In a few works containing a comparatively deep study of the apposition, one of the most discussed questions is of determining the direction of syntactic dependence in such expressions. In other words, which of the constituents has the status of a syntactic head and which has the status of a dependent and whether attempts of hierarchization in the constructions of this sort are at all justified. Below I discuss a few observations regarding approaches to the analysis of the internal syntax of appositive constructions in the languages of the sample.

As I have already noted in (Logvinova 2019a) in some cases authors use the criteria of morphosyntactic locus to determine the status of the members of an appositive phrase. In a grammar of German (Duden 2009: 988) the following arguments are presented in favor of the fact that the head of the appositive constructions is a common noun:

§ the article and the anaphoric pronoun agree in number and gender with the common noun (die_f Stadt_f Rom_n `the city of Rome', cf. das<sub>n</sub> schöne Rom<sub>n</sub> `beautiful Rome' and ich liebe die Stadt_f Rom_n, weil sie_f (*es_n) <...>, literally `I love the city of Rome, because she (* it)... `);

§ the common name is the morphosyntactic locus of the group, since only it is marked by case when the syntactic role changes:

(24) German (< Germanic < Indo-European, (Duden 2009: 988), glosses are mine)

die Scherze mein-es alt-en Freund-es Ferdinand

art joke 1sg.poss-gen old-gen friend-gen Ferdianad

`A joke of my old friend Ferdinand'

The same criteria are considered in I. Markoviæ's analysis of the apposition in Croatian: nominal predicates and anaphoric pronouns in Croatian, as well as in German, agree in gender with the common noun, rather than proper names:

(25) Croatian (< Slavic < Indo-European, (Markoviæ 2008: 130, glosses are mine))

a. Rijeka Dunav duz?a/* duz?i je od rijek-e Save

river(f) Danube(m) bigger(f) / bigger(m) is from river-gen Ñàâà

`The river Danube is bigger than the river Sava'

b. Knez?evina Lihtens?tain dala je velik-e skijas?-e.

principality(F) Liechtenstein give.pst(f) is great-pl skier-pl

U njoj /?njemy(m) se rodil-a H. Wenzel.

â she.loc he.loc herself give_birth.pst-f Õ. Venzel

`Liechtenstein gave a lot of great skiers. That is where H. Venzel was born'.

Szõke (Szõke 2017) argues that in Hungarian appositives the element that has the role of the syntactic head is the common noun. She establishes her argumentation on the following example:

(26) Hungarian ((Szõke 2017: 169), glosses are mine)

A Bront¸ nõvér-ek fiatalon haltak meg.

def Brontë sister-pl young die;pst;3pl as_well

`Sisters Bront¸ died young'.

In (26) only the common noun in apposition is marked for plural (nõvérek), while the proper name (Bront¸) bears no marking. The verb (haltak) shows the agreement with the third person plural subject.



3.2 Attributive strategy

The second most common strategy in the sample is subordinate or attributive. With this strategy, one of the members of the construction of specification gets the distribution of an adnominal modifier and the other one serves as the head of the construction. Following (Grašèenkov 2018: 331), attribution is understood as a “process by which a certain element acquires the ability to act as a modifier in a nominal phrase. In this strategy one part of the construction of specification is structurally dependent on another.

Attributive strategy comes in two types:

I. morphosyntactic means used in a construction of specification are the same as in a possessive construction;

II. morphosyntactic means used in a construction of specification are different from those used in a possessive construction but follow some other technique of attributivization used in that language.

I will first discuss the first type and then comment briefly on the second one.

3.2.1 Attributivization parallel to possessive constructions

As has been previously shown in (Logvinova 2019a), in four Indo-European languages from the sample (namely -- Irish, Albanian, Latvian and Finnish) constructions of specification with toponyms (and some other types of names) are structurally parallel to genitival expressions. All the named languages use dependent marking in possessive constructions (in the sense of (Nichols 1986)). This means that the marker of syntactic dependency (genitival marker in this case) is placed on the dependent constituent in these languages. The following examples illustrate that the dependent role is ascribed to proper names in such construction, as they receive the same marking as prototypical genitive dependents:

(27) Irish (< Celtic < Indo-European, [Doyle 2001:163])

a. teach Mhaìire

hous Marie\gen

`house of Marie'

b. (examples provided by B. An Píobaire, personal communication, the glosses are mine)

habhainn Shionna In (23) a, b and c the bold font marks the lenition of the first consonant.

river Cynon\gen

`Cynon river', lit. `river of Cynon'

c. sliabh Chorrán₁ Tuathail₂

mountain Carrauntoohil_1,2\gen

`Carrauntoohil mountain', lit. `mountain of Carrauntoohil'

d. pláinéad Mars/ *Mhars

planet Mars.nom *Mars\gen

`planet Mars'

(28) Latvian (< Baltic < Indo-European (Mathiassen 1997: 166-167))

a. ma?sa-s gra?mata

sister-gen book

`sisters book'

b. Daugava-s upe

Daugava-gen river

`the river Daugava'

(29) Albanian (< Indo-European, (Newmark et al. 1982: 136)

a. djal-i i Agimit

son-def art In Albanian article is a part of the genitive form. Agimit

`Agimit's son'

b. lim-i i Devollit

river-def art Devoll

`river of Devoll'

c. qytet-i i Vlorës

city-def art Vlore

`the city of Vlore'

d. muaj-i i Korrikut

month-def art July

`the month of July'

(30) Finnish (<Finnic < Uralic, (Karlsson 1999: 95), NRC)

a. Auli-n auto on keltainen.

Auli-gen auto is yellow

`Auli's car is yellow'

b. Viipuri-n kaupungi-ssa

Vyborg-gen city-inn

`in the city of Vyborg'

(apud Logvinova 2019)

Nubian (Africa) is also said to be a dependent-marking language. Again, as in the examples above, the proper name in (31) is marked in the same way as genitival dependent and, thus, has the status of a formal possessor.

(31) Nubian (< Nubian < Eastern Sudanic, (Abdel-Hafiz 1988: 94; Bell 2015))

a. berti-n ur

goat-gen head

`the goat's head'

b. Aìbiri-n muuleìe

Abir-gen mountain

`the mountain of Abir'

All the other languages predominantly using subordinate strategy in constructions of specification with toponyms are head-marking languages. This means that a noun modified by another noun bears the marker indicating its governing status. Among such languages are Chuvash (32), Arabic (33), Berber (34), Lamang (35) and Hausa (36). In the languages of this group, the marker found in possessive constructions is not in fact the marker of possession (as is Genitive in many European languages), but rather an indicator of structural dependency of one nominal constituent on another. This type of (head-) marking is very often called izafe (especially in connection to Iranian and Turkic languages) or construct state (in relation to Semitic languages) (Plungjan 2010: 200-2004), but is argued to be characteristic of the majority of African languages (Creissels 2009). The distribution of the markers of this class, as well as the range of semantic relations coded by them is different from language to language. Any comparative analysis of the markers of this sort is, of course, beyond the scope of the present paper (and appears to be hardly undertaken anywhere else). The core function of the markers of this type is in many cases to code possession, but not in any way only possession. What is important for this discussion, is that the presented languages make clear difference between dependent and governing constituents the the constructions of specification.

(32) Chuvash (<Turkic < Altaic, personal field notes)

a. poskil-ën ën-i tar-za

neighbor-gen cow-p_3 go-cv_sim

`neighbor cow ran away'

b. v?l var?av? xol-i ?om-?e për jal-da

3sg Warsaw city-p_3 near-loc one village-loc

?oral-n?

be_born-pc_pst

`She was born in the city of Warsaw'.

(33) Arabic (Arabic < Semitic < Afro-siatic, (Ryding 2005: 206), R. Mamedšaxov, personal communication)

a. 'ab-uu Hasan-i-n

father-nom Hasan-gen-def

`the father of Hasan'

b. ?-madiinat-u l-qaahairat-i

izf-city.s.f-nom def-Cairo.s.f-gen

`the city of Cairo'

(34) Berber (Middle Atlas) (< Berber < Afro-Asiatic (Putten 2013: 142, 420)

a. ?ali? siìmihò d=iwi?-s n=ammi

name kid pred.=son-3s of=uncle

`(I want)' kind Ali, he is the son of my uncle'.

b. gan iwiìnan n=am?Ìd?n as?aÌl n=awil?n

there one:ms of=man village of=Aujilan

`There one was a man in the village of Aujila'.

(35) Lamang (Lamang < Lamang (Wolff 2015a: 77; 270; Wollf 2015b: 541), glosses are mine)

a. oìgaìa daÌdaÌ

goat;ass father

`the father's goat'

b. lw-aìa GwoÌzoÌ

town-ass Gwoza

`the town of Gwoza'

(36) Hausa On the -n (called `genitive affix or genitive linker' and having an allomorph -r) affix in Hausa it is noted in (Creissels 2009: 5-6) that it cannot be undoubtedly considered to be a marker of syntactic headness as in languages with `canonical' instances of `construct state', i.e., in languages like Arabic (see example (33)), as it also appears on attributive constituents in constructions like fari-n kàree `white dog' (where fari is a masculine form of the adjective `white'), where the constituent that gets the marker is not the semantical head of the construction. A possible solution proposed by D. Creissels is to consider -n (and its allomorphs), appearing on attributive constituents to be another class of suffixes coinciding in form with the `genitive linker'. (< West Chadic < Afro-Asiatic (Jaggar 2001: 332), Sketch Engine/hausaWaC15)

a. ku?i-n AuduÌ

money-poss Audu

`Audu's money'

b. gari-n Kaduna

city-poss Kaduna

`the city of Kaduna'

Sango (37), Swahili (38) and Gude (39) use separate words to indicate attributivization. In (Nichols, Bickel 2013) this type of marking is considered to be head-marking as well.

(37) Sango (Ubangi < Niger-Congo, (Samarin 1963: 44, 170))

a. da tiì zo soì

house poss person this

`The house of this man'

b. fadeì k?ìt?ìr?ì tiì Bangui ak? k?n? aìnde

fast city poss Bangui it be_big later

`The city of Bangui will be big in the future'.

(38) Swahili (< Bantoid < Niger-Congo, Sketch Engine/ SwahiliWaC, glosses are mine)

a. mke wa Rais

wife cm.poss Rais

`the wife of Rais'

b. mji wa Dresden

city cm.poss Dresden

`the city of Dresden'

c. visiwa vya Zanzibar

island cm.poss Zanzibar

`the islanf of Zanzibar'

(39) Gude (< Biu-Mandara, < Afro-Asiatic, (Hoskison 1983: 63), Bible in Gude, glosses are mine)

a. min? nga Musa

wife poss Musa

`Musa's wife'

b. v?ran? ?ga Afisa

village pos Efes

`village of Efes'

In the three examples above a possessum is followed by a possessive particle and comes before its possessor. The order of the constituents in a construction of specification in those languages is such that a proper name appears after the possessive particle, and, thus, is in structural position of a formal “possessor”.

In languages like Iraqw (40) and Mano (41) when nouns are modified by other nouns (or adjectives and relative clauses in case of Iraqw) the head of the constructions receives the low tone. The examples below show that there is the same mechanism is involved in constructions of specification:

(40) Iraqw (< Southern Cushitic < Afro-Asiatic, (Mouse 1995: 195, 203))

a. garmoìo niinaì

boy:m:con small:m

`a small boy'

b. tloomaì-r kwaraì

mountain:con-f Kwara

`Moint Kwara'

(41) Mano (< Eastern Mande < Mande, (Khachaturyan 2015: 45), M. Khachaturyan, personal

communication, glosses by M. Khachaturyan)

a. gí lèè (the free form is lçç)

belly woman:cs

`pregnant woman'

b. NaìzaÌl?Ì-p?Ìl?Ì (the free form is p?ìl?ì)

Nazareth-place:cf

`the city of Nazareth'

c. Z?ìdaÌ? yiÌiÌ (the free form is yiìiì)

Jordan river:cs

`the Jordan river'

In (Logvinova 2019) I showed that a remarkable detail about the attributive coding is that it is distributed unevenly between the different contexts within one language and among different languages. Namely, the attributive strategy appears to be impossible in some contexts where it is replaced by juxtaposition. I introduced the matrix, which I present below with some corrections and additions. The problem of introducing all the considered above languages in the matrix is the impossibility to find the needed contexts for the majority of them.

Table 2

Distribution of the attributive and juxtapositive strategies in the most common specification contexts in nineteen languages The table should be read as follows. [A] means that in this language this context is encoded attributively. By contrast, [J] shows that the particular type of the construction reveals juxtapositive strategy. The [A / J] mark is used if some variation is observed. The symbol [c] is used to indicate composites: [c (A)] and [c (J)] are used for cases where one of the parts of the composite has attributive form and for cases with a simple concatenation of the bases, respectively. The designation [diff.], found in only one cell, means that, unlike all other contexts, in this case, Irish uses a different type of the possessive construction. Gray cells mean that the respective information is lacking.

Language Languages given in the italic font were already presented in (Logvinova 2019).

month The names of columns give the first element in the construction of specification. That is, in the first column constructions like the month X are considered.

city

mount

state

galaxy

river

island

festival

planet

book

`mister'

other animate common nouns

Pashto

A

A

A

A

A

A

A

A

A

A

J

J

Arabic

A

A

A

A

A

A

A

A

A

A

J

J

Hebrew

A

J

A

A

A

A

J

A

A

J

J

J

Swahili

J

A

A

J

A

J/A

A

A

J

J

J

Chuvash

A

A

A

A

A

A

A

A

A

J

J

Hausa

A

A

A

J

A

J/A

J

J

J

Finnish

c(À)

A

A

A

A

c(J)/J

A

c(A)

A

J

J

J

Estonian

c(A)

A

A

A

A

A

A

c(A)

J

J

J

J

Latvian

A

A

A

A

A

A

A

J

J

J

A

J

Irish

A

A

A

A

A

A

A

diff.

J

J

J

J

French

A

A

J/A

A

A

J

A

A

J

J

J

J

Albanian

A

A

A

A

J

A

A

A

J

J

J

J

Basque

A/J

A/J

J

J

A/J

A/J

A

J

J

J

J

English

A

A

J

A

J

J

J

J

J

J

J

J

Russian

J

J

J

J

A/J

J

J

A/J

J

J

J

J

Mano

A

A

A

A

J

J

J

Logoti

A

A/J

A/J

J

Berber

A

A

A

A

J

J

Gude

A

A

A

A

J

Information available

14

19

19

15

13

19

15

17

15

15

15

18

Attributive strategy

12

16

14

11

9

13

10

10

5

2

1

0

85%

84%

74%

73%

69%

68%

66%

58%

33%

13%

7%

0%

more attributive less attributive

The languages in which the attributive strategy is used in the most contexts are at the top of the table (although note that for the languages in the very bottom, information on the number of contexts is unknown). The “most attributive contexts” are found to the left. As can be seen the Table 2, in the majority of languages considered, attributive strategy is not used in constructions of specification with personal names of humans. This is universally true about all the studied languages that reveal attributive strategy of the first type. The two exceptions are Latvian and Persian, which exploit attributive coding in constructions with appellatives, such as English mister.

The fact that possessive-like coding is avoided with animate referents is not unexpected. The motivation for this is most probably to escape ambiguity. If a hypothetical language called “anti-English” had the same means to express possession and specification with animate referents, both expressions would have form like a brother of Tom/ Tom's brother or the professor of Johnson/ Johnson's professor. Animate referents (and especially entities referred by proper names) are believed to be most prototypical possessors (Koptevskaa-Tamm 2002: 148, Lander 2004: 311), since they are salient in in the discourse and therefore serve as better candidates to establish the reference of the possessum. Due to this fact constructions given above are most naturally interpreted as possessive. Since clearly inanimate referents, such as months and cities (and, probably, their names), are less expected to be “true” possessors, the ambiguity is highly unlikely and the restriction eliminates.

Also in (Logvinova 2019) I showed that in addition to the differences observed in how different languages use attributive strategy and what distinctions can be observed in one language in synchrony, the change from attributive to juxtapositive syntax can be observed in the diachronic dimension. Thus, Dobrovie-Sorin (2013: 798) mentiones that genitive construction was quite common in Romanian in constructions of specification with toponyms before, but now is peripheral and regarded as old-fashioned. The same is true about Chechen (Yakovlev 1940: 122-123), for which the appearance of juxtaposition is said to be contact-induced phenomena resulting from the interferation with Russian. The same observation comes from my own field work. In the texts collected from older speakers of Urmi dialect of Northeastern New-Aramaic one finds the constructions of specification built with the help of a `possessive' clitic t- is used, while the same type of expression is elicited from a younger speaker (about 20 years old), juxtaposition appears to be the only option:

(42) Urmi dialect of New Aramaic (< Semitic < Afro-Asiatic, M. Ovsjannikova, personal communication)

a. y?ìmma... ?t d-a braìta

mother rel obl-dem1.sg girl

`the mother of this girl'

b. guv maìt-?t aìrz?n, m?n mdiìt-?t uìrmiya, m?n

in village-rel Arzni from city-rel Urmia from

gu iraìk

in Iraq

`In the village of Arzni, from the city of Urmi, from Iraq'.

3.2.2 Attributive strategy different from that found in possessive constructions

The two languages that reveal attributive strategy but use morphosyntactic means different from those found in possessive constructions are Georgian and Kannada. In Georgian nouns in preposition to proper names in constructions of specification behave like adjectives in attributive position, namely, nouns ending in vowels show no case endings and consonant-final words use a reduced set of case endings:

(43) Georgian (< Kartvelian, (Aronson 19991: 22), glosses are mine)

a. gazet-i K'omunist'-i b. mdinare Rion-i

paper-nom Communist-nom river Rion-nom

`the newspaper “The Communist”' `the river Rion'

gazet-? Komunist'-s mdinare Rion-s

paper Comminist-dat river Rion-dat

`to the newspaper “The communist”' `to the river Rion'

In Kannada the attributive strategy applied in the construction of specification is parallel to the one used to derive adjectival-like modifiers from nouns (44):

(44) Kannada (< Dravidian < Indo-European, (Jensen 1969: 50, 242), glosses are mine)

a. agala-v-âda < agale

wide-?-be.pc `wideness'

b. arasan-âda hero?danu

king-be.pc Herod

`the king Herod'

However, according to a native speaker (Prathap Parashar, personal communication), simple juxtaposition both in the constructions with toponyms and personal names of humans is preferable, while the structures presented in (44) should be translated literary `the one who become X':

(45) (Prathap Parashar, personal communocation)

a. sahodara Agaranige

brother Aagarnige

`the brother Agharnige'

b. Bengalooru nagara-dalli

Bangalore city-loc

`in the city of Bangalore'

The examples from Georgian and Kannada are interesting in two aspects. First, in both cases attributive strategy is found in expressions with personal names of humans. However, since the attributivization process here is different from the one observed in possessive constructions, the restriction on animate referents discussed above is unrelevant.

Second, both in Kannada (in examples like those in (44)) and Georgian the constituent that undergoes attributivization is a common noun, not a proper name as in cases discussed above. The exception to this is Basque. As I discuss in more detail in (Logvinova 2019), in this language the morpheme that can appear (but is not obligatory) in the constructions of specification is -ko, which is used to converse the postpositional groups into noun modifiers:

(46) Basque ((Villasante 1978: 140; de Rijk 2008: 89), glosses are mine)

a. Ninibe-ko hiri-a

Ninib-link city-abs

`the city of Ninib',

b. bezero-entza-ko mahai-ak

client-des-link table-PL

`the tables for clients'

As can be seen from Basque, the role of a syntactic head is occupied by a common noun.

All in all, among the languages clearly distinguishing the head and the dependent role in a construction of specification, 26 designate the head role to a common noun and 2 give the syntactic priority to a proper name. Table 3 summarises the findings on headness observed in the present section.

Table 3

The headness in attributive constructions of specification

The constituent with the role of a syntactic head	Type of attributivization
	the same as in possessive constructions	different from the possessive constructions
proper name	0	2
common noun	25	1

The striking feature about the attributive strategy is that it appears in abundance in Africa and Western Eurasia, but is not seen anywhere else. Picture 2 shows the geographical distribution of strategies used to code specification with toponyms in the languages of the sample.

Picture 2 The geographical distribution of strategies used to code specification in constructions with toponyms in the languages of the sample This map is made with the help of the R-library introduced in (Moroz 2017).

The map is to be read as follows: (i) the green dots (labeled “juxtaposition”) represent the cases with juxtaposition of constituents, (ii) the blue dots (labeled “attribution”) stand for languages that mainly exploit attributive strategy identical to the one found in possessive constructions, (iii) orange dots (labeled “attributive 2”) are used to refer to languages with the attributive strategy that are seen in the constructions of specification being different from those found with possessive expressions (Georgian and Basque, Kannada is labeled juxtapositive since this type of expression is more preferable to native speaker); (iv) red dots mark the languages with the constructions of specification being parallel to relative constructions. The dots in violet and brown represent the languages that use the so called “presentative” marker and a special morpheme found only in appositions. These cases are discussed in a more detail below.

The fact that African and West-Eurasian languages are less tolerant to simple juxtaposition of nouns and prefer attributive coding to concatenation is curious and I have no explanation for this by now.

3.3 Other strategies to express the meaning of specification

3.3.1 Constructions of specification parallel to relative clauses

In the language Movima (isolate, South America) constructions of specification are said to be parallel to relative clauses.

(47) Movima (Haude 2006: 328)

ja' iy'bikwe? to? jan-ne=`ne, kinos ay'ku di'

just pro.2pl very who-be.person=f art.f.a aynt rel

Seida to? jan-ne=`ne, jankwa=us

Celida very who-be.person=f say=m.a

`She is just so much like you all, she is so much like aunt Celida!, he said'

(48) (K. Haude, personal communication)

n-as lo:los di' San Wakin

obl-art.n lo:los rel San Joaquin

`in the village of San Joaquin', lit. `in the village which is S. J.'

(49) (Haude 2006: 572)

kus tami:ba di' majniwa=`nes ya:ni:kay

art.m.a baby rel child_of=art.f nephew,niece

`the baby who is my niece's son'

As can be seen from the examples (47)-(49), the same relativizer di' is used both in specificational constructions with proper names and in clauses relativizing on subjects. Note that in Movima the element that serves as a head of a construction is a common noun, not a proper name. It is important to point out, however, that relativization is said to be the main device of modification within NP domain in that language. That means that although there are genuine adjectives that can serve as adnominal attributes, relativization can be used to express similar meaning as well:

(50) (ibidem: 115)

yey-na is to:mi [di' pa:luy]

want-dr art.pl water rel cold

`I want cold water'

It is questionable, whether the constructions of specification in the examples (47)-(48) really have clausal-like structure, and Haude provides no independent evidences for this fact, but argues that nouns serve as predicates in these constructions. An alternative would be to say that Movima simply uses attributive strategy as the languages discussed above. What distinguishes Movima from the previously discussed cases is that the very same marker is used to convert into adnominal modifiers both nouns and verbs, which is not true about the markers occurring by the `genuine' attributive constructions.

In Japanese, constructions of specification use an attributive form of a copula da which is identical in form to a possessive particle no. Iwasaki (Iwasaki 2013: 200) gives the following examples:

(51) Japanese (Iwasaki 2013: 200)

a. taroo no sensee

(name) gen teacher

`Taro's teacher'

b. sensee-no taroo

teacher-cop:att (name)

`Taro, who is a teacher'

Examples without a comma in translation are also common in descriptions, see an example from (Keiser 2001: 27):

(52) (Keiser 2001: 27, glosses are mine)

yokozuna no musashimaru

grand_champion cop:att Mudashimaru

`the grand champion Musashimaru'

Since no is a special form of the copula This statement is not universally accepted among the Japanese grammarians, and some researchers treat no in examples like (47)b as a the possessive morpheme, as in (51)a (see references in (Sode 2004)). I am leaning towards the copulative analysis, because if one accepts possessive analysis, the expression in (47)b should be literary interpreted as `Taro of a teacher'. In this case, a proper noun is the head of the construction (and a formal possessum), while a common noun is modifier (and a formal possessor). As I have shown above, this is not a typical situation for the attributive strategy parallel to possesion. An additional argument that Sode mentions is that no can be replaced by another copulative form dearu in constructions similar to (47)b., a phrase containing it should have a clausal structure. This assumption is the starting point of Sode's paper (Sode 2004), in which she suggests to treat Japanese modifiers with no like in (51)Îøèáêà! Èñòî÷íèê ññûëêè íå íàéäåí.b as relative clauses modifying a proper name (which is the syntactic head of the construction). Her argument is based on the possibility to add an adverbial modifier inside of the phrase and the fact that constructions with no are ambiguous, since they allow both restrictive and non-restrictive reading (exactly as relative clauses in Japanese usually do (ibidem)):

(53) (Sode 2004: 195)

itum izimekko no Atuo-wa mata sensei-ni sikar-are-ta

always bully cop Atuo-top again teacher-by scold-pass-pst

`Auto, who is always a bully, was scolded by a teacher again'. This example is, of course, questionable, since the proper name has a rather evaluative semantics (resembling the `quasi adjectives' introduced by Lee (Lee 1952)), but being not acquainted with Japanese, I rely on the author's opinion. The second problem with the examples from (Sode 2004) is that the translations that the author gives always show a comma between the members of the construction, so that they all appear to be translated as loose constructions. This can in fact reflect the authors intuition about them being semantically loose appositions. What is important is that in this case they should not show phonological detachment, as in the other place (Sode 2004: 2001), the author explicitly states that the intonational break is signaled by the comma in the transcription itself, not in the translation.

alternative: `Always the bully Auto was scolded by a researcher'.

Meanwhile, the possibility to add an adverbial modifier seems not to be a good test. In (Logvinova, in press) I give examples of adverbials of this type occurring in Russian loose appositions, which are not obligatory clausal in their scope:

(54) Russian (< Slavic < Indo-European (Logvinova, in press))

<...> Arkadij Gajdar, snac?ala bespos?c?adnyj krsasnyj komandir, potom

Arkadiy Gaidar at_first ruthless red commander then

sentimental'nyj detskij pisatel's <...>

sentimental children's writer

`<...> Arkadiy Gaidar, a ruthless communist commander at the first time (and) a sentomental childrens authors then'

What Sode mentions as well, is that constructions with no generally have an implication of a causal relation:

(55) Nihonzin-no Tanaka-san-wa misosiru-ga suki da

Japanese-cop Tanaka-Mr.-top miso.soup-nom like cop

`Mr. Tanaka, a Japanese, likes miso soup'.

(implied) `Tanaka likes miso soup. Maybe it is because Tanaka is Japanese'. (ibidem: 197)

The `causal' meaning is also common to what Heringa (Heringa 2012: 16) calls absolute constructions in English, treating them as free clausal adjuncts rather than appositions Heringa (Heringa 2012: 169) himself criticizes Sedo's approach and offers to analyze Japanese appositions as absolute constructions. I disagree with this view since no one of the authors writing on apposition in Japanese mentions this “causal” implication, and constructions with no are perceived as the very natural translation of simple Russian close appositions (Y. Konuma, personal communication).:

(56) A millionaire, Pontefract showered money everywhere. (ibidem)

Although both Sedo's arguments do not look as fully convincing to me, it appears to be the most comprehensive account of constructions of specification in Japanese. The most persuasive fact for me is that no is a copulative form and, thus, the modifiers introduced by it are not inter-NP modifiers as in case of attributivization discussed in Section 3.2, but rather have clausal scope.

Sometimes Japanese seems to allow specification to be expressed through juxtaposition I do not consider classifier-like words, cooccurring with Japanese proper names to be examples of constructions of specification. The names of cities in Japanese often add the element -si `city' (Kóobe-si `city of Koobe'), the names of mountains are usually followed by -san or -yama as an indicators of the class (Asama-yama `mountain Asama'), and with the names of rivers the classifier is gawa (Tone-gawa `Tone River') (Martin 1975/1991: 1048-1058). In (Martin 1975/1991: 1048-1058) these elements are described as “geographical affixes”. At least sometimes they can be morphologically integrated in the preceding stem (Y. Konuma, personal communication). It is important at the same time that at least kawa `river' and yama `mountain' may well be used as independent nouns. Since it is not clear if these elements are independent words or classifiers, they are disregarded in this study. (that is, no is not obligatory) (Y. Konuma). Observing the difference between the constructions with no and without no, Sode mentions that the first type, but not the second, allows the contrastive reading (57), and the first noun in the constructions without no has to be definite (ishi Tanaka `Tanaka, THE doctor of doctors'):

(57) Syuto *(no) Wasinton zya-naku-te, daitooryoo *(no) Wasinton kot-o

capital cop Washington is-not-and president cop Washington thing-acc

itteru-n-desu

saying-nomin-cop

`I am talking about Washington the president, not Washington the capital city'.

It is difficult to interpret the above facts within the existing theory of apposition. Since the structures without no are less typical and are said to be the result of no deletion (Martin 2003: 1048), I regard juxtaposition to as possible, but not a core strategy in Japanese.

Therefore, for both languages that are said to use relativization to express specification in this study I cannot claim with certainty that the process involved is the genuine relativization. However the decision to distinguish these languages from the languages with attributive strategy is mainly motivated by the fact that the modifiers introduced by the elements found in the constructions of specification in these languages are said to have clausal scope.

3.2 A special marker of apposition

An example of a special marker of `apposition' that is not found in any other part of grammar comes from the language Urarina, an isolate found in South America. According to (Olawsky 2006), in this language it is possible to express specification by juxtaposition of constituents, both in constructions with personal names of humans and toponyms:

(58) Urarina (ibidem: 165-169)

t?a?e kanii ermano santjago rai be-?

also gpf brother psn for tell-imp

`Also, tell brother Santiago, [that i wasn't able to get his rifle repaired]'.

(22) leihii nese betafwahee

one town pln

`one (certain) town [called] Bethpage'

At the same time at least in some cases with personal names and widely with toponyms a special morpheme -b is used:

(59) Maria b-eene

Maria ascm-woman

`the woman Maria'

(60) egipto b-atane

pln ascm-land

`the land of Egypt'

(61) kuniniku b-atane ku-fwa?ana t?r?-i? sini-akaan?

pln ascm-land asc-harbor arrive-prt sleep-1pl/ex

`We arrived at the port of Coninico and slept'.

(62) sinai b-it?atane-?

pln ascm-mountain-loc

`at mount Sinai'

The functions of this morpheme are unclear. This prefix is reported to occur almost exclusively with nouns followed by proper names in constructions of specification (the only exception is the noun (b)arasihije `star' which can be used with or without the initial b- when it is not preceded by a proper name). This morpheme is also unproductive (although the list of nouns allowing it is not provided) and is associated with contrast between traditional and innovative usage of language. Besides nouns, it can also be used with the verb (b)asihja?a `steal'. The author comes to a conclusion that the semantic or grammatical contribution of this marker is not clear and offers to gloss it an “associative marker”.



3.3.3 Representative particle

In the Niuean language (< Oceanic < Austronesian), constructions of specification with place names (63)-(64) use a preposition ko which (according to (Massam, Smallwood 1997; Massam 2006)) also occurs in nonverbal predication with the meaning of identification, with topics, titles and free standing noun phrases.

(63) i tonga ne o: atu ai ke he motu ko Niue

loc.p Tonga pst go,pl dir res goal loc island pres Niue

`One upon a time, a long time ago, there was a group of young Tongans who arrived at Niue Island'.

(64) hifo a ia ke he maaga ia ko Alofi

go abs.art he to art village that pres Alofi

`He goes down to that village Alofi'.

(65) ko e kamuta a au

pres art carpenter abs.prop.art I

`I am a carpenter'.

(66) ko e pusi

pres art cat

“The Cat” (title)

Messam and Smallwood (Massam, Smallwood 1997) consider ko to be a preposition coding a `default' case which occurs on NPs that are not in a thematic relation with a verb or an event -- that is: not connected to the external syntax of the phrase in any way. That is remarkable in comparison to the cases of attributive coding, which involves subordination of one constituent to the other.

3.3.4 Unclear cases

In Luganda a particle used in constructions of specification with toponyms is -e, which is different from a possessive particle -a:

(67) Luganda (Bantoid < Niger-Congo, (Kamoga, Stevick 1935))

mu kibu?ga ku-'e Ho?ima

in town cm-pr This gloss is stands for `prominence'. Hoima

`(I live) in the city of Hoima'

In (67) -e is attached to the class-marker ku, which agrees with the head of the construction (kibuga). The main function of -e is to “bring a noun or a pronoun in prominence”:

(68) (Ashton et al. 1954: 21)

Mukasa yE muzimbi

`Mukasa is the builder'

However, there are other functions of this very particle. Among them is a function of relativization on the direct object (69) and “emphasizing” in predicative phrases that makes “the subject of the predication to be more closely connected” to the NP in predicative position:

(69) (ibidem: 240)

Ekitabo kuE nguze kiri wa?

`Where is the book I've bought?'

(70) a. (kino) kitabo b. (kino) kyE kitabo

(This) (is) a book (This) it is the book

Two problems arise as one tries to give an analysis to the facts observed in Luganda. The first is evidently lack of detailed descriptions for the functions of -e. The second (although resulting from the first) is that all the three mentioned “functions” can hold in constructions like the one given in (67). If -e in (67) is a genuine relativizer, Luganda should be similar to Movima (see example (47)). If -e is a kind of presentative particle, then it appears to be close to Niuean.

In Ma'di one finds the n?? particle between the members of a construction of specification.

(71) Ma'di (< Moru-Ma'di < Central Sudanic, (Blackings, Fabb 2003: 672))

iÌtoì p?? a?druìpiÌ Ì n?? t?aìraìguÌleì tr?Ì

Ito prpl brother-spec pr Caragule comt

`Ito (Hare or Rabbit) and his brother Caragule'.

According to (Blackings, Fabb 2003: 272, 295) there are two types of n?? in Ma'di. The first n?? is found in possessive constructions and the second one is called “grammatical” in the grammar and appears in non-verbal prediction. An example in (72) shows the contrast between the possessive and the grammatical n??. Notice that ?ìp?ì can be either a proper name (Opi) or a common noun denoting `chief'. The “grammatical” n?? can only be used after names and other definite expressions. By contrast, the “possessive” n?? can only follow an indefinite. ?ìp?ì in (72) is not followed by the low tone suffix, which is a specific definite determiner in the language, so it is either the proper name, or the indefinite NP meaning `chief'.

(72) (ibidem: 272)

?ìp?ì n?? ?a?raì

chief pr child

`a chief's child'

`Opi is a child'

In (71) ni? is certainly not the possessive marker, as it is placed after the definite constituent. The only remained option is to consider n?? in (71) to be the grammatical n?? found after subjects in non-verbal predications. (Blackings, Fabb 2003) do not consider the grammatical n?? to be a kind of copula, but rather believe it to be a genuine pronominal subject in expressions of this sort, while the definite noun phrase preceding it is a kind of adjoined modifier. Without being able to give a more in-depth analysis to the nature of n??, it is reasonable to leave it at that. Nevertheless, the parallelism between appositive and non-verbal predicative expressions would not be unexpected, as it was argued already in (S??axmatov 1925-1927/2015: 360) and in the greater detail in (Burton-Roberts 1975) that appositive constructions are similar in their semantics to or even derived (in the transformational sense) from predicative expressions.

Following D. Zubova (Zubova 2018) and (Rogava, Keras?eva 1966: 66), I argued (Logvinova 2019) that in Adyghe, constructions of specification involve the use of a translative case and thus represent a kind of a typological rarum:

(73) Adyghe (< Northwest Caucasian < Northwest Caucasian, (Rogava, Keras?eva 1966: 66), glosses are mine with the reference to (Vydrin 2008))

a. ìûæúî-ó õúóãúý

stone-adv become;PST

`(he) became a stone'

b. îôèöåð-ýó Èâàíîâ

officer-adv Ivanov

`officer Ivanov'

It is noted in (Jakovlev, As?xamaf 1941: 393) the -ýó affix can also be used to mark the attributive function of adnominal modifiers. Adyghe lacks a clear distinction between nouns and adjectives, and the words with more “qualitative” meaning generally follow nominals they modify: êàðàíäàø (pencil) ïëúûæ (red) `a red pencil'. However, in some cases, a modifier can appear before the noun it modifies, receiving the -ýó suffix: ïëúûæ-ýó êàðàíäàø `red pensil' (ibidem: 79). Having examples like that one could say that Adyghe uses attributive-like strategy to express specification. This is not an undoubtedly correct solution since Jakovlev and As?xamaf (ibidem) mention that attributive constructions of this sort are not frequent in Adyghe and NPs with modifiers of this type can only occupy the position of a direct object. However, Rogava and Keras?eva (Rogava, Keras?eva 1966: 66) give examples of constructions of specification in subject position. Vydrin (Vydrin 2008: 433) considers Adyghe forms in -ýó (-ew) to be depictives. I have found no typological corelates to Adyghe suffix with a similar distribution and struggle to include it in the existing typology. A remarkable thing about Adyghe, is that similarly to Georgian and Kannada discussed above, it designated the head function to a proper name, rather than to a common noun.



3.4 Conclusions

In this Section, the basic morphosyntactic strategies for coding the meaning of specification in the languages of the world were discussed. The content of this chapter can be summarized in the following points.

1. The majority of the languages studied use simple juxtaposition to express specification -- it is the only available strategy in 65 out of 95 languages.

2. The second most used is attributivization. Attributivization wild-spread among the languages of Africa and Western Eurasia and appears in two types: similar and different to the one applied in possessive constructions.

3. Languages that use attributivization in constructions with toponyms in most cases opt for juxtaposition in constructions with personal names of humans. The exception from this rule is observed in some of the languages with attributive strategy different from the one applied in possessive constructions.

4. In most of the languages using attributive strategy, the role of the syntactic head in a construction of specification is designated to a common noun. The exception to this generalization again comes from the languages showing attributive strategy other than one found in possessive constructions.

5. Juxtaposition and attributivization are in competence: the languages allowing attributivization in some contexts and do not use it in others. Languages can switch from attributivization to juxtaposition, but the change in the other direction is not observed.

6. In contrast to languages opting for attributivization, Niuean language in constructions of specification uses the same particle as the one usually cooccurring with nouns in non-argument positions (in citations, among others), which shows that in Niuean one of the members of a construction is clearly treated to be “excluded” from the syntax.

7. In two languages from the sample (Movima and Japanese) constructions of specification are (presumably) parallel to relative clauses.

8. There is only one language in the sample (Urarina) that has a special morpheme used almost exclusively in the constructions of specification.

9. All in all, among the languages clearly distinguishing between head and dependent constituents in the constructions of specification (overall 32), the majority (28, including all languages with attributive strategy parallel to the one found in possessive constructions, Basque, Movima and Niuean) designate the head role to the common noun and 4 languages (Georgian, Kannada, Adyghe and, supposedly, Japanese) give the syntactic priority to the proper name.

10. 

4. Case marking in appositive constructions in typological perspective. The problem of constituency

4.0 Why interaction of apposition with case marking is worth considering

In this chapter I observe case marking patterns in constructions of specification made up with the juxtapositive strategy (appositional constructions in the syntactic sense) in languages having the grammatical category of case. The particular interest to interaction of case marking with the syntax of apposition comes from the two following considerations.

First, the syntactic scope and constituency of nominal apposition as a whole, and its elements in particular, as has been briefly shown in the introductory part, is a long-discussed problem in the literature, which, however, did not get a satisfying solution till now. The members of appositional constructions are either believed to constitute two (or more) separate NPs (the majority of the existing literature (Keizer 2007: 39)) or to be two nominal elements inside one NP (Keizer 2007). In each of the two cases the two following possibilities exist: the elements can be either in a hierarchical relation to one another (such as head -- modifier relation) or be juxtaposed without any further syntactic development. In the latter case (which I call juxtapositional analysis), appositives are expected to be structurally equal in a sentence, presumably not to form any constituent of a bigger scope, and to have the same relations with the external syntax of the clause. The juxtapositional analysis is schematically given in Picture 4. In case of hierarchization, the elements of apposition are expected to form a constituent of a larger scope and are not supposed to be in the same relation to the external syntax, as the function of a morphosyntactic locus would be attributed to the one of them playing the role of a syntactic head. The two possibilities are schematically given in the Pictures 4 and 5.



Picture 4 The juxtapositional analysis Both schemes are mine. X is the head, the indexes NP₁ and NP₂ do not imply any structural superiority of one element over another. Picture 5. “Single NP” analysis

The grammatical case, as a syntactic category, is defined exactly at the level of nominal phrases, and in the case languages an NP must receive case marking from its head at least once. Therefore, the case marking in a nominal phrase is indictive of its status as a syntactic unit. If appositive phrases are in fact treated as juxtaposed separated NPs, their members are supposed to receive case marking independently. Moreover, since their structural position is the same, the marking is expected to be the same This does not imply that the marking should be formally identical. as well. If that is true at least in some languages or it is even held as a universal, then one could consider it to be an argument in favor of juxtapositional analysis. Conversely, the “single NP analysis” predicts that appositions can receive a single marker per phrase.

Second, as is known from the data of Slavic languages (see Section 4.3 for more detail), case marking in appositive phrases can show a remarkable variation as the marker can be found on either each of the two constituents or only on one of them with the second preserving its “unmarked”. As an example, consider the possibility to preserve the Nominative form (also treated as unmarked form used in citations) of a proper name in Russian expressions like v gorod-e_loc Moskv-e_loc. vs. v gorod-e_loc Moskv-a_nom `in the city of Moscow' A member of an NP being unmarked for case is unusual for Russian, as it generally requires adjectival and pronominal modifiers to have the same case features as their heads (a minor exceptions is represented by a closed group of adjectives).. The avoidance of marking on each of the members of an apposition is predictable since it can be seen as “redundant”. What is really intriguing is what kind of factors can influence the presence or absence of “redundant” case marking in cases like that, how much variation is in general allowed in this domain and whether some languages are stricter than the others with regard to this phenomenon.

In the following sections I discuss the two issues in more detail. The section 4.3 is dedicated to the variability of case marking in appositive constructions in Slavic languages, while the preceding sections are devoted to the problem of constituency.

4.1 Patterns of case marking in appositive constructions of specification and constituency

In (Logvinova 2019b) I claimed that when considering the distribution of case markers in appositive phrases in relation to the problem of constituency, one should take into account the possibility of case agreement in the NP in principle. In the result I proposed the following scheme:

Picture 6 The correlation between case agreement in NP and case marking in appositive phrases

According to the scheme, the first scenario (1) does not tell anything about the constituency, because there can be more than one reason for case agreement. Members of apposition can either constitute two separate concatenated NPs each of which gets a case marker independently (see Picture 4 above) or it can be due to the process of case agreement in the NP. In the second scenario (2), members of an appositive phrase are expected to form a single NP but the status of the unmarked constituent is unclear. The third (3) scenario is the most straightforward: the one-time marking implies that the whole expression is in fact one syntactic unit. The fourth (4) scenario is also clear with the result that nominals in apposition are clearly two independent constituents.

The first type of noun phrase architecture is reported to be characteristic feature of the Australian languages (Louagie, Verstraete 2016). In many Australian languages, elements expected to be constituents of an NP receive independent case marking and can even be separated from each other by external elements of the clause:

(74) Kalkatungu (< Northern Pama-Nyungan < Pama-Nyungan (Blake 1983: 45), apud. (Louagie 2016)) cipa-yi t?uku-yu yaun-tu ya?i icayi

this-erg dog-erg big-erg white-man bite

`This big dog bit/bites the white man.'

On the basis of examples like (74) it is argued that there is no noun phrase level in Australian languages (although see (Louagie 2016) for counterarguments). 4 out of 5 Australian languages in our sample (Bininj Gun-Wok, Martuthunira, Wardaman and Warrongo) seem to have exactly this type of NP arrangement, while they allow multiple appearance of case feature within one NP and mark both constituents of an appositional phrase for case, as the following example from the language Martuthunira shows:

(75) Martuthunira (< Western Pama-Nyungan < Pama-Nyungan, (Dench 1987: 193))