LINGUIST List 9.1262

Sat Sep 12 1998

Review: (part 1): Beerman et al. Rightwards mvmt.

Editor for this issue: Andrew Carnie <carnielinguistlist.org>


What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Andrew Carnie at carnielinguistlist.org

Directory

  1. Steven Schaufele, _Rightward Movement_ review -- I

Message 1: _Rightward Movement_ review -- I

Date: Wed, 09 Sep 1998 17:56:55 -0700
From: Steven Schaufele <fcosw5mbm1.scu.edu.tw>
Subject: _Rightward Movement_ review -- I


[Editor's note: This is part one of a two part review. The second 
part appears in the next issue of LINGUIST]



Beerman, Dorothee, David LeBlanc, & Henk van Riemsdijk, eds. (1997) 
Rightward Movement (Linguistik aktuell 17) Amsterdam: Benjamins. 406
pp.

This is a proceedings volume collecting together papers presented in 
Oct. 1995 at the Tilburg Conference on Rightward Movement. The 
editors mention two or three papers that were presented at the 
conference but were not including in this proceedings volume. I say 
`two or three' because, although they mention three authors (Kayne, 
Koike, and Truckenbrodt), Koike actually has a paper in this volume; 
i do not know why his name is included in the list of authors whose 
papers were not included. At the back of the book, a complete list 
of addresses (snail- and e-mail) of the contributors is included.

The Editors' Preface (with van Riemsdijk named as principal author) 
provides a very nice summary of the relevant issues and their 
background in the literature. Special note is taken of repercussions 
of recent work in Minimalist Program (e.g., Kayne 1994) for the whole 
notion of `rightward movement' (hereafter RM). This reviewer would 
note in particular that, although it is commonly supposed that, as 
stated in the preface, `In a minimalist approach, movement is 
exclusively triggered by checking .... Given this new line of 
thinking, Rightward Movement simply cannot be triggered, hence it 
cannot exist', some of the contributors -- e.g., Alphonce & Davis, 
Buring & Hartmann -- demonstrate that it is in fact possible to 
develop hypotheses within the Minimalist framework that would enable 
triggering of such movement.

Of the 14 papers in the collection, it is hardly surprising that 7 
deal with `extraposition' (defined in various ways), and four of 
those are concerned primarily with one aspect or other of 
extraposition in German. I will discuss these extraposition papers 
as a group before considering the others.

Josef Bayer's paper `CP-Extraposition as Argument Shift' (pp. 37-58) 
begins with a very nice, neat summary of problems with classical 
extraposition account (via RM) for postverbal CPs in V-final 
languages (focussing particularly on Bengali, Hindi, and German), 
noting that these problems disappear under a Kayne-type analysis. 
However, he goes on to note definite empirical problems with a Kayne-
type analysis. He then proposes an analysis according to which a 
complement is right-adjoined to the maximal projection of its 
governing head (in this case VP), leaving behind a trace as sole 
sister of that head (V). If that trace is then deleted and the tree 
is pruned, the V ends up with a complement to its right which it can 
theta-mark. Alternatively, Bayer suggests, at least some 
`extraposed' CPs may be base-generated to the right, co-indexed with 
a (dummy/deletable) pronominal, by inheritance from which they are 
licensed. Bayer notes that this would help account for the fact that 
in some OV languages (e.g., Bengali), certain classes of CPs are 
*always* `extraposed'. If we assume that directionality is relevant 
to selection, then the resulting VP constitutes a barrier to a CP on 
the `non-canonical' side of the head, which would account for the 
scope effects Bayer notes earlier as being problematic for both the 
traditional account and the Kayne-type account.

In `Rightward Scrambling' (pp. 186-214), Anoop Mahajan argues on the 
basis of various relations sensitive to c-command that postverbal 
nominal arguments in Hindi are merely constituents left behind while 
everything else has moved leftwards. This analysis supersedes the RM 
analysis he proposed in an unpublished paper ten years ago and is 
deliberately consistent with an analysis based on Kayne's Linear 
Correspondence Axiom (hereafter LCA). This reviewer notes that many 
of Mahajan's arguments necessarily presuppose certain possibly 
dubious tacit assumptions, e.g., that RM must necessarily involve 
adjunction specifically to IP (contrary to the approach proposed in, 
e.g., Muller's, Wiltschko's, and Rochemont & Culicover's papers). 
For instance, the string in Mahajan's (33), which he marks (???) 
could be generated -- and its unacceptability accounted for -- by 
right-adjunction of the direct object to VP rather than IP. Nowhere 
does Mahajan actually address what in this reviewer's opinion is the 
most basic issue with questions like this: Is there or is there not 
any evidence of a *gap* corresponding to the postverbal material?

Michael S. Rochemont & Peter W. Culicover in `Deriving Dependent 
Right Adjuncts in English' (pp. 279-300) discuss various 
constructions in English, all of which might be included under a 
rather broadly-defined concept of `extraposition'. Distinguishing 
between the extraposition of relative clauses on the one hand and 
Heavy-NP Shift and Presentational-There Insertion on the other, they 
argue that Relative-Clause Extraposition is best treated as (1) base-
generated and (2) right-adjunction to the governing category (VP, IP, 
or CP) of the antecedent to the extraposed RC. They make an effort 
to conjure up plausible analyses of RC-Extraposition involving Kayne-
style leftward-movement but note that none of the possibilities they 
consider are quite satisfactory. Even the `best' option, involving 
movement of both the RC and its antecedent to distinct Spec 
positions, fails to provide any motivation for either the posited 
movement or the highly ramified structure such an analysis requires.

Expanding on earlier work of their own (Rochemont & Culicover 1990), 
Rochemont & Culicover argue that Heavy-NP Shift and Presentational-
There Insertion are best treated as instances of movement to a right-
adjoined A'-position. They demonstrate that, quite apart from the 
problems discussed in Rochemont & Culicover 1990, any attempt to 
analyze such constructions by means of exclusively leftward movement 
involves the extremely unattractive movement of what is not, in fact, 
by any stretch of the imagination a recognizable constituent. In the 
end, they acknowledge that there are some empirical problems shared 
by both the rightward-movement account they apparently prefer and the 
`movement to high specifier' account that would be more consistent 
with a Kayne-type approach, but that the latter raises some provoking 
theoretical problems that are absent from their rightward-movement 
account. They conclude by saying that `the question whether 
rightward movement exists or not ... is not an empirical one.'

Daniel Buring & Katharina Hartmann's paper `The Kayne Mutiny' (pp. 
59-80) presents an excellent argument for the empirical bankruptcy of 
the Kayne Antisymmetry hypothesis. Making crucial use of 
reconstruction at LF and of Binding-Theoretic statements referring to 
(undeleted) traces, B&H's argument is built upon the prediction that, 
if extraposition is a consequence of RM, it ought to be possible for 
a proper binding relation *not* to exist between an NP and a CP later 
in the sentence -- if the NP happens to be in a hierarchically lower 
position, from which neither it nor any of its daughters is able to 
c-command the CP (such a lack of binding relation is a priori 
impossible in a Kayne analysis, according to which any NP to the left 
of a CP must ipso facto c-command it). They then demonstrate that 
such binding failures are in fact attested, and are indeed not all 
that difficult to come up with in a language like German. (At the 
end of section 2, they acknowledge some confusing results with regard 
to coreference options, concluding that these `require further 
investigation'.) They further demonstrate (section 3) that the Kayne 
analysis actually does serious violence to many standard assumptions 
about movement, including (similarly to Rochemont & Culicover) issues 
of what qualifies as a (movable) constituent and under what 
circumstances a constituent may be `stranded'. (It's from this 
surreptitiously iconoclastic character of Kayne's hypothesis that 
they get their clever title.) And they demonstrate that verb-
topicalization ought to be impossible in a Kayne analysis, although 
of course it's quite common in German.

In order to account for the complications with regard to island-
constraint violations, etc. that have presented problems for earlier 
versions of a RM-analysis of extraposition in German and similar 
languages, B&H propose (p. 72) a generalization according to which 
finite clauses may never be governed by either V or I. This provides 
an actual motivation for CP-Extraposition, since presumably in its DS 
position a complement clause is governed by the matrix verb, and in 
order to escape that government must be right-adjoined to some higher 
phrasal node, presumably IP. This is in direct conflict with Bayer's 
analysis, according to which the extraposed CP ends up being governed 
by the matrix verb as a result of the deletion of its own trace and 
tree-pruning; which analysis is to be preferred ought to be an 
empirical problem.

Hubert Haider's paper `Extraposition' (115-152) argues on the basis 
of the extraposition of comparatives and the c-command relations 
essential thereto in English and German (mostly German) that 
extraposed constituents remain embedded in their DS mothers. Haider 
further argues that extraposed relative and argument clauses must 
also be VP-internal, since although they aren't subject to the same 
c-command relations themselves, they always come *before* extraposed 
comparatives which are. Broadening his scope in Section 2 to other 
examples of German extraposition, Haider demonstrates that they can't 
result from movement and must therefore be base-generated. But, on 
the basis of scope, c-command, and absence of island-effects, he also 
argues against an analysis in terms of base-generated adjunction.

Haider agrees with Kayne in assuming exclusive Leftward Movement; 
however, he allows for either head-initial or head-final base 
structures, and invokes head movement while Kayne invokes phrasal 
movement. Haider presents several predictions that Kayne's LCA 
theory would have for a language such as German, which he then 
demonstrates are all falsified by the actual data: (1) Phrases to the 
left of the verb should be in Spec-positions, and should therefore be 
islands (2) VP-adverbials and predicates should end up in postverbal 
position, since there's nothing to trigger their movement (3) VP-
topicalization ought to involve the movement of a functional 
projection containing a trace of the finite verb.

In `Extraposition as Remnant Movement' (p. 215-246), Gereon Muller 
offers a very neat analysis of extraposition in German as right-
adjunction to a variety of phrasal nodes, including CP as well as VP 
or IP, thereby accounting for various otherwise problematic details 
with regard to island effects in both leftward- and rightward-moved 
constituents. The paper includes a very interesting and useful 
comparative discussion of the adequacy of a variety of different 
proposed constraints for excluding unacceptable strings while 
allowing acceptable ones.

Martina Wiltschko's paper, `Extraposition, Identification and 
Precedence' (pp. 358-396), a summary of her 1995 Wien dissertation, 
discusses extraposition in German, focussing on the relation of 
*Identification* between the `identifyee', the (pro)nominal element 
(NP or DP) in the canonical position within the clause and the 
`identifier', the extraposed constituent. Both identifyee and 
identifier provide linguistically necessary information: The 
identifyee occupies a canonical (theta-)position, therefore 
satisfying syntactic requirements, while the identifier provides 
necessary semantic content to licence the identifyee's definiteness. 
Given that the identifyee *introduces* a discourse referent, it must 
(on the basis of Heim's (1980) Novelty Condition) precede the 
identifier. Wiltschko also argues for a Locality Constraint on 
Identification, according to which the identifier must c-command the 
identifyee, without any intervening XP; thus, the identifier must be 
right-adjoined to the minimal maximal projection dominating the 
identifyee. In Wiltschko's view, these two constraints together 
account for the fact that identifiers are always extraposed. She 
acknowledges that this analysis apply only to *restrictive relative 
clauses*, not to other types of modifiers. Attractive as the paper 
is in many ways, it suffers somewhat from the necessary exclusion of 
many supporting arguments, for which the interested reader is 
referred to the full-length dissertation.

Two of the papers are concerned primarily with parsing theory and the 
development of adequate parsing technology. Both of these papers, 
coming from different points of view, argue for a data-driven, 
bottom-up parsing strategy as against a hypothesis-driven top-down 
strategy. In `On Movement and One-Pass No Backtrack Parsing' (pp. 
301-330), Chris Sijtsma recognizes that `natural' (i.e., single-pass, 
no backtracking, faithful to derivation) bottom-up parsers are less 
restrictive than natural top-down parsers, which of course from the 
point of view of strict generative theory is a point against them, 
but assumes that there is enough variation among actual languages 
that a bottom-up parsing strategy is to be preferred. This reviewer 
finds such a conclusion attractive, but worries that Sijtsma has 
provided so little in the way of empirical demonstration to back it 
up; indeed, for such a mathematically-oriented paper (at least 
relative to this reviewer's experience), there is extremely little in 
the way of solid argument presented; most of the time, Sijtsma merely 
asserts that the proof of any given theorem is either self-evident or 
readily derivable; in a few cases, he refers to demonstrations 
elsewhere in the literature. Another issue both of these papers 
consider very seriously, without, however, either of them coming up 
with a very satisfactory solution, is the proper size of the look-
ahead window for an adequate parser. Sijtsma asserts (pp. 305-6) 
that any grammar with a look-ahead window greater than 1 is 
functionally equivalent to a grammar that looks ahead just one 
symbol, but then goes on to say, `In practice we still need ... 
parsers that look ahead more than one symbol.' In subsequent 
discussion it becomes clear that he is unclear just how large a look-
ahead window is empirically adequate. Likewise, Alphonce & Davis, 
while currently working with a look-ahead window of `at most two 
chunks' (p. 25), are clearly dissatisfied with this characterization. 
It is clear that this issue needs more thought, if not further 
research, devoted to it.

A fundamental claim of the paper by Carl Alphonce & Henry Davis, 
`Motivating Non-directional Movement' (pp. 7-36), is that Linear Precedence 
constraints, indeed LP phenomena of any kind, have no relevance for 
syntax at all; essentially, they claim that, from the point of view 
of all syntactic levels including LF, constituents are organized 
hierarchically in terms of dominance relations but not linearly in 
terms of precedence relations. In Alphonce & Davis' view, all 
precedence relations are imposed at PF, making them essentially 
matters of performance rather than competence. In the opinion of 
this reviewer, this is a very interesting and possibly attractive 
idea. Unfortunately, contrary to the promise contained in the 
abstract, this claim is not so much argued for as assumed within the 
paper. Nor is it made clear -- to this reviewer, anyway -- that it 
is explicitly argued for anywhere else, unlike the skipped arguments 
behind Wiltschko's paper and the citations given in Sijtsma's paper. 
Alphonce & Davis merely demonstrate that it is possible to develop an 
analytic approach -- more precisely, a parsing program -- that has no 
need for any kind of explicit syntactic constraints, at any level 
(whether UG or language-particular), making reference to linear 
order.

At the end of their abstract, Alphonce & Davis claim that they are 
motivated by a conviction that `it is a priori desirable to eliminate 
as much redundancy as possible between different components of the 
system.... if some phenomena has [sic] an independent processing 
explanation we hold that syntactic theory should not have to offer 
any explanation for it.' This approach is all very well in a purely 
formal mathematical system, but it is fairly common knowledge that 
redundancy is in fact a sine qua non of biological systems (cf. e.g. 
Gould 1993) and of natural-linguistic systems as well (cf. e.g. Hock 
1986, ch. 9 & 12; this fact is also acknowledged by Chris Sijtsma in 
his paper, p. 314). The mere fact that one can develop a parsing 
program that has no need to appeal to syntactic LP constraints, 
therefore, in no way demonstrates that such constraints have no place 
in human natural-language competence.

Much of Sijtsma's paper is devoted to developing points (regarding, 
e.g., the proper type(s) and subcategorization frame(s) of PPs) that 
are clearly relevant to his primary concern, which is developing an 
adequate automated parsing grammar, but are tangential to the focus 
of the collection. In arguing, contrary to Kayne, that UG does not 
stipulate one universal tree-structure for all languages, Sijtsma 
gets a fair amount of mileage out of replacing the assumption that 
node-labels are atomic with the assumption that they are merely 
shorthand for feature-bundles. Though he doesn't mention this, this 
replacement has actually been implicit in X-Bar Theory ever since the 
early 70's. With regard to directionality of movement, Sijtsma 
argues that rightward movement must be allowed by UG, with this 
caveat: In deriving SS from DS, leftward movement is unrestricted but 
rightward movement of modifiers (which don't leave obvious gaps) 
should not exceed the look-ahead buffer; on the other hand, in 
deriving LF from SS rightward movement is unrestricted but leftward 


[Editor's note: This review is continued in the next issue of LINGUIST]


- 
Steven Schaufele, Ph.D., Asst. Prof. of Linguistics, English Department

Soochow University, Waishuanghsi Campus, Taipei 11102, Taiwan, ROC

(886)(02)2881-9471 ext. 6504 fcosw5mbm1.scu.edu.tw

http://www.prairienet.org/~fcosws/homepage.html



 ***O syntagmata linguarum liberemini humanarum!***

 ***Nihil vestris privari nisi obicibus potestis!***
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue