CSLI Publications logo
new books
contact us
for authors
LFG Proceedings
CSLI Publications

Adnominal Clauses with Genitive Subjects across Asia

Peter Sells


In this paper I will consider the internal properties of clausal and clause-like modifiers of nouns, focussing on clauses with genitive-marked subjects, schematized in the structure in (1). Data from a variety of Altaic and East Asian languages point to interesting subtle differences in the conditions under which genitive subjects appear. In LFG, building on familar ideas of constructive case (Nordlinger 1998), this can be put in terms of what properties of the surrounding environment --- c-structure and f-structure --- are dictated by a genitive subject in a given language. LFG provides a formal representation of a typological spread, within which the languages mentioned here are local points.

(1) NP
        / \
Mod.Clause N
     /   \

1. So-called 'Ga/No Conversion' in Japanese has been the subject of a great deal of research (e.g., Ueda 1966, Harada 1971, Bedell 1972, Hiraiwa 2002, among others). The form no canonically marks possessors. The case alternation is shown on the simple examples in (2). (3) shows that the genitive subject is not a structural possessor of the larger NP, as it is clearly internal to the adnominal clause. In (3), the adnominal clause contains a multiple ga construction, of which the second phrase is the grammatical SUBJ (underlined), and which is marked with genitive.

(2) a. [Taroo-ga katta] hon
[Taroo-NOM buy.PAST] book
'the book that Taroo bought'
b. [Taroo-no katta] hon
[Taroo-GEN buy.PAST] book
'the book that Taroo bought'
(3) [kayoobi-no zyuu-zi-ga tugoo-no yoi] hito
[Tuesday-GEN 10-hour-NOM convenience-GEN good.PRES] person
'the person(s) for whom Tuesday at 10 is convenient' (Kanazawa 1990)

Martin (1975, 662) notes: “Throughout the history of Japanese the two particles ga and no have shared functions with each other, and the actual distribution of the functions today varies from dialect to dialect”. I propose that the non-possessor use of no above is to mark the subject of an adnominal clause (see Kanazawa 1990), as in (5), alongside the simple characterisation of ga in (4):

(4) -ga
(SUBJ ↑)    ('I am subject of my clause')
(5) -no
(SUBJ ↑)    ('I am subject of my clause')
(GF SUBJ ↑)    ('the clause of which I am subject is subordinate in a larger structure')
CAT((GF SUBJ ↑), N)    ('the category of that larger structure is N')

2. Examples from Altai (from Ackerman and Nikolaeva ms) show a somewhat similar construction, but with different overt marking. There is evidence --- presented in Kornfilt (2008) but not shown here --- that the genitive phrase in (6)a is internal to the modifying clause, even though it has the same form as a simple possessor, shown in (6)b.

(6) a. [bis-tin kel-gen] d'oly-bys
[we-GEN come-PART] road-1pl.poss
'the road on which we came'
b. bis-tin balyg-ybys
we-GEN fish-1pl.poss
'our fish'

However, the possessor agreement on the head noun is not obligatory, and is not necessary to license a genitive subject:

(7) [sler-dinat-kan] tülkü
[you-GEN shoot-PART] fox
'the fox you shot at'

In other words, the analysis of Japanese no also applies here - genitive marks the subject of an adnominal clause. Nominative subjects are also possible, when there is no agreement on the head noun.

Returning to the structure with agreement on the head noun, this must be true possessor agreement in (6)b. If so, what is its function in (6)a? This agreement can express the subject of a possessive relative:

(8) [tut-kan] balyg-ybys
[catch-PART] fish-1pl.poss
'the fish we caught'

Ackerman et al. (2004) propose an HPSG analysis of the participle in which it modifies a noun whose specifier is coindexed with the subject of the participle. In LFG, this might be expressed as in (9): assuming that the adnominal clause is an ADJ of the head noun, and that the relative clause involves a null TOPIC coindexed with the head noun, internally associated with the OBJ of the verbal participle:

(9) Altai verbal participle
(↑ OBJ) = (↑ TOP)('the object of the clause is the topic')
(↑ TOP INDEX) = ((ADJ ↑) INDEX)('the topic's index is the noun's index')
(↑ SUBJ) = ((ADJ ↑) POSS) ('the subject is the noun's possessor' (shown by (8))

3. Genitive subjects are found in Turkish (nominalised) embedded clauses, as well as relative clauses. The subjects of such clauses are routinely genitive, subject to the following conditions (Haig 1998, 164):

(10) a. Subjects of adnominal clauses are genitive; unless they are are low in 'control' (e.g., animacy) or 'individuation',
in which case they are nominative.
b. If an adnominal clause has a genitive subject, the predicate has the DIK nominaliser ('PP' for Possessive Participle in (11))
and shows possessive agreement with the genitive subject;
c. If the adnominal clause has no genitive subject, the Free Participle ('FP' in (11)) suffixed with -(y)An is used,
and there is no agreement within the clause.
(Hence the FP is typically used for subject relatives ---- there is no subject within the clause, and so no genitive subject.)

Consider the examples in (11) from Göksel and Kerslake (1995, 444-5):

(11) a. [ön-ün-den köpru geç-en] ev-ler
[front-3sg.poss-ABL bridge(NOM) be.situated-FP] house-PL
'the houses in front of which a bridge is situated' (subject is low in individuation)
b. [ön-ün-den köpru-nün geç-tiğ-i] ev-ler
[front-3sg.poss-ABL bridge-GEN be.situated-PP-3sg.poss] house-PL
'the houses in front of which the bridge is situated' (subject is individuated)

For Turkish, I propose (12), in which (ii-iii) are like Japanese and (iv) is like Altai, with the crucial information 'migrated' from the predicate to the case marker:

(12) Turkish genitive case
i. (↑ CASE) =c GEN('I need to be assigned Genitive case (by the Possessive Participle)')
ii. (SUBJ ↑)('I am the subject of my clause')
iii. (GF SUBJ ↑)('the clause of which I am subject is subordinate in a larger structure')
iv. ↑ = ((ADJ SUBJ ↑) POSS) ('information from me is shared with the POSS argument of the modified noun')

Considering these languages from Turkey to Japan, it can be shown how LFG formalises the grammatical properties of genitive subjects and the adnominal clause predicates that they occur with. As suggested here, the properties of genitive subjects are in fact remarkably consistent across Asia.


pubs @ csli.stanford.edu 
CSLI Publications
Stanford University
Cordura Hall
210 Panama Street
Stanford, CA 94305-4101
(650) 723-1839