Non-canonical possessive constructions in Negidal and other Tungusic languages: a new analysis of the so-called “alienable possession” suffix

1 Introduction

So-called (in)alienable possession is described as a crosslinguistically common feature whereby two distinct kinds of possession – inherent and permanent versus loosely associated and non-permanent – are formally distinguished in adnominal constructions (e.g., Chappell and McGregor 1996; Haspelmath 2017; Nichols 1988; see also the overview of the literature in the introduction to this issue). This category has also been widely noted in the Tungusic languages (Boldyrev 1976; see Avrorin [1959: 155–163] specifically for Nanai, Boldyrev [2007: 122–139] for Evenki, Nikolaeva and Tolskaya [2001: 135–141] for Udihe, Novikova [1960: 145–152] for Even, and Pevnov and Khasanova [2006: 503–504] for Negidal), which are also included by Nichols (1988: 591–592) in her classification of (in)alienable possession types. The Tungusic distinction is most commonly illustrated with the contrast between inherently possessed body parts (1a) on the one hand and body parts of dead animals that have entered the possession of some human (1b) on the other, with the latter being formally marked by a suffix -ŋ(V). Such examples appear to show that this formal opposition in possession marking is indeed one of inalienable versus alienable possession.

However, in Negidal, a critically endangered language spoken in the Far East of Russia, the so-called alienable possession suffix [which takes the form -ŋ(i)] occurs in contexts that cannot be explained by any form of loose association or “socially or economically conferred ownership” (Nichols 1988: 568), such as with human referents or with the sun, casting doubts on an analysis in terms of alienability. Nikolaeva and Tolskaya (2001: 135–141) also point out that in Udihe the cognate suffix occurs in many more contexts than simply alienable possession; they describe the suffix as having five distinct functions (see Section 4.3 for details).

We here address the question whether an analysis of the suffix -ŋ(i) in terms of marking alienable possession is warranted by the Negidal data. As will be seen from the label poss (standing for ‘non-canonical possessive construction’) with which we gloss this suffix, and as is laid out in detail in Section 4, we provide an analysis in which the function of -ŋ(i) is not to mark alienable possession, nor does this suffix have various different functions as postulated for the Udihe cognate by Nikolaeva and Tolskaya (2001: 135–141). Rather, the seemingly diverse uses of this suffix in Negidal can all be subsumed under a single function, namely to flag the occurrence of non-canonical possessive constructions. These data illustrate how apparently straightforward examples of the so-called alienability contrast might upon close examination turn out to have very different underlying motivations, thus raising the possibility that other cases of supposed (in)alienability might in actual fact also have different causes (see also Rose, this issue).

We extend our discussion to the closely related suffix -ŋi, which marks possessors in constructions with elided head nouns (hence the gloss psr “possessor of elided head”). As will be detailed in Section 7, such constructions are non-canonical from a syntactic perspective and thus complement our understanding of non-canonical possessive constructions not only in Negidal, but also in other Tungusic languages.

The remainder of the article is structured as follows: after a brief introduction to the language and the data that form the basis of this investigation (Section 2), we introduce possessive constructions in Negidal in Section 3. In Section 4 we discuss the uses of the suffix -ŋ(i), which was previously considered a marker of alienable possession. In order to situate the Negidal data in a broader context we provide a brief overview of the cognate suffixes in other Tungusic languages (Section 5). Section 6 is devoted to the description of the suffix -ŋi, which marks the possessor in possessive constructions with elided possessee in Negidal as well as in other Tungusic languages. The article concludes with a discussion of the different non-canonical possessive constructions that are marked by these two clearly cognate suffixes (Section 7).

2 Negidal: the language and the data

Negidal is a critically endangered Northern Tungusic language spoken in the Far East of the Russian Federation on the Lower Amur river and one of its major tributaries, the Amgun’. At least two dialects used to be distinguished: Upper Negidal and Lower Negidal (Cincius 1982: 3). However, Lower Negidal appears to be extinct by now, and only a handful of speakers of Upper Negidal remain (Pakendorf and Aralova 2018), among whom only two elderly women are still fluent in the language.

Like its relatives, Negidal is a fairly agglutinative language with rich, exclusively suffixing morphology. It has nominative-accusative alignment, with obligatory indexation of the S/A argument on the verb; objects, in contrast, are not cross-referenced on verbs. Nevertheless, both S/A and object (pro)nominal arguments are commonly omitted if their referents are retrievable from the context. Due to vestiges of vowel harmony and consonant assimilation at morpheme boundaries, the surface forms of suffixes show some variation (Pakendorf and Aralova 2020).

This study is based exclusively on a fully annotated corpus of oral speech of Upper Negidal (Pakendorf and Aralova 2017). This comprises about 200 texts of diverse genres (folklore, autobiographical anecdotes, procedural texts, and conversations) numbering approximately 76,500 words in total. Nine elderly speakers – four of whom are by now deceased – are represented in the corpus, eight women and one man. Of these, the man and four women (a mother and three of her daughters) were/are fluent speakers, while the others, including another daughter, show differing levels of attrition (Pakendorf and Aralova 2018; corpus description^[1]).

For our analysis of the contexts of use of the suffixes -ŋ(i) and -ŋi (cf. Section 6 for details on the distribution and morphonological behavior of these suffixes) we extracted all the examples found in the corpus, namely 656 for -ŋ(i) and 31 for -ŋi, and coded them for host (class of lexeme, such as human, animal, plant, mass noun, demonstrative, adjective, or numeral) and specific lexeme (e.g., ‘person’, ‘wood’, ‘good’, ‘this’, etc.); whether the item carrying either of these suffixes carried further derivational, case, or possessive marking and if so, which kind (cf. Section 3.1); the context of use of the suffixes; and the syntactic position of the item carrying them (see coding sheet: Aralova and Pakendorf 2023). We furthermore investigated all the lexemes belonging to particular semantic classes (kinship terms, body parts, animals, plants, food, personal names) to assess to what extent they occur with or without possessive marking in the corpus, and we scanned concordances of the possessive suffixes to gain an understanding of their functions in Negidal.

3 Possession in Negidal

Since an understanding of possessive marking is necessary in order to understand the function of -ŋ(i) and -ŋi, in this section we continue with a brief description of possessive constructions in Negidal (3.1) and the possessive classes (3.2) observed in the corpus.

4 The suffix -ŋ(i) in Negidal

As mentioned in the preceding section, the suffix -ŋ(i), which we gloss as poss, ‘non-canonical possessive construction’, occurs with two categorically opposed classes of items: the obligatorily possessed body part terms and the heterogeneous class of non-possessible items. It always precedes a possessive marker,^[5] either the possessive suffixes summarized in Table 1 or the proprietive suffix -ʨi. Below we provide a detailed discussion of the contexts of use of the suffix, making a distinction between the possession of nouns (Section 4.1) and the possession of formally non-possessible parts of speech (Section 4.2). The use of -ŋ(i) with such disparate classes of items can be explained by viewing it as a flag of non-canonical possessive constructions, as we argue in Section 4.3. Our analysis contrasts with that of Pevnov and Khasanova (2006: 503–504), who discuss the suffix -ŋ(i) in semantic terms, as a marker of alienable possession, as is common in descriptions of Tungusic languages (cf. Cincius 1982: 20); they contrast it with “not-indirect possession” (nekosvennaja prinadležnost’) that occurs with body parts, kin terms, domestic animals and cultural artefacts.

A word of caution is in order: as pointed out in Section 2, we are basing our study purely on our analysis of the Negidal corpus. Given the nature of the corpus (spontaneous oral speech produced by individuals for whom Negidal has ceased to be the primary language of everyday use and who show differing degrees of attrition), counterexamples to our generalizations do occur, and not all of the examples are clear (∼4 % were coded as ‘unclear’ for various reasons, see coding sheet). Nevertheless, since our database of examples is quite extensive, we feel confident in our overall analysis.

5 The marking of non-canonical possessive constructions in other Tungusic languages

We here summarize information on possessive constructions in four relatives of Negidal: Evenki, Even, Udihe and Nanai. Even and Evenki, which are spoken by small speech communities dispersed over the length and breadth of Siberia, are very closely related to Negidal, while Udihe and Nanai, which are spoken in geographical proximity to Negidal, are genealogically distant (Atknine 1997; Janhunen 2012). Our discussion is based on our interpretation of published descriptions and does not necessarily reflect the analysis of or the terms used by the original authors.

Judging from the available sources, the suffix -ŋi in Evenki functions in a very similar manner to -ŋ(i) in Negidal. As in Negidal, kinship terms and body parts appear to be obligatorily possessed, since they practically always occur with possessive suffixes, whereas terms denoting humans, environmental features, plants, wild animals or birds, and nouns denoting uncountable entities (‘water’, ‘bread’, ‘wood’) cannot carry possessive suffixes directly. Associative possession of such entities has to be licensed by the suffix -ŋi (Boldyrev 2007: 126–133; Nedjalkov 1997: 145). Furthermore, when adjectives, participles or demonstratives carry possessive suffixes because the head noun is elided, this possessive marking also has to be licensed with -ŋi (Boldyrev 2007: 133–136).

In Even, too, the suffix -ŋ occurs in the same contexts as -ŋ(i) in Negidal: with proper nouns, terms referring to people, the environment, wild animals and food, as well as adjectives, numerals, participles and demonstratives when the head noun is elided (Novikova 1960: 147–150). Interestingly, Novikova (1960: 141, 147) provides examples of this suffix attached to kin terms, with a reading of social rather than biological kinship (22).

In Udihe, too, possessive constructions are similar to those found in Negidal: while kinship terms and body parts tend to be obligatorily possessed, ownership of a body part by someone other than the inherent ‘possessor’ is marked by -ŋi, as is ‘possession’ of ‘land’ and other landscape terms, terms denoting humans, the words for ‘tree’ and ‘money’ (and, somewhat unexpectedly, ‘cow’), as well as substantivized adjectives (Nikolaeva and Tolskaya 2001: 135–141).

As in Negidal, in Nanai body part and kinship terms (except in vocative use) are obligatorily possessed, while proper nouns and meteorological terms as a rule do not occur with possessive suffixes (Avrorin 1959: 119–127). The ‘indirect possession’ suffix -ŋgo/-ŋgu occurs frequently with non-possessible nouns that are associatively possessed, and can also be added to adjectives and interrogative pronouns (Avrorin 1959: 160–161).

In summary, the types of nouns and other lexemes that require addition of the suffix -ŋ(V) to license possessive marking overlap to a large extent across Negidal, Evenki, Even, Udihe, and Nanai: kinship and body part terms are obligatorily possessed, while terms denoting humans, environmental features, uncountable masses and materials cannot take possessive suffixes unless these are preceded by -ŋ(V). Adjectives, participles, and demonstratives that take possessive marking because they substitute for an elided head noun need -ŋ(V) to license the possessive marking. From this perspective, the analysis we propose for Negidal could be extended to -ŋ(V) in other Tungusic languages, namely that this suffix has as its single function the marking of non-canonical possessive constructions. Like the possession of a body part by an individual who is not the inherent possessor, a kinship relationship that is not a true biological relationship is unexpected and not canonical, explaining why such a relationship is flagged with -ŋ in Even (22), resulting in an apparent marking of alienable possession.

6 Possessor marking in constructions with elided possessee in Negidal and other Tungusic languages

The description of non-canonical possessive constructions in Negidal would not be complete without describing another suffix, namely -ŋi, which was briefly mentioned in the introduction. As opposed to -ŋ(i), which is found on the possessee and its modifiers, -ŋi occurs exclusively on possessors in adnominal possessive constructions with an elided head noun; hence we gloss it psr, ‘possessor of elided head’. This suffix differs slightly in form from that found with possessees: while in the latter case, the base form is -ŋ with -i added as an epenthetic vowel in certain morphonological environments,^[11] the suffix that occurs with possessors has the base form -ŋi. This difference can clearly be seen when comparing two accusative-marked forms, one a possessee (23a) and one a possessor with an elided head noun (23b). In (23a) the bilabial fricative of the accusative case marker assimilates to the velar nasal of -ŋ(i), whereas in (23b) the accusative suffix retains its original shape following the vowel of -ŋi. Nevertheless, the two suffixes are clearly diachronically related (cf. Sunik 1982: 65) as well as functionally similar, as will be discussed in Section 7.

Our analysis of -ŋi is based on 24 clear examples in the corpus (see coding sheet, which also contains several examples with calques from Russian and several unclear cases). This suffix is attested in those cases when the possessee is omitted from an adnominal possessive construction: compare (24a) with the overt head noun sobgo ‘fish skin’ with (24b), where -ŋi substitutes for the head of the possessive phrase ‘skin of a catfish’. Although in examples such as (24c) the possessee is present in the sentence, the possessor having been added as an afterthought, it is not present in its canonical position in the possessive construction, namely following the possessor.^[12]

Since possessive noun phrases are not restricted to any specific position in the clause we find -ŋi-marked possessors in subject (24b), direct object (24c), and predicate position (25), although the latter are rare in the corpus (see coding sheet).

Interestingly, the split in marking a modifier depending on whether it is a possessor (marked with -ŋi) or another type of modifier [marked with -ŋ(V)] is found not only in Negidal, but is also present in some, albeit not all, other Tungusic languages. For Evenki, Bulatova and Grenoble (1999: 14) postulate a formal distinction of the suffix that marks possessors from the so-called marker of indirect possession: they analyze the morpheme that attaches to possessors as having a long vowel (-ŋiː) as opposed to the short vowel they determine for the suffix that marks possessees. However, this distinction is not mirrored in Nedjalkov’s (1997: 123–125) analysis, which appears to identify the two suffixes as one and the same form, an analysis that also seems to be taken by Boldyrev (2007: 122–139).

In contrast, in Even there is a clear formal distinction between the suffix that attaches to the possessee, which consists solely of the velar nasal -ŋ (26a), and the suffix that attaches to possessors in constructions with elided head, which takes the form -ŋi (26b); Cincius 1947: 144, 148; Novikova 1960: 147–152). The latter can occur in subject, predicate, direct object, and attribute position and is called a “special possessive form of nouns” by Novikova (1960: 150–152).

For Udihe, Nikolaeva and Tolskaya (2001: 141, 634) analyze the suffix -ŋi that occurs with possessors as clearly being identical to the “alienable possession suffix” and analyze it as expressing predicative possession: “The suffix -ŋi- here signifies that the possessive relationship is the main predication in the sentence. For obvious semantic reasons, only the alienable possessive relationship can be predicated by the construction in question.”

In contrast to Udihe, and like Even and Negidal, Nanai, too, makes a formal distinction between the suffix that marks possessees (-ŋgo/-ŋgu, [27a], [27b]) and that which marks possessors (-ŋgi, [27c], [27d]; Avrorin 1959: 155–163, 186–192), which Avrorin calls the “predicative-possessive form”. The same split is furthermore found in Uilta, a close sister of Nanai (Ozolinja 2013: 128–133, 253). In Nanai, possessors are marked with the suffix -ŋgi when they occur in predicative position, when they are postposed to their head (as an afterthought), or when the head is elided because it is retrievable from the preceding discourse (Avrorin 1959: 187–189).

Unlike the Tungusic languages spoken in Russia, Written Manchu (attested in northern China) has a genitive case form that marks the possessor in adnominal possessive constructions, but it lacks suffixes that attach to the head noun in such constructions, having neither the direct possession markers nor the suffix -ŋ(V). However, it does have a suffix -(n)iŋgə^[13] that marks the possessor in predicative position or when the possessee is elided (Avrorin 1956: 99, 2000: 74, 86) and that thus corresponds to the suffix -ŋi found in Negidal.

In the Tungusological tradition, the suffix -ŋi is often considered a marker predominantly of predicative possession (Avrorin 1956: 94, 1959: 186–188; Nedjalkov 1997: 123–124; Nikolaeva and Tolskaya 2001: 634–635). In contrast, we analyze the function of this suffix as marking possessors of elided head nouns, with the predicative position of the NP being a secondary aspect that emerges from discourse. Our analysis is fully compatible with the data presented by Avrorin (1956: 99) on Manchu, Avrorin (1959: 186–189) on Nanai, and Novikova (1960: 150–152) on Even; a parenthetical insertion by Avrorin (1956: 97) indicates that the same might hold for Evenki.

We are thus faced with formally similar (and most probably related), yet distinct suffixes in at least three Tungusic languages that appear to have distinct functions, namely the marking of possessees on the one hand and the marking of possessors on the other. Nevertheless, as we discuss in the concluding section, these discrepancies can be reconciled under a unified analysis as markers of non-canonical possessive constructions.

7 Discussion and conclusions

To summarize, the Tungusic languages have one (Evenki in one analysis, Udihe) or two (Evenki in an alternative analysis, Negidal, Even, and Nanai) formally very similar suffixes that occur in possessive constructions. One of these has the form -ŋi in all the languages except for Nanai, where it is -ŋgi, the other is -ŋ in Even and Negidal (with insertion of an epenthetic schwa or [i], respectively) and -ŋgo/-ŋgu in Nanai. These suffixes are most probably cognate (cf. Sunik 1982: 65), with the protoform arguably being -ŋi – and, as shown by the Manchu data – the original function possibly being the marking of possessors in adnominal constructions in which the head noun is omitted. In the languages that have two formally distinct suffixes, the formal distinction corresponds to a syntactic distinction, since the suffix -ŋi occurs exclusively with possessors in constructions with elided head nouns, while the suffix -ŋ(V) occurs with possessees. However, while the formal split corresponds to the syntactic split between possessor- and possessee-marking, there is also non-correspondence between form and syntactic function, since like -ŋi the suffix -ŋ(V) occurs with modifiers of elided possessees. Thus, in some languages, two distinct suffixes are found in adnominal possessive constructions with an elided head noun. A diachronic analysis of these suffixes in the Tungusic languages might help elucidate how these distinct functions evolved but has to be left for a later stage.

In addition to the shared origin of these suffixes, they can also be analyzed as sharing an overarching function, namely the flagging of non-canonical possessive constructions. In Section 4.3, we argued that the suffix -ŋ(i) that occurs in possessive constructions in Negidal has erroneously been analyzed as a marker of alienable possession, following a Tungusological tradition. As our analysis shows, this suffix signals either the possession of a body part by a non-inherent possessor, or the possessive marking of non-possessible items, both semantically non-possessible nouns and formally non-possessible parts of speech, such as proper nouns, numerals, demonstratives, and adjectives. The suffix -ŋi that marks the possessor seems to have a similar flagging function, though with a different type of non-canonical construction. As pointed out in Section 3.1, it is quite common to omit the possessor from possessive constructions, since in adnominal constructions this is indexed by the personal possessive suffixes, and its identity can thus be retrieved fairly straightforwardly. The canonical Negidal adnominal possessive construction can thus be taken to consist of a head noun carrying possessive suffixes and an optionally expressed possessor; the inverse, namely an overt possessor and an elided head noun is unexpected and non-canonical. Therefore the function of -ŋi marking possessors of elided head nouns can be subsumed synchronically under the overall function that we propose for -ŋ(i), namely the marking of non-canonical possessive constructions. Judging from the available descriptions, this analysis could be extended to the cognate suffix(es) in Negidal’s sister languages.

Non-canonical possessive constructions are clearly a complex, multilayered and interlaced phenomenon, with semantic, formal, and syntactic factors playing a role (Figure 1). Regarding nouns, possession can be seen as non-canonical mainly for semantic reasons: the suffix is used either with body parts that are possessed by a non-inherent, i.e., non-canonical, possessor, or with nouns that refer to entities that cannot be possessed, such as humans or different natural entities. In the case of other parts of speech, the use of the non-canonical possession marker has purely formal reasons: all non-nouns are marked with -ŋ(i) in possessive constructions, irrespective of whether they are associatively possessed, refer to particularized individuals, or stand in for a directly possessed noun. Syntactic considerations are important with respect to constructions that are structurally non-canonical. Although possession of formally non-possessible modifiers such as adjectives and participles needs to be licensed with -ŋ(i), while possessors carry -ŋi, these seemingly disparate constructions share the omission of the head noun of an adnominal possessive construction. We summarize the different types of non-canonical possessive construction in Figure 1.

Figure 1:

Types of non-canonical possessive constructions in Negidal.

Our study thus demonstrates that descriptions of the (in)alienability distinction supposedly found in Tungusic languages should be treated with caution. While this distinction appears to emerge straightforwardly from the oft-cited examples involving ‘alienably possessed’ body parts, close examination of all the contexts of use of the ‘alienable possession’ marker shows that this interpretation is merely a secondary result of the particular type of non-canonical construction involved. While it would of course be rash to conclude from this single case that all instances of inalienable versus alienable possession identified worldwide are erroneous, our study does indicate a need for caution when interpreting such data (cf. the introduction to this special issue, and in particular Rose, this issue).