
lv-pro.d:a
ronfeelv:pred #rc.h:p
000.h:s
ssim
lv
ra-rarared-cryv:pred
’And he heard someone crying.’
3.4 Direct speech
The occurrence of direct speech (or thought, content) is usually marked in Vera’aby means of a quotative marker so ’say’. It is analysed as a particle where it followson a verb of speech or thought etc and receives the gloss <other>, as in (93).
(93) veraa anv 047
##
nart
ln
maru-ruuncle-3dlnp.h:s ds-pro.h:poss
netam2:3sg
lv-pro.h:s ds
teksayv:pred
medat
adp
duru3dl
pro.h:g
soquot
other
##ds
eiinterj
other
[...]
’Their uncle said to them: ’Hey,...”
As in this example, direct speech often comprises more than a single subordinateclause, and hence all clauses constituting direct speech are treated as independentclauses and receive the <##ds>. Moreover, clauses containing complements thatresemble direct speech are not analysed as transitive constructions here (cf. fn 6),thus the subject of such a clause, expressing the ’utterer’, is glossed as <:s ds>,where the < ds> tag signals that the clause has a direct speech complement whichmay in other languages be analysed as a transitive object expression.
The quotative marker may also function as a predicate, as in (94a), and is thenglossed <other:pred>. In (61b), the quotative marker functions as the predicateand occurs without a subject, which is a common way of signalling a shift of speaker-addressee roles in reported conversation.
(94) veraa as1 040
a.
##
epers.art
ln
DolD.np.h:s
soquot
other:pred
’After a while Dol said:...’
b.
##ds
onoother
no1sg
pro.1:s
manpfv
lv
qe’finishv:pred
’...’Oh, I’m done.’
29 of 38
Multi-CAST veraa annotation notes
c.
##ds
nart
ln
giekavanp:a
manpfv
lv
manstimulatev:pred
no1sg
pro.1:p
”I’m already drunk on the kava.”
(95) veraa as1 011
a.
##
000.h:s
soquot
other:pred
’(He) said:...’
b.
##ds
babutother
ruwatwo.peoplenp.h:voc
medat
rn
=n=art
=rn
’isiruosame.sex.sibl-3dlrn ##ds
kumru2dl
pro.2:s=k=tam2
=lv
vanvanred:gov:pred
aloc.sp
adp
viewhereother:g
’...’Hey, you two brothers, where are you going?”
c.
##
000.h:s
soquot
other:pred
’(He) said:...’
d.
##ds
kamadu1dl.ex
pro.1:a
=k=tam2
=lv
siksikred:searchv:pred
no-maduposs.dom-1dl.ex
ln-pro.1:poss
eart
lnrawhermaphrodite.pignp:p
’...’We are looking for an intra-sex pig for us.”
3.5 Predicate or clause repetition
It is quite common in Vera’a narratives to stress the duration of an action orprocess, or the intensity of a property, by repeating the predicate. Though thistype of repetition is of course part of the way of speaking in the language, and thusby no means ’wrong’ or ’inferior’, it is nevertheless not considered for the analysis ofargument realisation, following the conventions of the GRAID manual. Repeatedclause constructions are thus glossed <#nc>, as in (96).
(96) veraa anv 081
##
epers.art
ln
ragapeoplenp.h:a
anedem1.a
rn
=k=tam2
=lv
siksearchv:pred
duruo3dl
pro.h:p #nc
siksearchx
duruo3dl
x
#nc
siksearchx
duruo3dl
x
[...]
’Then everybody was looking for them, looking for them, looking for them,looking for them, on and on...’
30 of 38
Multi-CAST veraa annotation notes
3.6 Complex predicates versus clause chaining
As mentioned above, a VC in Vera’a may consist of more than one word, andfurther verbs (serial verb constructions), but also adverbs, or directional particles,may occur in the VC in addition to the head verb. Thus, we deal with only onesingle predicate in these cases, and thus only the head verb receives the <v:pred>gloss, other constituents being treated as additional sub-constituents, glossed <rv>,as in (97–98). Note that in (97), a object pronoun occurs evidently inside the VCand is thus glossed <rv pro.h:p>.
(97) veraa iswm 171
##
dir3pl
pro.h:a
=ek=tam2
=lv
qerepushv:pred
ba’aintorv
di3sg
rv pro.h:p
sarbushwardsrv
leloc
adp
=n=art
=lnmogiposs.house-3sgln-pro.h:poss
=n=art
=ln
nimehousenp:g
’Then they pushed her into her house.’
(98) veraa paww 072
##ds
nik2sg
pro.2:a
gastat
lv
moroswantv:pred
kurdevourrv
kamaduo1dl.ex
pro.1:p
”You want to eat us.”
Complex predicates clearly di↵er from chained clauses, even where this is not soobvious on first sight, as in (99) already discussed above.
(99) veraa mvbw 052
a.
##
000.h:s
’eqeldescendv:pred
suwdownrv
mahitherrv
’[...He climbed the tree, picked a few (fruits),] (then) came down,...’
b.
##
000.h:a
bulstonev:pred
munmunoshatterrv
’...smashed them open... ’
c.
##
000.h:a
letransferv:pred
000:p
medat
adp
di3sg
pro.h:g
’...smashed them open and gave (some) to him.’
The analysis as a clausal chain rather than a complex predicate follows from com-binatory rules applying to di↵erent categories of words, for instance a directionalmarker ma always occurs at the right margin of a VC.
31 of 38
Multi-CAST veraa annotation notes
A Appendix
A.1 Notes on morphological glossing
Morphological glossing of Vera’a MultiCAST texts follows the Lepzig GloggingRules (LGR, Bickel et al. 20081). Below is a list of Vera’a-specific and standardLRG glosses used in glossing of Vera’a texts.
Table 4: Morphological glosses for Vera’a
exponent category description comment
1
2
3
me abil1 ability cf. A.1
mas abil2 cf. A.1
den abl ablative prep ablative preposition
=(e)n art common article inherently enclitic, introduces
common NP
’amen,
’alen
ass associative associative prepositions
’a ass.sp specific associative associative prepositions
e cc clause-combining
particle
may be same as DEM3 and/or disc
birin com comitative comitative preposition
cor correction construction restart after false start
-n cs constuct su�x possessive su�x accommodating
personal NP possessors
me dat dative dative preposition
’i del delimitative
aktionsart
post-verbal delimitative marker, not
part of TAMP system, cf. A.1
ne dem1 basic dem 1 cf. A.4 for explanations and com-
ments on demonstratives
ane dem1.a pref. basic dem 1 cf. A.4
gen dem2 basic dem 2 cf. A.4
agen dem1.a pref. basic dem 1 cf. A.4
e dem3 cf. A.4
anei dem4.a pref. dem. 4 cf. A.4
nei dem4 cf. A.4
-ge dis dissociative dissociative possessive su�x,
possessor unspecified
e disc discourse particle discourse-structuring function;
probably new paragraph, theme
dl dual
sa emph emphatic emphatic particle, can have focus-
marking e↵ect
ex exclusive
me fut future TAM
marker
Predicates refer to events in posterior
to CT; cf. A.1
e gen.neg1 general negation 1 cf. A.1
ros gen.neg2 general negation 2 cf. A.1
1available online at http://www.eva.mpg.de/lingua/resources/glossing-rules.php
32 of 38
Multi-CAST veraa annotation notes
exponent category description comment
hes hesitation hesitation phenomenon (some parti-
cle, pause, dots, etc)
mak imm immediacy predicate expresses soa immediately
anterior or posterior to CT
in inclusive
mas inabil2 inability cf. A.1
interj interjection used to cover various types of
interjection
le loc locative prep locative preposition
(’e)kene loc.dem1 locative adv 1 cf. A.4 for explanations and
comments on demonstratives
(’e)kenen loc.dem2 locative adv 2 cf. A.4
kenei loc.dem4 locative adv 4 cf. A.4
a loc.sp specific loc. prep specific locative preposition; marks
specific locative expressions
sene man.dem1 manner adv 1 cf. A.4 for explanations and
comments on demonstratives
asene man.dem1.a pref. manner adv 1 cf. A.4
esene man.dem1.e pref. manner adv 1 cf. A.4
segen man.dem2 manner adv 2 cf. A.4
vag- mult multiplicative derives iterative adverbs
nmlz nominalisation usually reduplication in nouns;
occasionally simply ’red-’
nsg non-singular reduplication in nouns; occasionally
simply ’red-’
vo num numeral prefix fossilised prefix for cardinal numerals
ne num.art numeral article introduces numeral phrases (NumPs)
’en ny.neg2 ”not yet” negation occurs in right periphery of VC
na- ord ordinal quantifier seems to resemble possessive classifier
’e part partitive article restricted to particular types of poss.
constructions
e pers.art personal article not specified for sexus
ero pers.art.f personal article
female
specialised form for female referents
man pfv perfective predicates refer to events anterior to
CT; cf. A.1
pl plural category of person marker, cooccurs
with person, clusivity gloss
’ere pl plurailizer free particle, used on its own
bo- poss.bed poss. clf ’bedding’ classifier for ’bedding possession’
(bed, pillow, sheets, etc.)
no- poss.dom poss. clf ’domestic’ classifier for ’domestic possession’
(animals, crops, personal etc.)
mo- poss.drink poss. clf ’drink’ classifier for ’drink possession’ (water,
kava, etc.; juicy fruit)
go- poss.eat poss. clf ’eat’ classifier for ’eating possession’ (food,
diseases)
mu- poss.gen poss. clf ’general’ classifier for unspecified possessive
relationship
33 of 38
Multi-CAST veraa annotation notes
exponent category description comment
mo- poss.house poss. clf ’house’ classifier for ’housing possession’
(house, door(way), window etc.)
bolo- poss.val poss. clf ’valuable’ classifier for ’possession of
customarily valuable items’
ko- poss.ves poss. clf ’vessel’ classifier for ’vessel possession’
(canoe, boat, truck, plane)
mas proh1 prohibitive cf. A.1
ros proh2 cf. A.1
so prosp prospective marker might overlap with complementation,
quotative, etc.
ver- rec reciprocal reciprocal prefix
red reduplication di↵erent functions, non-singular,
imperfective, distributive
a rel relativizer
mal rem.pst remote past cf. A.1
sg
=s sim simultaneous predicate expresses soa simultaneous
with other soa
ga stat stative TAM
marker
predicates express habitual, generic
soas, properties
=m tam1 TAM 1 cf. A.1
ne, =k, e tam2 TAM 2 cf. A.1
va’ane temp.dem1 time adv 1 cf. A.4 for explanations and
comments on demonstratives
va’agen temp.dem2 time adv 2 cf. A.4
ge thing placeholder word has either context-retrievable specific
or non-specific reference
tl trial probably genuinely trial rather than
paucal
A.2 Tense, aspect, mood, polarity marking
Vera’a has a fairly complex system of tense, aspect, mood and polarity marking.Table 5 provides an overview, arranging the total of 13 morphemes in two sets ofmarkers, one with a�rmative and one with negative polarity.
The TAM2 morpheme is the only one showing complex allomorphy, and theallomorphic variation is conditioned here by the person and number of the sub-ject. See A.3 on person markers for the forms involved. Note that the functionalaspects of TAMP marking in Vera’a requires more research. It seems, however,that for those categories with an informative label, the core set of functions canbe described as such. Particularly problematic in this regard are the two mostfrequently occurring TAMP markers labelled TAM1 and TAM2 here. In everydaycommunication, TAM1 marked predicates seem to designate realis, known states-of-a↵airs situated in the past, or having come into being gradually in the present.TAM2-marked predicates on the other hand designate states-of-a↵airs that are newto the addressee, ongoing or situated in the future. The use of these markers in nar-ratives is an even less understood issue, but it seems likely that it can be accountedfor in terms of shifting of deictic centre/CT within a narrative. More research isexpected to elucidate these issues. Note that in work by Alexandre Francois, the
34 of 38
Multi-CAST veraa annotation notes
a�rmative negative
=m tam1 (e) ... ros general=k, e, ne tam2 negativemak immediacyga stativeme future=s simultaneous
man perfective ... ’en ’not yet’mal remote past
mas ... ros prohibitive
me ... mas ability mas ... mas disability
Table 5: Vera’a free personal pronouns
Vera’a category TAM2 is analysed as ’aorist’, see for instance Francois (2009) on thedevelopment of aorist markers from person prefixes in North Vanuatu languages.
A.3 Person markers
Person markers in Vera’a are glossed only for the three categories of person, numberand clusivity. Vera’a does seem to possess a genuine trial, the respective formprobably being restricted to reference of three people. Morphological glossing doesnot reflect the syntactic function of person forms. Person su�xes always expresspossessors, and these possessive su�xes are distinguished from free person forms bythe presence of a hyphen. Free forms can occur in a variety of syntactic functions,noted by GRAID glossing. Tables 6 and 7 provide the paradigms of free personmarkers and possessive su�xes.
singular dual trial/paucal plural
1st incl — (gi)du(o) (gi)do’ol (gi)de1st excl no ka(ma)du(o) ka(ma)m’ol ka(ma)m2nd nik(e) kumru(o) kimi’ol kimi3rd di(e) duru(o) dir’ol dir(e)
Table 6: Vera’a free person markers
singular dual trial/paucal plural
1st incl — -du(o) -do’ol -de1st excl -k -madu(o) -mam’ol -mam2nd -m -mru(o) -mi’ol -mi3rd -gi -ru(o) -r’ol -re
Table 7: Vera’a possessive su�xes
All dual forms, as well as some forms within the free paradigm show variation
35 of 38
Multi-CAST veraa annotation notes
singular plural& dual
trial
1st =k =k =k2nd e =k =k3rd ne =k =k
Table 8: Vera’a TAM2 person marker
in the presence vs absence of a final vowel. This variation is conditioned solely bythe prosodic environment of the forms. The omission of the initial syllable in non-singular inclusive free forms as well as that of the medial syllable in non-singularexclusive free forms, on the other hand, is restricted to a partiular syntatic slot,namely the pre-VC subject position.
As indicated above, the TAM2 marker shows complex allomorphy conditionedby the number and person of the subject. The marker thus constitutes a personmarker, be it with a quite ’deficient’ paradigm / high degree of syncretism, makingonly rudimentary person and number distinctions. The paradigm is given in Table 8.
According to Francois (2009), these forms historically derive from bound subjectindexes that were prefixed to the verb in the respective proto-language of Vera’aand other closely related languages of the region. All the non-singular forms k arerelated to the 1st person form which would have spread throughout the paradigm.In the trial, it seems, we find an alternation between k and zero, also noted byFrancois (2009). Note, however, that overt zero TAMP markers may have di↵erentorigins, as discussed in Section 2.2.3 above.
A.4 Demonstrative forms
Vera’a has a large set of demonstrative forms which are systematically related toa 3-way system of basic demonstratives—DEM1–3—in the sense of Himmelmann(1997). Related to these are di↵erent types of adverbs. Apparently also formallyrelated to all these forms is a set of interrogative forms. Tables 9 and 10 summarisesthese forms.
Possibly also related to these forms is a fourth set of demonstrative forms,glossed DEM4, see Table 11. Their exact status is, however, not entirely clear atpresent. At least some of these forms may in fact be free variants of DEM1 forms,while others clearly seem to resemble temporal adverbs, for example enei ’now’.Di↵erent sets of demonstrative forms show prefixing by two types of element. The
Table 9: Vera’a demonstratives
dem1 dem2
plain a-prefix e-prefix plain a-prefix e-prefix
basic set ne(’le) ane(’le) gen(e) agen(e)
manner adv sene asene esene segen(e) asegen(e) esegen
time adv 1
time adv 2 va’ane va’ane va’ane va’agen va’agen va’agen
locative adv (’e)kene akene (’e)kegen(e) (’e)kegen(e)
36 of 38
Multi-CAST veraa annotation notes
dem3 interrogative
plain a-prefix e-prefix
basic set e vie avie
manner adv sivie asivie esivie
locative adv kivie kivie kivie
Table 10: Vera’a demonstratives and related forms
dem4 / temporal adverbs
plain a-prefix e-prefix
basic set nei anei enei
manner adv esenei
locative adv (’e)kenei (’e)kenei (’e)kenei
Table 11: Possible additional set of Vera’a demonstratives
a-prefix is probably the specific locative preposition a accreted to the respectiveplain forms of the basic demonstratives or manner adverbs. The a-prefixed formsof the basic set seem to be preferred with adnominal uses, though occasionally theplain forms are found in this function too. Both the plain and the a-prefixed formsof the basic set occur on clause level, namely clause-finally, with di↵erent functions:the plain forms seem to have reinforcing-assertative function (”You do know thatthis is true!”), while the latter has the function to mark the proposition of theclause as a common ground package to which further information will be amendedin following propositions.
References
Andrews, A. 2007. The major functions of the noun phrase. In: Timothy Shopen(ed.), Language typology and syntactic description. Cambridge: Cambridge Uni-versity Press. 132–223.
Bickel, B., Comrie, B. & Haspelmath, M. Revision 2008. The Leipzig GlossingRules: Conventions for interlinear morpheme-by-morpheme glosses.http://www.eva.mpg.de/lingua/resources/glossing-rules.php
Francois, A. 2009. Verbal aspect and personal pronouns: the history of aorist mark-ers in north Vanuatu. In: Alexander Adelaar & Andrew Pawley (eds.), Aus-tronesian historical linguistics and culture history. A festschrift for Robert Blust.Canberra: Pacific Linguistics. 179–195.
Haig, G. & Schnell, S. 2014. Annotations using GRAID (Grammatical Relationsand Animacy in Discourse). ms.
Schnell, S. 2011. Pronominal reference in Vera’a discourse. Paper presented atthe International Workshop on the Languages of Vanuatu. Australian NationalUniversity, Oct 20–23, 2011.
Schnell, S. 2012a. Explaining formal variation in subjects and objects in Vera’a.Paper presented at the New Ways of Analysing Variation Asia-Pacific 2 con-ference, National Institute for Japanese Language and Linguistics, Tokyo, 1–4August 2012.
37 of 38