Definiteness and Verb

[Dunstan Brown]

A couple of weeks ago we (Dunstan Brown, Greville Corbett, Julia Barron)
posted the following question to the ALT discussion list and the
Linguistlist (ask a linguist) list:

It appears to be the case that we would not expect a language to have a
verb which agrees in definiteness with one of its arguments. In saying
this, we do not intend to refer to verbs which agree in other properties
only in the environment of a definite NP, i.e. where definiteness is a
prerequisite for agreement. Does anybody know of any counterexamples and
also does anyone know if this claim has been made before?

Our thanks to all those who have responded - Matthew Dryer, Martin
Haspelmath, Deborah D. Kela Ruuskanen, Maria Koptjevskaja-Tamm, Andrew
Spencer, Larry Trask and especially Edith Moravcsik.

We thought it would be useful to summarise the discussion so far.

Several people cited Hungarian as the main example of a language which
appears to  display definiteness agreement on the verb. The following
examples are from Edith Moravcsik:

   - Intransitive verb:  Iskola'ba ja'r-nak.
                         to:school go   3P
                        'They go to school.'
   - Transitive verb:
     -- indefinite object: Egy ko~nyvet olvas-nak.
                            a   book:ACC read  3P
                            'They are reading a book.'
     -- transitive object: A   ko~nyvet olvas-sa'k.
                            the book:ACC read  3P
                           'They are reading the book.'
(~ is used  for Umlaut.)

The verbal suffix thus differs in form when the object is definite.

There are three contrasting analyses:  
Firstly (i.e. Edith Moravcsik)  that as the verbal inflection differs
only in the environment of a definite object, then the inflection itself
must be a fused subject (person, number) agreement and object
(definiteness) agreement marker, there being no other property of the
object which is referred to by the affix, such as gender or number. This
may be represented schematically as:

     I hit-S1SUBJ.DEFOBJ the girl.
     I hit-S1SUBJ        a girl.
     I hit-S1SUBJ.DEFOBJ the boy.
     I hit-S1SUBJ        a boy.

Secondly, (i.e. Martin Haspelmath) that such contrasts indicate that the
verb agrees with its object under the condition of definiteness, i.e.
when the object is definite, but otherwise it does not. This does not
mean that it agrees in the feature definiteness, but that agreement is
possible only in the environment of definiteness.  This may be
represented as follows:

I hit-S1SUBJ.S3OBJ the girl.
     I hit-S1SUBJ        a girl.
     I hit-S1SUBJ.S3OBJ the boy.
     I hit-S1SUBJ        a boy.

Thirdly, (i.e. Andrew Spencer) that Hungarian verbs *don't* agree with
their objects, rather they take a special kind of subject agreement in
the presence of 'definite' objects and then only with 3rd person objects
(except for the 1stSUBJ - 2(fam)OBJ
-lek/lak).

Martin Haspelmath suggested that if Hungarian verbs agreed in
definiteness with their objects then they should agree with person
pronouns which are inherently definite. Edith Moravcsik responded by
proposing that the definiteness feature that determines verb inflection
in Hungarian is a grammatical, rather than semantic, one: not all
semantically definite objects count as "definite" for purposes of
selecting verb inflection (e.g. 1st and 2nd persons). She also suggested
that the Hungarian examples differ from the kind of constructional
contrasts cited by Mathew Dryer in that the latter show differences due
to relational properties, whereas agreement features are typically
non-relational. If definiteness is not a relational feature then it
qualifies as being an agreement feature.

[We also found an interesting account of the Hungarian subject/object
marking strategy in Daniel Abondolo, (1988) Hungarian Inflectional
Morphology, Budapest:Akademiai Kiado, which addresses the issue of the
interaction of person with implicit and explicit objects.]

Clearly there is not a consensus on this topic.  If agreement is such
that:
X agrees with Y in Z under condition C
then everybody seems to agree on X (the verb) but views on what could
constitute Y, Z, and C vary. (e.g. Y is variously subject or object, Z
is nothing or definiteness etc. and C is the crucial part for some and
not for others)

We are working on an agreement database, and so the question of
definiteness  agreement is an important one for us, both in terms of how
one would wish to analyse the data and also of how to organise it so
that one can recover it irrespective of theoretical persuasion.

It does however appear unlikely that we will find a clear example of an
affix which unambiguously signals definiteness and no other property or
set of properties of the arguments of the verb. We will keep looking.

Thanks again.
Julia Barron.


--
Dr Dunstan Brown
Lecturer in Linguistics and Russian Language
LIS
University of Surrey
Guildford
Surrey GU2 5XH
Tel: +44 1483 259957
Fax: +44 1483 259527
Email: d.brown at surrey.ac.uk