Date: Tue, 22 Oct 1996 10:29:03 -0400 (EDT)
Message-Id: <199610221429.KAA11995@sarah.albany.edu>
In-Reply-To: <v02130501ae8cc18fc570@[128.120.185.208]>
To: Linguistics Conference <LINCONF@tamvm1.tamu.edu>
From: George Aaron Broadwell <g.broadwell@albany.edu>
Subject: Reply to Farrell from Broadwell
BROADWELL REPLY TO FARRELL
[Let me preface these comments with apology for
my delay in responding. I have been home sick for the
last few days and unable to reply]
Reply to comments from Patrick Farrell
Farrell correctly points out that a certain amount
of the argumentation in this paper is based on
the theoretical assumptions of the principles and
parameters approach.
An important assumption is the one embodied in the Projection
Principle (Chomsky 1981), which claims (approximately)
that syntactic rules don't change the subcategorization
features of lexical items.
Possessor Raising involves pairs like the following:
a.) [John im-ofi]-at illi-tok.
John III-dog-NM die-PT
b.) [John]-at [ofi]-at im-illi-tok.
John-NM dog-NM III-die-pt
'John's dog died.'
The only meaning difference that I am aware of between
the (a) and (b) versions is that 'John' is more topical
in (b).
Since the verb 'die' selects only a single argument, the projection
principle forbids an analysis in which 'John' is an argument of
(b) but not of (a).
Carden, Gordon, & Munro (ms) show that Chickasaw
possessor raising can apply iteratively, yielding
examples like the following:
c.) [Jan im-aaimpa' iyy]-at oppolo.
Jan III-table leg-NM broken
d.) [Jan im-aaimp]-at [iyy]-at oppolo.
Jan III-table-NM leg-NM broken
e.) [Jan]-at [im-aaimp]-at [iyy]-at oppolo.
Jan-NM III-table-NM leg-NM broken.
'The leg of Jan's table is broken.'
While it might seem plausible to think that 'John' is
somehow an argument of 'John's dog died', I think it
is far less plausible to assume that both 'Jan' and 'table'
are arguments of the verb 'broken' in (e).
Farrell's second point is tied to the first -- if we allow
a rule like possessor raising to add a new subject
to the sentence (presumably demoting the old one
in the process), then we could account for the
SR facts by referring to the grammatical relation
subject.
[For the record, let me point out that the
work on Choctaw by Davies that Farrell cites does
not contain any analysis of the sort of examples
of switch-reference and possessor raising that I
discuss in this paper. ]
The account I propose captures the similarity
between a subject and a raised possessor
configurationally. Both are in [NP, IP] positions
and both are potential antecedents of a SS
marker due to their syntactic position.
I doubt that the facts discussed here will change
anyone's mind about the validity of the projection
principle or the status of grammatical relations in
the theory. That's not really the point of the paper.
Readers who do not accept the background
assumptions of the paper will not accept the
conclusion either.
So, in deference to Farrell, let me restate the
argument of the paper. If you accept a theory
with the projection principle, then you are
pretty much forced to the analysis
of possessor raising given in my paper.
The interaction of possessor raising with switch-
reference shows that the binding requirements of
switch-reference markers are best stated in purely
structural terms -- thematic notions like agent or
co-argument do not play any necessary role.
--------------------------------------------------------------------------
George Aaron Broadwell, g.broadwell@albany.edu
Anthropology; Linguistics and Cognitive Science,
University at Albany, SUNY, Albany, NY 12222 | 518-442-4711