Adnominal possession in the languages of Wallacea: a survey

2.1 Basic possessive word order

The basic order of possessor and possessum throughout much of Linguistic Wallacea sees the possessor precede the possessum. A typical example of Psr-Psm order comes from Wamesa, a member of the South Halmahera-West New Guinea branch of Austronesian spoken along the south-western coast of Cenderawasih Bay. Wamesa has three main possessive constructions, all of which see the possessor preceding the possessum: a direct, morphologically marked inalienable construction for most body parts and kin terms (1a); an indirect, default or alienable construction using the verb ne ‘to have’ as a linker (1b); and a third construction with no possessive marking used primarily for part/whole constructions (1c).

While the Psr-Psm order is typical of all Papuan languages, not just those within Linguistic Wallacea (“the recognised Papuan order” as Reesink 1999: 81 calls it), it is not the typical order of possessor and possessum in Austronesian. For most languages of the Austronesian family, the possessor follows the possessum in the basic possessive construction. It is only in the vicinity of New Guinea that we find a concentration of Austronesian languages that “reverse” the typical ordering to Psr-Psm (Map 3). It is considered so important within the family that Himmelmann (2005) names one of his two major non-Oceanic Austronesian language types “preposed possessor languages.”

Map 3:

Basic ordering of possessor and possessum in the New Guinea area (adapted from Schapper 2020a: 488).

The reversal of the basic possessive order in Austronesian languages is widely agreed to be the result of contact with or shift from Papuan languages (Donohue 2007: 352–355). East of New Guinea it is only the Austronesian languages closest to the New Guinea mainland that have the Papuan possessive order (see inset of Map 3). But the change in possessive ordering in Austronesian languages of Linguistic Wallacea is striking for how far it extends from New Guinea, and speaks to the wide-ranging Papuan influence that underpinned the formation of the linguistic area.

It is important to note that the reversal of possessive ordering in Austronesian languages within Linguistic Wallacea was not a one-time change that occurred in a common ancestor. Rather, it is clear from the distribution of the order over established subgroups that it has erratically diffused into the Austronesian languages of the area. For example, the languages of the Sumba-Hawu subgroup, though they fall within Linguistic Wallacea based on other features, display the conservative Austronesian postposed possessor order, while their closest relatives in the Flores-Lembata subgroup (Fricke 2019: 229–233) have a mixture of preposed and postposed orders (discussed further in Section 2.2.1). At the same time, the preposed possessor order has diffused beyond the boundaries of CEMP languages. Banggai, spoken at the western perimeter of Linguistic Wallacea and subgrouping with western Austronesian languages in Sulawesi, also has preposed possessors (Schapper and McConvell forthcoming).

2.2 Splits in possessive word orders

The basic Psr-Psm order of possessive constructions in Linguistic Wallacea is well known in the literature. What is little discussed are the numerous deviations from this pattern whereby languages may allow the possessor to be postposed in certain contexts. As can be seen in Map 2, most languages with splits in possessive word orders are geographically clustered, which is potentially indicative of local processes of language contact and change. Here we will discuss the splits in possessor posing in Linguistic Wallacea under two broad labels: splits in possessor posing involving nominal versus pronominal possessors (Section 2.2.1); and splits conditioned by the semantics of the possessive relationship (Section 2.2.2).

3.1 Possessum-marking

While possessum-marking is the norm in Linguistic Wallacea, a major divide between Austronesian and Papuan languages is found with respect to where these markers attach to the possessum. In Papuan languages, bound possessive markers are overwhelmingly prefixal or proclitic. In Austronesian languages, bound possessive markers are typically suffixal or enclitic. In both Papuan and Austronesian languages, bound possessum-marking forms are inflectional, agreeing in person and often number with the possessor, typically in the inalienable possessor construction. Map 4 shows that this division is attested quite consistently across much of the region.

Exceptions to the Papuan prefixing pattern are found in Bomberai languages such as Kalamang (already illustrated in 4) and Iha (15). The aberrant suffixal pattern is not found in Iha’s closest relative, Mbaham (16). This seems to be because the Mbaham prefixes are innovative forms originating in personal pronouns marked with the genitive case marker -man (see Section 3.2), which then morphologized, i.e., tam- < *ta-man ‘2sg-gen’.

The dominance of suffixal markers for possessors in the Austronesian languages is a continuation of the pattern from their ancestor, Proto-Malayo-Polynesian (PMP), whereby pronominal enclitics indexing the possessor, typically known as genitive pronouns or genitive enclitics in the Austronesianist literature, bind to the possessum. Across Linguistic Wallacea, reflexes of the PMP genitive clitics grammaticalized into inflectional suffixes for possessors in inalienable possessor constructions. Deviations from this pattern are largely a result of further grammaticalizations.

In two regions of southern Linguistic Wallacea, we find the emergence in Austronesian of prefixal or proclitic markers for possessors, either in addition to suffixes or in place of them, in two regions. The first is in the languages of Ambon and western Seram, where the majority pattern is for suffixes to be used for the possessor of an inalienable noun (17a) and prefixes to be used to encode the possessor of an alienable noun (17b).

In a few languages nearby, possessive suffixes have been lost and possessors for all nouns are encoded with prefixes. Larike is an example of a language of this kind (18).

See Florey (2005) for a detailed account of the various processes of affixal decline and loss in Central Maluku possessive constructions.

The second region involves the languages of Southwest Maluku and eastern Timor. As can be seen in Table 1, these show a cline of movement away from possessive suffixes towards prefixes by way of preposed or proclitic possessive markers (Schapper and Zobel forthcoming). In Leti, possessive suffixes for 1st person plural clusivity specifications have collapsed together with 3rd person singular. They are now disambiguated by an additional proclitic possessive marker that is used alongside the possessive suffixes. Closely related Luang has extended the use of the proclitics to all persons, including making their use obligatory for person-number combinations where no ambiguity in the possessive suffixes exists. Tugun has moved further away still from possessive suffixes, with only one suffix -n ‘3sg’ used on a small number of originally inalienable nouns where they have a non-human possessor (Hinton 1991: 70–71). Otherwise, all nouns in Tugun have their possessor encoded with a preposed free possessive marker, some of which have proclitic or prefixal forms (Hinton 1991: 67–68). Dadu’a represents the logical end point of the loss of possessive suffixes and the move to prefixal possessive marking, with all nouns having their possessor encoded with a possessive prefix.

In northern Linguistic Wallacea, very few Austronesian languages mark possessors by means of a suffix alone. Complex patterns of possessive affixes, many of which represent innovative forms rather than PMP reflexes, cross-cut areal and genetic subgroupings. Van den Berg (2009: 346–347) reconstructs only possessive suffixes for Proto-SHWNG, suggesting that what he calls the “bewildering variety of forms and positions” (p. 342) found in the present-day SHWNG languages emerged only after the diversification of the family with a multitude of independent, parallel innovations, many of which may be attributable to contact with Papuan prefixing-type languages.

Languages that retain the suffixing type are Moor, Umar (19a), and Maˈya (19b). Patani uses suffixes with optional proclitics. (See Rødvand, this volume, for discussion of Patani.) Sawai (19c) has been reported by some sources (Kamholz 2014: 131–132; van den Berg 2009: 332) to use a combination of prefixes and suffixes, and by others (Whisler 1996: 48; Whisler and Whisler 1995: 663) to only use suffixes; this may indicate dialectal variation. While the suffixing type is conservative, the affixes themselves vary: Those found in Raja Ampat and South Halmahera paradigms are retained from PMP, but the Moor and Umar forms are each independently innovated. In some Maˈya dialects, suffixation is accompanied by tone loss (Remijsen 2001: 155). Superscripts indicate tone.

The Biakic and Yapen languages, as well as Ambel, Buli, and neighboring, non-SHWNG Irarutu, have progressed to using a combination of prefixes and suffixes or enclitics. While the systems are quite similar, the forms themselves are again mostly non-cognate – this is another case of multiple parallel innovations. Table 2 gives singular and plural paradigms for Wamesa and Biak. In Wamesa, the enclitics used with first- and third-person singular possessors are actually determiners, and indicate (literal or metaphorical) distance rather than strictly possessor person. (See Gil, this volume, for further discussion.) Biak has distinct inalienable possessive paradigms for kin terms, paired body parts, and non-paired body parts; the one given here is for non-paired body parts. The non-singular prefixes are identical, or nearly so, to the subject-marking prefixes that appear on verbs (see Section 5.1), and mostly non-cognate between branches.

Ambel uses a combination of a suffix and high-tone suprafix with singular possessors, and person/number-marking prefixes combined with a possessive suffix –n with non-singular possessors (3rd person inanimate possessors are indexed by a prefix i- and no suffix, regardless of number). Ambel has several subtly different direct possessive paradigms, one of which is given in Table 3 for the root yói ‘heart’. Diacritics represent tone.

The SHWNG languages of southern and eastern Cenderawasih Bay (except Moor), Kurudu, Yaur, Yerisiam (20a), Waropen (20b), and Warembori (20c), have all lost their suffixes and only use prefixes to mark possession. Given their position, it is unlikely that any of these can be traced back to the PMP possessive enclitics; many are transparently repurposed verbal subject markers. Yaur and Yerisiam are closely related, and their nearest relative, Umar, is strictly suffixing. The other prefixing languages belong to an additional three primary subgroups of SHWNG.

Finally, the Raja Ampat language Matbat has dispensed with prefixes and suffixes altogether and relies on a combination of infixes, vowel alternations, and tonal changes on the possessum, which index the person and number of the possessor (with extensive syncretism). The infix appears after the penultimate vowel of the root, or after the single vowel of monosyllabic roots. Only the 1pl.incl form is considered to be a retention, reflecting what used to be a possessive suffix. Table 4 gives paradigms for two Matbat inalienable roots. Superscript numerals represent tone. The inflections shown here appear along with proclitics indexing the possessor, as in (21).

The diversity of forms within SHWNG is unique in Linguistic Wallacea, where most language groups show more internal consistency and developments can be accounted for by discrete innovations in a single branch. In SHWNG, parallel innovations, including the loss of possessive suffixes and the development of possessive prefixes, often by repurposing verbal agreement morphology, occurred independently multiple times throughout the tree, leading to disparate structures in even closely related languages. While contact with and shift from Papuan languages may well provide a partial explanation, the full story is necessarily more complex, and begs further research.

3.2 Possessor-marking

Only five languages that we know of in Linguistic Wallacea have constructions in which the possessor is the locus of possessive marking. In these, it is never the only possessive construction in a language but rather always exists alongside possessum-marking constructions.

Most commonly, the possessor-marking construction is required in certain contexts. We saw already in (10b) that Maybrat makes use of possessor-marking for postposed alienable possessors, but that the inalienable construction is possessum-marking. Iha (22a) has a possessor-marking enclitic and Mbaham (22b) a possessor-marking suffix that are used when the possessor is encoded by a full noun (as opposed to possessum-marking person inflections seen in (15) and (16) above). The Iha and Mbaham markers are described as genitive case markers and each is part of a larger system of case markers.

In Sula, whilst 1st and 2nd person possessors are encoded by means of a free pronoun preposed to the possessum noun (23a), a 3rd person possessor, be it nominal or pronominal, must be marked by an enclitic =in (23b). This enclitic likely originates in the PMP genitive case marker *ni.

Moor, an Austronesian language of southern Cenderawasih Bay, has a unique form of split possession marking that coincides – as in Maybrat – with possessor posing. The most frequent possessive construction involves a preposed possessor and a suffix -ío on the possessum signaling the possessive relationship (24a). A second, less common structure has a postposed possessor marked with the suffix -ìjo and an obligatory article on the possessum (24b).

4.1 Lexicalization and the semantics of inalienability

The range of nouns that are classified inalienable varies from language to language. Nouns denoting body parts are the most typical members of the inalienable class, though frequently not all members of the semantic field will be included. In Wooi, for example, most body parts and products are inalienable, but huhu ‘breast’ and rerawa ‘skin’ are alienable (Sawaki 2017: 140). Kin terms are classified as inalienable in many languages, but in Papuan families of the area in particular it is frequently observed that only a small number of kin terms belong to the inalienable class. For example, in Bunaq, less than a third of kin terms are inalienable (Schapper 2022: 246–247), while in Yawa only ‘mother’ and ‘father’ are inalienable (Jones 1993). Kin terms are also particularly prone to irregular marking or suppletion in their possessive paradigms (Baerman 2014; Schapper and McConvell forthcoming). Many languages additionally treat associative nouns (e.g., ‘voice’, ‘name’, ‘breath’, ‘shadow’, ‘feeling’, ‘friend’), locational nouns (e.g., ‘top’, ‘side’), and parts of wholes (e.g., ‘edge’, ‘fruit’) as inalienable. In some languages, nouns for culturally important items, such as ‘brideprice’, ‘canoe’, and ‘netbag’ are also inalienable, though these are far more variable from language to language.

Lexical classes of inalienable nouns, while generally strict as to their membership, can nonetheless allow some leeway in the possessive marking of some nouns. In our experience, however, the flexibility is usually very limited and does not take away from the basically lexicalized nature of the possessive classification system. (See Grimes’ 1991 description of Buru possession as a fluid possessive classification system for a possible counter-example.) In some languages, a small set of nouns may be used with either alienable or inalienable constructions depending on the semantics of the possessive relationship. For example, in Ambel (Arnold 2018: 320–321), some body, plant, and animal parts appear in inalienable constructions when they are a part of the possessor (e.g., “my bone, which is part of my body”) and in inalienable constructions when the relationship is one of ownership (“my bone, which comes from a pig and which I have in my house”). In some cases, the choice of construction affects the interpretation of the meaning of the possessed noun, for example making the difference between ‘voice’ and ‘story’. (See also Section 4.5 for flexibility with regards to edible possession.)

Elsewhere, categorial leeway may be due to language shift. For example, only a limited set of nouns may be inalienably possessed in Wamesa as it is currently spoken, but any noun, even classic inalienables such as close kin terms and basic body parts, may appear in an alienable construction with no change in meaning (25), and occasionally alienable and inalienable constructions are combined, suggesting lexicalization of the inalienable form and bleaching of its possessive meaning for some speakers. This flexibility is not attested in earlier sources (Saggers 1979; van den Berg 2009), and appears to be a result of declining language use. Florey (2005) describes a similar grammatical restructuring accompanying language shift in the Austronesian languages of Seram in central Maluku.

A handful of languages have “double” possessive marking strategies that allow morphologically bound, inalienable nouns to occur with a free possessive marker normally used with alienable nouns in contexts where the possessum has been “alienated” from its possessor. An example of this from Blagar is presented in (26). A similar structure is described for the related Papuan language Bunaq (Schapper 2022: 352–354).

Such cases underline that in many languages, boundness is a strict morphological property of the lexical class of inalienable nouns, irrespective of semantics (cf. the languages of Cenderawasih Bay where this strictness does not exist, Section 4.2.2).

4.2 Direct versus indirect possession

The most widespread pattern for encoding alienability distinctions in both the Papuan and Austronesian languages of Linguistic Wallacea involves contrastive direct/indirect possessive constructions: Inalienably possessed nouns have their possessors marked “directly”, with the person and number of the possessor affixed on the noun expressing the possessum, while alienably possessed nouns have “indirect” possessors indicated by free possessive markers (pronouns, verbs, or particles), often inflected for the person and number of the possessor. An example of the direct/indirect possessive contrast is given for a Papuan and an Austronesian language in (27) and (28), respectively.

Cross-linguistically, contrastive direct/indirect possessive constructions to encode inalienability are rare. They are concentrated in two main clusters: Wallacea and Oceania on the one hand, and South and Central America as well the southern part of North America on the other hand (Bugaeva et al. 2021). The concentration of languages with the contrast within Linguistic Wallacea points strongly to it being an areal feature.

Early claims of a genetic explanation for the wide presence of direct/indirect possessive contrast in eastern Austronesian languages cannot be sustained. The absence of the distinction or even its remnants in the Austronesian languages of south-western Linguistic Wallacea and its presence in at least one genetically “Western” language, Banggai (Schapper and McConvell forthcoming) undermine Blust’s (1993: 259) reconstruction of the feature to his Central-Eastern Malayo-Polynesian subgroup of Austronesian. Donohue and Grimes (2008: 142) further show that the diversity of forms by which the contrast is marked in the Austronesian languages of southern Linguistic Wallacea, the putative CMP languages, points to erratic diffusion, rather than inheritance. This point is later conceded in Blust (2009). In the north of Linguistic Wallacea, however, it is widely agreed that the direct/indirect possessive contrast entered the Austronesian languages at an early stage. Van den Berg (2009: 353–354) argues that it was already present in Proto-Eastern Malayo-Polynesian (EMP), the hypothetical ancestor of Proto-SHWNG and Proto-Oceanic spoken in western New Guinea.

The geographic area in which the direct/indirect possessive contrast first appears in Austronesian languages is almost exactly the same area in which Papuan languages displaying similar systems are found. This strongly suggests that the emergence of the direct/indirect possessive contrast in the Austronesian languages was a response to regional norms within Linguistic Wallacea. Donohue and Schapper (2008) point out that the contrast between direct possession for inalienable nouns and indirect possession for alienable nouns can be reconstructed to Proto-Timor-Alor-Pantar and that the presence of such a feature at an early time in a Papuan family in the vicinity of where Austronesian languages first begin to display such typologically unusual systems is hardly likely to be chance. Donohue and Schapper trace the origin of direct/indirect possessive contrasts in Austronesian to Papuan influence, arguing that it arose through calquing Papuan constructions alongside more general pressure from Papuan languages to use preposed possessors (discussed in Section 2). Van den Berg (2009) agrees that Papuan contact was likely responsible for the direct/indirect possessive contrast entering Proto-EMP.

A final point to emphasize is that the direct/indirect contrast crosses genetic lines amongst the Papuan languages in Linguistic Wallacea, and includes members of the East, West, and South Bird’s Head, TAP, Mairasic, and Yawa-Saweru families as well as Papuan isolates such as Mpur. The cross-family presence of this geographically-limited feature supports arguments that Linguistic Wallacea is an ancient linguistic area whose formation predates the Austronesian dispersal (Donohue 2004a; Schapper 2010, 2015). This means also that the direct/indirect possessive contrast in the Austronesian languages of Linguistic Wallacea should be traced back not to a single Papuan source but to several.

Subtopics relevant to the direct/indirect possessive classification are discussed in the following sections. Section 4.2.1 looks at historical changes to direct/indirect systems that give rise to different affixal paradigms marking alienable and inalienable nouns. Section 4.2.2 discusses three types of zero-marked possessive constructions.

4.3 Possessive classification without inflectional difference

Possessive classification without inflectional difference is where the possessive markers for the alienable and inalienable classes of nouns are the same. The difference between classes lies in that inalienable nouns must occur with a possessor expressed in some form, whereas alienable nouns can occur without any expression of a possessor. A typical statement linking the expression of a possessor with (in)alienability is found in Bowden’s (2001: 266–267) description of Taba, an Austronesian language of South Halmahera: “[a] differentiation between alienable and inalienable possessive categories is not obligatorily marked by the use of different forms in Taba, as is common in many closely related languages. However, some of what could perhaps be called the most ‘inalienable’ kinds of possessive relationships (e.g., expressions referring to part-whole relationships) are distinguished in Taba by obligatory possessive marking.”

While a difference in the obligatoriness of the grammatical possessive marker is the most typical form taken by this kind of possessive classification, it is not the only type. Members of the “second” inalienable class of nouns in the languages of Alor (discussed further below) obligatorily express a possessor in one of two ways. Either they must occur with a possessive inflection with the same form as used by an alienable noun, or in a possessive compound where the possessor is expressed with a preceding noun. These two constructions are illustrated on the basis of Wersing in (37).

In the same vein, van Staden (2000: 125–126) describes some differences in the licensing of person-number prefixes and the reflexive prefix ma- as demarcating the difference between alienable and inalienable nouns in Tidore.

For many cross-linguistic definitions of possessive classification, differences in the obligatoriness of the expression of a possessor without inflectional difference such as described here would not be counted as a form of possessive classification. However, within Linguistic Wallacea, recognizing this obligatory/optional possessive contrast as a form of possessive classification is important for understanding how possessive classification systems change and adapt in the context of paradigm loss. In Linguistic Wallacea, languages with obligatory possession can be found in concentrations in North Halmahera and in the region around eastern Timor. In both regions, the obligatory/optional possessive contrast is shared across proximal Papuan and Austronesian languages, suggesting a role for contact.

Languages with an obligatory/optional possessor contrast are those where either direct or indirect possessive markers are synchronically absent, typically as the result of diachronic loss of one set or the other. In the case of Taba, the direct possessor affixes found in its near relatives have been lost. The modern Taba possessive marker ni is used obligatorily with inalienable nouns and optionally with alienable ones. It is a reflex of the indirect, general possessive classifier that has cognates in its relatives in South Halmahera (Bowden 2001: 233; see Section 4.5 on general possessive classifiers). Nearby in the Papuan languages of North Halmahera, comparison with their likely relatives on the West Bird’s Head suggests that the obligatory/optional possessive contrast has arisen through the loss of the direct possessive prefixes that are still found in West Bird’s Head languages. A similar situation is described by Schapper (this volume, Section 3) for the Timor-Alor-Pantar languages of eastern Timor. In these, the direct/indirect contrast has been lost due to the loss of reflexes of the Proto-TAP direct possessive prefixes. Instead, the original indirect possessive markers have expanded and are now used for nouns denoting both inalienable and alienable nouns, but while inalienable nouns are obligatorily possessed, alienable ones are only optionally so.

The loss of original indirect possessive markers appears to have occurred in other languages where the same set of affixes is used for both alienable and inalienable nouns, with the already described differences in obligatoriness in the appearance of the possessive affix. For instance, all nouns in Kalamang can take a possessive suffix, but one is obligatory on a small set of kin terms and a larger set of nouns used in part-whole relations (Visser 2020: 160–162). Although little is known about the morphological history of West Bomberai languages, it seems likely that these too originally had preposed indirect markers for possessors of alienable nouns ‒ reflected now as prefixal possessive morphology in Mbaham (see Section 3.1), but that these have been lost in Kalamang.

In some languages, however, it is not clear that a direct/indirect contrast was ever present. For example, there is no evidence that Leti or that its relatives in the Leti-Luang subgroup of southern Maluku ever had indirect possessive markers. Still, Leti displays the obligatory/optional form of possessive classification, in this case by means of a possessive suffix on the possessum noun. There are no nouns that cannot host a possessive suffix: Body parts and other nouns denoting typical inalienables are always inflected for a possessor (e.g., iran-nu nose-3sg.poss ‘his/her/its nose’; but *iran), whereas nouns denoting alienables such as vatu ‘stone’ are not required to, but can (e.g., vat-nu stone-3sg.poss ‘his/her/its stone’; vatu ‘stone’).

The repeated (re-)creation of possessive classification systems on the basis of an obligatory/optional contrast, both where paradigms of earlier indirect/direct systems were lost and where no earlier possessive classification system was known to have been in place, speaks to the significant areal pressure exerted on languages in Linguistic Wallacea to differentiate alienable from inalienable nouns. In fact, this pressure can also be observed in languages that still have (extensions of) earlier indirect/direct systems. In the languages of Alor, for example, we see the emergence of a third possessive class from what seems to be a new cycle of grammaticalization of possessive markers on nouns denoting “inalienables”. Illustrated in Schapper (this volume, Section 3), these languages have a small closed class of nouns which are obligatorily possessed like inalienable nouns, but their possessor is expressed with an alienable possessive prefix. This class exists alongside the set of inalienable nouns and the alienable nouns whose possessors are marked with different paradigms of possessive prefixes. Similarly, in Yawa, Jones (1993) describes the direct possessive construction on inalienable nouns denoting body parts and the kin terms ‘mother’ and ‘father’ (38a), while the indirect construction, using possessive pronouns, is considered alienable and applies to all other nouns (38b). However, possessive marking is obligatory with all kinship terms (outside of direct address), even those whose possessor is encoded indirectly (38c).

Such examples could be taken to suggest that splits of inalienable nouns into different inflectional classes, including potentially those described by Bach (2018: 177–195) and Arnold (this volume), might have arisen through different cycles of grammaticalization of possessive markers to form bound possessive classes with different nouns denoting inalienables at different times.

4.4 Free possessive linkers

A minor strategy for encoding alienability contrasts in Linguistic Wallacea involves a free uninflecting possessive linker between (pro)nouns expressing a possessor and a possessum. As is cross-linguistically common, the possessive linker only appears between a possessor and a possessum where the possessum noun is classified as alienable. These linkers are not inflectional and express no person/number information about the Psr or Psm. This strategy is only known from the Papuan language Abun and the Austronesian languages of the Central Flores subgroup.^[9]

In Abun, the possessor of inalienable nouns is introduced without a marker of possessive relationship (39a), as in the juxtapositional zero-marked constructions already discussed in Section 4.2.2. With alienable nouns, the possessor is introduced with an invariant free possessive linker bi (39b).

The Central Flores languages also use free possessor linkers to mark alienable nouns. For Keo, the best described Central Flores language, the possessive linker is used optionally to introduce alienable nouns without any obvious change in meaning based on the presence or absence of the linker. With inalienable nouns the possessive linker is typically not used, but where it is there is a change in the meaning of the possessive relationship. Consider the examples in (40). Baird (2002: 217) writes: “When the ko[’o] construction is used the speaker appears to focus on the relationship between the two entities, that the possessed actually belongs to or is part of the possessor. Whereas when the juxtaposed construction without ko[’o] is used, the emphasis is on the object as a ‘whole’ entity rather than as expressing a possessive relationship.”

It is worth noting that Abun and the Central Flores languages are the most isolating of all the languages in the region. While most other languages mark arguments on verbs and some possessors on nouns with bound person/number inflections (cf. Section 5.1 on the isomorphism of these paradigms), these languages are entirely devoid of such inflectional morphology. There is, potentially, a correlation between the lack of inflection and the use of free optional possessive linkers for differential possessive marking in these languages. Areal pressure to differentiate alienable and inalienable nouns may have led these two groups, in parallel, to develop possessive classification by means of a possessive linker.

4.5 Edible/general possessive contrasts

A final pattern of differential possessive marking involves a two-way contrast between edible (in some sources referred to as “alimentary”) and general possession in the domain of alienable possession found in Austronesian. This feature is prominently described in the literature on Oceanic languages (e.g., Lichtenberk 1985), but it is also present in the three South Halmahera languages Sawai, Patani, Buli (see Rødvand this volume) as well as Waropen, spoken in Cenderawasih Bay, and the putative CMP language Selaru. Remnants of a previous edible/general contrast may be visible in the Raja Ampat language Ambel (Arnold 2018: 298). The widespread pattern is for both the general and edible possessive classifiers to be indirect markers of the possessor, as seen with Buli in (41).

The choice of possessive classifier is often flexible depending on the specific relationship between possessor and possessum (cf. Lynch 2001). In Selaru, for example, the general possessive classifier wasi indicates that the pig is not food, but rather is either still alive or not yet prepared as food (42a), while the edible possessive classifier hina indicates that the pig is already considered food, i.e., it has been killed and cooked (42b).

These classifiers denote more than simple edibility. Lexicalization also appears to play a role in such indirect possessive classification systems. Edible rather than general possession is still used in Selaru for cash crops that are not intended to be eaten and in some cases are not even edible (Coward 2005: 52). In other cases, the use of a noun in association with eating or food is enough to trigger use of the edible possessive classifier in Sawai (Kamholz fieldnotes, cited in Gasser et al. forthcoming) and Patani (Rødvand this volume), as with gardens or plates. In Sawai, for example, whether a possessed noun takes the edible classifier depends on context: meja ‘table’ is marked with the edible classifier when it refers to a dining table, but not when it refers to a school table. The edible classifier is also used with hand implements in Sawai (Whisler 1996), as in (43).

While it is best known from languages of the Oceanic subgroup of the Austronesian family, where it reaches particular prominence in those languages’ grammars, the edible/general possessive contrast first appears in Austronesian languages in Linguistic Wallacea. It is therefore tempting to regard it, like other innovative possessive constructions in the region, as a result of Papuan contact. However, there are no known Papuan languages that have such a semantic contrast in their possessive structures. As Schapper (forthcoming) points out, it is not legitimate to simply attribute any innovation in the vicinity of Papuan languages to contact with them or shift from them; specific argumentation is required as to the Papuan origin and directionality of transfer in such cases. Nonetheless, in favor of a Papuan transfer scenario is perhaps the fact that the edible/general possessive contrast appears to have arisen multiple times within the Austronesian family within Linguistic Wallacea. The Selaru classifiers are not cognate with those found in SHWNG. The South Halmahera edible classifiers are cognate with those in Waropen, but cognate forms are also found across SHWNG in languages lacking an edibility contrast, used for basic indirect possession. An edibility contrast could plausibly be inherited from Proto-SHWNG, and since lost in the majority of languages, and indeed this is what van den Berg (2009: 349) hypothesizes. According to van den Berg’s (2009) reconstruction, however, the Proto-SHWNG edible possessive classifier *na is cognate with the Proto-Oceanic general possessive classifier *na, while the SHWNG general possessive *ri and the Proto-Oceanic edible classifier *ka are unrelated. The lack of form-function matches for items encoding the edible/general possessive contrast across CMP, SHWNG and Oceanic strongly points to parallel innovations.

5.1 Isomorphism of possessor and verbal inflections

It is common worldwide for the morphemes that index the person/number of possessors to match those used for verbal agreement. In her survey of 270 genetically and geographically diverse languages, Siewierska (1998: 11, 17) finds that of those that exhibit person marking on both possessive constructions and verbs, 83% show full or partial correspondence between the possessive and (some set of) verbal agreement markers: 43% show an exact match, 9% are mostly but not fully identical, and an additional 31% show some similarity. In only 17% of the eligible languages in her samples was there no correspondence between the two categories of markers. Identity of possessive and verbal affixes appears variably in Linguistic Wallacea, with areal sub-patterns. (See also Sawaki 2021.)

Where languages with possessive/verbal identity have multiple verbal agreement paradigms, the possessive affixes generally coincide with paradigms used for patients, experiencers, or subjects of stative verbs rather than agents. In the Austronesian languages of Aru, possessive suffixes match partly with those of stative verbs, but not active ones (Schapper fieldnotes). In Yawa, inalienable prefixes match the subject markers on a small class of stative intransitive verbs (Jones 1986: 44); Saweru inalienable prefixes also mark unaccusative subjects or the object of a transitive verb (Donohue 2001: 328, 2004b: 3–4). In Mairasi, there are roughly twenty different sets of verbal agreement markers; the inalienable possessive prefixes match the set used with certain psychological, perception, intransitive, and stative verbs, as well as to index indirect objects (Peckham 1982: 80–81). In the languages of Alor, both the alienable and inalienable possessive prefixes coincide with verbal prefixes indexing Subject or Patient roles, but not Agents (Schapper fieldnotes).

This tendency breaks down in two scenarios. Where the verbal paradigms distinguish purely morphological classes, which one is repurposed for possession is purely arbitrary, rather than traceable to syntactic/semantic principles. In Warembori, Donohue (1999: 24, 26) gives two (nearly identical) sets of possessive affixes and five (very similar) verbal paradigms; possessive Set II is identical to verbal Set V, and possessive Set I differs only in that it takes its 3sg prefix from verbal Set IV. In other cases, there may be more possessive paradigms than verbal ones. In Meyah, the possessor prefixes on body part terms are the same as those marking verbal subjects, but the kinship terms have a different set of prefixes for singular possessors (Gravelle 2002: 125–126, 131; see also Arnold, this volume, on split inalienables).

In New Guinea, Mansim, Maybrat, Mpur, Tehit, and Waropen are described as straightforwardly identical in their inalienable and verbal subject-marking affixes (Dol 2007: 49–51, 68, 84; Flassy 1991: 16, 37–38; Held 1942; Odé 2002: 52; Reesink 2002a: 287). This is also the case for basically all of the TAP languages, with the exception of Eastern Timor languages, which have largely lost their original person inflections on verbs (Schapper 2020b: 397–400; Schapper and Zobel forthcoming). Full identity is exemplified here in Table 5 by Mpur, a Bird’s Head isolate, where the prefixes used to mark the person, number, and gender of the possessor on inalienable nouns are the same as those used to index the subject on the class of verbs to which jap ‘stay’ belongs.

Partial identity frequently results from morphological restructuring of the possessive system, differential morpho-phonological rules, or sound change leading to a breakdown of similarity. The SHWNG Yapen languages are an example of the former. Here, non-singular possessors are marked with a prefix on the possessum identical (or nearly so) to the verbal subject indexing prefixes, plus an invariant possessive suffix or suffixes. Singular possessors, however, use a totally different system, relying solely on suffixes and/or enclitics. These languages inherited a suffix-only system from Proto-SHWNG; the addition of the prefixes with non-singular possessors is an innovation. The same is true of the Biakic languages, where for example in Biak the non-singular possessive and verbal subject-marking prefixes are identical when the possessum is a body part, but slightly different otherwise. (See also Section 3.1.) Examples from Ambai are given in Table 6. Silzer (1983) does not give full paradigms; therefore the 1pl.incl possessive form is given for ‘head’ instead of ‘hand’, and the 3du form for ‘sit’ instead of ‘paddle’.

In Inanwatan (de Vries 2004: 29, 36), it is primarily the morpho-phonological processes that apply to the affixes which differ between verbal and possessive prefixes, rather than the affixes themselves. The prefixes used for inalienably possessed nouns are very similar to those used for verbal object marking, but only the possessive prefixes undergo vowel elision before certain roots.

Further breakdown of identity is found for example in Sougb, shown in Table 7. Here the 1pl.excl prefix is the same between paradigms, and several other forms show some similarity, while yet others are quite different (Reesink 2002b: 206–7, 218). These forms suggest a common historical origin, with erosion of similarity, including limited morphological replacement, over time.

Languages with no identity of possessive and verbal agreement affixes of course include those where one or the other category is absent, as in Kalamang, Abun, the North Halmahera languages, and the Austronesian languages of eastern Timor. Another common type uses prefixes for one category and unrelated suffixes for the other, as in Moor, some SHWNG languages of Raja Ampat and South Halmahera, and many Timor languages. In northwest New Guinea, these two sorts of languages account for all attested instances where no connection is evident. Elsewhere disparate paradigms do exist, for example in the South Halmahera language Buli (Table 8), where the possessive affixes resemble the possessive and verbal subject markers found elsewhere in SHWNG, but the verbal subject prefixes are different.

In Hatam, the possessive/verbal subject markers are further extended to a special adjectival construction. Here, adjectives may coindex the noun they modify via a prefix marking person and number (44a). Reesink (1999: 57) reports that nouns marked in this way convey “a more permanent quality” than their unmarked counterparts (44b). (This differs importantly from the construction described in Section 5.3 in that the Hatam forms may be used both attributively and predicatively.) This prefix is identical to the paradigm used for inalienable possession (44c) and matches the verbal subject agreement paradigm for all cells except 3sg, which is null for verbs.

5.2 Identity of possessives with relatives, verbalizers, reifiers, and case marking

In the Biakic (SHWNG) languages of New Guinea, there is a morpheme ve that functions as a possessive marker, a relativizer, a verbalizer, and what Gil (2017) calls a reifier, which attributes the properties of one lexical item to another. Ve encompasses a broad range of other functions as well, including as verbs meaning ‘do, give, say, want, become’, dative, allative, future tense, etc., and appears with a subset of those functions in several nearby languages. (See Gil 2017 for a full description.) The reifying function in Biakic is similar in some ways to the attributive functions of possessors in TAP languages (see Section 5.3), but the attributes involved can stem from more parts of speech than just adjectives. Instances of multifunctional ve- are attested in Roon (Gil 2017, this volume), where for example the word for ‘snake’ is ro- ve- kwan nmlz-VE-long ‘thing that is long’, as well as in Dusner (Dalrymple and Mofu 2012), and in Biak (van den Heuvel 2006: 179–184). (VE- is also found in a number of other SHWNG languages with a range of grammatical functions, but its possessive use is limited to Biakic.)

The examples in (45) come from Dusner. In this language, adjectives are analyzed as stative verbs (45a), and ve is required in attributive constructions (45b). Outside of possessive constructions (45c), ve is usually glossed as a relativizer (45d) and may be used with headed or headless (45e) constructions. It can also convert other parts of speech, including loan words like nadi ‘prayer’, into verbs (45f).

Gil also considers the Mpur morpheme bi to be part of the larger areal VE complex, related to the Austronesian forms via borrowing with phonological adaptation. The prefix bi- is a verbalizer, used like Biakic ve- to convert loans and words of other classes into Mpur verbs (46a). A homophonous root bi functions as a possessive pronoun, used in alienable constructions (46b).

Maybrat also has homophonous morphemes for possession and relative clauses, though with a different form. As a possessive marker, ro- attaches as a prefix to the possessor in an inalienable construction (47a; see also 10). As a relativizer, it appears unbound (47b).

5.3 Possessive-attributive identity

Multiple languages in the south-eastern corner of Linguistic Wallacea have “possessive-like attribute constructions”, similar to those found in many Oceanic languages (Ross 1998). These are attribute constructions that display the typologically unusual phenomenon of having adjectival attributes optionally encoded in the same way as possessions (as opposed to possessors). For example, the noun denoting the referent of the NP acts as if it were a possessor, while the attribute is marked as if it were a possessed noun. Thus, in Welaun, the basic attribute construction for simple adjectival attributes has the attribute following the head noun with no morpheme marking the relationship (48a). A second, more marked, attribute construction includes the 3rd person possessive marker -aan ‘3.poss’ on the attribute (48b), paralleling the encoding of a possessed noun in an adnominal possessive construction (48c).

As pointed out by Schapper (forthcoming), possessive-like attribute constructions are cross-linguistically rare but show a striking distribution in Linguistic Wallacea that is likely indicative of contact. This feature is found throughout the Papuan languages of the Timor-Alor-Pantar family with particular clustering in the languages of eastern Timor, northern Pantar, and western Alor (Schapper this volume). Analogous structures are found in Austronesian languages nearby to TAP languages. They are known from central and eastern Timor including Welaun (above), Naueti (Veloso 2016: 52–54), Kemak, and others discussed in Schapper (2020b: 408–409) as well as Lamaholot lects like Lamaholot Lewotobi (Nagaya 2015). In Kedang, a close relative of Lamaholot, possessive marking of attributes is synchronically lost, but was present in an earlier stage of the language (Schapper and Zobel forthcoming): -n, going back to an earlier possessive suffix, is still used in Kedang as an optional marker of attribution on adjectives ending in a vowel (Samely 1991: 84). It is an open question whether the direction of influence here is from Papuan to Austronesian, as is commonly assumed for morphosyntactic features, or vice versa.

5.4 Possessive-reflexive-article identity

In the Papuan languages of North Halmahera, the prefix ma- marks a possessive relationship between two nouns but has a wide range of other functions as well. The typical range of these functions is illustrated on the basis of Gamkonora (a dialect of Sahu) in (49). Beyond marking a possessive relationship between two nouns (49a), the functions of ma- include its use as an article used on nouns with already established or discourse retrievable referents (49b), as a middle marker on intransitive verbs with middle semantics (49c) and denoting property concepts (49d), and a reflexive marker on transitive verbs (49e) and nouns (49f).

The multifunctionality of ma- has drawn attention in the literature on North Halmahera languages under a broad range of labels (see, e.g., Holton 2006; Reesink 1996). The best label for ma- is perhaps ‘relational noun marker’ in that ma-, at its broadest, functions to signal that the marked item has a relation to a referent that is established in the discourse. It seems most likely that ma- began as a reflexive marker that could appear on nouns and verbs (as is typical for much of the morphology in the area – see Section 5.1), and then extended to being used to mark nouns with already established referents and as a middle marker on intransitive verbs.

Holton’s (2006) suggestion that North Halmaheran ma- is a borrowing of a reflex of the Austronesian stative voice prefix, PMP *ma- seems to be based on little more than the formal similarity of the prefixes in the two groups of languages. Ma- does not have a range of functions in any Austronesian languages we are aware of that is similar to those found in the North Halmaheran languages. The formal similarity is likely a matter of chance, enhanced by the tendency for languages to make use of reduced sets of phonemes in affixal morphology.