Intersubjectivity and engagement in

Following Evans et al. (2018a, 2018b), I use “engagement” to refer to grammatical encoding of the relative accessibility of an entity or state of affairs to the speaker and addressee. I refer to what is thereby encoded as the “engagement function”. How neatly does that function map on to grammatical categories of particular languages? Here I address that question with respect to the Papuan language Ku Waru, focusing on spatial and epistemic demonstratives, and definiteness and indefinite marking. I show that forms within each of those word/morpheme classes do serve engagement functions, but in cross-cutting and partial ways. I show how the engagement function is also achieved through poetic parallelism, prosody, gaze direction and other aspects of bodily comportment. In the examples considered, the engagement function is realised through interaction between those extra-linguistic features and the grammatical ones. The main thing that is added by grammatical engagement marking is an explicit signalling of the intersubjective accord that has been achieved on other bases. I hypothesize that that is true of engagement overall, and conclude by suggesting some ways to test that hypothesis and to advance the understanding of engagement more generally.


Introduction
The publication of San Roque's (2018a, 2018b) study of what they call engagement is a landmark event, for the ways in which it advances the comparative study of linguistic aspects of one the most distinctive of human capacities: intersubjectivity. Notwithstanding their shorthand description of engagement as "grammaticalised intersubjectivity", what Evans, Bergqvist and San Roque (hereafter EBS) focus on is one dimension of intersubjectivity, namely "the relative accessibility of an entity or state of affairs to the speaker and addressee" (Evans et al. 2018a:113). Hereafter in this article, in order to refer specifically to that dimension of intersubjectivity, and to distinguish it from the posited grammatical category by which it is encoded, I will refer the former as the "engagement function" and will use the term "engagement", as EBS do, to refer to the grammaticalized encoding of that function.
In my view one of the most important questions raised by EBS's work is, how neatly does the engagement function map on to grammatical categories of particular languages? Their opening discussion of Andoke auxiliaries is impressive for the extent to which the categories they signal do seem to show such a mapping, but EBS (2018a:119) acknowledge that engagement "borders on many more familiar linguistic categories" and "much of the time actual languages run some of these dimensions together". So how common are pleasingly perspicuous Andoke-like engagement systems versus ones with less clear borders, and more running together of relevant dimensions? EBS's study provides much valuable evidence for addressing that question, but it is inevitably subordinated to their emphasis on establishing engagement as a viable category for comparative-typological study.
In this article, I will provide a complementary perspective by focusing on a range of engagement-related categories from a single language: Ku Waru. Ku Waru is spoken in the Western Highlands Province of Papua New Guinea. It belongs to the Trans-New Guinea family of Papuan languages, and, more immediately, to a dialect continuum within the Chimbu-Wahgi branch of that family that includes what Ethnologue classifies as four distinct languages: Melpa, Mbo-Ung, Imbonggu, and Umbu-Ungu. In those terms, Ku Waru belongs to the Mbo-Ung language (ISO code mux).
The formal elements within Ku Waru that I will focus on are: 1) spatial demonstratives; 2) epistemic demonstratives; 3) a pair of postpositions that are used for marking definiteness and indefiniteness. Like the "level shifting" Abui demonstratives and Marind verbal prefixes discussed by EBS (2018b:159-162), Ku Waru epistemic demonstratives and definite marking operate both at the level of the noun phrase and the clause, providing further evidence of the way in which the grammar of engagement can bridge those two levels. I will show that, of all the formal devices listed above, it is definite and indefinite marking that map most closely onto the engagement function, suggesting that that area of grammar in general may be an especially fertile field for the exploration of engagement, not just in "in article systems in western European languages" (EBS 2018a:117), but also in the many languages around the world in which categories of definiteness and/or indefiniteness are attested.
Then, focusing on videoed interactions between Ku Waru children and adults, I will show how the engagement function is realized, not by grammatical devices alone but by a complex interplay between them and other aspects of interaction including gaze direction and facial expressions; intonation and prosody; and patterns of parallelism of the kind that Dubois (2014) has treated under the rubrics of "resonance" and "dialogic syntax". Finally I will return to the question of degrees of systematicity in how the engagement function is realised and offer some conclusions about it in light of the Ku Waru evidence. Before introducing the relevant grammatical devices in Ku Waru I will first provide some background details concerning its syntax.

Background details of Ku Waru grammar
In some kinds of NP in Ku Waru the head noun comes last, with modifying elements before it. Those include possessive NPs such as yunu=nga kolya 'her place' in (2) and na=nga ung=uyl 'my speech' in (23), and relative clauses such as in nanu ti-d ku 'the money that I have given' in (25) and nunu lyi-lym yi=yl 'the man who takes you' in (31). In other kinds of NPs, including ones with modifying adjectives, the head noun comes first. Examples are wik autiyl 'many weeks' in (33) and kang kumulaya 'first-born son' in (34).
Like many other Trans-New Guinea Papuan languages, Ku Waru is a rigorously verb-final one: the main verb in the clause always comes last. Final verbs have portmanteau suffixes encoding tense, mood and aspect, and the person and number of their subject. There are systems of verb serialization and clause chaining, both of which make use of specially marked "medial" or "non-final" verbs. Those verbs do not inflect independently for TAM, but share their TAM value with the final verb with which they are in construction. Non-final verbs do inflect for person and number -albeit less fully than final verbs -and for whether their subject is the same or different subject from that of the final verb in the clause (switchreference). In bivalent clauses the word order is usually SOV but sometimes (about 10% of the time) it is OSV. There is an "optional" ergative postposition =n(i) that is used on most but not all subject NPs of bivalent clauses (see Rumsey 2010:1663-1667 for further details). These various aspects of clause level syntax in Ku Waru are illustrated by (1) -(4). For an example of switch-reference marking see (35).

Spatial deixis
Spatial deixis in Ku Waru is organized in terms of two dimensions, horizontal and vertical, and three distinctions within each. Within the horizontal dimension the distinctions pertain to the relative distance between the speaker and/or addressee and the referent. Within the vertical dimension they locate the referent at a higher elevation than the speaker/or addressee, lower, or at the same level. The basic terms in the system are the ones that pertain to the horizontal dimension. They can occur by themselves when the referent is at roughly the same elevation as the speaker and/or addressee. The terms are: ya near to speaker (and possibly, but not necessarily, near to the addressee) i near to both speaker and addressee ne at a middle distance from the speaker and addressee adi near to the addressee but not to the speaker wi far from both speaker and addressee When the referent is at a higher or lower elevation than the speaker and/or addressee, that is obligatorily indicated by the addition of the words ola 'up' or manya 'down' after the horizontal deictic. Examples (all of them elicited in a controlled setting) are (5) -(11).
(5) me=yl ya manya le-lym taro=def here down be.in.place-hab:3sg 'The taro is here [on the ground in front of the speaker and addressee, who are standing next to each other].' (6) tauwu=yl ya manya le-lym banana=def here down be.in.place-hab:3sg 'The banana is here [on the ground in front of the speaker, who is standing, with the addressee about 7 meters off to the side].' (7) me=yl i manya le-lym taro=def this.here down be.in.place-hab:3sg 'The taro is here [on the ground in front of the speaker and addressee, who are standing next to each other as in (5)].' (8) me=yl adi manya le-lym taro=def there down be.in.place-hab:3sg 'The taro is there near you [said by a man who is standing about five meters from the taro, which is on the ground beside the addressee, who is standing].' (9) me=yl ne manya le-lym taro=def over.there down be.in.place-hab:3sg 'The taro is over there [said by a woman who is standing at the same distance from the taro as in (7), but with the addressee standing beside her].' (10) aussik=ayl ne manya angalyi-lym hospital=def over.there down be/stand-hab:3sg 'The hospital is over there [visible in the distance about two kilometers from where the speaker and addressee are standing, looking down over the valley in which it is located, with a mountain ridge visible behind it].' (11) tawa=yl wi ola mol-lym tower=def there.distant up be/stay-hab:3sg 'The tower is up over there [visible in the distance about ten kilometers from where the speaker and addressee are standing, looking up to the hill on which the tower is located.]' Examples (5) and (6) illustrate the point that the use of ya does not depend on the proximity of the referent to the addressee, only to the speaker. Example (7) shows the use of i in the only kind of context where it is used, i. e., when the referent is close to both the speaker and addressee. Unlike ya, i is not used when the addressee is significantly further away from the referent, as in (6). Conversely, neither i nor ne are used when the speaker is significantly further away from the referent than the addressee, as in (8). In that context only adi is used. Based on these considerations we can see that, in EBS's terms, i, ne, wi and adi qualify as engagement markers, whereas ya does not. In terms of EBS's cross-cutting dimensions of referent accessibility to speaker and addressee, i, ne and wi are +speaker, +addressee; and adi is -speaker, +addressee.
Before leaving the subject of spatial deixis it is relevant to point out that the distance values within the system are not absolute ones, but rather, are always calibrated in relative terms within a presupposed frame of reference or "deictic field" (Buehler 1990). For example, the values of i, adi, ya and ne in (5) -(9) were calibrated within a framework that was set up by my having placed the banana and taro on the ground, along with an orange, on level ground, in a row, approximately 7 meters apart from each other, and then positioning Ku Waru two speakers in on various positions relative to them. In that context, when the objects of our attention were viewed from behind the end of the row, the values of i/ya, ne and wi were in effect mapped on to a tacitly understand deictic field that was limited to the roughly 16 x 3 meter space of interaction around those objects, so that 14 meters from the speaker counted as "distant" and 7 meters away as "middle distance". By contrast, in the context of (10) and (11), where we were looking out onto a valley from a high place on one edge of it, the entire valley become the relevant deictic field, so that the 2 kilometers to the hospital counted as middle distance relative to full 10 kilometer distance to the high ground on other side of the valley where the tower was located.
The context of (5) -(9) was an experimental one as described above, but the same pattern was evident a few days later when these same two Ku Waru speakers and I were standing on a boardwalk at the edge of a pond counting the turtles in it. The far edge of the pond was about 10 meters away. Unprompted by any questions from me, the speakers referred to the turtle right in front of us as i manya, to one at the far edge as wi manya, and to one in the middle as ne manya. In this case the relevant deictic field was understood to be the pond, with the use of the deictic terms calibrated accordingly, at approximately the same scale as for (5) -(9).
The relativity of deictic fields is a well-known phenomenon that has been extensively explored in the literature (see for example Hanks 1990Hanks , 2005 and references therein). What I want to point out here in connection with the Ku Waru case, and the concept of "engagement", is that questions of relative accessibility of the referent to speaker and addressee are themselves relative to a more basic framework of shared understandings of the ground of interaction -in this case the relevant deictic fields within which the accessibility is distributed.

Epistemic demonstratives
The Ku Waru system for what we call "epistemic qualification" makes use of three demonstratives, each of which is used to characterize referents or situations with respect to the interlocutors' shared knowledge of them. Two of those demonstratives, i and adi, are identical in form to ones that are included among the deictic expressions discussed above, in related senses. The third term kani, is unique to the epistemic series. The formal basis on which the epistemic demonstratives may be defined as a different word class from the spatial demonstratives, notwithstanding the overlap with respect to two of their members (i and adi) is that it is only the epistemic demonstratives that can appear after a final verb, as discussed in §2. All three epistemic demonstratives may occur with a wide range of nominal suffixes and postpositions, including case-marking ones as in (17) and (19), the number markers =ma plural and =sil dual, and the definite marker as described in §3.3. While the definite marker takes the expected form =yl when occurring with adi and kani, with i there is a root-specific allomorph =lyi, yielding ilyi.
i characterizes the referent as immediately present to both the speaker and the addressee. An example is (12), which comes from a speech that was made at a public event at which compensation was being paid by one group to another for injuries they sustained during the course of a tribal war when coming to the aid of the group that is presenting it (see Merlan and Rumsey 1991:34-156 for more background). The speaker is chastising a particular group of men from his tribe for not contributing to the payment2.  In both of these examples the referent of i is immediately present and perceptible. In such contexts the use of i as an epistemic demonstrative is conterminous with its use as a deictic demonstrative. Often the referents it points to are visible, but (13) shows that they can be audible instead. Indeed, one of the most common uses of i is as what we call a resumptive pronoun, pointing to something that has just been said or quoted.
(14) eni=ni pu-k disisin te-ang i nyi-kimul 2/3pl=erg go-nf:2/3pl decision do-opt:2/3pl ip say-1pl:ppr ' "You go and make a decision" we are saying.' A more idiomatic English translation of this sentence would be 'We are saying "You go and make a decision"'. But in keeping with Ku Waru's verb-final syntax, the framing verb nyi-'say' always comes after the reported locution that it frames. When there is a resumptive pronoun it comes after the framed utterance and before the framing verbs as in (14). For further details see Rumsey (2010).
kani, which we gloss with abk 'absent but known', is used in reference to things that both speaker and addressee are presumed to be familiar with, but which are not immediately present to them. Examples are (15) and (16) Example (17) illustrates the use of kani with a nominal suffix -the definite marker (to be discussed below) -followed by a nominal case-marking postposition, the genitive, which in this instance has the sense of "about" or "for". The referent of kani is the "fight plan" which is purportedly known to be under discussion between the two tribes, but not at the event at which this speech was being given (which was too public for such discussion).
adi in its epistemic function is used in reference to things which are not immediately present to the speaker or addressee and are unknown or relatively unfamiliar to the speaker. In this function we gloss it as "abu", which stands for "absent and unknown". Examples are (18)  akuna=ko te-amiyl hear-pfv:2/3pl=perhaps that=add do-opt:1pl 'Or if you think it should be heard somewhere elsewhere else, then let's do it that way.' As is often the case with adi, in (18) it has scope over the clause it occurs in. In (19) and (20) it has nominal referents: the person to whom the addressees are directed to speak in (19) and the proposed alternative venue for the court case in (20). At both the clause level and NP level the epistemic demonstrative adi entails a relatively low degree of access to the referent or predicated situation on the part of the speaker. Correspondingly, as discussed in §3.1 when adi functions as a spatial deictic it entails greater proximity of the referent to the addressee than to the speaker. But in the epistemic uses of adi as exemplified by (18) -(20) there is not necessarily any corresponding asymmetry of access. It is not the case, for example, that the addressees in (18) are presumed to know more about the burial than does the speaker, or that the addressees in (20) are presumed to know more about the alternative venue for the court, which is in any case purely hypothetical. On the contrary, although it is not (yet) fully grammaticalized, with adi there is generally a default presumption that speaker and addressee share the same low degree of accessiblity to the referent or predicated event, was clear from the context in (18) for example.
In conclusion regarding the Ku Waru epistemic demonstratives, in terms of EBS's (2018a, 2018b) typological axis of intersubjective distribution, the three of them comprise a motley crew. The third one, adi, has only an incipient or presumptive engagement value: -speaker, -addressee. The other two, -i and kani, are full-fledged engagement markers, indicating accessibility to both the speaker and the addressee, but between the two of them they instantiate only one of the four types within EBS's typology: +speaker, +addressee. Furthermore, within that one type, the choice between the two forms (i vs kani) marks a difference not in degree of accessibility per se, but in nature of the access, as immediate and perceptual or based on prior shared experience.

Definite and indefinite marking
Before discussing the grammar of definite and indefinite marking in Ku Waru it is relevant to note that EBS (2018a:117) have singled out the extensive literature on "definiteness contrasts expressed in article systems in western European languages" as one of the few exceptions to their generalization that there has been a "lack of attention paid to grammaticalised epistemic relations between speaker and addressee". Given that that is the case, in order to advance the study of engagement, one is led to ask, how widespread are "definiteness contrasts" in the languages of the world? While these systems have indeed been much studied in European languages, the notions of definiteness and indefiniteness have also been applied to many others. For example in his article on definite articles in the World Atlas of Language Structures (WALS) Matthew Dryer (2013a) surveys a total of 620 widely scattered languages and finds what he classifies as a definite article in well over half of them. There is of course plenty of room for disagreement in how to define such a category in cross linguistic terms, and Dryer's definition it has been criticized as being too expansive (Davis et al. 2014). For present purposes however his definition is a useful one for the way in which it focuses on intersubjective aspects of definiteness and indefiniteness, both in the Western European cases and in the Ku Waru one that I will be considering here. Dryer (2013a) defines "definite article" as "a morpheme which accompanies nouns and which codes definiteness or specificity". As to what he means by "codes definiteness", in keeping with the large literature on this subject (e.g. Lyons 1977, Hawkins 1978, Clark and Marshall 1981, Dryer says There are, broadly speaking, two functions associated with definite articles. One of these is an anaphoric function, to refer back to something mentioned in the preceding discourse. The other is a nonanaphoric function, to refer to something not mentioned in the preceding discourse but whose existence is something that the speaker assumes is known to the hearer. This assumed knowledge may be based on general knowledge (as in the sun) or it may be based on inferences that the hearer can make in context (for example, inferring from mention of a house that the house has a door, thus making it possible to use a definite article in referring to the door of the house) (Dryer 2013a).
While I think Dryer's use of the term "articles" is this context is infelicitous, I agree with how he characterises their functions, which I take to be consistent with a broad range of literature on definiteness. Both of the functions he points to involve engagement. This is obviously true of the non-anaphoric function because, as Dryer says, it entails assumptions on the speaker's part about what is known to the hearer. But the anaphoric function also involves engagement, since the use of a definite article entails not only previous mention, but also an assumption on the speaker's part that that previous mention has registered with her addressee and can be drawn upon by her in order to interpret the anaphoric reference. In other words, both the anaphoric -or more broadly, endophoric -and non-endophoric uses of definite marking involve what Clark and Marshall (1981) call "co-presence" of speaker and addressee with respect to the understood referent.
Consistent with other literature on this topic (e.g., Lyons 1977, Hawkins 1978 Dryer acknowledges that there is overlap in many languages between definite marking and demonstrative reference, with the same morphemes being used for both functions, sometimes with prosodic differences between them. He uses the word "article" broadly to include both words and affixes. According to this definition, in Dryer's (2013a) survey of 620 languages he found that 377 of them have definite marking.
Dryer defines "indefinite articles" as follows: A morpheme is considered here to be an indefinite article if it accompanies a noun and signals that the noun phrase is pragmatically indefinite in the sense that it denotes something not known to the hearer, like the English word a in a dog. This includes the use of the numeral for 'one' as an indefinite article and affixes on nouns signaling indefiniteness (Dryer 2013b).
For present purposes this definition is in need of some expansion. As Dryer is fully aware of (Dryer 2014:236-237), more is involved in the meaning of the indefinite article a(n) than whether the referent of the associated NP is known to the hearer. The indefinite article a(n) is polysemous. In one of its senses it entails not only that the referent is unknown to the addressee, but that it is known the speaker. In the other sense it does not. An example of that polysemy is the sentence "A student cheated on the exam". The difference between the two senses can be seen by considering that sentence in relation to two alternative possible follow-up sentences: "His name is John" and "We"re all trying to figure out who it was" (Heusinger 2002:245). In the literature on definiteness and indefiniteness (e.g., Lyons 1977:190, Givón 1978, von Heusinger 2002) that difference has been accounted for by positing a cross-cutting dimension of "specificity", whereby the first sense of the indefinite article in the above example would be classified as +indefinite, +specific and the second as +indefinite, -specific. A similar polysemy is shown by the Ku Waru marker that we gloss as "indefinite" (indf). Following Dryer and other literature on this topic I will use "indefinite" as a cover term for both senses, and for the marker itself. When it is called for I will distinguish between the two senses as "indefinite (specific)" and "indefinite (non-specific)". It has often been claimed that western European languages are unusual with respect to the role played in them by definite and indefinite articles. Given that Dryer's findings were based on a very broad sample, of which only a small minority were from Western Europe (as can be seen on the maps in Dryer 2013a, 2013b), if we accept his findings as even roughly indicative, we have to conclude that western European languages are not so unusual with respect to their overall incidence of definite and indefinite marking systems. Rather, as pointed out by Dryer in a later publication (2014), what is unusual about Western European languages is the obligatory character of definite/indefinite marking. He says that most languages that have two articles, one restricted to definites and one restricted to indefinites, distinguish three categories formally: those marked definite, those marked indefinite, and those not marked as definite or indefinite. And in most languages with one article, say an article that is restricted to definites, this article does not occur in all definite noun phrases, so the absence of an article does not mean that the noun phrase is indefinite (Dryer 2014: e234).
That is exactly the case in Ku Waru, to which I now turn. Ku Waru has both what Merlan and I call a definite marker, and an indefinite marker, both of which are postpositions. That is, they occur on the last word of the syntactic constituent over which they have scope. As we will see further below, those environments include both NPs and clauses.
The definite marker is =yl ~ =iyl ~ =uyl ~ =ayl. Examples showing these various allomorphs are (21) - In common with the deictic and epistemic demonstratives, the main function of definite marking is to help establish reference. While deictics do that by locating the referent within a shared space of sensory perception, as discussed above, definite markers do it through endophoric linkage or shared background knowledge.
A simple example of anaphorically based definite marking can be found in (35), where there is a reference in the first clause to an arrow, el, without either definite or indefinite marking, and another reference to the same arrow in the following clause with definite-marking.
A syntactically more complex example of endophorically based definite marking, in a standard environment for such linkage for it, is (25).
1sg.emph give-pfv:1sg money=def 'This money here is the money that I myself have given.' This is an instance of the copular construction, which in Ku Waru as in many other languages, does not make use of a "be" verb, but of verbless juxtaposition, in this case of an NP subject comprising the first four words, and a relative clause, shown in square brackets, with ku=yl as its head (i.e., in a structure that could be more literally glossed as 'the I-myself-gave-it money'). Note that the first instance of ku 'money' in (25) is not marked with the definite. That happens only in the following relative clause, after its reference has been specified with the immediately preceding words nanu tid 'I myself gave' (unlike in English, where the head noun of the RC comes first, and is cataphorically marked as definite). Other examples of anaphorically based definite marking may be found in all three lines of (31). Now let us consider some examples of definite marking which is based on shared background knowledge. As in other languages with definite marking, one of the main ways in which this happens in Ku Waru is through part-whole relationships. Below are three examples of this, all which come from the same public warfare-compensation event that I referred to above. One of the defining features of such events is that they involve named tribes or clans as the relevant transacting parties, each of which is represented by one of or more leaders acting on its behalf. The main way in which the leaders do so is by giving speeches, as referred to by one of them in (26).
(26) mujika-laulku=ma langi supingi tu-ng (tribe names)=pl food quickly hit/do-2/3pl:pfv el-ung nyi-k nuimka=yl pal-um fight-talk speak-nf:2/3pl neck=def split-pfv:3sg 'The Mujika-Laulku gulped food and talked 'fight talk' [oratorical genre] until their throats were sore.' Before this remark at the compensation event there had been no previous mention of a neck or necks, but for purposes of definite marking the references to people (of the Mujika and Laulku tribes) entail the presence of necks, not only because all people have them, but because the "neck" is a considered to be particularly important instrument for Ku Waru orators, and is often used a metonym for "voice". Another example of definite marking based in shared background knowledge is (27) In this case the background knowledge that is presumed by the use of the definite marker on the tribe names Epola and Alya is that in the context of this kind of event, there will be a least one man present from each of the tribes who is one its leading 'big men' (yi nuim, literally 'neck man'). Also, note that the definite marker also occurs on the final verb in this example. This use of definite marking will be discussed below.
A third example from another speech given at the same event is (28).
(28) ul pul=uyl i=lyi na pilyi-p tudu-kur affair base=def ip=def 1sg hear/think-nf:1 not.know-ppr:1sg 'I don't know what was the source of this affair [the fight for which compensation is being paid].' Before this remark there had been no mention of an ul pul, ('source of the affair'), but in this particular situational context it is clear that the affair in question is the present compensation event and the previous events of conflict and ally-recruitment that led up to it (see Merlan and Rumsey 1991:48-52, 123-128). Especially in affairs or this kind there is a strong presumption among Ku Waru people that there is a single precipitating incident -a source or 'base' from which they conflict has grown, as plants grow from seeds or cuttings. It is the assumed background knowledge of that fact about events in general which allows the expression ul-pul to take the definite marker in a context where it has no anaphoric antecedent.
Having introduced the definite marker and exemplified it mostly in non-anaphoric uses, I will now introduce the indefinite marker and then show it interacts with the definite one.
(29) olyo=nga lku=na abu=ti mol-ym=i 1pl=gen house=loc woman=indf be/stay-hab:3sg=q 'Is there a woman in our house?' (30) na ing=ti naa nyi-bu 1sg word/speech=indf neg say-fut:1sg 'I won't say anything.' As in many languages this indefinite marker is closely related to the word for 'one', which is ti. It can be seen as a cliticised variant of that word. Both as a clitic and as an independent word (=)ti is also used to mean 'another one', 'a different one', as will be exemplified below.
Having introduced the indefinite marker I now turn to a further consideration of anaphoric definite marking in Ku Waru, and exemplification of how it interacts with indefinite marking. The excerpt in (31) comes from a transcript of a paternity dispute centred on a woman who has become visibly pregnant after leaving her husband and going back home to live with her parents (see Merlan and Rumsey 1986 for more details and a full transcript). The speaker is a man from her husband's tribe who is putting a hypothetical scenario to the accused woman as part of a concerted effort to extract a confession from her. In the first line a hypothetical man is introduced with indefinite marking on the word for man, yi-ti 'a man'. In the second line that same hypothetical man is referred to with definite marking. This shows that the identifiability of the referent by the speaker is not presumed by the use of definite marking -only identifiability by the reported speaker within the imagined world that is being conjured up by the narrator.
In the third line the same expression as in the first line, yi=ti=kin, is used again, this time with the sense 'another man'. Apparently the two events referred to over the four lines are being presented as a slippery slide into the presumed act of adultery. In the second reference to the man in that line, kang-yi=yl=kin 'young man' is marked with the definite, showing that he is the same man as the one referred to with the indefinite earlier in the line. In the fourth line another man is introduced, the woman's husband. He is referred to with the same word for man yi that is used in the previous lines. Although that word is again marked with the definite marker, it is clear that it is a different man from the fact that it occurs with a specifying relative clause 'the one who takes you'.
As mentioned in §3.3 connection with the observations by Dryer (2014), one of the ways in which Ku Waru differs from western European languages with respect to definite and indefinite marking is that, in common with most of the other languages in his sample that have both kinds of marking, the choice between those two is not the only option: NPs can occur without being marked as either definite or indefinite. An example occurs in the first line of (31): kang-yi (as distinct from kang-yi=yl=kin as in the third line and the alternative well-formed expression kang-yi=ti=kin 'with a young man'). Other examples may be found (in 21) (lku) and (26) (langi). This potential for formal alternation between definite marking and zero-marking (as opposed to indefinite marking) makes the definite category a functionally more marked one than it is in languages such as English where the only choice is between definite and indefinite. When one looks at the use of Ku Waru definite marking in face to face interaction it can be seen to be used at least in part for "monitoring and adjusting intersubjective settings" (EBS 2018a:113) through the alternation, for a given referent, between zero-marking (i.e., the absence of either definite or indefinite marking) and definite-marking.
There are many examples of this in videos that have been taken of parent-child interactions during Francesca Merlan's and my ongoing study of Ku Waru children's language socialization. The following one comes from an interaction between two adults and a girl, who was then two years and nine months old. It can be viewed online at https://vimeo.com/260881253, password: Kailge. The adults in the interaction are the girl's mother, who is sitting behind her, and her uncle. The uncle has just given her a surprise gift, a small toy truck (called kar in Ku Waru, trucks being much more common than cars there). At the beginning of the clip the girl is holding the toy truck in front of her and looking down at it. The interaction then proceeds as in (32) Before this exchange the three speakers had been talking about the gift that the girl had been given by her uncle, but none of them had yet referred to it with the word kar. The uncle is presumably trying to elicit that word from the girl with his question aki-yl nabola-yl 'What's that?' Instead of replying with kar she says something that no one we've asked could recognize as a word. (Its initial consonant is not among the speech sounds of Ku Waru, nor is the diphthong ɛʊ). The mother then answers for her daughter and the girl repeats after her mother: kar, 'truck'. The answer is perfectly grammatical in that form, but note that it is not marked as either definite or indefinite. In this context its status as a referring expression is presumably less salient than it would be if marked as either definite or indefinite, since what the uncle is really asking for is the word for truck. After the mother and the girl have both provided that word, the uncle follows up by using it as a referring expression, appropriately marked with the definite. As can be seen from the video, he does so in the context of redirecting the girl's attention from the quasi-metalinguistic act in which he has elicited the word kar, back to the referent itself -the toy truck.
The peak of the preceding metalinguistic phase comes right after the mother provides the "correct" elicited form kar. Having already looked up from the toy truck to her uncle when making her "incorrect" reply [∫ɛʊ], after then hearing the mother's "correct" reply kar the girl leans forward to her uncle and likewise says kar with an animated facial expression, looking pleased with herself. As soon as the uncle asks kar-ayl kanakun-i 'Do you see the truck' she looks down at it again. Definite marking plays a part in this insofar as it helps to transform the interaction between the uncle and his niece from a dyadic face-to-face one, back to a triadic one in which they are both focusing on the truck but also attending to each other's attention to it. This is a prime example of the use of definite marking in the engagement function of "monitoring and adjusting intersubjective settings" (EBS 2018a:113).

Clause-level uses of the definite marker and epistemic demonstratives
As mentioned in §3.3, in addition to its use on NPs the Ku Waru definite marker is also used on final verbs, with scope over the entire clause (as is also true of some of the other engagement markers discussed by EBS 2018a, 2018b, and taken account of in their typology). In that environment its function is not to help identify referents, but rather to mark the propositional content of the clause as a matter of mutual knowledge between speaker and addressee. An example is (27), where the definite marker occurs not only on the tribe names epola and alya as discussed above, but also on the verb tebulu 'We will do', thereby highlighting the fact that what the speaker and his fellow tribesmen are about to do -publically present compensation to their allies for their support on the battlefield and the injuries they have suffered -is exactly what is expected under the circumstances, and a matter of common knowledge among members of the audience from the donor group and the recipient one. Another example of clause-level use of definite marking, from the same compensation event as (12) -(13) and (15) -(19) and (25) -(28), is (33).
(33) ekepu wik autiyl o-ba pu-kum=iyl now week many come-nf:3sg go-ppr:3sg=def 'Now (as we know) many weeks have come and gone.' ekepu wilyala yi=yl kul-um=iyl now up.there man=def die-pfv:3sg=def 'Now (as we know) the man up there [a well-known leader] has died.' Here again, as indicated within parentheses in the free translations, the speaker's use of definite marking on the verb in each clause indicates that the matters that are predicated within them are ones that are well known to his audience.
In common with the definite marker, the epistemic demonstratives can also occur with scope over the entire clause. This has already been exemplified by (18), where adi occurs in bare form immediately before the verb. More commonly, when it has scope over the clause it occurs immediately after the verb. The same is true of the other epistemic demonstratives kani and i, as discussed in §2. But interestingly, none of them ever occurs in bare form in that position. Rather, when following a final verb each of them is suffixed with the definite marker.
The use kani=yl at both the NP level and the clause level is exemplified in (34).

(34) [na=nga [[kang] [kumulaya]] kani=yl] li-rim=kin
1sg=gen boy first.born abk=def be.in.place-rp:3sg=com 'When my eldest son was lying dead wilyi kang=ma=n=a uj kes baim ti-ring kani=yl up.there boy=pl=erg=voc wood coffin buy do-rp:2/3pl abk=def those fellows up there bought the coffin.' At the NP-level, kani-yl in the first line indicates that the boy who died was well known, by the speaker of course, and also by his addressees, but that he is, obviously, not present. Clause-level kani-yl at the end of the second line indicates that the buying of the coffin for him by the men he refers to was an event that is well known to them, but not immediately perceptible to addressees or himself. The point of is not to announce it as new information but to call people's attention to it as commendable act. An example involving NP-level use of kani-yl and clause-level use of i-lyi within the same clause chain is (35).
(35) el kani-yl moju-lka arrow that.abk=def undergo-sr:1/3sg el=ayl=nga kompensesen te-kir i=lyi arrow=def=gen compensation do-ppr:1sg ip=def 'You got wounded by that arrow and it's for that arrow that I'm paying compensation.' The two-word NP at the beginning of this sentence refers to a wound that was suffered by one of the men in the tribe to whom compensation is being paid. The use of kani (in this case suffixed with definite marker =yl) indicates that that particular injury, which had happened a few months before, was known to both the speaker and his addressees but the arrow, or the injury for which it stands, was not present and perceptible to them at the time of speaking. By contrast, the clause-level use of i after the final verb (again suffixed with the definite marker, which is obligatory in this context) highlights the fact that the giving of compensation that for that wound is underway and perceptible at the time of speaking.
An example involving both adi=yl and i=lyi at clause level is (36).
(36) tu-rum adi=yl hit-rp:3sg abu=def 'He struck before.' altepa to-ba again hit-fut:3sg 'Now he wants to strike again.' oba ya yi=ma ekepu [[ ku tausen] i=lyi] te-kumul i=lyi so here man=pl now money thousand ip=def do-ppr:1pl ip=def 'So [that's why] we are now are making this payment of one thousand kina.' The first line refers to a tribe who are enemies of the speaker's tribe and his addressees'. The attack he refers to happened approximately forty years earlier, and is generally known of (as marked by the definite ending that occurs with adi), but not an event that the speaker (or in this case his audience) knows much about, as marked by adi itself. By contrast, the payment of 1000 kina that is referred in the last line is happening very visibly at the time of speaking, partly in order to strengthen the alliance between the speaker's tribe and the recipients', in view of the threat of another attack against them that is referred to the second line.

Ku Waru engagement marking in action
In the discussion so far I have been treating Ku Waru engagement mainly on the basis of examples from texts -in particular from transcripts of multi-participant interaction at two public events. Those were both recorded in the early 1980s without the use of video. Much more recently we have been making extensive use of video in our research on Ku Waru children's language socialization. Drawing on that material, in §3.3 I included within the discussion of definiteness an example from filmed interaction between a young child and two adults. I showed how alternation between presence and absence of NP-level definite marking in that context functioned in close synchronization with gaze direction, facial expressions and bodily orientation to achieve intersubjective coordination. Here I will take that further by turning to a somewhat more extended example of videoed interaction involving an older child, and clause-level use of the definite marker. The video is available online at https://vimeo.com/257625252 Password: Kailge. (Note that here and in (32) above I am not using the videos to make any point about children's speech in particular, but rather about embodied human interaction more generally, and the use of definite marking within it. For that purpose it would have been good to be able include videos of interaction between adults, but unfortunately for that purpose, we don't have any. There is however, nothing specifically child-like about the children's use or non-use of definite marking in either example (32 or 37)).
As will be seen in the video, the boy is sitting in his father's lap, facing away from him toward the video camera, which is being operated by his uncle. In the lead-up to this stretch of interaction, the uncle has suggested that he is going to give his nephew some money. Instead of responding directly to his uncle, the boy has put his hand into his father's pocket and started feeling around for coins there. As he does so, 12.8 seconds before the video begins, the father says the words shown in (37).
(37) mare pe-lym=ja kan-abiyl some be/lie-hab:3sg-maybe see-opt:1du 'Maybe some is there; let's (you and I) see.' After that, but still before the excerpted video segment begins, instead of money in the pocket the boy finds a bit of dried tobacco leaf there, which he pulls out and hands to his father. He then puts his hand back into the pocket and starts searching for money again. At that point the video segment begins. In order to show how the father's utterance in (37) has formed a starting point for what the child goes on to say in the video, it is repeated in the first line of (38) below. The video begins after the three dots shown beneath that line. This short stretch of interaction is a classic case of joint attention, in which all three participants are focusing on the father's pocket and the child's attempt to find money there. Within it the use of definite marking on the verb in line e is pivotal in that it marks the point of transition from shared uncertainty about whether there is a coin in the father's pocket to shared certainty that there is. As in (32), here again the use of definite marking is tightly integrated with several other linguistic and corporeal aspects of the interaction that can be seen the video. These include: 1. Gaze direction and facial expressions. After looking in the general direction of the father's pocket during lines b-d, immediately after line e in which the child in effect announces that he has found a coin he looks toward his uncle and smiles, communicating to the uncle his satisfaction that he has found what he was looking for. 2. Intonation and prosody. As the child feels around in his father's pocket during lines b and c he speaks those lines at a relatively high, level pitch, with elongated final vowel in line c, a prosodic feature which in Ku Waru as in many other languages (Tedlock 1983:211) is used iconically to signal that the action or state of affairs being referred to -in this case the state of uncertainty about whether there is a coin in the father's pocket -is prolonged. In line e, after feeling the coin, the child pronounces the word pelym=iyl 'it is there indeed' with a falling intonation on the final syllable, which is iconic of the resolution of that uncertainty, which is also explicitly indicated by the suffix (=iyl) on which the fall pitch in pitch takes place. I will have more to say about what enables this in §4 below. 3. Use of the suffix =ja. The child's use of this suffix in lines b and c is inherently intersubjective in that it entails that neither he nor his uncle know yet whether there is a coin in the father's pocket -only that there might be. The uncle affirms this entailment on both counts by his repetition of the child's utterance ti pelym=ja in line d, thereby aligning his stance with that of his uncle in that respect. 4. Parallelism between lines, both across conversational turns and within the child's single turn that extends across lines b-c. By "parallelism" here I am referring to the meaningful interplay of repetition and variation as theorized by Jakobson (1960), Silverstein (2004), Evans (2012), Fox (2014) and others. Particularly important among those others, for present purposes, is the work of Jack Dubois and his colleagues on what they call "dialogic syntax" (Dubois 2014) and "resonance"  across lines of talk, which they represent with the format that I have used in (38), in which the repeated elements and the positions where there is variation are vertically aligned across lines of text. As can be seen from (38), this kind of display, which they call a "diagraph", allows one to see how the process of intersubjective coordination unfolds not only through the use of specific morphemes whose meanings explicitly relate to aspects of that coordination, but also through patterns of repetition and variation that place those morphemes in salient relationships to each other across lines of text. A prime example of this in (38) is the relation between =ja in line lines b-d and =iyl in line e as discussed above. It is precisely the parallel relationship between those two morphemes in lines d and e that allows the prosodic contrast between them to iconically signal a movement from uncertainty to shared certainty.

Conclusions
Starting with EBS's definition of engagement as grammaticalized marking of the relative accessibility of an entity or state of affairs to the speaker and addressee, in this article I have addressed the question of the extent to which engagement comprises a systematic, well bounded domain within particular languages, with Ku Waru as my test case. I have focused on three sets of formal elements in Ku Waru that seem to be the most relevant ones: definite and indefinite markers, spatial demonstratives, and epistemic demonstratives. Looking at them in terms of EBS's axis of intersubjective distribution, I showed that within each of those three domains the meanings of the terms involve EBS's posited dimensions of epistemic access, but that in none of the domains are the terms distributed across the full set of four possible feature specifications. Table 1 shows their distribution in summary form.
As for EBS's other typological axis, of the scope of engagement marking, it is interesting to note that two of the three kinds of formal elements shown in table 1 -epistemic demonstratives and definitive markersare used as engagement markers at both the NP level and at the clause level, as discussed and exemplified in §3.4. In other words, all of them are "level shifting" markers in the sense discussed by EBS (2018b:159ff).
That is, not only does Ku Waru have engagement markers that function at each of the levels, all of the markers that function at the clause level (i.e., the definite marker and all of epistemic demonstratives) are the same as the markers that function at the NP level (unlike, for example, Andoke auxiliaries, which function only at the clause level). In this respect the syntactic distribution of Ku Waru engagement markers is part of a larger picture in that, as pointed out in §2, there are several other markers with the same distribution, including all of the Ku Waru case-marking postpositions, which are used both on NPs and on final verbs to form subordinate clauses with related senses to the ones they have when occurring with NPs (for which Merlan and Rumsey 1991:340-341). The two levels of scope that are discussed above are distinguishable on syntactic grounds, as pertaining to the NP vs the clause. So far in this discussion (as in much of EBS's) I have treated the differences in syntactic level as corresponding to a semantic/pragmatic distinction between a focus on "entities" vs a focus on "states of affairs" (EBS 2018b:150), the latter being associated with propositions (164). Based on the latter, semantic/pragmatic classification EBS posit a third scopal level of engagement marking: the metapropositional. At that level, "attention is coordinated not necessarily towards an event itself, but rather to the evidence for it. This represents a similar shift in level as that from entity (typically, the province of demonstratives in the noun phrase) to state of affairs" (EBS 2018b:150). As can be seen from §3.4, the descriptive part of the above quote applies well to the use of the epistemic demonstratives at the clause level in Ku Waru. At that level i indicates that the speaker and addressee both know about the event or state of affairs described in the clause, and furthermore, and that the basis on which they know about it is co-presence with it and immediate perception of it. The epistemic demonstrative kani also indicates shared knowledge, but that that knowledge is not based on co-presence and immediate perception, but rather on prior shared experience. In that respect the functions of i of kani at the clause level are transparently related to their functions at the NP level, where the shared knowledge in question concerns an entity rather than an event of state of affairs. This suggests that, for Ku Waru in any case, the distinction between the marking of knowledge distribution as between speaker and addressee, and the marking of its evidential basis, should not be seen as one that calls for a third scopal level as proposed by EBS. Rather, for Ku Waru the distinction in evidential basis is best seen as one that cross-cuts the scopal distinction between NPs/entities and clauses/propositions. The same may be true of other languages, given that, as noted by EBS, sometimes "evidentials may take scope over a referent...rather than a state of affairs" (150). And treating the matter in that way would bring the relevant semantic/pragmatic criteria more into line with the syntactic ones, since EBS's posited level of the metapropositional does not correspond to a distinct syntactic level, but is realised at the same syntactic level as is the propositional, i.e., the clause.
Returning now to my opening question about of degrees of systematicity in how the engagement function is realised, recall that in §3.2 I have described the epistemic demonstratives as a "motley crew" with respect to EBS's axis of intersubjective distribution, in that they are all located within one of the four cells, with the partial exception of adi, and that within that cell (the +speaker, +addressee one), the distinction between i and kani implements a further distinction, in the kind of relevant access to the referent, as between shared immediate perception and shared experience but not co-presence with the object. When we add definite and indefinite marking ( §3.4) and the spatial demonstratives ( §3.1) to the picture it becomes more filled out as can be seen from table 1. But unlike in the case or Andoke auxiliaries or Kogi epistemic marking, in Ku Waru there is no single set of elements that is definable on formal grounds and realizes all four possible kinds of engagement. Rather, to the extent that it is possible to identify a system of engagement along the axis of intersubjective distribution in Ku Waru, it must be done on largely functional grounds, by following EBS's procedure of identifying a typological design space based on "canonical, neatly cut-anddried categories" such the Andoke and Kogi ones, and then seeing how well it is filled out by functionally corresponding categories from other languages. In this case the answer seems to be, reasonably well, but with nothing like the Andoke and Kogi degrees of overall formal and functional coherence. Not only is there no single formally definable set of elements that realizes all four possible kinds of engagement as in Andoke and Kogi, there is a very uneven distribution of elements across the four cells of the table, with the majority of them clustering in a single cell: +speaker, +addressee. Furthermore, neither of the cells with the value -addressee are full-fledged instances of engagement makers, since indefinite marking takes account of both speaker's and hearer's knowledge only in its non-specific sense, and the epistemic demonstrative adi does so only on a presumptive basis (i.e., in the absence of contextual cues to the contrary).
Nonetheless, it is interesting to note that there are at least presumptive or context specific forms of engagement marking within all four cells. In view of the above we can see that this study has suggested two potentially useful questions for future comparative studies of engagement that are based on close examination of particular languages: 1) are there any paradigmatically contrasting sets of forms with the language that map on to each of the four cells within the table? 2) To the extent that the forms are unevenly distributed across the table, which cells are more heavily populated and which ones less so? I suggest that answers to question 1 will have a strong effect on the extent to which engagement becomes established as a viable typological category, and that answers to 2 will provide interesting new questions about the intersubjective basis for engagement marking.
Regarding the latter, it should be noted that among the functional domains of language, engagement marking is very unusual for the extent to which it operates in conjunction with the sorts of paralinguistic and poetic features that are exemplified in §4 and in EBS (2018a:126-128). If extensive comparative study were to show -as I suspect it would -that the relatively low degree of overall systematicity of engagement making that I have found within Ku Waru is more common than not, I would suggest that that fact is related to the thorough-going integration between engagement and other non-linguistic aspects of conversational interaction, and mutual redundancy between it and them. Such redundancy is well illustrated by the interaction in (38), in that the intersubjective concord which is signalled by the child's use of clause-level definite marking in line is also achieved by all three of the paralinguistic and corporeal aspects of the interaction that are discussed in §4 (gaze direction and facial expressions, intonation and prosody, and parallelism). The main thing that is added by engagement marking in that interaction is the degree to which it to explicitly signals that intersubjective accord that has been achieved on other bases. I suggest that this is true of the way in which engagement marking often works in the languages that have it -maybe even in the majority of instances of its use. In order to test that hypothesis, and to advance the understanding of engagement more generally, what is needed are further studies of the overall range of engagement markers within particular languages and of how they work in combination with other linguistic and non-linguistic resources for achieving intersubjectivity.
presentation there, to Tony Woodbury, Francesca Merlan, Naomi Peck and the anonymous Open Linguistics referees for helpful feedback on intermediate versions of the article, and to John and Wapi Onga for their invaluable advice regarding aspects of Ku Waru grammar while on a visit to Canberra at the time I was revising it. Thanks also to Australian Research Council for funding the field research on which it is based.

Non-standard abbreviations
abk absent but known abu absent & unknown en endophor ip immediately present nf non-final verb rp remote past