When the alienability contrast fails to surface in adnominal possession: bound nouns in Harakmbut

1 Introduction

This article investigates to what extent the notion of alienability can account for the distinct behavior of classes of nouns in the underdescribed language Harakmbut, more precisely the Arakmbut (Amarakaeri) dialect (glotto-code amar1274),^[1] spoken in south-east Peru (departamentos of Madre de Dios and Cusco). Harakmbut is still regarded as an unclassified Amazonian language (Wise 1999: 307; Dryer and Haspelmath 2013), although Adelaar (2000, 2007 has made a case for a genetic link with the Brazilian Katukina family, which may be further linked to Macro-Jê. More information on its genetic affiliation, internal classification, vitality and sociolinguistic context is presented in Van linden (2023: 441–444).

The notion of alienability entered the linguistic scene in the early twentieth century in the context of studies on possession. It was proposed to explain patterns of distinct morphosyntactic marking, or coding splits, in the grammatical environment of adnominal possession, e.g., in Melanesian languages (Lévy-Bruhl 1914) and North American languages (see Nichols 1988), and to account for restrictions in terms of possessee types in the grammatical environment of external possession in Indo-European languages (e.g., Bally 1926). These phenomena were explained in terms of the conceptual difference between inalienable and alienable possession: the former involves “inextricable, essential or unchangeable relations” between possessor and possessee, while the latter involves associations between possessor and possessee that are of a less permanent and inherent type (Chappell and McGregor 1996: 4). Prototypical possessees of the former type include body-part terms and kinship terms, while those of the latter type comprise artifacts. Within adnominal possession, then, the observed pattern is that inalienable possession is typically coded with less “morphosyntactic material” than alienable possession (Payne 1997: 105; see also Haspelmath 2017). Rather than viewing the alienability contrast as a semantic explanation for coding splits, Nichols (1988) proposed regarding it as a formal contrast at work on two different levels of linguistic organization: at the word level, yielding two different noun classes (inalienable vs. alienable nouns), and at the phrase level, yielding two different adnominal possessive constructions, with the inalienable construction featuring less morphosyntactic marking than the alienable one. This article will investigate the explanatory potential of each interpretation of the alienability contrast in examining the class of bound nouns in Harakmbut, and their morphosyntactic behavior in different grammatical environments.

The starting point of this article is indeed the morphological distinction between independent and bound nouns in Harakmbut. Unlike independent nouns, as in (1a), bound nouns require a noun prefix to obtain independent nominal status (wa(ʔ)- or e(ʔ)-), as in (2a), and mainly refer to inalienably possessed entities, such as body parts, plant parts, and landscape parts, as well as kinship terms and basic shapes or qualities of entities. These two noun classes are sometimes called non-obligatorily versus obligatorily possessed nouns, or alienable versus inalienable nouns, and they are ubiquitous in Amazonian languages (Krasnoukhova 2012: 87–88). Considering the two adnominal possessive constructions in the Harakmbut Examples (1b) and (2b), one may conclude that the distinct morphosyntactic patterning exhibited by members of the two noun classes can be explained in terms of the alienability contrast.

The pattern with the bound noun in (2b) shows a tighter morphosyntactic bond between the genitive-marked possessor and possessee – in fact, the two form a single phonological unit – than the two-word pattern with the independent noun in (1b). The examples are representative of their respective noun classes in patterns with human possessors, and might thus trick the analyst into thinking that Harakmbut has an alienability split. However, it will turn out that bound nouns can also enter the two-word possessive construction shown in (1b) when they take their noun prefix, and that some bound nouns (e.g., some kin terms) do not enter the one-word possessive construction. In such cases, then, the two noun classes use the same possessive construction. As observed by Krasnoukhova (2012: 87–88), this is very common among Amazonian languages having a class of inalienable nouns. She found that such languages even more often use the same construction with inalienable and alienable nouns (22 out of 41 languages) than a different construction (18 out of 41 languages; Krasnoukhova 2012: 87–88). I will thus argue that Harakmbut does not show a possessive split based on the distinction between bound and independent nouns. In fact, the data will point to a possessor-governed coding split with body-part possessees: animal possessors require a different coding than human possessors.

Corroborating evidence for the finding that adnominal possession is not determined by the alienability contrast in Harakmbut comes from data showing the distinct morphosyntactic behavior of bound nouns in grammatical environments other than adnominal possession. For instance, the data show that the bound nouns that enter the one-word possessive construction in (2b) can also fuse phonologically with other types of prenominal modifiers, such as interrogative, demonstrative, or quantifying modifiers, unlike independent nouns. As is the case in adnominal possession, the one-word strategy is invariably in competition with the two-word strategy. Moving beyond adnominal modification, the article will also hone in on noun-noun compounding and noun incorporation, in which bound nouns also behave differently from independent nouns. All in all, the article concludes that while the alienability contrast could be argued to motivate semantically the noun class system in Harakmbut, it falls short in accounting for the distinct morphosyntactic behavior of bound nouns in adnominal modification and beyond. Rather, this behavior is determined by the morphological property of boundness, which also takes effect in constructions that do not involve the semantics of possession.

As the prefixes wa(ʔ)- and e(ʔ)- serve not only to give independent nominal status to bound nouns but also to derive nouns from verb stems (or action nominals from predicates-cum-arguments) (see Van linden 2019), this article will also take nominalization into account. It will investigate the distinction between bound nouns and deverbal nouns in the various grammatical environments examined here, focusing on deverbal nouns as possessees in adnominal possessive constructions, but also as heads in other adnominal modification structures, and as N1 or N2 in noun-noun compounds. It will become clear that deverbal nouns always use the two-word strategy in any type of adnominal modification, featuring the nominalizing prefixes, but that they drop their prefix when occurring as N2 in noun-noun compounds, just like bound nouns. Unlike the latter, however, deverbal nouns are not incorporable into verbs.

In terms of data, this article draws on earlier work on Harakmbut, which has mainly focused on the Arakmbut variety (Hart 1963; Helberg Chávez 1984, 1990; Tripp 1976, 1995), as well as fieldwork conducted by the author. The first-hand data recorded in the field include elicited data and a collection of seven texts representing spontaneously produced language recorded in the native communities of Puerto Luz, San José del Karene and Shintuya, all with Arakmbut consultants, in the summers of 2010, 2011 and 2016. Additional elicited data were also collected at a distance. The practical orthography used in this article is IPA-based and different from the community spelling (see Van linden 2020: 9, Note 2).

The discussion is organized as follows. Section 2 discusses the distinction between bound nouns and independent nouns at the word level, concentrating on the distinct semantic domains covered by the members of either class. It also introduces verb-based nominalization. Section 3 moves to the phrase level, and focuses on adnominal possessive constructions, classifying them in terms of types of possessees (independent, bound and deverbal nouns) and possessors (human, animal and inanimate), and also considering alienable construal of bound nouns denoting body parts. Section 4 hones in on the morphosyntactic behavior of bound, independent and deverbal nouns in adnominal modification constructions other than adnominal possession. Section 5 leaves the phrase level and explores the morphologically complex word-level phenomenon of noun-noun compounding as well as the clause-level phenomenon of noun incorporation, again distinguishing between bound, independent and deverbal nouns. The article winds up with a brief summary and its conclusions on the relevance of the alienability contrast in Harakmbut in Section 6.

2 Bound, independent and deverbal nouns

This section takes a closer look at the two noun classes in Harakmbut, bound and independent nouns, and investigates whether they could be regarded as reflexes of the alienability contrast in terms of the semantic domains they cover. It will also introduce the nominalizing function of the two noun prefixes wa(ʔ)- and e(ʔ)- on verb stems.

Common nouns in Harakmbut divide into two morphologically distinct classes, viz. bound and independent nouns.^[2] While independent nouns can occur on their own without any additional morphology (see (1a)), bound nouns do need a noun prefix in their citation form, viz. wa(ʔ)- (3) or e(ʔ)- (4).^[3] As these noun prefixes enable bound nouns “to stand alone without possession”, they serve the function of what Nichols (1988: 597) has called absolutivization: they derive independent nouns from bound ones. In Chousou-Polydouri et al.’s (this issue) terms, they are markers of the unpossession construction.

Typically, bound noun stems combine with a single noun prefix exclusively, as in (2a) and (3), but some combine with both prefixes, yielding two different independent nouns whose referents show similarity in shape. For example, the bound root in (4) can also combine with the prefix wa(ʔ)- to denote the body part ‘rib’. Another exception is given in (5), in which the wa(ʔ)- form denotes a body part (5a), while the e(ʔ)- form denotes a plant part (5b), similar to (4).

As can be seen in Table 1 below, there are also plant parts that exclusively take wa(ʔ)-, but overall the generalization can be upheld that referents of bound nouns with e(ʔ)- belong to the world of vegetation. This generalization supports an analysis of the absolutivizing prefixes as derivational rather than inflectional (see also Danielsen (2007: 119–120) on the unspecified possessor prefix e- in Baure, Arawak, Bolivia).^[4]

Let us now turn to the semantic domains encompassed by the referents of bound nouns, which are listed in Table 1 with a non-exhaustive set of examples. The table also includes independent nouns belonging to the same semantic domain if available. What may strike the reader immediately is that bound nouns predominantly denote entities that are – in conceptual terms – inalienably possessed, such as body parts, plant parts, landscape parts, kinship terms, social and spatial relations, attributes, basic shapes, substances, and other parts of wholes (see Rose and Van linden, this issue). The Harakmbut data thus support the implicational hierarchy for membership of the ‘inalienable’ class of nouns proposed by Nichols (1988: 572), in which kin terms and/or body parts outrank part-whole and/or spatial relations, which in turn outrank culturally basic possessed items; the latter are invariably lexicalized as independent nouns in Harakmbut.

Interestingly, Table 1 also includes items that are less easily categorized as conceptually inalienably possessed and which some languages treat as non-possessible, viz. animals (see Lehmann 1998) for wild animals in Yucatec Maya; see also Chousou-Polydouri et al. (this issue). For this semantic category, independent nouns far outnumber bound nouns in Harakmbut. In relation to kinship terms, it should be noted that the independent nouns in Table 1 are in fact terms of address which have come to be used as reference terms (see Tripp 1995: 175–185) (see also Bril [this issue] on Kanak languages). With respect to landscape parts, bound nouns tend to refer to elements determining the physical shape of a landscape, while independent nouns refer to types of soil cover.

By and large, we can conclude that the class of bound nouns is semantically homogeneous and that the two noun classes can be regarded as a reflex of the alienability contrast surfacing at the word level, albeit not a perfect one. Specifically, since not all conceptually inalienable entities are lexicalized as bound nouns (e.g., kinship terms, bodily excretions) and since a few bound nouns denote conceptually non-possessible entities (animals), the data rather support Nichols’ (1988: 574) claim that inalienability is a lexical category rather than a semantic property. Yet, according to the criteria used by Chousou-Polydouri et al. (this issue) in their crosslinguistic study, the class of bound nouns in Harakmbut would rate as semantically coherent.

In addition to covering certain semantic domains, Nichols (1988: 564) also notes that bound nouns form a closed set, which may vary considerably in size across languages. In terms of the criterion of loan integration, bound nouns in Harakmbut form a closed class indeed: loan words invariably enter the class of independent nouns. For instance, even though there is a Harakmbut bound noun for ‘grandmother’, -mama,^[5] the Spanish loan abuela is also often used, and it is grammatically treated as an independent noun in spite of its inalienable semantics. The issue of productive compounding or derivation is not generally regarded as a criterion for open versus closed classes. However, if it were, the class of bound nouns could be regarded as (semi-) open because Harakmbut has many morphologically complex lexical items that contain bound noun roots and take a noun prefix, such as (6) and (7), and because such processes of word formation are productive (see also Section 5.1). Note that (6) combines (5a) and (3), with (3) being the head of the compound. The noun prefixes occur in (6) and (7) because the first elements in these complex items are bound nouns, namely -mbaʔpih ‘finger’ and -wẽ ‘liquid’ respectively.

The prefixes wa(ʔ)- and e(ʔ)- serve not only to derive independent nouns from bound ones; they also serve in verb-based nominalization. Nominalization with wa(ʔ)- is restricted to participant nominalization, viz. for instrumental and objective nominalization (see Comrie and Thompson 2007: 338–342), as in (8) and (9) respectively. It is primarily used to produce nouns for NP-use (Van linden 2019: 465–467).^[6]

Nominalization with e(ʔ)-, in turn, is used for both event nominalization and participant nominalization (see Van linden 2019: 468–484). The type of participant nominalization attested is that of objective nominalization, illustrated in (10). Just as in (9), the nominalized form in (10) denotes the result or the typical or ‘cognate’ object of the action designated by the verb; it can be used as the head of an NP.

In terms of semantic domains, instrumental nominalizations will typically denote items that are – in conceptual terms – alienably possessed, such as artifacts (8a) or accessories (8b), while objective nominalizations are often inalienably possessed. Example (9a) is an attribute, just like eʔ-mbaʔaʔ ‘work’. Other examples are in Table 2; note that the prefix e(ʔ)- is also used in the citation form of verbs (Van linden 2019: 482–483).

In conclusion, deverbal nouns feature the same prefixes as bound nouns and may even show inalienable semantics similar to bound nouns. However, they are not analyzed as members of this noun class because of their verbal origin and their distinct morphosyntactic behavior in several grammatical environments. The latter will be exposed in the remainder of this article.

3 Adnominal possession

This section moves to the phrase level and discusses adnominal possessive constructions. It is organized according to types of possessees: independent nouns (Section 3.1), bound nouns (Section 3.2), and deverbal nouns (Section 3.3), and will focus on different types of possessors (human, animal and inanimate). It will argue that Harakmbut does not manifest an alienability split; it merely shows a coding split for body-part possessees that is based on the human versus animal nature of the possessor.^[7] In addition, there is no pronoun/noun split.

4 Other types of adnominal modification

We now move to non-possessive types of adnominal modification to reveal that the coding strategy exclusive to bound nouns in adnominal possession, i.e., the one-word, prefixless strategy, is also observed for other types of modifiers. This finding corroborates our conclusion that this coding strategy is not an alienability phenomenon. Independent nouns and deverbal ones, by contrast, will always use the two-word strategy with these types of adnominal modifiers. We will take a closer look at interrogative, demonstrative and quantifying modifiers.

The generalization that emerges is that bound nouns show the same two coding strategies as observed for possessive modifiers when combined with other modifiers that obligatorily precede the nominal head in continuous noun phrases: they either attach to a noun prefix and follow the modifier in a separate word, or they directly attach to this modifier, dropping the noun prefix (see Van linden 2023: 453–454). Independent nouns and deverbal nouns will always follow the modifier as a separate word. Consider Examples (27)–(29).

In (27), the interrogative modifier kate modifies the independent noun aypo ‘food’, and the modifier-head structure consists of two words. The same goes for (28), featuring the deverbal noun wawadn ‘seat’ as head. In (29), by contrast, kate modifies the bound noun -nda ‘fruit’, which attaches to the modifier; the modifier-head structure forms a single stress domain, with the stress falling on the penultimate syllable, i.e., the last syllable of the modifier – stressed syllable nuclei are underlined (remember that clitics do not belong to the stress domain). Alternatively, the modifier-head structure can also use the two-word strategy kate wa-nda=pi without a difference in meaning.

The same behavior is found with demonstrative modifiers, for example, the distal form ken ‘that, ‘those’ in (30)–(32).

In (30) ken modifies the independent noun wettone ‘woman’, while in (31) it modifies the deverbal noun wawadn ‘seat’. Both types of head noun only allow the two-word modification construction. In (32), in turn, ken modifies the bound noun -siʔpo ‘child’, forming one phonological word with it. As an alternative to (32), the speaker could have used the two-word strategy ken wasiʔpo without a difference in meaning.

In the last set of examples featuring quantifying modifiers, the (a)-examples include independent nouns showing the two-word strategy, the (b)-examples feature deverbal nouns showing the same strategy, and the (c)-examples illustrate bound nouns using the one-word strategy. Example (33) contains a numeral ((a) and (c) come from a list of ingredients for ceviche), and (34) an indefinite quantifier.

The data discussed in this section bear out that the peculiar behavior of bound nouns in adnominal possessive constructions (discussed in Section 3.2), allowing for two coding strategies, is not limited to the expression of possession. Rather, it is a general characteristic of adnominal modification structures in which the modifier precedes the nominal head (note that adjectives do not always precede their head, see Van linden 2023: 457–458). This observation bolsters our conclusion that Harakmbut does not show an alienability contrast at the level of the phrase.

5 Beyond adnominal modification

Having looked at the (simple) word level (Section 2) and the phrase level (Sections 3 and 4), we now venture into yet other levels of analysis for which alienability phenomena have been documented in the literature (see Rose and Van linden, this issue), viz. the complex word level, looking at noun-noun compounding (Section 5.1) and the clause level, discussing noun incorporation (Section 5.2).

6 Conclusion: what can the alienability contrast account for in Harakmbut?

Having looked at the nature and behavior of independent, bound and deverbal nouns at various levels of linguistic organization in Harakmbut, we are now in a position to assess the relevance of the alienability contrast in that language. The description of the two-way noun class system (Section 2) revealed that membership of common nouns in these two morphologically determined classes is to a great extent motivated by the conceptual distinction between inalienably and alienably possessed items. While bound nouns predominantly include the former type, independent nouns primarily comprise the latter. There are exceptions at either end, which adduces evidence for Nichols’ (1988: 574) position that inalienability is a lexical category at the word level rather than a semantic property (see also Nichols and Bickel 2013). Yet, going by the criteria proposed by Chousou-Polydouri et al. (this issue), the two noun classes would qualify as semantically coherent in terms of inalienable versus alienable.

Although the alienability contrast could be argued to motivate the noun class system in Harakmbut, it was found to be irrelevant to adnominal possession, or to phrase-level phenomena at large. While languages showing bound versus independent nouns typically show coding splits in adnominal possession determined by membership of the possessee in these noun classes (see Rose, this issue; Rosés Labrada, this issue), the Harakmbut data only pointed to a coding split according to humanness for a set of bound nouns (i.e., body parts). In adnominal possession constructions, but also in other types of adnominal modification constructions, bound nouns were shown to use the same (two-word) coding strategy as independent nouns. They differ from the latter in also showing another, one-word coding strategy, but only in a lexically skewed way. That is, some bound nouns do not accept one-word modification constructions. While the distinct behavior of bound nouns in adnominal modification, compared to that of independent and deverbal nouns, can be explained by their having a nominal root or stem and being morphologically bound, the reasons for the differential acceptance of the one-word, prefixless strategy across the class of bound nouns remain unexplored.

With no alienability split at the phrase level, there was little hope of finding reflexes of the alienability contrast in the complex-word phenomenon of noun-noun compounding. Bound and independent nouns were found to merely differ in preference for the N1 or N2 position; neither class is excluded from either position. Deverbal nouns, in turn, behave identically to bound nouns in dropping their prefix in N2 while keeping it in N1. However, we did note differences with the two common-noun classes in that deverbal nouns only combine with independent nouns in N1, or bound nouns in N2; the reverse combinations are excluded.

Finally, we investigated noun incorporation and found that the explanatory potential of the alienability contrast for this clause-level phenomenon is similar to that for the (simple) word-level phenomenon of the noun class system. Inalienable semantics could be argued to determine the incorporability of nouns, but there are also exceptions.

More generally, this article has shown that although a language may manifest alienability oppositions at the lowest level of organization, viz. the word, this does not necessarily entail the presence of an alienability split in adnominal possession. This is in line with Krasnoukhova’s (2012: 87–88) finding that South American languages with inalienable and alienable nouns do not always use different possessive constructions with either class. In fact, they more often use the same possessive construction. In Harakmbut, then, the relevance of the alienability contrast seems to be limited to having motivated the morphological distinction between bound and independent nouns, which in turn motivates the distinct behavior of bound and independent nouns in various grammatical environments. The most intriguing aspect of this distinct behavior is the choice that bound nouns offer to speakers: when do speakers choose the two-word strategy and when do they select the one-word strategy? Is the competition between the morphosyntactic patterns discourse-related, semantically motivated, or lexically determined? These questions are left for further research.