Predicative possession in the languages of the Circum-Baltic area

2.1 The semantics of possession

Defining possession is a decidedly difficult task. Baron and Herslund (2001: 1) state that possession is a “very elusive notion”; Seiler (2001: 27) admits that “possession is a domain loaded with paradoxes”. The main problem in the research on possession is how to deal with the immense variety of meanings conveyed by possessive expressions. How is it possible to lodge under one conceptual roof such different expressions like Harry has blue eyes and Harry has a problem, apart from their being coded by the very same verb to have? Can they both be said to be instances of possession? And preliminarily: by which semantic criteria can we define an expression as possessive? To answer these questions, several approaches have been proposed. Seiler (1983: 4) evokes extra-linguistic concepts, defining possession as the relationship between the EGO and his or her kin, body-parts and belongings. Lehmann (1998: 6) refers to the relationships evoked by Seiler as what forms the “personal sphere” of an individual: possession is, thus, the relation between EGO and the entities that fall under his/her “personal sphere”. In cognitive terms, Langacker (2009: 81–102) defines possession as a prototypically organized category that clusters around three “conceptual archetypes”: kinship, part-whole relationships and ownership. What all instances of possession have in common is their schematic structure: they are reference-point constructions, where one of the participants (the possessor) is conceptualized as the reference point, used to define the domain where the second participant (the possessed entity, or possessee) is found. This “target domain” very much resembles Lehmann’s concept of “personal sphere”.

Both Lehmann and Langacker evoke the notion of control as the key semantic element of possession. However, control has been understood differently in studies on possession. Stassen proposes a definition of control as “power” whereby the possessor can make use of the possessee whenever and for whatever s/he wants (Stassen 2009: 14–15; see also Seliverstova (2004 [1990]: 579). As such, this definition applies only to those cases where the possessee is an alienable, concrete entity, but it would not adequately describe the relationship a possessor has with inalienable or abstract possessees (Stassen 2009: 17). Langacker (2009: 83–84) therefore proposes a broader definition of control as the exclusive possibility of accessing the possessee, where access can be understood as experiential or social (only I can use my arm; only I (and my siblings) can refer to our father as our father). Such a broad conception of control would also include cases such as I have a problem: in this case, the possessor is the only one in whose sphere of experience that particular event is found and can be referred to as being “a problem”. Control qualifies thus as the necessary ingredient of every possessive relation (see Stolz et al. 2008: 18).

From the above, a working definition of possession emerges: a possessive relationship between two entities occurs when one entity (the possessor) functions as a reference point for identifying the domain where the second entity (the possessee) is located and when the first entity exerts control over the second. The characteristics of the two entities involved in the relation and the properties of the relation itself can, of course, change: hence the variety of different possessive relations that are found in languages. Consequently, in expressions like Harry has a car, Harry has blue eyes and Harry has a problem, the relations that the first participant enters into with the second participant are obviously different. Nevertheless, they can all be considered as possessive, because in all of them the first participant functions as reference-point to define the domain where the second participant is found, and in all of them the first participant exerts control over the second: only Harry can use his car and eyes, only Harry experiences that particular problem.

As I discussed in Section 1, Heine (1997: 33–35) labels these different possessive relations as “possessive notions”: the possessive notions form the domain of possession and any linguistic construction that expresses each one of them, or any combination thereof, can be considered possessive. Heine also suggests that possession may be considered as a prototypically organized domain: the prototypical, and more central, notion would be ownership, or legally acknowledged property. This is characterised by a [+human] possessor ^[1] (henceforth: PR), a [–human; –animate+alienable; –abstract] possessee (henceforth: PE), and a [+permanent] relation. In the following, I will also use “prototypical possession” and “ownership” as interchangeable terms. Other kinds of possessive notions are derived by changing these parameters (see Heine 1997: 33–41). Though retaining Heine’s approach to the definition of the domain of possession, I will employ a slightly different list of possessive notions. In particular, I will not make reference to the notion of inalienable possession used by Heine to term body-parts and kinship relations, as in I have blue eyes/two sisters, where the possessive relations cannot be terminated by the possessor (one will always have eyes and sisters; see Heine 1997: 34). The alienability/inalienability of the possessee has been acknowledged to play a crucial role in many languages, where they can trigger different morpho-syntactic treatments (see Chappell and McGregor 1996). However, in the languages analyzed in this study, the alienability/inalienability parameter does not play an important role in accounting for possession splits: rather, other factors – such as the distinction between modified and non-modified body-parts: I have eyes (and thus I can see) vs. I have green eyes – are more prominent. I therefore prefer to distinguish body-parts possession and social possession. In the following sections, I will consider the notions listed in Table 1 as forming the domain of possession:

A study of linguistic possession, however, cannot be limited to the notions that belong directly in the domain of possession. In the following sections, three other conceptual domains will be taken into consideration: attribution, experience and location. In particular, within the domain of attribution, I will take into account body-part relations where the body-part is modified by an attribute – I suggest the label “physical descriptions” for them – and expressions of age (see Dixon 2010: 263). Physical descriptions, as in John has blue eyes, differ from body-parts possession in that the part is modified by an attribute: thereby, the aim of the whole expression is to describe the appearance of John, not just to list the body-parts he has (as it would be in Human beings have two legs). ^[2] Within the domain of experience I will analyze expressions conveying the meanings of suffering from an illness, of experiencing physical sensations of “hunger” and “coldness”, and psychological states such as “sorrow”, “shame” and “fear”. Finally, within the domain of location, I will analyze a special class of part-whole relations, namely those where the part is spatially included in the whole.

My conception of possessive relations differs from the one presented by Heine, who assumes that all the notions in Table 2 belong to the very domain of possession (see Heine 1997: 36). In earlier studies (see Mazzitelli 2015) I largely held the same opinion; however, I now think that the situations represented in Table 2 may be more fruitfully analyzed in relating them to other domains, namely attribution, experience and location. The main reason for including them in this analysis is that in some languages they can indeed be coded by means of constructions used to express prototypical possession. In English, for instance, physical descriptions, the experience of suffering from a disease and part-whole relations are typically coded by means of the verb have (to have blue eyes, to have a fever, the house has three rooms). The Romance languages allow nouns denoting “hunger”, “coldness”, “fear”, “shame” and expressions of age in the role of possessees, too, as in French j’ai faim/froid/honte ‘I am hungry/afraid/ashamed’ (see Manzelli et al. 2002). Therefore, the possessive linguistic encoding of these notions cannot just be ignored, but must be accounted for and, insofar possible, explained (see Stolz et al. 2008: 510). As I will show further, though, in the languages of my sample these relations have (quite comprehensibly) the highest chance of being coded through non-possessive constructions.

2.2 The syntax of possession: Possessive constructions

The previous section has examined the semantic component of possessive expressions and of expressions that convey meanings pertaining to the neighboring domains of experience, attribution and location. The present section considers the formal requirements of the constructions that will be analyzed in the following sections. First of all, I will limit the analysis to instances of predicative internal possession. In the studies about possession, a distinction is usually made between attributive and predicative possession. In instances of attributive possession, the possessor and the possessee are coded in the same constituent: John’s car. In instances of predicative possession, instead, the relation is asserted by means of a predicate, and the possessor and possessee form two constituents: John has a car (see Heine 1997: 44). Within predicative possession, again, a distinction is made between “have” and “belong”-constructions (I have a book vs. The book belongs to me; see Heine 1997: 29–33; Mazzitelli 2015: 36–42). In this paper, I will not be concerned with “belong”-constructions, nor will I deal with instances of so-called “external possession” (see Konig and Haspelmath 1997; Haspelmath 1999).

In order to identify the domain of the enquiry, a syntactic definition of “possessive construction” is needed. Syntactically, Heine (1997: 45–47) restricts the range of possessive constructions to seven syntactic combinations, which he calls “source schemas”.

The event schemas are both the representation of the cognitive path and of the concrete diachronic linguistic development that has led constructions devoted to the expression of diverse conceptual domains (the source domains: Location, Action, Goal, and so on) to their grammaticalization as constructions expressing possession. An example of a possessive construction derived from the Location schema is found in Finnish, where the possessor is encoded with the adessive case (1a), used also to express locative relations (1b; see Huumo 1996).

Danish displays a construction derived from the Action schema (2), featuring the verb have ‘have’, which can be traced back to Old Norse hefja ‘take’ (Baldi and Cuzzolin 2005: 31):

Heine’s schemas are a valuable tool for cross-linguistic comparison, in that they identify a set of syntactic requirements that a construction must fulfill in order to be identified as possession. I will therefore define the domain of the present enquiry as restricted to constructions that fulfill the following semantic and syntactic requirements: (i) they express one of the possessive notions or one of the relations pertaining to one of the neighboring domains of possession (location, attribution and experience), as listed in Tables 1 and 2, and (ii) their syntactic structure corresponds to one of Heine’s schemas, as presented in Table 3 (except for the Equation schema, which only gives rise to “belong”-constructions, not examined here, as in The car is Peter’s; Heine 1997: 65).

In the following, however, I will not refer to constructions using Heine’s labels, because they do not sufficiently highlight some crucial differences. The typical way of expressing possession in Finnish, for instance, is by encoding possessors by an adessive case marker, which is otherwise used to express location at a particular place. As will be shown in Section 4.3, Finnish codes inanimate possessors in part-whole relations with the inessive case, whose main function is to express location within a place. Since both constructions – the adessive and the inessive – are instances of the Location schema, they have to be distinguished by using different names. I will therefore use the following labels: constructions featuring a transitive predicate equivalent to English have are labeled as “have”-constructions; ^[3] in all other cases, the name of the construction is derived from the morphosyntactic encoding of the possessor. In the case of Finnish I will thus distinguish between an adessive construction and an inessive construction. Finally, I will refer to expressions like English I am thirsty as “adjectival constructions”.

Some final remarks are due. First, I will disregard verbs of possession such as English own, possess, Finnish omistaa ‘own, possess’, or Danish beside ‘own, possess’. Usually, these verbs specifically express ownership, even if they may sometimes be used to convey non-ownership meanings as well. Of course, a complete investigation of the domain of possession should also include these verbs. An interesting question could be, for example, how far outside the domain of ownership their semantic scope can extend. For lack of space, though, I decided to focus on constructions that show a more “vague” possessive semantics, such as English have (Heine 1997: 35). Secondly, I will not take into account the possible use of possessive constructions to encode conceptual domains such as modality or temporality. Indeed, possessive constructions can be grammaticalized into means of expression of modality or tense, as in the case of English constructions have to and have + past participle (Past Perfect). In Estonian, the construction mul.adon.prs.3sg ‘at me is=I have’ is used to express possession (see Section 4.1 below), and it has also grammaticalized to convey a perfect meaning in combination with the participial form: mul.adon.prs.3sgõpitud.pst.ptcp ‘learnt’, lit. ‘at me is learnt’=‘I have learnt’ (Metslang 2009: 55). Again, a study on possession should also consider these semantic extensions of possessive constructions and their possible grammaticalization. In this paper, though, for simplicity’s sake, I will leave aside the use of possessive constructions as expressions of TAM categories, as well as other uses such as the English to have a+deverbal nouns, as in to have a walk (see Wierzbicka 1982).

4.1 Prototypical possession: Ownership

As far as the coding of prototypical possession, i. e., ownership, is concerned, four groups, corresponding to as many typological types, can be distinguished (see also Koptjevskaja-Tamm and Wälchli 2001: 677). Following the terminology proposed in Heine (1997), these groups instantiate the Action, Location and Goal schemas:

1. Action schema. The first group includes German (West Germanic), Swedish and Danish (North Germanic), Polish (Slavic) and Lithuanian (Baltic). In these languages, a “have”-construction, derived from the Action schema, is used. The possessor is coded as the subject of the predicate “have” and the possessee as its direct object; in languages that exhibit case morphology, the possessor receives nominative marking and the possessee accusative marking (3):

2. Location schema. The second group comprises Finnish, Estonian, Olonec and North Karelian, Votic (Finnic) and Russian (East Slavic). In these languages, ownership is expressed by means of constructions derived from the Location schema. In Finnic, possessors are coded with the adessive case, as in Estonian (4). In Russian, possessors are encoded by a prepositional phrase whose head is the preposition u ‘at’+NP.gen (5):

In all these languages, the constructions employed to express possession maintain their locative function: Estonian ta.nomon.prs.3sgmaal.ad ‘s/he is in the country’ (Erelt 2009: 99); Russian učitel’.nom.sgstoit.prs.3sgu doski.gen.sg ‘the teacher stands by the blackboard’ (Cienki 1995: 75).

3. Goal schema. The third group only includes Latvian, where a dative construction is used.

4. Action+Location schemas. Finally, the fourth group comprises Belarusian. Here, both the u+NP.gen construction and a “have”-construction featuring the verb mec’ ‘have’ are used to express ownership, as in (7a) and (7b):

As this overview shows, the Germanic and the Uralic languages in my sample do not show any signs of intrafamilial diversity as all languages of the same family belong to the same typological type. The Slavic and Baltic groups, on the other hand, present a more complex picture, where closely related languages belong to different types. This situation has often been explained with reference to language contact (see Section 8). East Slavic (in particular Russian) would have grammaticalized the u +NP.gen construction to express possession under influence of the Finnic languages. Similarly, Finnic influence might have played a crucial role in preventing the development of a “have” verb in Latvian (Vykypěl 2001: 221; Wiemer 2004a: 104). ^[7] It can also be posited that the emergence of “have” (or at least the strengthening of its use) in Lithuanian may be due to, or might have been facilitated by, contact with speakers of the neighboring languages, Polish and Belarusian. ^[8]

4.2 Non-prototypical possession: Body-parts, social and abstract possession

4.3 Inanimate possession: Part-whole relations and inanimate abstract possession

The notions of part-whole relations, or inanimate inalienable possession, as defined in Heine (1997: 35), and inanimate abstract possession strongly deviate from the possessive prototype denoting ownership, where the possessor is a human being. As I stated above, I would include in the domain of possession only part-whole relations such as The table has four legs, where the part is constitutive of the whole (and not included within it). Such relations are expressed with the same strategy used for ownership in most languages in my sample, as illustrated by the following German example (14):

Finnic languages show two different patterns. Estonian shows the same construction that serves the expression of prototypical possession, with adessive encoding of the possessor (see (14); Metslang 2009: 65). Finnish and Karelian, however, encode the part-element with the inessive case, whose basic function is to express “location inside something” (Sulkala and Karjalainen 1992: 242; Zajkov 1999: 42; about its uses in part-whole relations, see Sulkala and Karjalainen 1992: 227; Stolz et al. 2008: 455): ^[11]

Votic seems to allow both patterns. The two informants interviewed for this study gave discordant answers about the acceptability of the adessive case for possessors of parts. Our first informant judged the adessive encoding in the example (18a) below as grammatical, while the second informant did not accept it and proposed an inessive encoding (18b):

Russian displays two competing constructions for the expression of part-whole relations: it allows both the prototypical possessive u+NP.gen construction (19a) as well as the verb imet’ ‘have’ (19b), which would usually not appear in expressions of ownership:

The notion of “inanimate abstract possession” includes those cases in which an inanimate possessor is associated with an abstract possessee, as in English Love has many names. In such cases, the part-whole interpretation is ruled out, and instead we are confronted with a metaphorical use: ^[12] the inanimate possessor is given human privileges, i. e., the possibility of possessing something. All languages in my sample can express this notion by means of their ownership constructions, as in the following Latvian example from Le petit prince:

As in the case of the part-whole relations analyzed above, Russian employs both u+NP.gen and imet’ ‘have’ to express the notion of inanimate abstract possession (Weiss and Rakhilina 2002: 178). In Lithuanian, this use is restricted (Šukys 1998: 60). In Polish, Belarusian and Russian, the dative case is also occasionally used in this context, though mainly in fixed expressions such as “NP.dat was/is/will not be an end” (see Mrázek 1990: 46; Janda and Clancy 2002: 97):

As for Votic, again our two informants gave different judgments about the grammaticality of the adessive encoding of inanimate possessors. The one informant considered it acceptable (22a), while the other suggested replacing it with an inessive encoding (22b):

Contrary to what happens in expressions of part-whole possession, Finnish and Karelian can use their ownership construction – thus with the adessive encoding of the possessor – to express inanimate abstract possession, as the following North Karelian example shows (23):

5.1 Attribution

5.2 Experience

5.3 Location

As I explained in Section 2.1, I consider expressions of part-whole relations where the part is spatially included in the possessor, e. g., English The house has three rooms, as part of the location domain. With regard to such expressions, a difference can be observed in the languages in my sample.

All the languages that employ a “have”-construction to express possession can also employ it to express part-whole relations involving spatial inclusion, as in the following Swedish example:

In all these languages, however, locative expressions can also be used, as in the English phrase In the house there are three rooms.

Instead, the languages of the area that employ adessive constructions for expressing possession, prefer locative constructions with inessive semantics to code part-whole relations with spatial inclusion. Russian and Belarusian can either use their “have”-verbs (aligning here with the other “have”-languages of the area) or a construction where the whole is encoded by a prepositional phrase headed by the prepositions v ‘in’ (Russian) and u ‘in’ (Belarusian)+NP.loc: ^[23] Russian kvartira.nom.sgimeet.prs.3sgvannuju.acc.sg ‘the apartment has a bathroom’ or v kvartire.loc.sgest’.prsvannaja.nom.sg ‘in the apartment there is a bathroom’ (Clancy 2010: 149). The u+NP.gen construction is not allowed (or at least disliked; see also Mazzitelli 2015: 114 about Belarusian) in this context.

In Finnish and Karelian, the adessive encoding is blocked; instead, an inessive construction is used: Finnish asunnossa.in/*asunnolla.adon.prs.3sgkylpyhuone.nom.sg ‘in the apartment there is a bathroom’ (Eeva Sippola, p.c.). Conversely, Estonian allows adessive encoding, as in (45), though, according to my corpus data, the inessive is more common:

6.1 Coding possession

In Table 4, the strategies used to code the notions belonging to the domain of possession are shown. Splitting constructions are marked in bold.

As the table shows, possession splits are largely absent in the languages in my sample. In Finnish, Karelian and Votic, inanimate entities are represented as locations in expressions of part-whole relations, receiving an inessive encoding (see Stolz et al. 2008: 456) – this is also the only grammatical split (as intended in Stolz et al. 2008: 43), that is the only one which is obligatory: adessive encoding of the possessor would be ungrammatical. Nevertheless, if the relation denotes abstract possession, inanimate entities are represented as adessive-encoded possessors, too. It seems, thus, that the relevant criterion for splitting is not (only) the [–human] or [–animate] feature of the possessor; rather, it is the part-whole character of the relation that triggers a locative understanding.

In Russian, Belarusian and Lithuanian, the dative construction furnishes the speakers with an alternative strategy in the field of abstract and inanimate abstract possession. However, as seen above, this construction is not always available, and it is not in free variation with the ownership construction, given the many lexical and semantic constraints it is subjected to. Dative constructions can still fulfill a possessive function in these languages (especially in Lithuanian), albeit confined to non-prototypical possessive notions.

A further split is found in Russian, where the verb imet’ is attested alongside the u+NP.gen construction for the expression of abstract relations and part-whole relations; unlike the Finnic split mentioned above, the Russian use of imet’ is never obligatory (unless required by syntax; see Weiss and Rakhilina 2002) and it is also allowed in cases of inanimate abstract possession.

Based on what has been said in the previous sections, the situation in the Circum-Baltic area can generally be described as “ownership is enough”. Apart from the few cases described above, the ownership constructions (“have”, u+NP.gen or dative constructions) are employed to express all other notions in the domain of possession. The languages of the Circum-Baltic area, thus, have adopted the European majority solution as shown in Stolz et al. (2008), that is, to code all possessive notions in terms of prototypical possession.

6.2 Coding attribution, experience and location

As was to be expected, an analysis of the coding strategies for the other three domains investigated in the previous sections (Table 5) presents a more complicated picture. Ownership constructions are marked in bold.

As shown in Table 5, ownership representation appears less frequently, except in case of physical descriptions and diseases. These two notions are commonly expressed in terms of possession in most languages of the area.

Conversely, expressions of age (attribution), sensations of “hunger” and “coldness” and emotions like “fear” and “shame” are rarely, if ever, given a possessive coding. The reluctance to code them in terms of possession is self-explanatory: they do not pertain to the domain of possession. Instead, it is their possessive coding – whenever it appears – that needs an explication.

In the first place, it needs to be clarified that possessive coding does not imply a possessive meaning: the notions of “age”, “hunger” or “fear” remain semantically non-possessive. In fact, even in languages that admit a possessive coding for them, they are often expressed through alternative strategies. Let us take the example of age: even in the Romance languages, where this notion is commonly expressed in terms of prototypical possession (see Stolz and Stolz 2009), there are also adjectival forms to encode it, such as the Italian adjectives dodicenne ‘twelve-years-old’ and ventenne ‘twenty-years-old’. Similarly, in English one can say He has blue eyes but also He is blue-eyed; it is possible to say This house has three rooms but also In this house there are three rooms.

I believe that the answer to the question, why attributive, experiential and locative expressions can receive a linguistic possessive coding, is to be found in the conceptual structure of the domains of attribution, experience, location. Like possession, they can all be qualified as non-agentive. The most salient participant of an attributive, experiential, locative and possessive relation qualifies, according to Van Valin and LaPolla. (1997: 139–147), for the semantic role of unprototypical actor (possessor, experiencer, location) or undergoer (theme of an adjectival predication as in John is thirty years old; see Stolz et al. 2008: 20). Though formally coded as the nominative subject of a two-place verb such as “have” or the oblique argument of “be”, the logical subject is, semantically, never an agent (a prototypical actor):

I argue that the fundamental conceptual operation that serves the construction of predication in all four domains is the same: the referent is ascribed a given property (Table 6). The basic structure can be represented as “[referent] [relator] [property]”. Properties may be material or abstract entities (“a car”, “three rooms”, “free time”; parts of a whole are included in this class, too), physical/mental attributes (“blue-eyed”), bodily sensations or feelings (“hunger”, “shame”). Whenever the ascribed property is (or is realized as) an entity, it can be represented linguistically as a possessee: the slot [referent] is realized as a possessor (either as a subject of “have” or as an oblique, if the language has no “have”-verb), the [relator] is “have” or “be” and the [property] is the possessee.

Of course, there are also other possibilities, as the Lithuanian example in (51) shows:

Here, the possessed entity (“blond hair”) is not realized as a possessee (according to one of Heine’s schemas); instead, it is part of a nominal predicate. In this case, even if the property is indeed realized as an entity (i. e., a noun phrase), its encoding and syntactic status make it part of a nominal predication – and not of a possessive construction.

6.3 Diseases as an intermediate class

At this point, a further question needs to be answered: what makes some experiential notions – such as the experience of suffering from a disease – more suitable to be expressed in terms of possession than others – such as the bodily sensations of “being cold”, “warm”, “hungry”?

Fedriani et al. (2013: 411), commenting on the situation in the Romance languages and in other languages of the Mediterranean area, claim that physical feelings are more apt than mental feelings (“sorrow”, “anger”, etc.) to be encoded through possessive constructions. The former are more likely to be conceptualized as “objects” one can possess, because they are much less abstract and more tangible, in that they involve easily observable physical reactions (sweating, freezing, changes in the body’s temperature, etc.).

Their observations also apply to the Circum-Baltic area, where the mental feelings of “being sorry, afraid, ashamed” are usually not encoded in terms of possession. “Hunger” and “cold”, on the other hand, are usually represented by means of a possessive construction, although, just like diseases, they imply physical reactions and they can also be “cured”, for example by eating something or putting on warm clothes. In my opinion, this anomaly can be explained by the fact that nouns denoting diseases (“flu”, “cancer”, “fever”) express, all in all, abstract concepts. “Having the flu” or suffering from “cancer” is a more complex experience than feeling a basic sensation like “hunger” or “coldness”. “Flu”, “cancer” and “AIDS” are specific illnesses, with which a lay person is usually not familiar. It is true that “having the flu” and “having a problem” are two different experiences; nonetheless, I would dare to say that diseases can be very easily grouped together with classes of abstract nouns. Further proof of these two notions being coded similarly is provided by Finnish. In Finnish, the possessor/experiencer of a disease can also be encoded with the inessive case. The inessive construction, however, can also be used with abstract properties:

This strategy is found in Russian, too: bylo.pst.n.sgv nej.locčto-to.nomprivikatel’noe.nom.sg ‘there was in her something attractive’ (Arutjunova and Širjaev 1983: 169). The fact that Finnish can use the same construction for abstract properties as well as for diseases contributes another argument to the hypothesis that “diseases” form a special class of concepts which are more easily classified as abstract concepts than as experiential notions (such as “hunger” or “thirst”). Interestingly, the genitive encoding of the experiencer, used for sensations of hunger and coldness as well as for the emotion of “being sorry”, is blocked here: *minun.genon.prs.3sgkuume.nom.sg ‘mine is fever=I have fever’ (Eeva Sippola, p.c.). This also speaks in favor of an interpretation of diseases as pertaining to the domain of abstract possession (and not only of experience).

A final interesting remark about the syntax of physical states in Finnish must be added. In Finnish, canonical possessees (as well as canonical subjects of existential predications) usually take the partitive case when negated (see (53a); Sulkala and Karjalainen 1992: 113) but “hunger” and “thirst” do not (see (53b); Uotila Arcelli 1975: 179):

Nouns denoting diseases, conversely, take the partitive case under negation, just like canonical possessees; yet the experiencer can receive inessive encoding as well (while this possibility is blocked in the case of typical possessors). Finnish, thus, distinguishes systematically between possessive and experiential expressions (Table 7).

In each one of the experiential notions taken into account here, Finnish has at least one morpho-syntactic feature that distinguishes these expressions from typical possessive sentences (see Croft 2001: 115). ^[24]

8.1 Finnic – East Slavic

Contacts between Finnic and Slavic-speaking populations can be traced to the Late Common Slavic period (fifth century AD), just before the Slavic migration (Wiemer 2004a: 103; Le Feuvre and Petit 2011: 258–262). The striking convergence between Finnic and East Slavic (particularly in Russian) with regard to the grammaticalization of adessive constructions into means of expressing possession has already attracted the attention of many scholars during the last decades (see, e. g., Veenker 1967; Bátori 1980: 150; Dingley 1995). A contact-induced origin of the “loss” of the verb “have” in Russian and the spreading of the u+NP.gen construction has often been proposed (see, e. g., Veenker 1967: 118; Le Feuvre and Petit 2011). However, evidence from Old Church Slavonic, as well as from Old Czech, Old Serbian and Middle Bulgarian (see Mirčev 1971; Vasilev 1973; Grković-Major 2011: 51) “demonstrate[s] that u+genitive was already used to encode predicative possession throughout the dialects of Late Proto-Slavic, though it was a peripheral construction that was restricted in its usage” (McAnallen 2011: 156). It follows that Finnic influence should be traced back to the Late Common Slavic period, and that it is not restricted to the Russian dialects (Vasilev 1973; Le Feuvre and Petit 2011: 258–262). In these Russian dialects (and through them also in Belarusian and Ukrainian), the prolonged contact with Finnic-speaking populations led to a strengthening of the u+NP.gen construction. In all other Slavic languages, which lost contact with Finnic populations after the migrations in the fifth–sixth centuries, the possessive use of u+NPgen has declined.

8.2 Russian, Votic and Karelian

Contact between the speakers of Votic and Karelian and speakers of Russian can be traced back to the very beginning of the Finnic-Slavic contact (fifth century) and has continued ever since (see Suhonen 1992; Pugh 2001; Laakso 2003). The impact of Russian on these languages has been considerable, both on the lexical and the morphosyntactic level (Suhonen 1992; Pugh 2001; Kuznetsova et al. 2015: 154). In the field of predicative possession, a clear example of a contact-induced phenomenon is the use of the allative case to mark possessors of expressions of age, a replica of the Russian use of a dative construction (see Section 5.1.2 above). Ojajärvi (1950: 97–100) points to an expanding use of the adessive/allative case instead of the genitive in Karelian, due to Russian influence: Mi šiuli.ad.allon nimi? ‘What is your (lit. ‘at you’) name?’. In Finnish, a genitive would be found: Mikä sinun.gennimesi on? ‘What is your name?’ (Ojajärvi 1950: 100).

8.3 Estonian and Latvian

Contact between speakers of the Baltic dialects, which will form the basis for Latvian, and speakers of the Finnic languages – in particular Livonian, spoken in the territory of present-day Latvian – are attested from the twelfth century BC (Balode and Holvoet 2001: 9). Convergences in several aspects of phonology and morphosyntax between Estonian and Latvian have already been underlined in Stolz (1991) and Metuzāle-Kangere and Boiko (2001). As already mentioned in Section 4.1, the fact that Latvian has never developed a “have”-construction (in contrast to Lithuanian) might also be ascribed to contact with Finnic-speaking populations. The dative case in Latvian and the adessive case in Estonian are, in many aspects, functionally equivalent, even though some differences can be observed (see Metuzāle-Kangere and Boiko 2001).

As Table 8 shows, Estonian and Latvian display a significant degree of consistency with reference to their sister languages Finnish and Lithuanian. In the treatment of age and bodily sensations, Baltic and Finnic behave differently from each other.

Estonian, though, differs from Finnish in coding psychic states and part-whole relations. Instead, it agrees with Latvian, which, in its turn, agrees with Lithuanian. It can therefore be assumed that it was Estonian which adapted to the Baltic model. Latvian and Estonian display an even greater level of convergence, which singles them out from their respective genetic groups. As mentioned above, negated possessees are coded with a genitive in Baltic with a partitive in Finnic. In Lithuanian, on the other hand, experiential nouns such as gėda ‘shame’ remain in the nominative case even when negated (since they have lost nominal properties, see Section 5.2.4). Conversely, in Latvian and Estonian, they receive genitive and partitive encoding, respectively, maintaining their nominal status: ^[25]

It can be questioned whether the semantic extension of the adessive construction to encode experiencers and beneficiaries (Huumo 1996: 88–90) could also result from language contact. It has been argued that in Karelian and Votic the same phenomenon might be due to Russian influence (Ojajärvi 1950; Suhonen 1992). Huumo (1996: 90) explains the extensive use of the adessive in Estonian by referring to the loss of the case ending -n, which was distinctive for the genitive, and to the “general tendency to replace the functions of the old -n case with the adessive”. It could be a functional replication of the Low and High German extensive use of “have” – but both Low and High German use dative constructions for encoding beneficiaries. The Estonian adessive could therefore be regarded as a case of syncretism, where the adessive case is seen as functionally equivalent to “have” and to the dative case. However, a functional convergence with the Latvian dative case seems to be more plausible as the Estonian adessive construction and the Latvian adessive construction also display a formal similarity. One could think of an areal convergence, possibly also involving Livonian, a practically extinct Finnic language spoken in the territory of present-day Latvia (see Balode and Holvoet 2001), which pushes towards a unified encoding of prototypical possessors, inanimate possessors (see Metslang 2009: 65) and beneficiaries/maleficiaries.

8.4 German and Polish

As already mentioned in Sections 5.3.1 and 5.3.3, German includes some expressions featuring the “have”-construction, which are absent in Danish and Swedish, namely Hunger haben ‘to have hunger=to be hungry’, and Angst haben ‘to have fear=to be afraid’. The presence of such expressions in German is probably due to the influence of French, where the use of “have” to code experiential notions is widely diffused. The extension of “have” as a strategy for expressing experiential notions can also be observed in Polish. This is the only Slavic language in my sample where the “have”-construction is firmly established as the major possessive strategy. Its use is even attested in the domains of attribution (expressions of age) and experience (diseases), alongside a diminished use of the dative case in comparison to its Eastern neighbors and genetic relatives Belarusian and Russian. The expansion of the sphere of use of Polish “have” could be the result of a convergence with German (and, through German, with the languages of the Charlemagne-Sprachbund;van der Auwera 1998), which shares a millenary history of contact with Polish (Hentschel 2009). True, the uses of Polish mieć ‘have’ and German haben ‘have’ often do not coincide (unlike German and Czech; see Clancy 2010: 241–246), so a direct replication is to be excluded. Still, the tendency to make increasing use “have”- instead of “be”-might be seen as areally determined.

8.5 Belarusian and Lithuanian

We know that speakers of the northwestern Belarusian dialects and of the southern Lithuanian dialects have been in contact for centuries, which has resulted in some phenomena of convergence (Wiemer 2004b). Yet, the impact of such convergences on the standard languages has been minimal: they have mostly affected the dialects.

As shown above, Belarusian is, typologically, a mixed language, for it uses both a construction derived from the Location schema as well as a construction derived from the Action schema to express ownership. However, the use of the constructions differs in several respects: “have” is more specialized for the expression of the notions that fall into the domain of possession (as well as for part-whole relations), and is less suitable for notions belonging to the domain of experience. In Lithuanian, “have” does not have any competitors for expressing possession; at the same time, it is not used to express experiential and attributive notions. There is, therefore, a remarkable degree of similarity in the way these two languages expand their “have”-verbs beyond the domain of possession proper. However, I would not ascribe this similarity to direct contact between the two languages but would suggest considering Belarusian and Lithuanian as being in the middle of a convergence area: they share with their northwestern neighbor Polish the capacity of using a “have”-verb for prototypical possession, while sharing with their eastern neighbors (Russian, Latvian) the tendency to encode experiencers and other similar semantic roles as obliques or, at least, not as subjects of “have”. Contact with Polish (which has exerted a considerable influence on both languages from the sixteenth century onwards, especially on Belarusian,) and, via Polish, with German might have helped to strengthen the role of the “have”-construction, while contact with Russian and Latvian (and, indirectly, with Finnic-speaking populations) might have helped to preserve the non-“have”-constructions (u+NP.gen, topicalized genitive, dative constructions). I am inclined to consider the similar treatment of the verb “have”, which both languages avoid in expressions of physical descriptions, diseases and experiential notions, as due to macro-contact (thus, as a manifestation of an areal convergence), and not to direct contact between the two languages only.