Simpson, Andrew (2000) Wh-Movement and the Theory of Feature-
Checking. John Benjamins, hardback ISBN: 1-55619-856-6,
xii+244pp, $75.00.
Reviewed by Edward John Garrett, Department of Religious
Studies, University of Virginia
In "Wh-Movement and the Theory of Feature-Checking", Andrew
Simpson presents an in-depth analysis of wh-constructions
from a diverse variety of languages, and draws forceful and
innovative conclusions regarding the licensing of wh-
phrases. The book is a substantially revised version of his
1995 doctoral dissertation from the School of Oriental and
African Studies in London. It is divided into three
chapters: Chapter 1 introduces the problem of wh-in-situ,
and puts forth numerous arguments against the traditional
view that wh-phrases in-situ must covertly move to a +wh+Q
Comp at Logical Form (LF). Chapter 2 introduces novel data
from Iraqi Arabic, which shows that in some languages wh-
phrases may be licensed non-locally, checked "at a
distance" from their +wh+Q Comps. The data shows that
crucial reference must be made to the position of wh-
phrases at overt syntax (Spell-Out), and that such cases
simply cannot be analysed with LF movement. Chapter 3
examines partial movement constructions (PM) in languages
such as German and Hungarian. Simpson argues that PM
supports Chapter 2's thesis that wh-phrases may be licensed
non-locally. In Chapter 3 he also works out the details of
his argument that the degree of locality of wh-phrase
licensing in a given sentence depends on the exact nature
of the +wh+Q Comp that occurs in the sentence. Furthermore,
even the same language may have different +wh+Q Comps, and
therefore may exhibit different kinds of locality
restrictions in different sentences.
The overall force of Simpson's book is to question the
need for LF as a separate level of representation, and to
argue for a principled, but parametric, treatment of wh-
phrase licensing which allows locality constraints to vary.
The book would seem to be most directed at syntacticians,
and particularly those in the Minimalist tradition - little
mention is made of other traditions, and Chomsky's recent
work figures prominently. However, since the data is cross-
linguistically wide-ranging and much of it new, the book
should also appeal to language typologists and empirically
oriented semanticists.
I suspect that the conclusions of Chapter 1 will be fairly
uncontroversial. Simpson correctly notes that arguments
for LF have hinged crucially on the "LF parallelism
assumption", the idea that LF movement dependencies
essentially parallel those in overt syntax. He carefully
investigates a variety of construction types, including
sentences with 'only' and wh-in-situ, parasitic gaps,
antecedent controlled deletion, weak crossover, and others,
in each case showing that wh-in-situ behave differently
from overtly moved wh-phrases. If there is LF movement of
wh-phrases, then, it does not parallel overt wh-movement at
all - and indeed Simpson proposes a no-movement
approach to wh-in-situ.
Chapter 2 begins with thought-provoking new data from Iraqi
Arabic (IA). IA exhibits an unusual constellation of
properties. Generally, both wh-movement and wh-in-situ are
possible. However, there are some cases where movement is
allowed, but wh-in-situ is not. For example, while a wh-
phrase may move across a -Q tensed clause to a +wh+Q Comp,
it may not stay in-situ inside a -Q tensed clause. On
standard accounts of wh-movement, this is most unexpected,
since LF movement should be possible wherever overt
movement is possible. The fact that the wh-phrase can move
overtly shows that nothing blocks it from moving, but if it
stays in-situ it is not licensed. Simpson draws several
conclusions from this data: Significantly, there must be an
overt licensing requirement on wh-phrases in IA. In
particular, a wh-phrase may be licensed in a non-local
"checking domain", provided that no tensed clauses
intervene between the +wh+Q tensed clause and the wh-
phrase. The IA data constitutes strong evidence against the
"checking uniformity hypothesis" of orthodox Minimalism,
namely that all features are checked locally in a Spec-Head
or Head-Spec relation.
Simpson argues that it is a feature of parametric
variation among languages as to how far the checking domain
of a +wh+Q Comp spreads. In English, the limit is the
sentence, since wh-phrases can occur in-situ at any
distance from a +wh+Q Comp. In Bulgarian, strict Spec-Head
locality is enforced.
If English wh-phrases don't need to move to be licensed,
then why do they move? Reflection on this question leads
Simpson to the second part of his thesis: the
"triggering hypothesis" (TH). According to TH, the reason
wh-movement occurs in languages like English is to
disambiguate a +Q Comp as either definitely +wh or
definitely +yes/no. This idea may be hard for some to
swallow, but it does find some support in Cheng's work on
the typology of question syntax.
The arguments of Chapters 1 and 2 seem fairly solid, and
although many may disagree with the details, I think that
Simpson's crucial arguments - e.g. that there is no LF
wh-movement, and that there is non-local feature-checking -
will be well-received. I am much less persuaded by the
arguments concerning partial movement.
Again, the thrust of the argument in Chapter 3 is that wh-
phrases must be licensed in their overt, Spell-Out
positions. The problem of PM, then, comes to the questions
of (a) what causes the wh-phrase to move to -Q Comp? and
(b) what job is done by the dummy wh-phrases? I am not
convinced by either of his answers.
Simpson argues that the dummy (expletive) wh-phrase in
a PM constructions triggers a special type of +wh+Q Comp
which enforces a weaker than normal locality restriction on
wh-licensing. Therefore, wh-phrases in PM constructions
need not move all the way to the root +wh+Q Comp, and
instead need only move a short distance to a -Q Comp, to
find themselves within the checking domain of a +wh+Q Comp.
Simpson's argument is partly motivated by his
observation that PM in German is subject to a kind of
tense-related locality similar to the tense-related
locality on wh-licensing in Iraqi Arabic (see above).
The PM chapter, however, suffers from several weaknesses.
First, there is no discussion whatsoever of possible
meaning or use distinctions between PM and non-PM
constructions, where the two freely alternate. It is
assumed that the two kinds of constructions are identical
in meaning and use, but no evidence is given one way or the
other. As long as investigations of PM continue to fail to
address this issue, they must remain cast in a significant
shadow of doubt.
This lingering doubt is particularly bothersome in
connection with Simpson's claim that "PM cannot
possibly be analyzed as involving checking of any other
non-wh features" (p. 154). The confidence of this assertion
notwithstanding, in fact the only non-wh movement that is
explicitly argued against is PM as focus movement.
Certainly, however, other possibilities come to mind,
including for example, PM as movement for scope, not of the
wh-feature, but of the nominal part of the wh-phrase. The
troubled relation between PM and negation in fact might
suggest a scopal solution. In any case, any evidence of
meaning or use distinctions between PM and non-PM would put
the issue in a totally different light, so it is
regrettable that this line of investigation is not
explored.
A second problem concerns Simpson's critique of
Horvath's LF pied-piping approach, in which the phrase
containing the partially moved wh-phrase moves at LF to
replace the wh-expletive in Spec, CP. Although Simpson
does present several good arguments against Horvath's
theory, he also takes these arguments to be general
arguments against any "indirect dependency approach". But
this is surely mistaken: one of the insights of Horvath's
approach, as well as of pied-piping approaches in general,
is that what matters is not the contained phrase (e.g. the
wh-phrase itself), but the container phrase (e.g. the
clause that contains the wh-phrase). The idea is that a wh-
feature might move to percolate its wh-feature to its
containing clause, and that it might be the clause which is
in need of wh-licensing, not the wh-phrase. As far as I can
tell, none of Simpson's criticisms of the LF pied-
piping approach apply to a percolation without LF movement
approach, which certainly shares the spirit of the pied-
piping approach. In fact, a percolation without LF movement
approach is very much in line with Simpson's own ideas,
since what is then required is the overt, Spell-Out
licensing of a wh-featured clause.
Finally, Simpson's chapter on PM seems to turn a well-
supported view of expletive-associate relations upside
down. Normally - and certainly in Horvath's approach -
there is a direct relationship between the wh-expletive and
the partially moved wh-phrase (or its containing clause).
In Simpson's theory, on the other hand, the crucial
relationship is between the +wh+Q Comp and the wh-phrase.
The wh-expletive triggers a special kind of Comp, which
then creates a special relationship with wh-phrases, which
results in PM. To my mind, this adds a second level of
difference where one was enough: Why do we need a different
kind of +wh+Q Comp if we already have a different kind of
creature in the wh-expletive? Why not maintain, more
simply, that the wh-expletive in PM constructions has a
special relationship with the partially moved wh-phrase or
containing clause, and just keep Comp out of it? Given the
importance of expletive-associate relations to syntax, and
work already done by Horvath, Mahajan, Dayal, and others on
this subject, why jump boat so soon?
Before closing, I want to add a brief remark in connection
with Simpson's concluding thoughts. In spite of
appearances to the contrary, he suggests, the collapse of
the checking uniformity hypothesis does not constitute a
significant challenge to the uniformity of syntax. That is,
the fact that different languages define their wh-checking
domain in different ways, and that even within the same
language we may have different checking domains in
different sentences, does not challenge the fundamental
uniformity of syntax. Although this low-level variation
does exist, "uniformity which can be taken as highly
significant is assumed to be present in the cross-
linguistic organization of functional structure and the
requirement that functional heads successfully establish
licensing relations with corresponding elements generated
in lower lexical domains" (p. 231). While I agree with
Simpson that his findings aren't irrecoverable blows to the
uniformity of syntax, at the same time I don't think that
his findings tend to support the uniformity of syntax
either. Rather, I see them as a small dent in the
uniformity machine. The machine is less beautiful as a
result, but it will be some time still before we know for
sure if the damage portends worse to come, or if it is only
cosmetic.
About the Reviewer
I received my PhD from the Department of Linguistics at
UCLA in January 2001. My dissertation was on evidentiality
in Tibetan, and my research tends to hover at the
intersection of semantics, pragmatics, and philosophy.
Currently I am working at the University of Virginia as a
linguist and computer programmer for a Tibetan language-
learning project funded by the Department of Education.
|
