Not obligatory Bound pronoun variation in Gurindji and Bilinarra* Felicity Meakins
University of Queensland
This is the first quantitative study of bound pronoun variation in an Australian language. Bound pronouns in Gurindji and Bilinarra (Ngumpin-Yapa, PamaNyungan) are obligatory for first and second persons, categorically absent for the third person minimal, and used 73% of the time to cross-reference third person non-minimal referents and minimal third person oblique referents. A total of 1095 tokens of referents were coded for three predictors: the grammatical relation of the referent, whether the referent was human and whether a co-referential nominal was also present in the clause. A number of properties of the referent significantly decreased the appearance of a bound pronoun including if the referent was non-human, non-human and an object, or also cross-referenced by a nominal. This variation has a number of implications for the function of bound pronouns in discourse and characterisations of non-configurational languages.
A lot of attention has been given to ‘null anaphora’ or the common omission of nominals in Australian languages, in particular Warlpiri (Ngumpin-Yapa, PamaNyungan).1 Null anaphora is one of the properties of non-configurational languages (Hale, 1983) which has generated a lengthy debate in the formal literature * I am enormously grateful to the Gurindji and Bilinarra people who taught me their language and explained its intricacies: Violet Wadrill, Biddy Wavehill Yamawurr, Topsy Dodd Ngarnjal, Ena Oscar Majapula, Sarah Oscar Yanyjingali, Connie Ngarrmaya, Ivy Hector and Mildred Hector. I am also grateful to Erika Charola, Helen and Norm McNair, Patrick McConvell, Rachel Nordlinger and Caroline Jones for the use their Gurindji and Bilinarra data which has supplemented my own in this paper. Thanks also to Mary Laughren who, as always, kept me on the straight and narrow with any claims about Warlpiri. 1. Also described for many other Australian languages including Jiwarli (Austin, 2001), Wambaya (Nordlinger, 1998) and Jingulu (Pensalfini, 2004). Asia-Pacific Language Variation 1:2 (2015), 128–161. doi 10.1075/aplv.1.2.02mea issn 2215–1354 / e-issn 2215–1362 © John Benjamins Publishing Company
Bound pronoun variation in Gurindji and Bilinarra 129
about whether nominals can occupy argument positions in a clause (Austin & Bresnan, 1996; Baker, 1996; Jelinek, 1984; Nordlinger, 1998; Pensalfini, 2004; Simpson, 1991). Other research has focussed on how information structure governs the variable use of nominals (Hale, 1992; Laughren, 2002; Legate, 2002; Simpson, 2007; Swartz, 1991). Much of this work assumes that the other argument marking system in languages such as Warlpiri — the bound pronoun system — is invariable, that is, the pronouns are either obligatory or categorically absent. In fact bound pronouns are not always found where they are expected in Warlpiri (see §3.4.3). Indeed variation in the application of bound pronouns has been observed in two other languages related to Warlpiri which also exhibit null anaphora — Jaru (Tsunoda, 1981, p. 140 onwards) and Bilinarra (Meakins & Nordlinger, 2014, pp. 238–242). Bound pronouns in these languages are obligatory for first and second persons as in (1), categorically absent for the third person minimal as in (2),2 and used only variably for third person non-minimal3 referents and minimal third person oblique referents, for example indirect objects, beneficiaries, oblique possessors and human goals, as in (3) and (4).4,5 (1) Gula=rnaA=ngguO nyunduO=ma gayi ba-rru neg=1min.s > 2min.o 2min=top chase hit-pot ‘I can’t chase you.’ (Bilinarra: AN: RN90–002b: 09:45 min) (2) Garu-lu=ØA=ØO gayi ba-ni jindagu jamud child-erg=3min.s = 3min.o chase hit-pst one turkey ‘A kid chased one bush turkey.’ (Bilinarra: SH: FM08_a087: 10:00 min) 2. The lack of 3rd person singular/minimal pronouns is not unusual across Australia (Dixon, 2002, p. 363). 3. The term ‘minimal’ is mostly synonymous with ‘singular’, but for an explanation of a minimal/unit augmented/augmented system, see footnote 9. 4. All examples will be referenced with the following information: Language (Gurindji or Bilinarra), Speaker (two or three letter initial) and Source (recording, published paper or story). The Gurindji speakers were: Violet Wadrill, Ronne Wavehill Wirrba, Dandy Danbayarri, Biddy Wavehill Yamawurr, Pincher Nyurrmiari†, Blanche Bulngari†, Ena Oscar Majapula, Sarah Oscar Yanyjingali, Topsy Dodd Ngarnjal, Connie Ngarrmaya and Lily Bennett†. The Bilinarra speakers were: Anzac Mungayarri†, Ivy Hector†, Kathleen Juduwurr, Doris Warnmal† and Mildred Hector. In most cases, the name of the recording includes the linguist name or initials. The linguists were: Felicity Meakins, Erika Charola, Helen and Norm McNair, Patrick McConvell, Rachel Nordlinger and Caroline Jones. Some examples will also include the start time of the utterance in a recording. For example, “Gurindji: VW: FM10_23_1b: 1:24min”. 5. I mark grammatical roles in the examples using Dixon’s (1979) syntactico-semantic distinctions of A (transitive subject), s (intransitive subject) and o (transitive object), as well as IO (indirect object), obl (oblique) and adj (adjunct).
130 Felicity Meakins
(3) Guya-ngga=ma=luA gujingga-la gayi ba-rra thus-loc=top=3aug.s click-loc follow hit-prs ‘That’s where they chase him clicking boomerangs together.’ (Bilinarra: AN: RN90–008b: 05:44 min) (4) Dimana-yawung-gulu=ØO gayigayi ba-rra bulugiO horse-prop-erg=3aug.o chase.redup hit-prs cattle ‘He chases the cattle on his horse.’ (Bilinarra: IH: RN90–014a: 20:45 min)
For both Jaru and Bilinarra, Tsunoda (1981), and Meakins and Nordlinger (2014) claim that the interplay between animacy and grammatical relations affects the use of these pronouns. This paper extends work on the variable application of bound pronouns in Ngumpin-Yapa languages by quantifying the tendencies observed in Bilinarra and a closely related language, Gurindji. Statistical modelling produces more precise predictions about the patterns of bound pronoun use. The paper begins with an overview of the bound pronoun systems of Bilinarra and Gurindji (§2). The systems are described in relation to the larger argument marking system of these languages and previous observations of bound pronoun variation in Bilinarra and Jaru. A quantitative analysis is then undertaken on 1095 tokens of non-minimal third person referents and minimal third person oblique referents (§3). They are coded for the presence of a bound pronoun, humanness, grammatical role and the presence of a co-referential nominal. In 27% of cases, bound pronouns are not used to express a referent but they are almost obligatory where the referent is human. The likelihood of their presence decreases if the referent is not human, particularly if the referent is non-human and a direct object, or if a co-referential nominal is present. These results are interesting for a number of reasons. Firstly, variability in the application of bound pronouns means their function is not restricted to argument marking. As discussed in §3.4.1, the convergence of a number of tendencies, such that referents are less likely to be cross-referenced if they are non-human (and in particular are non-human objects) and are also expressed by a nominal, is suggestive of topicality, that is, bound pronouns mark topical referents. Topicality explains much of the use of these pronouns, however the unexpected use of bound pronouns is shown in §3.4.2 to express something about a referent beyond topicality. The strong association of these pronouns with humanness means that they can be co-opted by speakers as another expressive device in their discursive toolkit. For example, the presence of bound pronouns is used to attribute sentience to inanimates and animals in creation myths and their absence undermines the agency and humanity of early white colonists in oral history narratives.
Bound pronoun variation in Gurindji and Bilinarra
Finally, as will be discussed in §3.4.3, the use of bound pronouns follows cross-linguistic tendencies described by Genetti and Crain (2003), for example, the pronouns are avoided for non-human referents and, because objects are generally non-human, pronominal objects are also avoided. The fact that (i) Gurindji and Bilinarra bound pronouns are not obligatory and (ii) these bound pronoun systems pattern like most of the languages examined in Genetti and Crain’s (2003) study which are configurational suggests that the behaviour of bound pronouns cannot be posited as a feature of non-configurationality.
Figure 1. Languages of the Victoria River District (Meakins & Nordlinger, 2014, p. xxxiii)
132 Felicity Meakins
2. Gurindji and Bilinarra pronominal clitics Gurindji and Bilinarra are members of the Ngumpin subgroup of the NgumpinYapa (Pama-Nyungan) family of languages (McConvell & Laughren, 2004). They are spoken in the southern Victoria River District of the Northern Territory, Australia (see Figure 1). Both languages are highly endangered. Gurindji has around 30 first language speakers remaining and Bilinarra no longer has any full speakers. The every day speech practice of middle-aged Gurindji and Bilinarra people is now code-switching between their traditional language and Kriol; and anyone under 40 years of age speaks the newly-developed mixed language, Gurindji Kriol (Meakins, 2008). Nonetheless shift in the use of Gurindji and Bilinarra is not considered to be a factor in phenomenon under investigation. The variable application of bound pronouns is a feature of the speech of the oldest speakers who grew up speaking Gurindji or Bilinarra. Gurindji and Bilinarra are agglutinating languages which exhibit a mix of dependent-marking and head-marking: argument relations are marked on nonobligatory nominals in the form of case inflections. Nominals are marked according to an ergative-absolutive system where transitive subjects are marked ergative as in (5) and intransitive subjects and objects are unmarked as in (6) and (7). Where nominals are present, they are often cross-referenced by bound pronouns (more specifically pronominal clitics), which generally attach to an (often) second position catalyst (in the case of Gurindji) as in (5) and (6), or to the first element of the clause (in the case of both Gurindji and Bilinarra) as in (7). Pronominal clitics align according to a nominative-accusative system.6,7 (5) Ngu=wulaA=rlaIO warluO pirrkap ma-ni nyawa, cat=3ua.s = 3obl fire make do-pst this [yalu-ngkungarlmurang-kulu]A, Nangala-ngku=purrupurruA that-erg trirel.kin.term-erg subsect-erg=and ‘Your two cousins who are my sisters made this fire for (damper) — and Nangala too.’ (Gurindji: VW: FM09_a13_3a: 1:18 min) (6) [Yawarlwarl Wanyil]S ngu=wulaS karrinyani, yala-ngka=ma crested.pigeon nail.tail.kang cat=3ua.s be.impf.pst that-loc=top ‘Both Pigeon and Nail-Tail Kangaroo lived there.’ (Gurindji: VW: FM10_23_1b) 6. Note that true head-marking in the sense of Nichols (1986) should involve pronominal/ agreement marker attachment to verbs. Bound pronominal enclitics can occur with inflecting verbs in Gurindji and Bilinarra, however, generally only when the verb is in second position. 7. For a discussion on pronominal clitic placement, see McConvell (1980, 1996) and Meakins and Nordlinger (2014).
Bound pronoun variation in Gurindji and Bilinarra 133
(7) Nyila=ma=rnaluA ma-ni-rra [yawu=ma jarrwa]O that=top=1aug.exc get-pst-impf fish=top lots ‘We used to get heaps of fish.’ (Bilinarra: IH: Meakins & Nordlinger, 2014, p. 145)
Pronominal clitics distinguish number (minimal, unit augmented and augmented)8 and person (1st, 2nd and 3rd) with 1st person non-minimal pronouns also making an inclusive/exclusive distinction.9 A combined pronominal clitic paradigm for Gurindji and Bilinarra is given in Table 1 (using the Gurindji orthography10 for simplicity). Note that third person minimal subjects and objects are not encoded with bound pronouns and a special third person minimal oblique form exists: =rla. Where two referents are referred to, they are often encoded in a single synchronically unanalysable complex pronoun11 such as =rnayinangulu ‘we do it to them’ in (8) and =yinangulu ‘they do it to them’ in (9). These paradigms are extensive; for Gurindji see Meakins et al.(2013, pp. 41–43), and for Bilinarra see Meakins and Nordlinger (2014, pp. 252–253) or Meakins (2013, pp. 44–45). (8) Ngu=rnayinanguluA,O karrap nya-nga-ni, wamamala-luA cat=1aug.exc.s > 3aug.o watch see-impf-pst girl.redup-erg ‘We used to watch them as young girls.’(Gurindji: VW: Meakins, 2014, p. 288) (9) Warrguwarrguj-ba=yinanguluA,O ma-ni garuO pick.up.redup-ep=3aug.s > 3aug.o get-pst child ‘They picked up the kids.’(Bilinarra: IH: Meakins, 2013, p. 153) 8. The minimal/unit augmented/augmented system is essentially the singular/dual/plural, respectively except for the inclusive forms where the minimal/unit augmented/augmented system has three (not two) inclusive forms covering the minimal =rli or =ngali (= two people — speaker and hearer), unit augmented =rliwula or =ngaliwula (= three people — speaker, hearer and one other) and augmented =rlaa or =ngala (= speaker, hearer and two or more others) categories. 9. Note that a series of free pronouns also exists, exemplified by nyuntu ‘2min’ in (1) and ngayu ‘1min’ in (17). These pronouns are a subclass of nominals and are used for emphasis. They will not be examined further in this paper. 10. The main difference between Bilinarra and Gurindji orthography is the use of /b, d, g/ and /p, t, k/, respectively, for the stop series. Like most Australian languages, these languages have a single stop series with voicing not a distinctive feature. 11. Historically, the person and number information of the subject can be analysed as split with an epenthetic ngu linking the object and number information of the subject: subj.pers — obj/ obl — ngu — subj.num. Another rule whereby 1st minimal clitics are positioned before 2nd and 3rd person pronouns, irrespective of grammatical function, also applies (1min > 2, 3) as shown in (Meakins & Nordlinger, 2014, p. 255 onwards).
134 Felicity Meakins
Table 1. Form of pronominal clitics in Gurindji and Bilinarra Subject
This allomorph is only found in Bilinarra (Meakins & Nordlinger, 2014, p. 227)
Of particular relevance for this paper is the variable use of bound pronouns. It is highly unusual for both a bound pronoun and the cross-referenced nominal to be present in a clause, for example (10) and (11) below where both the transitive subject and object nominals, as well as the complex bound pronoun =yinangkulu express the arguments of the clause. (10) Ngu=yinangkuluA,O [nyila=ma kartipa=ma]O karrayin-ta cat=3aug.s > 3aug.o that=top whitefella=top from.west-loc nya-nga-ni [yarrulan-tu kujarra-lu]A intake-impf-pst young.man-erg two-erg ‘The two young men were watching the whitefellas from the west.’ (Gurindji: RWH: EC98_a027) (11) Waringarri-luA=ma ngu=yinangkuluA,O warrior-erg=top cat=3aug.s > 3aug.o turtturt ma-ni kirri-walijaO=ma grab.redup do-pst woman-pauc=top ‘The warriors grabbed the women.’ (Gurindji: DD: EC98_a012)
Most clauses omit one or more nominal arguments (as seen in (8) and (9) above) which is referred to as ‘null anaphora’ (cf. Hale, 1983) and has been discussed in detail in the formal literature on non-configurational languages (see §3.4.3). Less 12. In the interests of glossing brevity, I gloss this pronominal series as o, but it marks obliques (e.g., benefactives, malefactives, animate goals and oblique possession) and (some) adjuncts (e.g., purposives), as well as objects.
Bound pronoun variation in Gurindji and Bilinarra
well observed is the omission of bound pronouns, for example in (12) the subject kartiya ‘whitefella’ is not referred to by a subject pronoun =lu ‘they’ (although it is clear from previous clauses that the speaker is referring to various managers and overseers on the cattle station where the Gurindji worked). In (13) the singular dative-marked oblique ‘with my grandmother’ is not cross-referenced by the bound pronoun =rla ‘with her’. (12) Kamparrijang-kulu, kula=yinaO jiya-rna-ni kartiya-luA=ma before-erg neg=3aug.o burn-impf-pst whitefella-erg=top ‘The white people didn’t give them anything before.’ (Gurindji: BW: FM09_17_1c) (13) Alrait kutij ngu=wulaS karrinya [ngayiny-ku=ma alright stand cat=3ua.s be.pst 1min.dat-dat=top jaju-wu=ma]OBL mother.mother-dat=top ‘Alright, the two little boys were standing with my grandmother.’ (Gurindji: BW: FM14_a206)
This type of variation is only found with respect to the distribution of the third person non-minimal pronouns=wula ‘3ua.s, those two’, =wuliny ‘3ua.s, the two of them’, =lu ‘3aug, they’, =yina ‘3aug, them’ and the third person minimal oblique =rla ‘3obl, for/to/with him/her/it etc’. Elsewhere the use of pronouns is categorical. For example, first and second person pronouns are obligatory as in (1), (16) and (17) and third person minimal pronouns expressing ‘he/she/it’ are never found (i.e., they do not exist) as in (2), (14) and (15). Non-minimal object pronouns are also categorically omitted when a dative-marked oblique occurs in the same clauses, as demonstrated in (18). The sequence =wuliny=ngku ‘the two of them to you’ is not permissible.13 (14) [Nyawa=ma jawi=ma]O=ØA=ØO, janginyina-luA ba-ni this=top fire=top=3min.s = 3min.o lightning-erg hit-pst ‘This fire, well lightning started it.’ (Bilinarra: AN: RN90–006b: 11:24 min) (15) CH-ngkuA=ma=ØA=ØO paraj pu-nya partaj-ja, kirrawaO name-erg=top find pierce perch-loc goanna ‘CH found a goanna perched (on a fence post).’ (Gurindji: SO: FM10_29_3a) 13. This rule against sequences of two object bound pronouns was first observed for Warlpiri. Hale (1973, pp. 334–335) noted that only the IO can be cross-referenced in a sentence with both a DO and IO. In Walmajarri, in the case of the verb ‘give’, only the IO is cross-referenced (Hudson, 1978, p. 26). In Jaru, a DO can be registered if it is higher on the animacy scale (animate > inanimate; 1st > 2nd > 3rd; pl > du > sg) (Tsunoda, 1981, p. 146).
136 Felicity Meakins
(16) Kula=rnangkulu ma-nku punyu-k neg=1aug.ex.s > 2min.o do-pot good-fact ‘We can’t make you better.’ (Gurindji: VW: FM10_27_1a) (17) Ngayu-warijS jarrakap ngu=rnaS=ngkuIO 1min-alone talk cat=1min.s = 2min.o ma-rna-na ngayuS, nyununyIO=ma talk-impf-prs 1min 2min.dat=top ‘I’m on my own talking to you.’ (Gurindji: VW: FM09_14_1a) (18) Hey lamparra-marnany-ju=ma=ØA=ØO=ngkuIO jayi-ngku, hey father.in.law-trirel.kin-erg=top=3min.s = 3ua.o = 2min.o give-pot [kujarra kirri kujarra]O nyununyIO two woman two 2min.dat ‘Hey your father-in-law who is my son will give two girls to you.’ (Gurindji: VW: FM10_a155)
Note that Gurindji and Bilinarra also contain a generic construction where groups of humans are referred to as a singular entity. In these constructions, it appears that the augmented form is optional as in (12), but in this construction, =Ø is not an optional augmented pronoun but actually the third minimal pronoun. For example in (19), non-Indigenous people are cross-referenced by =Ø “he/she/it” and Gurindji people are cross-referenced by =rla ‘to him/her/it”. In the second clause, they are referred to by =rli “we two”. The same speaker code-switches into Kriol in the following clause, shown in (20), and uses the Kriol singular pronoun im, clearly demonstrating the intended generic construction. These examples are excluded from the analysis in §3 and criteria for distinguishing generic constructions from true variation is given in §3.1. (19) Nyamu=ØA=rlaIO=nga ma-nyja kartiya=maA yalu-wuIO rel=3min.s = 3obl=dub talk-imp whitefella=top that-dat “Waku ngu=rliS karru jaliji” but cat=1ua.inc.s be.pot age.mate ‘The white man should have talked to black man and said, “Let us be friends”.’ (Gurindji: JMK: MCCONVELL_P04–012073) (20) a.
Jam-bala policeman-tu turlk-karra pa-na-ni indef-adj policeman-erg shot-cont hit-impf-pst ‘Some policemen shot them.’ b. Jam-bala im put-im prisoner, send-im-wei indef-adj 3sg put-tr prisoner send-tr-away ‘Others they made prisoners and sent them away.’ (Gurindji-Kriol CS: JMK: MCCONVELL_P04–012073)
Bound pronoun variation in Gurindji and Bilinarra 137
Variation in the application of bound pronouns in Ngumpin-Yapa languages has been previously observed in the grammars of Bilinarra (Meakins & Nordlinger, 2014) and Jaru (Tsunoda, 1981). For Bilinarra, Meakins and Nordlinger (2014, pp. 238–42) note that bound pronouns generally only cross-reference human participants, regardless of whether they are an argument or adjunct. For example, the object pronoun cross-references a malefactive which is not a core argument as in (21). Inanimate and animal participants are generally not cross-referenced by bound pronouns. For example, although the human subject is cross-referenced by =lu ‘they’, the ‘two coolamons’ are not cross-referenced by an object pronoun =wuliny ‘the two of them’ in (22). Cases where non-humans were cross-referenced were observed, however, as in (23). (21) Jangala-luA garrwa-ni=ma=yiO Jarrarda=yiOBL ma-ni subsect-erg hold-pst=top=1min.o love.magic=1min.o do-pst ‘Jangala married me. He did love magic on me.’ (Bilinarra: Meakins & Nordlinger, 2014, p. 238) (22) Birrgab=ma=luA=ØO=nga ma-n.gu [gawarla=ma gujarra=ma]O, make=top=3aug.s = 3ua.o=dub do-pot coolamon-top two=top jindagu-nginyi garndi-nginyi one-source wood-source ‘They might make two coolamons out of one piece of wood.’ (Bilinarra: Meakins & Nordlinger, 2014, p. 239) (23) Wuugarra=rnayinanguluA,O garra, frightened=1aug.exc.s > 3aug.o be.prs ngaja=ngandibanguluA,O baya-wu admon=3aug.s > 1aug.exc.o bite-pot ‘We’re frightened of (the dogs because) they might bite us.’ (Bilinarra: Meakins & Nordlinger, 2014, p. 241)
Table 2 sums up the interplay of animacy and grammatical relations in the use of pronominal clitics in Bilinarra. The table captures the generalisation that the higher the animacy of the referent and the closer the grammatical relation is to being a core argument in the clause, the more likely it will be cross-referenced by a bound pronoun. The limited nature of the Bilinarra data did not reveal variation in the cross-referencing of human referents, as will be shown in §3, where human referents are shown to be not expressed by bound pronouns in some cases in Gurindji. It is assumed, however, that the variation would emerge with a larger Bilinarra corpus.
138 Felicity Meakins
Table 2. Pronominal cross-referencing in Bilinarra (Meakins & Nordlinger, 2014, p. 242) Human
This type of variation in the use of bound pronouns in Ngumpin-Yapa languages was first observed by Tsunoda (1981) in Jaru. He noted that animate nominals were cross-referenced often and inanimate nominals were cross-referenced less often, although cross-referencing was possible, as shown in (24). He also observed that core grammatical roles such as subjects, objects and indirect objects were cross-referenced more often than adjuncts. These generalisations were captured in Table 3 where ‘yes’ means ‘yes, but variably’.
Table 3. Pronominal cross-referencing in Jaru (Tsunoda, 1981, p. 142) Sentence part
Case marking of nominal
Inanimate Case marknoun ing of bound pronoun
Subject (St, Si)
ABSOLUTIVE DATIVE DATIVE-1 DATIVE-2
Yes Yes / /
Yes / Yes Yes
/ / / /
DATIVE-1 LOCATIVE ALLATIVE ABLATIVE ABLATIVE-1.2
/ Yes Yes / Yes
Yes Yes Yes Yes /
/ Yes ? ? /
INSTRUMENTAL DATIVE-1 DATIVE-2 LOCATIVE ALLATIVE ABLATIVE
/ / / / / /
/ / / / / /
Never Never Never Never Never Never
Note. / indicates ‘not relevant’; … indicates ‘data not available’; ? indicates” ambiguity’
Bound pronoun variation in Gurindji and Bilinarra 139
(24) Jangi-nggu mawun-du nga-anuO [murrgun burnu]O bajan-a one-erg man-erg cat-3pl.o three tree cut-prs ‘One man is cutting down three trees.’(Jaru: Tsunoda, 1981, p. 141)
These studies of Bilinarra and Jaru bound pronouns demonstrate that variation is evident in the application of bound pronouns, and that this variation clearly relates to the animacy and grammatical role of the referent. The following section will model the relationship between bound pronouns, animacy and grammatical roles quantitatively in order to more rigorously demonstrate this relationship, and the more complex combined impact of animacy and grammatical roles on the use of bound pronouns. 3. Variation in the application of pronominal clitics in Gurindji and Bilinarra This section builds on previous observations about the variable use of bound pronouns in Bilinarra (and Jaru). The use of and motivations for the third person nonminimal pronouns: =wula ‘3ua.s, those two’, =wuliny ‘3ua.s, the two of them’, =lu ‘3aug, they’, =yina ‘3aug, them’; and the third person minimal oblique =rla ‘3obl, for/to/with him/her/it etc.’ is probabilistically modelled. This study augments the Bilinarra data with Gurindji data to quantify the variation found in these closely related languages. The addition of Gurindji data is necessary because Gurindji is a better documented language with more data available. The combination of data from both languages is justified due to the close genetic relationship between the languages. The model has a number of implications for characterisations about the function of bound pronouns beyond argument marking (§3.4.1 and §3.4.2) and properties of non-configurational languages (§3.4.3). 3.1 Data The data for this study are 1095 tokens of non-minimal third person and minimal third person oblique referents extracted from 80 Gurindji and Bilinarra narratives, procedural texts and dictionary elicitation sessions. Sixteen speakers are represented in the data (see footnote 5 for details). All are first language speakers of Gurindji or Bilinarra.14 14. The Bilinarra data and the Gurindji data collected by Erika Charola is available in the Jaminjungan and Eastern Ngumpin corpus in the DoBeS archives (http://dobes.mpi.nl/projects/ jaminjung/). Gurindji data collected by Meakins is stored in the ELAP archive as a part of the Gurindji Kriol corpus. (http://elar.soas.ac.uk/deposit/0273) and Gurindji narratives recorded by Helen and Norm McNair are available through AUSIL (http://ausil.org.au).
140 Felicity Meakins
Clauses which contained instances of non-minimal third person bound pronouns and the minimal third person oblique =rla were easily extracted through a corpus search, however finding examples of referents with no co-referential bound pronoun required a manual search of the 80 texts. An added difficulty lay in the interpretation of the non-occurrence of a bound pronoun (not to mention when neither a pronoun nor a nominal were present). The non-occurrence of a bound pronoun also encodes third person minimal, as shown in Table 1, therefore it was necessary to devise separate morpho-syntactic criteria to determine whether a referent was in fact non-minimal. Thus only clauses which also contained the following (1–5) were included: 1. Additional number morphology (e.g., -rra ‘pl’ — only found on co-referential demonstratives (25); -walija ‘pauc’ (26); -kujarra ‘dual’; -kuwang ‘pair’): (25) WalnginS ngu=yi=rla mila-ngka=ma karrinyana na nyawa-rraS fly cat=1min.o = 3obl eye-loc=top be.prs.impf foc this-pl ‘This lot of flies are getting in my eyes.’ (Gurindji: VW: FM10_a133) (26) Ngu=rnalu yuwa-na-na nyila=ma kuruwarran-walija=ma cat=1aug.ex.s put-impf-prs that=top cook.rock-pauc=top ‘We put the rocks on the fire.’ (Gurindji: VW: FM10_a148)
2. Reduplication of the co-referential nominal (27) or coverb (28) (which encodes pluractionality through reduplication): (27) Jitirnijitirni ngu=rla jayi-nga-na nya nya nya incomplete.redup cat=3obl give-impf-prs this this this ‘She gave her half-ground seed, like this this and this.’ (Gurindji: VW: FM10_23_1b) (28) Ngu=lu yala-ngka=rni pa-na-ni karnti-ku=warla cat=3aug.s there-loc=only hit-impf-pst stick-erg=foc tampatampang purrp dead.redup finish ‘They killed lots of fish with a stick right there.’ (Gurindji: PN: McNair: K1H)
3. Multiple nominals in an NP: (29) Ngu=n-ku=rla ngaji-wu, jalak yuwa-rru, wayita, an cat=2min.s-ep=3obl father-dat send put-pot pencil.yam and nyampayila kamara, yangunungku whatsitcalled black.soil.yam bush.potato ‘Do you want to send some pencil yams, whatsitcalled, black soil yams and potatoes to Dad.’ (Gurindji: VW: FM10_a155)
Bound pronoun variation in Gurindji and Bilinarra 141
4. Numerals kujarra or gujarra ‘two’ as in (30), jarrwa ‘lots, non-mass’ as in (31), jangkarni ‘lots, mass’ as in (32): (30) Nyundu=ma=n garrwa-la gujarra warlagu 2min=top=2min.s hold-prs two dog ‘You have two dogs.’ (Bilinarra: MH: RN90–015b: 27:56 min) (31) Nyila=ma=rnalu ma-ni-rra yawu=ma jarrwa that=top=1aug.exc get-pst-impf fish=top lots ‘We used to get heaps of fish.’(Bilinarra: IH: Meakins & Nordlinger, 2014, p. 145) (32) Jangkakarni15 ngu=rnalu ma-na-ni, muying, big/many.redup cat=1aug.ex.s get-impf-pst black.plum ngamanpurru=ma conkerberry=top ‘We used to get really big black plums and conkerberries there.’ (Gurindji: VW: FM09_16_2)
5. Context: In (33) Violet Wadrill describes how Gurindji people used to collect young budgerigars to eat. In the first subordinate clause and second main clause, the budgerigars are cross-referenced with the bound pronoun =yina ‘them’, however in the final clause the pronoun is missing but a plural reference is assumed on the basis of the preceding clauses. In (34) the bound pronoun =yinangulu ‘they do it to them ‘refers to the makers and recipients of left-handed boomerangs. The plurality of boomerangs is assumed because traditionally men had personal weapon kits rather than shared weapons. (33) Nyila=ma ngu=rlaa=rla pa-na-ni tikap, that=top cat=1aug.inc=3obl hit-impf-pst chop mapan-ta-waji kuya, nyamu=yina nantananta=ma karrwa-na-ni hollow-loc-agent thus rel=3aug.o little=top hold-impf-pst ‘We would chop at the hollow tree, like this, which held the baby ones.’ Ngu=rnayinangulu=rla paya-rna-ni tikatikap. cat=1aug.exc.s > 3aug.o = 3obl bite-impf-pst chop.redup ‘We would keep chopping them out.’ ma-na-ni ngu=rnalu yirr nantananta=ma. do-impf-pst cat=1aug.exc.s pull.out little=top ‘(Then) we would pull out the little ones.’ (Gurindji: VW: FM10_a149) 15. ‘Big’ is used in conjunction with non-count nouns to express ‘many’. In this example, the reduplication of jangkarni also contributes to this interpretation.
142 Felicity Meakins
(34) Ngu=yinangulu pirrkap ma-na-na wartiwarti-wu cat=3aug.s > 3aug.o make do-impf-prs left.handed-dat ‘They make (boomerangs) for left-handers.’ (Gurindji: BW: FM10_a143)
3.2 Procedure A generalised linear mixed model (GLMM) with logistic link function (glmr; glm2 package in R)16 (Marschner, 2011) was applied to the 1095 tokens to see what affected the use of bound pronouns in Gurindji and Bilinarra. All tokens were coded for a dependent variable (appearance of a bound pronoun), as well as three independent variables (the grammatical relation of the referent, whether the referent was human and whether a co-referential nominal was also present in the clause). Each dependent variable had a number of levels which had categorical values, that is, Y/N or A (transitive subject), s (intransitive subject), o (object), obl (oblique).17 Speaker was coded as a random effect. •
Fixed effects: – Dependent: • Pronominal present o Y, N – Independents: • Grammatical relation o A, s, o, obl • Human o Y, N • Nominal present o Y, N Random effect: – Speaker
The GLMM analysis is appropriate for data in which the dependent variable is binary — that is, the bound pronoun either appears or doesn’t — so a normal distribution of data points is not possible. The use of a logistic link function is necessary when the independent variable levels are categorical, that is, Y/N, A/s/o/obl, rather than a numeric range. The GLMM analysis, like other multilevel logistic regression models, also allows an analysis of the effect of individual variables as 16. Source: http://cran.r-project.org/web/packages/glm2/glm2.pdf 17. An oblique is any nominal where the nominal referent is marked dative e.g., indirect objects; oblique arguments e.g., animate goals; and adjuncts e.g., benefactive, purposives (see Meakins & Nordlinger, 2014, pp. 361–363 for more explanation of these categories).
Bound pronoun variation in Gurindji and Bilinarra 143
well as the combined effect of variables at their different levels. This turns out to be important with this dataset. This analysis also takes into account both fixed and random effects in one procedure. In particular, the specification of ‘Speaker’ as a random effect means the model takes into account that speakers disproportionately contribute to the data under analysis (with differing numbers of tokens) and that individual speakers behave more like themselves than other speakers. This analysis is commonly used in Probabilistic Syntax studies to quantify grammatical variation (for example, Bresnan, 2007; Bresnan & Ford, 2013; Meakins & o’Shannessy, 2010) and is increasingly replacing Varbul/Rbrul in quantitative sociolinguistics as a more sophisticated statistical method for modelling social motivations for the use of linguistic features. 3.3 Results The output of the analysis is given in Table 4. The significant results are bolded. The analysis shows that a bound pronoun occurs 73% (n = 798) of the time, but is less likely if the referent is not human (p < 0.01), a nominal is present (p < 0.05) or the referent is not human and an object (p < 0.001):18 Each of these significant variables will be discussed in the following sections. Interestingly, whether the referent is an object is not enough alone to predict the non-use of a bound pronoun, although it is the only level of grammatical relation with a positive z-value which means that a bound pronoun is less likely to be found cross-referencing an object than a subject (A or s) or oblique, but not significantly so. The model explains a good amount of variation in the data set (conditional R2=0.55). Note that conditional R2 (rather than a marginal R2) calculates variance based on both fixed and random effects and therefore takes account of all factors which are contributing to variation in the data set (Nakagawa & Schielzeth, 2013, p. 136). The estimated probability of not finding a bound pronoun cross-referencing a non-minimal referent is given below in Figure 2. The plot is calculated on ilogit values of the predictors which maps the range (−∞,∞) to the range (0,1), that is, it turns categorical predictors (Y/N, A/s/o/obl) into continuous data (values 0–0.99). The plots clearly show that the probability of a human referent not being cross-referenced with a bound pronoun is almost 0%. This means that humans 18. Note that drop1 was run twice to incrementally remove non-significant interactions, however the resultant model was not significantly different (p = 0.22) from the original model with all level of variables and interactions included. The model was also run without interactions, however the model with interactions performed significantly better (p < 0.01).
144 Felicity Meakins
Table 4. Output of generalised linear mixed model analysis on 1095 tokens Random effects
Analysis conducted on 1095 tokens, 16 speakers Fixed effects
p < 0.001
p < 0.01
Gram Role (s)
Gram Role (o)
Gram Role (obl)
p < 0.05
Interactions: Human (N):Gram Role (s)
Human (N):Gram Role (O)
p < 0.001
Human (N):Gram Role (obl)
Human (N):Nominal (Y)
Gram Role (s):Nominal (Y)
Gram Role (o):Nominal (Y)
Gram Role (obl):Nominal (Y)
Hum (N):Gram Role (s):Nom (Y)
Hum (N):Gram Role (o):Nom (Y)
Hum (N):Gram Role (obl):Nom (Y)
are almost categorically cross-referenced, regardless of their grammatical role or whether a co-referential nominal is present (the two left plots). Non-human referents (i.e., inanimates and animals) are significantly more likely to not be crossreferenced. The plots also show a smaller effect of the presence of a nominal. If the referent is expressed by a nominal, a bound pronoun is slightly less probable (the two upper plots). The plots also clearly demonstrate that an object on its own is not enough to decrease the likelihood of a bound pronoun. This effect is only significant when the referent is also non-human, as shown by the dramatic rise in the line in the two right plots. Each of the significant predictors’ effect on the realisation of bound pronouns will be discussed in the following sections. 3.3.1 Humanness Overall a bound pronoun occurs in 73% of cases, but is significantly less likely if the referent is not human, as shown in Table 5.
Bound pronoun variation in Gurindji and Bilinarra 145
Predicted probability of non−presence of bound pronoun
Human*Grammatical_Role*Nominal effect plot
Nominal: Y Human: 1Y
Nominal: Y Human: 2N
Nominal: N Human: 1Y
Nominal: N Human: 2N
0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
Figure 2. Probability of not finding a bound pronoun according to the humanness, grammatical role and the nominal presence of the referent19 Table 5. Occurrence of bound pronouns according to the humanness of the referent Bound pronoun present N Human
In 58% of cases, bound pronouns were not found cross-referencing a non-human referent, compared with only 9% of human referents. For example in (35) an inanimate transitive subject, ‘small (drops of rain)’, is not cross-referenced by a bound pronoun, but the human object is expressed by =yina ‘them’. Similarly in (36) a bound pronoun is not found cross-referencing an animal referent.
19. The letter abbreviations on the axis relate to the variable levels described at the beginning of §3.2. The numbers are an artefact of R which requires strict ordering of variable levels to determine the intercept. Plots are generated from these variable levels.
146 Felicity Meakins
(35) Ma-na-ni ngu=ØA=yinaO [yapakayi-lu yapakayi-lu]A get-impf-pst cat=3aug.s = 3aug.o small-erg small-erg ‘More and more small (drops of rain) were getting them.’ (Gurindji: VW: FM10_23_4) (36) Jipujipu=ØA ma-na-ni nganta warlu=maO extinguish.redup=3aug.s get-impf-pst dub fire=top [yalu-ngku=ma mungku-ku=ma]A that-erg=top red.faced.turtle-erg=top ‘Those turtles might have been extinguishing the fire.’ (Gurindji: VW: FM10_23_4)
This pattern is not unusual. It follows the cross-linguistic tendency for pronouns to be largely restricted to human referents, and avoided for non-human referents. Genetti and Crain (2003) refer to this as the Avoid inanimate pronouns constraint, which is discussed further in §3.4.3. Cases where the use of pronouns are probabilistically unexpected but appear nonetheless are discussed in §3.4.2. These include examples of humans not cross-referenced by bound pronouns (37) and examples of non-humans which are cross-referenced (38). (37) [Marlurluka kajijirri]S ngu=ØS=ngku=rla ma-rni old.men.redup old.women.redup cat=3aug.s = 2min.o = 3obl talk-pst ‘The old men and women told you to (take it back).’ (Gurindji: VW: FM10_27_1a) (38) Baya-rni=luA=rla, dambang, ngamayiO=ma bite-pst=3aug.s = 3obl dead mother=top ‘(The snakes) bit the mother (cat) to death.’ (Bilinarra: KJ: CJ: Cat and snake)
3.3.2 Presence of a nominal The presence of a co-referential nominal significantly decreases the likelihood of a bound pronoun being found, as shown in Table 6. In 41% of cases where a coreferential nominal is found, the bound pronoun does not appear, compared with 27% (n = 297) overall. Table 6. Occurrence of bound pronouns according to whether a nominal is also present Bound pronoun present N Nominal
599 496 1095
Bound pronoun variation in Gurindji and Bilinarra 147
While this effect seems large, it is largely driven by the fact that, regardless of grammatical role, most non-overt nominals have human referents (75%, n = 450), as shown in (39)–(41). The plot in Figure 2 better demonstrates the combined effect of the presence of nominals with the other variables (humanness and grammatical role). (39) Nyila-rra-lu kujarra-lu kartiya-luA=ma=ØA ngurrku pa-ni that-pl-erg two-erg whitefella-erg=top suspect hit-pst ‘Those two whitefellas started to suspect something.’ (Gurindji: DD: EC99_ a028) (40) [Marlurluka kajijirri]S ngu=ØS=ngkuA=rlaADJ ma-rni old.men.redup old.women.redup cat=3aug.s = 2min.o = 3obl talk-pst ‘The old men and women told you to (take it back).’ (Gurindji: VW: FM10_27_1a) (41) Jiwarrjiwarr ya-nta-ya-nta=luA=ØO kajirriO gather.redup go-imp-go-imp=3aug.s = 3aug.o old.women ‘You mob gather together the old women.’ (Gurindji: DD: EC98_a013)
Of course the converse of this tendency is also true, that is, the absence of nominals increases the presence of bound pronouns which suggests that, probabilistically at least, the bulk of argument marking falls to the bound pronouns which has implications for theories of non-configurationality, as discussed in §3.4.3. 3.3.3 Non-human objects In Gurindji and Bilinarra, 51% of object referents are not cross-referenced compared with 11–19% of subjects (A and S) and indirect objects and obliques (obl) as shown in Table 7. Table 7. Occurrence of bound pronoun according to the grammatical role of the referent Bound pronoun present N Gram role
The effect of the object role itself is not significant, but rather it is the combination of humanness and grammatical role which significantly decreases the likelihood
148 Felicity Meakins
of a bound pronoun (p < 0.001). This is one advantage of GLMM modelling — the ability to test the combined effect of variables. Thus bound pronouns do not appear in 75% of cases where the referent is a non-human object, as shown in Table 8. Table 8. Combined effect of humanness and grammatical role on the occurrence of a bound pronoun Gram role
S O OBL Total
Bound pronoun present N
Some examples of this effect are given below. In (42), a bound pronoun is not found cross-referencing ‘two dogs’, and in (43) the ‘cooking rocks’ are also not cross-referenced on the catalyst. Both objects are clearly non-minimal — warlagu ‘dog’ co-occurs with the numeral gujarra ‘two’ and the kuruwarrany ‘cooking rocks’ are suffixed with a paucal marker -walija. (42) Nyundu=ma=nA=ØO garrwa-la [gujarra warlagu]O 2min=top=2min.s = 3ua.o hold-prs two dog ‘You have two dogs.’ (Bilinarra: MH: RN90–015b: 27:56 min) (43) Ngu=rnaluA=ØO yuwa-na-na [nyila=ma cat=1aug.exc.s = 3aug.o put-impf-prs that=top kuruwarrany-walija=ma]O cooking.rocks-pauc=top ‘We put those rocks (on the fire).’ (Gurindji: VW: FM10_a148)
Interestingly this omission of bound pronouns by two non-configurational languages (Gurindji and Bilinarra) essentially patterns the same as the configurational languages (Genetti & Crain, 2003). This similarity is problematic for the characterisation of non-configurational languages, as discussed in §3.4.3.
Bound pronoun variation in Gurindji and Bilinarra 149
3.4 Discussion The quantitative analysis demonstrated that, though bound pronouns were common in discourse (occurring 73% of the time), a number of properties of the referent significantly decreased the appearance of a bound pronoun including if the referent was not human, non- human and an object, or also cross-referenced by a nominal. The following sections firstly demonstrate that these factors are suggestive of an information structure analysis, specifically topicality. Nonetheless, the strong association with humanness means that the non/use of bound pronouns provides speakers with a discourse device to express attitudinal stance to referents (§3.4.2). Finally, these results have implications for characterisations of non-configurational languages (§3.4.3). 3.4.1 Topicality The previous sections have demonstrated that referents in Bilinarra and Gurindji are less likely to be cross-referenced if they are non-human (and in particular are non-human objects) and are also expressed by a nominal. This convergence of factors, which affects the use of the pronominal clitics =lu ‘they’, =yina ‘them’, =wula ‘those two’, =wuliny ‘the two of them’, and =rla ‘for/to/with him/her/it’, is suggestive of topicality that is, bound pronouns mark topical referents. Cross-linguistically the presence of cross-referencing affixes is commonly conditioned by factors related to information structure, as in studies of objects in Ostyak (Nikolaeva, 1999, 2001) and object indexing in Yimas (Foley, 1992). In grammatical systems which show this type of variable marking, only arguments with topical referents (i.e., referents which have been previously established in discourse) get cross-referenced, while those with new referents do not. For example, Dalrymple and Nikolaeva (2011) observe that cross-referencing of objects is generally associated with old information. In Bilinarra and Gurindji, humans are more likely to be subjects (A or S) than objects and humans are more likely to be cross-referenced with pronominal clitics. Secondly, pronominal clitics are less likely to be found cross-referenced by nominals. Thus, the most common pattern found for a Bilinarra or Gurindji clause is for a subject to be human, not expressed by a nominal, but cross-referenced by a bound pronoun. For example, 82% (n = 268) of lexical As are human; 92% of these are cross-referenced by a bound pronoun, and only 19% of these human As have a co-referential nominal. Conversely o is most likely to be non-human, expressed by a nominal and not expressed by a bound pronoun, as exemplified in (44) and (45). (44) Ngu=lu kampa-rna-ni turturl-arra jarlarlka=ma cat=3aug.s cook-impf-pst roast-cont cat.fish=top ‘They used to roast catfish.’ (Gurindji: VW: FM09_13_3a)
150 Felicity Meakins
(45) Jarrwa=rnalu ma-ni-rra gambij warrija-wu many=1aug.ex.s get-pst-impf egg crocodile-dat guwarlambala-wu gambij turtle-dat egg ‘We used to get crocodile and turtle eggs.’ (Bilinarra: IH: Fishing text)
Where objects are cross-referenced, this is usually because they are humans (hence the association of animacy and bound pronouns). This association is compatible with the notion that humans are most often topical, as in (46)–(47). (46) Ngu=rna=wuliny ka-nya luny na karlarra marru-ngkurra kuya cat=1min.s = 1ua.o take-pst deposit foc west house-all thus ‘Then I would take them west and leave them there.’ (Gurindji: CN: FM10_28_2a) (47) Nyamu=rna=wuliny=nga ka-ngka wartuj rel=1min.s = 1ua.o=dub take-imp out.of.sight ‘I would have taken them out bush and hidden them.’ (Gurindji: BW: FM14_a206)
The pattern found in (44) and (45) follows Du Bois’ (1987) Avoid lexical A (i.e., avoid expressing a transitive subject with an NP) constraint which has a basis in the animacy of referent, as claimed by Everett (2009) and Haig and Schnell (in press). This effect also demonstrates a compounding of Genetti and Crain’s (2003) two constraints: Avoid inanimate pronouns and Avoid pronominal Ps. This convergence of factors and resulting pattern is likely the result of the fact that humans are more likely to be topical and therefore A subjects, and therefore not expressed by a nominal and or cross-referenced by a pronominal clitic. 3.4.2 Unexpected use of bound pronouns as a discourse device Despite the strong correlation between humanness and the use of bound pronouns in Gurindji and Bilinarra, bound pronouns are not found cross-referencing 9% of human referents and are found cross-referencing 42% of non-human referents. These cases are probabilistically unexpected and therefore require some explanation. This section suggests that the role of bound pronouns in Gurindji and Bilinarra extends beyond the standard functions of pronouns, such as indexing arguments, tracking referents through extended discourse and providing number information for unmarked nominals. It demonstrates how these pronouns and their sensitivity to animacy can be brought into service in discourse, enabling speakers to express an attitudinal stance to referents. In §3.3.1, the inherent animacy of referents was shown to affect the application of bound pronouns. Non-human referents were found to be cross-referenced by
Bound pronoun variation in Gurindji and Bilinarra
bound pronouns significantly less often than human referents, further exemplified by (48) and (49). (48) [Ngarin=ma=ØS jarrwa]S karrinyani brija-ngka=ma meat=top lots be.impf.pst fridge-loc=top ‘There was always heaps of meat in the fridge.’ (Gurindji: CN: FM10_28_2a) (49) Ma-na-ni ngu=lu=ØO jarrwaO do-impf-pst cat=3aug.s lots ‘They used to get lots of (the budgerigars).’ (Gurindji: VW: FM10_a133)
In fact the use of bound pronouns to cross-reference humans is almost categorical (91%) and therefore the presence of a bound pronoun can force a human interpretation of a nominal. For example, in a poorly designed elicitation session, I was berated by a Gurindji speaker for producing (50). I was attempting to express dual number for an inanimate, tuku ‘mussel’, forgetting its sexual connotations. Instead the use of the pronoun produced the reading of a human possessed body part. (50) Ngu=rna=wulinyO patpat20 ma-ni tuku-wuO cat=1min.s = 3ua.o feel.for do-pst mussel/clitoris-dat Attempted: ‘I felt around for two mussels.’ Actual: ‘I groped for those two’s clitorises.’
Yet the cases involving a human referent where a bound pronoun is not found and the cases involving a non-human referent where a bound pronoun is found need some explanation. These cases will be discussed in turn. Firstly, all of the non-use of bound pronouns to refer to human referents appear in Gurindji stories of non-Indigenous brutality. In particular bound pronouns are notably absent in stories of the invasion of Gurindji country by European colonists which began in the 1860s. Aboriginal people were massacred so that colonists could seize control of their land to establish cattle stations. In stories of these times,21 the Gurindji shock at the behaviour of early colonists is expressed in a number of ways, but perhaps most subtly through the use of bound pronouns. The strong association of bound pronouns with humans allows the story tellers to convey the savagery and animalism of the early colonists through their absence, as illustrated in (51)–(53). (51) Ngu=ØA=yinaO papart pu-nga-ni cat=3aug.s = 3aug.o massacre pierce-impf-pst ‘(Whitefellas) used to massacre (Aboriginal people).’ (Gurindji: DD: EC98_ a013) 20. This is a Gurindji word, despite the similarities with ‘pat’ from English. 21. See Charola and Meakins (in press) for a compilation of Gurindji historical narratives.
152 Felicity Meakins
(52) Nya-ngka=yiOBL=ØO na kartipaO look-imp=1min.s = 3aug.o foc whitefella ngaja=ØA=ngalaO kartipa-luA paraj pu-ngku lest=3aug.s = 1aug.inc.o whitefella-erg find pierce-pot ‘Watch the whitefellas for me now, otherwise they might find us.’ (Gurindji: RWH: EC98_a027) (53) Nyamu=ØA=yinaO pa-ni ngurra-ngurlung nyanuny-ngurlu=rni rel=3aug.s = 3aug.o hit-pst home-abl 3min.dat-abl=only wurlurturr exterminate ‘When (the whitefellas) brutally removed them from his22 land.’ (Gurindji: PN: MacNair: D2A)
The killings continued until station owners realised that Aboriginal people would make a good source of free labour. In stories of early station times, the lack of bound pronouns also has the effect of denying the managers and other non-Indigenous officials their humanity. Indeed many were perpetrators of inhumane policies. Station managers provided Aboriginal workers and their families with little food and poor accommodation, access to water and sanitation, described in (54), and children with European fathers were taken away from their Gurindji families to Croker Island, some never reunited (55). They formed what is now known as the Stolen Generation. The absence of these pronouns is in stark contrast to the Gurindji referents who are cross-referenced by pronouns in these examples. (54) Kamparrijang-kulu, kula=ØA=yinaO jayi-rna-ni before-erg neg=3aug.s = 3aug.o give-impf-pst kartiya-luA=ma whitefella-erg=top ‘The white people didn’t give them anything before.’ (Gurindji: BW: FM09_17_1c) (55) Long time you know lurtju-kari ngu=ØA=yinaO ka-nya island-other cat=3aug.s = 3aug.o take-pst [pilyingpilying-walija=ma ngaliwany=ma]O half.caste-pauc=top 1aug.inc.dat=top kula kajupari lurtju-kari neg close island-other ‘A long time ago (welfare officers) used to take our pilyingpilying children to an island. The island wasn’t close.’ (Gurindji: VW: FM14_a205)
22. This refers to a particular Aboriginal owner of an estate.
Bound pronoun variation in Gurindji and Bilinarra
The converse use of pronouns also requires some explanation. Non-humans are not expressed by a bound pronoun 58% of the time in the data set, as was exemplified in (48) and (49). Therefore cases where non-humans are cross-referenced are marked. The use of pronouns to cross-reference non-humans seems to accord them human-like attributes such as agency, in the case of transitive subjects (56), or affectedness in the case of intransitive subjects (57) and objects (58): (56) Namata ngu=yinangkuluA,O paya-rna-ni kuya no.matter cat=3aug.s > 3aug.o bite-impf-pst thus ‘It didn’t matter if (the ants) were biting them like this.’ (Gurindji: VW: FM10_a148) (57) Nyamu=luS karrinyani wangiyip=ma [yawu=ma nyila-rra=ma]S rel=3aug.s be.impf.pst dizzy=top fish=top that-pl=top ‘When the fish were dizzy then (from the silt).’ (Gurindji: VW: FM09_13_3a) (58) Nyila=ma=rnaA=wulinyO ba-ni warlagu-gujarraO=ma, ngayiA that=top=1min.s=3ua.o hit-pst dog-du=top 1min ‘I hit those two dogs.’ (Bilinarra: AN: RN90–004b: 25:20 min)
The cross-referencing of non-humans is particularly high in creation myths, referred to as Puwarraja in Gurindji and Bilinarra, or Dreaming stories in Australian English. In these stories, Dreaming creatures traverse the as-yet-unformed landscape laying down its features. These Dreaming creatures take on many forms. They are animals, humans, plants or natural phenomena such as rain or lightning, and are responsible for the creation of hills, rocks, waterholes and trees. In these stories, the use of the bound pronoun animates the Dreaming creatures or imbues them with human-like abilities such as higher order mental capacities. Bound pronouns are used to bring inanimates to life, such as in (59) where a stand of bullwaddy trees springs into existence on Bilinarra country. Dreaming animals become cognisant of the desires of others, as shown in (60), where some birds bury an echidna they have just wronged in a place of the echidna’s choosing. They also have an awareness of other minds, able to deceive for instance, as shown in (61). (59) GamanyjiS garrawarra gudidij-ba=luS waninya gan.gula, bullwaddy east stand.redup-ep=3aug.s fall.pst up nganayirla-ngga Guwadga-ngga. whatsitcalled-loc placename-loc ‘Bullwaddy trees came to stand up in the east at Guwadga.’ (Bilinarra: IH: FM: BIL04.track01.01aud: 08:06 min) (60) Ga-nya=luA, nyawa=na=luA wumara-ngga na take-pst=3aug.s this=seq=3aug.s rock-loc seq
154 Felicity Meakins
yuwa-ni, walyag, [murlu-nggu jurlag-walija-nggu=ma]A put-pst inside this-erg bird-pauc-other=top ‘They took him. Then this big mob of birds put Echidna in the rocks.’
(Bilinarra: Meakins & Nordlinger, 2014, p. 239)
(61) Nyawa nyawa ngu=wulaA wulawulaj jiya-rna-na punyunyu=ma this this cat=3ua.s hide.redup burn-impf-pst good.redup=top ‘Here and here the two (brolgas) hid the good (yams from their grandsons).’ (Gurindji: BP: McNair: Al1B)
Bound pronouns are also used in other genres of narrative to confer higher levels of sentience to inanimates and animals. For example, in the following two excerpts from a historical tale, the speaker amplifies the horror human victims felt watching as clouds possessed by an evil force form over them (62) and as red-faced turtles swarm into their shelters with an air of malevolent deliberation (63). (62) Kankula murla=maS=yinangkuluS,OBL=rla kurlng=parla up this=top=3aug.s > 3aug.o = 3obl cloud.form=foc yuwa-ni, maarn=maS put-pst cloud=top ‘Then clouds started forming over them.’ (Gurindji: VW: FM10_23_4) (63) MungkuS=ma=yinangkuluS,OBL ya-na-ni turtle.sp=top=3aug.s > 3aug.o go-impf-pst ‘The turtles were coming for them.’(Gurindji: VW: FM10_23_4)
Other human characteristics are expressed by the use of bound pronouns, for example, animals become possessors of homes (not just inhabitants) as in (64) and insects and clumps of vegetation are individuated in (65) and (66). (64) Jurlaka-walija-wuOBL ngu=rnaA=yinaOBL ngurra paraj pu-nya bird-pauc-dat cat=1min.s = 3aug.o home find pierce-pst ‘I found the home of some birds.’ (Gurindji: VW: FM14_a228) (65) YirrpanO ngu=rnaA=yinaO paraj pu-nya termite cat=1min.s = 3aug.o find pierce-pst ‘I found some termites.’ (Gurindji: VW: FM14_a228) (66) KurlpapS ngu=luS karrinyani clump cat=3aug.s be.impf.pst ‘(The grass) was (growing) in clumps.’ (Gurindji: VW: FM10_a141)
Thus the strong association of bound pronouns with human-ness provides speakers with an anthropomorphising device which allows them to attribute human-like properties to animals and inanimates. It also provides speakers with an additional
Bound pronoun variation in Gurindji and Bilinarra
attitudinal device in discourse to express their degree of empathy with referents. This discourse use operates in tandem with its other morpho-syntactic functions of argument marking and number cross-reference. 3.4.3 Bound pronouns and theories of non-configurationality In §3.3.2, bound pronouns in Gurindji and Bilinarra were found to be present only 73% of the time. This optionality was shown to relate to the presence of a nominal. The relationship between bound pronouns and nominals has occupied an important place in formal syntactic theories of non-configurational languages. Hale (1983) describes non-configurational languages as having the following three properties: pragmatically determined word order, discontinuous noun phrases, and the common omission of nominals — called ‘null anaphora’ -coupled with the compulsory presence of bound pronouns. Based on these properties, an early suggestion was that nominals cannot occupy argument positions in a clause, with bound pronouns fulfilling this role instead (Jelinek, 1984). Several criticisms of this approach have come from the LFG (Lexical Functional Grammar) literature (Austin & Bresnan, 1996; Nordlinger, 1998; Simpson, 1991). One argument against analysing bound pronouns as arguments are situations where bound pronouns are not found where an argument is required. For example, Simpson (1991, pp. 339–345) notes that in Warlpiri particular arguments are not registered as pronominal clitics on the auxiliary element. These include allative complements, objects of cognate object verbs, and one of the two objects of ditransitive verbs. Additionally no third person bound pronouns exist in Warlpiri, but Simpson (1991, pp. 154–155) interprets an unregistered subject or object by default as third singular. In all of these situations, the bound pronoun is found to be categorically absent. Situations where bound pronouns are found to be variably absent are not discussed for Warlpiri, despite the ramifications for Hale and others’ characterisation of non-configurational languages. For example, in (67) the plural object karli-patu ‘boomerangs’ is cross-referenced by -jana ‘them’, but (68) miyi-pardu ‘ceremonial foods’ is not cross-referenced. (67) Ngajulu-rlui kapi-rnai-janaj karli-patuj jarnti-rni 1sg-erg fut-1sg.s-3pl.o boomerang-pl trim-npst ‘I will trim the boomerangs.’(Warlpiri: Hale, 1982) (68) Muturnamuturna-rlu kala-lu miyi-pardu-ku old.woman.redup-erg pst.hab-3pl.s veg.food-dim-dat kunarri-kingarnti-rli-ji karla-ja ceremony.food-preparative-erg-top dig-pst
156 Felicity Meakins
yurturlu-nyayirni yarla-ju heap-really yam-top ‘The old women would go and dig up big numbers of yams in preparation for the presents of food.’ (SIL Warlpiri dictionary)
In Gurindji and Bilinarra, bound pronouns were found to be present only 73% of the time. This optionality was shown to relate to the presence of a nominal. When a nominal was absent, usage of the bound pronoun increased to 84%. The tendency for the absence of nominals to increase the presence of bound pronouns does suggest that, probabilistically at least, the bulk of argument marking falls to the bound pronouns. Nonetheless this still leaves 16% of sentences where neither a nominal nor bound pronoun is present to mark particular arguments, as in (69) and (70) where ‘the flies’ and ‘the two welfare officers’ are not expressed by a nominal or bound pronoun. (69) Paya-rna-na ngu=ØA=ngantipaO wululup-kula=ma bite-impf-prs cat=3aug.s = 1aug.ex.o swarm.redup-loc=top ‘(The flies) bit us when they swarm into our eyes.’ (Gurindji: VW: FM10_ a133) (70) Yuwayi yala-ngkurra ngu=ØA=yinaO ka-nya yes that-all cat=3aug.s = 3aug.o take-pst karu-walija=maO Jinparrak-nginyi=ma child-pauc=top place.name-source=top ‘That’s right, (the two welfare officers) took a lot of children from old Wave Hill Station to that place.’ (Gurindji: VW: FM14_a205)
These omissions are probably not detrimental to formal theories of argument marking in non-configurational languages. Presumably the solution devised for the categorical absence of bound pronouns works for variable absences as well. For example, within the Minimalist Framework, Baker (2000) solves the categorical absence of bound pronouns by suggesting that all argument positions are filled by phonologically null pronouns (pro). Bound pronouns, where they occur, bear a licensing/agreement relationship to these pros (AGR) and nominals act as secondary predicates. The LFG solution to this problem is to also downplay the status of pronominal clitics to agreement markers and to posit a null pronominal element pro which fulfils the argument requirements of the verb if a free nominal is not present to act as an argument (Simpson, 1991). Thus the variable use of bound pronouns exemplified by (69) and (70) presents no difficulties for the formal characterisation of non-configurational languages. Nonetheless, the fact that the omission of bound pronouns in Gurindji and Bilinarra, two non-configurational languages, essentially patterns the same
Bound pronoun variation in Gurindji and Bilinarra 157
as configurational languages is problematic. As was discussed in §3.3.3, the use of bound pronouns follows Genetti and Crain’s (2003) Avoid inanimate pronouns and Avoid pronominal Ps constraints. Because bound pronouns pattern according to the same principles as configurational languages, this strongly suggests that null anaphora is not a defining feature of non-configurationality as first suggested by Hale (1983). 4. Conclusion This paper has taken Meakins and Nordlinger’s (2014) observation about the variable application of third person non-minimal bound pronouns and the third person minimal oblique pronoun in Bilinarra and modelled the probability of the pronoun use in both Bilinarra and Gurindji according to a number of predictors including humanness, grammatical role and the presence of a co-referential nominal. It was found that if the referent of the bound pronoun was not human, nonhuman and a grammatical object or if a co-referential nominal was also present, the likelihood of the appearance of a bound pronoun was significantly decreased. This pattern of use was shown to be common cross-linguistically, conforming to Genetti and Crain’s (2003) constraints on pronouns, that is, Avoid pronominal P and Avoid inanimate pronouns. Thus Genetti and Crain’s constraints seem to apply to both configurational and non-configurational languages, which suggests that the use of bound pronouns in non-configurational languages (cf. Hale, 1983) is not a useful criterion for these languages. Bound pronouns are not merely obligatory or categorically absent as characterised by formal theoretical work on these languages, but instead many apply variably and abide by cross-linguistic norms. Although third person non-minimal bound pronouns and the third person minimal oblique pronoun were found to be variable, human referents are almost always cross-referenced. This pattern largely relates to topicality, however in a small number of cases, the pattern was shown to be aberrant, that is, non-human referents are cross-referenced less frequently but again many cases exist where animals and inanimates are expressed by a bound pronoun. In these aberrant cases, speakers convey attitudinal information. In the case of the absence of human pronouns, human characteristics such as animacy, control and empathy were denied to the referents. The bulk of these instances are found in historical narratives of the early brutal years when non-Indigenous settlers established themselves on Gurindji land. The opposite was also found in creation stories where inanimates and animals were accorded human-like characteristics through the use of bound pronouns.
158 Felicity Meakins
Abbreviations a admon acc agent all aug aux cat dat dem dim du dub ep erg exc fact foc fut gen imp impf inc lest loc
transitive subject admonitive accusative agentive allative augmentative auxiliary catalyst dative demonstrative diminutive dual dubitative epenthetic ergative exclusive factive focus future genitive imperative imperfective inclusive lest locative
neg nmlz npst nom now o obl pauc pl poss pot prop prs pst redup s sg source top ua 1 2 3 – = >
negative nominaliser non-past nominative focus object oblique paucal plural possessive potential proprietive present past reduplication subject singular source topic unit augmented first person second person third person morpheme break clitic break acting on
References Austin, P. (2001). Word order in a free word order language: The case of Jiwarli. In J. Simpson, D. Nash, M. Laughren, P. Austin, & B. Alpher (Eds.), Forty years on: Ken Hale and Australian languages (pp. 205–323). Canberra: Pacific Linguistics. Austin, P., & Bresnan, J. (1996). Non-configurationality in Australian aboriginal languages. Natural Language and Linguistic Theory, 14, 215–268. DOI: 10.1007/BF00133684 Baker, M. (1996). The polysynthesis parameter. Oxford: Oxford University Press. Baker, M. (2000). The natures of nonconfigurationality. In M. Baker, & C. Collins (Eds.), The handbook of contemporary syntactic theory (pp. 407–438). Oxford: Blackwell. Bresnan, J. (2007). Is syntactic knowledge probabilistic? Experiments with the English dative alternation. In S. Featherston & W. Sternefeld (Eds.), Roots: Linguistics in search of its evidential base (pp. 77–96). Berlin: Mouton de Gruyter.
Bound pronoun variation in Gurindji and Bilinarra 159 Bresnan, J., & Ford, M. (2013). Predicting syntax: Processing dative constructions in American and Australian varieties of English. Language, 86(1), 186–213. Charola, E., & Meakins, F. (Eds.). (in press). Yijarni: True stories from Gurindji country. Canberra: Aboriginal Studies Press. Dixon, R.M.W. (1979). Ergativity. Language, 55(1), 59–138. DOI: 10.2307/412519 Du Bois, J. (1987). The discourse basis of ergativity. Language, 63(4), 805–855. DOI: 10.2307/415719 Everett, C. (2009). A reconsideration of the motivations for preferred argument structure. Studies in Language, 33(1), 1–24. DOI: 10.1075/sl.33.1.02eve Foley, W. (1992). The Yimas language of New Guinea. Stanford: Stanford Univeristy Press. Genetti, C., & Crain, L.D. (2003). Beyond preferred argument structure: Sentences, pronouns and given referents in Nepali. In J. Du Bois, L. Kumpf, & W. Ashby (Eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam: John Benjamins. DOI: 10.1075/sidag.14.10gen Haig, G., & Schnell, S. (in press). The discourse basis of ergativity revisited. Language. Hale, K. (1973). Person marking in Walbri. In S.R. Anderson & P. Kiparsky (Eds.), A festschrift for Morris Halle (pp. 308–344). New York: Holt, Rinehart and Winston. Hale, K. (1982). Some essential features of Warlpiri verbal clauses. In S. Swartz (Ed.), Papers in memory of Lother Jagst: Work paper of SIL-AAB (pp. 217–315). Darwin: SIL. Hale, K. (1983). Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory, 1(1), 5–47. DOI: 10.1007/BF00210374 Hale, K. (1992). Basic word order in two “free word order” languages. In D. Payne (Ed.), Pragmatics of word order flexibility (pp. 63–82). Amsterdam: John Benjamins. DOI: 10.1075/tsl.22.03hal Hudson, J. (1978). The core of Walmatjari grammar. Canberra: AIAS. Jelinek, E. (1984). Empty categories, case and configurationality. Natural Language and Linguistic Theory, 2, 39–76. DOI: 10.1007/BF00233713 Laughren, M. (2002). Syntactic constraints in a ‘free word order’ language. In M. Amberger & P. Collins (Eds.), Language universals and variation (pp. 83–130). Westport, CT: Praeger. Laughren, M. (2011). Reconstructing first and second singular pronouns across northern PamaNyungan languages. Paper presented at the TELC Meeting at Charles Darwin University, Darwin. Legate, J. (2002). Warlpiri: Theoretical implications. Unpublished doctoral dissertation, Massachusetts Institute of Technology, Boston. Marschner, I. (2011). glm2: Fitting generalized linear models with convergence problems. The R Journal, 3(2), 12–15. McConvell, P. (1980). Hierarchical variation in pronominal clitic attachment in the Eastern Ngumpin languages. In B. Rigsby & P. Sutton (Eds.), Papers in Australian linguistics No. 13:Contributions to Australian linguistics (pp. 31–117). Canberra: Australian National University. McConvell, P. (1996). The functions of split-wackernagel clitic systems: Pronominal clitics in the Ngumpin languages. In A.L. Halpern & A.M. Zwicky (Eds.), Approaching second: Second position clitics and related phenomena (pp. 299–332). Stanford, CA: CSLI Publications. McConvell, P., & Laughren, M. (2004). Ngumpin-Yapa languages. In H. Koch & C. Bowern (Eds.), Australian languages: Reconstruction and subgrouping (pp. 151–177). Amsterdam: John Benjamins. DOI: 10.1075/cilt.249.11mcc
160 Felicity Meakins Meakins, F. (2008). Unravelling languages: Multilingualism and language contact in Kalkaringi. In J. Simpson &G. Wigglesworth (Eds.), Children’s language and multilingualism: Indigenous language use at home and school (pp. 247–264). New York: Continuum. Meakins, F. (2013). Bilinarra to English dictionary. Batchelor, Australia: Batchelor Press. Meakins, F. (2014). Nominals as adjuncts or arguments: Further evidence from language mixing. In R. Pensalfini, D. Guilleman & M. Turpin (Eds.), Language description informed by theory (pp. 283–315). Amsterdam: John Benjamins. DOI: 10.1075/slcs.147.12mea Meakins, F., McConvell, P., Charola, E., McNair, N., McNair, H., & Campbell, L. (2013). Gurindji to English dictionary. Batchelor, Australia: Batchelor Press. Meakins, F., & Nordlinger, R. (2014). A grammar of Bilinarra: An Australian aboriginal language of the northern territory. Berlin: Mouton de Gruyter. Meakins, F., & O’Shannessy, C. (2010). Ordering arguments about: Word order and discourse motivations in the development and use of the ergative marker in two Australian mixed languages. Lingua, 120(7), 1693–1713. DOI: 10.1016/j.lingua.2009.05.013 Nakagawa, S., & Schielzeth, H. (2013). A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution, 4(2), 133–142. DOI: 10.1111/j.2041-210x.2012.00261.x Nikolaeva, I. (1999). Object agreement, grammatical relations, and information structure. Studies in Language, 23(2), 331–376. DOI: 10.1075/sl.23.2.05nik Nikolaeva, I. (2001). Secondary topic as a relation in information structure. Linguistics, 39(1), 1–49. DOI: 10.1515/ling.2001.006 Nordlinger, R. (1998). Constructive case: Evidence from Australian languages. Stanford, CA: CSLI Publications. Pensalfini, R. (2004). Towards a typology of configurationality. Natural Language and Linguistic Theory, 22, 359–408. DOI: 10.1023/B:NALA.0000015794.02583.00 Simpson, J. (1991). Warlpiri morpho-syntax: A lexicalist approach. Dordrecht, Holland: Kluwer. DOI: 10.1007/978-94-011-3204-6 Simpson, J. (2007). Expressing pragmatic constraints on word order in Warlpiri. In A. Zaenen, J. Simpson, T.H. King, J. Grimshaw, J. Maling, & C. Manning (Eds.), Architectures, rules, and preferences: Variations on themes by Joan W. Bresnan (pp. 403–427). Stanford, CA: CSLI Publications. Swartz, S. (1991). Constraints on zero anaphora and word order in Warlpiri narrative text. Darwin: SIL. Tsunoda, T. (1981). The Djaru language of Kimberley, Western Australia. Canberra: Pacific Linguistics.
Author’s address Felicity Meakins School of Languages and Cultures University of Queensland Brisbane QLD 4072 Australia [email protected]
Bound pronoun variation in Gurindji and Bilinarra 161
Abstract (Gurindji) Nyawa-ma kamparrijang milimili bound-pronouns ngumpit-nginyi jaru-ngayiny. Nyamu-lu jarrakap marnana Gurindji-ma Bilinarra-ma, sometimes ngulu marnana bound-pronounjawung, sometimes bound-pronoun-murlung. Ngulu jarrakap yuwanana -rna-wu-ma, -ngkuwu-ma nyampa-kayirnikayirni. Kula-rnalu jarrakap marnana nyantu-wu-ma. Ngulu jarrakap yuwanana -lu, -yina, -wula, -wuliny, an -rla 73%. Nyawa-ma milimili-ma ngurna warrkuj manana jarrwalut jaru (1095). Ngurna yuwani bound-pronoun-la nyamu bound-pronoun wanyjiwu ngumpit-ku, nyamu bound-pronoun ngantuku manana nyampa-kayirnikayirni, nyamu nominal-rningan karrinyana. Ngulu yusim marnana bound pronoun-ma nyamu kula-lu ngumpin wanyji, nyamu kula-lu ngumpin an object an nyamu-lu yusim marnana nominal-rningan.