definiteness agreement in Hungarian

[Martin Haspelmath]

I had understood Dunstan Brown's query as excluding the Hungarian
situation, which can be analyzed as showing object agreement, but only
when the direct object is definite.

Such restrictions on agreement are of course not rare -- many languages
lack object agreement under similar circumstances (sometimes also
involving animacy), e.g. Spanish, Swahili, etc.

In Hungarian, an additional argument for saying that cases like

A   ko"nyvet olvas-sa'k.
the book:ACC read-3P("def")
'They are reading the book.'

involve object agreement under the condition of definiteness, rather
than definiteness agreement, is that the "indefinite" (or in my terms,
object-agreement-less) verb form is chosen when the object is a
non-third-person pronoun, e.g.

engem olvas-nak.
me:ACC read-3P("indef")
'They are reading me.'

Since 'me' is clearly definite semantically, this fact is puzzling on
the definiteness-agreement view, whereas it is expected on the
object-agreement view. (However, it must be added that the
object-agreement paradigm in Hungarian is defective, i.e. in most cases
no object-agreement forms exist for non-third-person objects.)

Martin Haspelmath