Definiteness

[Spencer, Andrew J]

One interesting example of the kind of thing Dunstan and colleagues may have
in mind is the situation in Latvian.

Latvian adjectives agree with nouns in number, gender and case, but they
have indefinite and definite sets of inflections. Num/Gend/Case agreement is
obligatory both within the NP (as attributive modification) and externally
(in predicative use with a copular verb). Even as secondary predicates, e.g
in resultatives, adjectives agree with nouns in Num/Gend but also in Case:

paint house-ACC red-ACC.FEM.SG 'paint (the) house red'

Within the NP the adjective has to be marked according to the definiteness
of that NP:

new house = 'a new house'
new-DEF house = 'the new house'

If there is a definite determiner (e.g. demonstrative or possessive) in the
NP, the adjective must be marked definite:

this/Gunars-GEN new-*(DEF) house	'this/Gunars' new house'

This looks for all the world like definiteness agreement within the NP.
However, it's not obvious that this is the best way of treating the
phenomenon. Notice that we have to say that the adjective agrees with the
determiner or possessive for definiteness, while agreeing with the noun for
(at least) Gender. We may or may not wish to say such a thing.

Definiteness behaves differently from Num/Gend/Case in predicative
constructions, in that the predicative adjective agrees with the subject
noun in everything *except* definiteness. Thus you would say:

this/Gunars-GEN new-DEF house is red-FEM.SG.NOM

not: *this/Gunars-GEN new-DEF house is red-FEM.SG.NOM-DEF

[I assume you would also have to say 'paint new-DEF house red' and not
'paint new-DEF house red-DEF but I would have to check that with
informants].

Of course, where a predicative adjective is used as a part of a zero-headed
(or conversion) construction, the adjective can be definite:

Imants bija pirmajs
Imants was first-DEF
'Imants was the first'

Thus, it might be better to say that definiteness, a property of the whole
phrase, is simply marked wherever it can be within the NP.

On the other hand, do we want to say that an adjective agrees with its noun
for Case? After all, this two is really a property of the whole NP isn't it.
Arguably the same could be said of Number. So maybe Case and Number are also
categories which have to be marked several times (in some languages) within
the NP, perhaps in as many places as is allowed. Of course, in Hungarian we
get just single marking for both Case and Number. If this reasoning is
correct then the only true agreement is Gender agreement because this is the
only property that can be ascribed inherently to the noun and not to the
construction (phrase) as a whole. This might make it a little easier to
understand such things as the absurdly complex patterns of Number/Case
marking with numerals in Balto-Slavic. If that is mere 'marking within the
NP' rather than some sort of agreement we might end up with a simpler theory
of agreement.

Incidentally, would Dunstan et al. wish to say that 'we wouldn't expect a
verb to agree in Case with one of its arguments'? This seems a reasonable
hypothesis to pursue - it's not clear to me what the answer is. In simple
situations it would be impossible to distinguish between agreement for
(Nom/Acc or Erg/Abs) Case and agreement for grammatical function (SUBJ/OBJ
or Ext/Int argument). You'd therefore need to look at quirky case. Are there
any *strong* instances of verbs having different subject or object agreement
markers depending on inherent case assignment? A hypothetical example would
be an Icelandic-prime in which the subject agreements were different for
Nom, Acc, Gen and Dat subjects. A weaker instance of this might be the
Latvian situation mentioned above, in which adjectives as secondary
predicates have the same Case as their secondary arguments (and other cases
involving Dative marked subjects).

Where does this leave Number?

How does all this relate to Suffixaufnahme?

Andrew Spencer




*******************************
Andrew Spencer
Dept. of Language and Linguistics
University of Essex
Colchester, CO4 3SQ, UK

tel:	+44 1206 872188
fax:	+44 1206 872085
email:	spena at essex.ac.uk
http://privatewww.essex.ac.uk/~spena
*******************************


> -----Original Message-----
> From:	Kersti Börjars [SMTP:k.borjars at MAN.AC.UK]
> Sent:	Friday, 19 November, 1999 20:21 PM
> To:	LINGTYP at LISTSERV.LINGUISTLIST.ORG
> Subject:	Definiteness
>
> I was interested in Dunstan's distinction: between 'coincidence of form'
> vs. 'agremeent'.
>
> >If we take it to be a feature of NPs, then an NP
> >may be marked as definite in more than one place, and we would
> >have a coincidence of form, rather than one agreeing with the other.
>
> Now, that seems a really tricky distinction to make. Does it depend on
> whether or not one of the elements can be seen as "inherently" having the
> feature, so that since nouns could be said to *have* gender, gender
> agreement within the noun phrase exists. Is this it?
>
> The argument you give I can't quite follow: from the (possible) lack of
> definiteness agreement between verbs and objects you conclude that
> definiteness marking within the noun phrase is not agreement. Why is this?
>
> Kersti
>
> **************************************************
>
> Kersti Börjars
> Department of Linguistics
> University of Manchester
> Manchester M13 9PL             k.borjars at man.ac.uk
> UK	       http://lings.ln.man.ac.uk/html/keb/
> Tel: 44+(0)161-275 3042
> Fax: 44+(0)161-275 3042
>
> ***************************************************