Ian Roberts Li9 – 17 November 2010
1
SYNTAX (LI/LT 9)
Ian Roberts
Movement Theory II: Wh-movement
1.
Introduction and recap: types of movement
(1)
A-movement (passives, raising, unaccusatives) involves:
a.
movement of a DP;
b.
leaves a trace/copy at the extraction site;
c.
moves to a position c-commanding the trace;
d.
moves to the closest possible Specifier;
e.
obeys the Structure Preservation Principle;
f.
(almost) always moves the DP to a subject (SpecIP) position.
(2)
Head movement (subject-aux inversion, V2, V-to-I) involves:
a.
movement of a head (X);
b.
leaves a trace/copy at the extraction site;
c.
moves to a position c-commanding the trace;
d.
moves the head to the closest available head position;
e.
obeys the Structure Preservation Principle.
(3)
Structure Preservation Principle:
*Movement of Xn to Ym where n  m for X’-theory.
2.
Wh-movement
(4)
a.
b.
c.
d.
c.
d.
(5)
Movement of one wh-phrase per wh-interrogative is obligatory in English:
a.
*Has Bill bought what?
b.
*I wonder [ Bill has bought what].
(6)
“Echo-questions” show where the wh-phrase originates:
Which book did Bill buy (which book)?
I wonder which book Bill bought (which book).
What a book Bill bought (what a book)!
It’s amazing what a book Bill bought (what a book)
the book which Bill bought (which book)
the man who bought the book
[direct question]
[indirect question]
[exclamative]
[indirect excl]
[object relative]
[subject relative]
Bill bought WHAT?!
(7)
Indirect echo questions are impossible:
*I wonder Bill bought WHAT?!
(8)
Wh-movement (interrogatives, exclamatives, relatives) involves:
a.
movement of an XP containing a wh-determiner/pronoun;
b.
leaves a trace/copy at the extraction site;
c.
moves to a position c-commanding the trace;
Ian Roberts Li9 – 17 November 2010
d.
e.
f.

2
DOES NOT move the wh-phrase to the closest available position (see
below);
obeys the Structure Preservation Principle.
always moves the wh-phrase to SpecCP position.
since SpecCP is an A’-position, wh-movement is also known as A’movement.
We are obviously very close to a unified theory of movement, but what about (8d)?
(9)
a.
b.
c.
d.

Which books did Bill buy (which books)?
Which books did you force Bill to buy (which books)?
Which books did Harry say you had forced Bill to buy (which books)?
Which books was it obvious that Harry said you had forced Bill to buy
(which books)?
The wh-expression can be fronted over an indefinitely large amount of
material: it is an unbounded dependency.
However, and this is a really crucial observation, wh-expressions cannot be fronted
over just any sequence of material:
(10)
a.
b.
c.

*Which book did Bill buy War and Peace and?
*Which book did that Bill bought surprise everyone?
*Which book did John fall asleep and Bill buy?
The enterprise of accounting for the facts in (9) and (10) – i.e. the
construction of a theory of unbounded dependencies – is central to the
concerns of generative grammar for three reasons:
I. We require the theory to be descriptively adequate (i.e. it must account for the facts
of English and other languages, and any typological generalisation which can be
observed). Therefore, given our observations in (9) and (10), unbounded
dependencies like wh-question formation are constrained somehow, and the
formulation of the constraints is an empirical matter.
II. As Ross (1986:6) points out, whatever constraints are formulated are likely to hold
more widely than just in English:
The constraints on variables which I will propose are often of such a complex
nature that to state them as constraints on rules in particular languages would
greatly increase the power of transformational rules .. So, from my
investigations of the few languages I am familiar with, I will tentatively
assume that the constraints I have arrived at are universal.
Another reason to think that the theory of unbounded dependencies is directly
connected to UG comes from the nature of the data. Given the rather exotic nature of
the data relevant to the formulation of the constraints, it is implausible that these
constraints are acquired. Therefore they must be innate, part of the language faculty.
Ian Roberts Li9 – 17 November 2010
3
III. The existence of a class of constraints on unbounded dependencies reduces the
class of possible languages. The constraints limit the functioning of transformational
rules, and therefore introduce an element of greater restrictiveness into the theory of
UG. As we said above, this is a desirable step towards the overall goal of explaining
the knowledge and acquisition of language, since language acquirers thus have fewer
hypotheses to consider in the process of grammar construction. So the formulation of
constraints on wh-movement is a major part of the project of achieving explanatory
adequacy.
3.
Constraining Wh-Movement: the theory of locality as it applies to
unbounded dependencies
3.1
The A-Over-A Principle
(11)
a.
b.
(12)
The A-Over-A Principle (Chomsky (1955, 1964)):
If a transformation applies to a structure of the form
[S … [A … [A … ] S
for any category A, then it must be interpreted so as to apply to the maximal
phrase of the type A.
(13)
a.
b.
Mary saw the boy walking towards the railway station. (ambiguous)
Who did Mary see walking towards the railway station?
(not)
*[DP Who ] did Mary see [DP ([DPWho]) walking towards the railway
station ] ?
-- violates (12)
[DP Who ] did Mary see ([DPWho]) [CP PROMary walking towards the
railway station ] ?
The ban on extraction of like from inside like has some good consequences, e.g.:
(14)
a.
b.
No extraction from relatives:
*Here’s the snowball [DP which ] I chased [DP the boy who threw
(which) at our teacher ].
(Ross (1986:12))
No extraction from complements to nouns:
*Where’s [DP the bikini ] which Tom mentioned [DP the fact that Sue
had worn (the bikini)] ?
(Ross (1986:13))
Problems:
(14)
Wh-movement from gerunds (Chomsky (1964:47-50)):
a.
Who would you approve of [DP my seeing (who) ] ?
b.
What are you uncertain about [DP giving (what) to John ] ?
Ian Roberts Li9 – 17 November 2010
AP-movement:1
a.
Handsome though I believe Dick is, I’m still going to marry Hermann.
b.
*Handsome though I believe (the claim) Dick is, I’m still going to
marry Hermann.
(15)

3.2
4
Here we see that AP-fronting from a complex NP is impossible, but the AOver-A Principle can’t account for it.
Island Constraints (Ross (1967/1986))
Ross (1967) introduced the concept of “island”, a piece of structure out of which
movement is impossible, e.g. the Complex NP Constraint (CNPC):
(16)
I.
Complex NP Constraint (CNPC):
No element contained in a sentence dominated by a noun phrase with a lexical
head noun may be moved out of that noun phrase by a transformation.
(Ross (1986:76))
The CNPC accounts for two main classes of facts:
(i)
the impossibility of extraction from relatives, as in (17):
(17)
*Which writer did you write [DP a play which [ was about (which writer) ]] ?
(ii) the impossibility of extraction from sentential complements to nouns like claim,
fact, story, etc., as in:
(18)
*Which writer did you believe [DP the claim that [
writer) ]] ?
we had met (which
The second island constraint discussed by Ross is the Coordinate Structure
Constraint (CSC), stated as follows:
(19)
II.
Coordinate Structure Constraint (CSC):
In a coordinate structure, no conjunct may be moved, nor may any element
contained in a conjunct be moved out of that conjunct.
(Ross (1986:99))
This constraint rules out the following kinds of example:
Is this really wh-movement? It’s definitely unbounded:
a.
Handsome though I believe Fred said Dick is, I’m still going to marry Hermann.
b.
Handsome though I believe it is obvious that Fred said Dick is, ..
c.
Handsome though I believe it is obvious that everyone knows Fred said Dick is,..
Also, though could be a C, with the AP in SpecCP. And cf. the very similar:
(iv)
However handsome you think Dick is, …
(v)
No matter how handsome you think Dick is, …
These contain the adjectival wh-word how. So “wh-movement” is an operation that extends beyond the
constructions listed in (8).
1
(i)
Ian Roberts Li9 – 17 November 2010
(20)
a.
b.
c.
5
*The lute which Henry plays and sings madrigals is warped.
*The madrigals which Henry plays the lute and sings sound lousy.
*The nurse who polished her trombone and the plumber who computed
my tax was a blonde.
(Ross (1986:98))
Ross observed that this constraint is in fact more powerful than the A-Over-A
Principle, as there is no natural way that the A-Over-A Principle can account for this
class of facts.
[An important proviso to the CSC is that certain rules can apply to coordinate
structures as long as they apply “across the board”, in that identical elements must be
affected in each conjunct. One example of across-the-board (ATB) movement is
Relative Clause Formation:
(21) Students who fail the final exam or who do not do the reading will be
executed].
III.
The Left Branch Condition (LBC)
No NP which is the leftmost constituent of a larger NP can be reordered out of this
NP by a transformational rule.
(Ross (1986:127))
The LBC says that, starting from (22a), only the biggest DP (DP1) can be extracted:
(22)
a.
b.
c.
d.
I played [DP1 [DP2 [DP3 Mick’s ] friend’s ] favourite guitar ]
*[DP3 whose ] did you play [DP1 [DP2 ([DP3whose]) friend’s ] favourite
guitar ] ?
*[DP2 whose friend’s ] did you play [DP1 ([DP2 [DP3whose] friend’s ])
favourite guitar ] ?
[DP1 whose friend’s favourite guitar ] did you play {[DP1 [DP2 [DP3whose]
friend’s ] favourite guitar ] ?
Ross (1986:145) notes that this condition does not seem to hold in certain languages,
e.g. Russian and Latin:
(23)
a.
b.
Čju
ty čitaješ
knigu?
Whose you are-reading book
“Whose book are you reading?”
Cuius legis
librum?
Whose you-are-reading book?
“Whose book are you reading?”
(Russian)
(Latin)
One could add to this list French combien (“how many”):
(24)
Combien as-tu
lu de livres?
How-many have-you read of books
“How many books have you read?”
Ross suggests that the grammar for each language contains a “conditions box” where
language-particular constraints on large classes of rules can be stated. Thus, the
Ian Roberts Li9 – 17 November 2010
6
English conditions box features the LBC, while the Latin and Russian ones
presumably do not.
IV.
The sentential subject constraint (SSC)
(25)
No element dominated by an S may be moved out of that S if that node S is
dominated by an NP which itself is immediately dominated by S.
(Ross (1986:149))
This is a rather roundabout way of saying that you can’t extract from a clause which is
in the subject position of another clause.
(26)
*Which politician was [IP [DP [CP that the police would arrest (which
politician) ]]] expected ?
Chomsky (1973) extended this to all complex subjects (the Subject Condition: no
extraction from subjects):
(27)
?*Which politician were [IP [DP the supporters of (which politician)]
arrested ]]?
V.
The wh-island constraint (Chomsky (1964:37ff.))
No extraction of a wh-phrase out of a clause introduced by another wh-phrase:
(28)
a.
b.
(29)
*How were you wondering [ which car [ you should fix (which car) (how) ]] ?
(30)
He told me about a book which I can’t figure out …
… whether to buy or not.
… how to read.
… where to obtain.
… what to do about.
(Ross (1986:19))

3.3
Looking at (29) and (30), there appear to be “weak” and “strong” wh-islands
(see Cinque (1991)).
Subjacency (Chomsky (1973))

(31)
??Which car were you wondering [ how [ to fix (which car) ]] ?
?*Which car were you wondering [ how [ you should fix (which car) ]]
The logical next step in research on island phenomena was to attempt to
characterise what the various island constraints have in common. In this way,
it was hoped that an intensional characterisation of the notion of island could
be arrived at, in place of an extensional list.
Successive-cyclic movement:
a.
John seems [IP (John) to have been arrested (John) ]
b.
[CP1 What did [IP1 Harry say [CP2 (what) [IP2 you had forced Bill [CP3
(what) [IP3 to buy (what) ]]]]]] ?
Ian Roberts Li9 – 17 November 2010
7
 In other words, wh-movement is only apparently unbounded.
Chomsky (1973:243) put this as follows:
(32)
COMP-to-COMP Condition: a phrase in COMP can only move to a higher
COMP [for “COMP” read “SpecCP”].
Subjacency can be stated as follows (this is a rather simplified version of the
formulation in Chomsky (1973:247f., 271f.)):
(33)
No rule can relate X and Y in the following structure:
… X … [B … [ B … Y … ]
where X is separated from Y by more than one bounding node B.
Chomsky proposed that bounding nodes were all and only cyclic nodes, i.e. S and NP
(these correspond to our IP and DP).
To see how subjacency works, consider the derivation of a sentence involving
extraction from a relative clause:
(34)
*Who did Mary read the book which we gave to?
Moving which on the lower cycle gives rise to an intermediate structure like (35):
(35)
Mary read [DP [DP the book ] [CP which [IP we gave to who ]]]
The only way to derive (34) from (35) is to move who up to the matrix COMP
position in one step. However, such a movement violates subjacency, as two IP-nodes
separate the base position of who from the matrix SpecCP, as can be seen in (35).
[What if who moves first to the lower SpecCP and on to the matrix one, followed by
movement of which to the lower SpecCP? This violates the Strict Cycle Condition:
(36)
Strict Cycle Condition: No rule can apply to a domain dominated by a
bounding node A in such a way as to affect solely a proper subdomain of A
dominated by a node B which is also a bounding node.
So you can’t apply a rule first to the whole CP and then just to the most deeply
embedded one].

(37)
How Subjacency accounts for the complement cases of the CNPC and of the
wh-island constraint. Relevant examples are given in (37):
a.
b.

*Which race did [IP you hear [DP the announcement [CP that [IP John
had won (which race) ]]]] ?
*Who did [IP you wonder [CP which books [IP to give (which books)
to (who) ]]] ?
In (37a), which race moves from the lower SpecCP to the matrix one,
crossing an DP node and an IP node.
Ian Roberts Li9 – 17 November 2010


8
In (37b), who moves to the matrix SpecCP directly from its base position over
the filled lower SpecCP The one-step movement of who to the matrix SpecCP
crosses two S-nodes, and so violates Subjacency.
Subjacency thus appears able to unify the CNPC and the wh-island constraint
(putting aside the problems for the latter noted by Ross and illustrated in
(30)).
The SSC can also be derived from subjacency. Schematically, the SSC prevents
extraction of the lower DP in the following configuration:
(38)
[IP [DP [IP … DP …] … ] … ]
Clearly, any such extraction will cross at least one IP-node and a DP-node (and this is
true for the more general Subject Condition also).

Subjacency can also take care of the Left Branch Condition. An DP
immediately dominated by another DP must cross that DP and the next IP
node up in order to get to the nearest SpecCP. It will thus necessarily cross
two bounding nodes and therefore violate subjacency.

What about the CSC? Assuming that conjoined categories form a larger
category of the same kind as those conjoined (i.e. that John and Mary is a DP
just like John and Mary), then extraction of a conjoined DP will violate the
Subjacency Condition, as would extraction of a DP out of a conjoined IP.
However, subjacency cannot directly handle extraction of an NP from a
coordinate VP.
In general, then, subjacency went some way towards unifying Ross’ island
constraints. The account relies on the idea that the bounding nodes are NP/DP and
S/IP. The next step is to try to see if we can give an intensional definition of the
bounding nodes. This brings us to the Barriers theory of Chomsky (1986).
3.4
Barriers
A new class of islands, the adjunct islands (Huang (1982)).
(39)
*Which warning did you leave [ despite (which warning)] ?
-- extraction from an adjunct is impossible.
(40)
Condition on Extraction Domains (CED)
No extraction from non-complements, i.e. subjects and adjuncts.
This suggests a relational definition of which categories count for subjacency. And
cf.:
(41)
(42)
a.
b.
a.
b.
*Who did you meet John [AP angry at (who) ] ?
Who did you make John [AP angry at (who) ] ?
*Which city did you meet a man [PP from (which city} ] ?
Which city did you see the destruction [PP of (which city) ] ?
Ian Roberts Li9 – 17 November 2010
9
Chomsky (1986b:14) first defines barriers in terms of “blocking categories” (BC):
(43)
X is a BC for Y iff X is not L-marked and X dominates Y.
(44)
X is a barrier for Y iff (a) or (b):
a.
X immediately dominates Z, Z a BC for Y;
b.
X is a BC for Y, X  IP.
In order to understand how these definitions work, we need a definition of L-marking
(Chomsky (1986b:15)):
(45)
X L-marks Y iff X is a lexical category that θ-governs Y.
Θ-government is the relation between a lexical head and its sister.
(46)
Subjacency:
Movement cannot cross more than one barrier.
(47)
Subject Condition:
[CP
[IP [DP … DP …] … ] … ]







Subjects of finite clauses are not directly θ-marked, since they are not
complements.
Therefore they are not L-marked.
Therefore, by (37), a subject DP in a finite clause is a BC for anything
contained in it, and
therefore a barrier, by (44b).
Moreover, the IP immediately dominating the subject is a barrier, by (44a).
So any element moved out of a subject DP into the nearest SpecCP position
will violate subjacency, as two barriers will be crossed.
This derives the Subject Condition.
[The Adjunct Condition, as in (39), is derived in a precisely analogous way. Adjuncts
are not complements, and so they are not L-marked. Since they are not L-marked they
are BCs and therefore barriers for extraction of material from inside them. Moreover
the IP immediately dominating an adjunct is also a barrier, given (44a). So extraction
from adjuncts is impossible].
(48) The CNPC (relative clauses):
Mary read [NP [NP the book ] [CP which [IP we gave to who ]]]




Relative clauses are not arguments of the DPs they modify.
Because of this, CP in (48) is not L-marked, so it is a BC and a barrier, and
the NP immediately dominating it is therefore also a barrier for material
extracted from inside CP.
As a result, who cannot move to the matrix SpecCP in one step without
violating subjacency.
Ian Roberts Li9 – 17 November 2010
10
[The complement case of the CNPC is more problematic, precisely because the CP
complement to the head noun of the complex NP, being a complement, is L-marked,
and therefore neither a BC nor a barrier. Neither is NP a barrier. Chomsky (1986b:36)
suggests that the CP complement of N may be an inherent barrier. It may thus be that
only one barrier is crossed, leading to a “weaker” violation].
(49)

4.
Wh-island violations:
*Who did [IP you wonder [CP which books [IP to give (which books)
to (who) ]]] ?
The presence of which books in the embedded SpecCP forces who to cross at
least one barrier, namely CP (which inherits barrierhood for material extracted
from within IP from the non-L-marked node IP). Since the main-clause IP is a
BC not a barrier, given (42b), it may be that only one barrier is crossed here,
leading to a weaker violation. Given the examples in (30), this may be the
correct conclusion.
Conclusions




(50)
wh-movement is local after all!
islands > subjacency > barriers: successive refinements of theory of locality
many empirical issies still open (the CSC/ATB, types of wh-islands, etc.)
parametric variation (cf. Ross’ “conditions box” and see Rizzi (1982))
Movement
a.
moves an X or XP;
b.
leaves a trace/copy at the extraction site;
c.
moves to a position c-commanding the trace;
d.
moves to the closest possible Specifier;
e.
obeys the Structure Preservation Principle;
References (check in the textbooks for the sections/chapters on wh-movement)
Carnie, A. (2002). Syntax. A Generative Introduction. Oxford: Blackwell.
Culicover, P. (1997). Principles and Parameters. An Introduction to Syntactic
Theory. Oxford: OUP.
Freidin, R. (1994). Foundations of Generative Syntax. Cambridge, Mass.: MIT Press.
Haegeman, L. (1995). An Introduction to Government and Binding. London:
Blackwell.
Haegeman, L. and J. Guerón. (1999). English Grammar: A Generative Perspective.
London: Blackwell (chapter 2 on Movement and Locality).
Ouhalla, J. (1994/1999). Introducing Transformational Grammar. London: Arnold
Poole, G. (2002). Syntactic Theory. London: Palgrave.
Radford, A. (1988). Introduction to Transformational Syntax. Cambridge: CUP
Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford:
Blackwell.
Roberts, I. (1996) Comparative Syntax, Chapter 4.
Roberts, I. (2007) Comparative Grammar, Volume 4: Wh-Movement, Introduction.
Ross, J.R. (1967) Constraints on Variables in Syntax, MIT PhD dissertation, Chapter
4.