Introduction: the limits of the explanatory potential of the alienability contrast

4.1 Highlights of the issue

The contributions to the present special issue re-assess the explanatory power of the alienability contrast by pushing the boundaries of the types of possessive situations to be studied, specifically by widening the scope of relevant data in three respects. First, they pay attention to a wider variety of possessive relationships. As mentioned in Section 2, studies on possession traditionally focus on the canonical relations of ownership, part-whole and kinship relations, while possession as a grammatical concept encompasses many more relation types. The articles in the special issue will not limit themselves to these canonical possessive relationships, but also address, among others, associative possession, social relationships, and participants-in-event relations (e.g., the articles by Aralova and Pakendorf, Bril, Creissels, Rose, Van linden). Second, the articles take into account diverse types of possessors, while the literature is often restricted to human possessors only. This bias towards human possessors is apparent from most definitions of possession. Hollmann and Siewierska’s (2007: 410) definition is a case in point: “Inalienable possession is generally seen as involving a fairly stable relation over which possessors have little or no control, alienable possession as comprising a variety of less permanent, more controlled relationships” (emphasis added). As inanimate possessors do not show variability in terms of more or less control, these have often been overlooked. In their description, the analysts will move beyond this anthropocentric bias in the literature on possession and extend their attention to non-human possessors (e.g., Rose, Rosés Labrada, Van linden). This is actually crucial for having a complete view of the expression of possession in individual languages. Rosés Labrada, for example, shows that in Piaroa inanimate possessors make use of a fully distinct possessive construction from the ones recruited by animate possessors. Third, we are not limiting ourselves to canonical possessees either, and include for example ‘intentional parts’ (e.g., ‘soul’), cultural aspects of life (e.g., ‘job’), and event-denoting nouns as possible possessees (e.g., Bril, Creissels, Van linden).

The present special issue also differs from previous studies with a crosslinguistic approach by including three studies on South American languages, one from a major and well-studied family, the Arawakan family (Rose), one from a small family, Jodï-Sáliban (Rosés Labrada) and, third, an unclassified language, Harakmbut (Van linden). In doing so, it fosters further reflection on possessive classifiers, derivation among possessive noun classes (i.e., lexical classes of nouns depending on their behavior regarding possession), and the behavior of various noun classes in compounding and noun incorporation.

The articles in the issue discuss the manifestations of the alienability contrast at different levels: (i) the word level with the question of possessive noun classes (Aralova and Pakendorf, Bril, Chousou-Polydouri et al., Rose, Rosés Labrada, and Van linden), compounding (Van linden), and proprietive markers (Aralova and Pakendorf), (ii) the phrase level with the major question of adnominal possessive constructions (Aralova and Pakendorf, Bril, Chousou-Polydouri et al., Creissels, Rose, Rosés Labrada, and Van linden) but also that of non-possessive adnominal modifiers (Van linden), (iii) the clause level with questions regarding noun incorporation (Van linden), and (iv) the sentence level, with complex sentences featuring possessed modal nouns taking complement clauses (Bril). In fact, several contributions to the issue (Aralova and Pakendorf, Bril, Chousou-Polydouri et al., Rose, Rosés Labrada, and Van linden) build their analyses of the expression of possession on drawing a clear line between possessive noun classes and adnominal possessive constructions (i.e., the word and the phrase level).

One word-level phenomenon that will be addressed in several articles is that of noun classes; common nouns fall into distinct classes depending on how they behave morphosyntactically in adnominal possessive constructions and in unpossessed contexts. Some languages discussed in the present special issue show more than two possessive noun classes, including at least bound nouns, or obligatorily possessed nouns, and two sets of independent nouns, viz. those that are optionally possessed nouns, and those that cannot directly carry possessive personal markers (often called non-possessible or non-directly possessible nouns, see Footnote 4), as in the articles by Aralova and Pakendorf, Bril, Rose, and Rosés Labrada. The binary alienability contrast only sets the first class apart from the remaining ones. This first class typically includes nouns that are – in conceptual terms – inalienably possessed. Interestingly, this class often includes items that have inanimate possessors, like ‘(the tree’s) leaf’, ‘(the landscape’s) river’, and ‘(the path’s) half’. The remaining classes, in turn, typically comprise nouns that are alienably possessed or non-directly possessible, but the alienability contrast leaves the distinctions among them unexplained unless a gradable view of alienability is adopted. As the definition of alienability is highly abstract, semantic analyses in terms of the alienability contrast are difficult to evaluate, which is further aggravated by the circumstance that variable treatment of individual items as alienable or inalienable is often dismissed as being culturally specific. Overall, the contributions thus tend to support Nichols’ (1988: 574) view of alienability as a lexical property of nouns, but at the same time also show that possessive classes need not be limited to two in a single language.

At the phrase level, the articles of the issue illustrate two types of split in adnominal possession, governed by either the possessee or the possessor. Interestingly, in three languages a superficial look at the data suggests an analysis in terms of an alienability split, lexically constrained by the possessee, but the bigger picture points to a split conditioned by the animacy of the possessor (Creissels, Rosés Labrada, and Van linden). In one of these languages, the animacy of the possessor plays a role for one noun class only (Van linden), while in the other two, the possessor-determined split takes precedence over a possessee-determined split (Creissels, and Rosés Labrada).

Focusing on possession-determined coding splits, the articles largely confirm Haspelmath’s (2017) universal presented in Section 2, although Cristofaro’s article presents counterexamples to it. It should also be noted that languages that show more than two possessive noun classes somewhat spoil the picture. In some such languages, nouns that cannot be directly possessed can nevertheless enter some possessive constructions. These constructions then show more morphological marking than the basic possessive construction (by including for instance derivational marking, generic possessive nouns, possessive classifiers, or non-canonical possession marking). Being more marked morphologically than obligatorily possessed nouns, they thus verify again the alienability contrast when compared to the latter. But at the same time, the alienability contrast does not account for the distinction between the nouns that cannot be directly possessed and those that can (but do not have to), as they both include referents that are semantically possessible but for which possession is marked differently.

Interestingly, the word-level and phrase-level phenomena discussed in the previous paragraphs (possessive noun classes and adnominal possessive constructions) are usually “explained” on the basis of the alienability contrast. What is remarkable is that some languages investigated in the current issue, for example, Piaroa (Rosés Labrada), Zuanga-Yuanga (Bril), Negidal (Aralova and Pakendorf) and Mojeño Trinitario (Rose), require a distinction between the word level (noun classes) and the phrase level (adnominal possession) because there is no one-to-one correspondence between the categorization at these two levels. Consequently, the explanatory power of the alienability contrast is weakened by being used ad hoc for two phenomena that do not coincide. The articles by Aralova and Pakendorf, and Rose even go one step further and propose accounts in which alienability is, at best, epiphenomenal and in any case marginal. These contributions propose alternative accounts to the traditional analysis of the expression of possession in specific language families, viz. in Arawakan and Tungusic respectively, that could be fruitfully extended to other languages of those families.

As language-specific work on the alienability contrast often deals with various levels of organization, as detailed above, crosslinguistic comparison of adnominal possessive constructions is not straightforward. For instance, one could choose to compare the semantic categories picked out by alienability contrasts, or rather focus on their morphosyntactic patterns. The matter of crosslinguistic comparison is addressed in the issue by Chousou-Polydouri et al. They use a specially designed database to look at alienability phenomena from multiple perspectives, considering noun possession classes and adnominal possessive constructions independently of each other. Their methodology also allows for the inclusion of languages with non-possessible noun classes and multiple non-default possession classes alongside languages with a more traditional inalienable/alienable noun class distinction.

Finally, the present issue approaches the questions of the functional motivation for alienability splits and their emergence. At the synchronic level, the article by Rose offers a quantitative survey of the distribution of Mojeño Trinitario nouns of different possessive noun classes as possessed or not in discourse, following the method used in Haspelmath (2017) for Biblical Hebrew, Ancient Greek, and English (three languages which are known not to show an alienability contrast). Even though there is a general association between possessive noun classes and frequency of use in possessive constructions, the results show that nouns cannot be successfully predicted to belong to one class or another depending on the relative frequency of their possessed occurrences in discourse. At the diachronic level, the articles by Creissels and Cristofaro investigate the semantic distinctions that may have played a role in the constructions at the origin of possessive coding splits, going against explanations in terms of iconicity (Haiman 1983) and predictability or frequency (Haspelmath 2017). All of these articles thus contribute to the special issue’s conclusion that possessive splits are a multi-factorial phenomenon and that an all-encompassing explanation is, and will likely remain, elusive.

All the contributions to the present issue therefore concur in showing that the explanatory power of the alienability contrast is not sufficient to explain the complexity of the data. At best, it can be regarded as a heuristic tool for exploring the data. In other words, if used as the only explanatory principle, the alienability contrast could actually hamper an adequate description of the data, which are sometimes even better described without recourse to alienability at all. The data presented in the issue show that noun classes and all morphosyntactic phenomena often explained in terms of the alienability contrast cannot be reduced to it; other factors also come into play, such as specific semantic characteristics of the possessee (whether it is edible, for example), of the possessor (animacy, for example), or of the possessive relationship (such as different kinship relations), as well as language contact and diachrony. As hinted at in several places above, we believe that some of the limits and problems of the alienability contrast stem from an ill-defined concept of possession. While many accounts focus on some ‘canonical’ possessive situations (very often the possession of a body part by a human possessor), the articles in the issue reveal that the reality of grammatical possession is much wider in the production of speakers of various languages.

4.2 Summaries of the contributions

We start the special issue with a crosslinguistic survey of differential adnominal possession by Chousou-Polydouri, Inman, Huber and Bickel. The article first carefully describes the methodology used for the survey, namely a questionnaire forming the structure of a detailed multi-level and relational database. This questionnaire investigates both the word and phrase levels, distinguishing possessive noun classes and adnominal possessive constructions. Each noun class is lexically defined by having a unique way of being possessed and not possessed (i.e., by its noun valency and valency-changing mechanisms). The semantic domains covered by each class in individual languages are also surveyed. The article then offers the first results of this approach applied to an areally and genetically diverse sample of 120 languages. Important results of the survey include that about half of the languages have several possession classes, and these are mostly semantically coherent. Classes of conceptually inalienable nouns are rather frequent (but only about a third consist of obligatorily possessed nouns), while classes of conceptually non-possessible nouns are rare. Furthermore, the findings on the semantics of nouns that make up different possessive noun classes, analyzed as semantic networks, confirm the existence of a universal semantic core for the classes of inalienable nouns (precisely kinship and body parts) and non-possessible ones (essentially animals). However, the existence of lexical possessive classes is not universal, as half of the languages have no possessive noun class distinction.

The next six contributions address the role of the alienability contrast in the analysis of individual languages. Bril’s article goes back to the roots of the alienability contrast, studying adnominal possession in a number of Oceanic languages in the footsteps of Lévy-Bruhl (1914). However, the data from Zuanga-Yuanga and other northern Kanak languages of New Caledonia prove too diversified to be meaningfully captured by the binary contrast, both at the word and phrase level. Bril shows that the possessive noun classes are not based on the alienability contrast. For instance, nouns from a single semantic domain (e.g., kin terms and body parts) are spread across different noun classes. There are four main possessive noun classes (indirectly possessed free nouns, nouns possessed with classifiers, directly possessed bound nouns, and directly possessed free nouns) and a fifth small class of non-possessibles; these classes are determined by their members’ access to both possession and non-possession constructions. In addition, there is a set of nouns denoting parts of wholes – not distinguished as a separate noun class by Bril – that behave differently: they have access to three distinct possessive constructions depending on the semantic type of part-whole relation expressed, from more inherent to more temporary or contingent. The article thus also addresses the question of what types of semantic relations can be coded by adnominal possessive constructions. In addition to the prototypical relationships, including different kinds of part-whole relations, the data point to participants-in-event relations for deverbal-noun possessees and to that of modal agents (Verstraete 2005: 1402) for modal-noun possessees like ‘duty’ or ‘will’.

The issue then crosses the Pacific, presenting three contributions on South American languages, which have often been neglected in earlier work on alienability. First, Rosés Labrada analyzes the mismatch between possessive noun classes and adnominal possessive constructions in Piaroa (Jodï-Sáliban, Venezuela and Colombia). In this language, there is a possessor-determined split: only animate possessors are indexed on the possessee via affixation; inanimate possessors occur in a juxtaposition construction with their possessees. In constructions with animate possessors, two different possessee-determined splits – and therefore, Rosés Labrada argues, four possessive classes – can be identified in Piaroa. One split relies on a contrast between obligatorily-possessed nouns (kinship and body-part terms) and optionally-possessed ones (other semantic categories), and thus can be said to be motivated by the alienability contrast. The other split is dependent on the type of possessive construction, namely, direct versus indirect possession, a given possessee noun can occur in. All obligatorily-possessed (“inalienable”) nouns and some optionally-possessed (“alienable”) nouns occur in the direct construction. Other alienable nouns, such as those denoting food and drinks, by contrast, occur in the indirect construction, which features possessive (or genitive) classifiers. Importantly, the alienability contrast cannot account for the distribution of possessees across the different Piaroa possessive constructions: alienability yields two classes that only partially overlap with the two classes that stem from the contrast between direct and indirect possession. Rosés Labrada therefore concludes that two concurrent systems of possessive classification must be posited to account for the data.

The second contribution on South American languages, by Rose, focuses on Mojeño Trinitario (Bolivia) but has implications for the whole Arawakan family. While existing accounts propose that the alienability contrast is at work in Arawakan languages at the word level, distinguishing between alienable and inalienable noun classes, and at the phrase level, through the competition between two adnominal possessive constructions, Rose breaks radically from this tradition, by bringing also a third class of non-(directly) possessible nouns into the picture and by attributing a central role to derivation among noun classes. Applying her account to the whole family would leave most Arawakan languages without a class of optionally possessed (or “alienable”) nouns, as these would be stems derived from non-directly possessible noun roots. With respect to the class of obligatorily prefixed (or “inalienable”) nouns in Mojeño Trinitario, Rose finds that it is a large class, semantically more heterogeneous than the class of optionally prefixed (or “alienable”) nouns and that it is receptive to loans, characteristics that go against Nichols’ (1988) generalizations (see Section 2). More generally, Rose identifies as the main theoretical problem behind investigating the explanatory power of the alienability contrast that it is unclear to which levels of linguistic organization it should apply a priori. At any level considered, its explanatory potential in Mojeño Trinitario – and by extension Arawakan languages – is weak.

The last South American language reported on is Harakmbut (unclassified; Peru). Van linden offers a detailed description of two noun classes (independent and bound) and two adnominal possessive constructions (one-word vs. two-word). At the word level, the distinction between bound and independent (or free) nouns is rather well accounted for by the alienability contrast, with most bound nouns being semantically inalienably possessed and independent nouns being construed as alienable. Interestingly, this article also brings into the discussion the behavior of deverbal nouns, which do not neatly behave as either bound or independent nouns. At the phrase level, while at first sight, the distinction between the two adnominal possessive constructions could also look like a perfect illustration of the binary contrast of alienability, the intricate interplay between the noun classes and the possessive constructions leads to a much more complex picture of possession in Harakmbut. Furthermore, the distribution of the two noun classes also differs in contexts other than adnominal possession, such as in noun-noun compounding at the word level, with interrogative, demonstrative, or quantifying modifiers at the phrase level, and noun incorporation at the clause level. The alienability contrast does not satisfactorily explain the distinct morphosyntactic behavior of the two noun classes, neither in adnominal possession nor in the other morphosyntactic contexts.

The next article moves to Eurasia, and is also fieldwork-based; Aralova and Pakendorf draw on a large corpus of oral speech of Upper Negidal (Tungusic, Far East of the Russian Federation). Their contribution goes against traditional accounts of differential possession marking in Tungusic, according to which the suffix -ŋ(V) marks alienable possession, e.g., on body parts belonging to a non-inherent possessor. Distinguishing three possessive classes, they argue instead that the suffix -ŋ(V) signals non-canonical possession. It is found on members of two classes only, viz. on obligatorily possessed body-part terms (yielding an “alienable” interpretation), and on non-possessible items. The latter class is argued to not only include nouns denoting entities like humans, wild animals, plants and substances, but also non-nominal items (e.g., numerals, demonstratives, adjectives and participles). Importantly, what all of these share is that they require the suffix -ŋ(V) to accept possessive suffixes in adnominal possession or the proprietive suffix in predicative possession. The types of semantic relation coded by adnominal possessive constructions, with or without the suffix -ŋ(V), include associative possession, which establishes a relation of salience between two discourse participants. A semantic relation found with -ŋ(V)-marked constructions only is that of particularization, which delimits a mass quantity (e.g., ‘your water’ for ‘the water you need in the house for washing and drinking’). Aralova and Pakendorf also extend their analysis to a cognate suffix found only on possessors and build the case that both of them (when formally distinct in individual languages) flag non-canonical possession across the Tungusic family. The alienability account in prior literature, by contrast, would leave the majority of occurrences of -ŋ(V) unexplained.

A similar conclusion is arrived at for Mandinka (Mande, Gambia, Senegal, and Guinea Bissau) in Creissels’ article, which represents the African macro-area in the present issue. The language shows no noun classes but has a coding split in adnominal possession. While the differential possessive marking for body parts, kin terms and artifacts may look as if conditioned by alienability, extending the scope to non-prototypical possessive relations (privileged relationships of any kind) bears out that the split for ordinary nouns is governed by animacy. That is, inanimate possessors by default require the direct possessive construction (juxtaposition), whereas animate possessors require the indirect possessive construction (with linker), except for body parts or (a set of) kin terms. These exceptions are accounted for by the etymology of the possessive marker lá, according to which the indirect construction originally meant ‘the Y located near X’, and for kin terms also by specific cultural practices. Extending the scope to non-prototypical possessees, viz. deverbal nouns, the article further argues against an alienability split: the split between the two possessive constructions is determined by the syntactic role of the possessor in the corresponding clause (notional objects use the direct construction, while notional subjects use the indirect one). All in all, the data show that the Mandinka coding split is possessor-governed and that semantic features such as animacy and control may play an important role in the genesis of possessive splits.

The issue concludes with Cristofaro’s article, which concentrates on the alienability contrast as an explanandum itself and discusses data on the genesis of possessive splits. By investigating the diachronic development of overt markers involved in adnominal possession splits in a great variety of languages, this contribution shows that the usual explanation of these splits by the alienability contrast is inconsistent with these diachronic processes. The origin of these markers goes back to some use in constructions for which the alienability contrast was irrelevant, such as localization, for example, as is the case for Mandinka (Creissels, the present issue). Moreover, the synchronic distribution of these markers is best explained by the semantic properties of their source constructions, rather than by some general functional explanation that would apply a posteriori to the result of unrelated historical developments. Cristofaro’s conclusions thus seriously challenge the idea that possessive splits reflect general preferences for the use of overt marking for alienable possession.