Categorizing possession in Zuanga-Yuanga and other Kanak languages (New Caledonia): a typological perspective

1.1 Defining adnominal possession

Crosslinguistically, splits between direct and indirect adnominal possessive constructions correlate with the weight and cohesion of encoding. “If a language has an adnominal alienability split, and one of the constructions is overtly coded while the other one is zero-coded, it is always the inalienable construction that is zero-coded, while the alienable construction is overtly coded” (Haspelmath 2017: 199). This prediction is borne out in Oceanic and Kanak languages. Another feature is the “cohesion scale” defined as follows: “In no language will the linguistic distance between X and Y be greater in signaling inalienable possession, in expressions like ‘X’s Y’, than it is in signaling alienable possession” (Haiman 1983: 793, quoted in Haspelmath 2017: 207).

The reasons behind such coding splits and the explanations for it are varied and are discussed in the wide literature on this topic, among them (Aikhenvald 2013; Chappell and McGregor 1989, 1996; Haspelmath 2017; Lichtenberk 1985a, 2004, 2009a, 2009b; Nichols 1988; Nichols and Bickel 2005, 2013). Some definitions correlate dichotomous possessive constructions to the semantics of control, permanence and inherent relation versus their opposite feature, while others like Nichols (1988) consider alienability splits as a structural phenomenon and a label for types of constructions correlating with noun classes on the grounds that “no account of the semantics of possession types will accurately predict the membership of the ‘inalienable’ set of nouns either within one language or crosslinguistically” (Nichols 1988: 568). Haspelmath (2017: 193–194) also predicts that the construction types are attributable to frequency asymmetries and that “additional marking is found when the possessive relationship is less predictable”.

The possessive constructions in ZY are now defined and analyzed against this background.

1.2 Defining adnominal possession in ZY

Lichtenberk (1985a: 125–126) acknowledges that in many Oceanic languages, possessive constructions are not fully predictable from the semantics of the relationship, but he still considers that the “possessive systems of most Oceanic languages are by and large semantically based”. Pawley and Sayaba (1990: 167–168) hold similar views, stating that in Fijian languages “certain nouns belong to strict and semi-arbitrary noun classes for purposes of possessive-marking, others show marking consistently following semantic^[3] […] relations holding between possessed and possessor”. On the other hand, Geraghty (1983: 242) writes, “while there may well have been a time in the history of the Fijian languages when the inherent semantic property of inalienability alone determined how a noun was possessed, the situation has now changed in most Fijian languages, and semantically inalienable nouns are not all possessed in the same way”.

Oceanic languages actually display a cline of tight or loose correlations between possessee noun classes, types of possessive constructions, and the semantics of alienability.

For instance, in Manam (Lichtenberk 1985b) and Paamese (Vanuatu, Crowley 1996), direct or indirect possessive constructions and semantic distinctions (inherent vs. contingent) are strongly correlated, independently from noun classes. Fijian languages according to Pawley and Sayaba (1990: 167–168) have an intermediate position with some “strict and semi-arbitrary noun classes” restricted to one single type of possessive marking on the one hand, and constructions varying with the kind of conceptualized relations “for instance expressing an opposition between possession of goods for consumption and goods as property”.

Like other Oceanic languages, the Kanak languages of New Caledonia – among which, Zuanga-Yuanga (ZY) and Nêlêmwa (NEL) – have a pattern of direct or indirect possessive constructions. ZY and NEL (like some other Kanak languages) also have possessive classifiers used for nouns denoting food, drink, and possession of animals and plants.

In ZY, directly possessed nouns are marked by possessive suffixes or by directly postposed possessor nouns, while indirect possession is marked by prepositions or linkers. Direct and indirect possessive constructions correlate with fairly rigid possessee noun classes,^[4] which usually belong to a single structural class and allow few alternate constructions. Only a limited number of nouns occurring in parts-of-whole relations have alternate constructions with semantic differences denoting inherent versus contingent relations (discussed in Section 5). Another essential feature of ZY is that direct or indirect possessive constructions cannot be fully predicted from lexical-semantic categories such as kinship terms and body parts, which are heterogeneous. There is thus some broad correlation between direct and indirect possessive constructions and the notion of (in)alienability, but this correlation is somewhat inconsistent. Consequently, following Nichols (1992: 117), it is considered that “the term inalienable, […] refers not to a semantic constant having to do with the nature of possession, but to whatever set of nouns happens to take inalienable possession marking in a given language”.

Haspelmath (2017: 198) defines an alienability split for possessed nouns as follows: “a construction used characteristically with kinship and/or body-part possessed nouns is an inalienable possessive construction, while a construction that is characteristically not used with kinship and/or body-part possessed nouns is called an alienable possessive construction. When a language has two constructions of this kind, we can say that it makes an alienability contrast, and we can call the nouns occurring in the inalienable construction inalienable nouns, and those occurring in the alienable construction alienable nouns.”

The challenge in ZY and NEL is that body parts and kinship terms occur in direct and indirect constructions for grammatical and structural reasons (explained in Sections 3.3.2 and 4),^[5] but they do not express clear semantic distinctions. This supports the notion that the split in possessive constructions in ZY primarily correlates with possessee noun classes and secondarily with the semantics of possession. Directly possessed nouns are inalienable and indirectly possessed nouns are alienable in that structural sense. The terms inalienable or alienable are thus taken as labels for structural types of possessive marking, not as consistent semantic characterizations of possession.

In ZY, the main possessed noun classes are (i) bound nouns, (ii) free nouns and (iii) a few non-possessible nouns. Bound nouns have a construct∼absolute suffix that allows them to occur without a possessor and which is replaced by a direct possessor. Free nouns are possessible nouns that do not have the absolute suffix, they subdivide into those which are:^[6]

1. possessed with classifiers (classifiers are themselves directly possessed nouns)

2. directly possessed

3. indirectly possessed by prepositions (all loan words belong to this class);

summarizes the above.

The analysis proceeds as follows. Section 2 presents the general morphosyntactic features of direct and indirect possessive constructions in ZY. Section 3 discusses the main possessed noun classes and their possessive constructions, starting with bound nouns (Section 3.1); nouns possessed with classifiers (Section 3.2); and directly possessed free nouns are analyzed in Section 3.3, while indirectly, prepositionally possessed free nouns are presented in Section 3.4. Section 4 discusses kinship terms showing them to be a split class; Section 5 presents alternate constructions in relation to parts-of-whole (meronymic) possessive constructions, and Section 6 discusses the possessive construction of compound nouns comprising non-possessible nouns. The remaining sections present these same adnominal possessive constructions with other types of possessees, viz. nominalizations (Section 7), spatial nouns (Section 8), some of which originate in body-part nouns; quantifier and partitive nouns (e.g., ‘part of’, ‘amount of’) (Section 9), some of which are subtypes of parts-of-whole relations; and modal nouns (Section 10). Section 11 concludes the article.

2.1 Morphosyntactic features

The pronominal possessive suffixes in the Zuanga-Yuanga dialects are represented in Table 2. One of the differences between the three dialects is the loss of all final consonants in GO, which entailed the loss in GO of the original possessive suffixes for all three singular persons, which are cognate with the Proto-Oceanic possessive pronouns (1sg *-ŋku, 2sg *-mu, 3sg *-ña, Pawley 1973),^[8] and which are retained in PA and BO as 1sg –ny [ɲ], 2sg -m, 3sg –n. In the GO lect, these former possessive pronominal suffixes were replaced by the paradigm of object pronoun suffixes (see Table 2). Compare kiò-ny [PA BO] and kiò-nu [GO] ‘my belly’.

Except for the first three singular possessive suffixes, the other possessive suffixes and suffixed object pronouns are identical for all other persons in the three lects GO, PA, BO.

Another noteworthy and specific feature of the GO dialect is the existence of trial pronouns for 1st, 2nd and 3rd persons, with clusivity distinctions for 1st person (Table 2). Suffixed possessive pronouns do not mark gender, only number.

In ZY as in NEL, direct constructions are marked by personal possessive suffixes (1a), or by directly postposed possessor nouns (1b). Indirect constructions are marked by prepositions and oblique pronominal suffixes (see the paradigm in Table 2) as in (2a)–(2b). Thus, direct possessive constructions display a tighter structural relation than indirect constructions.

Semi-direct possessive constructions do not occur with a distinct possessee noun class, but are an alternate construction for some nouns expressing parts-of-whole relations of a more transient, contingent nature (see Section 5). They are marked by a now accreted, suffixed linker –a, possibly cognate with POc *ka2, which is a marker of uncontrolled possessive relation (Pawley 1973). It is called semi-direct, because the possessor, if pronominal, is suffixed to the accreted linker as in (3b), if nominal, it is postposed as in (3a)–(3c).

2.2 Stem allomorphies of directly possessed nouns and diachrony

Some directly possessed nouns have allomorphic stems but are otherwise directly possessed. Some of these allomorphies result from diachronic processes such as the attrition of the non-possessed stem, or an accreted linker on the possessed stem (see Section 2.2.4). These allomorphic stems distinguish possessive from modifying constructions (see Section 2.2.3).

3.1 Bound nouns and the construct suffix –n

Bound nouns are a subset of directly possessed nouns^[15] that only belong to this class and are defined by their having an absolute suffix in their stand-alone form (i.e., without a possessor).^[16] They are only found in the PA and BO lects and their construct∼absolute suffix generally takes the form of a final consonant, often –n in various northern Kanak languages. This morpheme could be cognate with the non-personal genitive relator POc *ni (Ozanne-Rivierre 1991: 332; Ross 2001: 261). In GO, the loss of all final consonants, including the construct suffix, entailed the wholesale loss of bound nouns in the GO lect, which, as a consequence, only has free nouns as shown in (15a), to be compared with the PA, BO bound nouns:

The construct suffix allows the class of bound nouns in PA and GO.

1. to stand alone, in their absolute form,^[18] see (15b), (17a),

2. to occur in compound nouns (16a), (16c).

Moreover, it does not show agreement in number, and thus has properties of absolute and construct morphemes (see Creissels 2017); it is glossed CST for convenience. This morpheme is replaced by a nominal possessor (16b), (16d), (17b) or by a pronominal possessive suffix if the possessor is human (16e), (17c).

In (16a)–(16c), the nouns for ‘thumb’ and ‘index’ with the absolute suffix –n, occur in compound nouns in which thoomwã or êmwèn are adjoined modifiers, not possessors, thus (16a) does not mean **the woman’s finger, and (16c) does not mean **the man’s finger; the direct possessive constructions are respectively shown in (16b)–(16d) where the possessor is directly postposed to the possessee.

In (17a), the absolute suffix on puxu-n ‘base, bottom’ (referring to an inanimate entity) also stands in a paradigmatic opposition with a definite nominal possessor (17b) or with a personal possessive suffix (17c).

In PA and BO, there is only one construct suffix -n used for inanimate possessors and for non-specific, generic possessors, as in hii-n mãni kòò-n [BO] ‘limbs’ (lit. arms and legs), whereas in neighboring northern languages, like NEL and Nyelâyu, there are two construct suffixes. In NEL, they are –n (for non-specific human possessors) and -t (for non-specific and non-human possessors): e.g., shi-n ‘hand’ versus shi-t ‘paw, wing, tentacle’ (Bril 2013). In Nyelâyu, the construct suffixes are respectively –k (+ non-spec. human possessors) and -t/-r (+ non-spec. inanimate possessors): e.g., ye-k ‘hand’ versus ye-r ‘sleeve’ (Ozanne-Rivierre 1998: 36).^[19]

Apart from some body parts, the construct suffix also occurs with some kinship terms of reference, such as kêê-n mãni kibu-n [PA] ‘fathers and grandfathers’,^[20] which are mostly used in ceremonial speeches to refer to the generic notion of ancestors, not to any specific personal parenthood. Kinship terms are a split class further discussed in Section 4.

In PA and BO, the CST suffix is homonymous with the third-person singular possessive suffix –n of human or animate possessors (inanimate possessors do not have any pronominal indexation). A disambiguating feature is that in contrast with possessive pronominal suffixes, the CST suffix does not agree in number with the possessor; this is best seen in the context of dyadic kinship terms (see Examples (18) to (20)).

Dyadic kinship terms are marked by è- … -n with the construct suffix –n in PA and BO, while the suffix is lost in GO. These terms refer to a few types of dual or plural relationships such as parent and children, grandparents and grandchildren, maternal uncle and nephew/niece, parents-in-law and children-in-law, spouses.

If used predicatively, these kinship terms have dual or plural subject pronouns depending on the number of persons, but the construct suffix -n remains unchanged, as in (19)–(20).

There are equivalent forms in other northern languages: a- … -n in Nyelâyu (Ozanne-Rivierre 1998), a(m)- … -n in NEL (Bril 2000: 84–85). The stem is usually the lower term of the dyad, i.e., pòi(-n) ‘child’ (see (19)), peebu(-n) ‘grandchild’ in ZY (see (20)), with some counterexamples such as è-pööni-n ‘maternal uncle and nephew/niece’ [PA], where the stem pööni(-n) ‘maternal uncle, nephew, niece’ is the higher term of the dyad.

The CST –n also occurs with some quantifier nouns like jiu-n ‘the whole’, allowing their absolute use in reference to inanimate entities as in (21a), compared with the plural personal possessive suffix jiu-laa in (21b) referring to the number of people in (21b) (see Section 9 for other cases).

The CST –n is found with the time noun mwaji-n [PA] ‘time, delay’.

It also occurs with one property-denoting noun, mudo-n [BO] ‘decay’,^[21] it stands in a paradigmatic opposition with a directly postposed noun as in mudo hõbwòn [PA] ‘old clothes’ and mudro mwa [GO] ‘a decayed house’ (lit. the decay of the house). NEL has a similar construction with the bound noun mobwa-t ‘decay’, e.g., mobwa mwa ‘a decayed house’ (Bril [2000: 236]; see Ross [1998] for cases in other Oceanic languages).

The CST –n occurring on some modal nouns allowing their non-possessed, absolute use, is analyzed in Section 10.

3.2 Nouns possessed with classifiers

Possessive classifiers are frequent among Eastern Oceanic languages, which have between four and twenty of them (Lichtenberk 1983, 1985a: 105).^[22] ZY and NEL are in the high range with respectively ten and thirteen classifiers, Iaai has twenty-three of them (Ozanne-Rivierre 1976), while Micronesian languages have up to twenty or more.

The reconstructed POc possessive classifier morphemes are: *ka-1 for food, *ka-2 for uncontrolled possession,^[23] *na-/*no- for controlled possession, *ma- for drink (Pawley 1973: 153–169). Lichtenberk (1985a: 117) and Lynch (1997) consider *na-/*no- as a general classifier. Lynch (2001: 150) adds *ta- and/or *sa- as other candidates. Of these, only POc *ka-, *kani ‘eat’ is reflected as the starch food classifier, cè [GO], caa- [PA BO].

In ZY, the use of classifiers is restricted to nouns belonging to specific semantic domains (such as food and drink) and to nouns that cannot be possessed in any other way. A list is given in Table 3 for ZY and some other northern Kanak languages.

Possessive classifiers in ZY and NEL are morphemes with an obligatory direct possessor (generally marked by a pronominal suffix) and an adjoined possessed noun, with the pattern CL-poss.suffix possessee. The adjoined possessee is thus indirectly possessed, in relation to a nominal classifier which is itself directly possessed as shown in (23).

Classifiers in ZY and NEL check many of the properties of classifier systems (Franjieh 2012: 196): (i) they do not occur with all nouns; (ii) they are semantically assigned; (iii) they can be used anaphorically; (iv) they mark nouns as non-generic.

There is no general possessive classifier in ZY, nor in NEL or YAL, but there is one in Tîrî, hêê- ‘belongings’ (Osumi 1995: 144–156) and one in Iaai aɲi-n (Ozanne-Rivierre 1976: 188). In many Kanak languages,^[25] possessive classifiers denoting the “possession” of food types are extremely common, with a shared set of basic categories, such as starch food, proteins, fruit and leaves, drink, and sometimes more categories. Food types are assigned to a superordinate category such as starch, protein, and drink, and possessed by the corresponding possessive classifiers. These are sortal classifiers that express a ‘genus’/kind, an inherent feature of the possessed entity, common to other members of that class, except for w(h)aza-[GO], w(h)ala- [PA BO] which is only used for ê ‘sugar-cane’, as in whaza-nu ê [GO] ‘my sugar-cane’ (for consumption). Yet, they do not constitute a gender system and do not index any female, male or thing/inanimate distinctions.^[26]

All other possessive classifiers are relational and denote a type of relation to the entity, such as carried objects, weapons, share of food brought for ceremonies, seedlings or cuttings for planting, catch at hunting, and domestic animals. Interestingly, animals and plants are classified not according to some inherent property, but according to relations entertained with them (see Table 3).

Sortal classifiers such as those used for food types tend to be rigidly assigned and are usually exclusive of each other, with a few exceptions like nu ‘coconut’, which can be categorized as a drink or as protein food according to the consumable part considered.

Relational classifiers are much more fluidly assigned since they categorize a noun according to a type of relation rather than to an inherent property, as pointed out by Lynch (1973: 76). Thus, yams as cuttings for farming are possessed with the relational classifier êê-, as a carried load with the relational classifier phò-, and as starch food with the sortal classifier cè-.

Relational classifiers vary according to conceivable relations, e.g., animals as catch at hunting and fishing, as food, as domestic or raised animals; plants are categorized as cuttings for cultivation, as shared goods in ceremonies, or as food.

Sortal and relational classifiers often originate from directly possessed nouns and are themselves directly possessed nouns,^[27] called “possessed classifiers” by Aikhenvald (2000: 126); consider for instance:

In ZY, only phò- ‘load’ and pòi- ‘child’ are used independently as directly possessed nouns as in (26), (27); all other classifiers are only possessive classifiers.

Phò- refers to any kind of transported thing, including hunted game or fished animals; its semantics also include the share of goods brought at ceremonies and it is used as a deontic noun referring to duties and obligations (see Section 10). Their classifier and nominal functions are distinguished by syntactic context; a classifier has a possessive suffix and is followed by the adjoined possessed noun, like phò-la nõ [BO] (lit. cl:load-their fish) ‘their load/catch of fish’, while as a possessed noun phò is directly followed by the possessor as in phò kamyô [GO] ‘the load of the truck’ which denotes a container-contained relation.

Possessive and numeral classifiers are most generally distinct in ZY and NEL, except for phò- ‘load’ in ZY and fha-t ‘load’ in NEL, which have both functions (Bril 2014: 178).

Classifiers in ZY can be used anaphorically if the possessed entity is referential and contextually retrievable, in such a case, the possessee does not need to be mentioned. For instance, the construction in (28) can be reduced to êê-ny [PA] ‘my cuttings’ (for planting) if the referent is immediately retrievable.

In (29), the food type is known to be starch, but the specific referent is not mentioned since it is contextually referential. On the other hand, in the interrogative sentence in (30), the smell is identified as being that of meat or fish, triggering the classifier hò-, but the specific referent of the class is unknown and is precisely what needs to be identified.

This stand-alone property of classifiers is permitted by their nominal properties, they are the head and the superordinate term, while the adjoined possessee is the subordinate term. Besides, in the case of relational classifiers, the semantics of the classifier is the most important information to identify the use of the subordinate noun, as shown below with ê ‘sugarcane’.

An additional argument is that in some cases like (32), the possessive classifier is predicative and has an argument nhye hînu, the classifier and the possessee do not form an inseparable complex NP.

Most classifiers are in common use in ZY, except for the classifier of weapons p(h)ai-, which mostly occurs in story-telling, in relation to traditional weapons like spears, arrows and bows, sling-shot stones, clubs, axes, e.g., phai-ny bulaivi [PA] ‘my club’. The classifier can be used for modern weapons such as jige ‘gun’, e.g., pai-nu jige [GO] ‘my gun’, but jige is more commonly used either as an indirectly possessed noun, jige i je [GO] ‘his gun’, or as a directly possessed noun in [PA BO] with a different vowel, jigèle-n [PA], jigali-n [BO] ‘his gun’.^[28] Even jitua ‘bow’ is used as an indirectly possessed noun jitua i je [PA] ‘his bow’, as in NEL.

Thus, processes of language change trigger categorical changes. While the classifier for weapons phai- is falling into disuse, some classifier nouns undergo semantic extension like pòi- ‘child’,^[29] which also expresses ownership of domesticated animals or cattle pòi-ã vaaci [PA] ‘our cattle’, and extends to ownership of created objects such as photos, as in (32).

On the other hand, the possessive classifiers for food types appear to be the most stable.

Classifiers in ZY and NEL are not repeaters (auto-classifiers), but Iaai (Ozanne-Rivierre 1976: 191) has various repeaters that extend to the possession of other entities categorized in the same class.^[30] For instance, the Iaai noun uma ‘house’ is an auto-classifier in umwe-k uma ‘my house’ and a classifier for all types of houses, umwe-k ito, ‘my round house’; the noun nu ‘coconut tree’, has the same twin function, nuu-k nu ‘my coconut tree’ and also occurs with all cultivated plants, e.g., nuu-k koko ‘my yam plant’ (lit. cl-my yam); similarly for hu ‘boat’, huu-k hu ‘my boat’, and for all types of boats huu-k galu ‘my row-boat’.

3.3 Directly possessed free nouns

In contrast with bound nouns, free nouns can be used in their base form without any CST suffix and without possessive marking as in (33a). These nouns are directly possessed and are marked by possessives suffixes (33b) or by directly postposed possessor nouns; consider for instance: yaaza [GO] ‘name’, yaaza-nu [GO] ‘my name’, yaaza ẽnõ [GO] ‘the child’s name’.

3.4 Indirect, prepositionally possessed free nouns

The class of indirectly possessed free nouns is marked by prepositions, i (possibly from POc *qi, Ross 2001) for animate possessors as in (40), by (x)o, (w)o for inanimate possessors as in (41) and for indefinite/collective possessors. Similar constructions occur in other Oceanic languages (see Lichtenberk et al. 2011).

The greatest number of loan words, if possessible, belong to the class of indirectly possessed nouns, independently of their lexical meaning; most of them denote artifacts and goods (i.e., loto ‘car’, burei ‘bouteille’, mwani ‘money’, etc.). Some loan nouns for food and drink are possessed with the adequate classifier (kafe ‘coffee’, pomitee ‘potato’, phwalawa ‘bread’, aari ‘rice’, etc.).

Another type of indirect possessive construction is marked by the linkers ne ∼ ni and occurs with nouns mostly denoting parts-of-wholes with inanimate possessors. In ZY as in various other extreme-north Kanak languages, the possessive linkers are homophonous with locative prepositions ni ∼ ne ‘in, on’. This construction has been mentioned for some metaphorical body parts (see Section 3.3.2) such as ‘calf’ and ‘biceps’ literally meaning ‘the banana of/on his leg/arm’ (42) and indirectly marked with the possessive linkers ne ∼ ni.^[33]

These possessive linkers ne ∼ ni are further discussed in relation to alternate parts-of-wholes constructions in Section 5.3.

These linkers could reflect the POc associative linker *ni (Lichtenberk 1985a; Lynch 1996), or a former genitive *NV linker. Cognate forms are attested in various New Caledonian languages. Compare ‘the bark of the tree’ in three northern languages in (43): the nasal linker is reflected as -n in Nemi (NEM); in NEL, it is integrated in the head noun and marked by regressive nasalization of its final vowel; it is dropped in GO and PA, and the part-of-whole relation is directly marked.

Cognate possessive linkers are ne in Caac, la in Nyelâyu (44a), le in Nemi (44b), they are homophonous with locative prepositions;^[34] some of these morphemes host possessive suffixes.

In Cèmuhî (Rivierre 1980: 152–157), various possessive linkers, nè-, ko-, hê-,^[35] indicating inclusion and parts-of-whole relationship, host possessive suffixes. cèm nè- only marks relationships between inanimate nouns, as in ZY; Cèmuhî and Nemi ko ∼ xo ‘on’ mark indirect objects and possession, as in (45). The indirect possessive linkers (x)o, (w)o in ZY have the same origin, and similar functions, but do not host possessive suffixes and are restricted to inanimate possessors (see (41)).

5.1 Alternate direct and semi-direct parts-of-whole relations

Some directly possessed nouns have an alternate semi-direct construction denoting some parts-of-whole or some property of the possessee; they generally express relations of a more transient, more contingent nature than the direct possessive constructions, and they are marked by the suffix –a (with the possible insertion of a glide -(w) between adjacent vowels).

For instance, mõlò ‘life’ is directly possessed as mõlò-nu [GO] ‘my existence, my life’, while the constructions with -(w)a expresses a distinct type of relation and has a slightly different meaning, as mõlò-(w)a ‘way of life’ in (51).

Similarly the noun dou ‘envelope, moult’ is directly possessed as a body-part in dou-n [PA] ‘his (bodily) envelope’, and in dou pwaji ‘the crab’s discarded carapace’; but it is marked by the suffix -a in dou-(w)a hêgi [dowa hê^ŋgi] [GO PA] ‘the casing of the valuables’ referring to a man-made case. Also consider the meronymic relations in (52).

In NEL, the suffix –a is only used with non-human modifiers (Bril 2013), but Zuanga-Yuanga has no such restriction; the coding split with parts-of-whole is not conditioned by the possessor’s animacy, but by the kind of relation at hand.

5.2 Alternate semi-direct and classifier constructions for parts-of-whole relations

A few nouns have an alternate semi-direct construction or an indirect classifier construction. One such case is hênu [GO], hînu [PA] ‘image, shadow’ (POc *qanunu ‘shadow, reflection, soul’); its semi-direct construction with a possessive pronominal suffix or a nominal possessor, as in (53a) to (53d), often expresses subordinate, transient parts-of-whole relations or temporary images, reflections, shadows, shades; in (53d) the eye shade is temporary protection for the eye, distinct from the directly possessed body part ci-mee-nu [GO] (skin-eye-poss.1sg) ‘my eye-lid’.

On the other hand, hênu is possessed with the classifier noun pòi- which denotes some creation, an image taken by the photographer in (54).

These distinctions are common in Oceanic languages; in Motu (Lichtenberk 1985a: 109), a similar distinction is expressed with a classifier when some form of authorship is implied, and by direct possession otherwise.

5.3 Alternate semi-direct and indirect constructions of parts-of-wholes

A few nouns, mostly expressing parts-of-wholes, have alternate semi-direct and indirect possessive constructions with the possessive linkers ne ∼ ni (discussed in Section 3.4), expressing different relations and semantics.

Consider for instance the noun gu [GO], gun [PA, BO] ‘noise, din’; its indirect construction as gu ne chiò [GO] ‘the noise of the bucket’ (chiò ‘bucket’) refers to a noise caused by some external agent and involving the head noun. On the other hand, a noise inherently emanating from an entity and considered as a part of it, is marked with –a as in gun-a loto [GO BO] ‘the noise of the car’; it can also be directly possessed as in (56), but the form gunè-n might result from the accreted possessive linker ne hosting a possessive suffix as in other Kanak languages:

Accreted linkers are the source of stem allomorphies, some of which were discussed in Section 2.2.4; they are possible causes for changes of possessive constructions and possessee classes.

Metaphorical body-parts were mentioned to be indirectly marked by a linker ne ∼ ni in Section 3.4. Another case is that of wa [GO], wal [PA] ‘liana, rope, band’, whose semantics extends metaphorically to body-parts meaning ‘tendon, artery, vein, muscle’, indirectly marked by ne ∼ ni as in (57a)–(57b). This stands in contrast with directly possessed compound nouns denoting artefacts or body-ornaments, in which wa(l) has its basic meaning ‘rope, band’ as in (58). Metaphorical relations are thus marked as more distant, while non-metaphorical meanings occur in directly possessed compound nouns.

Similar alternate constructions occur in NEL, between indirectly marked metaphorical body parts in (59a) and the semi-direct possession of body ornaments in (59b).

On the other hand, it is directly possessed when used with its basic sense of ‘rope, liana’:

To summarize, these alternate direct, semi-direct and indirect constructions and their semantics only occur in parts-of-whole relations and are restricted to a few nouns. They show some correlation between their formal coding and the semantic feature of the relations at hand. The semi-direct constructions with the accreted suffix -a tend to express contingent, transient part-of-whole relations, while the indirect constructions with the prepositional linkers ne ∼ ni express extraneous and metaphorical part-of-whole relations.