1 Introduction

Possessive constructions are among the most debated linguistic phenomena, and their defining characteristics have been widely discussed from various theoretical perspectives (Nikiforidou 1991; Langacker 1995, 2003, 2009; Heine 1997a, 1997b; Baron et al. 2001; Coene & D'hulst 2002; Stassen 2009; McGregor 2009; Dixon 2010; Aikhenvald & Dixon 2013). However, due to the inherent semantic plasticity and grammatical diversity of these constructions, no single theoretical explanation has yet been proposed to account for their cross-linguistic characteristics in a uniform way. Various meanings ranging from prototypical senses (such as ownership) to peripheral ones (like association) are anchored around the possessive construction. Nevertheless, possessive linguistic devices are so semantically polysemic and grammatically applicable to numerous structures that they have been described as vacuous entities (Bendix 1966; Bach 1967) or explained as the result of subjectification (Langacker 2009, 85) and grammaticalization (Heine 1997a). Apart from its semantic and syntactic heterogeneity, the possessive construction is subsumed under various typological classifications. All of its constituting elements, including the possessor (henceforth Pr), the possessee (henceforth Pe), and the possessive relationship (henceforth Prel) are subject to typological variations. Accordingly, two main lines of investigation seem necessary for reaching a unified account for the divergence and multiplicity of the possessive construction: First, a typological study to determine the range of linguistic forms the possessive construction may take in different languages, and second, a cognitive appraisal to go beyond the linguistic expressions with the purpose of exploring the underlying conceptual schemata the possessive construction may evoke.

To reach the primary goals of our study, we will elaborate on possessive constructions in Kurdish (the central variety, also referred to as Sorani) and Hungarian from a cognitive perspective. Kurdish belongs to the Western branch of Iranian languages, categorized as a branch of the Indo-Iranian and Indo-European language families (MacKenzie 1961; Oranskij 1999). Hungarian is a member of the Finno-Ugric group of the Uralic language family. Since Uralic languages are not generally considered to be associated with Indo-European languages, Kurdish and Hungarian are to be thought of as genetically unrelated languages. Although possessive markers express various semantic and syntactic functions in Kurdish, ¹ they have not been adequately addressed, particularly from a cognitive perspective. A substantial body of the available literature on possession-related phenomena in Iranian languages has been devoted to the Ezafe linker as a functional morpheme that appears in Noun Phrases to overtly codify an intrinsic relationship between the nominal head and other linguistic elements in its domain such as complements, possessors, and modifiers (Samiian 1994; Ghomeshi 1997; Karimi 2007, 2013; Samvelian 2005; Kahnemuyipour 2014). Having said this, it should also be noted that Ezafe cannot be considered as a prototypical possessive marker because it has various syntactic functions going beyond the prototypical scope of possessive codification. Besides, most of these studies relegate the function of Ezafe in possessive constructions to an item that is only required in formal computations. Accordingly, as Lotfi (2014) points out, a significant body of the related literature on possession in Iranian languages is suffering from a formal reductionism, which leads to the systematic neglect of the cognitive aspects of the Ezafe and possession. To avoid such reductionism, we will address both the typological and cognitive aspects of possessive constructions with equal degrees of elaboration. Moreover, the linguistic similarities between various kinds of Kurdish and Hungarian possessives have encouraged the authors to consider Hungarian data and examine if they could be cognitively explained. Incidentally, the same reductionist approach is generally observed in the available literature on Hungarian possessive constructions, as the formal treatment of different types of possessive constructions has been the focus of an extensive body of related studies (Szabolcsi 1981, 1989, 1992). We will demonstrate that extending the cognitive argumentations that were suggest for Kurdish possessives in the first place to Hungarian data would lead to convincing results. This type of research can also enrich the current literature on Hungarian possessive constructions.

The remainder of this paper is structured as follows: Section 2 analyzes nominal possession in Kurdish and Hungarian. In this section, the heterogonous functions of nominal possessive constructions in these languages are explained according to the reference-point model, presented by Langacker (1995, 2009). Predicative possession in Kurdish and Hungarian is addressed in Section 3. It is argued that predicative possessive constructions in these languages represent a non-standard type of topic-possessive construction in which the Pe serves as the sentential trajector, and the Pr acts as the clausal topic. Furthermore, we will show that Kurdish employs another predicative possessive construction, which is considered as an instance of the be-possessive pattern, consisting of an intransitive copular verb and hi as the possessive-marker morpheme. The copular verb is a schematic structure that requires other components to elaborate it. The schematic position elaborated by other elements is called e-site in cognitive grammar (Langacker 2008, 198). Having this terminology at our disposal, we will argue that the Pe, acting as the target of the relationship evoked by hi, elaborates an e-site of the copular verb in this construction. There is also another specific type of predicative possession in Kurdish which involves the Ezafe linker and a copular verb (haya). In this construction, the trajector of the Ezafe and that of the intransitive verb are identified and unified in the final stage of the conceptual integration of the clause. Along with the main argumentation line, which is mainly concerned with Kurdish data, similar instances of predicative possession in Hungarian are described and accounted for. In the final part of the third section a brief discussion will be presented on the nature of the so-called ergative construction in Kurdish and the conceptual relation they have with predicative possessive constructions. It is claimed that there is an inheritance link between the possessive construction and the so-called ergative construction in Kurdish grammar. This inheritance link serves as a basis for a cognitive treatment through which the possessor of predicative possessive constructions is reanalyzed as the agent of so-called ergative clauses. Finally, Section 4 presents the conclusions and implications of the paper.

2 Nominal possession in Kurdish and Hungarian

Languages employ typologically different mechanisms to encode the possessive relationship within Noun Phrases. The simplest mechanism is juxtaposition of the Pr or the Pe in a particular order without the intervention of any additional morphological element (see Croft 2003, 32–36; Dixon 2010, 267‒271; Aikhenvald 2013, 6‒8). As for Kurdish, the most fundamental linguistic strategy for indicating the possessive relationship between the Pr and the Pe within NPs is the concatenation process through which a cliticized Pr gets attached to a nominal Pe. As Croft (2003, 32) points out, possessive affixation may be categorized as an instance of juxtaposition, involving a gradual transformation of juxtaposed elements into bound morphemes. This pattern is attested in the history of Iranian languages: Old Iranian languages, as historical ancestors of New Iranian languages, have employed the juxtaposition of the pronominal Pr and the nominal Pe. In Old Iranian languages the Pr and Pe also agreed in case, gender, and number (Abolqhasemi 2006, 101–102). Along with other pervasive changes that occurred during Old-to-New transition of Iranian languages, the agreement between the Pr and the Pe disappeared, and the possessive pronouns became more grammaticalized and took the form of possessive clitics ² hosted by the Pe (Moiin 1984, 8).

The possessive determiner in modern Kurdish, like in other new Iranian languages such as modern Persian, is highly schematic and can express a wide range of associated meanings. The following examples ³ illustrate the semantic scope the Kurdish possessive determiner can possibly cover:

What distinguishes examples (1), (3), (4), (5) from (2), (6), (7), (8) is the distribution pattern of the (in)definite marker. The realization of the (in)definite suffix is associated with the alienability and inalienability of nominal possessive constructions in Kurdish. When a possessive construction is interpreted as inalienable, possessive clitics simply attach to the nominal heads without the addition of any other linguistic material. In other words, the possessive clitic and the nominal head can form a minimal unit, as in (1), (3), (4), and (5). In alienable possessives, on the other hand, the (in)definite marker prevents the possessive clitic from attaching to the nominal head directly, cf. (2), (6), (7), and (8); the alienable possessive construction, thus, includes more dependent morphemes (noun + (in)definite marker + possessive marker) compared to the inalienable version of the possessive construction (noun + possessive marker). That is to say, the alienable nominal head and the possessive clitic cannot form a minimal unit. The inalienable group includes kinship relationship (1), part-whole (3), and mental and physical attributes (as in (4) and (5)). However, the alienable group, for its part, involves ownership (2), and any kind of association (see (6), (7), and (8)).

The difference in the distribution pattern of (in)definite marker in alienable and inalienable possessive constructions can be straightforwardly accounted for if we utilize iconicity as a motivation and the role it can play in the ultimate realization of possessive constructions (see Aikhenvald 2013, 8). More specifically, it can be argued that the Pr and the Pe in the inalienable construction are too conceptually intertwined to be structurally separated from each other via any grammatical element. In other words, the intimate conceptual relationship is reflected in the internal structure of Noun Phrases and gives rise to a minimal linguistic structure. By contrast, the non-intimate alienable relationships include two participants which lack a firm cognitive link and are thus separated by the (in)definite marker. Through the intervention of this marker, the Pe is crystalized and separated from the Pr. Furthermore, the absence of the (in)definite marker in the inalienable possession may lie in the identifiability and inferability of inalienable Pes. Inasmuch as inalienable Pes are always inferable, using the (in)definite marker with them will result in redundancy. On the other hand, the referents of the Pe in the alienable construction may not be identifiable or inferable through the linguistic context; consequently, they should always be overtly marked as definite or indefinite. It should be noted that the aforementioned iconic relationship between (in)definiteness and (in)alienability should be viewed as a general tendency rather than a categorical rule. Although paternal terms are among those relations that may never be mediated by the (in)definite marker, there is a number of kinship terms whose expression as a Pe in a nominal possessive construction may occasionally involve the (in)definite suffix. For instance, the kinship terms that occupy the same rank in the family tree as the parents do (like uncles and aunts as brothers and sisters to parents) or those that outrank parental relations (like grandparents or grandmothers) can host the possessive clitic without the (in)definite marker if and only if they are singular and known (see (9) and (10)). However, in certain cases where these kinship terms are employed so as to refer to some particular individuals (to highlight one among some), they have to be used with the indefinite marker (as in (11) and (12)). Other kinship terms including son, daughter, grandson, granddaughter, niece, nephew, and cousin are always accompanied by an (in)definite marker (see (13) and (14)).

Having discussed examples (9)–(14), we can identify a continuum including identifiable inalienable possession (like parental kinship, whole-part), non-inferable alienable possession (ownership and association), and some cases in between (some affinal kinship terms).

A fundamental question that is due at the current stage of our analysis is how the semantic diversity of the Kurdish possessive determiner can be uniformly explained from a cognitive perspective. Several explanations inspired by the idea of metaphorical extension and mere association have been put forth to account for the possessor determiner's semantic diversity, yet Langacker (1995, 57–8; 2000, 175–176) finds them all unsatisfactory. Two reasons may be offered to justify this theoretical stance: first, several instances of the possessive construction (ŝirja-ka=t: diving-def-your; ŝoota-ka=y: shooting-def-her) cannot be regarded as a metaphorical or derived version of ownership. Second, the theory of association on its own cannot explain the asymmetries which are attested in possessive constructions. To explain such a diversity of possessive meanings from a unified perspective, Langacker (1995, 58; 2009, 81–85) presents the reference-point model, which is proposed to be immanent in all prototypical cases of possessive constructions. This model is a type of scanning of a discrete mental path along which a salient element is employed as an anchor so as to arrive at another element (Langacker 2008, 83). Differently put, the reference-point model refers to a mental ability that invokes one entity as the reference-point to establish a connection with another element, described as the target. Taking this model as our point of departure, we investigate prototypical instances of nominal possessive constructions in Kurdish below.

In archetypical possessive constructions, the Pr actively controls (socially, physically, or experientially) the Pe as a member of those entities (dominion) that it possesses. The following examples illustrate this point:

The Pr (=t: your) in (15), as the owner of the Pe (kəteb: book), exerts physical and social control over the Pe. Similarly, the Pr (=i: his or her) in (16), as the physiological whole, is capable of controlling the Pe (dast: hand), as its part. Finally, the parental Pr in (17) (=əm) may control the Pe in terms of sociocultural norms. Generally speaking, the Pr can manipulate and handle the Pe as a member of its own domain. The following schema illustrates this relation. The symbol D refers to the dominion, including those possessed and controlled by the Pr. Additionally, Pe1, Pe2 and Pe3 represent the possessed elements in Schema 1.

View Full Size

• Download Figure
• Download figure as PowerPoint slide

Besides, Pr's active and objective control can provide the conceptualizer (C) with a subjective mental path (mental access) from the Pr as the reference-point (R) to the selected possessed entity Pe, as the target (T) (Langacker 2009, 85). In other words, the conceptualizer may invoke the Pr as a reference-point to establish mental access to the Pe. Prs (=t: your; =əm: my) as the salient entities have a great potential to serve as the reference-points for the Pes as targets (kəteb: book; koŕ: son). Therefore, in addition to the fact that the Pr (=t or =i) can manipulate the Pe (kəteb or dast) objectively, it may also be used as a reference-point to provide subjective mental access to the target (Pe). As a result, it can be argued that the possessive determiner in Kurdish can evoke a reference-point relationship with two e-sites elaborated by the Pr as the reference-point and the Pe as the target, as shown in Schema 2. The e-sites refer to the schematic aspects of the relationship that Pr and Pe elaborate.

View Full Size

• Download Figure
• Download figure as PowerPoint slide

Consequently, it seems that the possessive determiner invokes two different kinds of mental relationships between the Pr and the Pe: the first one pertains to the objective control of the Pr over the Pe, and the second one refers to the subjective mental path from the Pr to the Pe. In Fig. 1, which illustrates the cognitive representation of ownership, kinship and whole-part relations in (15), (16), and (17), the possessive determiner and the nominal head are represented as R and T respectively. The thick arrow line shows the Pr's objective control over the Pe, and the thin arrow indicates mental access to the target via the reference-point.

View Full Size

Prototypical possessive construction

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

Apart from the prototypical cases discussed in (15)–(17), there are some possessive constructions that do not represent the Pr's active control over the Pe. For example, the nominal heads in the following examples, raʔis (boss) in (18) and hāwsā (neighbor) in (19), are not objectively controlled by their corresponding possessive determiners, =t (your) and =m (my). Comparing (15) with (18) can illuminate this point. In (15), the possessive determiner refers to a person who owns the Pe and can control it objectively. In contrast, in example (18) the possessive marker denotes a person who has a boss, yet cannot exert any objective control over him/her.

The fact that Prs are not the true controllers of Pes in non-canonical nominal possessive constructions can be linked to the passive nature of Prs in these constructions (see Langacker 2009, 85). Unlike prototypical instances of nominal possession in which Prs are active elements, in peripheral cases Prs are passive elements that just hold a possessive relationship with their corresponding Pes. Despite the absence of objective control, the subjectively-construed relationship between the Pr and the Pe is still preserved in examples (18) and (19). In other words, in this construction (Fig. 2), the objective relationship between the Pr and the Pe has faded away, yet the mental path between them has remained in such a way that the Pr can still be used as a reference-point to establish a mental connection with the Pe. As mentioned above, unlike example (15), in which the Pr (=t: your) has an objective and subjective (mental access) relationship with the Pe (kəteb: book), the Pr (=t: your) in example (18) has only a subjective relationship to the Pe (raʔisa: boss). In fact, the former element is used to gain mental access to the latter one without having any objective control over it. To put it differently, the conceptualizer may use the Pr (=t: your) as a reference-point to provide mental access to the Pe (raʔis: boss) as the target (Schema 3).

View Full Size

Schematic possessive construction

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

View Full Size

• Download Figure
• Download figure as PowerPoint slide

Therefore, it can be argued that the objective control of the Pr over the Pe has been lost in (18) and (19), but the cognitive path linking the Pr to the Pe remains unaffected. In Langacker's (2009, 85) nomenclature, this situation is described as subjectification; a process through which the gradual disappearance of an objective relationship results in a subjectively-construed one. Due to the fact that the reference-point model is highly schematic hinging on mental processing rather than any particular conceptual content, we can flexibly apply it to all non-prototypical instances of nominal possession. Fig. 2 depicts the cognitive schematic model underlying the non-prototypical instances of the possessive determiner. The lack of a thick arrow in Fig. 2 indicates that Pr has no objective control over the Pe.

To sum up, we can maintain that in Kurdish nominal possessives the possessive clitic evokes a schematic reference-point model in which the conceptualizer invokes one entity (the Pr) as the reference-point to establish a mental contact with another one (the Pe) as the target through any kind of association, including possession.

Nominal possessives in Hungarian are formed through the same concatenation strategy, based on which an attenuated possessive element is suffixed to a nominal head. Hungarian possessive suffixes come in two numbers and three persons as follows: -m functions as first person singular, -d as second person singular, -a or -e as third person singular, -nk as first person plural, -tök, -tok or -tek as second person plural, and -uk or -ük as third person plural. The following examples demonstrate this strategy in Hungarian:

In the case of plural Pes in Hungarian, the possessive suffix follows the plural suffix just as it was the case in Kurdish (cf. (2) and (7)):

The data show that both alienable and inalienable nouns in Hungarian can freely take possessive suffixes, unlike Kurdish in which alienable possession is basically formed through concatenation of the possessive affix to those nouns which already bear the (in)definite suffix (examples (2), (6), (7), and (8)). In other words, it could be concluded that iconicity is not the dominant motivation in determining the ultimate form of nominal possessives in Hungarian.

The fact that Kurdish and Hungarian nominal possessive constructions are structurally and conceptually alike can be viewed as an empirical motivation to account for the underlying characteristics of these constructions in both languages through similar lines of cognitive argumentation. In what follows we apply to Hungarian examples the cognitive analysis which was previously put forward for Kurdish nominal possessives. Taking Langacker's reference-point model as the main cognitive framework, it can be argued that in Hungarian, the Pr actively controls the Pe as a logical subpart of those entities which the Pr possesses. As Langacker argues (2009), this objective control, being exerted by the Pr on the Pe, may provide the conceptualizer (C) with the required mental path from the Pr as the reference-point (R) to the selected Pe as the target (T). Therefore, not only can the Pr (manifested as possessive suffixes in the Hungarian examples) control the Pe, it can also be employed as a reference-point for accessing the target Pe mentally.

Despite the attested similarities in the overall structure of nominal possession in Kurdish and Hungarian, these languages use different strategies for encoding possessive constructions, i.e., constructions in which both the Pr and Pe are realized as full noun phrases or pronominal expressions. Kurdish employs another form of concatenation mechanism in which the nominal Pr and Pe are linked together via a simple linker which is generally known as Ezafe in the relevant literature on Iranian Linguistics (Qharib et al. 1971, 39‒45; Moiin 1984; Anvari & Ahmadi-Givi 2011, 134–138). Ezafe is basically manifested as a close front vowel (specifically, -i) in Kurdish yet may turn into a glide (-y) when is used for introducing a topicalized element (see (36) and (37)). Besides ownership, the Ezafe linker in Kurdish covers a number of other semantic domains. The following examples demonstrate Kurdish possessive constructions and the range of meanings Ezafe may cover:

As evident from the examples above, Ezafe can express meanings such as possession, part-whole, participant-event, object-material, nominal collocation, and pure association. It may be safe to posit that Ezafe evokes a highly schematic relationship that can convey different kinds of semantic association. Yet, it should be noted that this schematic interconnection whose order is not reversible is not semantically vacuous. Any modification to the linear order of the constituents in this construction may result in ungrammaticality or meaning shift:

Following Lotfi (2014, 60), Veisi Hasar & Dehghan (in press) analyze this asymmetric order in the Kurdish Ezafe construction as a kind of intrinsic relationship evoked by Ezafe linker. This proposal can be theoretically solidified in light of Langacker (2008, 84), in which the conceptual irreversibility of entities is interpreted as an indication of intrinsic relationship between them. The intrinsic relationship concerns the cognitive proximity and experiential dependency of the parts involved. Langacker (1995, 68) describes the intrinsic relationship as another manifestation of the reference-point model in which the entities are asymmetrically related to one another. In other words, a primary participant (landmark) functions as the reference-point providing the conceptualizer with mental access to another one (trajector) as the target. Consequently, it may be argued that Ezafe as a non-grounding element profiles an asymmetric and intrinsic relationship between one salient entity and a related phenomenon. The reference-point in this construction is described as a landmark, and the target is considered as a trajector which is mentally accessed by a cognitively prominent element, namely the landmark. It is worth mentioning that both the landmark and the trajector are focal participants in a relational predicate; however, the latter is only conceived and assessed through the former since the landmark functions as a ground for the trajector. The following diagrams illustrate the conceptual schemata evoked by Ezafe linker (Ez) in (26) and (28) (Figs 3 and 4).

View Full Size

The conceptual schema evoked by Ezafe (example 26)

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The conceptual schema evoked by Ezafe (example 28)

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

In comparison to the nominal possessive construction in (15)–(17) in which the possessive marker (at the final position of the phrase) acts as the reference-point and the nominal head (at the initial position) functions as the target (kəteba-ka=t: your book), in these examples the pronominal or the nominal modifier (at the final position) plays the role of landmark which is used to establish a mental connection with the trajector (nominal head) appearing at the initial position of the phrase. Therefore, the landmark as the prominent and presupposed cognitive element appears at the final position of the Ezafe construction, and the trajector occupies the initial position.

Because Ezafe always introduces a presupposed prominent element, it can also introduce topicalized elements. In other words, the other cognitive function of Ezafe is related to the idea of topicality. The Ezafe marks a relationship between the topic (as the landmark) and the rest of the proposition (as the trajector). The following examples illustrate this point clearly:

In (36) and (37), the agent and patient have been topicalized via Ezafe. As Langacker (2008, 513) points out, the topic may act like a reference-point through which a mental access to a proposition is made possible for the conceptualizer. In harmony with this cognitive idea, one can argue that Ezafe invokes a kind of intrinsic relationship (as another kind of reference-point model) to hold between the topic and the proposition. Accordingly, the topic acts as a reference-point for understanding the proposition as the target. This mental subjective path is made possible by the mediation of Ezafe. The topicalized element functions as the landmark providing mental access to the information expressed in the rest of the sentence as the trajector. This information can only be accessed via the topicalized element. This intrinsic relationship is marked by the Ezafe linker (Schema 4).

View Full Size

• Download Figure
• Download figure as PowerPoint slide

A similar type of concatenation of nominal elements via a particular linker is attested in Hungarian. The possessive construction in Hungarian, however, is composed of other grammatical elements for which no counterpart can be found in the corresponding Kurdish examples:

As be inferred from (38) and (39), the possessive linker in Hungarian is placed after the nominal Pe (the head), unlike Kurdish in which the Ezafe element appears between the Pr and the Pe; moreover, a particular suffix expressing person-number agreement with the Pr attaches to the right periphery of the Pe. This very feature of Hungarian possessive constructions has led Szabolcsi (1981, 1989, 1992) to suggest an essentially configurational sentence-like account for Hungarian possessive constructions based on which the possessive linker parallels the head of sentences (the head of IPs) in that it can trigger overt person-number agreement with the nominative Pr. ⁶ In addition to this construction, Hungarian also has another type of possessive construction in which the Pr appears in dative case, instead of nominative, and precedes the article a(z) (Szabolcsi 1981; Dékány 2015). The following example, taken from Szabolcsi (1981, 263), is an illustration of the second possessive construction in Hungarian and includes the same Pr and Pe as example (39) above:

Just as Ezafe in Kurdish, the -ja/je suffix in Hungarian can be explained as a cognitive device capable of evoking an intrinsic relationship between two related concepts that participate in an asymmetric cognitive relationship. As a result, -ja/je is considered as a non-grounding element which may provide mental access from a prominent conceptual entity (Peter) to another concept (arm) that is identifiable only in terms of its immediate scope (the owner, or the whole). A comprehensive analysis of the cognitive characteristics of these constructions, which are not based on grounding elements, falls beyond the scope of the present paper and can be proposed as a potential line of inquiry for further research in this domain.

3 Predicative possession in Kurdish and Hungarian

Unlike nominal possession, the predicative one demonstrates the possessive relationship in the clause. Various kinds of typological classifications of predicative possession have been presented in terms of several parameters, including the nature of the verb (Aikhenvald 2013, 27; Dixon 2010, 299), the functional encoding of the Pr and Pe (see Stassen 2009), and the grammaticalized cognitive schemata (see Heine 1997a, 1997b). As far as Kurdish is concerned, the most frequent verb used for verbalizing the possessive relationship is the existential verb haya (bu in the past tense). It is essentially an intransitive verb involving one focal participant. ⁷ Regarding functional encoding, the Pe elaborates the focal e-site of haya, and also the Pr can be expressed as a possessive clitic indexed on the Pe ((41)–(43)). This existential verb turns into niya (‘not-exist’) in the negative construction, as exemplified in (43). The following examples can clarify this point:

What seems to be primarily at issue here is that the Pe is represented as the sentential subject and the Pr as a possessive clitic directly attaches to the Pe. It should be noted that the Pe is an obligatory element that cannot be deleted. Besides possessive clitics, the Pr may also be expressed as the sentential topic ((44)–(45)):

The presence of the topical Pr is not grammatically obligatory: it may be overtly expressed or remain covert on the basis of discursive considerations. When the topical Pr is covertly expressed, it can be recovered via the possessive clitics attaching the Pe (as in (41) and (42)). In other words, the recovery of deleted elements in Kurdish, as a pro-drop language, is made possible through the use of possessive clitics.

The ultimate realization of the Pr in the topic position in the predicative possessive construction can be accounted for in terms of the topic accessibility scale. Lambrecht (1994, 165) maintains that the topicality of a linguistic element is evaluated in terms of the accessibility of its referent in the discourse. When an expression has an active referent in its discourse model and thus occupies the highest possible position on the accessibility scale, it is regarded as a prototypical topic whose deletion won't pose any serious problem for the addressee. On the other hand, the expressions whose referents are brand new constitute less probable topics. These expressions are placed on the lower extreme of the accessibility scale and become practically impossible to identify once they get deleted. In addition to these extreme cases, there are some borderline expressions (topics) that have identifiable but inactive referents. The very fact that Kurdish is a pro-drop language and the implications brought about by the topic accessibility scale can ultimately constitute a persuasive explanation for the behavior of the topical Pr in the predicative possessive construction. Any anaphoric element (filling the topic position in the predicative possessive construction) which refers back to an active topic in the discourse can be easily deleted. Yet those full NPs that can be characterized as non-active topics are preferably present. For instance, consider the pronoun ʔaw ‘he’ in (46): as an anaphoric element, it refers back to a continuous active topic (monk) and thus can be deleted due to its identifiability. The Noun Phrase qaʃa-yak ‘a monk’ in (47), on the other hand, is an instance of a topical Pr that cannot be omitted. It is so because qaʃa ‘monk’ is a linguistic element with a brand-new referent that has not been previously introduced into the discourse.

It can be concluded that the possessive construction in Kurdish includes an existential verb which is elaborated by the possessee as its subject. Besides being indexed on the Pe, the possessor may also appear as the sentential topic according to discursive factors. The following examples illustrate this construction.

This structure in Kurdish is in complete harmony with the topic-possessive construction as proposed by Heine (1997a, 47), and illustrated in the following examples from the Lango and Cahuilla languages: as for X, Y exists: X owns Y.

A structural comparison of the predicative possession in Lango and Cahuilla ((50)–(51)), with its corresponding counterpart in Kurdish ((48)–(49)), clearly indicates that all of these three languages employ similar patterns for the ultimate realization of the possessor. As it is the case in Kurdish, the possessor in Lango and Cahuilla can appear either as an attenuated element being attached to the possessee and/or as a full NP at the topic position. Moreover, the concept of possession in the languages in question is denoted by an existential verb.

It is worth mentioning that the indexing of the possessor on the possessee is by no means a rare mechanism in the languages of the world. Examples (52) and (53) are further illustrations of this pattern in Toba Batak and Banggai. Stassen (2009, 70–71) classifies possessive constructions of this sort, which are characterized by possessor-indexing on the possessee, as non-standard types of topic possession. He maintains that possessor-indexing is common in languages that have a topic possessive construction.

The pattern of the possessor-indexing on the possessee in Kurdish predicative possessives is grammatically motivated. The existential verb of the possessive clauses, haya, is in its invariant default form and hence fails to agree in person and number with any of its arguments. Taking the role of the verbal agreement affix, the cliticized possessor provides the required grounding information concerning the speech event and its participants. In order to see how grounding information is conveyed through verbal agreement affixes in typical accusative constructions (in present tense) in Kurdish, consider the following examples. (54) and ((55)–(56)) exemplify intransitive and transitive constructions respectively.

Unlike in the examples above, the verb in the topic-possession construction lacks any kind of agreement. As a result, the possessive clitic provides the grounding information. It can be concluded that the main characteristics of topic-possession in Kurdish are as follows:

1. A. This construction includes an existential verb.
2. B. The Pe is the grammatical subject of the verb.
3. C. The Pr is attached to the Pe as a possessive clitic.
4. D. The Pr may be expressed as the sentential topic.
5. E. The possessive clitic replaces the verbal affixes in conveying the required grounding information.

The conceptual structure underlying the topic-possession construction in Kurdish cannot be fully explained unless the conceptual integration of the possessive marker and the existential verb are taken into consideration. First, the cognitive function of possessor-indexing (on Pe) will be taken into account. It can be argued that the possessive clitic (attaching to the Pe) evokes a kind of reference-point relationship elaborated by the Pr as the reference-point and the Pe as the target entity. The intransitive existential verb, on the other hand, profiles a relation including one e-site that is then elaborated by the Pe in the final conceptual integration. As a result, the target of the relationship evoked by the possessive clitic elaborates the e-site of the intransitive verb and accordingly turns into the trajector of the final construction. Because of the conceptual integration, the verb's existence domain is equated with the possession domain of the possessive clitic. The following example reveals this conceptual process:

The possessive clitic, =m, in (57) evokes a reference-point relationship between the Pr and the Pe. The Pr, =m, serves as the reference-point for the Pe (target), duĉarxa. Moreover, the intransitive existential verb, haya, profiles a relationship with one e-site. As a result of conceptual integration, the Pe as the target entity elaborates the e-site of the intransitive verb and thus becomes the sentential subject. The domain of existence is in fact identified with that of possession. Fig. 5 shows the conceptual integration of (57):

View Full Size

The conceptual integration of topic-possessive construction in Kurdish

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

As discussed earlier ((44)–(49)), not only can the Pr attach the Pe as a person-maker clitic, it may be also expressed as the sentential topic. In the latter case, the topic itself functions as a reference-point and takes the whole proposition as the target (see Langacker 2008, 513). In other words, the conceptualizer employs the topic as a reference-point to gain mental access to the whole proposition as the target. The following example shows this point:

In this example, the whole integrated conceptual structure of (duĉarxa=y haya in Fig. 5) is mentally accessed via the reference-point of ʔali. The following diagram illustrates this conceptual structure. For the sake of space, the diagram has been simplified and many details have been removed (Fig. 6).

View Full Size

The conceptual integration of topic-possessive construction in Kurdish (with an overt topic)

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

There is a second pattern for the predicative expression of the possessive relationship in Kurdish. In this clausal pattern, the predicate is an intransitive copular verb (a: be) whose e-site is elaborated by the Pe, i.e., the clausal subject. The possessive relationship is expressed by the morpheme hi. This morpheme does not have concrete content and is schematically used only for linking the Pr to the Pe. This construction is exclusively specified for the concept of ownership ((59)–(60)). In other words, as (61) and (62) illustrate, extending the semantic domain of this construction to other conceptual areas such as kinship, part-whole and association sounds unnatural.

It might be argued that the free morpheme hi in ((59)–(60)) invokes a reference-point relationship including two e-sites, one being elaborated by the Pr and the other by Pe. The Pr as a reference-point provides the conceptualizer with mental access to the Pe as the target entity. Moreover, the intransitive copular verb agrees in number and person with the Pe. It could be maintained that the e-site of the verb is elaborated by the Pe that then becomes the trajector of the final clause. In the ultimate stage of conceptual integration, the copular verb's domain is equated with the possessive domain of hi. This construction may be categorized as be-possessive, as presented by Langacker (2009, 98). The following diagram illustrates the conceptual structure of example (59) (Fig. 7).

View Full Size

The conceptual integration of be-possessive construction in Kurdish

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

Going through the underlying conceptual structure of both of the above-discussed predicative constructions, one can readily point to a shared mechanism that is attested in both patterns. In both constructions the Pr, which is the controller or a sort of agent, is demoted. Then the Pe, which is in fact the controlled entity, is promoted and becomes the subject of the clause. Both constructions include an intransitive predicate profiling a relationship elaborated by the passive entity, Pe, as the sentential trajector.

The third pattern for the clausal realization of possession in Kurdish resembles the second pattern in that the third one, too, is a construction whose predicate is an intransitive copular verb and could be categorized as another manifestation of be-possessives. Yet what distinguishes the current pattern from the previous one is the fact that the third pattern lacks any type of possessive marker. However, the construction in question employs the Ezafe linker to evoke the possessive-intrinsic relationship between the Pr and the Pe. This construction is used for expressing the meanings of possession ((63)–(64)), kinship (65), part-whole (66), and association (67) in the clausal format. The sentence consists of a copular verb (a: be), agreeing in number with its subject, and an Ezafe, invoking an intrinsic relationship between two entities. A relevant point that merits mentioning is that the trajector of Ezafe and that of the clause are co-equivalent. Due to the trajector equivalence, the Ezafe domain and the predicate domain are integrated in the final phase of conceptual integration.

In (67), the sentence consists of an intransitive copular verb agreeing in number with Ali, as the sentential subject. Besides, Ezafe evokes an intrinsic relationship (between hamkār and hasan) elaborated by hamkār as the sentential subject's counterpart and the trajector and another participant, Hasan, as the landmark. The trajector of the Ezafe is accordingly identified with the sentential trajector in the final construction. Moreover, the domain of be is equated with the domain of Ezafe in the final construction. The following diagram illustrates this conceptual integration (Fig. 8).

View Full Size

The conceptual integration of be-possessive construction in Kurdish (with an Ezafe linker)

Citation: Acta Linguistica Academica 68, 3; 10.1556/2062.2021.00157

• Download Figure
• Download figure as PowerPoint slide

The equivalence between the trajector of the verb and that of Ezafe cognitively motivates the conceptual integration of the Ezafe and the copular verb. This cognitive proximity motivates the identification of Ezafe domain and that of the copular verb. The Ezafe phrase turns into a predicative one when its trajector is apt to be identified with that of the copular verb.

Kurdish and Hungarian employ similar strategies for formulating the possession relationship in the predicative form. Since there is no specific verb with the meaning “have” in Hungarian, this language adopts an existential construction in which the intended possessive relationship between the Pe and Pr is encoded through the verb “be”. What is common in both predicative and nominal types of possessive constructions in Hungarian is that in both of them the attenuated form of the Pr, as a possessive suffix, is attached to the nominal Pe. However, in the predicative case, the mere presence of a form of “to be” brings about the intended clausal interpretation. In the following examples the third person singular and plural forms of “to be” in Hungarian possessive clauses are van and van-nak and are employed for singular and plural Pes, respectively.

As evident from ((68)–(71)), the van element in Hungarian resembles haya in Kurdish in that both of these units are originally derived from “to be”. Moreover, both of them appear as the predicate of possessive clauses and therefore provide the required basis for the predicative encoding of the possessive relationship in Hungarian and Kurdish. This structural similarity aside, it seems that Kurdish haya is more grammaticalized than its Hungarian counterpart, van, in that the existential element in Kurdish shows an invariant form for all kinds of Pes while its Hungarian counterpart can exhibit a full paradigm with six different inflected forms.

The Pr can also appear as an overt Noun Phrase in Hungarian possessive clauses. As it was the case in Kurdish ((46)–(49)), the overt realization of topicalized Pr in Hungarian is closely related to its level of accessibility: less accessible Prs whose identification requires more processing are more likely to be realized as overt NPs in the topic position of possessive clauses. However, Prs whose referents are active enough to be identified with a minimum amount of processing effort do not usually appear as overt NPs at the beginning of possessive clauses. ⁹ An interesting point about Hungarian is that in addition to topicalized Prs, focused ones can also be expressed overtly in possessive clauses. In (72) and (73) the overt Prs occupying the initial position of the possessive clauses are topicalized and focused elements respectively. It is also worth mentioning that the overt Pr in Hungarian possessive clauses takes a dative suffix which literally amounts to “to” in English.

As can be inferred from the above examples, the constituents of both clauses are essentially the same. Yet, what determines the focused reading of the overt Pr in (73) is the different order which the Pe and the verbal element appear in. Moreover, the attenuated Pr that is attached to the Pe as a possessive suffix provides the required grounding information about the participants. ¹⁰ This very fact was also elaborated on in the Kurdish data.

Apart from their formal similarities, predicative possessives in both Kurdish and Hungarian can receive the same types of cognitive explanations. In other words, the same line of cognitive argumentation which was put forward to account for different aspects of Kurdish predicative possessives could be justifiably extended to Hungarian predicative possessives. It could be argued that the possessive affixes in Hungarian predicative possessives, which are attached to the Pe, bring about a reference-point relationship elaborated by both the Pr, as the reference-point and the Pe, as the target entity. Given the intransitive nature of the existential verbal element realized as van, van-nak in Hungarian predicative possessives, there exists only one e-site, which is subsequently occupied by the Pe in the final conceptual integration. To put it differently, the only e-site in the conceptual profile of Hungarian predictive possessives is elaborated by the target, which is basically evoked by the possessive suffixes. The same entity then turns into the trajector of the whole construction. This line of cognitive argumentation can be vividly depicted in Fig. 5.

The last important point that is going to be addressed pertains to the complicated relationship between predicative possession and the ergative alignment in Kurdish. In fact, the topic-possessive construction in Kurdish ((41)–(45)) (as for X, Y exists) has a very intricate relationship with a contentious construction described variously as ergative or non-accusative ¹¹ by different scholars (Haig 2008; Karimi 2010, 2012, 2014; Dabirmoghaddam 2013). Kurdish exhibits a stem-based split alignment system: ¹² transitive constructions whose main verbs are derived from the present-tense stem show accusative alignment, yet in those transitive constructions in which the main verbs are derived from the past-tense stem, the so-called ergative alignment is attested. Inasmuch as Central and Southern Kurdish lack structural case morphemes, alignment shift in these varieties is mainly manifested through changes in the patterns of agreement. In accusative constructions (present tense), the verb agrees in person and number with the subject argument via inflectional suffixes which are attached to the main verb (see (75) and (77)). However, in so-called ergative constructions (past tense) the transitive verb does not agree with any of its arguments and therefore lacks inflectional endings. The patient appears before the verb, and the agent emerges as a possessive clitic hosted by the patient ((74), (76)). In these cases, agent-referring clitics provide the grounding information required for the interpretation of the clause and accordingly compensate for the absence of verbal inflections. In addition, the nominal or pronominal agents, ʔali in (76) and mǝn in (74), may appear as the sentential topic at the initial position of the sentence. However, it is not an obligatory element and may be omitted according to the principle of topic accessibility ((78)–(80)). (74) and (76) illustrate this construction in the past tense along with their accusative counterparts in the present tense ((75), (77)):

As maintained earlier, the overt realization of the Pr in topic-possession constructions is controlled by the principle of topic accessibility ((44)–(49)). The same pattern determines the ultimate form of the so-called ergative construction, i.e., the covert or overt manifestation of the agent in these constructions can be accounted for in terms of the aforementioned principle. Those agents that are pronominal and refer back to an active topic ((78)–(79)) can be expressed covertly. On the contrary, other agents such as kābrā-yek in (80) are preferably expressed overtly since they have not been mentioned earlier and their covert realization might result in opaqueness.

It can be concluded that although the agent can be expressed either overtly or covertly at the initial topic position of the so-called ergative clause, it is obligatorily indexed through the possessive clitics, which are hosted by the patient. In addition, the verb doesn't agree with any of its arguments. The ultimate form is like example (81):

Although it would be challenging to describe this construction as ergative, it can be readily subsumed under the category of theme-oriented constructions, as described by Langacker (1991, 386; 2008, 374). This categorization is motivated by the fact that the passive semantic role (like māʃin which is the theme in example (81)) elaborates the trajector position, i.e., the subject of the construction. The most important argumentation in favor of considering the theme as the trajector of this construction is based on the order of the clitics hosted by the verb. When both internal and external arguments in a past-tense transitive sentence are expressed covertly, they are realized as possessive clitics and are hosted by the verb in a particular order: the one referring to the patient attaches firstly to the verb, while the other one representing the agent follows the former, as in (83). This very fact shows that the verb takes the patient as its trajector; therefore, this construction as a whole can be considered as a theme-oriented one.

Moreover, we can use a special type of present-tense construction to empirically solidify our claim concerning the theme-oriented nature of so-called ergative constructions in Kurdish. These constructions include psychological predicates such as love, hate, ache, and itch with a non-strong agent (experiencer). Although the so-called ergative alignment in Kurdish is only attested in constructions whose main verbs are derived from past-tense stems, the aforementioned predicates have to be expressed via ergative-like constructions in both past and present tenses. ((84)–(87)) including these predicates show low degrees of transitivity, and are characterized by having non-strong agents. It can then be argued that these predicates tend to be verbalized within a theme-oriented system, which is manifested by the so-called ergative construction. The ergative pattern is a canonical syntactic framework for the clausal expression of events that contain verbs with low levels of transitivity and non-strong agents. In this specific usage, the so-called ergative pattern is not tense-restricted and can be attested in non-past tenses as well. However, when a predicate is accompanied by a strong agent, and the sentence is characterized as highly transitive, the sentence follows the default accusative pattern in the present tense (88).

Therefore, it can be argued that the so-called ergative constructions in Kurdish employ the theme-oriented pattern in which the patient occupies the trajector position. In addition to past-tense transitive constructions, there is a particular class of verbs that are characterized by low degrees of transitivity and non-strong agents employing the same theme-oriented pattern in all tenses. As one can observe in the (85) and (87), the themes, Karim and Hasan, are considered to be the trajectors of the sentences. The agents, however, appear as possessive clitics, =mān and =t, respectively, and are attached to the patient. They may also occupy the topic rather than the subject position, and appear in the initial position of the sentence as a full NP, (see (84) and (86)), if they meet the interpretive requirements of the topic accessibility principle.

It can be argued that the controversial nature of this construction may be due to both ignoring its relationship with the topic-possessive construction and disregarding the cognitive function of the possessive clitic in favor of formal criteria in the previous studies mentioned above. A thorough analysis of the examples that have been presented so far reveals structural similarities between the so-called ergative structure and the topic-possessive construction:

1. A. The verb is uninflected in both ergative and topic possessive constructions.
2. B. Taking the role of the verbal inflections, the possessive clitics provide the required grounding information.
3. C. The agent and the possessor emerge as the topic of the sentence.
4. D. The patient and the possessee act as the sentential trajector.

In both constructions, a passive participant, theme and Pe, takes the position of the clausal subject, and the agentive one, agent and Pr, emerges as the topic or a possessive clitic. The structural similarity between the so-called ergative and topic-possessive may be symptomatically accounted for through the inheritance link between them (see Goldberg 1995). It can be asserted that the eventual structure of the ergative construction has been construed in terms of the intrinsic possessive relationship. This argumentation is in harmony with the ideas of Benveniste (1966) and Trask (1979) who link the emergence of the ergative construction in Iranian languages to the reanalysis of the possessor as the agent.

A detailed analysis of the possessive clitics’ cognitive functions in ergative constructions can also help us to illustrate the aforementioned relationship more accurately. The possessive clitic is to accomplish the task of grounding instead of the absent verbal inflections by virtue of grammaticalization (see Creissels 2005). Therefore, the agent and patient have been linked to each other through a possessive relationship. The agent, then, emerges metaphorically as the possessor since it is the entity that controls the patient. In a similar vein, the patient is metaphorically realized as the possessee, functioning as the controlled entity. The possessive clitic invokes an intrinsic relationship between the agent (Pr) and the patient (Pe). The asymmetric intrinsic relationship between the Pr and the Pe can metaphorically illustrate the asymmetric relationship that holds between the agent and the patient. In the ultimate conceptual realization, the Pe (patient) and the Pr (agent) of the possessive relationship correspond to the trajector and landmark of the verb respectively. Therefore, the possessive domain is identified with the eventual domain of the verb in the final construction. Consequently, the cause-effect structure of the event is construed as the controller-controlled relationship of the possession. Now, we can explain the cognitive motivation that relates the topic-possessive to the ergative construction through an inheritance link. Both of the constructions in question are characterized by containing a theme (patient and possessee) elaborating the sentential trajector, and this semantic homogeneity motivates the inheritance link between them. In other words, because both of these constructions are theme-oriented, they can be related to each other conceptually. Finally, it should be noted that this cognitive explanation cannot account for all aspects of ergative construction in Kurdish, yet it can be considered as cogent evidence for the theory proposed by Benveniste (1966) and Trask (1979).

4 Conclusions

This paper offered a comparative analysis of possessive constructions in Kurdish and Hungarian. It indicated that the Kurdish language applies the morphological mechanism of concatenation to express the possessive relationship between the Pr and the Pe in NPs. The same process can be seen in Hungarian, in which the possessive marker (Pr) attaches to the Pe. Unlike Kurdish, in which the alienability or inalienability of the Pe affects the ultimate form of the nominal possessive constructions, Hungarian forms both alienable and inalienable possessives alike. Due to the underlying semantic heterogeneity of nominal possessive constructions in Kurdish and Hungarian, we proposed the same type of cognitive analysis to account for their characteristics in both languages. It is argued that nominal possessives evoke the reference-point model, in which the Pr functions as the reference-point and the Pe as the target element. The former provides a mental path to the latter thanks to the conceptual proximity existing between them.

However, in cases where the Pr and the Pe are to be realized as full noun phrases or as pronominal expressions, Kurdish and Hungarian use different mechanisms. Kurdish applies the Ezafe linker to establish a connection between nominal or pronominal Prs and Pes, while in Hungarian possessives a possessive linker attaches to the Pe and a suffix showing person and number agreement appears at the right periphery of the Pe. Despite their structural differences, the Kurdish and Hungarian linkers are explained as operators evoking an intrinsic asymmetric relationship between the Pr as the landmark and the Pe as the trajector.

The paper also revealed that Kurdish has two different kinds of predicative possessions. The basic one, categorized as the topic-possessive construction involves an intransitive existential verb with one focal participant which is elaborated by the Pe as the sentential trajector. In addition, the Pr, which is cliticized and gets attached to the Pe, may surface as an overt topic in the initial position of the clause provided it is evaluated as a less probable topic. As for the possessive clitic, it evokes a reference-point model in which the Pr, as the reference-point provides mental access to the Pe, as the target for the conceptualizer. The Pe, then, elaborates the focal entity of the existential verb. In the final phase of the conceptual integration, the domain of the existential verb is identified with the domain of the possessive suffix. The same cognitive explanation is suggested for the topic possessive construction in Hungarian which uses a specific type of existential verb. The verb van in Hungarian is used to encode the connection between the Pr and the Pe. The Pr element as an attenuated form is attached to the nominal Pe element. As it was the case in Kurdish, the presence of possessor in Hungarian is determined by the principle of topic accessibility. It is maintained that the possessive suffix in this construction may also evoke a reference-point model in which the Pr as the reference-point provides a mental path to the Pe as the target entity. In the final step of the conceptual integration, the only e-site of the intransitive verb of the clause is elaborated by the Pe. Consequently, the target element becomes the trajector of the possessive clause.

The second pattern for the predicative expression of possession in Kurdish (be-possessive) includes a free morpheme hi evoking a reference-point relationship between the Pr as the reference-point and the Pe as the target. This construction involves an intransitive copular verb with one e-site. We suggested that the Pe elaborates the e-site of the copular verb and accordingly turns into the sentential trajector. As the third pattern of the clausal realization of possession in Kurdish, we introduced a special kind of predicative construction in which the intransitive copular verb agrees in number with its sentential subject. This construction also includes an Ezafe linker evoking an intrinsic relationship between two entities. The integration of the copular domain with the Ezafe domain determines the ultimate form of the construction in question. Furthermore, it is posited that the integration of domains is triggered by the co-equivalence of the trajector of Ezafe and that of copular verb.

Finally, the paper focused on the so-called ergative construction in Kurdish to prove that it is conceptually connected to topic-possessives through an inheritance link. It is asserted that possessive clitics invoke an asymmetric and intrinsic relationship between the agent and the patient. The asymmetric relationship between the Pr and the Pe is metaphorically reflected in the asymmetric connection between the agent and the patient. Furthermore, it was maintained that the target/reference alignment of the possessive clitic is identified with the trajector/landmark alignment of the verb in the final construction. As a result, the domain of the verb is construed in light of possessive domain. Finding similar non-possessive constructions whose internal organization resembles that of possessive ones in Hungarian, and seeking to cognitively account for their structure in terms of those principles which have been already evoked for possessive constructions could constitute an interesting line of investigation for further research in the future.