Unaccusativity in HPSG

Fri Nov 26 12:25:48 UTC 2004

Hi,

> (research on test mismatches, cross-linguistic variation, etc.). See,
>  for instance, work by Delia Bentley, Antonella Sorace, Annie Zaenen,
>  and my own, on these issues.

Yes, Ryu, 1997 and Müller, 2002 also discuss test mismatches.

> I believe the right move in HPSG is to 'stick to the surface', as its
> basic philosophy dictates.

I am for argumentation and against dictatorship. Sorry, couldn't resist.
=;-)

> Trying to find a way to represent 'unaccusative' subjects as
> underlying objects (i.e. by making them the value of an 'ERGATIVE'
> attribute, or something like that) is, in my view, the wrong way to
> go. Better to try to explain the phenomena that some believe
> characterize the class of unaccusatives on the basis of their
> particular semantic or pragmatic properties. Hope this contributes to
> the discussion.

I agree that semantic/pragmatic facts play an important part in an
explanation of the phenomenon. There is also work by Ingrid Kaufmann
on the Unaccusative-Unergative Distinction.

> Aranovich, RaÃºl. 2003. 'Two Types of Postverbal Subjects in Spanish:
>  Evidence from Binding.' In Claire Beyssade, Olivier Bonami, Patricia
>  Cabredo Hofherr, Francis Corblin (eds.) Empirical Issues in Formal
> Syntax and Semantics 4. Paris: Presses Universitaires de
> Paris-Sorbonne. 227-242

Hm, in Section 6 you write:

> The evidence reviewed so far, summarized in (17), shows that the
> split between verbs that can (or cannot) occur in ISI is orthogonal
> to the unergative/unaccusative split.

But this means that your paper is not about a `treatment of
unaccusativity using HPSG'. Rather you point out the non-applicability
of the inversion cases as a test for unaccusativity.

As I said above, there are also other tests (like auxiliary selection in
German) that result in inconclusive categorization of verbs into the
unaccusative/unergative classes. The question is what to do about such
situations. I decided to drop the auxiliary selection test from the
unacc tests and focus on prenominal participles and passive instead.

For instance the participle of the transitive verb `reparieren'
predicates over the object while the participle of an unaccusative verb
like `ankommen' predicates over its subject:

(1) a.   der reparierte Wagen
          the repaired   car
     b.   der angekommene Zug
          the arrived     train
     c. * der geschlafene Junge
          the slept       boy

We could follow your advice and not call the verb `ankommen'
unaccusative, but then we had to invent another name for the phenomenon
and after all, the data in (1) does show that the subject of `ankommen'
  differs from the one of `schlafen' in some way.

To analyze the data in (1), we have to say something about SUBJ, COMPS,
SUBCAT, and/or ARG-ST of these predicates. Using a list valued feature
DA for representing information about a blocked argument is motivated on
several grounds. For instance the logical subject remains accessible in
the representation for `reparierte' (the adjectival participle of
`reparieren'):

+-
| DA    < NP_1 >
| SUBJ  < NP_2 >
| COMPS <>
| CONT repair(1,2)
+-

The logical object of `reparieren' corresponds to the noun the particple
predicates over, the logical subject of the participle cannot be
realized by an NP in the prenominal position, but it can be realized by
an agent expression, usually a `von' or `durch' phrase.

The von phrase (by phrase in English) should be treated as an adjunct
(Höhle, 1978, some of Höhle's data are also contained in Müller, 2003,
Section 5). In Müller, 1999, Chapter 15.3, I treated the von phrase as
an argument that was introduced by the auxiliary, but this leads to
problems in the analysis of Partial Verb Phrase fronting, since the
`von' phrase can be fronted together with participles (see also Müller,
1999, p. 376, and Müller, 2003, Section 3.4). Having the  designated
argument present in the representation of `reparierte' (and in
projections of this word) makes it availible for coindexation with the
NP in the von phrase. For details of the analysis see Müller, 2003,
Section 5. So using such a feature is not just a hack.

A perfect theory should find some principled way to explain how the
value of DA is determined or influenced by semantic/pragmatic factors,
if such a set of factors can be isolated.

I agree that having several pointer features (for external arguments,
internal arguments, and maybe even other features) should be avoided (as
Dan Flickinger, pc 2000 pointed out to me, features like ERG introduce
grammatical functions into HPSG and this is against what the philosphy
dictates =;-). Having more than one such feature is not necessary and in
many cases these features break other parts of the grammar (see again
Müller, 2002 and Müller, 2003 for discussion).

Best wishes

	Stefan

@Book{Hoehle78a,
   author      =	{Tilman N. H{\"o}hle},
   title	      =	{{Lexikalische Syntax: Die Aktiv-Passiv-Relation und
andere Infinitkonstruktionen im	Deutschen}},
   address     =	{T\"ubingen},
   publisher   =	{Max Niemeyer Verlag},
   series      =	{Linguistische Arbeiten},
   number      =	67,
   year	      =	1978
}

@Book{Kaufmann95a,
   author      =	{Ingrid Kaufmann},
   title	      =	{{Konzeptuelle Grundlagen semantischer
Dekompositionsstrukturen. Die Kombinatorik
		lokaler Verben und pr"adikativer Elemente}},
   publisher   =	{Max Niemeyer Verlag},
   address     =	{T{\"u}bingen},
   series      =	{Linguistische Arbeiten},
   number      =	335,
   year	      =	1995
}

@Article{Kaufmann95b,
   author      =	{Kaufmann, Ingrid},
   title	      =	{O- and {D}-Predicates: A Semantic Approach to the
Unaccusative-Unergative	Distinction},
   journal     =	{Journal of Semantics},
   volume      =	12,
   number      =	4,
   pages	      =	{377--427},
   year	      =	1995
}

@Book{Mueller99a,
   author      =	{Stefan M{\"u}ller},
   title	      =	{{Deutsche Syntax deklarativ. Head-Driven Phrase
Structure Grammar f\"ur das
		Deutsche}},
   publisher   =	{Max Niemeyer Verlag},
   address     =	{T\"ubingen},
   series      =	{Linguistische Arbeiten},
   number      =	394,
   note	      =	{\url{http://www.cl.uni-bremen.de/~stefan/Pub/hpsg.html}},
   year	      =	"1999"
}

@Book{Mueller2002b,
   author      =	{Stefan M{\"u}ller},
   title	      =	{Complex Predicates: Verbal Complexes, Resultative
Constructions, and Particle Verbs
		in {German}},
   series      =	{Studies in Constraint"=Based Lexicalism},
   number      =	13,
   publisher   =	cslip,
   address     =	{Stanford},
   note	      =
{\url{http://www.cl.uni-bremen.de/~stefan/Pub/complex.html}},
   year	      =	2002
}

@inProceedings{Mueller2003e,
   author      =	{Stefan M{\"u}ller},
   title	      =	{Object"=To"=Subject"=Raising and Lexical Rule. An
Analysis of the {German} Passive},
   editor      =	{Stefan M{\"u}ller},
   booktitle   =	{Proceedings of the HPSG-2003 Conference, Michigan State
University, East Lansing},
   address     = {Stanford},
   publisher   =	{CSLI Publications},
   pages = {278--297},
   year	      =	2003
}
http://www.cl.uni-bremen.de/~stefan/Pub/passiv-da.html

@TechReport{Ryu97a,
   author      =	{Byong-Rae Ryu},
   title	      =	{{Argumentstruktur und Linking im constraint"=basierten
Lexikon: Ein Zwei-Stufen-Modell für eine HPSG"=Analyse von Ergativität
und Passivierung im Deutschen}},
   type	      =	"Arbeitspapiere des SFB 340",
   number      =	{\Nr 124},
   institution =	{Eberhard-Karls-Uni\-ver\-si\-t{\"a}t},
   address     =	{T\"ubingen},
   year	      =	1997
}

--
Stefan Müller

Universität Bremen/Fachbereich 10      Tel: (+49) (+421) 218-8601
Postfach 33 04 40
D-28334 Bremen

http://www.stefan-müller.net

http://www.cl.uni-bremen.de/~stefan/Babel/Interaktiv/