Message-Id: <9611031635.AA06787@dragon.emich.edu>
Date: Sun, 03 Nov 96 10:28:54 -0800
From: Gil Rappaport <grapp@mail.utexas.edu>
To: linconf@tamvm1.tamu.edu
Subject: Comments on Lasnik's paper
Lasnik's admirably clear paper can be summarized as follows. LF
movement involves FEATURES, not CATEGORIES. Overt movement, on the
other hand, subject to PF requirements, pied-pipes categories along
with features. Since the semantic and referential properties affecting
binding and scope are associated with categories (not features),
binding and scope are affected by overt movement, not by LF
movement. The former includes familiar cases of A-movement (raising to
subject) and A-bar movement (Wh-movement); expletive constructions
(there is/are NP XP) behave like overt movement; therefore, assume
that they involve overt movement, of NP to Spec-of-Agr(o).
Questions to the author/comments:
-Bottom line: where ARE binding conditions satisfied, at S-Structure
or LF?
Favoring LF: a) there is no SS, b) semantic essence of binding
suggests LF, and c) even though LF movement doesn't affect
binding/scope, that does not mean that LF cannot be where binding
is. It then becomes moot whether it is at the input to LF (ne
S-Structure) or at its output (interface with the cognitive system).
Favoring SS: Chomsky's argument that Condition C must be
satisfied at S-Structure (5-7).
Aside on this argument: the argument depends on a (relevant)
difference between the S-Structure and LF of a given construction. The
assumption as presented is that (5) (with the relative clause in
Spec-of-Comp) resembles the LF of overt Wh-movement, as well as of QR
and covert Wh-movement, so that (6) and (7) illustrate the difference,
and it is the S-Structure which gives the configuration for a
Condition C violation. But Economy points to (5) as providing the
difference. According to P&P, the relative clause in Spec-of-Comp is
an illegitimate LF. As a result of pied piping, the operator is
CONTAINED IN (rather than simply IS) the category binding the
trace. This violates the Economy principle of Full Interpretation,
which entails that traces can be bound only by operators (a trace not
bound by an operator is not a well-formed LF object). The LF of (5)
should resemble `for which x, x a book, John read x that he
liked'. Then of course it isn't clear why Condition C should not be
violated in (5) at LF. Which is the point.
I've never found the Barss facts like (8) grammatical, and can't let
them by. (11) is ungrammatical to me as well. A simpler construction:
*John asked [which picture of himself did Sue take]? The weaker Barss
facts are, the weaker the evidence for S-Structure binding is.
After (43) and (44), the author writes `However, under the feature
movement analysis of LF movement, in (44), even if there is LF
WH-movement, only the formal features of `which picture of himself'
would move, leaving the anaphor in the lower `governing category'. I
believe that `(44)' should read `(43)'. Or else I don't understand at
all.
Stjepanovic's analysis of structural and inherent case is cited in
support of the notion that both types of case assignment result from
raising to Spec-of-something, and feature checking with the Spec,
feature Spec-Head agreement. If Agr(o) is the landing cite for Acc,
Dat, Inst, and Gen. complements of the verb, how is it ensured that
the right case (ultimately a lexical property of the verb) is used? I
don't insist on the old version of head-complement checking, but at
least then it could be said that NPs are inserted with a case feature
and a V has a matching seeking feature for that particular case:
_pomogne_ `help' would have the seeker feature [Dative-]; only a
complement N with the feature [Dative] could send its case feature on
up to the higher head to check the case feature; a [Genitive]
complement would lead to crashing. How is this work done in the
Stjepanovic system, assumed here?
Gilbert Rappaport
Dept. of Slavic Languages
Univ. of Texas at Austin