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

4. Analysis

The influence of pragmatics on the employment of the Thaayore ergative morpheme is significant in several respects. To begin with, it shows the much discussed ‘animacy hierarchy’ (attributed to, and emergent from the ideas of, Silverstein 1976¹⁶) to be played out not in Kuuk Thaayorre morphosyntax, but rather in pragmatics. In Kuuk Thaayorre morphosyntax (as described in 1.2), there is a split between pronouns (which are marked on a nominative-accusative pattern) and nouns (which follow the ergative-absolutive pattern). Within the nominal category, however, there is no division with respect to which semantic categories of noun may receive ergative marking in the appropriate syntactic context. As shown in §3, though, there is a difference in the level of markedness of subjects (which filters through to optional ergative marking), according to their level of animacy. This is exemplified by comparing (16.1) (which has an unmarked human subject) with (16.4) (in which the ergative-marked subject is a dog):

(16.1') parr_r nhul thamr puut nhaanham

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

‘the boy looks in the boot’
(16.4') Kuta-ku nhul glass nhaanham

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

‘the dog looks in the jar’

The relative animacy of subject and object also plays a role as the comparison of (17) with (18) shows:
(17) nhul parr-an kuta mi’irr

3sg(erg) child-erg dog(acc) pick.up:npst

‘the boy picks up the dog’
(18) parr_r nhul thatr mi’irr yuur-un

child(#erg) 3sg(erg) frog(acc) pick.up:npst hand-dat

‘the boy holds the frog in his hand’
While the subject of both clauses is human, in (17) parran ‘child’ receives the ergative marking its syntactic function warrants, while in (18) the marker is omitted. This can be attributed to the fact that in (18) the highly animate subject participant is acting upon a significantly less animate entity (the frog), which has not, throughout the preceding text, been accorded much in the way of personality. In (17), by contrast, the boy is acting upon an only slightly less animate being (the dog), who has been a major protagonist (alongside the boy) throughout the text¹⁷. The pragmatic employment of ergative marking thus reveals a cline of animate - inanimate nouns. This cline is also manifest in existential and ascriptive constructions, as described in Gaby (2006). In that case, too, dogs and ‘social animals’ form an intermediate category between humans, animate non-humans and inanimate entities, in terms of the posture and movement verbs with which they can combine. Silverstein (1976)¹⁸ similarly proposes a ‘global’ case-marking system for Dalabon, in which the suffixation of -yi to a transitive subject NP purportedly depends on “the Agent being below or at the same feature-level as the Patient” (Silverstein 1976:129).
The distribution of ergative marking cannot be attributed to animacy alone (as per semantically-based case marking), however. Example (19) (taken from the very beginning of the Frog Story), for instance, shows the expected ergative marking of the subject of a transitive clause with exactly the same configuration of participants as in (18) (i.e. a boy acting upon a frog):
(19.1) Thatr nhul glass-ak nhiinhin

frog(nom) 3sg(nom) glass-dat sit:rdp:npst

‘A frog is sitting in a jar’

(19.2) Parr-an pul kuta-ku nhaanham nhunh thatr

boy-erg 3du(erg) dog-erg see:p.ipfv 3sgacc frog(acc)

‘A boy and a dog are looking at the frog’

The opening clause introduces the frog, suggesting to the addressee that this will be the main protagonist. The enduring protagonists of the story (i.e. the boy and the dog) are at this point a truly ‘unexpected’ subject of the second clause, this being their first mention. It is hardly surprising, then, that they each receive (syntactically appropriate) ergative marking, highlighting their role as agents of the event. The fact that the presence or absence of the ergative morph cannot be predicted by the relative animacy of participants alone, reinforces the argument that ergative marking is (partially) pragmatically conditioned, rather than simply based on semantics.
The tendency for the ergative morpheme to be omitted where the subject is highly animate runs counter to what we might predict from Hopper and Thompson’s (1980) theory of transitivity. According to their parameters of volitionality and agency (parameters E and H respectively), we would expect that clauses with human subjects would rate more highly on the transitivity scale (and therefore be more likely to include ergative marking) than those in which the subject is lower in animacy. Instead, we find the reverse. This can be accounted for in two ways. Firstly, it has often been suggested that the primary function of ergative markers is to distinguish the two arguments of a transitive clause (cf. Comrie 1989, Heath 1976, Wierzbicka 1981). The use of Kuuk Thaayorre ergative marking in intransitive clauses (discussed in §3.1) can be seen as distinguishing the subject not from an object (as there is none), but from other potential subjects. As such, the ergative morpheme could be analysed as marking the syntactic function of subject rather than the syntactic case of transitive subject (making the label ‘ergative’ somewhat misapplied). This possibility is explored further in §5.
DuBois’s (1987, 2003) account of ‘preferred argument structure’ offers a second, alternative analysis of how ergative case morphemes could come to be associated with marked subjects more generally. DuBois finds that speakers tend to avoid representing agents lexically, and tend not to introduce new participants as agents. Accordingly, where an agent is overtly realised, it must be pragmatically marked for some reason (and hence likely to be morphologically marked). This explains how the majority of lexical agents (i.e. transitive subjects) could come to be ergative-marked for pragmatic reasons, rather than purely because of their grammatical function (cf. Gaby under review for discussion of the diachronic origins of the Kuuk Thaayorre ergative morphs).
5. Theoretical implications

As raised in §4, several alternative analyses of the function(s) of the ergative morpheme suggest themselves. First of all, it would be possible to propose two homophonous (sets of) morphs, one of which is a straightforward ergative case marker, the other being a discourse marker of ‘unexpected subjects’. This analysis is attractive because it allows us to retain a tripartite case system, while also allowing the distribution of the ‘unexpected subject’ marker to be conditioned wholly by pragmatics. The downfall of this analysis, though, is the fact that the ergative morph may be omitted from transitive subjects. If there is a fully grammaticised ergative case morpheme, this should be present in all syntactically ergative contexts. If the ‘unexpected subject’ marker is an entirely separate morpheme, we would be required to independently explain the deletion of the ergative morpheme in contexts where a transitive subject is expected. Such an analysis is clearly not parsimonious. Also problematic for this analysis, is the extreme formal irregularity of the morphs in question, shared exactly by both putative functions. We would not normally expect such complete isomorphism between distinct morphemes in the context of such irregularity.

Alternatively, the (hitherto labelled) ‘ergative’ morpheme could be analysed as monosemous, encoding simply ‘unexpected subject’. This ‘unexpected subject’ function could be framed as a marked nominative case (versus the unmarked nominative form used in pragmatically neutral or expected contexts)¹⁹. Under such an analysis, the morpheme has some characteristics of a case marker (in that it marks a dependent noun for the relationship it bears to the head – cf. Blake 1994²⁰). However, it is unlike case marking in that it is only applied where there is a need to clarify or emphasise the syntactic function of the subject argument, rather than being obligatory in particular syntactic contexts. The statistical correlation between ergative marking and being the subject of a transitive clause (rather than an intransitive subject) can be attributed to the fact that there tends to be greater potential for ambiguity in clauses with multiple arguments.
To analyse this morpheme as a ‘marked nominative’ case does not, however, capture the strong association between this morpheme and the subject of a decontextualised transitive clause (as opposed to a decontextualised intransitive clause). This is borne out by the near-complete correlation between transitive subjects and ergative-marking (and conversely intransitive subjects and absence of the ergative) in decontextualized elicitation. Moreover, all consultants rejected as ungrammatical out-of-context intransitive clauses with ergative-marked subjects proposed by the author. The vast majority of proposed transitive clauses with unmarked subject NPs were also corrected to contain the ergative marker. A final problem with the ‘marked nominative’ analysis is the fact that the distribution of, and motivating factors for, ergative marking differs significantly between transitive and intransitive clauses.
I therefore conclude that the ergative morpheme is indeed associated with the syntactic ergative case, in spite of its distribution being influenced by pragmatics. As already stated, its function of marking a dependent NP for its relation to the head predicate establishes the morpheme’s status within the case system. The question becomes, then, how exactly to characterise the function of this case morpheme in light of its distribution. In order to tackle this question, let us return to the arguments put forward by Mel’čuk (1979), Goddard (1982) and Blake (1994) in distinguishing (morphological) case form from (syntactic) case function. Applying their analysis to Kuuk Thaayorre (ignoring, for the moment, the pragmatically-conditioned distribution of the ergative morpheme), we can see how three core syntactic cases (ergative, nominative and accusative) emerge from the comparison of pronouns and nouns, each of which display only only two case forms:

	Pronoun (‘3sg’)	Noun (‘man’)
Ergative	nhul	pamal
Nominative	nhul	pam
Accusative	nhunh	pam

Table 3. Comparison of syntactic case function and morphological case form

Despite the syncretism of ergative and nominative pronominal forms, and nominative and accusative noun forms, the distinction between each of the three syntactic cases is morphologically encoded by at least some subpart of the system and is therefore analytically maintained for the case system as a whole.
If we then extend this analysis to differentiate between the three relevant pragmatic contexts (expected subject, pragmatically neutral, unexpected subject) in which subject arguments are verbalised, it remains equally clear that the ergative, nominative and accusative cases are a syntactic reality in Kuuk Thaayorre. Note that we are now concerned only with noun forms:

	Expected subject	Pragmatically neutral	Unexpected subject
Ergative	pam	pamal	pamal
Nominative	pam	pam	pamal
Accusative	pam	pam	pam

Table 4. The interaction between syntactic case and pragmatics in determining case form

This time, however, there is syncretism between all three core cases where the subject is expected (i.e. where the syntactic function of a NP is predictable it is formally unmarked, regardless of case), between nominative and accusative cases in neutral contexts, and between the nominative and ergative cases where the subject is unexpected. Again, each of the three core syntactic cases is somewhere differentiated from the other two. The distribution of morphological forms, however, is conditioned by the interaction between syntactically-assigned case (varied in table 4 by row) and pragmatics (varied by column). This should not be a particularly shocking proposition, as Australianist linguists have long accepted that (semantic and/or pragmatic) features of animacy can condition the distribution of case morphs (since Silverstein 1976 and Heath 1976). We are familiar, too, with languages in which case markers simultaneously encode non-case information (e.g. number in Latin, or modal information in Kayardild (Evans 1995)²¹). Analogously, then, the Kuuk Thaayorre ergative case morpheme explicitly encodes syntactic case, while its (non-)employment signals the degree of pragmatic markedness²².
The homophony of core case forms in Kuuk Thaayorre (as presented in table 4) does not increase the potential for mis-mapping grammatical relations to argument NPs for two reasons. Firstly, because a NP in nominative case – isomorphous with the accusative case in neutral and expected contexts and with the ergative case in unexpected contexts – is never copresent with a transitive subject or direct object in the same clause. Secondly, the presence of two unmarked NPs (in nominative and accusative case respectively) in transitive clauses with an expected subject does not give rise to ambiguity since the pragmatically expected context is, by definition, one in which the mapping of grammatical relations to argument NPs should be obvious.
In summary, then, I argue that Kuuk Thaayorre has a tripartite core syntactic case system (distinguishing ergative, nominative and accusative cases). The morphological expression of these cases is determined jointly by: (a) the underlying syntactic case of the argument; (b) the word class of the nominal form (i.e. whether it is a pronoun or a noun); and (c), if it is a noun, the pragmatic status of the argument (i.e. how easily the addressee is expected to match it to the relevant syntactic function).
6. Pragmatic case marking in other languages

There are a growing number of languages in which pragmatic factors have been demonstrated to co-condition ergative inflection. This section surveys the most detailed recent analyses of such systems, highlighting both similarities and differences between the distribution, functions and inferred diachronic development of optional ergativity cross-linguistically.

From his detailed analysis of a large corpus of Gooniyandi texts, McGregor (1998) identifies several etic characteristics of clauses in which the agent is ergatively marked. Of particular relevance to the data described in §3, he finds that ergative marking may be prompted by an “unforeseen Agent given narrative context” or when the “Agent contrasts with other (potential) Agents” (1998:503). McGregor also finds that the level of agentivity of the subject participant in large part determines the presence or omission of ergative marking, though his definition of agentivity as "goal-directed" does not seem to be a relevant parameter for Kuuk Thaayorre pragmatic case.
Warrwa also permits the omission of the ergative postposition in transitive clauses where the subject is “both expected and low in agentivity” (McGregor 2006: 393). It possesses two ‘ordinary ergative markers’ as well as a ‘focal ergative marker’ that signals an unexpected subject (in transitive, intransitive, verbless and other clauses). There is thus a pragmatic contrast between case morphemes with the same syntactic function. McGregor argues that both the syntactic ‘ergative’ case and pragmatic ‘focus’ are inherent to the meaning of the Warrwa focal ergative postposition, an assertion that might also be made of the Kuuk Thaayorre ergative morpheme. The Warrwa situation is especially complex, however, as there is the contrast between an elided subject NP, an unmarked subject NP, a subject NP marked with an ordinary ergative morph as well as a subject marked by the focal ergative morph. Nevertheless, there are many commonalities in the conditions under which ergative marking may be employed or omitted in Warrwa and Kuuk Thaayorre, particularly with respect to the marked status of subjects that are ‘unexpected’ due to their not being the main protagonist of the relevant episode. In Warrwa, though, McGregor (2006: 410) finds that ‘potent’ agents are more likely to receive focal ergative marking (with human agents being more potent than lower animate or inanimate agents), while in Kuuk Thaayorre (and Gurindji Kriol, see below) a subject participant that rates lowly on the animacy scale is more likely to be ergative-marked.
The Jingulu focus markers (a subset of which are homophonous with, and most likely derived from, the ergative case markers) can be attached to oblique as well as core arguments (rather than just the subject, as in Kuuk Thaayorre) and may be suffixed to noun stems already inflected for case (Pensalfini 1999:6). Significantly, the Jingulu focus markers operate in parallel to the case system, with ergative case marking remaining obligatory for transitive subjects. This is very different to the Kuuk Thaayorre case, in which pragmatic considerations are embedded in the core case-marking system itself. The focal function of Jingulu case-markers is analysed as a recent innovation, occuring as part of the process of language obsolescence. Pensalfini (1999:26-27) argues that, in pro-drop languages such as Jingulu, the presence of noun phrases is pragmatically marked and "overt nominal arguments are therefore generally associated with focus". He proposes that more recent learners of Jingulu might have initially reanalysed the case system as nominative-accusative (by analogy with English), and ascribed a focus-marking function to the ergative morpheme. This phase of acquisition would then leave an imprint on the fully proficient speakers’ variety of Jingulu.
Like Kuuk Thaayorre, Warrwa and Gooniyandi, the emergent mixed languages Gurindji Kriol and Light Warlpiri allow the optional omission of ergative-marking in transitive clauses (Meakins & O’Shannessy 2004). In Gurindji Kriol, ergative morphology is also found in some intransitive clauses, and is associated with the pragmatic functions of contrastive focus and topic marking. The rigidification of word order in Gurindji Kriol seems to have contributed to the reanalysis of the ergative morpheme (Meakins & O’Shannessy 2004) since the semiotic burden of the ergative morpheme is lessened by the coding of grammatical relations in the ordering of constituents. Such an explanation is not available for Kuuk Thaayorre, as the ergative morpheme is used pragmatically even in the conservative speech of elders, for whom constituent order remains flexible. A final point of interest is the increased tendency in both Gurindji Kriol and light Warlpiri (as in Kuuk Thaayorre) for less animate participants to receive ergative marking (Meakins & O’Shannessy 2004).
While Pensalfini (1999) and Meakins & O’Shannessy (2004) attribute the association of syntactic case markers with discourse function to the forces of language contact, this is not so clearly the case with Kuuk Thaayorre. Pormpuraaw (home to the vast majority of Kuuk Thaayorre speakers) was established as an Anglican mission in 1938. Prior to this, almost all of the Thaayorre led a traditional life with little or no contact with English speakers. The pragmatic usage of the ergative morpheme, however, is found in old texts and throughout the speech of Thaayorre people of all ages, many of whom were raised monolingual. Moreover, Gaby (under review) proposes that the function of marking pragmatic focus preceded the syntactic function of marking a transitive subject for at least some of the Kuuk Thaayorre ergative allomorphs.
7. Conclusion

This paper has shown that if the preceding discourse and/or interlocutors’ world knowledge do not lead the addressee to expect a particular participant to be represented as subject, then the speaker is likely to mark this subject argument as ‘unexpected’ by affixing the ergative morpheme. As shown in §3.1, this may lead to the ergative marking of subjects in intransitive clauses. Conversely, where the previous discourse and/or world knowledge leave the addressee in no doubt as to the assignation of grammatical relations to arguments, ergative marking may be omitted. In both instances, it is clear that the employment of case morphology is motivated not by syntax, but by pragmatics. §5 considered the theoretical implications of the data presented herein, concluding that Kuuk Thaayorre should be analysed as a language with a syntactic ergative case (alongside the nominative and accusative syntactic cases), but in which the distribution of case morphology is co-conditioned by pragmatics.

It has been suggested by Pensalfini (1999) and Meakins & O’Shannessy (2004) that discourse functions of case markers arise from situations of language obsolescence in the context of contact with a dominant language. The pragmatic use of the Kuuk Thaayorre ergative morpheme, however, appears to have originated at a stage of the language’s relative fortitude. The fate of this morpheme, under the pressure of enduring contact with English, remains to be seen.

Abbreviations

1	first person
2	second person
3	third person
abl	ablative case
acc	accusative case
dat	dative case
du	dual number
erg	ergative case
erg^	ergative inflection of an ‘unexpected’ intransitive subject
#erg	omission of ergative inflection of an ‘expected’ transitive subject
Excl	exclusive of addressee
foc	Focus
imp	imperative mood
nom	nominative case
npst	non-past tense
p.ipfv	past imperfective tense/aspect
p.pfv	past perfective tense/aspect
perm	Permissive
pl	plural number
rdp	Reduplication
sg	singular number
vbr	Verbaliser
the	addressee-proximal demonstrative
that	distal demonstrative