ngunhu-trarrathathu-rnu There appears to be a misprint in the original source. puni-lha mimi-nkgu Kanbuny

that.nom-du send-passp go-past uncle-eff Karnuny-eff

`Those two went off sent by their uncle Karnuny'

Reasoning above, in what follows the data on these languages is not taken into account in calculations.

Overall, there are 58 languages in the sample that represent case marking either in the form of a morphological case or a clitic. The data on most of these languages (41) was already considered in (Logvinova 2019). The results were that in 21 languages without case agreement in NP, appositional phrases required only one case marker for the whole group, as in the following example from Ladakhi:

(76) Ladakhi (< Bodic < Sino-Tibetan, Bible in Ladakhi, glosses are mine)

Yordan ltsa?s-poe / Yordan ltsa?s-pone

Jordan river-acc / Jordan река-all

`in the river Jordan' `towards the river Jordan'

This is the third (3) type scenario, as it is introduced above, which means that nouns in apposition in those languages are treated as members of a singular NP, not as two concatenated NPs. The only exceptions to this observation were Lak (77) and Koasati (78) (Logvinova 2019a):

(77) Lak (< Lak-Dargwa < Nakh-Daghestanian, (Kakvaeva 2010: 54; Kazenin 2013: 34))

a зузала Гьарун / зузала Гьарун-ул

worker G'arun / worker G'arun-gen

`the worker G'arun' `worker G'arun's'

b. <...> вазирна-чIа Низамулмулклу-чIa

vizier-loc Nizamulmulklu-loc

`by the vizier Nizamulmulklu'

(78) Koasati (< Muskogean, (Kimball 1991: 388, 479))

a. hahciм coba-faм o?aм-:ci-n

river big-loc be-3non:3g-sw

`They arrived at the big river <....>'.

b. cokfi-k am-oklaм-k im-hoм:pa-Vмhc-on

Rabbit-subj 1sposs-friend-subj 3stats-be_hurt-habit-sw:foс

`The case is that my friend Rabbit was injured'.

The majority of 20 languages with case agreement in the NP domain required a case marker to be present on each member of an appositive phrase, as in the following example from Georgian:

(79) Georgian (< Kartvelian, А. Konиixina, personal communication, transcription and glossses are mine)

a. ??? ?????????? b. ???-? ??????????-?

dzma aleksandre dzma-s aleksandre-s

brother Aleksandre brother-dat Aleksandre-dat

`brother Aleksandre ' `(to) brother Aleksander'

Only four languages, namely, Albanian, Romanian, German and Chechen, were different from that pattern having a case marker only on the common noun. The order of elements in such constructions in these languages is such that a common noun preceeds a proper name, therefore, the case marker cannot be considered to be a group morpheme. An example from Albanian is given below:

(80) Albanian (< Indo-European, М. Karagjozi-Kore, personal communication, glosses are mine)

president-it Ilir Meta

president-dat Ilir Meta

`(to) the president Ilir Meta'

It is reported for Old Albanian (Dindelegan, Maiden 2016: 368-371) that the repetition of the case marker was possible, but not obligatory and by the end of the OR period appositions without agreement became more frequent. Appositions displaying dative or vocative agreement are mentioned to be found even in the present-day language (ibidem), but the corresponding examples are absent from the corpus (roTenTen15). The place of case marking in German depends on the absence or presence of the definite article in the phrase: Prinz Adalbert-s_GEN Pony vs. das Pony des Prinz-en_GEN Adalbert `the Pony of Prince Adalbert' (Kunkel-Razum, Mьnzberg: 2009: 993) See more detailed examples in (Logvinova 2019a).. The fact that Balkan languages (Albanian, Romanian) pair with the Slavic pattern (see examples from Russian above and a more detailed discussion in Section 4.3) leads to believe it to be an areal phenomena.

The list of languages with case marking not discussed before includes Akuntsu, Anaamuxra, Arrernte (Mparntwe), Cocama-Cocamilla, Epena, Haida, Huambisa, Mohawk, Movima, Northwest Sahaptin, Nubian (Dongolese), Quechua (Huallaga), Tariana, Tundra Nenets, Bininj Gun-Wok and Warrongo (17 all in all). The two of them (Bininj Gun-Wok and Warrongo) are Australian languages for which case marking on each member of NP is common (if an NP is at all existent), as was discussed above. Exactly 7 of the remaining languages use postpositional clitics to express case and do not show case agreement in the NP domain. These languages are Arrernte (Mparntwe), Cocama-Cocamilla (81), Epena (82), Haida, Nubian (Dongolese) (83), Quechua (Huallaga) (84), Tariana (85) and Teso. The relevant data on position of case markers was not available for the language Haida (as it was impossible to analyze the examples from the available Bible script properly). In all the other cases the postpositional clitic only appears on one of the members of an appositive phrase as the following examples show:

(81) Cocama-Cocamilla (Yopбn 2010: 844)

raepe, ikian, arkare na chira-n eee rakuna ritama-tsu-i

then this mayor qt1 name-nzr *** Lagunas town-abl-ndr

`<...> en then the one that is called the mayor, who came from the Lagunas town <...>'

(82) Epena (Choco < Choco, Epena Bible, glosses are mine with reference to (Hams 1994))

a. Nazaret p'uuru-dee-pa

Nazareth town-loc-from

`from the city of Nazareth'

b. p'uuru Nazaret-de

town Nazareth-loc

`to the city of Nathareth'

(83) Nubian (Nubian < Eastern Sudanic (Bechhaus-Gerts 2011: 33))

baraka mulki abuuna Dauud-ni-ga

blessing kingdom our_father David-gen-obj

`blessed be the kingdom of our father David'

(84) Quechua (Huallaga) (< Quechuan < Quechuan (Weber 1983))

Nazaret marka-n-kuna-man

Nazareth town-3p-plur-goal

`to their city of Nathareth'

(85) Tariana (< Northern Maipuran < Arawakan (Aikhenvald 2003: 630))

nese-na kwaka Venezuela-se nese-na kwaka rio

then-rem.p.vis what Venezuela-loc then-rem.p.vis what river

Orinoco-naku di-wa pa:-pua Kunukunuma kepitana-pua

Orinoco-top.non.a/s 3sgnf-enter one-cl:river Kunukuma rel+name-cl:river

`There, in Venezuela, what's its name, out of the Orinoco river, comes out another (river) called Kunukunuma'

The only Australian language in the sample without obligatory case marking on each member of an NP is Arrernte. According to (Wilkins 1989: 102) the order of NP constituents in Arrernte is rigid and case clitic is always placed on the right edge of the phrase. As can be seen from example (86), the same patten is found with nouns in apposition as well:

(86) Arrernte (Mparntwe) (< Central Pama-Nyungan < Pama-Nyungan (Wilkins 1989: 22), Bible in

Arrente; glosses are mine, with reference to (Wilkins 1989))

a. <...> ape brother Bennett-le awerne-rle nhenhe-rle akwete re

and brother Bennett-erg poor_thing-rel here-rel still 3sgs

ne-me

be-npp

<...> `and brother Bennett, who is still here poor thing'.

b. apmere Nazareth-ele

place Nazareth-loc

`from the city of Nazareth'

c. apmere ikwere-atherr-enhe Nazareth-werne

place 3sgdat-two-poss Nazareth-all

`to their home town of Nazarath'

Among the previously not discussed languages having morphological case are Akuntsu, Anaamuxra, Huambisa, Mohawk, Tundra Nenets and Northwest Sahaptin. Except for Northwest Sahaptin and Tundra Nenets, the languages in this group do not show case agreement in NP. The available data on the Mohawk and Akuntsu is ambiguous and requires further clarification. As can be seen from the examples below, in appositive constructions the two remaining languages exploit one case marker for both constituents:

(87) Huambisa (Pen?a 2015: 683), Bible in Huambisa

a. t??ii nui?=ka sai-ru t?suna=hai? hintina=t?su=k?

okay there=foc brother.in.law-1sg Tsuna=com trail=infer=restr

?ia-ka-ara-tai

go.pl-intens-pl-infer

`... oh, then, with my brother-in-law Tsuna, they must have gone'.

b. yaakat Nazaret

yaakta-rin Nazaret-num

town-1pl/2pl.3.poss Nazareth-loc

`to their city of Nazareth'

(88) Anamuxra ((Ingram 2001: 300-302)

a. Arari vwa-a-ki

pn settlement-nd-obl

`at the village of Arari'

b. Xzxivu ani

pn river-obl

`along the Xzxivu river'

In Sahaptin noun modifiers show obligatory agreement in case with their heads (Jansen 2010: 290). The same is true of the nouns in constructions of specification:

(89) Northern Sahaptin (< Sahaptian < Penutian, (ibidem))

wiмya-pa niмxyaмawi-pa p?мt'x?anuk-pa

far.away-loc Nixyaawi-loc mountain-loc

`far away at the Nixyaawin mountain'

The scenario found in Sahaptin, technically, does not allow any conclusions about the constituency in constructions of this sort.

In Tundra Nenets, case agreement in adjectives is always optional and is dependent on a number of conditions (Nikolaeva 2014: 153), but in apposition only one case marker per group is possible (ibidem: 171): Ivanov l'ekar?-m_ACC `doctor Ivanov (acc)'. Given that case agreement is not obligatory in general and the natural motivation is to avoid redundant marking, that appears only logical. In this study it is treated as an example of the third (3) scenario.

To sum up, 7 languages are excluded from calcualtions: 4 Australian languages (Bininj Gun-Wok, Martuthunira, Wardaman and Warrongo), 2 languages for which the data on case distribution in appositive phrases is absent or is not reliable (Mohawk and Akuntsu) and the language Movima which uses a strategy other than apposition to express specification. The remaining 51 case-marking languages of the sample are distributed between the types introduced in Picture 6 as follows

Table 4

The 4 types of distribution of case feature in appositive phrases

Case agreement in NP	Are both appositives marked for case?
	yes	no
yes	18	3
no	1	29
Total	51 The remaining 7 languages are: 4 Australian languages not taken into account in this section (Bininj Gun-Wok, Martuthunira, Wardaman and Warrongo), 2 languages for which the data on case distribution in appositive phrases is absent or is not credible (Mohawk and Akuntsu) and the language Movima which uses a strategy other than apposition to express specification.

As Table 4 shows, the majority of the languages considered follow either the third (3) scenario and have only one case marker for the whole appositive phrase or the first (1) scenario and mark both members of appositive expression for case. The third one clearly shows that at least in some languages constructions of specification are treated as any other NPs with modifiers.

Regarding the first scenario, it has been said above that it appears to be controversial due to the fact that the presence of the case marking on both constituents can be either the result of the agreement process within NP or separate case marking of concatenated NPs. Examples from Hungarian and Khwarshi can serve as a possible solution to this obstacle. Hungarian toponyms in illative contexts get a case ending different from one going with common nouns, namely, superessive instead of illative (To?rkenczy 2002). Due to this fact, a sortal name and a toponym may receive different case marking in appositive constructions:

(90) Hungarian (< Ugric < Ugric, huTenTen12, glosses are mine, with the reference to (To?rkenczy 2002))

leg-kiss-ebb vбrosб-ba Pбlhбzб-ra

sup-small-sup city-ill Pбlhбza-supess

`to the city of Pбlhбza'

The same split is observed in an example from Khwarshi:

(91) Khwarshi (< Avar-Andic-Tsezic < Nakh-Daghestanian, (Khalilova 2009: 297))

aл-a Iq?q?o dah hadam goli

village-in Инхоквари.cont few people be.PRS

`There are few people in the village, in Inkhokwari'.

In (91), the common noun meaning `village' is in inessive case, while the proper name is marked for contessive. Examples like (90) and (91) give grounds to think that the appearance of more than one case marker per group can be not due to the process of agreement (in which case one would expect case markers to be the same), but due to the separate assignment of case to each nominal. If that is true about all the cases of double case marking in appositive phrases (even in languages showing identical markers on the surface), that should be considered an evidence in favor of the juxtapositional analysis. However, whether two rather exotic examples can stand for the whole class is a question.

4.2 Towards a better understanding of constituency

In the previous section it was argued that languages tend to treat nouns in apposition as a single NP rather than juxtaposed NPs, as in languages without case agreement a case marker appears only once per appositive phrase. However sometimes apposition shows characteristics of constituents of a level lower or higher than NP.

At least in some languages with preposition of proper names to common nouns (cf. Abkhaz Гагра а?ала?ь `the city of Gagra', see more details on the constituent order patterns in Chapter 5) appearance of other modifiers in the same NP is not tolerated. Introduction of a modifier results in the change of the order of constituencies so that a proper name is placed after a generic term. This rule is explicitly stated at least in the grammars of Lezgian (Haspelmath 1993: 256) and Chuvash (Aљmarin 1903: 5).

(92) Lezgian (< Lezgic < Nakh-Daghestanian, (Haspelmath 1993: 256)

a. Kristina wax `my sister Christine'

b. Q'wed lahaj wax Kristina xudoz?nik ja.

two ord sister Kristina artist cop

`My second sister Christine is an artist'.

c. *Kristina q'wed lahaj wax

(93) Chuvash (< Turkic < Altaic, (Aљmarin 1903: 3, 5), the glosses are mine)

a. Mikulaj timлr?л

Nikolaj blacksmith

`the blacksmith Nikolaj'

b. pujan xu?? Xvetлr

rich merchant Fedor

`the rich merchant Fedor'

According to the available data, the same phenomenon is observed in Hungarian, Armenian and Korean. These languages place adjective modifiers before their heads. A conventional order of a common noun and a proper name in appositional constructions with toponyms in both languages is PC (proper -- common). But when a common noun is modified by an adjective, the order changes as in Lezgian. A remarkable detail about this permutation is that when placed linearly after its generic term, proper name requires its own case marker in Hungarian and Armenian:

(94) Hungarian (< Ugric < Ugric, KonText.cz)

a. eмs onnan Gid vбrosб-ba men-tek

and from_there Gid city-ill go-prs.3pl

`And they come from there to the city of Gid'.

b. (huTenTen12)

leg-kiss-ebb vбrosб-ba Pбlhбzб-ra

sup-small-sup city-ill Pбlhбza-supess

`to the city of Pбlhбza'

(95) Armenian (< Indo-European, М. Mkrtc?jan, personal communication, glosses are mine)

a. ?? ???-??? ?? ?????? ?????-???

es apr-um em Yerevan k'aghak'-um

I live-pc.pres be.cop Yerevan city-loc

`I live in the city of Yerevan'.

b. ??????? ?????*(-???) ?????-???

gexecik k'aghak'*(-um) Yerevan-um

beautifui city-loc Yerevan-loc

`in the beautiful city of Yerevan'

Although in Korean order changes as well, only one case marker is required:

(96) Korean (A. Pukhovskaia, personal communication)

a. ј­їп µµЅГїЎј­

seoul dosi-eyse

Seoul city-loc

`in the city of Seoul'

b. Е« µµЅГ ј­їпїЎј­

keun dosi seoul-eyse

big city seoul-loc

`in the big city of Seoul'

The constraints on modifiers inside the appositive phrase suggests that appositive phrases have the status smaller than that of a full NP. Examples (92) and (93) show that in soma languages it is no longer possible to consider both members of apposition to be a single constituent as it gets heavier. According to criteria for constituency introduced above, the appositive phrases in (92)b and (93)b, should be treated as two separate concatenated NPs. However, they still formally qualify as “close” appositional constructions on the criterion of intonational integrity. Loose appositions (that are characterized by prosodic detachment) in Hungarian and Armenian, as well as in some other languages from the sample with one-time case marking in NP (namely, Basque, Chuvash and Yukaghir) require a case suffix to appear on each of the two nouns:

(97) (huTenTen12)

Egy kis vбros-ban, Kleуnai-ban szбll-t meg egy szegйny parasztnбl

one little city-loc Kleonai-loc stay-pst.3sg even one poor farmer

`He stayed in a one small city, in Kleonai, with one very poor'.

Yet intolerance to modifiers is not universal. In Basque, although attributive modifiers precede their heads, proper names remain in preposition and do not get their own case marker:

(98) Basque, ((Hualde, de Urbina 2011: 166), N. Zaika, personal communication, SEG; glosses by N.

Zaika)

a. Erroma hiria

Rome city

`the city of Rome'

b. Estatu Batu-eta-ko Savannah hiri-ra hel-du dira

state unite-pl-adj Savannah city-all arrive-pfv aux

`They arrived in the city of Savannah in the USA'. lit. `They arrived in the city of Savannah of the USA'.

c. Orleansville Algeria-ko hiri-an sortua

Orleansville Algeria-adj city-inn born

`(the one) born in the Algerian city of Orleansville'

In (96)b a genitive-like modifier On the functions of Basque affix -ko see (de Rijk 2008: 91-95). Estatu Batuetako precedes the whole appositional phrase, while the city name is adjacent to the common noun. In (96)c, by contrast, a modifier of the same Algeriako type breaks the appositional phrase, nevertheless, a case marker appears only once per group Adjectival modifiers in Basque come after the nominal head, but examples with “true” adjectives added to appositional phrases are absent from the corpus, so that it is impossible to give a minimal pair to the examples from Hungarian..

In Ingush adjectival modifiers precede their heads and a proper name in constructions with toponyms comes before a common noun. Addition of an attributive modifier to an appositional phrase does not result in apposition break and the order of constituents remains the same as in phrases without adjectival modifiers:

(99) Ingush (Nakh < Nakh-Daghestanian, М. Dahkilgova, personal communication)

a. Магас г1ала

Magas cuty

`the city of Magas'

b. xозача Магас г1ала

beautiful Magas city

`the beautiful city of Magas'

On the other hand, an important addition to the proposed theory of constituency and case marking in appositions comes from the constructions with coordination. In all the languages with a group case morpheme (languages marking case only once per NP), for which it was possible to find the required examples, the coordinated nominals generally receive case markers independently, as can be seen from the following examples in Basque:

(100) Basque (EPG, glosses are mine) In fact, Basque allows to use only one case morpheme per coordinated phrase (de Rijk 2008: 840).

Tbilisi-ren eta Batumi-ren arteko harreman-ak <...>

Tbilisi-gen and Batumi-gen between relation-pl.def

`The relationships between Tbilisi and Batumi'

However, when placed in apposition, the coordinated nominals do not show any case morphemes and the only case marker that appears is placed on the left most constituent of the phrase, i.e. on the common noun:

(101) (Yandex The research request was formulated as follows: " * eta * hirietan". There were more than 5 analogous expressions with different proper names in the result.)

Mendoza eta Cordoba hiri-etan

Mendoza and Cordoba city-pl.inn

`in the cities Mendoza and Cordoba'

The same is true about Hungarian, Chuvash, Yukaghir (100) and Armenian (101).

(102) Yukaghir (L. Kurilova, personal communication, glosses are mine)

a. Якутскай-?а Москва-?а

Yakutsk-loc Moscow-loc

`in Yakutsk and Moscow'

b. Якутскэй Москва гуорат-пэда-?а

Yakutsk Moscow city-pl-loc

`in the cities of Yakutsk and Moscow'

(103) a. ????? ??????????-??? ?? ???????-???

Sankt Peterburg-um ew Moskvay-um

Saint-Petersburg-loc conj Moscow-loc

`in Saint-Petersburg and Moscow'

b. ????? ?????????? ?? ?????? ?????-???-???

Sankt Peterburg ew Moskva k'aghak'-ner-um

Saint-Petersburg conj Moscow city-pl-loc

`in the cities of Saint-Petersburg and Moscow'

Regarding the problem of constituency, the fact that apposition can include coordinated expression can be considered as an argument for apposition to be a constituent of a bigger scope than just an NP.

4.3 Case marking in Slavic appositive constructions

This part of the present study is different from the rest of the study both in its tasks and methods. However, since it also addresses the problem of case marking in constructions of specification with proper names it is included in the present section. This part is inspired by the framework of intragenetic typology, as it is understood particularly in (Kibrik 1998). The intragenic approach is distinguished by its orientation to the survey of linguistic variation on the scale of a group of genetically-related cognate languages. Although being in a certain way opposed to the classical methodology of linguistic typology, this approach allows to give a more detailed description of complex linguistic parameters and to examine the possible continuity of change.

The languages of the Slavic family can all be attributed to the languages generally using the juxtapositional strategy with the preferred CP (common -- proper) order of constituents. It is important to recognize that the choice of Slavic language family in this case only partly stems from its suitability for the present theoretical objectives. Any other satisfactorily described language family could be chosen. The Slavic family is not a bad candidate in this case. First and foremost, all Slavic languages with the exception of Macedonian and Bulgarian have case marking on nouns and pronouns. Furthermore, the choice of the Slavic family is due to my native knowledge of Russian and the resulting modest capacity to read and understand original texts and the theoretical literature in other languages of the family, as well as the availability of relatively big electronic corpora for most of the Slavic languages.

14 extant Slavic languages are traditionally divided into three branches, namely, Western, Eastern, and Southern. In this section, I am going to analyze the material of six of them: Ukrainian, Belorussian (Eastern branch), Polish, Czech (Western branch), Croatian, and Slovene (Southern branch). The data on Russian is discussed in detail in (Logvinova 2018). The literature on apposition in these languages is scarce (but see, for example, (Kulik 1961, Kljuc?kovsky 1962, 1963; Mizak 1966;) for Ukrainian, (Biryla, S??uba 1986) for Belarussian, (Bartnika et al. 2004) for Polish) with some exception for relatively newly released studies of Croatian appositions (Markoviж 2008, Sesar 2013, Belaj 2014), focused mainly on the search for semantic grounds to define the head of the construction. To my knowledge, no quantitative study on apposition in Slavic has ever been conducted.

4.3.1 Methodology

The data discussed in this section comes from several electronic linguistic corpora available on the Sketch Engine platform. For all languages except for Croatian, the corpora from TenTen Corpus Family The TenTen Corpus Family is a group of web text corpora made to meet the same general standards, which ensures their comparability. For more detail see: https://www.sketchengine.eu/documentation/tenten-corpora/. were used. In the case of Croatian, there is no available TenTen Corpus for this language, that is why Croatian web-corpus (hrWaC Croatian corpus from the web (hrWaC ), URL: https://www.sketchengine.eu/hrwac-croatian-corpus/. ) was used. For statistical hypothesis testing, I predominantly used the multiple regression model (further -- MLR), as it allows to define in what way (positive or negative) and with which force several explanatory variables predict the outcome of a response variable. The significance level (the p-value) of ? 0,05 is taken as significant.

4.3.2 Case agreement in appositions in six Slavic languages

The problem of case agreement in close appositions was widely discussed with respect to the data from Russian. In Russian, certain types of proper names in apposition to common nouns can either have the same case as the preceding generic term or preserve the default nominative form. That is shown in the examples below:

(104) Russian (< Slavic <Indo-European (RNC: М. L. Gasparov. Zanimatrl'naja Grecija, 1998), glosses are mine)

a. Na ostrov-e Krit-e c?tili pes?c?er-u <...>

on island-loc Crete-loc honor;pst;pl cave-acc

`(they) honored a cave on the Crete island'

b. <...> razorila moguc?ee carstvo na ostrov-e Krit

ruin;pst;f mighty;n kingdom(n) on island-loc Crete-loc

`(it) ruined mighty kingdom on the Crete island'

In both examples in (102) an appositional phrase ostrive Krit(e) is dependent on the preposition na which requires it's dependent NPs to be in the Locative case. However, as the example (102)b shows, the case marker can be omitted on the proper name. This also appears to be true about other Slavic languages. See an example from Ukrainian

(105) Ukrainian (< Slavic < Indo-European (RNC: I. I. Akimus?kin. Tropoju legend, 1961), https://studopedia.su)

a. <...> jakij bes?ketuvav na ostrov-i Rodos-i

which rampage;pst;m on island-loc Rhodes-loc

`<...> who was rampaging on the Rhodes island'

b. kolosal'na statuja bog-a soncj-a Gelios-a na ostrov-i Rodos

enormous;f statue(f) god-gen sun-gen Helios-gen on island-loc Rhodes

`an enormous statue of the god Helios on the Rhodes island'

As in the above examples from Russian, a proper name in an appositive phrase in Ukrainian can either agree (103)a in case with the presiding common noun, or preserve the “default” Nominative form (103)b.

A bulk of literature (S?vedova et. al 1980: 58; Rozental 1989: 265-267; Graudina 1976: 138-145; Golub 2016: 278-279; Matushansky 2012, Superanskaja 2013, Logvinova, in press), discussing this this phenomenon in Russian, provides evidence that the agreement is conditioned by several factors, among which are the following:

– the lexical category of the common noun;

– the number congruency of constituents (i.e. the constituents have the same feature of grammatical number);

– the gender congruency of constituents;

– frequency of a proper name.

The underlying theoretical premise for the last of the factors listed above (i.e. frequency) was outlined before in (Logvinova 2018). The main assumption is that case marking of proper names in appositive structures may be conditioned by the degree of familiarity of the names for the speaker in the way that unfamiliar names tend to be preserved in their default nominative form rather than excessively display the case marking provided that it is already marked on the sortal term. The relative frequency of the particular proper noun in a representative collection of texts in the language appears to be a possible empirical indicator of the degree to which speakers are familiar to a particular name.

Regarding the factor of the lexical category of the common noun, it seems that generally Slavic languages make the same contrasts between nouns preferring and disfavoring case agreement of the adjoined proper name. Table 5 below gives an overview on the frequency of case agreement of proper names with different common nouns. The table is organized according to the following principles. In each column for the same expression (translated in different languages) it is calculated on the basis of the corpus, how many times this expression revealed case agreement or absence of case agreement and what is the percentage of cases with agreement. The choice of proper names is conditioned by their potential frequency and, as a consequence, their presence in the used corpora. In case of the expressions river + X and mountain +X (where X is a proper name), two categories of names are considered -- those having the same grammatical gender feature as the common noun and those with different gender. The cases with agreement frequency ?50 % are given in gray. As can be seen from the table, all the languages under discussion reveal tendency for agreement of proper name after the noun `city' and after the noun `river' when the grammatical gender of the proper name is congruent (feminine in this case). In other cases, agreement is generally avoided. The language showing agreement in most of contexts is Croatian.

Table 5

Case agreement of proper names after different common nouns in seven Slavic languages

Language

c The titles of columns give the following information: [C] =context -- the considered expression; [u/m] = unmarked / marked -- the number of cases with a proper name not agreeing with a common noun before slash and the number of cases with agreement after slash; [% m] = % of marked forms -- the percent of forms showing agreement. A cell is in gray when the percent of cases with agreement is higher than 50%. In case of cells containing [--], the required information was not found in the corpus. The note [(c)] given in lowercase after some names, mean that the name is congruent in grammatical gender with the preceding common noun. *** In case of feminine proper names of mountains different proper names are used.

u/m

% m

c

u/m

% m

c

u/m

% m

c

u/m

% m

c

u/m

% m

c

u/m

% m

c

u/m

% m

c

u/m

% m

city `the city of Moscow'

river `Volga river/ Danube river'

mountain `the mountain Sinai/ the mountain X***'

desert `the Sahara Desert'

lake `the lake Baikal'

island `the Zanzibar island'

planet `the planet Earth'

star `the star Sirius'

Russian

город Москва

1*103/3*105

99%

река Волга(c)

6*103/ 8*103

57%

гора Синай

5*103/ 400

7%

пустыня Сахара

2565/ 972

28%

озеро Байкал

>14*103/ 302

2%

остров Занзибар

699/ 24

4%

планета Земля

>3*104/ 174

<1%

звезда Сириус

513/ 39

7%

река Дунай

1500/ 194

11%

гора Белуха

582/ 429

42%

Ukrainian

місто Москва

4/16

80%

річка Волга(c)

70/ 96

57%

гора Синай

207 /8

3%

пустеля Сахара

342 /150

30%

озеро Байкал

488/9

2%

острів Крит

624/ 42

6%

планета Земля

>2*103/ 33

3%

зірка Сіріус

31/ 2

6%

річка Дунай

462/ 28

6%

гора Говерла

440/ 278

78%

Belarussian

горад Масква

2/20

90%

рака Волга(c)

0/3

100%

гара Сінай

20/ 0

0%

пустыня Сахара

4/ 3

42%

возеро Байкал

20/0

0%

востраў Зыслаў

33/ 2

6%

планета Зямля

46/ 0

0%

--

--

--

рака Дунай

8/0

0%

--

--

--

Polish

miasto Moskwa

57/ 129

70%

rzek№ Woіg№(c)

12/70

85%

gуr№ Sinai

54/ 0

0%

pustynia Sahara

102/ 71

41%

jezioro Bajkaі

636/3

1%

wyspie Zanzibar

136/ 3

2%

planeta Ziemia

1082/ 831

41%

gwiazda Sirius

3/0

0%

rzek№ Dunaj

165/1

<1%

gуra Cantoria

53/ 35

40%

Czech

mмsto Moskva

455/ 388

46%

шeka Volga(c)

6/7

53%

hora Sinaj

1800/ 102

5%

pouљќ Sahara

145 /67

31%

jezero Bajkal

863/10

1%

ostrщvekZanzibar

4/ 0

0%

planeta Zemм

3/48 + obl

94%

hvмzda Sнrius

166/ 0

0%

шeka Dunaj

1200/ 615

32%

hora Radhoљќ

861/32

4%

Slovene

mesto Moskva

10/ 11

52%

reka Sava(c)

17/ 2660

99%

gora Sinaj

82/ 0

0%

puљиava Sahara

31/ 11

26%

jezero Bajkal

120/3

2%

otoиek Zanzibar

47/ 0

0%

planeta Zemlja

539/ 537

50%

zvezda Sirius

25/ 0

0%

reka Nil

248/ 0

0%

--

--

--

Croatian

grad Moskva

3/ 100

97%

rijeka Volga(c)

0/ 69

100%

planina Sinaj

22 /4

15%

pustinja Sahara

14/ 49

77%

jezero Baikal

4/0

0%

otok Zanzibar

0/ 11

100%

planeta Zemla

110/ <2000 + obl

95%

--

--

--

rijeka Dunav

344/ 71

17%

planina Uиka(c)

22/82

79%

Why cities and rivers are in most cases different from the other contexts may seem puzzling. What is distinct about these types of expressions seems to be their higher frequency in the discourse. It seems reasonable, that cities and rivers are more common to be seen in the landscape of the Western Europe (where the majority of Slavic languages are spoken) than mountains, islands, deserts and lakes. Why this should be connected to the facts of agreement is not clear from the first sight, but as I will show in what follows frequency comes out to be the factor of a primary importance in relation to the discussed problems.

In what follows only the data on case agreement in appositions with the noun `city' is considered. This decision is conditioned by the following considerations. First, it is usually the field of the greatest variation in the agreement (as can be seen from the Table 4). Second, city names are abound in all the Slavic languages and represent all the possible variation in their grammatical characteristics (as opposed to other types of toponyms that are less frequent and less diverse). To discover the factors that can be significant for the agreement in appositions, I followed the same procedure as in (Logvinova 2018) for the Russian data. In short, for each language, the dataset was created containing the information on the relative frequency of each particular city name The information about the relative frequency is presented in the number of usages of the selected proper lexeme per million words in the corpus. The conventionalized name for this measurement is ipm, which stands for items per million. Since not all the corpora in TenTen family allow to search for lemmas, in some cases a more complex CQL-query was necessary to find all the forms of a particular name in the corpus. Generally, the query in this case had a form similar to the following: [word= "Донецьк" | word="Донецьк." | word="Донецьк.."], where | is used to search for alternative conditions in the same query and [.] stands for any symbol. Accordingly, the query given above will find all the possible forms of the Ukrainian city name Донецьк. Note that this query does not prevent from receiving in the search results a form Донецький, but I realized this mistake too late to make the required adjustment for every entry in the sample. At the same time this inaccuracy can be tolerated based on the two considerations: (1) such forms occur very rarely in comparison to those needed and (2) in cases like that the frequency of the derivates can as well serve as a reflection of the familiarity of this proper name. , its relevant grammatical features (such as grammatical gender and inherent number Inherent number of the noun is its grammatical number that is not conditioned by the context. For example, the inherent number of the toponym Moskva `Moscow' is singular, whereas the inherent number of the name C?eboksary `Cheboksary' is plural which can be figured out based on agreement of adjectives: krasiv-aja_(F.SG.NOM) Moskva and krasiv-ye_(PL.NOM) C?eboksary.) and the statistics about how often this name showed case agreement with the preceding common noun in an appositive construction. In contrast to (Logvinova 2018) this time, I also controlled for the factor of grammatical case, making different samples for each of the four cases considered. I will only discuss the data on Genitive, Dative, Instrumental and Locative cases. The Accusative is left aside because there are known problems in processing examples in Accusative as in masculine inanimate and neuter nouns the Accusative form is indistinguishable from the Nominative in some Slavic languages, so that the analysis of a considerable bulk of corpus data without manually filtering is impossible.

Further I discuss the data grouped within the genetic branch. Although this decision does not appear fully satisfying, there are no other grounds to establish any other groupings.

1. Ukrainian and Belarussian

In Ukrainian, the word for `city' is misto, which is neuter in grammatical gender. The dataset representing the Ukrainian sample is given in Table 1 in the Appendix 2. The result of applying a multiple regression model to the Ukrainian data predictably revealed no significance for the factor of feminine grammatical gender (i.e. there is no difference in how feminine and masculine names tend to behave when used in apposition to the sortal term misto). However, the result for neuter names is more unexpected, with a strong negative correlation (E= -39.6340) of this feature with the percentage of agreeing forms. This result can lead to an erroneous conclusion about gender congruency being unimportant in Ukrainian. Such a misinterpretation should be avoided, based on the two following facts. First, neuter city names are much less frequent than masculine or feminine (see Table 1 in Appendix 2), which results in the shortage of data for proper comparison. Second, in other types of appositions, for example, with the noun riиka `river' Ukrainian, just as Russian (Logvinova 2018), demonstrates a strong tendency for the agreement of feminine proper names from 1^st declinational class and the preservation of the nominative form for masculine proper names.

Table 6

The agreement of proper names with the sortal term riиka in the Genitive case in Ukrainian*

Name	Grammatical features	NOM	GEN	SUM	% of the agreed forms
Прут	masc	747	9	756	1
Дністер	masc	875	33	908	4
Дунай	masc	405	25	430	6
Рось	fem, 3rd Declination	607	40	647	6
Устя	fem, 1st Declination	184	40	224	18
Десна	fem, 1st Declination	264	106	370	29
Бистриця	fem, 1st Declination	165	106	271	39
Ворскла	fem, 1st Declination	195	119	314	38
Синюха	fem, 1st Declination	63	94	157	60
Амазонка	fem, 1st Declination	36	131	167	78

*The significance of the difference between groups is checked with the help of the t-test for independent samples (p=0,0039)

As in the case of appositions in Russian, the inherent plurality of the name turned out to be a factor hindering agreement (with the impact value being the highest among the factors across the measurements) and the frequency of the proper name was also confirmed significant.

With respect to the differences observed between cases, it appeared that the agreement in Genitive was on average more frequent (see Table 7). However, the pairwise comparison using t-test for dependent samples showed that although there was a difference in how often city names agree with the preceding sortal noun between cases, this difference is rather weak with the strongest contrast between Genitive and Instrumental.



Table 7

The mean and the median frequency of agreement in Ukrainian depending on case

Type of frequency	GEN	DAT	INS	LOC
mean frequency of agreement cases	62 %	51 %	42 %	49 %
median frequency of agreement cases	72 %	58 %	43 %	56 %

The Belarussian word for city is gorad, which is masculine similar to its Russian cognate. The results for the Belarussian dataset (Table 2 in Appendix 2) did not reveal any significance for any of the tested factors except for the factor of inherent plurality (agreement is hindered when the proper name considered is inherently plural as in case of Gorki_(pl)). This result is explained by the scarcity of data, which itself is due to the limitations of the Belarussian corpus The TenTen corpus (?64 million words) is not the only one electronic corpus for Belarussian. Another big corpus is Беларускі N-корпус (https://bnkorpus.info) (?163 million of words). But working with this corpus is difficult because there is no possibility to download the results. when compared to other corpora used. As Table 2 in the Appendix clearly illustrates, there were problems in retrieving information on particular names. For that reason, no justified comparison between cases is possible.

2. Czech and Polish

In Polish and Czech, the word for `сity' is miasto and mмsto respectively, both having the neuter grammatical gender. In both languages, the congruent gender feature on proper names (i.e. neuter proper names) positively correlated with agreement, but only in Czech, the correlation proved statistically significant (p=0.01660). Also in Czech, both frequency and inherent plurality of the proper noun influenced the percentage of cases with agreement (p < 0,05 in both cases), with the restriction that the estimated impact of the frequency parameter was rather low when compared to the others (E Here E stands for an Estimate value. = 0.06830). In the case of Polish, only the impact of inherent plurality was confirmed when tested with the MLR (E = - 61.209742, p=2.45e-09). As before with Belarussian, such an outcome appears to be a result of data sampling to a large extent.

What is remarkable about the Czech and Polish data is that in both samples there was a similar discrepancy in how frequent the case agreement was among the different cases. Namely, the percentage of agreeing forms in Genitive on average was significantly higher than in each of the three remaining cases. As an illustration, consider Table 7 indicating the data on the percentage of agreeing forms in Genitive, Dative, Instrumental and Locative in Czech and Polish only for masculine city names of comparable frequency.

Table 8

The percentage of case-marked forms in different cases for the selected city names in Czech and Polish

Czech	Polish
Name	Freq., ipm	GEN The figures in the columns named GEN, DAT, INS and LOC give the rounded result of computing the percentage of agreeing forms in appositions with a given name in the given case.	DAT	INS	LOC	Name	Freq., ipm	GEN	DAT	INS	LOC
Bohumнn	3,51	48	24	36	17	Sіupsk	9,92	78	0	16	48
Chotмboш	3,48	43	13	6	24	Gniezn	9,9	92	33	55	6
Tachov	2,86	47	26	23	11	Wіocіawek	9,72	48	0	0	0
Tachov	2,86	47	26	23	0	Zamoњж	9,15	49	0	0	3
Bechynм	2,3	0	38	17	52	Przemyњl	8,95	89	0	63	65
Dobшнљ	2,06	62	10	1	34	Koіobrzeg	8,55	15	8	13	14
Љternberk	2,05	43	31	15	13	Tczew	6,79	87	0	26	65
Ћamberk	1,88	46	15	14	13	Suwaіki	6,65	26	0	0	4
Љenov	1,62	27	0	8	0	Gіogуw	4,98	68	0	0	0
Slatiтany	1,38	8	0	0	0	Bкdzin	3,98	72	50	22	5
Volary	1,26	12	0	0	8	Lкbork	3,09	88	0	33	33
Mean		35	17	13	16			65	8	21	22
Median		43	15	14	13			72	0	16	6

As descriptive statistics at the bottom of Table 8 shows, the mean value for the percentage of agreed forms in both Czech and Polish samples was higher in Genitive than in other cases. This observation is proved statistically significant for the whole samples by pairwise comparison with the t-test for the dependent samples. In both languages, only Genitive showed statistically significant deviation from all the other samples, while the result for the other pairs was under the adopted significance level.

As there is an observable discrepancy between the patterns for agreement among cases in Czech and Polish, it is especially important to know how the agreement is realized in the Accusative. As has been already mentioned, there are certain difficulties in working with accusative contexts as the forms used in the Accusative are similar to those used in the Nominative for masculine inanimate and neuter nouns. However, there remains a possibility to check this type of context by working only with feminine proper names, for which the Accusative marking is always different from the Nominative in certain groups of nouns. To ensure the context is accusative only the appositive phrases after certain prepositions were considered: pro `about' for Czech and przez `through' in Polish, both governing only the Accusative. The results are shown in Table 9 below. The information about the percentage of the agreed forms in Genitive is provided for the purposes of comparison. As can be seen from the table, the agreement in Accusative was in all cases less frequent than the agreement in Genitive.

Table 9

Case agreement between feminine city names and a sortal noun in Accusative in Czech and Polish*

Czech	Polish
City name	NOM	ACC	SUM	% marked	% marked in GEN	City name	NOM	ACC	SUM	% marked	% marked in GEN
pro mмsto… `about the city...'	przez miasto… `through the city...'
Prah(a/u)	167	55	222	25	96	Warszaw(a/к)	32	14	46	30	60
Ostrav(a/u)	155	77	232	33	89	Czкstochow(a/к)	15	4	19	21	79
Jihlav(a/u)	57	24	81	30	83	Gdyn(a/к)	22	5	27	19	69
Opav(a/u)	40	17	57	30	86	Јomї(a/к)	33	0	33	0	61
Poliиk(a/u)	23	20	43	47	84	Piі(a/к)	5	1	6	17	84
Bнlin(a/u)	8	3	11	27	54
Mean				27%	82%					17%	71%
Median				30	85					19	69

*The difference between the samples for Genitive and Accusative in both cases proved to be statistically significant by applying the t-test for dependent samples

This result is partially similar to the data for Russian concerning which I previously argued (Logvinova 2018) that both Genitive and Locative were cases conducive to case agreement between names in apposition. One of the possible explanations for the observed asymmetry is frequency. In all of the considered samples (including datasets from other Slavic), Genitive was the most frequent case, i.e. there were more appositive structures of the type `city X' in Genitive than in any other case (excluding Nominative and Accusative, for which no data was collected). This is, however, not a special property of appositional constructions as Genitive is argued to be the most frequently used case after Nominative at least in some Slavic languages (see (Kopotev 2008: 146) for Russian and (Laskovsky 1989: 213) for Polish). The fact that Genitive being most frequent with appositions, allows more case agreement than other cases is appealing as it shows that the agreement is not only conditioned by the frequency of the particular city name, but also by the frequency of a construction itself.

3. Croatian and Slovenian

In Croatian, as well as in Slovenian, the case endings for Dative and Locative are all similar through the paradigms (except for minor exceptions), that is why the collecting data on those languages required hand sorting of the corpus data. In Croatian, the basic noun for `city' is grad which is masculine in gender, while for Slovenian it is mesto and it is neuter as in Ukrainian, Polish, and Czech. Slovenian appears to be different from the other languages under discussion in generally disfavoring the agreement between components of an appositive structure with the sortal noun for `city'. In contrast, in Croatian, the agreement is the preferable strategy, as can be seen in Table 10 below:

Table 10

The mean and the medial frequency of agreement in the four cases in Croatian and Slovenian

	type of frequency	GEN	DAT	INS	LOC
Croatian	mean frequency, %	96	96	90	93
	median frequency, %	99	100	100	100
Slovenian	mean frequency, %	19	6	8	5
	median frequency, %	9	0	0	0

Even inherently plural names, which showed strong resistance to the agreement in all the languages discussed above are not different in their propensity for agreement from inherently singular names in Croatian. Applying the MLR to the Croatian dataset reported no significance for any of the alleged independent variables. The same result was obtained for the Slovenian data. As can be judged from the Table 10, in Slovenian, just as in all the languages discussed above, agreement in Genitive was more frequent than in any other case,

4.3.3 Discussion

The fact that Genitive allowed more agreement than any other case in almost all the languages considered gives rise to the question if it is possible to provide a hierarchy of cases allowing more or less case agreement in appositions in Slavic. The fact that there was no statistical difference between other cases in each language taken separately can be disregarded at this point provided that in each of the six languages the difference between the cases is the same. The Table 11 below shows that this is not true, as cases appear to be ranged differently in different languages, so that no clear cross-language hierarchy is possible to the present moment.

Table 11

The comparison of mean and median frequency of agreement of proper names in apposition in four cases in the six languages studied

Language	type of frequency	Case
		GEN	DAT	INS	LOC
Ukrainian	mean, %	62 (1)	51 (2)	42 (4)	49 (3)
	median, %	72 (1)	58 (2)	43 (4)	56 (3)
Belarussian	mean, %	75 (2)	62 (3)	81 (1)	56 (4)
	median, %	93 (3)	95 (2)	100 (1)	69 (4)
Polish	mean, %	55 (1)	14 (4)	19 (3)	23 (2)
	median, %	61 (1)	0 (4)	9 (3)	10 (2)
Czech	mean, %	50 (1)	23 (2)	20 (4)	21 (3)
	median, %	48 (1)	21 (2)	15 (4)	18 (3)
Slovenian	mean, %	19 (1)	6 (3)	8 (2)	5 (4)
	median, %	9 (1)	0	0	0
Сroatian	mean, %	96 (1-2)	96 (1-2)	90 (4)	93 (3)
	median, %	99 (4)	100	100	100

The main conclusions that can be made from this Section are the following:

- almost for all the languages considered the factor of inherent plurality of the proper name was a factor hindering agreement in apposition;

– in all the languages the factor of gender congruency between the components in apposition and higher frequency of the proper name were proven to positively correlate with the percentage of the agreeing forms, but not in all the cases this correlation was proved to be statistically significant;

– in both Czech and Polish the agreement pattern in the Genitive case was considerably different from any of the four remaining cases, with appositions in genitive allowing agreement significantly more frequently -- this difference was also present in some other languages, but didn't reveal statistical significance;

– Slovenian and Croatian were different from the other languages in the sample as well as opposed to each other in their persistent propensity for hindering and preferring agreement respectively.



4.4 Conclusions

In this section the possible patterns in the distribution of case markers within the appositive constructions were observed both on the typological scale and within the Slavic stock. The main conclusions that can be drawn from the material discussed in this chapter can be summarized in the following statements:

1. Nouns in apposition are in most cases treated as forming a constituent of a level of NP, as in languages lacking case agreement in NP generally require the nouns in apposition to have a common case marker.

2. There is some evidence that appositions can have a syntactic scope smaller than a typical NP, as they do not allow introduction of new modifiers inside of the appositive structure without the break of constituency; on the contrary, appositions sometimes allow to include the whole coordinate structures inside them, which is an evidence of them being “bigger” than the typical NPs.

3. In languages allowing case agreement of members of apposition, it can be irregular and dependent on a number of factors. In Slavic languages, considered in this section, these factors appear to be frequency and congruency in such grammatical features as number and gender.



5. Order of constituents in constructions of specification in a typological perspective

This chapter considers the problem of linear arrangement of a common noun and a proper name in appositive constructions in the languages of the sample. Conclusions that will be made in this section partially reproduce the results already obtained in (Logvinova 2019), but are based on a bigger sample.