Pragmatically case-marked: Non-syntactic functions of the Kuuk Thaayorre ergative suffix

Pragmatically case-marked:

Non-syntactic functions of the Kuuk Thaayorre ergative suffix

Abstract: In Kuuk Thaayorre, ergative marking is of both syntactic and pragmatic import. Syntactically, ergative inflection marks a noun phrase as the subject of a transitive clause. Though this may be considered definitional of an ergative morpheme, Kuuk Thaayorre joins a growing number of languages in which ergative marking is documented to be ‘optional’; not obligatorily present in all transitive clauses. Conversely – and more unusually – the subject of a Kuuk Thaayorre intransitive clause may in some cases be ergative-marked. This chapter proposes that as well as signifying the ergative case relation, the ergative morpheme’s presence in an intransitive clause signals that the subject referent is ‘unexpected’, and its absence from a transitive clause signals that the subject referent is ‘expected’.

1. Introduction¹

Researchers have long recognised that both syntactic context and morphological form must be taken into account when identifying the case array of any language. This paper examines cases of mismatch between the syntactic licensing of ergative case and the distribution of the formal markers of ergativity in one Australian language, Kuuk Thaayorre. It argues that this discord between syntactic case assignment and morphological case-marking can only be explained if we allow that discourse context and pragmatics also play a role in determining the distribution of case markers.

At first glance, Kuuk Thaayorre appears to be a straightforwardly morphologically ergative language, possessing a set of morphs that attach to the subject of a transitive clause. However, these same morphs are sometimes found on the subjects of intransitive clauses, and may in other cases be omitted from the subjects of transitive clauses. This chapter argues that in such contexts Kuuk Thaayorre speakers employ ergative case morphology for pragmatic rather than syntactic ends. Specifically, the ergative morpheme may be affixed to any subject that is pragmatically marked, and omitted where the subject is pragmatically unmarked, regardless of clausal transitivity. For the purposes of this paper, a subject can be considered pragmatically (and consequently morphologically) unmarked if the addressee can be expected to correctly map the referent of an unmarked NP to the subject function purely on the basis of the preceding discourse and/or world knowledge. Where the identification of the subject participant is less straightforward, the subject is pragmatically marked. This question of pragmatic markedness crucially revolves around the mapping of referents to grammatical function, and not the retrievability of referents per se. A NP whose reference is ‘given’ in the particular context will typically be elided, rendering moot the question of ergative marking.
Following a brief discussion of the distinction between (syntactic) case and case form in §2, §3-4 outline the respective contributions of discourse context and world knowledge in determining the case marking of the subject. I argue that without reference to these two pragmatic factors, it is impossible to accurately predict the distribution of ergative case-marking. The fact that pragmatic context plays a role in determining ergative case-marking begs the question of whether the ‘ergative’ morpheme can rightly be considered a case in the first place. This question is explored in §5. Pragmatic functions of case morphology have been documented in a growing number of Australian languages (Pensalfini 1999, McGregor 1998, McGregor 2006, Meakins & O’Shannessy 2004 - see §7), and the data reported here contribute to the understanding of case as a multi-stratal phenomenon.
1.1 Language Background

Kuuk Thaayorre is a Paman language spoken on the west coast of Cape York Peninsula, Australia. Predominantly dependent-marking, Kuuk Thaayorre is located at the non-configurational end of the spectrum (Austin & Bresnan 1996), with extremely flexible ordering of constituents and common ellipsis of arguments (and other constituents). Word order within the noun phrase is more fixed, but permutations motivated by focus or speech style are common in natural speech. An argument (if not elided entirely) may take the form of one or more noun phrases apposed in the same case, as in (1)². Case is realised on the last constituent of the NP (or rather N' – cf. Gaby 2006: 290). Each NP constituent of a (complex) conjoined NP or inclusory construction inflects for case independently. The repetition of pronouns is common, in both full and reduced (encliticised) forms:

(1) ngali I. C. ngali yat kuthirr

1du:excl(nom) I. C.(nom) 1du:excl(nom) go:p.pfv two(nom)

‘I. C. and I went, the two of us’
Note that (1) contains an inclusory construction, in which a non-singular pronoun is apposed to a noun phrase denoting a subset of the participants represented by the pronoun³.
Syntactically, Kuuk Thaayorre possesses a tripartite case system, distinguishing nominative, ergative and accusative arguments. Morphologically, however, there is a split between the pronominal paradigm (in which the syncretism of nominative and ergative case forms results in a nominative-accusative marking pattern) and full noun phrases (in which nominative/accusative syncretism gives rise to a ergative-absolutive marking pattern). Following Goddard (1982), the syntactic case of unmarked nominals and pronouns is glossed in brackets. As this paper is concerned with the ergative case marker, our focus will be on the case marking of nouns and not pronouns⁴.
Case is marked on the phrase-final element of the noun phrase only, as is fairly common across the Australian continent (cf., for example, Diyari (Austin 1981), Yankunytjatjara (Goddard 1985), Ngarinyin (Rumsey 1982) and Wik Mungkan (Dench & Evans 1988:5)). Such phrasal marking might be taken to suggest that the case morphs are postpositional enclitics, yet their formal irregularity is more characteristic of inflectional affixes. Compare, for example, the forms of the following ergative morphs:

	Nominative	Ergative
‘meat’	minh	minhal
‘good’	min	minthurr
‘tooth’	kiin	kiina
‘dog’	kuta	kutaku
‘woman’	paanth	paanthu
‘eye’	meer	meere
‘cat’	thok	thokun
‘saltwater croc.’	pinc	pinci
‘child’	parr_r	parran
‘boomerang’	werngr	werngarr

Table 1. Comparison of nominative and ergative forms of ten Kuuk Thaayorre nouns

Neither phonological environment nor semantics (i.e. noun classes) can fully predict the form of the ergative morph⁵. As such lexically-determined allomorphy cannot be reconciled with an analysis of the ergative morpheme as an independent grammatical unit, I will continue to refer to the ergative markers as (inflectional) case suffixes⁶. It should be noted that it is these same lexeme-specific morphs that are used pragmatically to mark an ‘unexpected subject’, as described in §3.
In addition to the three core cases, Kuuk Thaayorre nouns and pronouns inflect for the genitive, dative and ablative cases. Instrumental adjuncts take ergative case. Comitative, proprietive and privative ‘cases’ are encoded by enclitics. It is important to note for later discussion that the ergative and comitative cases are formally distinguished.

Case	Noun form ‘woman’	Pronoun form ‘3sg’
Ergative	paanth-u	nhul
Nominative	paanth	nhul
Accusative	paanth	nhunh
Genitive	paanthak	nhangan
Dative	paanthun	nhangun
Ablative	paanthum	nhanganma
Comitative	paanth=kak	(nhangun)^7
Proprietive	paanth=(k)^8aak	-
Privative	paanth=(k)aar	-

Table 2. Case forms of nouns and pronouns

1.2 Data

The data presented in this paper were collected during three visits made by the author to the community of Pormpuraaw (home to the majority of Kuuk Thaayorre speakers) over seven months in total. The contexts in which they were elicited may be roughly classified as either: (a) elicited narrative; or (b) prompted natural speech. Elicited narratives in this corpus were gathered by negotiating a particular topic upon which the consultants would expound without further interference. Prompted natural speech arose from presenting consultants with a visual (usually video) stimuli to describe. Alternatively, a pair or group of consultants might be asked to play a game or attend to a task requiring verbal interchange. The responses of the consultant(s) in these cases are usually lengthy and fluent. However, the fact that the subject matter has been manipulated by the linguist may affect language use. Unmonitored natural speech might be assumed to represent a language most accurately (being language as it is really used), but it is also prone to speech errors. Decontextualised elicited data, on the other hand, may suffer from a particular form of ‘judgement hypercorrection’; with consultants rejecting sentences as ungrammatical because they are not provided with the discourse context in which they might be uttered. For these reasons, I have decided to focus upon elicited narratives and prompted natural speech, which are sufficiently fluent and lengthy (as well as being contextually anchored) to be pragmatically rich, while also being controlled enough to allow for error detection.

The data set analysed here is in places supplemented by that reported in Hall (1972) and Foote & Hall (1992).
2. Case vs case-form

It is widely recognised that the formal manifestation of case must be distinguished from syntactic case categories (e.g. Mel’čuk 1979, Goddard 1982, Blake 1994 and Spencer 2006). Clearly, though, there must be some alignment between the distribution of case morphs and the syntactic function of the arguments to which they attach if these morphs are to be labelled case-markers in the first place. In Kuuk Thaayorre, for instance, the ergative morpheme is labelled such because of the regularity with which it attaches to the subjects of transitive clauses but not the subjects of intransitive clauses. This can be seen in the elicited – and thus decontextualised – examples (2-4):

(2) pam ith Cairns-na yat

man(nom) that Cairns-dat go:p.pfv

‘that man went to Cairns’
(3) pam minhal patha-rr

man(acc) animal:erg bite-p.pfv

‘the animal bit the man’
(4) pam-al minh patha-rr

man-erg animal(acc) bite-p.pfv

‘the man bit the meat’
As we would expect of a morphologically ergative language, the unmarked form pam ‘man’ occurs as both the subject of an intransitive verb (yan ‘go’ in (2)) and the object of a transitive verb (path ‘bite’ in (3)), while the marked form pamal occurs with the function of transitive subject in (4). The distribution of marked and unmarked pronominal forms differs, however:
(5) ngay Cairns-na yat

1sg(nom) Cairns-dat go:p.pfv

‘I went to Cairns’
(6) ngay nhunh patha-rr

1sg(erg) 3sgacc bite-p.pfv

‘I bit him’
(7) nhul nganh patha-rr

3sg(nom) 1sgacc bite-p.pfv

‘he bit me’
In the pronominal paradigm, the direct object function is distinguished by the accusative case form (e.g. nganh in (7)) which stands in opposition to a single subject pronoun (e.g. ngay in (5) and (6)). Since each of these three core cases (nominative, ergative and accusative) is morphologically distinguished in at least a subpart of the nominal system⁹, they are analysed as syntactically distinct across the board.
However, a syntax-based account of ergative case-marking in Kuuk Thaayorre is challenged by examples such as (8). Here we find an ergative-marked subject of an intransitive clause. For clarity, the gloss ‘erg^’ will be employed where the morph occurs in such syntactically non-ergative contexts. In (9) the transitive subject is unmarked (where ergative marking would be expected – this absence of ergative marking in syntactically ergative contexts will be signalled by the gloss ‘#erg’).
(8) Parr-an pul kuta-ku ngok-eln wontr

child-erg^ 3du(nom) dog-erg^ water-dat fall:npst

‘the child and the dog fall into the water [together]’
(9) minh patp piinth.kat waawath

animal hawk(#erg) scrap(acc) rdp:search:npst

‘hawks fossick for scraps’
In examples such as these, the morphological case forms found in the clause are out of alignment with the syntactic case of the arguments they mark. Following from this, it can be seen that the syntactic case of arguments does not solely determine morphological marking. Instead, I contend, pragmatic considerations contribute to the inclusion or omission of the ergative morpheme. In §3, I outline the range of pragmatic factors that shape whether a subject is ‘marked’ or ‘unmarked’.

3. Pragmatic case

3.1 Marked intransitive subject

It has already been suggested that the ergative suffix may (occasionally) be attached to the subject of an intransitive verb to mark an ‘unexpected’ subject; one for which the referent is ambiguous or potentially difficult to retrieve. Most commonly, the subject is marked as unexpected because of a change in protagonist, or topic chain¹⁰. This fits with McGregor’s (1998:516) ‘Expected Actor Principle’¹¹, which states that:

The episode protagonist is - once it has been established - the expected (and unmarked) Actor of each foregrounded narrative clause of the episode; any other Actor is unexpected.
The use of ergative marking where the membership of the protagonist group is in flux is illustrated by (10). The key protagonists throughout the text from which (10) is excerpted are the speaker and his siblings, denoted by ngancin ‘we (plural, exclusive)’. However, in the second and third lines, the set of referents denoted by ngancin expands to include the parents of the original protagonists. In line 2, the enlarged referent group is signalled by an inclusory construction, composed of the unchanged pronoun ngancin ‘we (plural, exclusive)’ apposed to noun phrases singling out each of the parents. Crucially, each of these noun phrases is ergative-marked, despite the intransitivity of the clause as a whole:
(10.1) ngancin kanangkarr parr_r mant

3pl:excl(nom) long.ago child small(nom)

‘when we used to be small kids’
(10.2) ngul ngancin nganip-i ngancin,

then 1pl:excl(nom) father-erg^ 1pl:excl(nom)

mami ngathan-man ngancin…

mum 1sgPOS-erg^ 1pl:excl(nom)

ngancin irrkaw yat

1pl:excl(nom) to:West go:p.pfv

‘we, including Dad and Mum, went out bush’
(10.3) ngul ngancin parr_r mant ngancin

then 1pl:excl(nom) child small 1pl:excl(nom)

yuk-un thaangkm ngancin

tree-dat climb:p.ipfv 1pl:excl(nom)

‘then we small children were climbing a tree’
(10.4) thowolnam ngancin parr_r mant

play:p.ipfv 1pl:excl(nom) child small

‘we small children were playing’
(10.5) ngul ngancin… nganip-i ngancan thakarr

then 1pl:excl(nom) father-erg 1pl:excl:acc leave:p.pfv

pul nganam-u

3du(nom) mother-erg

‘then we… father and mother left us [to go fishing]’
The fact that the form of the subject pronoun (ngancin ‘we’) does not change - despite the variation in the set of individuals denoted by it across clauses - makes the tracking of discourse participants particularly problematic. I argue that it is for this reason that the speaker marks the unexpected protagonists (nganip ‘father’ and mami ‘mum’ in (10.2)) with ergative case, despite their function as subject of a strictly intransitive verb (yan ‘go’). In addition to this emphatic case marking, the frequent repetition of the subject pronoun (ngancin ‘we’) – four times in the single clause in line two – no doubt serves to reinforce the fact that these unexpected actors are now included within the reference of this subject pronoun, where they had previously been (and would be again) excluded from it. This is also made clear in subsequent clauses ((10.3) and (10.4)), where the reference of ngancin reverts back to the children alone. In both of these clauses, the full noun phrase parr_r mant ‘small children’ is included in addition to the pronoun, in order to ensure that the (once again) reduced reference set of the subject is understood. This repetition of a full noun phrase for an established discourse participant is marked (though not unusual) in Kuuk Thaayorre discourse, and can be considered alongside ergative marking as a complementary strategy for signalling a problematic subject-to-referent mapping.
Note also that although the ergative-marked noun phrases (nganipi and mami ngathanman (in 10.2) and nganipi and nganamu (in 10.5)) are apposed to unmarked subject pronouns (glossed as ‘nominative’), this does not necessarily signal case disagreement. Because the nominative and ergative cases are not formally distinguished in the pronominal paradigm, ngancin (in 10.2) and pul (in 10.5) could alternatively be glossed ‘erg^’ (i.e. syntactically nominative but formally ergative).
Turning now to example ((11), the ergative case is again employed where two participants, who had been acting independently in the preceding text, now come to act as a unit (i.e. a single protagonist group). This excerpt is taken from a description of the Frog Story¹², at the point at where the boy and the dog (who had been having independent adventures in the preceding text) both end up falling into the same pool:
(11.1) Minh-al thunpirr parr_r ngotonci-ntam

animal-erg propel:p.pfv child(acc) hill-abl

‘the bull throws the boy from the hill’
((11.2) Kuta yokun.manorrp wontr

dog(nom) same.way fall:npst

‘the dog also falls’
((11.3) Parr-an pul kuta-ku ngok-eln wontr

child-erg^ 3du(nom) dog-erg^ water-dat fall:npst

‘the child and the dog fall into the water [together]’
That the ergative morpheme in ((11.3) is not marking syntactic function is clear from its absence in ((11.2), where the unmarked NP kuta ‘dog’ functions as subject of the intransitive verb wontr ‘fall’ just as do the marked NPs kutaku ‘dog’ and parran in ((11.3).
Example (12) similarly includes ergative marking on a noun phrase functioning as intransitive subject:
(12) G. Y. nhul driver. Ngali mit rirk E.-nthurr

G.Y. 3sg(nom) driver 2du:excl work DO¹³:npst E.-erg^

‘G. Y. is the driver. We two work, [me and] E.’
The referent of the ergative-marked NP (E.) is a perfectly plausible subject according to world knowledge (i.e. highly animate, and likely to be engaged in the activity of working), but is ‘unexpected’ (in McGregor’s sense) due to his not having been mentioned previously. Accordingly, he is introduced by a NP placed within an inclusory construction in order to clarify that the speaker is not referring (by means of the second person dual exclusive pronoun, ngali) to himself and the protagonist of the preceding clause (G. Y.). The ergative case morph can therefore be viewed as emphatically marking the noun phrase as part of the subject, in contradistinction to the subject of the preceding clause.
As noted above, the ergative morpheme is not used to express the comitative case (as found in some other Australian languages, e.g. Wambaya (Nordlinger 1998)). The comitative relation that obtains between the speaker and ‘E.’ in (12) is expressed by the inclusory construction (i.e. the apposition of the non-singular pronoun ngali ‘we two, exclusive’ and a noun phrase denoting a subset thereof, i.e. E.). If the speaker wished to highlight the comitative relation, this could be done by attaching ‘comitative’ -kak to E., in place of the ergative morpheme. The ergative suffix on the subset noun phrase is unrelated to (and uncalled for by) the inclusory construction in this context.
3.2 Unmarked transitive subject

Just as the ergative morph may be affixed to the subject of an intransitive clause, ergative marking may be omitted from transitive clauses in certain pragmatic contexts. Once again, the key to predicting where the ergative morpheme may be omitted lies with the degree to which the identity of the subject corresponds to the addressee’s expectations; expectations built by the interaction of discourse context and world knowledge. Turning first to the role of world knowledge in constructing and decoding argument structure, a subject is more likely to be ‘expected’ where it is: (a) high in animacy (and significantly higher in animacy than its object); and (b) engaged in an activity with which an entity of its type is likely to be engaged. These two criteria are illustrated by (9) and (13):

(9') minh patp piinth.kat waawath

animal hawk scrap(acc) rdp:search

‘hawks fossick for scraps’
(13) pam peln mong werngr ulp thunpm

man(#erg) 3pl(erg) many boomerang(acc) the throw:p.ipfv

‘many men threw the boomerang’
It is clear that hawks are more animate than scraps and men are more animate than boomerangs. It is also entirely consistent with our knowledge of the world for hawks to be engaged in foraging and men to be engaged in throwing boomerangs (particularly in contrast with scraps foraging or boomerangs throwing). Additionally, the hawks and men respectively appear as actors in the preceding discourse. Accordingly, the association of minh patp ‘hawk’ (in (9)) and pam mong ‘many men’ (in (13)) with the grammatical function of subject is straightforward even in the absence of overt case marking. The SOV constituent order in both of these examples may also be significant. Further investigation (and the course of time) may reveal an increasing reliance on word order for the interpretation of grammatical functions in the absence of obligatory case-marking. It should be noted, however, that examples (9) and (13) were uttered by an elder of seventy-four, whose first language is (a conservative variety of) Kuuk Thaayorre, and whose speech generally exhibits flexible constituent order.
The omission of ergative marking due to discourse context is further exemplified by the narrative excerpt (14). The speaker here is describing his trip to Darwin shortly after cyclone Tracy. He tells of a taxi driver he met who had been injured in the cyclone while trying to lead a group of tourists to safety:
(14.1) glass-n ke’e-rr=unh.

glass-erg spear-p.pfv=3sgacc

‘he was cut by glass’
(14.2) Taxi driver, pam ngotn.

taxi driver man black

‘the taxi driver, a black man’
(14.3) Taxi driver glass-n ke’e-rr.

Taxi driver(acc) glass-erg spear-p.pfv

‘the taxi driver got cut by glass’

(14.4) Glass-n ulp aka ke’e-rr

glass-erg the here spear-p.pfv

‘that glass cut [him] here [points to arm]
(14.5) nhul taxi driver pam guide-m rirk-m

3sg(erg) taxi driver(#erg) man(acc) guide-vbr do-p.ipfv

‘that taxi driver was guiding people’
If the clause (14.5) were considered in isolation, it would be impossible to justify the omission of the ergative case marker from the subject NP. There is no difference in animacy between the subject (‘taxi driver’) and object (‘people’) of the clause, and since number is not marked on nouns, either of the NPs taxi driver and pam ‘man’ / ‘people’ could be in syntactic apposition to the subject pronoun nhul ‘3sg’¹⁴. However, given discourse context, it is straightforward to associate the enduring protagonist of the previous clauses (taxi driver) with the subject function.
In example (15), the same participant group is represented as subject in each clause, but the ergative marking present in the first two clauses is omitted in the third:
(15.1) paanth-u wanhwanhrrul yak ii theerngarr=unh

woman-erg how.many.people:erg snake there kill:p.pfv=3sgacc

‘How many women killed that snake?’
(15.2) paanth pinalam-thurr theerngarr yak ulp

woman three-erg kill:p.pfv snake(acc) the

‘three women killed the snake…’
(15.3) paanth ii peln pinalam=unh theerngarr woman(#erg) there 3pl(erg) three=3sgacc kill:p.pfv
yak ulp

snake(acc) the

‘[those] three women there killed that snake’
Here, as in examples (9) and (13), several pragmatic and semantic factors conspire to unambiguously identify each of the unmarked arguments with the relevant syntactic function. Firstly, women are higher in animacy than snakes, making it more congruent with world knowledge for the women to act upon the snake than the reverse. This factor is not particularly strong, however, as it is just as plausible for a snake to kill a woman as a woman to kill a snake (particularly in the taipan-infested territory of the Thaayorre). More critical, in this case, are the discourse context and verbal semantics. Taking first of all the discourse context, it is is clear that the preceding discourse (in which the women do receive ergative marking) establishes the respective roles of the participants sufficiently clearly for case marking to be omitted in the final clause (which may be regarded as a summary of the preceding clauses) without any resultant ambiguity. Additionally, the verb employed - theerng ‘kill’ - more precisely describes the act of killing by striking, and as such would not be used with a limbless creature (such as a snake¹⁵) as subject. Thus verbal semantics further disambiguates the mapping of participants to syntactic functions.
Finally, excerpt (16) from the Frog Story demonstrates the interplay of world knowledge and discourse context over a series of clauses:
(16.1) parr_r nhul thamr puut nhaanham

child(#erg) 3sg(erg) foot boot(acc) look:rdp:npst

‘the boy looks in the boot’
(16.2) “Thatr wanthan yat?”

frog(nom) to.where go:p.pfv

‘[He thinks:] “where has the frog gone?”’

(16.3) nhul thatr ngaathirr waawath-r

3sg(erg) frog(acc) still search:rdp-npst

‘he’s still looking for the frog’
(16.4) Kuta-ku nhul glass nhaanham

dog-erg 3sg(erg) glass(acc) look:rdp:npst

‘[now] the dog looks in the jar’

In (16.1), the human subject (already introduced in the preceding text) acts upon an inanimate object. Here world knowledge (strengthened by the interclausal topicality of the boy) is more than ample to establish the syntactic function of arguments. In clauses (16.2) (which implicitly reports thought) and (16.3), the boy remains the (implied or articulated) topic. In (16.4), though, attention shifts to the dog. This protagonist shift (possibly combined with the lower animacy of the dog) renders it an ‘unexpected’ subject, triggering ergative marking.