Ͻkere is doing something different in adnominal possession

. Cross-linguistically, some languages make a morphosyntactic distinction between alienable and inalienable adnominal possession, where alienable possession is more morphologically marked, and inalienable possession shows a tighter structural bond between the possessor and possessee. In this paper, I show that Ɔkere violates these cross-linguistic generalizations differently. I also show that two types of mó occur in the language ; one is a possessive marker, and the other is an independent pronoun. Again, I show that the nature of the possessive marker and the independent pronoun leads to a pro-drop in inalienable possession. The data and analysis in this paper favor proposing an overt possessive marker and a covert possessive marker. This paper adds to the literature on the exceptions to the cross-linguistic generalizations on adnominal possession by showing that the exceptions to the cross-linguistic generalizations may manifest differently in some languages.


Introduction.
Possessive split languages have a pattern, and Ɔkere violates the pattern.The pattern shows that possessive split languages make a morphosyntactic distinction between alienable and inalienable adnominal possession, where alienable possession is more morphologically marked, and inalienable possession shows a tighter structural bond between the possessor and possessee (Ultan 1978;Seiler 1983;Haiman 1983;Nichols 1992;Heine 1997;Haspelmath 2008;Myler 2018;Karvovskaya 2018).I provide this cross-linguistic generalization in (1) from Myler (2018: 1).I will refer to (1a) as the markedness generalization and (1b) as the structural proximity generalization in this paper.
If there is a contrast between alienable and inalienable possession with respect to the presence of morphological structure, alienable possession is always more morphologically marked.b.
Inalienable possession involves a tighter structural bond between possessee and possessor.
In Ɔkere we see that inalienable possession is more morphologically marked, and the structural bond between the possessor and possessee in inalienable possession is not tighter than that of alienable possession.I illustrate this in (2).
(2) a. ɛ -nyɛ ́ à kyá In (2), Ͻkere violates the markedness since there is an extra morphological marker in (2c).On the face of the data, the structural proximity generalization is difficult to establish because (2b) and (2d) have the same linear structure, but I will argue that Ɔkere violates the structural proximity generalization.To do that, I make the following points.
Ͻkere has two types of mó, one which is a possessive marker.b.
Ͻkere exhibits pro-drop in possessive constructions.c.
There are two Poss heads, one overt and one covert, and both form a constituent with the following noun.e.
The need to revisit the structural proximity generalization.
The rest of this paper is organized as follows.Section 2 provides background information about the Ͻkere language, its people, and the data collection procedure.Section 3 discusses the literature on adnominal possession focusing on the structure of this type of possession and the concept of alienable and inalienable possession cross-linguistically. Section 4 gives an in-depth analysis of adnominal possession in Ͻkere.In this section, the paper gives reasons why the claims of the paper are well grounded.Section 5 provides some typological implications and future directions.The conclusion to this paper is provided in section 6.

2.
The language and its speakers.Ɔkere belongs to the broader Niger-Congo family.It is a Guan language that falls under the Kwa language family.Ɔkere is a Hill Guan variety spoken in the Akuapem mountains by some 80,000 speakers alongside Akuapem-Twi (a dialect of Akan), as a second language.A large population widely knows the Ɔkere language, and people of the Ghanaian community as Kyerepɔn (Cherepon).

3.
The literature on adnominal possession.In many languages, there are no morphosyntactic distinctions in adnominal possession.When there is a distinction (grammatically) between alienable and inalienable possession, alienable possession involves a large set of members, and the set of inalienable nouns usually is a closed set (Nichols 1988cited in Heine 1997).This situation is attested in many languages, and Ɔkere also follows this pattern.Heine (1997: 10) shows that the nouns or concepts that are often categorized as inalienable cross-linguistically are "kinship roles, body parts, relational spatial concepts, parts of other items, physical and mental states, and nominalizations where the 'possessee' is a verbal noun."Languages differ in what they categorize as inalienable nouns, and as such, it is difficult to give a universal ranking of inalienable domains (Heine 1997).
In Paamese and Tinrin, for instance, kinship, body parts, and spatial relations count as inalienable, but in many Australian languages, kinship relations are alienable while body parts are inalienable (Heine 1997).Even in some languages, certain words which may fall under a category of nouns may behave differently from the group they fall within.We see this in Saker, a NE-New Guinea language, where kinship relations are inalienable, but words for 'husband,' 'wife, ' and 'child' are alienable. Chappell and McGregor (1996), following Lévy-Bruhl (1914), add that inalienability involves any form of a tight relationship between the possessor and the possessee (Heine 1997).In what follows, I provide examples of alienable and inalienable possession.

ALIENABLE AND INALIENABLE POSSESSION.
The constructions discussed in this section follow markedness and structural proximity generalization, which I rephrase here: when there is a split in adnominal inalienability marking, the alienable possession will have an overt possessive marker1 , while in the inalienable possession, the possessive marker is morphologically absent.Structurally, the relationship between the participants in inalienable possession is closer than inalienable possession.I provide illustrations below.
(4) Kabba (Moser 2004: 120-121  The constructions in (4-10) are consistent with the markedness and the structural proximity generalization in (3).In these examples, the possessees licensing the absent possessive marker are either kinship nouns (5&6) or body parts (4, 7-10), and those that license the possessive marker are non-kinship and non-body part nouns.These constructions have been given some motivation by Haiman (1983), who claims that the possessive marker is absent when the possessee is closer to the possessor and present when the possessee and the possessor are not close enough.Haiman (1983) asserts that 'arm' in (9b), for instance, is not seen to be separate from the body; likewise, 'garden' in (9a), so referring to 'arm' involves a smaller semantic distance compared to 'garden,' hence the presence of bi to indicate the distance between the possessor and possessee.Haspelmath (2008), on the other hand, argues that it is the frequency of occurrence that plays a role in the absence or presence of the possessive marker, in the sense that the greater proportion of the use of a noun in a possessive construction, the more likely is it for the possessive marking to be reduced.Thus, since 'arm' occurs more frequently than 'garden' in possessive constructions, the possessive marker is expected not to surface.Haspelmath (2008) further adds that the syntax of adnominal possession is the same in that overt possessive marker is expected and that languages do not usually express it when they are inherently possessed to avoid redundancy.This assertion is entertained in this paper such that the syntax for adnominal possession in Ɔkere will either have a null or overt possessive marker.The motivation for this is encouraged by Myler ( 2018), where possession relations on non-inherently relational roots are introduced by POSS head.This is discussed in detail in section 4. From the examples provided above in (11-14), possession involving kinship2 nouns are more morphologically marked than other possession involving other types of nouns.The data in ( 12) also tells us that pronominal possessors involving kinship nouns behave differently from nonpronominal kinship possession.In pronominal kinship, we have a juxtaposition of two lexical items, which on the surface looks like the possessor and the possessee, just like we have in nonkinship possession, but it is possible for pronominal possession involving kinship nouns to have the same morphosyntactic structure as the non-pronominal kinship possession in a focused context.I discuss this in section 4.1.

Inalienability in
4.1 THE POSSESSIVE MARKER, MÓ.In this paper, I argue that two types of mó exist in Ɔkere: an independent pronoun, mó, and a possessive marker, mó, which only occurs with inalienable possession.This position is motivated by Toivonen's (2000) analysis of Finnish possessive suffixes.
According to Aikhenvald (2013: 41), possessive markers cross-linguistically "may have additional meanings and extensions, not directly related to possession, nor ownership."It has been attested that the possessive marker can mark definiteness (Collinder 1957;Schlachter 1960;Tauli 1966;Fraurud 2001;Suihkonen 2005;Luo 2013).Looking at the literature on possessive markers, it makes sense that the possessive marker in Ɔkere can be described within the context of Aikhenvald (2013) that it has external features that it shares with other aspects of the Ɔkere grammar.
The possessive marker, mó, is homophonous with the 3SG object marker (see Table 1 below adapted from Oppong (2019)), a fact that is attested in other related languages.This is a common pattern in the area where the possessive marker and the 3SG are homophonous, also found in Fante (Boadi 2010) and Tongugbe (Kpoglu 2019).It is, therefore, not surprising that the possessive marker, mó, in Ɔkere, has the same number and person features as the third-person object pronouns.The possessive marker in Ɔkere agrees with the possessor in number and person, thereby causing it to spell out in different forms, just like the pronouns in Ɔkere.In other words, the possessive marker in Ɔkere, POSS, has unvalued number and person features, which get interpreted by the feature specifications of the possessor.Toivonen (2000) discusses independent pronouns and suffixes in Finnish possession and proposes a lexical split analysis (a single phonological form corresponding to two different functions, thereby having different lexical entries) for the suffixes and rejects a uniform account.Toivonen's (2000) analysis, involving a dual function for the possessive suffix in Finnish possessives, is relevant to the discussion of pronominal mó and POSS mó in Ͻkere; in fact, my proposed analysis will build on elements of it.
Possessive DPs in Finnish can be made up of two elements that both have person and number features: a suffix and an independent pronoun.The suffix is obligatory in first and second-person possessive DPs, but the independent pronoun is optional, but that is not the case for third-person possessive suffix.As a result, she rejects a uniform analysis for the suffixes.Her main empirical argument is based on the observation that the presence of an independent pronoun in the third person changes the interpretation.This is illustrated in ( 15) and ( 16).
(15) Finniah (Toivonen 2000: 590) Pekka pesee autoa-nsa.Pekka washes car-3Px 'Pekkai is washing his/heri/ * j car.' ( 16) Finniah (Toivonen 2000: 590) Pekka pesee hänen autoa-nsa.Pekka washes his/her car-3Px 'Pekkai is washing his/her * i/j car.' In the absence of the independent pronoun hänen, as in (15), a reflexive reading for the possessive suffix is required, but in its presence, a reflexive interpretation is disallowed, as in ( 16).According to the binding theory of Chomsky (1981), an 'anaphor' must be bound in its binding domain, and a 'pronoun' must be free.In (15), -nsa behaves like an anaphor, and in (16), it does not.A uniform analysis that did not treat -nsa as an anaphor would overgenerate for (15), allowing either an anaphoric or non-anaphoric interpretation.A uniform analysis that treatednsa as an anaphor would work perfectly for ( 15), but it would not allow the free reading in ( 16).Therefore, positing that the form -nsa has two different lexical entries is the best analysis for Finnish.In other words, there are two lexical items that have the same phonological form but separate lexical entries.Following Toivonen's arguments above, I present an argument for a split analysis here based on the data in Ɔkere.The situation in Finnish that Toivonen discusses is not the same as what we find in Ͻkere because, in Ͻkere, there is no obligatory reflexive interpretation when there is an overt pronoun.Therefore, we cannot use the reflexive vs. non-reflexive argument to decide whether there are two mó's or one mó, but nevertheless, an argument can be made on the basis of the facts.Suppose that there is only one mó which is a pronoun; then we would expect that it should be optional.If it's a pronoun, and its only duty is to refer to the R-expression, then its absence should not cause any problems or give different interpretations.As it occurs in Finnish, pronoun doubling in the first and second person possessive constructions allows an independent pronoun to be omitted.If Ͻkere is a language that permits pronoun doubling, then it may be expected that the pronoun can be optional without arriving at ungrammaticality, but that is not the case in Ͻkere.
Furthermore, you would expect it to be possible in all constructions if it's a pronoun.If it was only one mó, the pronoun, then we should be able to double the possessors in John mó ní 'John's mother' as it happens in kinship nouns and all other kinds of nouns illustrated below.
(17) John (*mó) àbá John POSS hand 'John's hand' (18) John (*mó) kyá John POSS building 'John's building' The constructions in ( 17) and ( 18) are ungrammatical.This goes against the claim that there is only one pronoun since pronoun doubling should be blocked by any possessee.If, on the other hand, it is a possessive marker that is only licensed with certain kinds of nouns, we can explain the obligatoriness of mó in inalienable possession because the syntax requires it.I, therefore, propose that we have two different lexical entries for mó: the referential pronoun mó in mó kyá 'his/her building' and the possessive marker mó in John mó ní 'John's mother.'This analysis explains the facts in Ͻkere in the sense that mó is obligatory if the possessed noun is a kinship noun.Since only inalienable possession can license mó, then mó is a lexical item that has a relationship with only kinship nouns.If mó is a pronoun in the language, then its distribution should not be constrained to a group of possessed nouns.That said, I will show in the ensuing sections that the possessive marker, mó, and the pronoun, mó, occur at different syntactic positions.
As illustrated in ( 17) and ( 18), the POSS heads should be different for the constructions to be grammatical.If POSS head must be different in ( 17) and ( 18), then it means the POSS heads in the two constructions do not bear the features of the POSS head in inalienable construction.Therefore, the POSS head in alie will have no unvalued features since the POSS head in inalienable possession bears unvalued features.I show the implication of POSS head having unvalued features in the subsequent paragraph.

ARGUMENT FOR PRO-DROP IN ϽKERE.
In this section, I make a case for my claims that the prodrop phenomenon is attested in Ɔkere.The analysis made here largely follows Roberts's (2009) defective goal approach.
In inalienable possession involving pronominal possessors, the pronominal possessor is dropped.The deletion of the pronominal possessor is akin to a deletion analysis of null subjects, which has received several proposals in the literature (compare Rizzi 1986;Huang 2000;Chomsky 2001;Holmberg 2005;Roberts 2009).The view that pro has valued features at SpecTP and behaves just like an overt pronoun proposed by Holmberg (2005) is entertained in the current paper on Ͻkere, where pro also bears valued features at SpecPossP.POSS in Ͻkere agrees with the possessor in inalienable possession and bears unvalued features [uNum:__, uPers:__], which need to be valued by pro because pro is valued since if POSS and pro have unvalued features, neither of them can value each other.Therefore, following Roberts (2009: 60), the agreement relation between POSS and the pro will be represented as shown in ( 19).The outcome of Agree in (b) shows a feature bundle being presented twice, so one of the copies must be deleted before Spell Out, following Nunes' argument of chain reduction (Nunes 2004) and Roberts' (2009:76) generalization that "[d]efective goals always delete/never have a PF realization independently of their probe."Following Roberts (2009), Livitz (2011:108) argues that pro is "silent cross-linguistically because it is a defective goal whenever it enters into an Agree relation." Using the defective goal account,4 Livitz (2011) argues that pro can be overt when it bears a feature absent in the probe that it agrees with.In other words, pro can be overt if its features are not a subset of the probe.Adopting the thinking of Livitz (2011), pro is overt if it has a feature that is not present in poss.This account aligns with pro in Ͻkere in that pro is overt when it has [+FOC] feature specifications.This is illustrated below.
(20) mó mó ní pro;foc POSS;3sg mother 'his/her, his/her mother.' (21) wó wó ní pro;foc POSS;2sg mother 'your, your mother' In (20), mó, the left edge bears focus and hence bears [+FOC] and takes the form of POSS based on the person and number features on poss.This makes it possible for the overt realization of the left edge pro in (20) and (21).The above analysis does not work in alienable possession since POSS is always null.For an alienable possession to bear focus, the focus element, in this case, the pronoun, will have to be introduced higher above the possessor because the possessor node will not be a landing site for the focus element since it is already occupied.
So far, the data has shown that Ɔkere violates the markedness and the structural proximity generalizations.The markedness generalization is violated because inalienable possession involves an overt possessive marker.This is in opposition to the markedness generalization.As it stands, the structural proximity generalization is also violated because, on the face of the data, the structural bond between the possessor and the possessee in alienable possession is closer than inalienable possession.The presence of the overt possessive marker that breaks the juxtaposition of the possessor and possessee in only inalienable makes it possible to conclude that Ɔkere is a counterexample to the structural proximity generalization just like markedness generalization.However, in the next section, I show that the structure of adnominal possession that I propose for Ɔkere will show a different violation of the structural proximity generalization.

THE SYNTAX OF ADNOMINAL POSSESSION IN ƆKERE.
As mentioned earlier, the data in Ɔkere make it challenging to postulate two different structures in Ɔkere following Myler's (2018) account.Since body parts are inherently relational nouns and bear npart-of, they can introduce a possession relation (Myler 2018), but the behavior of sortal (not being able to introduce a possession relation) and kinship nouns in the language suggests that poss is always present, but it only Spells out in kinship possessive constructions, a prediction made by Haspelmath (2008).This account violates the structural proximity generalization because positing that a poss head is always present in Ɔkere adnominal possession leads to a conclusion that inalienable possession is as tight as alienable possession.I refer to the violation of the structural proximity generalization as 'partial' and the markedness generalization as 'total' because of how Ɔkere violates these generalizations.
Based on the account given in this section, the syntactic trees for sortal nouns, body parts, and kinship are presented in ( 22), ( 23), and (24), respectively.In the case of pronominal possessors, inalienable will involve a pro-drop as discussed in section 4.3 while alienable possession will involve an overt pronoun.I illustrate this below.( 25

√ 𝑠é n{}
Again, as established in section 4.3, mó can be doubled when we have a pronominal possessor and a focused context.This gives rise to constructions like ( 27) and ( 29) and their corresponding tree structures in ( 28) and ( 30 Nichols (1992) and Myler (2018) made this observation in Dizi, an Omotic language spoken in Ethiopia, and Tzutujil, a Mayan language, respectively.The behavior of Dizi, Tzutujil, and Ɔkere are only a few of the many languages that do not follow the cross-linguistic generalizations.Even though the three languages violate the cross-linguistic generalizations, the generalizations cannot be abandoned statistically, an opinion held by Nichols (1992).
The violations caused by the analysis provided for Ɔkere as it stands mean that some languages may violate only one or aspects of the two generalizations.This is because a revised version of the structural proximity generalization stated here, the structural bond between the possessor and the possessee is always at least as tight or tighter with inalienable possession than with alienable possession, will work perfectly for languages that have (in)alienability distinction and follow the two cross-linguistic generalizations on adnominal possession.
If Ɔkere may violate only the markedness generalization when the structural proximity generalization is adopted, does that mean the markedness generalization should be seen as a statistical universal?On the other hand, does it also mean the structural proximity generalization becomes an absolute universal?I leave these questions for future research.

Conclusion.
From the investigations done in this paper, kinship nouns form a group of inalienable nouns whose morphological structure involves more marking than alienable possession.This contrasts with the markedness generalization, which predicts that alienable possession will be more morphologically marked.Regarding the structural proximity generalization, the account provided in this paper concludes that there is no difference between the syntax of alienable and inalienable possession in the sense that POSS is always present (overt POSS and covert POSS).This conclusion violates the structural proximity generalization and leads to the revised version of the structural proximity generalization.
As regards the possessive marker, the paper argued that two forms of mó exist.The paper also argued that pro-drop is attested in Ɔkere because the pronominal possessor is omitted in inalienable possession.The omission results from the fact that the pronominal possessor and the possessive marker in inalienable possession have the same form and bear the same number and person features and that the possessor is recoverable in the syntax when it bears focus.

Table 1 .
Table showing the pronominal system in Ɔkere 4.2 ARGUMENTS FOR A SPLIT ANALYSIS IN ϽKERE.