Cheng, Lisa Lai-Shen. (1997). On the Typology of Wh-Questions.
Outstanding Dissertations in Linguistics Series.
Garland Publishing, New York & London. 210 pages.
Reviewed by Kerstin Hoge, University of Oxford
This book (originally written as the author's 1991 Ph.D. dissertation)
attempts to account for the full range of cross-linguistic variation found
with single and multiple wh-questions. To this aim, Cheng proposes the
Clausal Typing Hypothesis which postulates a parameter that identification
(or 'typing') of a sentence as a wh-question is achieved either by a
question particle or by overt wh-movement. The first part of the book
(Chapters 2 and 3) discusses the Clausal Typing Hypothesis and its
predictions and implications for the analysis of wh-in-situ languages,
single movement languages, optional movement languages and multiple
fronting languages. Data from multiple fronting languages lead Cheng
to an investigation of the interpretation and inherent properties of
wh-words in a number of languages. This marks the transition to the
second part of the book (Chapters 4 and 5) which shifts the focus to
wh-words and quantificational phenomena in Mandarin Chinese.
Cheng's book provides an extensive overview of the differences exhibited
across languages with respect to wh-question formation and presents an
interesting attempt to unify two previously separate approaches to
wh-movement in arguing that both properties of C and of the wh-words
themselves are responsible for the observed typological distinctions.
Chapter 1 presents the general aims and outline of the book and briefly
introduces the phrase structure of Mandarin Chinese, which features
prominently in the book as an example of a wh-in-situ language.
Chapter 2 introduces the Clausal Typing Hypothesis. Cheng
observes that languages which allow for wh-in-situ in single questions
invariably use overt markings (such as particles) in matrix yes-no
questions. If a language has an overt yes-no question particle, it will
also have a wh-question particle, which may but need not be overt.
Question particles and overt wh-movement, both of which identify a clause
as an interrogative, are thus mutually exclusive. To account for this
generalisation, Cheng proposes that clauses are typed at S-structure
either by a wh-particle or by overt wh-movement but crucially not by both
in one language, cf. (1).
(1) Clausal Typing Hypothesis
Every clause needs to be typed. In the case of typing a
wh-question, either a wh-particle in C is used or else fronting of a
wh-word to the Spec of C is used, thereby typing a clause through C by
(=Cheng 1997:22, (9))
The two available strategies are illustrated by Mandarin Chinese and
Mandarin Chinese satisfies clausal typing by base-generating a
wh-particle in C. Consequently, all wh-phrases will stay in-situ. Overt
wh-movement is ruled out by the Principle of Economy of Derivation
(Chomsky 1989), according to which movement is more costly than Merge, a
'last resort' operation and applicable only when clausal typing could not
be achieved otherwise. At LF, wh-phrases move to SpecCP for scope,
selection and absorption purposes.
English uses overt wh-movement to satisfy clausal typing, i.e. C
acquires the [+wh]-feature of the XP in its specifier. The Clausal Typing
Hypothesis further requires Cheng to make the following assumptions with
respect to English: (i) subject wh-phrases must undergo overt movement
(contra the Vacuous Movement Hypothesis), (ii) no Q-morpheme or
[+wh]-feature is base-generated in C, and (iii) 'whether' and 'if' are not
Chapter 3 discusses optional and multiple wh-fronting languages
which present a challenge to the Clausal Typing Hypothesis. Optional
fronting languages appear to use both modes of clausal typing, i.e. they
allow for wh-fronting although they have a wh-particle (given the
possibility of wh-in-situ). Using Egyptian Arabic, Bahasa Indonesia and
Palauan as languages of investigation, Cheng argues that questions with
fronted wh-phrases in optional fronting languages do not instantiate overt
wh-movement but display clear similarities to cleft and topicalisation
structures. Thus, a 'fronted' wh-argument is base-generated as the
subject of a reduced cleft construction, while a fronted wh-adjunct has
undergone topicalisation. Optional fronting languages are then
straightforward in-situ languages which satisfy clausal typing by
base-generating a wh-particle in (matrix) C.
Multiple fronting languages appear to violate the Principle of
Economy of Derivation in that all wh-phrases must move at S-structure
although clausal typing is satisfied by single wh-movement. Extending
Nishigauchi's (1990) analysis of Japanese wh-words, Cheng suggests that
bare wh-words in multiple fronting languages are similar to indefinites,
i.e. they lack inherent quantificational force and simply introduce
variables into the semantic representation which must be bound by other
elements in the sentence. It is thus not surprising that the
interpretation of a wh-word in a multiple fronting language can vary with
its context. For example, Polish wh-words are interpreted as
interrogatives when fronted, receive a D(iscourse)-linked reading when
in-situ, are used as polarity items in yes-no questions and conditional
sentences, and form the morphological base of indefinites, filling out the
whole paradigms of person, place and time. Cheng now argues that the four
different readings arise because there are four different binders which
determine the quantificational force of the bare wh-word (WH) in these
contexts, as schematised in (2).
(2) Binders (and Readings) of Polish WH-Words:
a. WH is bound by an overt D(eterminer)
b. WH is bound by a null D with interrogative force
c. WH is bound by existential closure
d. WH is bound by a [+wh] C
Null determiners which contribute interrogative quantificational force
need to be licensed and identified by feature matching with a [+wh] C at
S-structure. It is thus to satisfy the licensing requirement of the null
determiner that all wh-phrases in multiple fronting languages move to a
position governed by C.
The chapter also contains a discussion of Rudin's (1988) proposal
to distinguish between multiple fronting languages which allow for
multiply filled SpecCP (e.g. Bulgarian, Romanian) and those where SpecCP
can host only one wh-phrase and all other wh-phrases are IP-adjoined (e.g.
Polish, Czech). Here, Cheng proposes to replace Rudin's ECP-style
account of superiority effects with an analysis that derives them from the
'shortest movement' condition, so as to account for the observed [subject
+ adjunct] and [object + adjunct] orderings of wh-phrases in the
Bulgarian-type languages. Moreover, she suggests analysing IP-adjunction
as Q(uantifier) R(aising) and thus as clausebound, thereby ruling out
multiple wh-extraction from embedded clauses in Polish-type languages.
In Chapter 4 Cheng examines wh-words in Mandarin Chinese and
analyses them as polarity items which require both a trigger and a binder
for interpretation, cf. (3).
(3) WH-Words in Mandarin Chinese
reading trigger binder
interrogative wh-particle wh-particle
polarity yes-no particle/ existential closure
negation (unselective binder)
universal dou 'all' dou 'all'
Thus, when a wh-word is interpreted as an interrogative, it is licensed by
a wh-particle which will also bind it and contribute interrogative force
to it. The polarity reading is triggered by a yes-no particle or
negation, and the binder of the wh-word is introduced by existential
closure which will apply to the elements inside VP. It follows that
wh-words in affective contexts cannot occur in subject position (SpecAspP)
as this position is external to VP and therefore not in the scope of
Cheng further argues that all indefinites in Mandarin Chinese lack
inherent quantificational force and must be bound by existential closure.
This has the consequence that indefinites cannot occur in subject position
unless the AspP is the complement of a modal (you 'have') which can act as
a binder for the indefinite subject.
The second half of Chapter 4 is taken up with a discussion of the
universal quantifier dou 'all', which Cheng analyses as a quantificational
adverb which is base-generated as an Asp'- or V'-adjunct. At LF, the
distributor dou adjoins to its associate. This movement is clause-bound
and leaves a trace which must be antecedent-governed. dou is a selective
binder and can only license one NP at a time, moving to the closest
element which can use a trigger.
Chapter 5 deals with multiple quantification and interactions
between wh- and quantifier phrases in Mandarin Chinese and English. In
Mandarin Chinese scope relations between quantified phrases always reflect
their surface order. Cheng explains this lack of scope ambiguities as
resulting from the fact that neither indefinites nor universally
quantified phrases undergo QR.
The interpretation of questions which contain universally
quantified phrases (mei-NPs) appears to be sensitive to the presence of
the quantificational adverb dou 'all'. Whereas questions which contain a
mei-NP without dou only have an individual reading, questions with a
mei-NP which is licensed by dou allow for both an individual and a
pair-list reading, cf. (4).
(4) mei-ge-ren dou mai-le shenme
every-CL-person all BUY-ASP what
'What did everyone buy?'
a. what is the thing such that everyone bought?
b. for every x, what is the thing that x bought?
(= Cheng 1997:161, (31))
The ambiguity does, however, not result from different scope relations.
Rather, Cheng argues, the pair-list reading is due to a distributive
reading of mei-NP which is created by its association with the distributor
dou. Scope reconstruction (Frampton 1990) of the wh-phrase to the site of
its intermediate (VP-adjoined) trace (which Cheng adopts in her analysis
of English wh-quantifier interactions) is not possible in (4) because dou
creates a barrier and the intermediate trace therefore deletes at LF.
Cheng's analysis of wh-quantifier interactions in Mandarin Chinese
rests on the assumption that wh-phrases move at LF. She discusses both
arguments in favour of LF-movement and Aoun and Li's (1993) claim that
in-situ wh-phrases do not move at LF since they can be modified by 'only'
which needs to be associated with an overt element in its c-command
domain. Cheng argues that wh-phrases which are modified by 'only' must
have a D-linked interpretation and do therefore not constitute evidence
against LF-movement. She further notes Reinhart's (1990) observation that
D-linked wh-phrases have to move to be properly interpreted and suggests
that such movement takes place at a post-LF level. The chapter concludes
with a discussion of the landing site of LF-movement which Cheng
identifies as SpecCP (contra Mahajan 1990).
Cheng's book occupies a place which marks the onset of the Minimalist
approach to syntactic theory. While she assumes the Barriers framework of
Chomsky (1986) and uses a disjunctive ECP to derive locality conditions on
dou and the lack of scope reconstruction in wh-quantifier interactions,
the Principle of Economy of Derivation is central to her argumentation.
Not only does it prohibit a language from using both strategies for
clausal typing, it also accounts for superiority effects in Bulgarian (by
forcing movement of the closest wh-phrase to occur first), and precludes
LF-lowering of indefinite subjects in Mandarin Chinese (since there is an
alternative derivation which generates a structure with a modal binder).
Moreover, it rules out multiple wh-movement for clausal typing and thus
requires Cheng to account for multiple fronting languages as resulting
from a licensing requirement of the wh-words. As seen earlier, multiple
fronting is shown to correlate with the ability of the wh-words to form
the morphological base of indefinite NPs.
However, it appears that not all languages that have indefinites
which are derived from wh-words display multiple fronting. German
prefixes a wh-word with irgend- 'some' to form an indefinite, filling out
the whole paradigms of person, place and time, as seen in (5).
wer 'who' irgendwer 'someone'
wo 'where' irgendwo 'somewhere'
wann 'when' irgendwann 'sometime'
was 'what' irgendwas 'something'
Moreover, a bare wh-expression can be used as a polarity item in yes-no
questions and conditionals, e.g. (6)-(7).
(6) Hast du wen angerufen?
have you whom called
'Did you call anybody?'
(7) Wenn du wen anrufen willst, musst du mir Bescheid sagen.
if you whom call want, must you me let-know
'If you want to call anybody, you must let me know.'
The morphological forms and possible readings of German wh-words thus
mirror the situation found with Polish wh-words, as described by Cheng.
German and Polish also have in common that neither language has a question
particle (under Cheng's analysis) and that they must therefore use overt
wh-movement for clausal typing. However, whereas Polish fulfils Cheng's
prediction and fronts all wh-phrases in multiple questions, German can
only move one wh-word to clause-initial position, cf (8).
(8) a. Wer hat wen angerufen?
who has whom called
'Who called whom?'
b. *Wer wen hat angerufen?
who whom has called
Yiddish presents the reverse problem in that it allows for
multiple wh-movement but does not derive its indefinites from the wh-words
in a morphologically transparent way, cf. (9).
(9) ver 'who' emetser 'someone'
vu 'where' ergets(vu) 'somewhere'
ven 'when' a mol 'sometime'
vos 'what' epes 'something'
To make things worse, multiple fronting is optional in Yiddish.
Wh-phrases can stay in-situ without necessarily having a D-linked reading,
(10) a. ver vemen hot ongeklungen?
who whom has called
'Who called whom?'
b. ver hot ongeklungen vemen?
who has called whom
'Who called whom?'
German and Yiddish thus appear to be direct counterexamples to Cheng's
analysis of multiple fronting languages.
Note also in this context that French, often cited as an example
of an optional movement language, falls outside Cheng's analysis. As
acknowledged in the book (1997:39, FN 4), French does not have a question
particle and is therefore not an in-situ language. Nevertheless it allows
for wh-words to remain in-situ in single matrix questions.
It might be possible to account for optional wh-movement in French
by postulating that French possesses both a non-overt yes-no particle and
a non-overt wh-particle. French would then satisfy clausal typing by
base-generating a wh-particle in C and all wh-phrases could stay in-situ.
Movement of a wh-phrase to clause-initial position could be due to another
attractor, such as Focus, which is not obligatorily present. Whatever
appeal such an analysis might hold, it points to the problem of allowing
for non-overt question particles. While we do not expect to find overt
wh-particles in languages without yes-no particles, what will stop us from
assuming that a language which allows a wh-word to stay in-situ has a
question particle even when there is no (phonologically) overt evidence
for such a claim? Cheng's proposal to relate wh-in-situ to the
availability of non-overt wh-particles may thus ultimately turn out to be
too permissive a theory.
Another problem arises with Cheng's approach to superiority
effects in Bulgarian-type languages. As noted, Cheng explains superiority
effects as violations of the Principle of Economy of Derivation. Given
that "adverbs are in the most embedded positions" (1997:81),
[adjunct + argument] orderings are ruled out because movement of the
argument is a shorter movement and must therefore take place before the
adjunct moves to SpecCP. However, as observed by Boskovic (1997),
Bulgarian questions containing more than two fronted wh-phrases display
superiority effects only with respect to the topmost wh-phrase, the others
are freely ordered, contrary to the prediction made by Cheng's approach.
These comments aside, Cheng's book is to be commended. It is an
extremely well written and clearly presented work which should be fully
comprehensible even to readers who are not familiar with the literature on
Mandarin Chinese syntax or Heim's (1982) theory of indefinites. The book
is ambitious in its scope, discussing a wide range of languages, and
has undoubtedly made an important contribution to the study of
Aoun, J. and A. Li. (1993). Wh-elements in situ: syntax of LF?
Linguistic Inquiry 24.199-238.
Boskovic, Z. (1997). On certain violations of the Superiority Condition,
AgrO, and economy of derivation. Journal of Linguistics
Chomsky, N. (1986). Barriers. Cambridge, Mass.: MIT Press.
Chomsky, N. (1989). Some notes on economy of derivation and
representation. In MITWPL 10: Functional Heads and Clause Structure
(reprinted in Principles and Parameters in Comparative Grammar,
ed. by R. Freidin, 417-454. Cambridge, Mass.: MIT Press).
Frampton, J. (1990). The fine structure of wh-movement and the proper
formulation of the ECP. Ms., Northeastern University.
Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases.
Ph.D. dissertation, University of Massachusetts, Amherst.
Mahajan, A. (1990). The A/A-bar Distinction and Movement Theory.
Ph.D. dissertation, MIT.
Nishigauchi, T. (1990). Quantification in the Theory of Grammar.
Reinhart, T. (1990). Interpreting wh-in-situ. Ms.
Rudin, C. (1988). On multiple questions and multiple wh fronting.
Natural Language and Linguistic Theory 6.445-502.
Kerstin Hoge is a D.Phil. student in linguistics at the University of
Oxford. Her research interests include syntactic theory and Yiddish.