Questioning the relevance of alienability in Arawak linguistics: an innovative analysis of possession in Mojeño Trinitario

2.1 The basic adnominal possessive construction

In the basic adnominal possessive construction, the noun for the possessee has to carry a person prefix indexing the possessor. This noun may be modified by the following noun phrase for its possessor.^[8] This head-marked construction is summarized in Figure 3, where optional modifiers are overlooked. It is exemplified in (10).

Figure 3:

The basic adnominal possessive construction.

Very often, as in (11), the possessor is not expressed as an overt noun phrase, either because it is already given in the context and the possessive prefix on the possessee suffices to retrieve its identity, or because the possessor is first or second person.^[9]

2.2 The indirect adnominal possessive construction

The indirect adnominal possessive construction follows the pattern of the basic adnominal possessive construction (summarized in Figure 3), in that a noun carries a person prefix indexing the possessor, and may be modified by a following noun phrase for its possessor. But here the noun that carries the person prefix is systematically the generic possessive noun -yeꞌe ‘belonging’,^[10] and is combined with another noun, that is semantically the possessee. The indirect adnominal possessive construction is summarized in Figure 4, where optional modifiers are overlooked.

Figure 4:

The indirect adnominal possessive construction.

Just like the possessee noun in the basic construction (Figure 3), the noun -yeꞌe is preceded by a determiner and carries a person prefix referring to the possessor. However, -yeꞌe does not add any particular meaning to the possessive construction. It is a dummy item coreferential with the noun that specifies semantically the identity of the possessee.

In general, the noun -yeꞌe precedes the noun expressing the possessee, and both are preceded by one determiner only, as illustrated in Example (12) where mayeꞌe paku is preceded by the non-human determiner to.

The fact that the generic possessive noun -yeꞌe and the semantically more specific noun for the possessee share a single determiner excludes an analysis in terms of juxtaposition of two noun phrases: both nouns are forming the core of a noun phrase, with -yeꞌe being a head and the possessee noun a modifier.^[11] This noun phrase then allows the embedding of a further noun phrase for the possessor.

Most frequently, the possessor noun phrase is not expressed because it is one of the speech act participants, as in (13), or because it is discourse-given, as in (14) from the Frog story (a rendering of a picture book about a boy and a frog [Mayer 1969]), where ma- refers to the main protagonist of the story.

In addition, and extremely rarely, the possessee may not be expressed. The meaning of -yeꞌe is then either semantically vague, or interpretable in context. It may be translated as ‘thing’ or with a possessive pronoun as in (15).

The noun -yeꞌe is itself an obligatorily prefixed noun without a clear semantic extension beyond the meaning of expressing a relation between two entities, hence its gloss as ‘generic possessive noun’ (gpn).^[12] It retains its nominal features: it is preceded by a determiner, takes possessive prefixes like all obligatorily prefixed nouns, and also combines with the plural marker -ono found on nominals (15) and the nominal irrealis suffix -ina (not illustrated here).

This section has presented the two Mojeño Trinitario adnominal possessive constructions, while the next section will present the three Mojeño Trinitario possessive noun classes. The distribution of the different classes of nouns in the two possessive constructions will be discussed in Section 5.

3.1 Obligatorily prefixed nouns

A first class of nouns comprises those roots which must carry a possessive prefix specifying the person, gender and number of the possessor. Table 3 offers a selection of these obligatorily prefixed noun roots. The first column shows the form used with a possessive person prefix, the fourth column gives the gloss, and the last one the semantic domain.

The first subset of nouns in Table 3 are always produced with a person prefix in the texts, and in elicitation sessions, speakers could not think of them without a prefix. An example is given in (16). Other nouns use one of several formal devices to occur without a prefix, i.e. unpossessed. The second column of Table 3 gives the form of these nouns when used without a prefix, while the third highlights the device used for the prefixless form, often a suffix glossed nposd as in (17). This derivation process will be dealt with in Section 4.1.

In terms of semantic domains (last column of Table 3), the class of obligatorily possessed nouns exhausts all kinship terms and spatial relations, and it includes most body parts and parts of wholes, most items of clothing and personal accessories, some bodily excretions, some attributes like ‘work’, ‘charge’ and ‘language’ and a few artifacts like ‘beer’, ‘arrow’, and ‘flute’. Note the presence of two loanwords from Spanish in this class: -kuñaru ‘brother-in-law’ and -kuñara ‘sister-in-law’.

3.2 Optionally prefixed nouns

A second class of nouns consists of noun roots that can take a possessive person prefix but do not have to. They can also appear bare. This is illustrated by the two occurrences of metsi ‘pot’ in (18), first with a possessive person prefix and then prefixless. A selection of optionally prefixed nominal roots is presented in Table 4. Stems derived from non-prefixable roots behave in a similar fashion (see Section 4.2).

Most roots in this class show two different forms depending on the presence or absence of a prefix, because of rhythmic syncope (Rose 2019a), such as -pokre ∼ pkure ‘canoe’ (the underlying form of which, pokure, is never attested synchronically). Most invariable roots are borrowings, like kareta ‘cart’ (<Sp. carreta). It is also noteworthy that many of the optionally prefixed nouns have historically incorporated a classifier (i.e., the root never occurs without the classifier): one can recognize the classifiers -mo clf.fabric in -poomo ‘leather’, -pe clf.blade in -esaviipe ‘machete’, -pa clf.needle in -ijruupa ‘spindle’ and -ve clf.pointed in enrove ‘fishing hook’ for example.

Semantically, the optionally possessed nouns can be conceived of as referring to alienably possessed items. Most refer to artifacts that can easily change possessors. However, this class also includes one noun that refers to a body part and one to a part-whole relation, two semantic classes of nouns considered as expressing inalienable relationships because they do not usually change possessors. The only body-part term in this class, -poomo ‘skin, leather’ is most often used of the body part taken off from an animal (leather is an important material in this cattle-farming society). Elicited data indicate that natural entities or landscape elements, animals and plants may also enter this class.

3.3 Non-prefixable nouns

The third class of nouns consists of roots that do not take possessive person prefixes. A selection of non-prefixable noun roots is presented in Table 5. The first column shows the bare form of a noun, the fourth column the gloss, and the last one the semantic domain. The second and third columns concern only the roots that can be derived into optionally prefixed noun stems (see Section 4.2). Additionally, stems derived from obligatorily possessed roots behave like non-prefixable nouns (see Section 4.1).

Semantically, non-prefixable noun roots refer to elements which are not normally possessed, i.e., are not seen as necessarily participating in the sphere of some other element: many refer to inanimate entities like natural elements and flora, but some also refer to animate beings (wild animals and general terms for humans). The class nevertheless also contains some items that can be conceived as possessed, referring to animates like domesticated animals and social relations, and to inanimates like food, clothing, artifacts, part-whole relations, bodily excretions and attributes. Note that many but not all of the non-prefixable nouns that can be conceived as being possessed are borrowings from Spanish, like ꞌresia ‘church’, padrino ‘godfather’, ꞌavitu ‘robe’, prata ‘money’ or moto ‘motorcycle’.

This section has described the three lexical classes of Mojeño Trinitario nouns based on their morphology: obligatorily prefixed, optionally prefixed and non-prefixable. How these classes associate with the adnominal constructions will be described in Section 5, and how they can or cannot be accounted for by an alienability contrast will be discussed in Section 6, after the possible derivation among these classes is presented in the next section.

4.1 Derivation of non-prefixable noun stems from obligatorily prefixed noun roots

Some obligatorily prefixed nominal roots can be derived into non-prefixable stems by various formal means, listed in the third column of Table 3: by adding a nposd suffix -re, -ko, or -ti,^[14] via a final vowel shift in the noun -sawara ∼ saware ‘tobacco’, or through suppletive forms or substantial formal changes for a few nouns. The stems derived through these means occur without a possessive person prefix, such as muꞌi-re ‘dress’ in (19), where a potential possessor is not at issue.^[15]

As for semantics, the sub-set of obligatorily prefixed nominal roots that can be derived into non-prefixable stems includes some affinal kin terms, body parts, clothing items and accessories, attributes and artifacts. The sub-set of obligatorily prefixed noun roots that cannot be derived into non-prefixable stems includes non-affinal kin terms, spatial configurations and parts of wholes. At least two of the derived noun stems show a meaning different from that of the root: chti-re ‘skull’ from -chuti ‘head’, and ꞌchojrii-ko-wo ‘story’ (21) from -echjiriiwo ‘language, speech’ (20). It is a common characteristic of derivation that the semantics of the derived item is not always predictable.

The three most frequent derivational devices are the nposd suffixes -re, -ko, and -ti. Each one is specialized for a certain set of nouns, but this distribution is not predictable.^[16] One can nevertheless propose a few semantic generalizations: -re is used for clothing and accessories, and for most (derivable) body parts; affinal kinship terms are derived with either -re or -ko; most lexical nominalizations take -ko.

4.2 Derivation of optionally prefixed nominal stems from non-prefixable noun roots

The other derivation process among noun classes derives non-prefixable noun roots into prefixable nominal stems. This is essentially done with two posd suffixes -ra and -ne (see the third column of Table 5). In (22), -ra makes the non-prefixable root krutsu ‘cross’ (from Spanish cruz) a prefixable form, which occurs prefixed in a basic possessive construction.

Derived prefixable stems most often occur within the basic possessive construction. However, I consider that the derived stems are optionally, rather than obligatorily, prefixed because two derived nouns occur without a possessive person prefix in my corpus, one outside of a possessive construction and the other in the indirect possessive construction (exemplified in (24)).

As for semantics, the sub-set of non-prefixable noun roots that can take a posd suffix includes terms for a diversity of elements: animals, plants and natural elements/phenomena,^[17] as well as terms for humans expressing (temporary) social relations (such as ‘employee’) rather than permanent identities (such as ‘man’). Some terms for human beings show a distinct meaning when derived (such as ꞌchane ‘person’ and -chanera ‘assistant, employee, partner’), as expected for derivation processes.

Just like the nposd suffixes, the two posd suffixes -ra and -ne are in complementary distribution, as they are specific for certain sub-sets of nouns. It is not feasible to motivate the distribution of the two suffixes on specific noun roots on the basis of phonological or semantic grounds in a straightforward fashion.^[18] One can only note that -ra tends to associate with nouns referring to animates (humans or animals), often with a sense of accompaniment or assistance. Both -ra and -ne are attested with borrowings, such as waka ∼ -waka-ra ‘cow, meat’,^[19] and kaja ∼ -kaja-ne ‘caja (a music instrument)’. Elicitation has shown that -ra is particularly productive on a large variety of nouns.

Finally, the corpus shows a single example of double derivation. A non-prefixable stem ꞌmotne-ko ‘work’ derived with a nposd suffix from the obligatorily prefixed root -emtone ‘work’ can additionally take a posd suffix to become prefixable. In the process, its meaning changes from ‘work’ to ‘employee’, with -ra expressing a social relation of assistance. In (23), the doubly-derived stem is used in the basic construction, and in (24), in the indirect possessive construction.

The interaction of the two types of derived stems described in this section with adnominal possessive constructions is described in the next section. Section 6 will discuss how derivational processes make the picture of the expression of possession in Mojeño Trinitario more difficult to explain by the alienability contrast.

5.1 Regular combination of noun roots with possessive constructions

Noun roots of different classes, when non-derived, most frequently enter a specific possessive construction. Obligatorily prefixed noun roots like -chicha ‘offspring’ exclusively enter the basic possessive construction as in (25). Optionally prefixed noun roots like pokre ‘canoe’ mainly enter the basic possessive construction as in (26). Non-prefixable roots like ꞌpogꞌe ‘land’ may exclusively enter the indirect construction as in (27). What has not been checked through elicitation work is whether some non-prefixable nouns can absolutely not enter the indirect construction and are therefore strictly non-possessible.

5.2 Exceptional combination of an optionally prefixed noun root with the indirect possessive construction

It has been asserted in the preceding section that (non-derived) optionally prefixed noun roots mainly enter the basic possessive construction. This combination is not exclusive: there is one corpus example of such a noun in the indirect construction, given in (28). Since this pattern has been reproduced in elicitation with other optionally prefixed nouns, I take it as a rare but grammatical combination. In the discourse from which (28) comes, the possessee is a very atypical member of the canoe category, namely a caiman. The only hypothesis I can put forward for this example is that the possession of a “caiman canoe” might be conceived of differently from the possession of a wooden canoe.^[20]

5.3 Obligatorily prefixed nouns in a doubly marked possessive construction

It has been stated in Section 2 that the two adnominal possessive constructions may combine in a single construction: in this doubly-marked possessive construction, the possessee noun takes a prefix for its possessor and is preceded by the generic possessive noun -yeꞌe that also takes a possessive prefix. This construction is analyzed as a single noun phrase, given the presence of one determiner only (the demonstrative jma in (29)), and the absence of a prosodic break. The interpretation of this construction requires further research.

Only obligatorily prefixed nouns are found in the doubly-marked possessive construction, which is rare: it occurs five times in the text corpus, and once in an elicited noun phrase, as in (29) and (30).

The five text examples all involve lexical nominalizations (with various patientive nominalizers), in which the nominalized verb obligatorily takes a possessive prefix encoding the “subject” of the verb. In this construction, the person prefixes on the deverbal noun and on -yeꞌe are always coreferential so that there is a single possessor entity. A tentative explanation for this double marking of possession is that in the basic construction, the possessive relationship is preferably interpreted as being agentive: the referent of the person prefix on the deverbal noun, in analogy with its use on verbs, tends to be interpreted as the agent of the verbal root.^[21] In contrast, the possessive relationship expressed in the indirect construction does not have this additional semantics. The doubly marked possessive construction would then express two types of possessive relationships simultaneously, with one of them being more “agentive” when deverbal nouns are involved.

For instance, in (29), the deverbal noun ni-ru ‘food’ is possessed both with a possessive person prefix and through the use of the generic possessive noun -yeꞌe. Given that this sentence was produced to describe a video clip (Vuillermet and Kopecka 2019), it is difficult to speculate what kind of special possessive relations the protagonists could have with their food. Of course, they could be conceived of as being both the eaters and the owners of the food.

This analysis of the doubly-marked possessive construction as encoding two different possessive relationships works as well for the single (elicited) example of doubly marked possession involving an underived nominal stem, given in (30).^[22] In this example, the person prefix on the noun and that on -yeꞌe are not coreferential, so this construction involves two possessors. The person prefix on the noun for ‘fruit’ can only refer to the plant it has grown on and an owner of the ‘fruit’ can only be referred to through the indirect possessive construction. This elicited example is the only case not involving derivation where one clearly sees a different contribution of the two possessive constructions.

5.4 Integration of derived prefixable noun stems in the basic possessive construction

We have seen that underived non-prefixable roots enter the indirect possessive construction only (5.1). However, once derived into prefixable stems, they can enter the basic possessive construction. Noticeably, most of the examples of the corpus involve borrowings, with the exception of ochenekra ‘path’ (illustrated in (33)) and the doubly-derived ꞌmotnekora ‘employee’ (seen in (23) and (24)).

Most of the time, there is no discernable meaning difference between the two possessive constructions involving the same root. For instance, the following pair of examples shows the non-prefixable noun root wasu ‘glass’ (borrowed from Spanish) in two very comparable sentences.^[23] In (31), the basic construction builds on the prefixable form of the noun, derived with the posd suffix -ra; in (32), the indirect construction builds on the non-derived form. There is no noticeable difference in meaning between the two constructions.

The two possessive constructions have been noticed to clearly encode two different possessive relationships with one non-prefixable root only, ꞌchene ‘path’. Within the basic possessive construction involving the derived prefixable form -ochene-k-ra in (33), the possessor of the path is its goal, i.e., the place to which the path leads (here, the field). Within the indirect possessive construction involving the non-prefixable form ꞌchene in (34), the possessor of the path is the person who usually uses it (and was probably responsible for making it in the first place).

5.5 Integration of derived non-prefixable noun stems in the indirect possessive construction

As stated in Section 5.1, obligatorily prefixed roots exclusively enter the basic possessive construction when underived. When derived into non-prefixable noun stems, they may either be used outside of any possessive construction as in (35) or enter the indirect possessive construction. This latter combination is attested with three deverbal nouns only, echjiriiwo ‘language’ as in (21), owsa ‘village’, and ajre ‘written document’ in (36).^[24]

For the former, the semantic distinction from its non-derived form in (20) is not due to the construction, but rather to the derivation, since the meaning of ‘story’ is also attested outside of a possessive construction as in (35). In fact, there is no real difference in terms of the semantics of the possessive relationship between the two possessive constructions in (20) and (21): both are inherent possessions.

By contrast, the derivation of ajre ‘written document’ into a non-prefixable stem ajúre-ko does not seem to affect the lexical meaning of the noun. When the obligatorily prefixed stem ajre occurs in a basic possessive construction, the person indexed by the person prefix is spontaneously interpreted as the agent of ‘write’; for instance, to n-ajre is interpreted as ‘my document (that I wrote)’. In contrast, the derived non-prefixable stem ajúre-ko does not have this implication in an indirect possessive construction: in (36), the possessive relationship is spontaneously interpreted as ownership.^[25]

This section has shown that there is no simple association between noun classes and possessive constructions. In particular, it has presented counterexamples to the assumption that crosslinguistically, the indirect possessive construction is used exclusively with nouns that cannot take any possessive inflectional morphology or cannot occur with a possessor (Bickel and Nichols 2013; Krasnoukhova 2012: 58). There is very little evidence for a systematic inalienable/alienable contrast when the two possessive constructions accept the same noun.

6.1 Alienability as a binary contrast

A preliminary issue in attempting to apply the alienability contrast to noun classes in Mojeño Trinitario is that this contrast is generally presented as binary, while there are three noun classes in the language, some with subclasses. It is therefore unclear what distinction the alienability contrast is supposed to explain.

The notion of alienability was initially motivated by coding differences between different classes of nouns within the syntactic domain of adnominal possession. Haspelmath expresses the most common expectation as follows: “Possessive constructions with inalienable nouns tend to show zero coding, short coding, bound coding, and/or obligatoriness, while possessive constructions with alienable nouns tend to show overt coding, long coding, free coding, and/or impossessibility” (Haspelmath 2017: 218). In Mojeño Trinitario, a coding contrast in terms of the number of morphemes is found between prefixable and non-prefixable noun roots: the former require only a possessive prefix, while the latter need either derivation or the use of a generic possessive noun. This overt differential possession marking draws a boundary within semantically alienable nouns (see Section 3), in contradiction with the predictions related to the alienability contrast.

Coding differences can also be covert: the contrast is not apparent in terms of number of morphemes, but only through the obligatoriness of the possessor in discourse. This is referred to as ‘obligatory possessive inflection’ by Bickel and Nichols (2013) and as ‘inherent’ possession by Payne (1997: 105–106). In the Arawak literature (see, for example, Aikhenvald 1999: 82; Michael 2012), the alienability contrast usually distinguishes between obligatorily possessed and optionally possessed nouns, because the terms inalienable and alienable are taken as synonyms of bound and free. Indeed, among the prefixable roots, the obligatorily prefixed nouns like -pórape ‘older sibling’ in (16) and the optionally prefixed nouns like metsi ‘pot’ in (18) are equally marked for their possessor, but while metsi can also occur without a prefix, -pórape cannot. Importantly, this covert differential marking is valid only among noun stems that can, but do not have to, be possessed. This means that this instantiation of the alienability contrast applies to a part of the possessive domain only, excluding the third class of so-called non-possessible nouns. Incidentally, we might wonder what the added value is of the terms inalienable and alienable, if they are used as synonyms of bound and free.

As a conclusion, there is definitely some arbitrariness as to which distinctions among noun classes the alienability contrast is taken to account for, especially if seen as binary. Another possibility, which steers clear from excluding or conflating noun classes, would consist in making the alienability contrast a ternary contrast itself, which would reflect the three-way lexical organization straightforwardly. Specifically, this approach implies allowing for “more alienable” and “less alienable” nouns, as does Haspelmath (2017: 213–214) with the term “super alienable” for the semantic group of nouns that rarely occur in possessive constructions (the non-prefixable nouns).

6.2 The alienability contrast as a semantic contrast between noun classes

The notion of alienability was originally meant to account for different semantic types of possession. Let’s now zoom in on the semantics of noun classes. The Mojeño Trinitario data only partly reflect commonly held views about the semantics of inalienable versus alienable noun classes. Heine (1997:10) presents as likely inalienable nouns those expressing the following meanings: (a) kinship roles, (b) body parts, (c) relational spatial concepts, like ‘interior’, (d) parts of other items, like ‘branch’, (e) physical and mental states, like ‘strength’, and (f) nominalizations, where the ‘possessee’ is a verbal noun, for example, ‘his singing’. The class of Mojeño Trinitario obligatorily prefixed nouns does include, as expected, kinship terms, spatial concepts, physical and mental states (subset of ‘attributes’ in my semantic categorization)^[26] and nominalizations (see Sections 3.1 and 4.1 above). However, we have seen that body parts and parts of wholes are not only lexified as obligatorily prefixed nouns but also marginally as optionally prefixed nouns, and attributes such as ‘age’ are also found among non-prefixable roots. The classification of Mojeño Trinitario nouns therefore only partly corresponds to the expectations spelled out above.

Another semantic generalization made on possessive classification is that “[s]emantics is involved in that there is usually a default or open class and a specified or determinate class with a semantic common denominator shared by most but not all members of the class and also sometimes found in non-members” (Nichols and Bickel 2013). These authors explain that the specified class is the one including kinship terms, body parts, parts of wholes, spatial relations and property nouns (my “attributes”). However, with respect to the Mojeño Trinitario data, I find it difficult to consider the class of obligatorily prefixed nouns to be “specified” and “with a semantic common denominator”, as it covers a good number of semantic domains (see the last column of Table 3). This is especially striking in contrast with the optionally prefixed noun class, which is almost entirely made of artifacts and is therefore semantically homogenous and specific (see the last column of Table 4).

Similarly, the Mojeño Trinitario data do not support Nichols’s (1988: 562) assertion that “the nouns that take ‘inalienable’ possession virtually always form a closed set, often a small one, while those taking ‘alienable’ possession are an open, hence infinite, set”. Table 6 presents the relative size of the three Mojeño Trinitario noun classes and their major subclasses. It is clear that obligatorily prefixed nouns (the class usually considered as consisting of inalienably possessed entities) do not form a marginal set in the current extension Mojeño Trinitario lexicon.^[27]

Regarding whether the classes are closed or open, the integration of borrowings and nominalizations into the noun classes must be examined.^[28] Table 3 to Table 5 show that nouns borrowed into Mojeño Trinitario have joined all three possessive classes: -kuñara ‘sister-in-law’ from Spanish cuñada is an obligatorily prefixed noun, kareta ‘cart’ from Spanish carreta is an optionally prefixed noun, and ꞌresia ‘church’ from Spanish iglesia is a non-prefixable noun. There are only two loanwords in the class of obligatorily prefixed nouns (-kuñara ‘sister-in-law’ and -kuñaru ‘brother-in-law’), and these are not derivable into non-prefixable stems. This contrasts with the numerous non-prefixable borrowed nouns which show a deep morphological integration by being derivable into prefixable stems. However, this does not mean that the class of obligatorily possessed nouns is closed, because it actually shows a good number of deverbal nouns, thus making this class open to new items. All three classes are therefore open, which goes against Nichols’s (1988) assertion above.

6.3 The alienability contrast motivated by frequency of possession?

The literature on the alienability contrast discusses how this contrast should be explained. Haspelmath (2017) argues that it should be explained “by predictability due to the higher relative frequency of possessed occurrences of inalienable nouns”. To support this claim, he counts how many times six inalienable and six alienable nouns of Biblical Hebrew, Ancient Greek, and English are possessed in 20 random corpus examples. The results show that inalienable nouns occur much more often possessed than alienable nouns in speech. A major criticism one can address to this study is that none of these languages show possessive noun classes. The present study offers to undertake a similar task on the basis of a language with several possessive noun classes, namely Mojeño Trinitario. It will both serve as an assessment of one particular motivation of the crosslinguistically recurrent possessive coding split, and as an empirical tool in investigating where the major alienability split is expected to lay in that language, following that motivation.

The study targets 12 obligatorily prefixed (including six derivable ones), 6 optionally prefixed, and 12 non-prefixable noun roots (including six derivable ones). Because my corpus is small but searchable, I have counted all occurrences of these roots and classified them as possessed (i.e., within any of the two possessive constructions) or unpossessed.^[29] The results are presented in Table 7. They only partly meet the expectations.

As expected, non-derivable obligatorily prefixed nouns (first set of nouns in Table 7) always occur possessed, since they require a possessive person prefix when they are used as the head of an NP.^[30] Derivable obligatorily prefixed nouns (second set of nouns in Table 7) generally occur possessed in discourse, most often via the basic possessive construction.^[31] This proportion is naturally lower than that of other obligatorily possessed nouns, which always require a possessive person prefix. The corpus includes rather few nouns carrying a nposd suffix, and these are not frequent in texts, except for a few very frequent derived forms like ꞌwósa-re ‘town’, ꞌsan-ti ‘field’, tapraa-ko ‘charge’ and ꞌrésa-re ‘beer’.

As for optionally prefixed roots (third set of nouns in Table 7), two occur about equally often possessed or non-possessed, while the four remaining ones occur much more rarely possessed.

Now for the twelve non-prefixable nouns (fourth and fifth sets of nouns in Table 7), seven of them rarely or never occur possessed in the corpus, as expected. Remarkably, five other ones occur possessed as often as those optionally prefixed roots that are less often used possessed.^[32] The derivable fauna term wiye ‘ox’ behaves similarly to the optionally prefixed nouns that are more often found possessed (‘pot’ and ‘canoe’ in Table 7).

In brief, only the class of obligatorily prefixed nouns behaves homogeneously and as expected, with member nouns being possessed at least 50 % of the time. If the frequency at which roots were found possessed were a predictor for a binary contrast (as expected for the alienability contrast), then the split between the two degrees of alienability would not correspond to the noun classes: optionally prefixed ‘pot’ and ‘canoe’, as well as the non-prefixable ‘last name’ should be obligatorily prefixed nouns. In the end, because these counts have shown that noun class membership in Mojeño Trinitario is not predictable from the relative frequency of possessed occurrences in discourse, it weakens Haspelmath’s (2017) claims on the motivation of the alienability contrast. Consequently, the fact that this methodology did not show a clear binary or ternary frequency distribution that would match a formal contrast undermines the operability of the ‘alienability contrast’ in Mojeño Trinitario.

6.4 The alienability contrast and derivation among possessive noun classes

This section investigates how the concept of an alienability contrast fits with derivation processes between possessive noun classes. The Arawak literature is inconsistent in whether the alienability contrast applies to roots or stems: obligatorily possessed nouns are considered inalienable based on the behavior of their roots, while so-called optionally possessed nouns are considered alienable based on the behavior of their derived stems (they are possessible only when derived with a suffix). This section aims to discuss how the notion of alienability articulates with derivation.

First, Nichols (1988: 597), in her pioneering article on alienability, explains that in some languages, some special morphological devices apply to bound nouns to enable them to stand alone as so-called ‘absolutive’ forms. She considers this process to be derivational and argues that “the existence of an absolutive does not weaken the claim that the bound nouns cannot occur without possession”. In short, she applies the alienability contrast to the root level. This line of analysis seems to be followed in most descriptions of Arawak nposd suffixes, and is sometimes explicitly described as ‘derivation’ (Michael 2012: 153; Terhart 2021: 179, among others).^[33]

In the present article, I extend Nichols’ line of analysis to the posd suffixes, which make some non-prefixable roots possessible through the basic construction found throughout the Arawak family (where these roots are overwhelmingly analyzed as ‘alienable’ or ‘optionally possessed nouns’). The status of the posd suffixes is unclear in current descriptions of Arawak languages. Danielsen (2007: 123) and Terhart (2021: 180–181) describe them explicitly as derivational, but – surprisingly – still consider the roots that take them as alienable but derived into inalienable stems with the posd suffixes. If here too the concept of alienability applies at the root level, the roots in question are not possessible without previous derivation. This is why I propose to analyze these roots in Mojeño Trinitario as non-prefixable rather than as prefixable.

Regarding the derivational processes themselves, the coding differences seem to fit the expectations. In Haspelmath’s terminology, the posd suffixes constitute an additional marker in the possession of super-alienable nouns, compared to alienable and inalienable nouns which only require a possessive person prefix. In turn, the nposd suffixes constitute an additional marker in the non-possession of inalienable nouns, compared to alienable and non-possessible nouns. In brief, derivational processes among noun classes can be seen as additional coding on nouns when occurring in non-prototypical possessive contexts.

It is yet unclear how to account for why only some noun roots are derivable. Most often, the derivation does not yield different lexical meanings between the root and the derived stem. I argue that the semantics associated with the derivation processes are essentially not a matter of lexical semantics but one of semantic valency. One could advance that the derivable noun roots are multi-relational, i.e., they can enter multiple relations with other entities (Durieux 1990; cited in Dahl and Koptjevskaja-Tamm 1998).

Some obligatorily possessed nouns (body parts, clothing items, accessories and artifacts) are logically multi-relational, and could be conceived as entering different types of possessive relationship: for instance, clothes and accessories have a relation with the person that wears them, but also with the person who made them. This type of multi-relationality is not evidenced in our small corpus, where the only derivable obligatorily possessed nouns that enter both possessive constructions are deverbal nouns. As argued above in 5.3 and 5.5, the multi-relationality of deverbal nouns seems to be triggered by the ambiguity of the possessive prefix, which can be interpreted as a participant of the situation expressed by the verbal root or as indexing the possessor of the deverbal noun.

Turning now to derivable non-prefixable roots, we have observed among them only one clearly multi-relational noun, the one for ‘path’. When derived as in (33), its semantic valency increases, involving an additional “possessor”, the place to which this path leads. In sum, multi-relationality could be a semantic particularity of (some) derivable roots. On this basis, one could see the two derivation processes under study as valency-changing mechanisms manifesting this potential multi-relationality in morphosyntax.^[34] The posd suffixes increase the valency of nouns, while the nposd ones decrease it (see Queixalós 2005, 2016 for the concept of noun valency). But there seems to be no clear semantic foundation to which nouns are derivable. Interestingly, the potential for derivation does not align with the alienability contrast: non-derivable nouns are found in all three classes, and derivable nouns in both obligatorily prefixed and non-prefixable noun classes.

6.5 The alienability contrast and the two adnominal possessive constructions

The application of the alienability contrast at the construction level also needs to be investigated. On the formal level, the indirect construction shows more coding devices than the basic construction (because of the presence of the generic possessive noun). The two adnominal possessive constructions are therefore good candidates for embodying the supposed alienability contrast. However, on the semantic level, it is not the case that the basic construction only encodes inalienable possession (it is also used with the artifacts expressed by optionally prefixed nouns), even if the indirect construction can generally be associated with alienable possession (it is mainly used with non-prefixable stems, considered to be alienable).

In fact, the following pair of elicited examples shows that the distribution between the two possessive constructions does not crucially depend on the particular semantics of the possessive relationship (nor on the identity of the possessee and possessor), but primarily on noun classes. The examples involve two synonymous nouns as possessees, one native and the other borrowed from Spanish,^[35] with the same first person possessor. The native word is an obligatorily prefixed noun and enters the basic possessive construction (37), while the loanword is a non-prefixable noun and enters the indirect construction (38).

In general, the distribution of the two constructions largely depends on the class of the noun root or the derived noun stem, as detailed in Section 5. There is little evidence for an inalienable/alienable contrast when the same noun can enter both possessive constructions. The most numerous cases of ‘competition’ between the possessive constructions can be seen as an effect of the origin of said nouns as deverbal nouns.

6.6 Summary on the alienability contrast in Mojeño Trinitario

To summarize the findings of this section, the alienability contrast can be applied arbitrarily to different parts of the distinctions between noun classes (6.1). The only strong case it makes is the distinction in terms of obligatoriness of inflection between obligatorily prefixed nouns and the others, aka as bound versus free nouns. It leaves the remaining differences between optionally prefixed and non-possessible nouns unexplained.

The alienability contrast does not predict which individual nouns join which particular noun classes or sub-classes, be it based on semantics (6.2) or frequency of occurring possessed (6.3). This article indeed offers frequency counts on how often nouns of different possessive classes occur possessed in discourse (Table 7) and concludes that these would predict an alienability contrast that does not correspond to the actual distribution in the three noun classes. Another important result of this article for the typology of possession is that the “inalienable” class of nouns cannot be seen as a closed marginal class semantically more homogenous than the other classes, contrary to Nichols’s (1988) general claims (6.2). As for derivational processes among noun classes, they can be seen as differential coding for non-prototypical possessive situations.

Finally, the distinction between the two adnominal possessive constructions (syntactic level) is also evaluated as a weak instantiation of the alienability contrast (6.5), because it only partially corresponds to the semantic expectations: while the structure with most coding material (the indirect construction) only expresses alienable possession, the least coded structure (the basic construction) encodes either inalienable or alienable possession.

6.7 Discussion

The main theoretical problem behind the alienability contrast as an explanation for the encoding of possession is that it is unclear to which sets of linguistic objects it should a priori apply. This is so because the literature applies it to either noun classes or constructions, and usually presents it as binary, sometimes as ternary. This vagueness has two logical consequences. First, I consider that if this functional explanation can apply at several levels at once in a language, and if these levels do not perfectly overlap, then the “explanation” is contradictory. Second, the explanation is more likely to be considered robust if applied as an a posteriori explanation to a particular set of data, at a level chosen by the analyst on a language-individual basis, but this can be called methodological opportunism (Croft 2001: 30).

The vagueness of how the alienability contrast should apply has several practical consequences. On the one hand, it means that discussing this contrast in a crosslinguistic perspective is not methodologically sound, unless one systematically separates the levels of analysis, as done by Chousou-Polydouri et al. (this issue). On the other hand, at the language-individual level, it may then leave unexplained some phenomena it is supposed to cover and has a very low predictive power. One could actually wonder whether invoking the alienability contrast to “make sense” of lexical or grammatical distinctions does not actually bring confusion into language-specific analyses. Two potential effects on language-specific analyses are the following: (i) by promoting a binary contrast, the alienability contrast makes the distinction among free forms (optionally prefixed and non-prefixable class) less salient, (ii) by merging formal and semantic considerations, it is an obstacle for a clear separation of lexicon, syntax, and semantics. For instance, “alienable” was originally a semantic concept, but is most often used to label a class of nouns, and sometimes a possessive construction. This results in many descriptions of Arawak languages presenting obvious contradictions. For example, nouns labeled as “non-possessible” (and as such distinct from “alienable” nouns) are still mentioned as entering a particular possessive construction, being hence obviously morphosyntactically possessible, and semantically alienable. At the same time, there has been until this article an apparent general resistance to labelling those roots that happen to be easily possessed only once suffixed as “non-possessible”.^[36]

To solve the problem of the vagueness of the domain to which the alienability contrast should apply, this article advocates a strict separation of the levels of analysis, as implemented with Mojeño Trinitario data. A major methodological issue when investigating the potential of the alienability contrast as an explanation for the competition between the adnominal possessive constructions is the difficulty of teasing apart the semantic input of noun classes, the effect of the derivation processes among these, and the semantic import of the two possessive constructions, and this is, I believe, what future efforts should concentrate on.