Noun phrase complexity and contiguity in a Papuan language

4.2.1 External possession and part-whole relationships

In several construction types, possessive relationships are not expressed by the standard possessive structure (with the postposition en, as in Maria en aha ‘Maria’s house’) but by independent NPs, which then can give the appearance of nominal discontinuity. This is most obvious in situations involving actions on a body part, which typically are expressed by casting both the body part and its owner as arguments of the verb, i.e., in an external possession construction (see Schultze-Berndt and Simard 2012: 1030–1032). Each of the arguments is indexed by a different index set on the verb: the affected body part is indexed by Undergoer affixes (or, rather, alternations, since Undergoer exponence is often non-concatenative) and the owner of the body part is indexed by means of the Dative prefix series (Olsson 2021: 203–212). The separate indexing is evidence that each participant is realized as an independent argument, which corresponds to their expressions as separate, often non-adjacent, noun phrases in the clause. An example of external possession is in (14).^[7] Here, the body part pa ‘head’ is indexed in the lexical verb stem, which appears in the gender form matching the Gender III feature of pa ‘head’. The person/number features of the possessor bongso ‘youngest sibling’^[8] are registered in the Dative prefix. Importantly, the seemingly discontinuous expression bongso…pa could not form a grammatical noun phrase on its own.^[9]

The same method of expressing possession by means of two independent noun phrases, with each participant being an argument of the verb, is also found with non-body part expression, as in the similar example in (15). The principle is the same, with the exception that possessors of non-body part possessums are indexed by means of the so-called Genitive prefix series, rather than the Dative seen in (14). Again, the seemingly discontinuous constituents manemna and Alamem could not merge to form a possessor-possessum noun phrase on its own.^[10]

4.2.2 Secondary predicates

Another type of structure that must be distinguished from discontinuous modification are secondary predicates (Schultze-Berndt and Himmelmann 2004). An expression such as kandi ‘raw’ in (16) must be analyzed as a secondary predicate, and not as a modifier that has been separated from its original position in a continuous expression kandi-awe ‘raw fish’, for essentially the same reasons given by Schultze-Berndt and Simard (2012: 1029). For example, secondary predicate structures as in (16) always express temporary properties, but unlike attributive adjectives, they never express permanent characteristics.

Another argument for distinguishing secondary predicates from discontinuous modification is that there are some expression types in Coastal Marind that can only function as secondary predicates, and not as attributive modifiers. One such type is participles with on-going activity readings, which are common as secondary predicates, as seen for ihwluk ‘crying’ in (17).^[11] Participles with ongoing activity readings do not occur in the attributive use, as opposed to participles with resultative state readings, such as kahwidluk ‘dead’, which are fine in the attributive use (kahwidluk-basik ‘dead pig’). See Olsson (2021: 76–83) for further discussion and exemplification.

Because of these differences, it is not possible to use data such as that in (16)–(17) as evidence for discontinuous NEs in Coastal Marind, because such structures are licensed by a separate construction, secondary predication. This construction involves two separate NPs, one of which is referential, e.g., awe in (16), the other being a depictive predicate, like kandi in (16).

4.2.3 Topics

A minor source of apparent discontinuous NEs can be found in some utterances with topic-comment articulation. Here I use ‘topic’ to mean a constituent found at the beginning of an utterance that is overtly distinguished from the rest of the utterance (the comment) by means of a pause, by intonation, or by overt morphological material such as a postposed demonstrative or the particle a (Olsson 2021: 138, 485–487). Semantically, the topic identifies some entity about which the rest of the utterance provides information (an aboutness topic), or, more broadly, it functions as a general “frame setting” topic that specifies the setting that is relevant for the rest of the utterance. A superficial look at a sentence such as (18) could suggest that it contains a discontinuous NE, with the possessor epe anem ‘that man’ separated from the possessum wah ‘mother’. These two expressions could indeed be merged to form a grammatical noun phrase, as kinship terms often allow possession without the postposition en. But it is clear from the intonational pattern that epe anem, as well as the intervening placename Alatep, are prosodically marked topics, as suggested by my translation. The pitch excursions marking the final syllable in each of the words epe, anem and Alatep are seen in Figure 1 and identify the two initial constituents as topics. The pitch movements of the first three words (the last of which stretches into the upper range of the speaker) contrast with the relatively flat intonation associated with the comment ‘his mother died (lit. was lying) there’.

Figure 1:

Annotated pitch track showing two topicalized phrases, epe anem ‘that man’ and Alatep

4.2.4 Measure phrases

Measure phrases are another frequent case of apparent discontinuity. In (19), the measure phrase inah plastik ‘two plastic bags’ specifies the amount of kanis ‘betelnut’ although the expressions are not adjacent.

The main reason for not treating measure phrases as part of discontinuous NEs is that measure phrases never seem to be part of the NP in Coastal Marind. Unlike English, which allows a measure phrase and the measured substance to be expressed within one NP (one bag of rice), in Coastal Marind all attestations of measure phrases are either separated from the expression denoting the quantified entity, or, if the measure phrase is added directly after the quantified expression, marked by a prosodic reset suggesting apposition. My interpretation of these facts is that measure phrases in Coastal Marind are expressed as adjuncts on the clause level, and that contiguous NPs lack a slot for a modifying measure phrase.

4.2.5 Gender agreement and constituency

The gender system of Coastal Marind offers a splendid, albeit rather extreme, illustration of Louagie and Reinöhl’s (this issue) reminder that gender agreement between nominals is often not a straightforward indicator of constituenthood and, conversely, that ‘non-conventional agreement’ (Louagie and Reinöhl, this issue) does not exclude syntactic unithood. Nominal agreement targets in Coastal Marind exhibit so-called external agreement (see e.g., Chumakina et al. 2019), whereby a target unexpectedly agrees with a non-local controller. In (20), the postposition lek ‘from’ appears in its plural form lik in accordance with the gender of the subject isahih ‘children’, despite being a locational adverbial and not a dependent of the subject NP. If agreement were local, we would have expected the postposition to agree with its complement NP, sakola ‘school’. (Note that this example features an actual discontinuous NE, ipe…isahih ‘the…children’, but this does not affect the external agreement on the postposition.)

An even more flamboyant example illustrating both agreement with a non-local controller and non-agreement between modifier and its head is in (21).^[12] The question-word Vn ‘which’ does not agree with the noun that is modifies, milah ‘village’, but takes its gender specification from the subject of the clause, despite the lack of syntactic dependency between the controller and target. This is very surprising, because un ‘which’ is part of a clear syntactic constituent with milah, yet the agreement target finds a long-distance controller in the higher subject NP. These agreement phenomena, which are restricted to a small (but systematic, nevertheless) set of environments in Coastal Marind,^[13] show how nominal agreement can be surprisingly disconnected from constituenthood.

The preceding section described a number of structures that do not qualify as instances of discontinuous NEs. In the next section I examine those discontinuous NEs that cannot be subsumed under any of these structures, and must be treated as actual discontinuous NEs.

4.3.1 Discontinuous NEs under constituent focus

Constituent focus is often invoked as a motivation for discontinuous NEs (e.g., Reinöhl 2020: 86; Siewierska 1984; see references in Schultze-Berndt and Simard 2012: 1038; Louagie and Verstraete 2016: 52). For Jaminjung, Schultze-Berndt and Simard (2012: 1034–1047) argue in detail that discontinuous expression is motivated when an NP has a focused sub-part (e.g., a modifier). The focused subpart is expressed in the beginning of the intonation unit, which is the normal position for focused constituents, while the non-focused, or given, part of the NP is expressed in a post-verbal position. Examples that provide exact parallels of this situation can be found in Coastal Marind, suggesting that focus on a nominal subconstituent is a motivation for discontinuous NEs in this language too. Consider (22), which was uttered by a member of a hunting party that had just returned to the village. The important game animals are wallabies, pigs and deer, and the speaker specifies how many of each were caught. In this context, the animals are the givens, and the numbers of each animal are the unknown variables that an interlocutor might ask for. The modifier hyakod ‘one’ is in focus, and appears in the immediately preverbal position, while rusa ‘deer’ appears at the end of the clause, after the verb.

Focus expression in Coastal Marind shows some typologically rare characteristics, and I will give a brief outline of the focus system before discussing its relevance for discontinuous NEs. A focused constituent must be expressed in the syntactic position immediately preceding the verb, as in (22) above, which is not uncommon cross-linguistically. A more uncommon feature is that this obligatory syntactic placement is accompanied by morphological marking, not on the focused constituent itself, but instead by prefixes on the verb, marking the role of the focused constituent. These prefixes are called Orientation prefixes in Olsson (2021: 253–296), and mark the verb as being ‘oriented’ towards a focused constituent in a certain role. The prefixes shown in the examples in this article are: the ‘Neutral’ prefix k(a)- (with a zero allomorph ø- in non-present contexts) which primarily marks a focused constituent in the S/A role, the ‘Object’ m(a)- marking the O role, ‘Directional’ k(a)- (making no tense distinctions) which primarily marks the role of goal (in motion events) or recipient (with transfer verbs), and the Restrictive s-, meaning ‘only’ (the Restrictive is purely quantificational and does not distinguish the grammatical role of the focused constituent). This intricate system of focus marking (combined with contextual and prosodic cues) often makes it straightforward to determine whether a speaker intends to convey focus on a constituent, which in turn makes it possible to show that some cases of discontinuity correspond to focus on one part of a complex NE.

A mini-dialogue (without discontinuous NEs) illustrating the workings of the Coastal Marind focus system is in (23), taken from discussion of a set of pictures (described in San Roque et al. 2012). In this picture, a person is seen getting beaten up by some others. The two speakers, A and B, are trying to decide whether it is the wife or the husband, seen in earlier pictures, that is getting beaten up. The expressions uhyum ‘his wife’ and eham ‘her husband’ occur in the focus positions in the two utterances.

The utterances in (23) would be appropriate answers to a question such as ‘Who is in the picture?’ or ‘Who is getting cut/beaten up?’, which can be understood as implicit in the context of (23). Content questions and their answers are typical contexts for the use of this focus structure. The conversational excerpt in (24) gives an explicit content question (line 1) and its answer (line 2). The focused expressions ta ‘who’ (line 1) and kaka Budi ‘elder brother Budi’ (line 2) are obligatorily placed in the immediately pre-verbal slots.

More generally, speakers combine pre-verbal placement and the use of the Orientation prefixes on the verb in contexts where some alternative is emphasized to the exclusion of other potential alternatives, e.g., in corrections to some previous statement.

Constituent focus targeting one part of a NE is a source of discontinuity in Coastal Marind. When a clause contains a nominal head and a modifying element, and the modifying element, rather than the head, is the focal part of the utterance (e.g., because it corresponds to the information asked for in a content question, or singles out some property among other alternative properties), speakers of Coastal Marind have the choice to place only the modifier in the pre-verbal focus slot, leaving the nominal head stranded elsewhere in the utterance. In their study of Jaminjung, Schultze-Berndt and Simard (2012) find that the types of modifers that occur separated from their head under contrastive focus are interrogative expressions (i.e., wh-phrases), possessors, quantifiers and attributive property expressions (2012: 1035–1036). Discontinuous expression of these four types are possible in focus contexts in Coastal Marind, and are attested in the corpus, with the exception of attributive property expressions, which so far remain unattested in spontaneous speech data. I will give some examples illustrating discontinuous NEs associated with constituent focus, before commenting on their significance.

Discontinuous interrogative expressions are found mostly in elicited data, but this is probably an artefact of interrogatives having been a common target for elicitation due to their complicated morphological paradigms (involving the verb prefixes h- and b- seen in the examples), and more discontinuous interrogatives would probably turn up with more conversational corpus data. Examples (25) and (26) illustrate discontinuous interrogative expressions. In (25), the interrogative phrase ta en ‘whose’ is in the preverbal position associated with focus, and the nominal head yaba-ember ‘big bucket’ added at the end of the utterance. It is likely that the presence of the adjective yaba ‘big’ is a contributing factor to the discontinuous construal: a contiguous NP ta en yaba-ember would be grammatical in the preverbal position, but speakers seem to prefer to narrow down the material that is placed in the focus position to the part that is most relevant for interpretation, in this case ta en ‘whose’.^[16] In (26), which was volunteered during an elicitation session, the interrogative intagi ‘how many’ is separated from awe ‘fish’.

Figure 2 shows the prosodic properties of (25). The focused interrogative phrase ta en ‘whose’ has the main prominence of the utterance, as seen by the waveform in the upper panel and the peak in the pitch track. After the utterance-initial discourse particle awi ‘what about’, which has its own rising pitch, there is a pitch rise during the focused interrogative ta en, and the high pitch is then maintained for the copula, before dropping sharply to mark the de-emphasized head, yaba-ember ‘big bucket’. This seems to parallel the pattern of de-accenting in partially focused discontinuous NEs in Jaminjung, described by Schultze-Berndt and Simard (2012: 1036).

Figure 2:

Annotated pitch track of Example (25).

Focused numerals are a source of discontinuous NEs in the corpus. A clear example of an emphasized numeral involving a discontinuity is (27), in which the focused expression hyakod ɣa ‘really one’ is separated from the noun basik ‘pig’.

Figure 3 shows the prosodic properties of (27). Again, the focused numeral phrase hyakod ɣa ‘really one, one entire’ is accompanied by a rise in the pitch, reaching a high plateau that is maintained throughout the verb, and then drops sharply before the head, basik upe ‘that pig’, which is made much less prominent.

Figure 3:

Annotated pitch track of Example (27).

Discontinuous NEs involving focused adjectival expressions appear to be possible, but are not attested in corpus data. The elicited utterance in (28) shows a focused adjectival expression, dislocated from its semantic head aha ‘house’. This structure was judged well-formed by other speakers, but remains unattested in the corpus, and I was not able to elicit it with adjectives other than ‘big’ and ‘small’. The limited occurrence of focused adjectives separated from their heads is unsurprising given the tight bondedness between adjectives and their heads noted in Section 3.

It should be noted that examples of discontinuous NEs in which only a subpart (such as the modifier) is focused are infrequent, and I have only found six clear examples of the phenomenon in corpus data. Note, however, that argument NPs in which only a subpart is focused, but which retain the standard modifier-head order of tight NPs (Section 3), are completely unattested in the corpus. Complex NPs in the focus position are not uncommon in the corpus, especially with expressions of temporal duration.^[17] Example (29) is from a story, and can be thought of as answering an implicit question ‘How long did you stay there?’. This example shows that complex NPs with the standard modifier-head order are possible in the focus position, but are used when the scope of the focus includes both the modifier and the head.

This suggests that constituent focus on subparts of NEs explains some of the discontinuous NEs in the Coastal Marind data, as was argued for Jaminjung by Schultze-Berndt and Simard (2012). Note, however, that a typological difference between the focus systems of the two languages probably results in a lower proportion of discontinuous NEs associated with constituent focus in Coastal Marind. The focus position in Coastal Marind is immediately pre-verbal, and not clause-initial like in Jaminjung, and it is common for the remainder of the NP to be placed clause-initially, preceding the focused pre-verbal modifier.^[18] This sequence of an initial (non-focused) nominal, followed by a (focused) modifier in the pre-verbal position does not result in a discontinuous NE, but simply a deviation from the standard modifier-head order of tight NPs. A clear example is (30), in which the emphasis on hyakod ‘(not a single) one’ motivates its placement in the pre-verbal position. (In negated sentences with focus, like this one, the negative morpheme mbya intervenes between the focused constituent and the verb.) The noun awe ‘fish’ is not part of the focus and so does not qualify for placement in the pre-verbal position (as part of a tight NP).

The preceding discussion has suggested that constituent focus on a subpart of the NE is a causal factor behind some instances of non-contiguous NEs in Coastal Marind. The structural prerequisites that seem to be involved are, firstly, the possibility of discontinuous NEs, and secondly, the existence of a dedicated slot in the architecture of the clause that hosts a focused constituent, and thirdly, a preference for placing only the focused subconstituent in the pre-verbal focus slot, rather than the entire constituent of which the focused material is a part.

4.3.2 All-new, presentational sentences

In their study of Jaminjung, Schultze-Berndt and Simard (2012) identify “sentence focus” as one of the contexts that frequently trigger discontinuous NEs. The data they cite are strikingly similar to Coastal Marind utterances such as (31) and (32), which also involve discontinuous NEs. These are “all-new” sentences introducing a new participant. Referring to the works of Sasse (1987) and Lambrecht (1994), Schultze-Berndt and Simard describe such utterances as “thetic” and involving focus on the whole sentence. The vagueness of the theticity notion makes its application to novel languages somewhat difficult, and in earlier work on Coastal Marind I reserved this label for certain structures involving an auxiliary with a postposed verb stem (Olsson 2021: 461). This structure is commonly used to answer the question ‘What happened?’, at least if the reply is cast as containing no presupposed information. The utterances in (31) and (32) represent a different type of thetic statements, in which a stative positional verb predicates the location and existence of some new participant, which I will refer to as presentational here.

The pitch track in Figure 4 shows the prosodic characteristics of the presentational sentence in (31). The head iwag ‘woman’ is given prominence since it is the most informative element of the utterance, and introduces a set of new participants. But the rest of the utterance differs prosodically from the ones involving focus seen before (Figures 2 and 3). Unlike clauses with focus, there is no high pitch plateau stretching across the focused preverbal constituent and the verb, and no sharp pitch drop marking the de-accented post-verbal material.

Figure 4:

Annotated pitch track of Example (31).

Presentational utterances of this type are common in the corpus, but the vast majority involves simple noun phrases, typically just a bare noun.^[19] I have only found nine instances of complex NEs (involving e.g., a numeral or property word) in such presentational sentences, and five of those show discontinuity of the same type as in (31) and (32). In each instance the pattern is the same, with the head noun, or some other item identifying the referent, such as isi ‘another (one)’ in (32), in the beginning of the utterance, followed by a demonstrative element preceding the verb, and the modifying element after the verb. This is exactly the pattern identified by the pattern identified by in their data on discontinuity under “sentence focus” (2012: 1041).

Why do discontinuous NEs occur in presentational, or thetic, utterances? For Jaminjung, Schultze-Berndt and Simard argue that discontinuous NEs are one of several strategies used to signal that an utterance is thetic, and they conclude that “NP discontinuity needs to be added to the list of crosslinguistically available strategies associated with sentence focus” (2012: 1046). For Coastal Marind, it seems that the discontinuous construals in (31) and (32) correlate with the all-new status of the referents, but it is unclear whether the discontinuous NEs can be said to ‘signal’ presentational or thetic utterances. The discourse functions of (31)–(32) are already conveyed by the use of positional verbs with preceding locational specifications, and these functions can be conveyed equally well in presentational sentences in which the subject is a bare noun (which cannot be given discontinuous expression). I suggest that an important feature of the discontinuous NEs in (31)–(32) is that the postverbal modifiers introduce additional information describing the entity that the utterance as a whole is about. I will return to the issue of aboutness in Section 4.3.4.

4.3.3 Utterance-final demonstratives

Utterance-final demonstratives that form a discontinuous expression with a noun earlier in the clause, with which they agree in gender or number (indicating that they are not locational adverbials), are common in corpus data. This phenomenon occurs mainly in face-to-face conversation and in reported speech in narratives, but the precise discourse function of these expressions is unclear. Utterances with final demonstratives are similar to the presentational sentences discussed in Section 4.3.2 as they mostly involve novel participants, but the information-structural status of the new participant seems to differ between instances, and the introduction of main participants does not seem to be the main function of these utterances.

Typical examples are in (33)–(34). The contexts suggest that the function of the utterance-final demonstratives involves an attentional element, i.e., that they serve to bring the speaker’s attention to some novel situation, as has been described for utterance-final demonstratives in other Papuan languages (Evans et al. 2018b; Kratochvíl 2011).

Example (35), taken from the same narrative as (33), is evidence that these clauses do not serve to introduce new participants into the discourse. Here, the same canoe as in (33) is mentioned again. The canoe is old information, but the mode of expression is identical to its first mention.

Examples (33)–(35) lack the focused-presupposed dichotomy of the utterances discussed in Section 4.3.1, but other utterances with final demonstratives do contain focused constituents in the pre-verbal position. Examples (36)–(37), taken from a narrative, involve first mentions of participants, but are in fact focus constructions, as shown by the presence of the Neutral Orientation prefix ka- on the verbs. They can be thought of as answering implicit questions of the type ‘Who is that coming there?’, and the context following (36) does indeed involve discussion of the person’s identity (Have a good look at him, oh, it is really him, etc.).

I conclude that the use of discontinuous NEs in (33)–(37), caused by the utterance-final demonstratives, does not primarily follow from the information-structural status of the participants, but seems to correlate with the attention-directing function of these utterances.

4.3.4 Aboutness and discontinuous NEs

In this section, I discuss a discourse feature that seems to be a central factor in the distribution of discontinuous NEs in Coastal Marind, which I refer to as aboutness. This feature holds between an utterance and a referent, so that an utterance can be described to be about a referent. Aboutness is an important notion in philosophy (Yablo 2014), library science (Hutchins 1978), and, in linguistics (Krifka 2008), where it is seen as one of the prerequisites for topicalization (Endriss and Hinterwimmer 2008; Reinhart 1981).

In Coastal Marind, aboutness is correlated with discontinuous NEs in the following way: when the head of a NE ranks high in aboutness (i.e., the utterance can be understood as providing information about the referent of the NE), further modifiers added to the utterance can be expressed outside the tight NP structure described in Section 3, possibly forming a discontinuous NE. This observation accounts for the discontinuous NEs with constituent focus described in Section 4.3.1, for a sentence such as (25) easily allows paraphrases with ‘about X’, e.g., “About that bucket: whose is it?”. The aboutness prerequisite is naturally compatible with the presentational sentences in Section 4.3.2, as those sentences provide information about the participant they introduce. A paraphrase of (31), for example, would be “About the women: they were sitting there and there were three of them”. The utterances with final demonstratives in Section 4.3.3 invariably involve participants that rank high in aboutness.^[20]

Importantly, I do not claim that aboutness explains why speakers sometimes produce discontinuous NEs. My claim is merely that referents that rank high in aboutness are possible targets for discontinuous expression (or expression as loose NEs more generally), whereas those that rank low in aboutness are not. The aboutness restriction is a possible explanation for the observation that discontinuous nominal expressions are unattested in adjunct roles. All instances of discontinuous NEs in the Coastal Marind corpus involve core arguments of verbs, a pattern that has been noted for other languages (Louagie and Verstraete 2016: 51). Adjuncts typically provide additional specifications about the state-of-affairs expressed by an utterance, and it is rarely the case that a clause and its core arguments primarily provide information about the referent expressed by some adjunct.

For an illustration of this, consider the locative adjunct ‘one place’ in (38). The discourse context suggests that contrastive focus on the numeral ‘one’ would be intended here, as animals fleeing a hunter are unlikely to fall in the same spot. Yet, the whole expression is placed in the focus position, rather than only the numeral, which was the construal that was seen with core arguments containing focused numerals in Section 4.3.1. Rather, the whole expression is placed in the pre-verbal position, without any discontinuity – just like numerous other parallel examples involving adjuncts in the corpus. The reason for this, in my view, is the relative lack of aboutness of the adjunct ‘one place’. The paraphrase “About the place: there was one, and the deer fell there” is pragmatically odd, while the translation and contextual clue (‘About deer …’) given in (38) seem natural.

This pattern is confirmed by studying all attestations of the numeral ‘one’ in attributive use in the corpus. Table 1 gives corpus counts for all 73 attestation of this word according to its ordering with respect to the head noun: numeral–noun (which is the order found in the tight NP), noun–numeral and non-contiguous orders. The table also presents the distribution of these orderings across argument and adjunct roles in the clause.

Table 1 shows that all deviations from the tight numeral–noun orders involve NEs in argument roles. Example (39) illustrates both of these orderings, noun–numeral (in its first clause) and non-contiguous ordering (in its second clause). The aboutness status of the nominal heads is clear. One could ask “what was that utterance about?”, and “some wallaby” would be a natural response. The remaining 28 instances of hyakod in these two orderings conform to the aboutness criterion.^[21]

Of the 43 instances of the order numeral–noun (i.e., that of tight NPs), 38 were in adjunct roles. As in the case of Example (38), I was unable to identify any indications that the clauses in which these adjuncts were found somehow could be interpreted as providing information about the adjunct. However, the remaining five instances of numeral–noun order are in argument roles. I will provide some brief comments on these examples, as they are a potential problem for the aboutness account, unless they turn out to be arguments that rank low in aboutness.

Three of the examples are rather complex, and I will not give them in full here, in interest of space. All of them involve transfer events (with three participant roles), and the relevant NPs arguable rank lower in aboutness than the other participants in the events (which are given zero expression, as they are contextually given).^[22] Here I focus on the remaining two examples, which are easier to summarize.

In the first example, given in (40), the expression hyakod yahun ‘one canoe’ is in an argument position, licensed by the Instrumental-Comitative applicative prefix ka- on the verb. While it does not seem unreasonable to claim that this utterance says something about a canoe, there is clearly a more central participant that this utterance is about, viz. some previously mentioned men, as indicated by the clause-initial topic. The phrase hyakod yahun ranks lower in aboutness, which makes non-contiguous expression less likely.

In the second example, given in (41), the speaker was talking about two men in another village who he suspected were married to the same woman. Like in the previous example, the utterance is about the referent of the NP (the woman) to some degree, but this referent is outranked by the referents of the subject of the clause (the men), which are discussed in the preceding and following turns. Note that this clause involves some disfluency, as marked by the use of the pro-word ago (in its Gender II form agu) instead of a lexical noun, but this has no affect on the ordering of the words.

Based on the preceding discussion, I suggest that aboutness is a necessary condition for discontinuous NEs (as well as the loose noun–numeral order) in Coastal Marind. The discontinuous NEs associated with partial constituent focus (Section 4.3.1), in presentational utterances (Section 4.3.2) and utterances with final demonstratives (Section 4.3.3) all rank high in aboutness. Adjuncts rank low in aboutness, which is reflected in the absence of discontinuous NEs in adjunct roles in the corpus. In Examples (40)–(41), I gave some counterexamples to the tendency for argument NPs with the modifier hyakod ‘one’ to be expressed either non-contiguously or with the noun–numeral order. I suggested that the use of the tight numeral–noun order in these examples follows from the relatively lower aboutness rank of the expressions, which are outranked by other, more topical, participants.

It is worth stressing that aboutness, as understood here, is only a condition for non-contiguous NEs, and does not explain why they exist in Coastal Marind in the first place. The explanation for why the grammar of Coastal Marind allows, and – in some circumstances – encourages discontinuous NEs is probably related to the lack of elaboration within the tight NP structure, a preference for distribution rather than accumulation of information, a fixed pre-verbal focus position, and perhaps the prosodic properties of the language (as suggested by Schultze-Berndt and Simard [2012: 1049] for Jaminjung). Language contact has perhaps played a role, as the neighboring Ngkolmpu language features discontinous NEs, although apparently to a more limited extent than Coastal Marind (Carroll 2020).